Texts in Computer Science

Marco T. Morazán

Animated
Problem Solving
An Introduction to Program Design
Using Video Game Development
Texts in Computer Science

Series Editors
David Gries, Department of Computer Science, Cornell University, Ithaca, NY, USA
Orit Hazzan , Faculty of Education in Technology and Science, Technion—Israel
Institute of Technology, Haifa, Israel
Titles in this series now included in the Thomson Reuters Book Citation Index!
’Texts in Computer Science’ (TCS) delivers high-quality instructional content for
undergraduates and graduates in all areas of computing and information science,
with a strong emphasis on core foundational and theoretical material but inclusive
of some prominent applications-related content. TCS books should be reasonably
self-contained and aim to provide students with modern and clear accounts of topics
ranging across the computing curriculum. As a result, the books are ideal for semester
courses or for individual self-study in cases where people need to expand their
knowledge. All texts are authored by established experts in their fields, reviewed
internally and by the series editors, and provide numerous examples, problems, and
other pedagogical tools; many contain fully worked solutions.
The TCS series is comprised of high-quality, self-contained books that have broad
and comprehensive coverage and are generally in hardback format and sometimes
contain color. For undergraduate textbooks that are likely to be more brief and
modular in their approach, require only black and white, and are under 275 pages,
Springer offers the flexibly designed Undergraduate Topics in Computer Science
series, to which we refer potential authors.

More information about this series at https://link.springer.com/bookseries/3191

Marco T. Morazán

Animated Problem Solving

An Introduction to Program Design
Using Video Game Development
Marco T. Morazán
Department of Computer Science
Seton Hall University
South Orange, NJ, USA

ISSN 1868-0941 ISSN 1868-095X (electronic)

Texts in Computer Science
ISBN 978-3-030-85090-6 ISBN 978-3-030-85091-3 (eBook)
https://doi.org/10.1007/978-3-030-85091-3

© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland
AG 2022
This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether
the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse
of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and
transmission or information storage and retrieval, electronic adaptation, computer software, or by similar
or dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors, and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or
the editors give a warranty, expressed or implied, with respect to the material contained herein or for any
errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.

This Springer imprint is published by the registered company Springer Nature Switzerland AG.
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
To my parents, Doris and Marco, who taught
me to love teaching and to realize that having
an education is not a privilege but a
responsibility.
Preface

Everybody engages in problem solving. It is a natural and inevitable part of life.

Historically, the link between problem solving and programming has been less
emphasized. When you write an essay, you are programming—at different levels
many times. You make sure ideas flow and arguments make use of the data you are
analyzing. You write several drafts of the essay. Each draft represents a refinement.
Every paragraph has a point and you avoid repeating yourself. All this is part of
programming, including computer programming. Programmers, people who solve
problems using a computer, go through the exact same steps to write a program.
The same steps are taken by a psychologist analyzing a patient and by a chemist
experimenting in a laboratory. Even a painter engages in programming. No? Does a
painter not want to elicit an outcome or an emotion in you? Indeed, how to achieve
this is a problem that must be solved by the artist. Consider the painting Sorrowing
Old Man (At Eternity’s Gate) by Vincent van Gogh (search for it on the internet).
Can you see the old man’s sorrow? Can you imagine the weight of the years on him?
If so, we can say that the painter successfully solved the problem. This brings us to
another important component of problem solving: testing. It is not only important
to solve a problem. It is equally important to test the solution to make sure it works
and in many cases to make sure that the solution is efficient.
This book is about systematic problem solving or if you like about systematic
reasoning. Unlike most textbooks about programming, this textbook is not about
tinkering with or hacking code. This book is about making a plan to solve a problem
and then implementing the solution. As we shall discover, it turns out that the
solutions to many problems are similar. This should not come as a surprise because
we often solve many problems using similar data. How do you do grocery shopping?
You make a list of items and check them off as you put them in your cart. How do
you manage your chores today? You make a list of chores and check them off as
you get them done. Pretty similar, no? Similarities give rise to abstraction to avoid
repetitions—or reinventing the proverbial wheel. This book, therefore, is also about
abstraction. Thinking abstractly is a powerful tool in problem solving.
In this textbook, all the solutions to problems are expressed as programs. It is
important to be somewhat precise about what a program is. A program is much more
vii
viii Preface

than just code written using a programming language. Remember that a program is
a solution to a problem. Therefore, a program has a design, code, examples of how it
works, and tests. That is, it communicates how the problem is solved and illustrates
that the solution works. If any of the mentioned components are missing, then we
have an incomplete program. Would you believe someone who simply told you that
n2 , where n is a nonnegative integer, is the sum of the first n odd numbers? Many
readers would be skeptical. What if they also provided the following examples:
02 = 0
22 = 1 + 3
42 = 1 + 3 + 5 + 7
It is very likely that most readers would now feel more confident that the claim is
true. It is the same in programming. We cannot simply say that here is a function
that does this or that. We need to explain how the function computes its value, and
we need to have examples that show how it works. The steps taken to design a
program in a systematic manner is called a design recipe. In this textbook, you shall
study many different design recipes. Each design recipe shall become a tool in your
problem-solving toolbox.
There are two problem-solving techniques that are emphasized throughout the
book: divide and conquer and iterative refinement. Divide and conquer is the process
by which a large problem is broken into two or more smaller problems that are easier
to solve and then the solutions for the smaller pieces are combined to create an answer
to the problem. Iterative refinement is the process by which a solution to a problem
is gradually made better—like the drafts of an essay. Mastering these techniques is
essential to becoming a good problem solver and programmer.
Finally, problem solving ought to be fun. To this end, this book promises that by
the end of it you will have designed and implemented a multiplayer video game that
you can play with your friends over the internet. To achieve this, however, there is a
lot about problem solving and programming that you must first learn. The game is
developed using iterative refinement. As we learn about programming, we shall apply
our new knowledge to develop increasingly better versions of the video game. In
fact, every skill you develop for problem solving and program design is transferable
to other (non-programming) domains and to other programming languages.

1 The Languages and the Parts of the Book

The book uses the Racket student languages to write programs. These languages
are chosen for several reasons. The first is that they have an error-messaging system
specifically designed for beginners. This means that unlike common programming
languages the error messages are likely to make sense to beginners. If you do not
understand an error message, do not hesitate to ask your professor or search for
help online. The second is that the syntax is simple and easy to understand. This
is important because the emphasis is always on problem solving and not on how
1 The Languages and the Parts of the Book ix

to correctly write expressions. The third is that the student languages progressively
become richer. At the beginning, you have fewer features at your disposal and,
therefore, the possible errors are fewer. The fourth reason is that the student languages
come with powerful libraries to create graphics, animations, and video games. These
libraries allow students to inject their own personalities in the development of games
and animations. You are strongly encouraged to be creative. Finally, the fifth reason
is that the Racket student languages are likely to put all students on the same playing
field. Most students will be learning the syntax of the programming language together
for the first time.
The book is divided into five parts. Part I focuses on the basics. It starts with
how to write expressions. Once expressions are mastered, the first abstraction lesson
introduces us to functions. In addition, this part introduces you to conditional expres-
sions that allow you to write programs that make decisions. Just this much knowledge
allows us to write interactive programs and puts us on our way to a multiplayer video
game. As you shall discover, decision-making is fundamental to solving problems
that involve information that has many varieties. For example, the whole numbers
may be positive or negative—two varieties—and how a whole number is processed
depends on which variety a given number belongs to. Think about how to compute
the absolute value of a whole number.
Part II introduces you to compound data of finite size. Compound data has multiple
values associated. For example, a point on the Cartesian plane is compound data of
finite size. There are two values: an x coordinate and a y coordinate. Being able to
define compound data of finite size to represent elements in the real or an imaginary
world is a powerful skill to develop.
Part III introduces you to compound data of arbitrary size. This is data that has
multiple values, but the number of values is not fixed. Once again, think about a
grocery list. Sometimes there are no items in the list and at other times there may
be 10, 6, or 17 items in the list. This is where you are introduced to structural
recursion—a powerful data-processing strategy that uses divide and conquer to
process data whose size is not fixed. The types of data that are introduced are lists,
intervals, natural numbers, and binary trees. The knowledge developed is used to
develop a video game that is more challenging for the player.
Part IV delves into abstraction. This section is where we learn how to eliminate
repetitions in our solutions to problems. In fact, we learn how different data can
be processed and different problems can be solved in exactly the same way. You
are introduced to generic programming, which is abstraction over the type of
data processed. This leads to the realization that functions are data and, perhaps
more surprising, that data are functions. In other words, the line between data and
functions is artificial—a fact that is not emphasized enough in Computer Science
textbooks. This realization naturally leads to object-oriented programming—a topic
that you are likely to study extensively.
Part V introduces you to distributed programming—using multiple computers to
solve a problem. This is a topic that until now has never been addressed in a textbook
for beginning programmers. The fact that you develop proficiency in program design
makes it possible for this topic, common in modern computer applications, to be
x Preface

discussed. If you have ever sent a text message or have ever played a game online,
then you have benefitted from and have used a distributed program. It is impossible,
of course, to discuss all the nuances of distributed programming in this textbook.
Nonetheless, you are introduced to a modern trend that is likely to be common
throughout your professional career and beyond.

2 Acknowledgments

This book is the product of over ten years of work at Seton Hall University build-
ing on the shoulders of giants in Computer Science. There are many persons and
groups who deserve credit for informing my work. The Racket community has been
unequivocal in its support for the techniques that I have developed. There is an un-
payable debt of gratitude owed to Matthias Felleisen from Northeastern University
for our discussions over the years about Computer Science education, Liberal Arts
education, and program design. My students and I have greatly benefitted from his
support. Other Racketeers who have deeply influenced me are Shriram Krishna-
murthi, Matthew Flatt, Robert Bruce Findler, and Kathi Fisler. This textbook is a
tribute to our debates and their published work.
I would also like to thank the Trends in Functional Programming (TFP) and
the Trends in Functional Programming in Education (TFPIE) communities. These
communities provided (and continue to provide) a venue to discuss and present
work advancing Computer Science education. I am grateful to many individuals
including Peter Achten, Jurriaan Hage, Pieter Koopman, Simon Thompson, and
Marko van Eekelen. Their insightful feedback has informed much of the material in
this textbook.
Finally, I would like to thank Seton Hall University and its Department of Com-
puter Science for supporting the development of the work presented in this textbook.
In particular, the support of John T. Saccoman, Manfred Minimair, and Daniel Gross
is appreciated. Most of all, I am grateful to all my CS1 students over the past decade
who have informed my Computer Science education efforts. It is likely true that my
students have learned a great deal in my courses, but it is an absolute certainty that
I have learned more from them. They have refined the delivery of every idea found
in this textbook. I am especially grateful to all my undergraduate tutors and teach-
ing assistants, including Shamil Dzhatdoyev, Josie Des Rosiers, Nicholas Olson,
Nicholas Nelson, Lindsey Reams, Craig Pelling, Barbara Mucha, Joshua Schappel,
Sachin Mahashabde, Rositsa Abrasheva, Isabella Felix, and Sena Karsavran. Without
my dedicated students at Seton Hall University and their insight into what students
understood, this textbook would have been impossible.
Contents

Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii
1 The Languages and the Parts of the Book . . . . . . . . . . . . . . . . . . . . . viii
2 Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . x

Part I The Basics of Problem Solving with a Computer

1 The Science of Problem Solving . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

3 Getting Started . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
4 Computing New Values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
5 Definitions and Interactions Areas Differences . . . . . . . . . . . . . . . . . 11
6 Saving Your Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
7 Error Messages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
7.1 Grammatical Errors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
7.2 Type Errors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
7.3 Runtime Errors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
8 What Have We Learned in This Chapter? . . . . . . . . . . . . . . . . . . . . . 19

2 Expressions and Data Types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

9 Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
10 Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
11 Strings and Characters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
12 Symbols . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
13 Booleans . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
13.1 Basic Boolean Operators in BSL . . . . . . . . . . . . . . . . . . . . . 32
13.2 Predicates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
14 Images . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
14.1 Basic Image Constructors . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
14.2 Property Selectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
14.3 Image Composers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

xi
xii Contents

14.4 Empty Scenes and Placing Images . . . . . . . . . . . . . . . . . . . 42

15 What Have We Learned in This Chapter? . . . . . . . . . . . . . . . . . . . . . 45

3 The Nature of Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

16 The Rise of Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
17 General Design Recipe for Functions . . . . . . . . . . . . . . . . . . . . . . . . . 51
17.1 The Design Recipe in Action . . . . . . . . . . . . . . . . . . . . . . . . 53
18 Auxiliary Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
18.1 Bottom-Up Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
19 Top-Down Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
20 What Have We Learned in This Chapter? . . . . . . . . . . . . . . . . . . . . . 68

4 Aliens Attack Version 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71

21 The Scene for Aliens Attack . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
22 Creating Aliens Attack Images . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
23 Shot Image . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
24 Alien Image . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
25 Rocket Image . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
25.1 Rocket Window Image Constructor . . . . . . . . . . . . . . . . . . . 82
25.2 Rocket Fuselage Image Constructor . . . . . . . . . . . . . . . . . . 84
25.3 Rocket Single Booster Image Constructor . . . . . . . . . . . . . 85
25.4 Rocket Booster Image Constructor . . . . . . . . . . . . . . . . . . . 87
25.5 Rocket Main Body Image Constructor . . . . . . . . . . . . . . . . 88
25.6 Rocket Nacelle Image Constructor . . . . . . . . . . . . . . . . . . . 90
25.7 Rocket ci Constructor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
26 Drawing Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
27 What Have We Learned in This Chapter? . . . . . . . . . . . . . . . . . . . . . 100

5 Making Decisions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101

28 Conditional Expressions in BSL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
29 Designing Functions to Process Data with Variety . . . . . . . . . . . . . . 104
30 Enumeration Types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
31 Interval Types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
32 Itemization Types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
33 What Have We Learned in This Chapter? . . . . . . . . . . . . . . . . . . . . . 125

6 Aliens Attack Version 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127

34 The Universe Teachpack . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
35 A Video Game Design Recipe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
36 Adding the Rocket to Aliens Attack . . . . . . . . . . . . . . . . . . . . . . . . . . 140
37 What Have We Learned in This Chapter? . . . . . . . . . . . . . . . . . . . . . 149

Part II Compound Data of Finite Size

Contents xiii

7 Structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
38 The posn Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
39 Going Beyond the Design Recipe . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
40 Revisiting in-Q1? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
41 What Have We Learned in This Chapter? . . . . . . . . . . . . . . . . . . . . . 166

8 Defining Structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167

42 Defining Structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
43 Computing Structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
44 Structures for the Masses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176
45 What Have We Learned in This Chapter? . . . . . . . . . . . . . . . . . . . . . 183

9 Aliens Attack Version 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185

46 Data Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
47 Function Templates and Sample Instances . . . . . . . . . . . . . . . . . . . . . 187
48 The run Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
49 Drawing the World . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
49.1 The draw-world Refinement . . . . . . . . . . . . . . . . . . . . . . . 191
49.2 Drawing Aliens . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
50 The process-key Refinement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
51 Processing Ticks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
51.1 The process-tick Handler . . . . . . . . . . . . . . . . . . . . . . . . 197
51.2 The Design of new-dir-after-tick . . . . . . . . . . . . . . . . 199
51.3 The Design of Auxiliary Functions for
new-dir-after-tick . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
51.4 The Design of move-alien . . . . . . . . . . . . . . . . . . . . . . . . . 206
52 Subtyping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
52.1 Checking Errors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
53 The game-over? Handler . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
54 Computing the Last Scene . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
55 What Have We Learned in This Chapter? . . . . . . . . . . . . . . . . . . . . . 223

10 Structures and Variety . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225

56 A Bottom-Up Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226
57 Code Refactoring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233
58 What Have We Learned in This Chapter? . . . . . . . . . . . . . . . . . . . . . 236

11 Aliens Attack Version 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239

59 Data Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 240
60 The draw-world Refinement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
61 The process-key Refinement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246
62 The process-tick Refinement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 251
62.1 The Refinement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 251
62.2 The move-shot Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253
63 The game-over? Refinement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 256
63.1 The hit? Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257
xiv Contents

63.2 The draw-last-world Refinement . . . . . . . . . . . . . . . . . . 259

64 What Have We Learned in This Chapter? . . . . . . . . . . . . . . . . . . . . . 261

Part III Compound Data of Arbitrary Size

12 Lists . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 265
65 Creating and Accessing Lists in ISL+ . . . . . . . . . . . . . . . . . . . . . . . . 266
66 Shorthand for Building Lists . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269
67 Recursive Data Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271
68 Generic Data Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 274
69 Function Templates for Lists . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275
70 Designing List-Processing Functions . . . . . . . . . . . . . . . . . . . . . . . . . 277
71 What Have We Learned in This Chapter? . . . . . . . . . . . . . . . . . . . . . 279

13 List Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 281

72 List Summarizing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 281
73 List Searching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 286
74 List ORing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 288
74.1 Determining If an Alien Is at the Left Edge . . . . . . . . . . . . 288
74.2 Determining If an Alien Is at the Right Edge . . . . . . . . . . . 291
74.3 Determining If an Alien Has Reached Earth . . . . . . . . . . . 292
75 List ANDing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 294
75.1 All Even in a lon . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295
75.2 Determining if a lon Is Sorted . . . . . . . . . . . . . . . . . . . . . . 296
76 List Mapping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 300
76.1 Moving a List of Aliens . . . . . . . . . . . . . . . . . . . . . . . . . . . . 300
76.2 Moving a List of Shots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 301
76.3 Returning a Different List Type . . . . . . . . . . . . . . . . . . . . . . 302
77 List Filtering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 304
77.1 Extracting Even numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . 305
77.2 Removing Hit Aliens . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 306
77.3 Removing Shots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 310
78 List Sorting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 314
79 What Have We Learned in This Chapter? . . . . . . . . . . . . . . . . . . . . . 318

14 Natural Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 319

80 Data Definition for a Natural Number . . . . . . . . . . . . . . . . . . . . . . . . 320
81 Computing Factorial . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 321
82 Computing Tetrahedral Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . 323
83 Making Copies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 327
84 What Have We Learned in This Chapter? . . . . . . . . . . . . . . . . . . . . . 329
Contents xv

15 Interval Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 331

85 Interval Data Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 332
86 Revisiting Factorial . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 334
87 Creating an Army of Aliens . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 336
88 Largest Prime in an Interval . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 342
89 What Have We Learned in This Chapter? . . . . . . . . . . . . . . . . . . . . . 347

16 Aliens Attack Version 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 349

90 New world Data Definition and Function Template . . . . . . . . . . . . . 349
91 The draw-world Refinement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 351
92 The process-key Refinement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 354
93 The process-tick Refinement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 358
93.1 The new-dir-after-tick Design . . . . . . . . . . . . . . . . . . 361
94 The game-over? Refinement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 365
94.1 The draw-last-world Refinement . . . . . . . . . . . . . . . . . . 367
95 A Bug Despite Hundreds of Tests Passing . . . . . . . . . . . . . . . . . . . . . 369
96 What Have We Learned in This Chapter? . . . . . . . . . . . . . . . . . . . . . 372

17 Binary Trees . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 373

97 Binary Tree Data Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 374
98 Traversing a Binary Tree . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 376
99 The Maximum of a (btof int) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 379
100 Binary Search Trees . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 382
100.1 A (listof cr) Representation . . . . . . . . . . . . . . . . . . . . . . . . . 383
100.2 A (btof cr) Representation . . . . . . . . . . . . . . . . . . . . . . . . . . 384
100.3 A (bstof cr) Representation . . . . . . . . . . . . . . . . . . . . . . . . . 386
101 Abstract Running Time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 388
102 The Complexity of Searching the Criminal Database . . . . . . . . . . . . 391
103 Balanced (bstof cr) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 393
103.1 Creating a Balanced Binary Search Tree . . . . . . . . . . . . . . 393
103.2 Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 396
104 What Have We Learned in This Chapter? . . . . . . . . . . . . . . . . . . . . . 398

18 Mutually Recursive Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 401

105 Designing with Mutually Recursive Data . . . . . . . . . . . . . . . . . . . . . . 403
105.1 Revisiting the Maximum of a (btof int) . . . . . . . . . . . . 403
106 Evaluating Arithmetic Expressions . . . . . . . . . . . . . . . . . . . . . . . . . . . 407
107 Trees . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 414
107.1 Creating a Search Tree for Tic Tac Toe . . . . . . . . . . . . . . . . 418
107.2 Can Win Tic Tac Toe? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 422
108 Project: Tic Tac Toe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 426
108.1 Data Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 427
108.2 Design draw-world . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 428
108.3 Design process-mouse . . . . . . . . . . . . . . . . . . . . . . . . . . . . 428
108.4 Design process-tick . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 429
xvi Contents

108.5 Design game-over? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 430

109 What Have We Learned in This Chapter? . . . . . . . . . . . . . . . . . . . . . 430

19 Processing Multiple Inputs of Arbitrary

Size . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 433
110 One Input Has a Dominant Role . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 433
111 Inputs Must Be Processed Simultaneously . . . . . . . . . . . . . . . . . . . . . 436
112 No Clear Relationship Between the Inputs . . . . . . . . . . . . . . . . . . . . . 438
113 What Have We Learned in This Chapter? . . . . . . . . . . . . . . . . . . . . . 442

Part IV Abstraction

20 Functional Abstraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 445

114 A Design Recipe for Abstraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 447
115 Functions as Values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 447
116 Abstraction Over List-Processing Functions . . . . . . . . . . . . . . . . . . . 450
116.1 List Summarizing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 450
116.2 List Searching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 456
116.3 List ORing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 459
116.4 List ANDing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 462
116.5 List Mapping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 464
116.6 List Filtering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 466
116.7 List Sorting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 467
117 Abstraction over Interval-Processing Functions . . . . . . . . . . . . . . . . 472
118 What Have We Learned in This Chapter? . . . . . . . . . . . . . . . . . . . . . 475

21 Encapsulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 477
119 Local-Expressions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 477
120 Lexical Scoping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 480
121 Using Local-Expressions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 483
121.1 Encapsulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 483
121.2 Readability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 488
121.3 Furthering Functional Abstraction . . . . . . . . . . . . . . . . . . . . 490
121.4 One-Time Expression Evaluation . . . . . . . . . . . . . . . . . . . . 492
122 What Have We Learned in This Chapter? . . . . . . . . . . . . . . . . . . . . . 497

22 Lambda Expressions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 499

123 Anonymous Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 501
124 Revisiting Function Composition . . . . . . . . . . . . . . . . . . . . . . . . . . . . 503
125 Curried Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 505
126 Designing Using Existing Abstractions . . . . . . . . . . . . . . . . . . . . . . . 511
126.1 Computing the Value of a Series . . . . . . . . . . . . . . . . . . . . . 511
126.2 Approximating π . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 514
127 What Have We Learned in This Chapter? . . . . . . . . . . . . . . . . . . . . . 516
Contents xvii

23 Aliens Attack Version 5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 517

128 Constants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 518
129 Structure Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 522
130 Encapsulating and Refactoring Handlers . . . . . . . . . . . . . . . . . . . . . . 522
130.1 The draw-world Handler . . . . . . . . . . . . . . . . . . . . . . . . . . 522
130.2 The process-key Handler . . . . . . . . . . . . . . . . . . . . . . . . . 525
130.3 The process-tick Handler . . . . . . . . . . . . . . . . . . . . . . . . 526
130.4 The game-over? Handler . . . . . . . . . . . . . . . . . . . . . . . . . . 532
130.5 The draw-last-world Handler . . . . . . . . . . . . . . . . . . . . . 533
131 Refactoring run . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 534
132 What Have We Learned in This Chapter? . . . . . . . . . . . . . . . . . . . . . 535

24 For-Loops and Pattern Matching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 537

133 For-Loops . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 538
133.1 for-loops . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 538
133.2 for*-loops . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 542
134 Pattern Matching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 546
134.1 Illustrative Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 547
134.2 Refactoring Using Pattern Matching . . . . . . . . . . . . . . . . . . 549
134.3 Designing Using Pattern Matching . . . . . . . . . . . . . . . . . . . 551
135 What Have We Learned in This Chapter? . . . . . . . . . . . . . . . . . . . . . 554

25 Interfaces and Objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 557

136 Interfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 558
136.1 Improving the Human Interface . . . . . . . . . . . . . . . . . . . . . . 561
136.2 Services that Require More Input . . . . . . . . . . . . . . . . . . . . 561
137 A Design Recipe for Interfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 565
138 Interfaces and Union Types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 566
139 An Abbreviated (listof X) Interface . . . . . . . . . . . . . . . . . . . . . . . . . . 567
139.1 Step 1: Values and Services . . . . . . . . . . . . . . . . . . . . . . . . . 567
139.2 Step 2: Interface and Message Definitions . . . . . . . . . . . . . 568
139.3 Step 3: Class Function Template . . . . . . . . . . . . . . . . . . . . . 568
140 The Empty (listof X) Class . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 571
140.1 Step 4: Signature, Purpose, Class Header, and
Message-Passing Function . . . . . . . . . . . . . . . . . . . . . . . . . . 571
140.2 Step 5: Auxiliary Functions . . . . . . . . . . . . . . . . . . . . . . . . . 572
141 The Non-Empty (listof X) Class . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 573
141.1 Step 4: Signature, Purpose, Class Header, and
Message-Passing Function . . . . . . . . . . . . . . . . . . . . . . . . . . 573
141.2 Step 5: Auxiliary Functions . . . . . . . . . . . . . . . . . . . . . . . . . 574
142 Step 6: Wrapper Functions and Tests . . . . . . . . . . . . . . . . . . . . . . . . . 575
143 What Have We Learned in This Chapter? . . . . . . . . . . . . . . . . . . . . . 579

Part V Distributed Programming

xviii Contents

26 Introduction to Distributed Programming . . . . . . . . . . . . . . . . . . . . . . . . 583

144 A Design Recipe for Distributed Programming . . . . . . . . . . . . . . . . . 585
145 More on the Universe API . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 586
146 A Chat Application . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 589
146.1 The Components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 589
146.2 Data Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 589
146.3 Communication Protocol . . . . . . . . . . . . . . . . . . . . . . . . . . . 592
146.4 Marshalling and Unmarshalling . . . . . . . . . . . . . . . . . . . . . . 593
146.5 Component Implementation . . . . . . . . . . . . . . . . . . . . . . . . . 593
146.6 Running the Chat Tool . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 600
147 What Have We Learned in This Chapter? . . . . . . . . . . . . . . . . . . . . . 601

27 Aliens Attack Version 6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 603

148 Refining the world Data Definition . . . . . . . . . . . . . . . . . . . . . . . . . . 603
149 The draw-world Refinement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 607
150 The process-key Refinement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 610
151 The process-tick Refinement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 615
152 The game-over? Refinement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 617
153 What Have We Learned in This Chapter? . . . . . . . . . . . . . . . . . . . . . 619

28 Aliens Attack Version 7 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 621

154 Components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 621
155 Data Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 622
156 Communication Protocol . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 623
156.1 Player-Sparked Communication Chains . . . . . . . . . . . . . . . 623
156.2 Server-Sparked Communication Chains . . . . . . . . . . . . . . . 624
156.3 Message Data Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . 626
157 Marshalling and Unmarshalling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 632
158 Component Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 638
158.1 Player Component . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 638
158.2 Server Component . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 646
159 A Subtle Bug . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 653
160 What Have We Learned in This Chapter? . . . . . . . . . . . . . . . . . . . . . 656

29 Aliens Attack Version 8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 657

161 The Components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 657
162 Data Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 658
163 Communication Protocol . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 659
163.1 Player-Sparked Communication Chains . . . . . . . . . . . . . . . 660
163.2 Server-Sparked Communication Chains . . . . . . . . . . . . . . . 661
163.3 Message Data Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . 662
164 Marshalling and Unmarshalling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 664
165 Component Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 665
165.1 Player Component . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 666
165.2 Server Component . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 669
Contents xix

166 A Subtle Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 683

167 What Have We Learned in This Chapter? . . . . . . . . . . . . . . . . . . . . . 684

Part VI Epilogue

30 Advice for Future Steps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 687

168 Advice for Computer Science Students . . . . . . . . . . . . . . . . . . . . . . . 687
169 Advice for Non-Computer Science Students . . . . . . . . . . . . . . . . . . . 688
 Part I
The Basics of Problem Solving
with a Computer
Chapter 1
The Science of Problem Solving

We all solve problems every day. Have you ever thought about how you go about
problem solving? Do you randomly go about trying potential solutions to a problem
or do you think about how to solve the problem? Most of the time you probably think
about the problem to find a solution. A natural question that arises is how do we think
about a problem to find a solution. In other words, what steps do we take to arrive to
a plausible solution? The solution is plausible until we test it and feel confident that
it works. If it does not work, of course, we go back to thinking about the problem to
obtain a refined solution. Understanding how to think about problems and solutions
is where Computer Science and programming are beneficial to everyone.
Computer Science is not the study of computers just like Chemistry is not the
study of test tubes nor Astronomy is the study of telescopes. So, why is it called
Computer Science? Although there is no clear answer to this question, the best
guess is that Computer Science is an umbrella term for many disciplines whose
primary tool is the computer, such as programming language theory, algorithmics,
software engineering, data mining, robotics, and artificial intelligence. If Computer
Science is not the study of the computer, then what is Computer Science and what do
computer scientists do? Computer scientists solve problems. Unlike biologists who
solve problems in Biology or diplomats who solve problems in Diplomacy, computer
scientists solve problems that are relevant to all fields of study and to all facets of
human life. Stated simply, Computer Science is multidisciplinary. Therefore, the
best way to describe Computer Science is to say that it is the science of problem
solving. Programming is how computer scientists express solutions to problems.
Although Computer Science is a relatively young discipline,1 it has developed many
effective techniques to solve problems. This is why everyone ought to learn to design
programs. Problem solving is an essential skill just like reading, writing, and doing
arithmetic. Your journey through this book will help you learn effective problem
solving techniques.

1
The first Computer Science Program was established in 1953 at the University of Cambridge in
the United Kingdom. The first Computer Science Department in the United States was established
at Purdue University in 1962.
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 3
M. T. Morazán, Animated Problem Solving, Texts in Computer Science,
https://doi.org/10.1007/978-3-030-85091-3_1
4 1 The Science of Problem Solving

Two of the best-known problem solving techniques are divide and conquer and
iterative refinement. Problem solvers use divide and conquer when a larger problem
is decomposable into smaller problems. Instead of solving an entire problem in one
huge step, the problem is divided into a set of smaller problems. These subproblems
are all solved independently, and then their solutions are combined to formulate the
answer to the large problem. For example, consider computing your quiz average.
You can divide this problem into two smaller problems: sum the quiz grades and
count the number of quizzes. Once these subproblems are solved, the results are
combined by dividing the former by the latter to formulate the quiz average.
Iterative refinement is used to develop a solution in steps. This is particularly
useful to manage the complexity of a problem. Instead of developing a full answer
at once, you solve a simpler version of the problem. Once that is done, you add
complexity to the problem and re-solve it. This process continues until you have
a full answer. For instance, consider the problem of developing a video game like
Pacman. You may first create a game that only has Pacman. Then you create a game
that has Pacman and food. The next version of the game has Pacman, the food, and
one ghost. Finally, the last version of the game has Pacman, the food, and multiple
ghosts.
You have been taught to solve problems most of your life. This means that you have
been taught to compute. For example, you have been taught to compute meaning from
English expressions. You know that Thelma is that pig has a different meaning from
That is Thelma’s pig. Other languages that you have been trained to do computations
with are Arithmetic and Algebra. For example, you know how to compute the value of
the following expression: 5 * (8 + 2). In essence, we use expressions to describe
computations and we evaluate expressions to derive meaning. You may have never
thought explicitly about divide and conquer or about iterative refinement, but you
have been taught to use these techniques. Consider writing an essay. The first step
is to create an outline. This is where you decompose your argument into different
points. Perhaps, every point is implemented as a different section. You combine
the different sections into an essay by making sure that your argument easily flows
from one section to the next. As you can observe, you are using divide and conquer.
Staying with essay development, you first write a rough draft and then repeatedly
make improvements until you are happy with the result. In other words, you are using
iterative refinement.
If we are going to express solutions to problems using a computer, we need to use
a programming language much like we use English to communicate. A programming
language allows us to communicate to the computer what we want it to do. There are
many programming languages. Each has its strengths and weaknesses. You will learn
many programming languages as you explore the world of Computer Science. We
will start with a programming language called Beginning Student Language (BSL).
BSL is a programming language specifically designed to teach beginners like yourself
how to design and implement solutions to problems. A solution to a problem written
in BSL (or any other programming language) is called a program. When a program
is evaluated, we obtain its meaning. That is, we obtain the solution to an instance of
a problem.
3 Getting Started 5

Fig. 1 DrRacket’s language menu

3 Getting Started

In order to write programs, we need a programming development environment.2

A programming development environment is to programs what a text editor is to
essays. Just like you write an essay using a text editor, we write programs using
a programming development environment. There is, however, a major difference
between a programming development environment and a text editor. You can do
more than write programs with a programming development environment. You can
also evaluate programs.
The first step you need to take is to download and install the programming
development environment called DrRacket using the following link:
https://download.racket-lang.org/
After installing DrRacket, run it and go to the Language menu. Click on Choose
Language... . A pop-up menu will appear and you need to click on Beginning
Student. Figure 1 illustrates what the language menu ought to look like. Click on
OK and then click on the RUN button toward the top right corner of DrRacket.
You are now ready to program! Figure 2 displays DrRacket’s interface (the colors
on your computer may vary). The top half is called the definitions area and the bottom
half is called the interactions area. We type programs in the definitions area and we
interact with our programs in the interactions area. After typing a program in the

2
A programming development environment is commonly called an integrated development envi-
ronment (IDE).
6 1 The Science of Problem Solving

Fig. 2 DrRacket’s interface

definitions area, you may click the RUN button to evaluate your program. Let us try
this out. Type the following string3 in the definitions area:
"Hello World! I am DrRacket."
Now click on RUN. In the interactions window you will see the string printed.
Congratulations! You just wrote your first program. When you clicked on the RUN
button, you told DrRacket to evaluate your program. The value of a string is just
the string itself. Once DrRacket knows the value of a program, it is printed in the
interactions area.

* Ex. 1 — Write and run a program that prints your name.

** Ex. 2 — Write and run a program that prints your age.

** Ex. 3 — Write and run a program that prints your name and your age.

So, what is a program? A program is similar to an essay. In an essay, you express an

argument in favor of or against a point of view. In a program, you express the solution
to a problem. They both have in common the use of expressions. In English, your
expressions are words that may be used to construct larger expressions like sentences
and paragraphs. In programming, your basic expressions are values (for now), like
strings and numbers, that may be used to create larger useful expressions. We will
start by studying expressions and discover how expressions lead us to functions—the
sentences of Algebra and programming.

3
A string is anything inside,"", double quotes.
3 Getting Started 7

Fig. 3 Initial grammar for expressions

expr ::= string
::= number
string::= "<character>*"
number ::= positive
::= negative
positive ::= mag
negative ::= -mag
mag ::= int
::= real
::= fraction
int ::= digit+
digit ::= 0|1|2|3|4|5|6|7|8|9
real ::= digit* .digit+
fraction ::= int/nonzero-int
nonzero-int ::= <1|2|3|4|5|6|7|8|9>int
program ::= expr*

To do so, we need to know how to write valid expressions. That is, we need a
mechanism to describe how to write expressions. We will use a grammar to describe
expressions. A grammar consists of a series of production rules. A production rule
tells a programmer, for example, how to type valid expressions.
Figure 3 displays our initial grammar for expressions. Each production rule con-
sists of three parts. The leftmost part is the syntactic category being defined. If this
is missing, it means that it is the same syntactic category as the production rule
above it. The middle part is always the symbol ::=. The symbol ::= may be read
as is or as may be substituted by. The rightmost part is the definition of the syntactic
category. The first two rules state that an expression, expr, is either a string or a
number. The next rule states that a string starts and ends with ". In between, you
may have 0 or more characters. The * means 0 or more.4 Observe that character
is in a different font and surrounded by < and >. This indicates that it is a syntactic
category we only describe verbally. You may think of a character as anything that is
produced by a keystroke such as letters, spaces, and punctuation marks.5 The next
rule states that a number is either a positive or negative. A positive is a mag
(magnitude) and a negative is a - followed by a mag.
There are three rules for mag. These state that a mag may be an integer, a real, or
a fraction. An int is 1 or more digits (the + means 1 or more and is called Kleene
plus). A digit is an integer in [0..9]. A real is 0 or more digits followed by a .
followed by 1 or more digits. Finally, a fraction is an int followed by a followed
by a nonzero integer. In the grammar, the | means or. Therefore, a nonzero-int is
any digit from 1 to 9 followed by an integer.

4
In a grammar, a * is called Kleene star, named after the American mathematician Stephen Cole
Kleene.
5
Characters inside a computer are represented by an ASCII code. For example, the ASCII code for
a is 97. You will learn more about ASCII codes in a Computer Architecture course.
8 1 The Science of Problem Solving

We are now ready for our initial grammatical definition of a program. The last
rule in Fig. 3 states that a program is 0 or more expressions. Why do we need 0
or more expressions? Given that a program may compute an arbitrary number of
values, a program may have 0 or more expressions.
Let us try to write a new program. To write a new program, open a new tab in
DrRacket by using Ctrl-T. Ctrl-T means pressing the Ctrl and the T keys at the
same time. Now write a program in the definitions area to compute the name of this
textbook, the number one million, and negative 8.7 as follows:
"Program By Design"
1000000
-8.7
The result in the interactions area after clicking RUN is
"Program by Design"
1000000
-8.7

* Ex. 4 — Write and run a program that prints your year of birth and your age.

** Ex. 5 — Write and run a program that prints your favorite color, your lucky
number, and the largest prime number less than 25.

4 Computing New Values

So far, all our programs have computed constant values, that is, values that we wrote
before running the program. In order for programs to be truly useful, we need to
be able to compute new values. That is, we need to be able to combine values to
create new values. Fortunately, BSL provides us with application expressions. An
application expression applies a function to one or more arguments. The function,
in essence, combines its inputs to create a new value and returns this new value.
Therefore, an application expression evaluates to the value returned by the function.
We extend the expr’s production rules in Fig. 3 to include application expressions
as follows:
expr ::= (<function> expr+ )
This production rule states that an application expression starts with an opening
parenthesis, followed by a function (the operator), followed by 1 or more expressions
(the operands), followed by a closing parenthesis. The function and each of its
arguments must be separated from each other by one or more spaces. BSL provides
us with many functions, for example, to combine numbers and to combine strings.
Table 1 lists some of the functions BSL provides along with sample uses.
At this point, you may feel that the syntax notation is a bit awkward. In your
Mathematics textbooks, for example, the basic operations (i.e., +, -, *, and /) are
4 Computing New Values 9

Function Sample application expression Value

+ (+ 1 2 3 4 5) 15
– (– 5 8 2) –5
* (* 39 45 29 2) 101790
/ (/ 20 4 2) 2.5
expt (expt 64 1/2) 8
string-length (string-length "Program By Design") 17
string-append (string-append "Hello " "World") "Hello World"

Table 1: Some numerical and string BSL functions

written using infix notation. This means that the operator is written in the middle
of the operands. For example, the sum of 1, 2, 3, 4, and 5 is written as 1 + 2 +
3 + 4 + 5. A function application, on the other hand, is written operator first and
then in parentheses the operands. For example, applying f to 3 is written as f(3).
In contrast, BSL uses prefix notation. That is, inside parentheses we first write the
operator and then the operands. Observe that the basic operations are functions, and
therefore, BSL treats them no differently than other functions (i.e., they are written
using prefix notation). Thus, the sum of 1, 2, 3, 4, and 5 is written as (+ 1 2 3 4
5) and applying f to 3 is written as (f 3). Always remember that when you want
to apply a function to some arguments, you must put in parentheses the function
first and then the arguments. As you can see, the translation from the syntax used
in Mathematics textbooks to BSL syntax is straightforward. One of the advantages
of BSL syntax is that it may be less typing as is the case for the sum of 1, 2, 3, 4,
and 5. More importantly, however, is that ambiguity cannot arise. For example, in
Mathematics syntax what is the value of 10 + 4 * 5? It depends if you add or you
multiply first. You probably remember operator precedence and believe the correct
value is 30. In BSL, we do not rely on remembering operator precedence. Prefix
notation and the use of parentheses always make it clear what arguments a function
is applied to. If the value of 10 + 4 * 5 is 30, then we write (+ 10 (* 4 5)) in
BSL. Observe that BSL syntax clearly communicates to the readers of our code that
the arguments to * are 4 and 5 and that the arguments to + are 10 and the value of
(* 4 5).
This last example explicitly shows us that application expressions may be nested.
In (+ 10 (* 4 5)), (* 4 5) is nested. There is no limit to the number or the depth
of nested expressions as subtly suggested by the above production rule for application
expressions. Always keep in mind that in an application expression for every opening
parenthesis, there must be a matching closing parenthesis. Furthermore, remember
that a closing parenthesis cannot simply be written anywhere in an application
expression. A closing parenthesis must appear after the last argument to a function.
Otherwise, the meaning of the application expression (i.e., what it evaluates to)
changes. For example, the following two expressions have different meanings:
(* (+ 4 1) 7 (- 15 5)) (* (+ 4 1 7) (- 15 5))
10 1 The Science of Problem Solving

Fig. 4 Accessing DrRacket’s Help Desk

In the first, the last argument to + is 1 and 7 is an argument for *. It evaluates to 25.
In the second, the last argument to + is 7 and 7 is not an argument for *. It evaluates
to 120. As you can see, the meaning of the above expressions is different. Always be
careful about where you close your parentheses.
There is one more thing you may be wondering. Do all functions take as input an
arbitrary number of expressions like + and *? The answer to this is no. It depends
on the function that is being applied. For example, string-length only takes one
input. If you are not sure if a function accepts an arbitrary number of arguments,
you have two options. You can try using it in a program or in the interactions area.
Consider string-append from Table 1. Can we give it more than two arguments?
Let us try it by typing after the prompt, >, in the interactions area the following:
(string-append "Hello" " " "World" "!" " " "I am DrRacket.")
After hitting Enter, the following value is printed:
"Hello World! I am DrRacket."
The answer to our question is that we can give string-append more than two
arguments. Below the output string, you get the DrRacket prompt again. DrRacket
is ready to evaluate another expression for you.
The second option is to consult DrRacket’s Help Desk as illustrated in Fig. 4.
A browser pops up and you can type BSL in the search box. After clicking on BSL,
5 Definitions and Interactions Areas Differences 11

you are taken to the documentation page for the BSL language. On this page a search
for string-append yields
(string-append s t z ...) → string

s : string
t : string
z : string

Concatenates the characters of several strings.

> (string-append "hello" " " "world" " " "good bye")
"hello world good bye"
The . . . means that the function takes as input an arbitrary number of arguments.
The arrow means returns. After the arrow, we find the type of value returned. Further
down, we see the type of each input: all are string type. The types of the inputs and
the type of the output define the signature of the function. It clearly informs anyone
who wishes to use string-append that if they provide strings as input, the function
returns a string as its value. In terms of high school Mathematics, you may think of
the input types as the domain of the function and the return type is the range of the
function. Finally, at the bottom we see a brief description of the function’s purpose
and an example of its use. Whenever you are unsure if BSL has a function or you are
unsure of how to use a BSL function, you may look for it in DrRacket’s Help Desk
to determine if it exists and, if so, the signature tells you the expected inputs and the
expected output.

5 Definitions and Interactions Areas Differences

As you have seen, expressions written in both the definitions and the interactions
areas can be evaluated. In the definitions area you need to click RUN. In the interactions
area you need to hit Enter. So, why are there two areas?
As stated before, the two areas are used for different purposes. In the definitions
area, we write programs. These are solutions to problems that we may save and
run multiple times. In the interactions area, we ask DrRacket to evaluate one-time
expressions, that is, expressions that we are not interested in running multiple times.
The difference may seem subtle, but as you advance through the first few chapters
of this book, the differences will become better delineated.
A good rule of thumb to follow is that if you wish to save the expressions you
write to solve a problem, then you write them in the definitions area. If you only
want to quickly have DrRacket evaluate an expressions for you, then write it in the
interactions window.
12 1 The Science of Problem Solving

Fig. 5 Saving your work

6 Saving Your Work

It is an unfortunate fact that computers may crash. When this happens your unsaved
work is likely to be lost. Therefore, it is important that you regularly save your work.
It is a good idea for you to create a directory for the programs you will develop as
you read this book. You may even want to create subdirectories for each chapter.
This is an effective way to keep your files organized.
To save your work go the File menu. If it is the first time saving your cur-
rent program, click on Save Definitions As... as illustrated in Fig. 5. After
this, find the directory you wish to save your work in and enter a filename. If you
choose the filename Exercise-2-Chapter-4, your work is saved in a file named
Exercise-2-Chapter-4.rkt. The .rkt extension is automatically added to indi-
cate that it is a Racket file. In reality, it is a BSL file, but it is saved as a Racket
file given that BSL is a subset of Racket. If it is not the first time saving the current
program, click on Save Definitions inside the File menu. Alternatively, you
may simply use Ctrl+S.

7 Error Messages

Errors are part of the problem solving process and even the most experienced problem
solvers write programs with errors in them. Do not panic when you get an error
message. Remember the famous saying errāre hūmānum est (to err is human6). The
error message is DrRacket’s attempt to help you diagnose the error.

6
The complete saying, coined by Alexander Pope in his Essay on Criticism, is to err is human; to
forgive, divine.
7 Error Messages 13

An important lesson to absorb is that error messages help us diagnose and correct
bugs7 in our programs. An error message itself does not diagnose the bug nor does it
tell us how to fix the bug. That is left to us as program developers, because only the
program developers really know the intended meaning of expressions in a program.
DrRacket approximates where an error occurs, but the programmer must diagnose
and remedy the bug.

7.1 Grammatical Errors

As you now know, essays and programs are both written following the production
rules of a grammar. The English grammar tells you how to write a valid English
expression. Similarly, the BSL grammar tells you how to write valid BSL programs.
When you write a valid BSL program and click on RUN, DrRacket evaluates your
program and prints the answer.
If you fail to write a valid BSL program and click on RUN, DrRacket prints an
error message. When DrRacket detects a grammatical error, an informative error
message is printed in the interactions window. The message details the grammatical
error. For example, type the following in the definitions area:
"Hello World! I am DrRacket.
After clicking RUN, the following error message appears in the interactions window:
read-syntax: expected a closing '"'
The error message states that it was expecting to find a closing double quote. As you
have probably already observed, the opening double quote does not have a matching
closing double quote. Always make sure to read error messages, because they are
usually helpful in determining the cause of errors.
Grammatical errors may also occur writing application expressions. Type the
following program:
(sting-append "I’m done" " " "and going home!")
After clicking RUN, the following error message is displayed:
sting-append: this function is not defined
DrRacket is telling us that it does not know the function sting-append. We can
clearly see that the name of the function is misspelled and we correct it as follows:
(string-append "I’m done" " " "and going home!")

7
The term computer bug stems from a moth found trapped in a computer relay in 1947. The bug was
removed and taped to the log book. The log book along with the moth are part of the Smithsonian
National Museum of American History collection in Washington D.C.
14 1 The Science of Problem Solving

Fig. 6 Grammatical Error

Another common grammatical error is failing to properly write an application

expression with balanced parentheses. For instance, type the following program in
the definitions area to compute (4 * 5) + (50 / 10):
(+ (* 4 5 (/ 50 10))
After clicking RUN, the following error message is displayed:
read-syntax: expected a ‘)‘ to close ‘(‘
DrRacket is informing us that there is a missing closing parenthesis. Observe that in
the definitions window, the opening parenthesis before + is highlighted as displayed
in Fig. 6. DrRacket is letting us know which opening parenthesis it could not match.
If we do not think carefully about the error, we may be tempted to fix the bug by
adding a closing parenthesis at the end as follows:
(+ (* 4 5 (/ 50 10)))
After clicking RUN, the following error message is displayed:
+: expects at least 2 arguments, but found only 1
This is a type error. We will discuss type errors in the next subsection. After some
thought, we realize that we placed the missing closing parenthesis in the wrong place.
DrRacket could not find a closing parenthesis for the first opening parenthesis. The
bug, however, is that the parenthesis before * was not properly closed. We fix the
problem as follows:
(+ (* 4 5) (/ 50 10))
After clicking RUN, the correct value for (4 * 5) + (50 / 10), 25, is printed in
the interactions area.
7 Error Messages 15

Fig. 7 Runtime error

Another common grammatical mistake is to place an expression inside paren-

theses when a function is not applied. This mistake is common among beginners
and it is one you need to be careful about. For instance, consider typing after the
DrRacket’s prompt:
(* (10) (30))
Figure 7 displays the error message after hitting Enter. Observe that 10 is highlighted
in pink. DrRacket is telling us that it did not find a function after the opening
parenthesis for the first operand. Instead, it found 10. 10 is not a function and,
therefore, should not be in the operator position. In fact, 10 is a number and a
number by itself is not written inside parenthesis when it is intended as an argument
to a function.

* Ex. 6 — Type and fix the following buggy expressions:

1.(+ (* 1 2)) 7 (* 5 10))
2.((* 9 6 4))
3.(string-len "Isa and Sena")
4.(sine (/ pi 2))
5.(squareroot (str-length "Yes!"))
16 1 The Science of Problem Solving

7.2 Type Errors

Another kind of error that we may make is called a type error. Type errors mostly
occur in two ways. The first is when the wrong number of arguments are provided
to a function. Let us revisit a program from the previous subsection.
(+ (* 4 5 (/ 50 10)))
After clicking RUN, the following error message is displayed:
+: expects at least 2 arguments, but found only 1
DrRacket is telling us that it believes that we are not correctly using +. We are not
providing the right number of arguments to +. This is a very common mistake done
by beginners and it is important that we keep this in mind as we implement solutions
to problems.
The second is when the wrong type of argument is given to a function. For
example, type the following after the prompt in the interactions area:
(string-append "My lucky number is: " 8)
After clicking RUN, the following error message is displayed:
string-append: expects a string as 2nd argument, given 8
The string-append function expects all its inputs to be strings. In this example,
the second argument is a number. To fix this bug, we must make the second argument
a string as follows:
(string-append "My lucky number is: " "8")
After clicking RUN, the string is printed in the interactions area. It is worth highlight-
ing that "8" is not the same as 8. The former is a string and the latter is a number,
that is, they are different types of data.
Providing a function with the wrong type of input is a very common mistake.
Therefore, type theory is an important and growing discipline within Computer
Science. As we progress through this textbook, we will learn design techniques to
help us avoid type errors. Furthermore, we will learn how to exploit the type of data
being processed to design solutions (and programs) to problems.

* Ex. 7 — Type and fix the following buggy expressions:

1.(string-length 9865)
2.(number->string "7615461")
3.(sqr "20")
4.(add1 (add1 (add1 "7"))))
5.(sub1 (add1 (string->number (string-length "100"))))
6.(string-append ’Animating "Programs" ’Textbook)
7 Error Messages 17

Fig. 8 Runtime error

7.3 Runtime Errors

You need to be aware of a third kind of error called a runtime error. Unlike gram-
matical and type errors, runtime errors are not detected until after an evaluation has
started. For example, type the following program in the definitions area:
(* 5 0)
(+ 5 0)
(/ 5 0)
(- 5 0)
The results after clicking RUN are displayed in Fig. 8. We can observe that the first
two expressions are evaluated and their results are printed. The third expression is
highlighted in pink signalling that an error has been detected during its evaluation.
The error message in the interactions window informs us that there has been an
attempt to divide by 0. The fourth expression is highlighted in black signalling that
it has not been evaluated.
Our programs may also contain runtime errors that do not generate an error
message. For example, write the following program to compute f(4), where f(x)
= 2x2 + 10x + 3:8
(+ (* 2 (sqr 4)) (* 10 4) 4)
After clicking RUN, no error message is generated and 76 is printed in the interactions
window. Observe, however, that 76 is not the value of f(4). These are probably the
hardest errors to fix, because no error message is generated. In this example, the bug
is rather easy to spot. The last 4 in the expression is a typo and should be 3.

8
sqr is a BSL function that squares its input. Look it up in the Help Desk!
18 1 The Science of Problem Solving

Fig. 9 Testing in BSL

To help protect ourselves against this type of error, most programming languages,
including BSL, allow programmers to write unit tests. A unit test validates that two
different expressions evaluate to the same value. In order to write unit tests, we need
to expand our grammar as follows:
program ::= {expr | test}∗
test ::= (check-expect expr expr)
Now, a program is 0 or more expressions or tests. The curly braces are used to
delimit the scope of the Kleene star and should not be typed as part of any program.
In BSL, we write a unit test by typing inside parentheses check-expect and two
expressions. The first expression is the one that we wish to test and it provides what is
called the actual value. The second expression provides what is called the expected
value of the first expression.
We can now rewrite our buggy program as follows:
(+ (* 2 (sqr 4)) (* 10 4) 4)

(check-expect (+ (* 2 (sqr 4)) (* 10 4) 4) 75)

Figure 9a displays DrRacket after clicking run. When tests fail, you get a pop-up
window indicating the number of tests evaluated and the number that passed. For
each failed test, the actual and expected values are displayed. In addition, the line
number of the error in your program is reported.9 In Fig. 9a, we see that 1 test was
evaluated and 0 tests passed. The only failed test is at line 3. We correct our program
as follows:
(+ (* 2 (sqr 4)) (* 10 4) 3)

(check-expect (+ (* 2 (sqr 4)) (* 10 4) 3) 75)

Figure 9b displays the result after pressing RUN. There is no pop-up window with
failed tests and DrRacket in the interactions area indicates that all tests pass.
It may or may not have bothered you that we had to type the same expression
twice. This may be tolerable when you only need to type one expression twice. Most
programmers, however, would find it extremely annoying and, more importantly,

9
You may configure DrRacket to display line numbers in the View menu.
8 What Have We Learned in This Chapter? 19

Fig. 10 Rectangle of width 3 and height 4

error-prone to have to retype multiple expressions multiple times. Furthermore,

evaluating the same expression multiple times (i.e., once for every time it is typed)
may make programs slower. In the next chapter, we will learn techniques to avoid
the repeated typing of the same expression to write tests.

** Ex. 8 — Fix the runtime errors in the following program:

"I have square with a perimeter of 20"
"The length of a side is: "
(/ 20 0)
"The area is:"
(+ 20 20)

(check-expect (* 4 (/ 20 0)) 20)

(check-expect (+ 20 20) (sqr (/ 20 0)))

** Ex. 9 — For a rectangle displayed in Fig. 10, write a program that computes
1.The rectangle’s area
2.The rectangle’s perimeter
3.The rectangle’s diagonal distance from the lower left corner to the top right
corner
Make sure you write unit tests to validate your computed values.

8 What Have We Learned in This Chapter?

The important lessons of Chap. 1 are summarized as follows:

• We write programs using a programming development environment.
20 1 The Science of Problem Solving

Fig. 11 Chapter 1 BSL grammar

program ::= {expr | test}*
test ::= (check-expect expr expr)
expr ::= string
::= number
::= (<function>expr+)
string ::= "<character>*"
number ::= positive
::= negative
positive ::= mag
negative ::= -mag
mag ::= int
::= real
::= fraction
int ::= digit+
digit ::= 0|1|2|3|4|5|6|7|8|9
real::=digit* digit+
fraction ::= int/nonzero-int
nonzero-int ::= <1|2|3|4|5|6|7|8|9> int

• DrRacket is the IDE used in this textbook.

– The definitions area is where programs are written.
– The interactions area is where one-time expressions are evaluated and where
error messages are printed.
– To evaluate a program, click on RUN.
• A programming language is used to write solutions to problems. We call such a
solution a program.
• The syntax of a programming language may be described using a grammar. The
BSL syntax developed so far is displayed in Fig. 11.
– A program is 0 or more expressions.
– An expression can either be a number, a string, or an application expression.
– Strings are written inside double quotes.
– A number is a positive or negative integer, real, or fraction.
– An application expression is used to apply a function to its arguments. In-
side parentheses you first write the function name and then its arguments all
separated by 1 or more spaces.
• Use DrRacket’s Help Desk to determine if a function exists or to learn how to
use a function.
• In order to minimize the danger of loss, frequently save your work.
• Error messages help us determine where a bug occurs but are unable to tell us
how to fix the bug.
• Unit testing helps us find runtime errors by validating that an actual value is the
same as an expected value.
Chapter 2
Expressions and Data Types

You have learned to write expressions using numbers, strings, and applications.
Although this has enabled you to write programs, we need a richer set of data
types to make problem solving easier. This chapter introduces a broader set of data
types that are part of BSL. These are divided into two broad categories: primitive
and compound. A primitive data type is any type of data that is indivisible. For
example, numbers are a primitive data type. A compound data type is one that
contains multiple pieces of data. A piece of compound data can be divided into its
component parts. For example, a coordinate on the two-dimensional Cartesian plane
is compound data. Every coordinate is composed of an x and a y value. The new
data types introduced in this chapter are Boolean, symbol, character, and image. This
chapter also revisits the number and string data types to present them in more detail.
In addition, this chapter introduces how to define constants to avoid having to type
the same expression multiple times in programs.
Consider computing the area and the perimeter of the rectangles in Fig. 12.
Thinking about the problem leads us to conclude that area is computed by multiplying
the width and the length of a rectangle and perimeter is computed by adding the
doubling of the width and the doubling of the length. Based on this analysis, Fig. 13
displays a solution in the form of a program. The first thing you notice is that
comments are used to document a program. Anything after a ; is a comment in
BSL. Specifically, comments are used to document how area and perimeter are
computed. This helps the programmer and anyone else who reads the program to
understand the solution to the problem. Comments are also used to identify the
computation performed for each rectangle. Finally, unit tests are written to make
sure that the computations yield the correct values. Observe that there is one test for
each computation.
The program in Fig. 13 contains many repeated expressions. This is considered
poor programming, because repetitions are boring and lead to errors. Quite frankly,
no one wants to do the same thing over and over again. Therefore, we need a
mechanism to avoid repetitions in our programs. One idea that is promising is to
compute a value once. Instead of (typing and) evaluating each expression multiple

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 21

M. T. Morazán, Animated Problem Solving, Texts in Computer Science,
https://doi.org/10.1007/978-3-030-85091-3_2
22 2 Expressions and Data Types

Fig. 12 Three rectangles

(b) Rectangle 2.

(a) Rectangle 1.

(c) Rectangle 3.

4 2

5
2

times, use the value of the expression multiple times. For this we need to learn a
little more BSL syntax to define variables. Variables are used to store the values of
expressions.

9 Definitions

To define and use variables, the BSL syntax is extended as follows:

program ::= {expr | test | def}∗
def ::= (define <variable name> expr)
expr ::= <variable name>
A program consists of 0 or more expressions, tests, and definitions. A definition
starts with an opening parenthesis, followed by the keyword define, followed by a
variable name, followed by an expression, and ends with a closing parenthesis. A
variable name consists of a number of characters without spaces.10 An expression
may be a variable name. This is how variables are used.
A definition binds the variable to the value of the expression. For example, type
the following in the definitions area:
(define DELTA-X 5)

(define i (* 4 25))

10
Some characters are not allowed such as #.
9 Definitions 23

Fig. 13 Version 1 of the area and perimeter program for the rectangles in Fig. 12
;; Area = width * height
;; Perimeter = (2 * width) + (2 * height)

;; Area for rectangle 1

"The area for rectangle 1 is: "
(* 3 4)
(check-expect (* 3 4) 12)

;; Perimeter for rectangle 1

"The perimeter for rectangle 1 is: "
(+ (* 2 3) (* 2 4))
(check-expect (+ (* 2 3) (* 2 4)) 14)

;; Area for rectangle 2

"The area for rectangle 2 is: "
(* 2 5)
(check-expect (* 2 5) 10)

;; Perimeter for rectangle 2

"The perimeter for rectangle 2 is: "
(+ (* 2 2) (* 2 5))
(check-expect (+ (* 2 2) (* 2 5)) 14)

;; Area for rectangle 3

"The area for rectangle 3 is: "
(* 2 2)
(check-expect (* 2 2) 4)

;; Perimeter for rectangle 3

"The perimeter for rectangle 3 is: "
(+ (* 2 2) (* 2 2))
(check-expect (+ (* 2 2) (* 2 2)) 8)

These two definitions bind DELTA-X to 5 and bind i to 100. After clicking RUN, typing
the expression and pressing Enter at the prompt in the interactions area returns 5
and typing the expression i and Enter at the prompt returns 100. Defining variables
is how typing and evaluating the same expression more than once is avoided. They
store the value of an expression allowing programmers to avoid multiple evaluations
of the same expression. Instead, the defined variable is used multiple times.
By convention, two types of variables are distinguished: those whose value may
change and those whose value may not change. The latter are called constants. Both
are defined using the def syntax. The convention in this textbook is to use uppercase
letters for constants and lowercase letters otherwise. In the above example, DELTA-X
is considered a constant and i is not considered a constant. In BSL, there is no such
convention and it is up to the programmers to use this convention.
Armed with the def syntax, the program in Fig. 13 may be rewritten as displayed
in Fig. 14. Observe that for each value computed a constant is defined. Instead of
using generic variable names like x and i, the constants are given names that inform
24 2 Expressions and Data Types

Fig. 14 Version 2 of the area and perimeter program for the rectangles in Fig. 12
;; Area = width * height
;; Perimeter = (2 * width) + (2 * height)

;; Area for rectangle 1

(define AREA1 (* 3 4))

(check-expect AREA1 12)

;; Perimeter for rectangle 1

(define PERIM1 (+ (* 2 3) (* 2 4)))

(check-expect PERIM1 14)

;; Area for rectangle 2

(define AREA2 (* 2 5))

(check-expect AREA2 10)

;; Perimeter for rectangle 2

(define PERIM2 (+ (* 2 2) (* 2 5)))

(check-expect PERIM2 14)

;; Area for rectangle 3

(define AREA3 (* 2 2))

(check-expect AREA3 4)

;; Perimeter for rectangle 3

(define PERIM3 (+ (* 2 2) (* 2 2)))

(check-expect PERIM3 8)

any reader what they represent. This is a good practice and it is a practice that
is followed in this textbook. It is a habit that every programmer is encouraged to
follow. The value of each constant is defined by an expression. Subsequently, the
variable is used to refer to the value of the expression instead of reevaluating the
same expression again. For instance, in the unit tests the constants are used. Observe
also that it is no longer necessary for the program to return each area and perimeter
value, which eliminates the need to clutter the interactions area with printed values.
If a value needs to be examined, type the name of the variable at the prompt and hit
Enter.
Which version of the program is clearer? Most people will argue that the program
in Fig. 14 is clearer. Any interested reader can see what the constants represent.
Furthermore, they can easily see that the tests are there to make sure that the values
of the constants are correct. Admittedly, for this rather small and simple program, it
may not seem significant. Nonetheless, it is important for you to realize now that as
the size of programs grows, it is harder for a programmer (and a reader) to remember
10 Numbers 25

Fig. 15 An error in unit testing using check-expect with real numbers

all the details. Therefore, the goal of a program is twofold. First, a program must
correctly solve a problem. Second, a program must clearly communicate the solution
to the problem. Both of these goals are equally important and you must adhere to
them as responsible problem solvers. It is never acceptable to get the code to work
without documenting the solution to the problem. Think about this carefully. If you
were working on a team programming project, would you trust code that you do not
understand?

10 Numbers

Recall that a number is either an integer, a real, or a fraction. Number is a primitive

type (of data) in BSL. Chapter 1 presented examples of expressions using integers and
how to write unit tests using check-expect. Now, consider writing a program to
compute the circumference of a circle with a radius of 7. Recall that the circumference
of a circle is given by 2πr. Type the following program to solve the problem:
;; The circumference of a circle with radius r is 2 * pi * r

;; The circumference of a circle with radius 7

(define CIRCUM-7 (* 2 pi 7))

(check-expect CIRCUM-7 43.98)

In BSL, pi stores the value of π. Figure 15 displays the failed test message after
clicking RUN. The message is informing us that a real number (called inexact
number in BSL) cannot be checked using check-expect. The reason is that numbers
26 2 Expressions and Data Types

in computer are not exactly the same as numbers in Mathematics. In a computer, a

real number is limited to a finite number, n, of decimal places. Any decimals after
the nth one is discarded. This means that computations using real numbers in BSL
may contain errors. This is why they are called inexact. Examining the value on
CIRCUM-7 at the prompt reveals that it is #i43.982297150257104.
DrRacket is nice enough to print the value prefixed with #i to warn that it
is inexact and, therefore, may contain an error. This problem with real number
computations is common among programming languages. Indicating that a value is
inexact, however, is not common among other programming languages. When the
magnitude of a real number is very small or very large, DrRacket prints it using
scientific notation. For example, #i4.3290784900439874e-010 corresponds to the
inexact number #i4.3290784900439874 x 10-10 .
In order to test computations involving real numbers, the grammar production
rules for test must be expanded as follows:
test ::= (check-within expr expr expr)
As with check-expect, the first two expressions are, respectively, for the value
being tested and the expected value. The third expression defines the error tolerance.
For the test to pass, the actual value must be at most the tolerance away from the
expected value.
The program to compute the circumference of a circle with a radius of 7 is updated
to
;; The circumference of a circle with radius r is 2 * pi * r

;; The circumference of a circle with radius 7

(define CIRCUM-7 (* 2 pi 7))

(check-within CIRCUM-7 43.98 0.01)

The unit test states that the value of CIRCUM-7 may be at most 0.01 away from 43.98
(the expected value). After clicking RUN, the only test passes.
In summary, use check-expect to test integers and fractions and use check-
within to test reals. The example above also demonstrates that bugs do not only
appear in expressions. Bugs may also occur in tests.

* Ex. 10 — In Fig. 14, the expression (* 2 2) appears several times. Rewrite

the program to eliminate its repetition.

** Ex. 11 — A horse riding academy has a rectangular field of 25 meters by

5 meters. They wish to convert it to a horse riding outdoor classroom. In the
middle of the field, they will fence off a 6 meters by 5 meters rectangle for
students to sit and listen to lectures. Write a program to compute the area of
the field left for horse riding. Make sure to write unit tests and document your
solution.
11 Strings and Characters 27

* Ex. 12 — Run the following program:

(define NUM1 #i0.51467e-99)
(define NUM2 #i999999999.42477796076938)
(define NUM3 #i0.8724e-84)

(check-within (* NUM1 NUM2 NUM3) (* NUM3 NUM2 NUM1) 0)

We know from Mathematics courses that multiplication is commutative. Why
does the test fail?

11 Strings and Characters

In Chap. 1, a string is described as anything that is inside double quotes. The word
“anything" suggests that there may be more than one piece of data in a string. In fact,
a string is our first example of a compound data type. Compound data is constructed
using several pieces of data. For example, the string "cat" may be constructed using
the strings: "c", "a", and "t". We build a string by using double quotes. Putting cat
inside double quotes glues together "c", "a", and "t" as a single piece of data. This
is exactly the same as typing (string-append "c" "a" "t"). In other words,
the use of double quotes is simply shorthand for using string-append. Functions
that build compound data, like string-append, are called constructors.
When data is compound, it is possible to access some or all of its components.
Functions that extract a component from an instance of compound data are called
selectors. For strings, we can use the following BSL function to extract substrings:
;; string N [N] → string
;; Purpose: Extract from s the substring starting in
;; position i and ending in position j-1
(substring s i j)
The signature is stating that the function substring requires at least two inputs. The
first is a string and the second is a natural number less than or equal to the length of
s.11 The [] indicates that the third argument is optional. The optional third argument
is a natural number, j, such that 0<=j<=length of s. Think of i and j as defining
the interval [i..j) and substring as extracting the substring from position i to
position j. Not providing the third argument is shorthand for j being the length of the
s. For example, (substring "Wow!" 2) is shorthand for (substring "Wow!"
2 4). Finally, it is important to note that strings’ positions start at 0 (not 1).
The following program illustrates that strings are compound data:

11
A natural number is an integer greater than or equal to 0.
28 2 Expressions and Data Types

(define CAT "cat")

(check-expect CAT (string-append "c" "a" "t"))
(check-expect (substring CAT 0 1) "c")
(check-expect (substring CAT 1 2) "a")
(check-expect (substring CAT 2 3) "t")
(check-expect (substring CAT 3 3) "")
(check-expect (substring CAT 0) CAT)
(check-expect (substring CAT 1) (substring CAT 1 3))
The first test illustrates that "cat" is shorthand for (string-append "c" "a"
"t"). The second, third, and fourth tests, respectively, demonstrate that substrings
of length 1 at the 0th, 1st, and 2nd positions of CAT are "c", "a", and "t". The
fifth test demonstrates that the substring starting at 3 and not including 3 is the empty
string. The sixth test demonstrates that CAT is the same as the substring starting at
0. Finally, the seventh test demonstrates that the substring starting at 1, "at", is the
same as the substring in positions [1..3).
In fact, strings are not really made up of substrings. Every element of a string is a
character (abbreviated as char). Therefore, substring in reality is a function that
converts the selected characters into a string for the programmer. This function was
created to make programming easier.
Expressions may evaluate to characters. Therefore, we need new syntax for char-
acters:
expr ::= \#<character constant>
To denote characters, they are typed by prefixing them with \# followed by a character
constant. For our purposes, a character constant is any single character you can
type using a keyboard. As numbers, characters are primitive data and cannot be
subdivided into components. To actually access the characters of a string, we may
use the following function:
;; string N → char
;; Purpose: Extract the iith char from s
(string-ref s i)
The index i must satisfy 0<=i<=length of s. The following program illustrates the
use of string-ref:
(define DOG "dog")

(check-expect (string-ref DOG 0) \#d)

(check-expect (string-ref DOG 1) \#o)
(check-expect (string-ref DOG 2) \#g)

* Ex. 13 — There are many useful BSL functions to manipulate strings. Be-
come familiar with, for example, string-ith and string-length by looking
them up in DrRacket’s Help Desk. Write a program that appends these 3
11 Strings and Characters 29

string: "Program", "By", and "Design". Include tests to illustrate the follow-
ing:
1.The appended strings result in "Program By Design".
2.The length of the appended strings is 17.
3."P" is the substring in positions [0..1].
4.\#P is the first character of the appended strings.
5."By" is the substring in positions [8..10].
6.The last character of the appended strings is \#n

* Ex. 14 — What happens if we add the following tests to the CAT program
above? Why does each test pass or fail?
(check-expect (substring CAT 3) "")
(check-expect (substring CAT 4) "")
(check-expect (substring CAT 0 3) CAT)
(check-expect (substring CAT 0 4) CAT)
(check-expect (substring CAT -1 1) CAT)
(check-expect (substring CAT 3 1) CAT)

* Ex. 15 — Consider the following program fragment:

(define MARCO "Marco")

(define OCRAM (string-append (substring MARCO . . .)

(substring MARCO . . .)
(substring MARCO . . .)
(substring MARCO . . .)
(substring MARCO . . .)))

(check-expect OCRAM "ocraM")

Complete the program by filling in each . . . with an expression. Make sure the
test passes.

* Ex. 16 — Consider the following program:

(define GREETING1 "hello") ;; lowercase greeting
(define GREETING2 "HELLO") ;; uppercase greeting

(check-expect GREETING1 GREETING2)

The test fails after running the program. Rewrite the test by applying a function
to either GREETING1 or GREETING2. Find a function to change the case of a
string.

* Ex. 17 — Write a program that adds the lengths of "Sena" and "Isa".
30 2 Expressions and Data Types

12 Symbols

Symbols are like strings that are not decomposable. That is, symbol is a primitive
type like number. Symbols are used when we are not interested in manipulating or
accessing the characters as can be done with strings. Symbols require extending the
BSL grammar explored. The following is a new expr production rule for symbols:
expr ::= '<character>+
An expression may be a symbol, which is written as a ' followed by one or more
characters with no spaces.
DrRacket denotes most symbols as a ' followed by the name of the symbol. For
example, entering (string->symbol "Apple") at the prompt returns a symbol
that is printed as 'Apple to the screen. Entering (string->symbol "1024")
returns a symbol, but it is printed differently. Numeric symbols are printed as ',
followed by |, followed by the number, and ending with |. Therefore, the symbol for
"1024" is printed as '|1024|.
Not surprisingly, a symbol is convertible to a string and vice versa. The following
program exemplifies this observation:
;; NAME PROGRAM

(define NAME 'Joshua)

(define NAME2 'Sachin)

(define STR-NAME (symbol->string NAME))

(define STR-NAME2 (symbol->string NAME2))

(check-expect STR-NAME "Joshua")

(check-expect STR-NAME2 "Sachin")

(define JOSH (string->symbol STR-NAME))

(define SACH (string->symbol STR-NAME2))

(check-expect JOSH NAME)

(check-expect SACH NAME2)
In this program two symbols are converted to strings using symbol->string and
then converted back to symbols using string->symbol. Read about these functions
in DrRacket’s Help Desk.
The functions symbol->string and string->symbol are interesting because
they are inverses of each other. Informally, this means that they reverse each other. In
the program above, for example, symbol->string converts 'Josh into "Josh" and
string->symbol converts "Josh" back to 'Josh. More formally, you may recall
from high school Mathematics that two functions are inverses if when composed the
original input is returned. You may recall it from your Mathematics textbook as
(f ◦ g)(x1 ) = x1 = (g ◦ f)(x1 )
13 Booleans 31

Do not panic if you do not remember or recognize function composition and inverses
right now. These will be discussed in more detail later in this book. We mention them
here, because they are important in programming. For example, you may compress
and uncompress your files, encrypt and decrypt your sensitive data, or marshal–
unmarshal your data for computer communication. All of these are achieved by a
pair of functions that are inverses of each other.

** Ex. 18 — Write a program that defines a symbol for your first name and
that computes a symbol that is your first name reversed without the last letter.
Hint: Make a plan for the values that need to be computed to solve this problem.
For example, the program first converts the symbol to a string, then computes a
new string for your reversed name without the last letter, and finally computes
a symbol from that new string.

* Ex. 19 — Explain what happens and why when the following tests are added
to the name program above:
(check-expect NAME STR-NAME)
(check-expect SACH STR-NAME2)

* Ex. 20 — The functions string->number and number->string are in-

verses of each other. Write a program that illustrates this fact.

13 Booleans

Boolean is a primitive type of data. There are only two kinds of Boolean values:
#true and #false. This type of data is important because it allows to perform
computations that depend on whether conditions hold or fail to hold. For example,
in Mathematics you have studied the absolute value function:

x if x ≥ 0
absV alx =
−x otherwise

How do you compute absVal(-100)? You plug in −100 for x, evaluate the condi-
tion, −100 ≥ 0, and determine that it is false. Since this condition is false, the value
of the function is given by −(−100). Evaluating this expression tells you that the
value of absVal(-100) is 100.
Two new expr production rules for Booleans are needed:
expr ::= #true
::= #false
These rules state that a # followed by either true or false represents a Boolean
value in BSL. Nothing else is a Boolean.
32 2 Expressions and Data Types

x y x∧y x∨y ¬x
false false false false true
false true false true true
true false false true false
true true true true false

Table 2: Truth table for the basic Boolean functions

Just like numbers have basic functions (i.e., +, -, *, and /) to create new numbers,
Booleans have the following basic functions: and (also denoted by ∧), or (also
denoted by ∨), and not (also denoted by ¬). A truth table, as the one in Table 2, is
used to illustrate the four basic Boolean functions. From the table, we can infer that
1. and is true when all inputs are true and is false otherwise;
2. or is true when any input is true and is false otherwise;
3. not is true when the input is false and is false otherwise.
The above is a complete description on the basic Boolean functions. It is noteworthy
that it is not always necessary to evaluate all the arguments to and or to or. Consider
the following two expressions:
(and #false (≤ x 20)) (or #true (> i 100))
After evaluating the first argument to and, the other arguments may be ignored,
because the result is known to be false. Similarly, after evaluating the first argument
to or, the other arguments may be ignored, because the result is known to be true.

13.1 Basic Boolean Operators in BSL

BSL takes advantage of the fact that not all arguments to and and or may have to
be evaluated to make them faster. Instead of evaluating all the arguments to and or
or, BSL stops evaluating arguments to and after the first that evaluates to #false.
Similarly, BSL stops evaluating arguments to or after the first that evaluates to #true.
To achieve this, and and or are not functions in BSL. Instead, they are new types
of expressions. The BSL grammar includes the following two production rules for
expr:
expr ::= (and expr expr expr∗ )
::= (or expr expr expr∗ )
These production rules tell us that and and or expect at least two values as input. In
BSL, not is a one-input function.
13 Booleans 33

Fig. 16 Unevaluated expressions in Boolean table validation program

We can now write a program to validate Table 2 as follows:

;; Basic Boolean Functions Validation Program

(define AND-FF (and #false #false))

(define AND-FT (and #false #true))
(define AND-TF (and #true #false))
(define AND-TT (and #true #true))
(define OR-FF (or #false #false))
(define OR-FT (or #false #true))
(define OR-TF (or #true #false))
(define OR-TT (or #true #true))
(define NOT-F (not #false))

(define NOT-T (not #true))

(check-expect AND-FF #false)
(check-expect AND-FT #false)
(check-expect AND-TF #false)
(check-expect AND-TT #true)
(check-expect OR-FF #false)
(check-expect OR-FT #true)
(check-expect OR-TF #true)
(check-expect OR-TT #true)
(check-expect NOT-F #true)
(check-expect NOT-T #false)
34 2 Expressions and Data Types

After running this program, all the tests pass. Observe, as displayed in Fig. 16, that
not all the expressions are evaluated. The expressions that remain unevaluated by
the tests are highlighted in black. These expressions are unevaluated, because their
value is not needed to determine the value of an and or an or expression. In fact,
there is no way to test the unevaluated code. As programmers, this means that, at
best, thorough testing gives us confidence in the program. Testing, however, never
establishes that the program is correct.

** Ex. 21 — Add the following definitions and tests to the basic Boolean func-
tions validation program:
(define AND-FERR (and #false (/ 5 0)))
(define OR-FERR (and #false (string-append "Is this "
"a Boolean?")))

(check-expect AND-FER #false)

(check-expect OR-FERR #true)
What happens with each test? Why?

13.2 Predicates

Functions that return a Boolean are called predicate functions. They are useful to
determine if the input meets some condition. In BSL, built-in types have a predicate to
determine if the input is of that type. These are the predicates for the types presented
in this chapter:
Type Predicate
number number?
string string?
character char?
symbol symbol?
boolean boolean?
image image?
For example, (string? "Hi there!") evaluates to #true and (boolean?
103167) evaluates to #false.
You have used predicates in high school Mathematics. For instance, <, >, ≥, ≤,
and = are all predicates. They are all functions in BSL with the following names: <,
>, >=, <=, and =. These functions all require 1 or more inputs. Typing (< 10) at
the prompt returns #true. Why does this make sense? It turns out that a numerical
predicate returns #true unless one of its input makes it #false. There is nothing
in < 10 that makes it #false, and therefore, this expression evaluates to true.
What does < 10121831 evaluating to #true mean? This is the BSL expression for
10 < 12 < 18 < 31. Remember that BSL expressions use prefix, not infix, notation.
It evaluates to #true, because 10 < 12 ∧ 12 < 18 ∧ 18 < 31 is true.
13 Booleans 35

Strings share an important property with numbers: they are ordinal. This means
that they may be ordered. Strings may be lexicographically ordered (i.e., alphabet-
ically ordered). Therefore, we can ask if one string is less than another or if one
string is greater than or equal to another. We cannot, however, use numerical pred-
icates with strings. The corresponding string predicates are string<?, string>?,
string<=?, string>=?, and string=?. Observe that the ? suggests that a question
is being asked. For example, (string<? "a" "b" "c") is asking if "a" comes
before "b" and "b" comes before "c".
Unlike strings, symbols are not an ordinal type. This means that there are no
functions to compare symbols like > or string<?. There are two predicates to test
symbols. The predicate symbol? tests if its input is a symbol. The predicate eq?
may be used to test if two symbols are equal. In fact, eq? does not exclusively
work with symbols. For example, we can write (eq? 1 1) and (eq? "Alpha
Centauri" "Alpha"). The first evaluates to #true and the second evaluates to
#false. Functions, like eq?, that work for many types of inputs are called generic
functions. We will explore generic functions later in this textbook. For now, we use
=, string=?, and eq?, respectively, to test for numeric, string, and symbol equality.
BSL predicates for Booleans include boolean? and false?. The first tests if its
input is a Boolean. The second tests if its input is #false. You may ask yourself,
why is true? not a predicate in BSL? The answer is that the creators of BSL cannot
possibly implement every conceivable function. Therefore, they must choose what
functions to provide. One criterion that is used to decide is whether or not a value
may be computed using other functions. Can we write expressions to determine if
a variable is #true? To answer this question, we must carefully think about what it
means to be #true. First, the value must be a Boolean. Second, the value must not
be #false. With this insight, we can write an expression to determine if a variable is
true using not, and, and false?. This is a sample program to test our observations:
(define A-STRING "This is not true.")
(define A-SYMBOL 'not-true)
(define A-NUMBER 87)
(define A-BOOL #true)
(define A-BOOL2 #false)

(define IS-TRUE-A-STRING (and (boolean? A-STRING)

(not (false? A-STRING))))
(define IS-TRUE-A-SYMBOL (and (boolean? A-SYMBOL)
(not (false? A-SYMBOL))))
(define IS-TRUE-A-NUMBER (and (boolean? A-NUMBER)
(not (false? A-NUMBER))))
(define IS-TRUE-A-BOOL (and (boolean? A-BOOL)
(not (false? A-BOOL))))
(define IS-TRUE-A-BOOL2 (and (boolean? A-BOOL2)
(not (false? A-BOOL2))))

(check-expect IS-TRUE-A-STRING #false)

(check-expect IS-TRUE-A-SYMBOL #false)
36 2 Expressions and Data Types

(check-expect IS-TRUE-A-NUMBER #false)

(check-expect IS-TRUE-A-BOOL #true)
(check-expect IS-TRUE-A-BOOL2 #false)
Observe that function composition is used to determine if a variable is #true. That
is, the values returned by boolean? and not are inputs to and and the value returned
by false? is input to not. We begin to see that function composition is useful to
design programs.

* Ex. 22 — BSL provides many useful predicates to us. What are the signature
and purpose of the following predicates:
1. odd?
2. even?
3. positive?
4. negative?
5. string-uppercase?
6. string-lowercase?
7. string-numeric?
8. string-alphabetic?
9. string-contains?

* Ex. 23 — Write a program to determine the number of letters in the symbol

'abracadabra.
** Ex. 24 — Logical implication (or simply implies) is important in program
solving. It means that if some condition, ANTECEDENT, is #true, then some
other condition, CONSEQUENT, is #true. Traditionally, implies is denoted by
⇒. This is the truth table for implies:
ANTECEDENT CONSEQUENT ANTECEDENT ⇒ CONSEQUENT
false false true
false true true
true false false
true true true
Write a program to validate the truth table for implies.

*** Ex. 25 — The nand (not and) boolean function has a property called func-
tional completeness. This means that any Boolean expression may be written
only using nand expressions. The truth table for nand is
A B (nand A B)
false false true
false true true
true false true
true true false
The value of nand is given by the expression (not (and A B)). We can
compute (not X) using a nand expression as follows: (not (and X X)).
14 Images 37

Fig. 17 Examples of outline and solid images

 An Outline Rectangle.  A Solid Circle.

Write a program to demonstrate how to compute and and or using only nand
expressions.

* Ex. 26 — Run the following program:

(define NUM1 #i0.51467e-99)

(define NUM2 #i0.8724e-304)

(check-expect (> (* NUM1 NUM2) 0) #true)

Why does the test fail?

14 Images

An image is a visual rectangular type of compound data. Every image consists of

pixels. A pixel is a picture element representing a colored dot. Every image has a
length and a width measured in pixels. In BSL, expressions may evaluate to an image.
This means that we need a new production rule for expr:
expr ::= <image>
Here <image> is a graphic that we either create or copy from elsewhere and paste
it into DrRacket. For images we shall not delve into the manipulation of pixels.
Instead, we shall focus on image manipulation as one whole value.
BSL provides us with a teachpack to create and manipulate images called
2htdp/image. To use the 2htdp/image teachpack, the first line in your program
needs to be:
(require 2htdp/image)
38 2 Expressions and Data Types

This teachpack contains many functions that are split into three broad categories:
Basic Constructors These are functions to create basic images like circles, rectan-
gles, ellipses, stars, and text.
Property Selectors These are functions to extract properties of images such as
width and length.
Image Composers These are functions that combine existing images to create new
images.
It is useful to understand two basic definitions before using the 2htdp/image
teachpack:
mode This refers to how basic geometric shapes are drawn: not filled or filled. Use
'outline or "outline" when an outline of a shape is desired. Use 'solid or
"solid" when a solid shape is desired.
image color This is a string or a symbol representing the desired color. Search for
image-color? in DrRacket’s Help Desk to see the list of predefined colors.
Figure 17 illustrates outline and solid geometric shapes. Figure 17a displays an
outline red rectangle created using the rectangle function. Figure 17b displays a
solid green circle created using the circle function.

14.1 Basic Image Constructors

To illustrate the use of the basic image constructors, consider the program in Fig. 18
to construct the images in Fig. 17. First, constants for the characteristics of the
rectangle and the circle are defined. Next, a constant for each image is defined. The
rectangle is created using the basic constructor rectangle. It requires four inputs: a
width, a height, a mode, and a color. The circle is created using the basic constructor
circle. It requires three inputs: a radius, a mode, and a color. This is followed by
tests. We can write the expected value of a test using a function application or an
image given that both are valid expressions in BSL. The tests that use images require
that the image be created first and then copied and pasted as part of the test.
There are too many basic image constructors to describe them all in this textbook.
Therefore, search for 2htdp/image in the Help Desk to learn about and experiment
with them. Be creative!

14.2 Property Selectors

There are two property selector functions:

image-width Returns the width of an image in pixels.
image-height Returns the height of an image in pixels.
14 Images 39

Fig. 18 Basic constructors in BSL

(require 2htdp/image)

(define RECT-WIDTH 35)

(define RECT-HEIGHT 55)
(define RECT-MODE outline)
(define RECT-COLOR red)

(define CIRC-RADIUS 25)

(define CIRC-MODE solid)
(define CIRC-COLOR green)

(define RED-OUTLINE-RECT (rectangle RECT-WIDTH RECT-HEIGHT

RECT-MODE RECT-COLOR))
(define GREEN-SOLID-CIRCLE (circle CIRC-RADIUS CIRC-MODE CIRC-COLOR))

(check-expect RED-OUTLINE-RECT (rectangle 35 55 outline red))

(check-expect RED-OUTLINE-RECT )
(check-expect GREEN-SOLID-CIRCLE (circle 25 solid green))

(check-expect GREEN-SOLID-CIRCLE )

Consider the problem of computing half the width, half the height, and the number
of pixels for the following images:
(define A-STAR-IMG (radial-star 10 5 35 'outline 'gold))
(define A-RHOMBUS-IMG (rhombus 60 75 'solid 'red))
The first step is to clearly identify what needs to be computed and how it is computed.
We need to compute three values as follows:
Half the Width Divide the image’s width by 2.
Half the Height Divide the image’s height by 2.
Number of Pixels Compute the image’s area.
The next question that arises is how are we to test our computations. We know that
every image is rectangular, but this does not tell us what is the width or the height
of a star image or of a rhombus image. When we do not know what a specific value
ought to be or it is too difficult to write a specific value, we can use property-based
testing to write tests. Instead of testing for the specific value of an expression, we
test properties that the value of the expression ought to have. For our problem, we
40 2 Expressions and Data Types

Fig. 19 Sample program using image selectors and property-based testing

(require 2htdp/image)

(define A-STAR-IMG (radial-star 10 5 35 outline gold))

;; Star computations
(define STAR-W (image-width A-STAR-IMG))
(define STAR-H (image-height A-STAR-IMG))
(define STAR-HALF-W (/ STAR-W 2))
(define STAR-HALF-H (/ STAR-H 2))
(define STAR-PIXELS (* STAR-W STAR-H))

;; Star tests
(check-expect (* 2 STAR-HALF-W) STAR-W)
(check-expect (* 2 STAR-HALF-H) STAR-H)
(check-expect (/ STAR-PIXELS STAR-W) STAR-H)
(check-expect (/ STAR-PIXELS STAR-H) STAR-W)

(define A-RHOMBUS-IMG (rhombus 60 75 solid red))

;; Rhombus computations
(define RHOMBUS-W (image-width A-RHOMBUS-IMG))
(define RHOMBUS-H (image-height A-RHOMBUS-IMG))
(define RHOMBUS-HALF-W (/ RHOMBUS-W 2))
(define RHOMBUS-HALF-H (/ RHOMBUS-H 2))
(define RHOMBUS-PIXELS (* RHOMBUS-W RHOMBUS-H))

;; Rhombus tests
(check-expect (* 2 RHOMBUS-HALF-W) RHOMBUS-W)
(check-expect (* 2 RHOMBUS-HALF-H) RHOMBUS-H)
(check-expect (/ RHOMBUS-PIXELS RHOMBUS-W) RHOMBUS-H)
(check-expect (/ RHOMBUS-PIXELS RHOMBUS-H) RHOMBUS-W)

know that an image is always rectangular. This means that twice half the width ought
to be the width, twice half the height ought to be the height, the number of pixels
divided by the image’s height ought to be the image’s width, and the number of
pixels divided by the image’s width ought to be the image’s height.
Now that we have a game plan, we can proceed to write our program as displayed
in Fig. 19. Observe that related definitions and tests are kept together. First, a data
item is defined like A-RHOMBUS-IMG. Second, the result of computations is defined
like RHOMBUS-HALF-W. Third, tests are written. Organizing code in such a manner
is a good practice because anyone who reads the code (including yourself 6 months
from now) will be able to understand the goals.
14 Images 41

Fig. 20 An image of nested squares

14.3 Image Composers

Image composers are used to create new images from existing images. The image
teachpack provides a myriad of such functions and you ought to experiment with
them as you read the documentation in the Help Desk.
Consider the problem of creating the image in Fig. 20. As always, the first step
is to determine what needs to be computed. The image consists of 6 nested squares.
The largest square is yellow and at the bottom. The smallest square is blue and on
the top. This suggests a divide and conquer strategy. First, compute all the needed
squares. Second, overlay the squares to create the desired image. We can test the
result using property-based testing. A program to implement this strategy is
;; Tests
;; NOTE: SQUARE0 is the largest square.
(check-expect (image-width NESTED-SQS) (image-width SQUARE0))

(check-expect (image-height NESTED-SQS) (image-height SQUARE0))

;; The squares
(define SQUARE0 (square 60 'solid 'yellow)) ;; largest square
(define SQUARE1 (square 50 'solid 'blue))
(define SQUARE2 (square 40 'solid 'yellow))
(define SQUARE3 (square 30 'solid 'blue))
(define SQUARE4 (square 20 'solid 'yellow))
(define SQUARE5 (square 10 'solid 'blue)) ;; smallest square

;; The Nested Squares

(define NESTED-SQS (overlay SQUARE5 SQUARE4 SQUARE3
SQUARE2 SQUARE1 SQUARE0))
The tests check that the dimensions of the resulting image are correct: the dimensions
ought to be the same as the largest square used. The program independently defines
42 2 Expressions and Data Types

the 6 needed squares from largest to smallest. Finally, the needed image is created
by using the function overlay always placing a smaller image over a larger image.
It is natural to wonder how the placing is done. Recall that an image consists of
many pixels. One of these pixels serves as an anchor point around which images
are placed. Unless otherwise specified, image composing functions use the pixel at
the center of the image as the anchor point. This means, for example, that overlay
places all the center points of the images one on top of another. In our program,
the anchor point of SQUARE1 is placed over the anchor point of SQUARE0. Next,
the anchor point of SQUARE2 is placed over the anchor point of the SQUARE1 and
SQUARE0 composed image, and so on until all the squares are overlayed. You may
contrast this behavior with the behavior of the application expression:
(overlay/xy img1 i j img2)
For overlay/xy, the images start aligned with respect to their top-left corner pixel
(not the center pixel). The function overlays img1 over img2 but moves img2 i pixels
to the right and j pixels down. If i is negative, img2 is moved to the left. If y is
negative, img2 is moved up. Experiment with image composers!
Another noteworthy characteristic of our program is that the testing is not very
satisfying. Consider what happens if we mistakenly define NESTED-SQS as follows:
(define NESTED-SQS (overlay SQUARE0 SQUARE1 SQUARE2
SQUARE3 SQUARE4 SQUARE5))
No tests fail when running the program despite that NESTED-SQS is the image of
a yellow square. Clearly, this is not the desired result and this testing shortcoming
ought to be addressed. To remedy such a situation, add unit tests using actual values
that are computed. In this case, first experiment with our program. Once the desired
image is computed, copy and paste it as the expected result for testing NESTED-SQS.
In this manner, any reader of your code can see that both your model and your result
are consistent with expectations. You must be very careful when employing such
a strategy to write tests. If a computed value is wrong, then its use in a test will
make the test pass. Your program, however, still computes the wrong value. It is best
to limit the use of this strategy to test computed images that are easy to visually
validate.

14.4 Empty Scenes and Placing Images

There are two image composing functions that are used extensively to create anima-
tions and video games:
empty-scene Creates an empty image of some given width and height. Option-
ally, the color of the empty image may be provided.
place-image Places the first given image at the given x and y coordinates in the
second given image.
As characters and elements change in a video game, the displayed image must change.
This is done by computing a new image. The empty image is used to create the base
14 Images 43

Fig. 21 Computer graphics coordinate system

0 1 2 3 ...
0
1
2
3
..
.

image of the video game or animation. Think of this base image as the empty canvas
of an artist. The function place-image is used to place elements.
To properly use place-image, a brief description of the computer graphics coor-
dinate system is needed. Figure 21 displays the coordinate system used in computer
graphics. The possible values of x grow from left to right. To place an image further
to the right, the x value is increased. To place an image further to the left the x value
is decreased. The possible values of y grow from top to bottom. To place an image
further down, the y value is increased. To place an image further up, the y value is
decreased. For a given image with dimensions WIDTH x HEIGHT, the valid x values
are in [0..WIDTH-1] and the valid y values are in [0..HEIGHT-1]. If any part of
a placed image falls outside these coordinate ranges, the image is cropped and parts
of the placed image do not appear in the result.
Consider the problem of creating an image with three different alien ships in a
black empty scene. To make the task more manageable, use a divide and conquer
strategy:
• Define the dimensions of the empty scene.
• Compute three different alien ships.
• Compute an image with one alien ship.
• Compute an image with two alien ships.
• Compute an image with three alien ships.
• Write tests.
There are an infinite number of ways a program can be written following this strategy.
Figure 22a displays a proposed partial solution. A black empty scene, 200 x 100,
and 3 different alien ships that differ in color are defined. After this, 3 images are
defined using place-image. Each successive image contains one more ship.
The resulting image with the 3 alien ships is displayed in Fig. 22b. Immediately,
you can see there is a bug in the program. The pink alien ship is cropped and only
part of it appears in the image. This example highlights the value of testing. The
program in Fig. 22a does not contain a test. Although the divide and conquer strategy
is a good one, it is always necessary to perform thorough testing to establish a degree
of confidence in a problem’s solution.
44 2 Expressions and Data Types

Fig. 22 Program for an image with 3 different alien ships in a black empty image
 Program.

(require 2htdp/image)

(define WIDTH 200)

(define HEIGHT 100)

(define E-SCENE (empty-scene WIDTH HEIGHT black))

(define ALIEN-SHIP0 (overlay (circle 7 solid gray)

(rectangle 23 3 solid gray)))

(define ALIEN-SHIP1 (overlay (circle 7 solid pink)

(rectangle 23 3 solid pink)))

(define ALIEN-SHIP2 (overlay (circle 7 solid white)

(rectangle 23 3 solid white)))

(define SCN0 (place-image ALIEN-SHIP0 15 30 E-SCENE))

(define SCN1 (place-image ALIEN-SHIP1 197 80 SCN0))

(define SCN2 (place-image ALIEN-SHIP2 (/ WIDTH 2) (/ HEIGHT 2) SCN1))

 Result Image.

** Ex. 27 — Debug and complete the design of the program in Fig. 22a. Make
sure to test all defined values.
* Ex. 28 — Write a program that creates an image of a banner with your name
on it.
* Ex. 29 — Write a program that creates an image of a triangle that is over an
image of the triangle flipped over the x-axis.

** Ex. 30 — Write a program that creates the image of a smiley face.

15 What Have We Learned in This Chapter? 45

Fig. 23 Sample starry night images

 Starry Night I.

 Starry Night II

*** Ex. 31 — Write a program that creates a starry night image. Your com-
position must have a representation of at least one moon, three stars, and a
stickman. Figure 23aa and 23bb display sample starry night images created by
students. Be creative and have fun!
a
Courtesy of Lindsey M Reams.
b
Courtesy of Shamil Dzhatdoyev.

15 What Have We Learned in This Chapter?

The important lessons of Chap. 2 are summarized as follows:

46 2 Expressions and Data Types

Fig. 24 Chapter 2 BSL grammar

program ::= {expr test def}*
def ::= (define <variable name> expr)
test ::= (check-expect expr expr)
::= (check-within expr expr expr)
expr ::= number string <variable name>
::= \#<character constant>
::= <character+>
::= #true
::= #false
::= <image>
::= (<function> expr+ )
::= (and expr expr expr*)
::= (or expr expr expr*)
string ::= "<character> *"
number ::= positive
::= negative
positive ::= mag
negative ::= -mag
mag ::= int
::= real
::= fraction
int ::= digit+
digit ::= 0|1|2|3|4|5|6|7|8|9
real ::= digit* .digit+
fraction ::= int/nonzero-int
nonzero-int ::= <1|2|3|4|5|6|7|8|9>int

• When solving a problem, outline the values that must be computed to guide the
development of a solution based on divide and conquer.
• In addition to number and string, BSL includes the character, symbol, Boolean,
and image types.
• Primitive type instances, like a Boolean, cannot be decomposed.
• Compound type instances, like a string, can be decomposed.
• The BSL syntax developed so far is displayed in Fig. 24.
– New expressions for characters, symbols, Booleans, and images.
– andand orexpressions are syntax for the Boolean functions and and or.
– not is a Boolean function in BSL.
– | means or. For example, the first production rule for expr states that it can be
a number, a string, or a variable.
• Computations involving numbers and images may contain errors.
• Use check-within to test the computation of a real number.
• Boolean functions are called predicates and allow programmers to determine if
the input satisfies a property.
• Model checking tests properties of a computed value.
• The 2htdp/image teachpack provides functions to create, compose, and extract
properties of images.
Chapter 3
The Nature of Functions

Every high school student has studied functions. What for? Have you ever wondered
where functions come from? Why do they exist? Certainly, many students can state
that a function is a mapping from a domain element to a range element. That,
however, does not tell anyone where functions come from. Perhaps, looking at an
example helps. Consider the following function:
f(x) = x2 + 3x - 10
Certainly, many students can state that x is an input element from the domain and
that the value of f(10) is 120. Now, we know something about functions. We can
use functions to compute values. This is done by "plugging in" 10 for x. That is, all
instances of x are substituted with 10 to obtain
f(10) = 102 + 3·10 - 10
Now, it is all a matter of evaluating function’s body (i.e., the expression to the right of
the equal sign). This is screaming to us that functions are likely to play an important
role in programming. After all, BSL, as well as other programming languages, is
exceptionally good at evaluating expressions. A program in BSL to evaluate f(10)
looks like this:
;; Evaluating f(x) = x^2 + 3x - 10
(define F-X10 (+ (sqr 10) (* 3 10) -10))
(check-expect F-X10 120)
We still, however, do not know where f(x) came from.
This chapter introduces the process of abstraction over expressions that leads to
functions. It also introduces the BSL syntax needed to define functions (like f(x)).
More importantly, it introduces the first design recipe to guide problem solvers in
function development. A design recipe is a series of steps, each with a specific
outcome, that defines a systematic process to implement functions. A design recipe
does not tell you the solution to a problem. That you must figure out. It does, however,
prove useful to take you from a problem statement and a blank IDE to a program that
solves the problem in a manner that is understandable to others.
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 47
M. T. Morazán, Animated Problem Solving, Texts in Computer Science,
https://doi.org/10.1007/978-3-030-85091-3_3
48 3 The Nature of Functions

Fig. 25 Program for f(x) = x2 + 3x - 10 at x = 10, 0, 1, 20, 100

;; Evaluating f(x) f(x) = x^2 + 3x - 10

(define F-X10 (+ (sqr 10) (* 3 10) -10))

(define F-X0 (+ (sqr 0) (* 3 0) -10))
(define F-X1 (+ (sqr 1) (* 3 1) -10))
(define F-X20 (+ (sqr 20) (* 3 20) -10))
(define F-X100 (+ (sqr 100) (* 3 100) -10))

(check-expect F-X10 120)

(check-expect F-X0 -10)
(check-expect F-X1 -6)
(check-expect F-X20 450)
(check-expect F-X100 10290)

16 The Rise of Functions

Writing a program to compute the value of f(10) has proven rather easy. Now you
are asked to also have your program compute f(0), f(1), f(20), and f(100).
After some thought and a bit of typing, your program looks as displayed in Fig. 25.
A different constant is defined for each value of f(x) obtained from a different value
of x. The tests validate that each value of f(x) has been correctly computed.
Now, you are asked to have your program compute all the values of f(x) for the
integers in [11..19]. This is another 9 different values of x and, frankly speaking,
many are unlikely to want to do a lot of the same typing over and over again. To
avoid this, some may decide to copy, paste, and edit code to account for the changes.
This is dangerous, because it is error-prone. It also means that for a new batch of x
values you will be doing a lot of copying and pasting again. There must be a better
way to get the job done.
Why would anyone decide to cut and paste to write more code? It turns out there
is something interesting about the expressions used to compute the different values
of f(x). Observe that they are almost identical (see Fig. 26). There is only one
difference from one expression to the next: the number plugged in. The elements
that vary among similar expressions are called variables. That is, we can abstract
away the differences among similar expressions by associating each difference with
a variable. For the expressions above, let us call the single difference x (as done in
the mathematical function f(x)). This reduces all the expressions above to a single
expression:
(+ (sqr x) (* 3 x) -10))
In order to exploit such an expression, a mechanism is needed to provide x with a
value.
Let us go back and see how mathematicians provided this mechanism. They chose
the syntax f(x) = e, where e is a mathematical expression. We say that f(x) is
the function header. A function header contains two parts: the name of the function
and the name of the input variables. The input variables are called the function’s
parameters. For f(x), f is the function name and x is the only parameter. We say that
16 The Rise of Functions 49

Fig. 26 Expressions for f(x)

(+ (sqr 10) (* 3 10) -10)
(+ (sqr 0) (* 3 0) -10)
(+ (sqr 1) (* 3 1) -10)
(+ (sqr 20) (* 3 20) -10)
(+ (sqr 100) (* 3 100) -10)

e is the function body. Mathematicians use a mathematical application expression to

bind parameters to values. For example, f(10) binds x to 10. To obtain an expression
that can be evaluated, every parameter is substituted with its binding in the body
of the function. The process of substituting every variable with its binding is called
β-reduction. For f(10), β-reduction substitutes every x with 10 in x2 + 3x - 10
to obtain 102 + 3·10 - 10. Observe that the β-reduction transformation yields an
expression that has no variables and may be evaluated.
BSL allows programmers to write their own functions much like mathematicians
do. To allow programmers to do so, there is a new def production rule:
def ::= (define (<name> <name>+ ) expr)
This production rule states that a function definition is written starting with an
opening parenthesis, the keyword12 define, and inside parenthesis the function
name and the name of 1 or more parameters.13 This is called the BSL function header.
The function header is followed by a BSL expression and a closing parenthesis (to
match the opening parenthesis before define). The expression after the function
header is the body of the BSL function.
Developing a function is not quite enough. It is not reasonable to assume that
anyone will know how to correctly use a function just by looking at it. Any reader
or user of a function needs to know the types of the parameters, the type of the
value returned, and the purpose of the function. In other words, functions need to
be documented with a signature and a purpose statement (just like it is done in
DrRacket’s Help Desk). For our current problem, x’s type is number and the type
returned by the function is number. Therefore, the signature is number → number.
The purpose of the BSL function is to compute the value of f(x).
The program in Fig. 25 can now be updated to compute all the values of f(x)
for the integers in [11..19]. Instead of defining 9 constants for the new function
values, a single BSL function for f(x) is defined and it is used 9 times to compute
the new values. The result of this process is displayed in Fig. 27. Observe that the
new tests use f instead of using a constant for the tested value. Now, the value being
tested is the value returned by f. Further observe that f’s signature and purpose are
written before f’s definition. This helps anyone reading the program to understand
the defined function.
Carefully think about what has been achieved. Ask yourself these questions:
• For the tests, do you prefer to use f or always use defined constants?
• Can this program compute any value of f(x)?
12
A keyword is a word that has a special meaning in a programming language.
13
A name is a symbol written without the '.
50 3 The Nature of Functions

Fig. 27 Program for computing values of f(x) using a BSL function

;; Evaluating f(x) f(x) = x^2 + 3x - 10

(define F-X10 (+ (sqr 10) (* 3 10) -10))

(define F-X0 (+ (sqr 0) (* 3 0) -10))
(define F-X1 (+ (sqr 1) (* 3 1) -10))
(define F-X20 (+ (sqr 20) (* 3 20) -10))
(define F-X100 (+ (sqr 100) (* 3 100) -10))

(check-expect F-X10 120)

(check-expect F-X0 -10)
(check-expect F-X1 -6)
(check-expect F-X20 450)
(check-expect F-X100 10290)
(check-expect (f 11) 144)
(check-expect (f 12) 170)
(check-expect (f 13) 198)
(check-expect (f 14) 228)
(check-expect (f 15) 260)
(check-expect (f 16) 294)
(check-expect (f 17) 330)
(check-expect (f 18) 368)
(check-expect (f 19) 408)

;; number number
;; Purpose: To compute f(x)
(define (f x)
(+ (sqr x) (* 3 x) -10))

• Instead of asking you to modify the program for every value of f(x) needed, can
others use the program to compute these values?
To answer the first question, consider a single test. Most individuals find more
readable and prefer to type
(check-expect (f 11) 144)
over
(check-expect F-X11 144)
(define F-X11 (+ (sqr 11) (* 3 1) -10))
The answer to the second question is yes given enough memory. As a program is
evaluated, memory is allocated to store values. Since memory is finite, it is possible
that a program may be asked to compute a value that requires more memory than
is available. This is something to keep in mind. In practice, however, people around
the world execute programs every day without running out of memory. The answer
to the third question is an unequivocal yes. Anyone running the program only needs
to type the correct application expression at the prompt in the interactions area.
Why were the tests using constants not rewritten to eliminate the need to define
the constants? The answer is that many times it is useful to see how a value is
17 General Design Recipe for Functions 51

Fig. 28 The general design recipe for functions

1. Outline the representation of values and the computation.
2. Define constants for the value of sample expressions.
3. Identify and name the differences among the sample expressions.
4. Write the function’s signature and purpose.
5. Write the function’s header.
6. Write tests.
7. Write the function’s body.
8. Run the tests and, if necessary, redesign.

computed to understand a program. Therefore, you should write both types of tests:
tests that use concrete values (like the last 9 tests) and tests that use defined constants
for expressions that show how a value is computed (like the first 4 tests).

17 General Design Recipe for Functions

Abstraction over expressions led to functions. The question now is how do we

systematically develop functions to solve problems. There are general steps that
ought to be followed to develop every function. Figure 28 presents the steps in the
form of a design recipe. Each step has a specific outcome that guides you to a
problem’s solution.
The first two steps are called problem analysis. Step 1 asks for how values in the
real or an imaginary (like in a video game) world are represented in the program. This
means that we must represent values in the real or imaginary world in a manner that
our programming language can manipulate. Furthermore, this representation must
be invertible. That is, a result must be translatable to the world element it represents.
It is important for the representation to capture the characteristics needed to carry
out the computation. For example, consider the problem of computing the area of
a rectangular field. A field in the real world cannot be processed by a computer.
Therefore, we need a representation of its characteristics that are needed to compute
the area. We can represent the width, length, and area of the field as nonnegative real
numbers. Step 1 also asks for the computation of values to be outlined. Returning to
the area of a rectangular field, we can state that the area is the product of the length
and the width. Step 2 requires the development of sample expressions that illustrate
how the needed value is computed.
Step 3 is the abstraction step. The differences among the sample expressions
must be identified and each is given a unique variable name. These variables are the
parameters of the function. Pick names that identify the value they represent. For
example, if a variable represents a number, then a-number is a better choice than z.
Step 4 requires that the function’s signature be written. To do so, the type of each
parameter and the function’s return type must be written with an arrow between
them. In addition, Step 4 requires a short description of the function’s purpose. This
purpose statement needs to be clear and concise so that others may understand what
52 3 The Nature of Functions

Fig. 29 The basic function template

#|
;; type type
;; Purpose:
(define (f-on-args var)
...)

;; Sample expression definitions for f-on-args

(define CONSTANT-1 expr)
...
(define CONSTANT-N expr)

;; Tests using sample computations for f-on-args

(check-expect (f-on-args expr) CONSTANT-1)
...
(check-expect (f-on-args expr) CONSTANT-N)

;; Tests using sample values for f-on-args

(check-expect (f-on-args expr) value-1)
...
(check-expect (f-on-args expr) value-m)
|#

is computed by the function. The function header is written to satisfy Step 5. Pick a
function name that helps identify the purpose of the function. For example, naming
a cubing function f is a poor choice. A better choice is naming the function cube. In
addition, the parameters from Step 3 must be written in the order that corresponds
to the types in the signature. For Step 6, tests are written using an application of
the function as the tested value and using both the constants from Step 2 and other
values not defined as constants. The body of the function is developed to satisfy
Step 7. β-reduction is performed on any one of the sample expressions. That is, the
differences are substituted with the correct parameter. Finally, Step 8 requires the
running of the tests. If all the tests pass, you may have guarded optimism in having
designed a correct function. If there are any errors, then you need to check your
answers to each step of the design recipe.
The design recipe suggests a function template may be used for function design.
A function template is like a skeleton for functions. It captures the main components
a problem solution needs. Figure 29 displays the basic function template. Every time
a problem is being solved, the template is copied and specialized. When an answer
for a step of the design recipe is formulated, the copied template is edited to reflect
the answer.
17 General Design Recipe for Functions 53

17.1 The Design Recipe in Action

To illustrate the use of the design recipe, consider the problem of creating name
banner images for Craig, Julia (Ohio), Steve, Little Nick, Big Nick, and Jeremy. The
name image should contain only the name in a green 36 size font.
The answers to each step of the design recipe are outlined as follows:
STEP 1
Represent each name as a string. The banner image is computed by applying the
text function to a name, 36, and 'olive.
STEP 2
Three constants for the value of sample expressions are

(define CRAIG (text "Craig" 36 'olive))

(define JULIA (text "Julia (Ohio)" 36 'olive))
(define STEVE (text "Steve" 36 'olive))

STEP 3
The only difference among the three sample expressions is the string representing
the name. The variable for this difference is name.
STEP 4
There is only one difference which is a string and the value computed is an image.
The signature and purpose statement are

;; string → image
;; Purpose: Create a banner for the given name
;; using font 36 and the color olive

STEP 5
The function header contains only one parameter, because a single difference was
identified in Step 3. The name make-banner is suggestive of the purpose of the
function. The function header is
(define (make-banner name)
STEP 6
The tests using the defined constants and sample values are below.

;; Tests using sample computations

(check-expect (make-banner "Craig") CRAIG)
(check-expect (make-banner "Julia (Ohio)") JULIA)
(check-expect (make-banner "Steve") STEVE)
54 3 The Nature of Functions

Fig. 30 Name banners program

(require 2htdp/image)

;; string image
;; Purpose: Create a banner for the give name using
;; font 36 and the color olive
(define (make-banner name)
(text name 36 olive))

; Sample expression definitions for make-banner

(define CRAIG (text "Craig" 36 olive))
(define JULIA (text "Julia (Ohio)" 36 olive))
(define STEVE (text "Steve" 36 olive))

;; Tests using sample computations for make-banner

(check-expect (make-banner "Craig") CRAIG)
(check-expect (make-banner "Julia (Ohio)") JULIA)
(check-expect (make-banner "Steve") STEVE)

;; Tests using sample values for make-banner

(check-expect (make-banner "Little Nick") )

(check-expect (make-banner "Big Nick") )

(check-expect (make-banner "Jeremy") )

;; Tests using sample values

(check-expect (make-banner "Little Nick")

)
(check-expect (make-banner "Big Nick")

)
(check-expect (make-banner "Jeremy")

STEP 7
The body of the function uses name instead of a concrete string:
(text name 36 'olive)
STEP 8
All tests pass.
Figure 30 displays the full program written in BSL inside DrRacket.
18 Auxiliary Functions 55

* Ex. 32 — Follow the steps of the design recipe to write a program to compute
the area of an ellipse.

** Ex. 33 — Follow the steps of the design recipe to write a program to com-
pute the image of a smiley face.

** Ex. 34 — Logical equivalence, denoted A ⇔ B, means that A ⇒ B and

that B ⇒ A. That is, it means that A is true if and only if B is true. The truth
table for logical equivalence is
A B A⇔B
F F T
F T F
T F F
T T F
Follow the steps of the design recipe to write a program to compute the logical
equivalence of two Boolean values.

* Ex. 35 — Follow the steps of the design recipe to write a program to deter-
mine if a string is of even length.

18 Auxiliary Functions

Seldomly, programs only need to compute many instances of a single value as the
program in Fig. 30. It is far more common to have to compute several different values
to solve a problem. A good designer develops a function for each different value.
Such a practice makes it easier for the programmer and for others to understand
the solution to the problem. It also endows programs with modularity. Modularity
enables re-usability of expressions and reduces code duplication. In other words,
modularity is tightly connected to abstraction.
If problem analysis reveals that different kinds of values need to be computed,
consider which values ought to be computed by an auxiliary function. If special
knowledge is required to compute a value or if different instances of the same value
are needed, then designing an auxiliary function is a good choice. On the other hand,
if the expression needed to compute a new value is short, simple, and used only once,
then designing a new function may add little clarity to the solution.
If a programmer determines that several functions are needed, she can follow
two basic strategies: bottom-up or top-down. Bottom-up is a programming style in
which the simplest (usually the smallest) functions are designed first, and then these
functions are used to design more complex functions. Top-down is a programming
style in which the complex functions are designed first, and then simpler functions are
designed. The most commonly used approach is top-down, because a programmer
rarely knows in advance all the simpler functions that are required. Following a top-
down approach allows the programmer to discover, as the design advances, needed
simpler functions.
56 3 The Nature of Functions

Fig. 31 Computing the area of a washer requires two radii

18.1 Bottom-Up Design

To illustrate the bottom-up approach, consider computing the area of a washer as

displayed in Fig. 31. The goal is to compute the area of the green section. This area is
the area difference of the outer circle with radius ro and the inner circle with radius
ri . There are two different values to compute: area and difference. This suggests two
different functions. Given that computing the difference depends on computing the
area, a bottom-up approach starts by designing the area function for circles.
The steps of the design recipe for the area of a circle function may be satisfied as
follows:
STEP 1
The area of a circle is given by the product of π and the square of the circle’s
radius.
STEP 2
Two constants for sample expression values are

(define C-AREA5 (* pi 5))

(define C-AREA18 (* pi 18))

STEP 3
The only difference among the sample expressions is the radius. The variable for
this difference is a-radius.
18 Auxiliary Functions 57

STEP 4
The signature and purpose statement are

;; R>= 0 → R>= 0
;; Purpose: Compute the area of the circle with the given
;; radius.

R>= 0 is any real number greater than or equal to zero.

STEP 5
The function header is
(define (area-circle a-radius)
Observe that the chosen function name suggests the purpose of the function.
STEP 6
The tests using defined constants and sample values are

;; Tests using sample computations

(check-within (area-circle 5) C-AREA5 0.01)
(check-within (area-circle 18) C-AREA18 0.01)

;; Tests using sample values

(check-within (area-circle 0) 0 0.01)
(check-within (area-circle 10) 314.15 0.01)

Observe that the tests include, 0, the lowest possible value for a-radius. It is
always good practice to test border values in the domain of a function.
STEP 7
The body of the function is

(* pi (sqr a-radius))

STEP 8
All tests pass.
With area-circle designed, implemented, and tested, the function for the area
of a washer is designed. The steps of the design recipe may be satisfied as follows:
STEP 1
The area of a washer is given by the area difference of its outer and inner circles.
STEP 2
Three constants for the value of sample expressions are

;; Sample expression definitions for area-washer

(define W-AREA6-3 (- (area-circle 6) (area-circle 3)))
58 3 The Nature of Functions

(define W-AREA15-9 (- (area-circle 15) (area-circle 9)))

(define W-AREA4-1 (- (area-circle 4) (area-circle 1)))

Observe that function composition is used to write the tests. The output of
area-circle is used as input to -. In addition, observe that the constant names
are suggestive of the value they represent.
STEP 3
There are two differences in the sample expressions: the radii of the outer and the
inner circles named, respectively, outer-radius and inner-radius.
STEP 4
The signature and purpose statement are

;; R>= 0 R>= 0 → R>= 0

;; Purpose: Compute the area of the washer with the given
;; radii.
;; ASSUMPTION: outer-radius >= inner-radius

An assumption statement is added to make clear that outer-radius must be greater

than or equal to inner-radius. Such statements help others to correctly use the
function.
STEP 5
The function header is
(define (area-washer outer-radius inner-radius)

The name of the function is suggestive of its purpose. Compare this name choice
with a name like g or area.
STEP 6
The tests using defined constants and sample values are

;; Tests using sample computations for area-washer

(check-within (area-washer 6 3) W-AREA6-3 0.01)
(check-within (area-washer 15 9) W-AREA15-9 0.01)
(check-within (area-washer 4 1) W-AREA4-1 0.01)

;; Tests using sample values for area-washer

(check-within (area-washer 0 0) 0 0.01)
(check-within (area-washer 10 0) 314.15 0.01)
(check-within (area-washer 7 4) 103.67 0.01)
18 Auxiliary Functions 59

Fig. 32 Area of a washer program

;; R > = 0 Real>= 0
;; Purpose: Compute the area of the circle with the given radius.
(define (area-circle a-radius)
(* pi (sqr a-radius)))

;; Sample expression definitions for area-circle

(define C-AREA5 (* pi (sqr 5)))
(define C-AREA18 (* pi (sqr 18)))

;; Tests using sample computations for area-circle

(check-within (area-circle 5) C-AREA5 0.01)
(check-within (area-circle 18) C-AREA18 0.01)

;; Tests using sample values for area-circle

(check-within (area-circle 0) 0 0.01)
(check-within (area-circle 10) 314.15 0.01)

;; R > = 0 R > = 0 Real>= 0

;; Purpose: Compute the area of the washer with the given radii.
;; ASSUMPTION: outer-radius >= inner-radius
(define (area-washer outer-radius inner-radius)
(- (area-circle outer-radius) (area-circle inner-radius)))

;; Sample expression definitions for area-washer

(define W-AREA6-3 (- (area-circle 6) (area-circle 3)))
(define W-AREA15-9 (- (area-circle 15) (area-circle 9)))
(define W-AREA4-1 (- (area-circle 4) (area-circle 1)))

;; Tests using sample computations for area-washer

(check-within (area-washer 6 3) W-AREA6-3 0.01)
(check-within (area-washer 15 9) W-AREA15-9 0.01)
(check-within (area-washer 4 1) W-AREA4-1 0.01)

;; Tests using sample values for area-washer

(check-within (area-washer 0 0) 0 0.01)
(check-within (area-washer 10 0) 314.15 0.01)
(check-within (area-washer 7 4) 103.67 0.01)

STEP 7
The body of the function is
(- (area-circle outer-radius) (area-circle inner-radius))

STEP 8
All tests pass.
The complete program is displayed in Fig. 32. An advantage of bottom-up design
is that functions may be tested as soon as they are implemented. For example, the
area-circle function may be tested as soon as it is written.
60 3 The Nature of Functions

It is possible to design a single function to compute the area of a washer. Such a

function contains repetitions and may look as follows:
;; R>= 0 R>= 0 → R>= 0
;; Purpose: Compute the area of the washer with the given
;; radii.
;; ASSUMPTION: outer-radius >= inner-radius
(define (area-washer outer-radius inner-radius)
(- (* pi (sqr outer-radius)) (* pi (sqr inner-radius))))

;; Sample expression definitions for area-washer

(define W-AREA6-3 (- (* pi (sqr 6)) (* pi (sqr 3))))
(define W-AREA15-9 (- (* pi (sqr 15)) (* pi (sqr 9))))
(define W-AREA4-1 (- (* pi (sqr 4)) (* pi (sqr 1))))

;; Tests using sample computations for area-washer

(check-within (area-washer 6 3) W-AREA6-3 0.01)
(check-within (area-washer 15 9) W-AREA15-9 0.01)
(check-within (area-washer 4 1) W-AREA4-1 0.01)

;; Tests using sample values for area-washer

(check-within (area-washer 0 0) 0 0.01)
(check-within (area-washer 10 0) 314.15 0.01)
(check-within (area-washer 7 4) 103.67 0.01)
This function is obtained by inlining area-circle. Both versions solve the problem.
The question now is which one is clearer? Most people would argue that the version
in Fig. 32 is clearer. In the first version, it is easier to see how the area of a washer
is computed. Even a person who has forgotten the formula for the area of a circle
can discern how the problem is solved. Would that same person as easily discern
how the problem is solved in the second version above? Nonetheless, inlining is an
important optimization done by compilers14 and that you will learn more about in a
Programming Language or a Compilers course.

* Ex. 36 — Use a bottom-up approach following the steps of the design recipe
to write a program to determine if f(a) equals f(b), where f(x) = x2 - 9.

*** Ex. 37 — Use a bottom-up approach, following the steps of the design
recipe, to write a program to implement not, or, and and using only nor (not
or). The truth table for nor is displayed in Fig. 33. For example, (or A B) =
(nor (nor A B) (nor A B)).
* Ex. 38 — Use a bottom-up approach following the steps of the design recipe
to write a program to determine if two strings of length 3 contain a.

14
A compiler is a program that converts programs written in a language like BSL into machine
code that is executable by a computer.
19 Top-Down Design 61

Fig. 33 The truth table for nor

A B (nor A B)
F F T
F T F
T F F
T T F

* Ex. 39 — Use a bottom-up approach following the steps of the design recipe
to design and write a program to compute the flag images of Ghana, Honduras,
Lithuania, Niger, and Syria.

19 Top-Down Design

To illustrate the top-down design process, consider the problem of computing images
like those in Fig. 34. Each image has 4 overlayed squares at a 45◦ angle. Instead of
first trying to figure out how to compute each square in the image composition, start
by writing expressions to compute the image composition. For unknown values,
like the length of a square, use a constant. The definitions of these constants are a
different problem from the problem of creating the desired image. In other words,
use a divide and conquer strategy.
Under this design approach, the steps of the design recipe may be satisfied as
follows:
STEP 1
The image is computed by overlaying squares of alternating colors and differing
lengths. Every other overlayed square, starting with the largest square, is rotated
by 45◦ .
STEP 2
Observe that it is already known that several squares of different colors and lengths
are needed. This immediately suggests that we abstract away the image compu-
tation for squares and design (later) a function for this purpose. Furthermore, we
know that this function requires two arguments (one for each difference): a length
and a color.
Constants for the value of sample expressions may be defined as follows:

(define IMG1 (overlay

(rotate 0 (make-sqr IMG1-LEN4 'brown))
(rotate 45 (make-sqr IMG1-LEN3 'gold))
(rotate 0 (make-sqr IMG1-LEN2 'brown))
(rotate 45 (make-sqr IMG1-LEN1 'gold))))
62 3 The Nature of Functions

Fig. 34 Nested squares at an angle

(define IMG2 (overlay

(rotate 0 (make-sqr IMG2-LEN4 'black))
(rotate 45 (make-sqr IMG2-LEN3 'red))
(rotate 0 (make-sqr IMG2-LEN2 'black))
(rotate 45 (make-sqr IMG2-LEN1 'red))))

Observe that we do not concern ourselves at this point with how make-sqr is
implemented or with how to define the different lengths. We assume that they can
all be defined.
STEP 3
There are 6 differences in the sample expressions: 4 lengths and two colors. We
name the differences: bot-len, len2, len3, top-len, botclr, and topclr.
STEP 4
The signature and purpose statement are

;; R>= 0 R>= 0 R>= 0 R>= 0 color color → image

;; Purpose: Create an image of nested squares at 45
;; degree angles using the given lengths
;; and colors.

STEP 5
The function header is
(define (make-nested-sqs bot-len len2 len3 top-len
botclr topclr)
19 Top-Down Design 63

STEP 6
The tests using defined constants and sample values are

;; Tests using sample computations

(check-expect (make-nested-sqs IMG1-LEN1 IMG1-LEN2
IMG1-LEN3 IMG1-LEN4
'gold 'brown)
IMG1)
(check-expect (make-nested-sqs IMG2-LEN1 IMG2-LEN2
IMG2-LEN3 IMG2-LEN4
'red 'black)
IMG2)

;; Tests using sample values

(check-expect (make-nested-sqs 50 35.35 24.99 17.67 'olive 'yellow)

)
(check-expect (make-nested-sqs 20 14.14 9.99 7.06 'pink 'purple)

STEP 7
The body of the function is

(overlay (rotate 0 (make-sqr top-len topclr))

(rotate 45 (make-sqr len3 botclr))
(rotate 0 (make-sqr len2 topclr))
(rotate 45 (make-sqr bot-len botclr)))

STEP 8
All tests pass.
At this point, we cannot run any tests because make-sqr and the length constants
are undefined. The answer written for Step 8 assumes that the tests pass after all
required functions and constants are defined. If all tests do not pass, then checking
the steps of the design is necessary.
Let us turn our focus now to the design and implementation of make-sqr. We
arrive to this function with some clear ideas about the abstraction. There are two
differences: side length and color. If we were to discover that more values are needed,
64 3 The Nature of Functions

then we have a signal that there is a bug in our abstraction. We proceed assuming
that the function can be designed with the stated differences.
The steps of the design recipe may be satisfied as follows:
STEP 1
The image of a solid square is computed using the square function and the given
side length and color.
STEP 2
Constants for the value of sample expressions may be defined as follows:

(define SQ1 (square 500 'solid 'lightbrown))

(define SQ2 (square 70 'solid 'darkblue))

STEP 3
There are 2 differences in the sample expressions as expected: a length and a
color. We name the differences: side-len and a-color.
STEP 4
The signature and purpose statement are

;; R>= 0 color → image

;; Purpose: Compute the image of a square of the given length and color
degree angles using the given lengths and colors.

STEP 5
The function header is
(define (make-nested-sqs botlen len2 len3 toplen
botclr topclr)

STEP 6
The tests using defined constants and sample values are

;; Tests using sample computations

(check-expect (make-sqr 500 'lightbrown) SQ1)
(check-expect (make-sqr 70 'darkblue) SQ2)

;; Tests using sample values

(check-expect (make-sqr 12 'orange) )

(check-expect (make-sqr 35 'skyblue) )

19 Top-Down Design 65

Fig. 35 The length of the next smaller square is the length of a hypotenuse

STEP 7
The body of the function is

(square side-len 'solid a-color)

STEP 8
All tests pass.
The next task is to define the constants for the lengths. For any image of nested
squares, the length of the largest square is arbitrary. For example, the length of the
largest square can be 100. The length of the next smaller square is not arbitrary as
it depends on the length of the outer square. How do we compute the length of the
next smaller square? The problem must be carefully analyzed. Figure 35 displays the
placement of a smaller square inside a larger square. The length of the outer square
is w. Observe that two adjacent midpoints and the center of the outer square form
a right triangle. The lengths of the sides forming the 90◦ are both w2 . The length of
the inner square is the length of the hypotenuse of the right triangle. We know that
the length of the hypotenuse is given by the Pythagorean theorem:
x2 = ( w2 )2 + ( w2 )2

x = ( w2 )2 + ( w2 )2

= 2 * ( w2 )2

2
= 2 * w4

w2
= 2
66 3 The Nature of Functions

A function can be defined to compute the length of inner squares. The only difference
in computing one inner square length from another is the length of the outer square.
Instead of abstracting over similar expressions, we have developed a mathematical
expression for the length of an inner circle that is easily translated to BSL. Both
approaches to converging on an expression for the body of a function are valid and
ought to be part of any problem solver’s toolbox.
The steps of the design recipe may be satisfied as follows:
STEP 1
The length of the next smallest square is equal to the length of the line between
the two midpoints of two adjacent sides of the
enclosing square of length w. The
2
length of the next smaller square is given by w2 .
STEP 2
Constants for the value of sample expressions may be defined as follows:

(define IMG1-LEN1 100)

(define IMG1-LEN2 (sqrt (/ (sqr IMG1-LEN1) 2)))
(define IMG1-LEN3 (sqrt (/ (sqr IMG1-LEN2) 2)))
(define IMG1-LEN4 (sqrt (/ (sqr IMG1-LEN3) 2)))

STEP 3
There is a single difference in the sample expressions as expected: a length. We
name the difference side-len.
STEP 4
The signature and purpose statement are
;; R>= 0 → R>= 0
;; Purpose: Compute the length of the next smallest square

STEP 5
The function header is
(define (compute-new-sqr-len side-len)

STEP 6
The tests using defined constants and sample values are

;; Tests using sample computations

(check-within (compute-new-sqr-len 100)
IMG1-LEN2
0.01)
(check-within (compute-new-sqr-len IMG1-LEN2)
IMG1-LEN3
0.01)
19 Top-Down Design 67

(check-within (compute-new-sqr-len IMG1-LEN3)

IMG1-LEN4
0.01)

Given that constants for the square lengths for a second image are needed, the
function being designed may be used to compute these. The tests using sample
values based on these constants may be written as follows:
;; Tests using sample values
(define IMG2-LEN1 150)
(define IMG2-LEN2 (compute-new-sqr-len IMG2-LEN1))
(define IMG2-LEN3 (compute-new-sqr-len IMG2-LEN2))
(define IMG2-LEN4 (compute-new-sqr-len IMG2-LEN3))

(check-within IMG2-LEN2 106.06 0.01)

(check-within IMG2-LEN3 75 0.01)
(check-within IMG2-LEN4 53.03 0.01)
(check-within (compute-new-sqr-len 0) 0 0.01)
(check-within (compute-new-sqr-len 8) 5.65 0.01)

STEP 7
The body of the function is

(sqrt (/ (sqr side-len) 2))

STEP 8
All tests pass.
An important point to highlight is when a defined function may be applied in BSL.
A defined function may be applied any time when it is part of a test. All other times,
however, a function must be defined before it is applied to any arguments. That is,
the definition of a function must appear in a program before its first use outside tests.

* Ex. 40 — Implement the nested squares program.

* Ex. 41 — Use a top-down approach following the steps of the design recipe
to write a program to compute the flag images of Armenia, Austria, Bulgaria,
Colombia, and Estonia.

* Ex. 42 — Use a top-down approach following the steps of the design recipe to
write a program to compute f(x) = 4(x3 - 1)3 - 6(x3 -1)2 + 2(x3 -1)
+ 8.
68 3 The Nature of Functions

* Ex. 43 — Use a top-down approach following the steps of the design recipe
to write a program to append two double copies of a string separated by a space.
For example, "abc" results in "abcabc abcabc".

*** Ex. 44 — Use a top-down approach following the steps of the design recipe
to write a program to compute ((a ⇒ b) ∧ (b ⇒ c)) ⇒ (a ⇒ c). i
⇒ j is ¬i ∨ j.

20 What Have We Learned in This Chapter?

The important lessons of Chap. 3 are summarized as follows:

• Functions are abstractions over similar expressions.
– Each difference becomes a function parameter.
– In the body each difference is replaced with the corresponding parameter.
– Functions add modularity and help avoid repetition bugs.
• The BSL syntax explored, displayed in Fig. 36, includes a new def production
rule to define functions.
– Recall that | means or.
– expr ::= #true | #false means an expression may be #true or may be
#false.
• The basic function design recipe is a roadmap for developing and implementing
functions.
• Auxiliary functions are used when specialized knowledge is needed to compute
a needed value or when different instances of the same value are computed.
• There are two general programming development styles.
– Bottom-up: simplest functions are developed first.
– Top-down: complex functions are developed first.
– Regardless of development style, the design recipe guides development.
20 What Have We Learned in This Chapter? 69

Fig. 36 Chapter 3 BSL grammar

program ::= {expr test def}*
def ::= (define <variable name> expr)
::= (define (<name> <name+>) expr)
test ::= (check-expect expr expr)
::= (check-within expr expr expr)
expr ::= number string <variable name>
::= \#<character constant>
::= <character+>
::= #true #false
::= <image>
::= (<function> expr)
::= (and expr expr expr*)
::= (or expr expr expr* )
string ::= "<character> *"
number ::= positive
::= negative
positive ::= mag
negative ::= -mag
mag ::= int
::= real
::= fraction
int ::= digit
digit ::= 0|1|2|3|4|5|6|7|8|9
real ::= digit* .digit+
fraction ::= int/nonzero-int
nonzero-int ::= <1|2|3|4|5|6|7|8|9>int
Chapter 4
Aliens Attack Version 0

It is time to put your newly acquired skills to work. This chapter develops the first
version of a video game that we shall call Aliens Attack. This chapter, however, will
not develop a full video game. You need to learn more about problem solving and
more BSL syntax to write a fully working video game. Therefore, the video game
is developed using the process of iterative refinement. That is, as you learn more
about problem solving and about BSL syntax, a more complete video game shall be
developed. This process culminates with a multiplayer game that you may play with
your friends over the internet. That is a promise. If you work hard and absorb all the
lessons contained in this textbook, then you will acquire enough skills to develop a
multiplayer video game.
Figure 37 displays an image of a single-player version of Aliens Attack. In the
scene there is an army of aliens attacking earth and there is a single rocket defending
earth. There is no placating the aliens and earth must defend itself. The rocket is at
the bottom of the scene over earth and may move left to right without going off the
edge of the image. The rocket may also shoot in an attempt to destroy the aliens.
The army of aliens starts toward the top of the scene. All the aliens move in the
same direction: right, left, or down. The aliens move right one step at a time until
an alien reaches the right edge of the scene. At this point all the aliens move down
a step. After moving down a step, the aliens move left one step at a time until an
alien reaches the left edge of the scene. The aliens then move down a step. This cycle
continues until an alien reaches and conquers earth (the player loses the game) or all
the aliens are destroyed (the player wins the game). Observe that, like the rocket, the
aliens may not go off the scene as they move. When the rocket shoots, a shot starts
at the position of the rocket and moves up until it hits an alien or it goes off the top
each of the scene.
The first version of Aliens Attack is not truly a video game. That is fine because
it will be refined. To start, the focus is on problems that you can solve with the skills
you have already developed. The development of Aliens Attack version 0 focuses on
the creation and placement of images to render the video game as an image. This is
a good starting point because you know the basics of creating and placing images
and of writing functions. The first goal is to design the scene where the images of
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 71
M. T. Morazán, Animated Problem Solving, Texts in Computer Science,
https://doi.org/10.1007/978-3-030-85091-3_4
72 4 Aliens Attack Version 0

Fig. 37 A sample rendering of a single-player Aliens Attack

the game are rendered. The second goal is to develop functions to create the images
of the rocket, the alien, and the shot. The third goal is to create functions to draw the
elements in a scene.

21 The Scene for Aliens Attack

Recall that graphics are created using a coordinate system in which the x coordinate
grows from left to right and the y coordinate grows from top to bottom. Also recall
that a w1 x h1 empty scene into which images are placed may be created as follows:
(empty-scene w1 h1 )
Creating an empty scene is only part of what is needed. The question that immediately
arises is how do we reason about this empty scene. Figure 38a displays the view of
the empty scene as a grid of pixels. Each square in the grid is a pixel and items, like
the blue square, are placed in the scene using pixel coordinates. The blue square in
Fig. 38a is at pixel coordinate (15, 19). This can work well but seems to mismatch
our goal. Our goal is to place images in a scene, not manipulate pixels. If we choose
a maximum size for each image in the video game, we can reason about the empty
scene as boxes where an image fits. Figure 38b displays this view of the empty scene.
A square is placed using image coordinates, not pixel coordinates. For instance, the
blue square in Fig. 38a is at image coordinate (3,3).
21 The Scene for Aliens Attack 73

So, why should we care about the perspective we adopt? It turns out that data
representation strongly influences how a problem is reasoned about. Consider, for
example, how is an alien hit by a shot is detected. We shall first analyze the problem
using the pixel perspective of the game’s scene. Assume that the image of an alien
has dimensions W x H as illustrated in Fig. 39 and the alien is located at pixel
coordinates (x, y). We define a hit as the shot’s position being inside the alien’s
image. That is, if the shot’s position is any pixel covered by the alien’s image, then
the alien is hit. In Fig. 39, for example, the shots at (xr1 ,yr1 ) and at (xr2 ,yr2 ) have
hit the alien. The shots at (xr3 ,yr3 ) and (xr4 ,yr4 ) have not hit the alien. Now,
how is a hit detected? After thinking about it, you can see that a hit means that the
shot can be at most half the alien’s image width (i.e., W/2) away on the x-axis and
at most half the alien’s image height (i.e., H/2) away on the y-axis from the alien’s
coordinates (i.e., (x, y)). This reasoning leads to a function to detect a hit alien
that looks as follows:
(define HALF-ALIEN-IMG-WIDTH (/ (image-width FUEL-IMG) 2))

(define HALF-ALIEN-IMG-HEIGHT (/ (image-height FUEL-IMG) 2))

(define (alien-hit? x-alien y-alien x-shot y-shot)

(and (<= (abs (- x-alien x-shot)) HALF-ALIEN-IMG-WIDTH)
(<= (abs (- y-alien y-shot)) HALF-ALIEN-IMG-HEIGHT)))
Now, consider solving the same problem using an image perspective of the game’s
scene. Under the image perspective, we do not have to worry about every pixel in
the scene. Images can only be placed in a box with image coordinates x and y. This
means that an alien is hit by a shot if both have the same coordinates. This reasoning
leads to a function to detect a hit alien that looks as follows:
(define (alien-hit? x-alien y-alien x-shot y-shot)
(and (= x-alien x-shot) (= y-alien y-shot)))

Fig. 38 Empty scene perspectives

(a) Pixel Perspective. (b) Image Perspective.
0 1 0 1
0 0

1 1
74 4 Aliens Attack Version 0

Fig. 39 Detecting a hit alien under the pixel perspective

H/2

W/2 (x,y) W/2

H/2

This function is much simpler and easier to understand than the first. The important
lesson here is that different representations, like pixel coordinates versus image
coordinates, may lead to vastly different solutions/programs. It is worth exploring
different representations to determine if any lead to a simpler and easier to understand
solution/program.
Given that we have carefully analyzed how to represent coordinates in the game’s
scene, we may be cautiously confident that using image coordinates leads to simpler
problem solutions. We can also surmise that we need to be careful about defining
the data representation in the program. For this, we need data definitions. A data
definition is written to describe a new type of data. For Aliens Attack, the following
data definitions and constants are used:
;; Character Image Maximum Dimensions
(define IMAGE-WIDTH 30)
(define IMAGE-HEIGHT 30)

;; Video Game Scene Dimensions

(define MAX-CHARS-HORIZONTAL 20)
(define MAX-CHARS-VERTICAL 15)

;; Empty Scene Constants

(define E-SCENE-COLOR ’pink)
(define E-SCENE2-COLOR ’black)
(define E-SCENE-W (* MAX-CHARS-HORIZONTAL IMAGE-WIDTH))
(define E-SCENE-H (* MAX-CHARS-VERTICAL IMAGE-HEIGHT))

#|DATA DEFINITIONS
A character image (ci) is an image which is at most
IMAGE-WIDTH x IMAGE-HEIGHT pixels
21 The Scene for Aliens Attack 75

An image-x is an integer in [0..MAX-CHARS-HORIZONTAL-1]

An image-y is an integer in [0..MAX-CHARS-VERTICAL-1]

A scene is a E-SCENE-W x E-SCENE-H image

A pixel-x is an integer in [0..E-SCENE-W-1]

A pixel-y is an integer in [0..E-SCENE-H-1] |#

;; Sample image-x
(define AN-IMG-X (/ MAX-CHARS-HORIZONTAL 2))
(define MIN-IMG-X 0)
(define MAX-IMG-X (sub1 MAX-CHARS-HORIZONTAL))

;; Sample image-y
(define AN-IMG-Y (/ MAX-CHARS-VERTICAL 2))
(define MIN-IMG-Y 0)
(define MAX-IMG-Y (sub1 MAX-CHARS-VERTICAL))

;; Sample empty scenes

(define E-SCENE (empty-scene E-SCENE-W
E-SCENE-H
E-SCENE-COLOR))
(define E-SCENE2 (empty-scene E-SCENE-W
E-SCENE-H
E-SCENE2-COLOR))
The define constants clearly communicate that for any ci in the game (i.e., rocket,
alien, and shot), the maximum width is 30 pixels and the maximum height is 30 pixels.
The scene fits 20 ci horizontally and 15 ci vertically. This means image coordinates
range from (0,0) to (19, 14). Examples of image-x and image-y values are defined
to make writing tests easier. These include the extrema values: the largest and the
smallest possible values. Pixel coordinates range from (0,0) to ((sub1 E-SCENE-W),
(sub1 E-SCENE-H)). Clearly, we need a mechanism to map image coordinates to
pixel coordinates. We shall return to this new problem after tackling the creation of
character images.
The data definitions clearly describe what is (and what is not) a ci, an
image-x, an image-y, a scene, a pixel-x, and a pixel-y. For example, 2 is
a valid image-x because 2 is in [0..(sub1 MAX-CHARS-HORIZONTAL)]. On
the other hand, 1000 is not a valid image-x because 1000 is not in [0..(sub1
MAX-CHARS-HORIZONTAL)]. These data definitions guide the development of func-
tions for the video game. Furthermore, the newly defined data types may be used in
signatures. That is, the new data types may be used to describe the inputs to and the
output of functions.
76 4 Aliens Attack Version 0

22 Creating Aliens Attack Images

Creating the images needed for Aliens Attack provides the opportunity to practice
the steps of the design recipe. In addition, it provides us with the opportunity to write
an application programming interface (API). An API defines the data formats, the
conventions, and the functions that may be used to create and access data. For Aliens
Attack images, the API is rather simple: the dimensions of any character image must
be 30 x 30 pixels or less and must be created using functions in the image teachpack.
One of the benefits of using or creating an API is information hiding. This means,
for example, that the implementation details of a function are hidden from any
programmer who uses the function. A programmer uses the function independent of
how the function is implemented. To be successful, the signature and the purpose of
the function must be clear. Otherwise, programmers will not know how to properly
use a function.
Observe that according to the data definitions, every ci is an image. Every image,
however, is not a ci. A natural question that arises is determining if an image is a
ci. Testing image values is useless, because everybody’s idea of a shot, an alien, or
a rocket is different. Furthermore, the data definition ci only specifies ci properties
(not values). This is a clear situation where property-based testing is needed. For
this, a predicate to determine if a given image is a ci is needed. Following the steps
of the design recipe, the predicate may be designed as follows:
STEP 1
Determine that both the image width is less than or equal to IMAGE-WIDTH and
that the image height is less than or equal to IMAGE-HEIGHT.
STEP 2
Definitions for the value of sample expressions are

(define IS-CI
(and (<= (image-width (circle 10 'solid 'red))
IMAGE-WIDTH)
(<= (image-height (circle 10 'solid 'red))
IMAGE-HEIGHT)))

(define NOT-CI
(and (<= (image-width (square 40 'solid 'blue))
IMAGE-WIDTH)
(<= (image-height (square 40 'solid 'blue))
IMAGE-HEIGHT)))

(define NOT-CI2
(and (<= (image-width
(rectangle 2 50 'solid 'blue))
IMAGE-WIDTH)
22 Creating Aliens Attack Images 77

(<= (image-height
(rectangle 2 50 'solid 'blue))
IMAGE-HEIGHT)))

Observe that sample expressions are developed for both images that are and that
are not ci.
STEP 3
The only difference among the sample expressions is the image tested. The
variable for this difference is an-img.
STEP 4
The signature and purpose statement are

;; image → Boolean
;; Purpose: To determine if the given image is a ci

STEP 5
The function header is
(define (ci? an-img)
STEP 6
Testing may be satisfied as follows:

;; Tests using sample computations for ci?

(check-expect (ci? (circle 10 'solid 'red)) IS-CI)
(check-expect (ci? (square 40 'solid 'blue)) NOT-CI)
(check-expect (ci? (rectangle 20 40 'solid 'blue)) NOT-CI2)

;; Tests using sample values for ci?

(check-expect (ci? (ellipse 10 22 'outline 'green)) #true)
(check-expect (ci? (rectangle 5 33 'solid 'yellow)) #false)

Observe that both images that are and are not ci are tested.
STEP 7
The body of the function is

(and (<= (image-width an-img) IMAGE-WIDTH)

(<= (image-height an-img) IMAGE-HEIGHT))

STEP 8
All tests pass.

Recall that this step is only satisfied when the tests are executed and pass.
78 4 Aliens Attack Version 0

* Ex. 45 — Design and implement a predicate to determine if an image is a

scene.
* Ex. 46 — Design and implement a predicate to determine if a number is an
image-x.

* Ex. 47 — Design and implement a predicate to determine if a number is an

image-y.

* Ex. 48 — Design and implement a predicate to determine if a number is a

pixel-x.

* Ex. 49 — Design and implement a predicate to determine if a number is a

pixel-y.

** Ex. 50 — A blank pink or black scene is rather dull. Personalize your

empty scene to provide a more interesting and more pleasant background.

Now that ci and non-ci images are distinguishable, attention turns to the con-
struction of cis. There are three image constructors that are needed: one for a shot,
one for an alien, and one for a rocket. The development presented here is only illustra-
tive. This is your video game and you are encouraged to be creative and personalize
the game to your liking. Understand the principles for developing the images and then
design functions for your own images. Remember that, unlike homework problems
in high school Mathematics, there may be many solutions to a problem.

23 Shot Image

The image of the shot is the simplest one in Fig. 37. The shot’s image is a radial
star and the color is orange. These, of course, may be different depending on your
preferences. The goal is to design a constructor for shot images. The steps of the
design recipe may be satisfied as follows:
STEP 1
A shot image is a ci of a radial star created using the function radial-star.
Observe that the data definition of a shot image is described in terms of the type
for character images previously developed.
STEP 2
Definitions for the value of sample expressions are

(define SHOT-COLOR 'orange)

(define SHOT-COLOR2 'skyblue)

23 Shot Image 79

(define SHOT-IMG (radial-star 8

(/ IMAGE-WIDTH 8)
(/ IMAGE-WIDTH 2)
'solid
SHOT-COLOR))

(define SHOT-IMG2 (radial-star 8

(/ IMAGE-WIDTH 8)
(/ IMAGE-WIDTH 2)
'solid
SHOT-COLOR2))

The constants are defined to facilitate changing the color of sample shots. Observe
that the outer diameter of the radial star is IMAGE-WIDTH/2 = 15. Therefore, the
image of a shot is a ci.
STEP 3
The only difference among the sample expressions is the color of the shot. The
variable for this difference is a-color.
STEP 4
The signature and purpose statement are

;; color → image
;; Purpose: Create a shot image of the given color

STEP 5
The function header is
(define (mk-shot-ci a-color)
STEP 6
The tests are below.

;; Tests using sample computations for mk-shot-ci

(check-expect (mk-shot-ci SHOT-COLOR) SHOT-IMG)
(check-expect (mk-shot-ci SHOT-COLOR2) SHOT-IMG2)
(check-expect (ci? SHOT-IMG) #true)
(check-expect (ci? SHOT-IMG2) #true)

;; Tests using sample values for mk-shot-ci

(check-expect (mk-shot-ci 'red) )

(check-expect (mk-shot-ci 'brown) )

(check-expect (ci? (mk-shot-img 'red)) #true)
(check-expect (ci? (mk-shot-img 'brown)) #true)
80 4 Aliens Attack Version 0

Observe that the last two sample computations and sample values tests are
property-based testing. They illustrate that the image returned by mk-shot-img
is a ci.
STEP 7
The body of the function is

(radial-star 8 (/ IMAGE-WIDTH 8) (/ IMAGE-WIDTH 2) 'solid a-color))

STEP 8
All tests pass.

24 Alien Image

The alien image in Fig. 37 is the composition of two images of the same color: an
X and a circle. Specifically, the circle is overlayed over the X. A function to achieve
this uses function composition. The steps of the design recipe may be satisfied as
follows:
STEP 1
The alien image is computed using function composition. The outputs of text
and circle are inputs to overlay. The circle image is overlayed over the X
image. The alien image must be a ci.
STEP 2
Definitions for the value of sample expressions are

(define ALIEN-COLOR 'black)

(define ALIEN-COLOR2 'orange)

(define ALIEN-IMG (overlay (text "X" 25 ALIEN-COLOR)

(circle (/ IMAGE-WIDTH 4)
'solid
ALIEN-COLOR)))

(define ALIEN-IMG2 (overlay (text "X" 25 ALIEN-COLOR2)

(circle (/ IMAGE-WIDTH 4)
'solid
ALIEN-COLOR2)))

Getting the alien images to be at most 30 × 30 pixels requires experimentation.

It is not clear from the code above that the resulting images have this property.
Property-based testing may be used to demonstrate this.
24 Alien Image 81

STEP 3
The only difference among the sample expressions is the color of the alien. The
variable for this difference is a-color.
STEP 4
The signature and purpose statement are

;; color → image
;; Purpose: Create an alien image of the given color

STEP 5
The function header is

(define (mk-alien-ci a-color)

STEP 6
The tests are below.

;; Tests using sample computations for mk-alien-ci

(check-expect (mk-alien-ci ALIEN-COLOR) ALIEN-IMG)
(check-expect (mk-alien-ci ALIEN-COLOR2) ALIEN-IMG2)
(check-expect (ci? ALIEN-IMG) #true)
(check-expect (ci? ALIEN-IMG2) #true)

;; Tests using sample values for mk-alien-ci

(check-expect (mk-alien-ci 'purple) )

(check-expect (mk-alien-ci 'lightbrown) )

(check-expect (ci? (mk-alien-img 'purple)) #true)
(check-expect (ci? (mk-alien-img 'lightbrown)) #true)

STEP 7
The body of the function is

(overlay (text "X" 25 a-color)

(circle (/ IMAGE-WIDTH 4) 'solid a-color))

STEP 8
All tests pass.
82 4 Aliens Attack Version 0

Fig. 40 Components of the rocket image

(a) Rocket Window.

(b) Rocket Fuselage. (c) Rocket Single Booster.

(d) Rocket Booster.

(e) Rocket Nacelle.

(f) Rocket Main Body.

25 Rocket Image

The rocket image is the most complex of the images in Fig. 37. It is composed of the
images displayed in Fig. 40. To design a function to create a rocket image, a divide
and conquer bottom-up strategy is followed. That is, the simplest components are
independently designed first. Then simple components are combined to create more
complex figures.
It is important to keep in mind that creating images for the components of a larger
image requires experimentation. Rarely, if ever, are the dimensions of a component
right after the first attempt. This is another reason why it is important to write sample
expressions. Sample expressions may be repeatedly refined until the desired visual
effect is achieved. The steps of the design recipe presented for each component
below are the final result after several refinements improving the dimensions of each
component. Do not be afraid to experiment. Experimenting with sample expressions
may lead to insights on how to solve a problem.

25.1 Rocket Window Image Constructor

Creating a rocket window image is a simple exercise with which to practice the steps
the design recipe.
STEP 1
A rocket window is a solid vertical oval.
STEP 2
Definitions for the value of sample expressions are
25 Rocket Image 83

(define WINDOW-COLOR 'darkgray)

(define WINDOW2-COLOR 'white)

(define WINDOW (ellipse 3 10 'solid WINDOW-COLOR))

(define WINDOW2 (ellipse 3 10 'solid WINDOW2-COLOR))

STEP 3
The only difference among the sample expressions is the color of the oval. The
variable for this difference is a-color.
STEP 4
The signature and purpose statement are

;; color → image
;; Purpose: Create rocket window image

STEP 5
The function header is

(define (mk-window-img a-color)

STEP 6
Testing may be satisfied as follows:

;; Tests using sample computations for mk-window-img

(check-expect (mk-window-img WINDOW-COLOR) WINDOW)
(check-expect (mk-window-img WINDOW2-COLOR) WINDOW2)

;; Tests using sample values for mk-window-img

(check-expect (mk-window-img 'darkblue) )
(check-expect (mk-window-img 'gray) )

STEP 7
The body of the function is

(ellipse 3 10 'solid a-color)

STEP 8
All tests pass.
84 4 Aliens Attack Version 0

In DrRacket, the code ought to be organized as follows:

;; color → image
;; Purpose: Create rocket window image
(define (mk-window-img a-color)
(ellipse 3 10 'solid a-color))

;; Sample expressions for mk-window-img

(define WINDOW (ellipse 3 10 'solid WINDOW-COLOR))
(define WINDOW2 (ellipse 3 10 'solid WINDOW2-COLOR))

;; Tests using sample computations for mk-window-img

(check-expect (mk-window-img WINDOW-COLOR) WINDOW)
(check-expect (mk-window-img WINDOW2-COLOR) WINDOW2)

;; Tests using sample values for mk-window-img

(check-expect (mk-window-img 'darkblue) )
(check-expect (mk-window-img 'gray) )
Observe that the steps of the design recipe are written in an organized fashion
keeping the function, the sample expressions, and the tests together. The constants
WINDOW-COLOR and WINDOW-COLOR2 ought to be defined in a section dedicated to
constants toward the top of the program.

25.2 Rocket Fuselage Image Constructor

STEP 1
The fuselage is a solid circle.
STEP 2
Definitions for the value of sample expressions are

(define FUSELAGE (circle (* 1/3 IMAGE-HEIGHT)

'solid
FUSELAGE-COLOR))

(define FUSELAGE2 (circle (* 1/3 IMAGE-HEIGHT)

'solid
FUSELAGE2-COLOR))

STEP 3
The only difference among the sample expressions is the color. The variable for
this difference is a-color.
25 Rocket Image 85

STEP 4
The signature and purpose statement are

;; color → image
;; Purpose: Create the fuselage image

STEP 5
The function header is
(define (mk-fuselage-img a-color)
STEP 6
Testing may be satisfied as follows:

;; Test using sample computations for mk-fuselage-img

(check-expect (mk-fuselage-img FUSELAGE-COLOR) FUSELAGE)
(check-expect (mk-fuselage-img FUSELAGE2-COLOR) FUSELAGE2)

;; Test using sample values for mk-fuselage-img

(check-expect (mk-fuselage-img 'olive) )

(check-expect (mk-fuselage-img 'lightgray) )

STEP 7
The body of the function is

(circle (* 1/3 IMAGE-HEIGHT) 'solid a-color))

STEP 8
All tests pass.

25.3 Rocket Single Booster Image Constructor

The single booster image is an equilateral triangle pointing down. Computing this
image requires function composition. The output of triangle is input to rotate.
STEP 1
A rocket single booster is an equilateral triangle image rotated 180◦ .
STEP 2
Definitions for the value of sample expressions are
86 4 Aliens Attack Version 0

(define SINGLE-BOOSTER (rotate 180

(triangle (/ FUSELAGE-W 2)
'solid
NACELLE-COLOR)))
(define SINGLE-BOOSTER2 (rotate 180
(triangle (/ FUSELAGE-W 2)
'solid
NACELLE2-COLOR)))

Observe that in Fig. 40, the single booster image is the same color as the nacelle
image. Therefore, constants NACELLE-COLOR and NACELLE2-COLOR are defined
and used to write sample single booster (and nacelle) expressions.
STEP 3
The only difference among the sample expressions is the color. The variable for
this difference is a-color.
STEP 4
The signature and purpose statement are

;; color → image
;; Purpose: Create single booster image

STEP 5
The function header is
(define (mk-single-booster-img a-color)
STEP 6
Testing may be satisfied as follows:

;; Tests using sample computations for mk-single-booster-img

(check-expect (mk-single-booster-img NACELLE-COLOR) SINGLE-BOOSTER)
(check-expect (mk-single-booster-img NACELLE2-COLOR) SINGLE-BOOSTER2)

;; Tests using sample values for mk-single-booster-img

(check-expect (mk-single-booster-img 'darkred) )
(check-expect (mk-single-booster-img 'gold) )

STEP 7
The body of the function is

(rotate 180 (triangle (/ FUSELAGE-W 2) 'solid a-color)))

STEP 8
All tests pass.
25 Rocket Image 87

25.4 Rocket Booster Image Constructor

Observe that the booster image in Fig. 40d is two copies of a single booster image
side by side. This suggests that a constructor for a booster image needs a single
booster image as input.
STEP 1
A booster image is two copies of a single booster image side by side.
STEP 2
Definitions for the value of sample expressions are

(define BOOSTER (beside SINGLE-BOOSTER SINGLE-BOOSTER))

(define BOOSTER2 (beside SINGLE-BOOSTER2 SINGLE-BOOSTER2))

STEP 3
The only difference among the sample expressions is the single booster image.
The variable for this difference is a-sb-img.
STEP 4
The signature and purpose statement are

;; image → image
;; Purpose: Create booster image

STEP 5
The function header is
(define (mk-booster-img a-sb-img)
STEP 6
Testing may be satisfied as follows:

;; Tests using sample computations for mk-booster-img

(check-expect (mk-booster-img SINGLE-BOOSTER) BOOSTER)
(check-expect (mk-booster-img SINGLE-BOOSTER2) BOOSTER2)

;; Tests using sample values for mk-booster-img

(check-expect (mk-booster-img (mk-single-booster-img 'darkred)) )
(check-expect (mk-booster-img (mk-single-booster-img 'gold)) )

STEP 7
The body of the function is

(beside a-sb-img a-sb-img))

88 4 Aliens Attack Version 0

STEP 8
All tests pass.

25.5 Rocket Main Body Image Constructor

Constructing an image for the rocket’s main body, as displayed in Fig. 40f, requires
three images: a window image, a fuselage image, and a booster image. It also requires
the use of function composition given that the fuselage goes above the booster and
the window goes on the fuselage.
STEP 1
The rocket’s main body is created by placing the fuselage above the booster and
then placing the window one quarter of the way down the height and half away
across the width of the fuselage.
STEP 2
Definitions for the value of sample expressions are

(define ROCKET-MAIN (place-image

WINDOW
(/ (image-width FUSELAGE) 2)
(/ (image-height FUSELAGE) 4)
(above FUSELAGE BOOSTER)))

(define ROCKET-MAIN2 (place-image

WINDOW2
(/ (image-width FUSELAGE2) 2)
(/ (image-height FUSELAGE2) 4)
(above FUSELAGE2 BOOSTER2)))

Observe that previously defined sample images are used to write the sample
expressions.
STEP 3
There are 3 differences among the sample expressions: the window, the fuselage,
and the booster images. The variables for these differences are, respectively,
a-window, a-fuselage, and a-booster.
STEP 4
The signature and purpose statement are

;; image image image → image

;; Purpose: Create the main rocket image
25 Rocket Image 89

STEP 5
The function header is
(define (mk-rocket-main-img a-window a-fuselage a-booster)
STEP 6
Testing may be satisfied as follows:

;; Tests using sample computations for mk-rocket-main-img

(check-expect (mk-rocket-main-img WINDOW FUSELAGE BOOSTER)
ROCKET-MAIN)
(check-expect (mk-rocket-main-img
WINDOW2 FUSELAGE2 BOOSTER2)
ROCKET-MAIN2)

;; Tests using sample computations for mk-rocket-main-img

(check-expect (mk-rocket-main-img
(mk-window-img 'black)
(mk-fuselage-img 'yellow)
(mk-booster-img
(mk-single-booster-img 'yellow)))

(check-expect (mk-rocket-main-img
(mk-window-img 'green)
(mk-fuselage-img 'skyblue)
(mk-booster-img
(mk-single-booster-img 'lightred)))

STEP 7
The body of the function is

(place-image a-window
(/ (image-width a-fuselage) 2)
(/ (image-height a-fuselage) 4)
(above a-fuselage a-booster))

STEP 8
All tests pass.
90 4 Aliens Attack Version 0

Fig. 41 Rocket image

25.6 Rocket Nacelle Image Constructor

The image of the rocket’s nacelle in Fig. 40e is deceptively simple. At first glance,
it looks like an ordinary rectangle. A closer examination of the rocket image, as
displayed in Fig. 41, reveals that the nacelle and the booster are the same color. In
addition, the nacelle must be the same width as the rocket’s main body and about a
quarter of the height of the main body. With these observations, the design of the
constructor for nacelle images follows.
STEP 1
The image of a nacelle is a rectangle that is the same width as the rocket’s main
body and a quarter of the height of the rocket’s main body height. The color of
the nacelle is the same color as the booster in the rocket’s main body.
STEP 2
Definitions for the value of sample expressions are

(define NACELLE (rectangle (image-width ROCKET-MAIN)

(/ (image-height ROCKET-MAIN)
4)
'solid
NACELLE-COLOR))

(define NACELLE2 (rectangle (image-width ROCKET-MAIN2)

(/ (image-height ROCKET-MAIN2)
4)
'solid
NACELLE2-COLOR))

Observe that the same constants used to construct single booster images are used
to construct the corresponding nacelle images.
STEP 3
There are two differences among the sample expressions: the main rocket image
and the color of the nacelle. The variables for these differences are, respectively,
a-rocket-main-img and a-color.
STEP 4
The signature and purpose statement are

;; image color → image

;; Purpose: Create a rocket nacelle image
25 Rocket Image 91

STEP 5
The function header is
(define (mk-nacelle-img a-rocket-main-img a-color)
STEP 6
Testing may be satisfied as follows:

;; Tests using sample computations for mk-nacelle-img

(check-expect (mk-nacelle-img ROCKET-MAIN NACELLE-COLOR)
NACELLE)
(check-expect (mk-nacelle-img ROCKET-MAIN2 NACELLE2-COLOR)
NACELLE2)

;; Tests using sample values for mk-nacelle-img

(check-expect (mk-nacelle-img
(mk-rocket-main-img
(mk-window-img 'blue)
(mk-fuselage-img 'green)
(mk-booster-img
(mk-single-booster-img 'yellow)))
'yellow)
)
(check-expect (mk-nacelle-img
(mk-rocket-main-img
(mk-window-img 'blue)
(mk-fuselage-img 'green)
(mk-booster-img
(mk-single-booster-img 'lightorange)))
'lightorange)
)

STEP 7
The body of the function is

(rectangle (image-width a-rocket-main-img)

(/ (image-height a-rocket-main-img) 4)
'solid
a-color))

STEP 8
All tests pass.
92 4 Aliens Attack Version 0

25.7 Rocket ci Constructor

The final image constructor needed is the rocket image constructor. This constructor
must place the nacelle over the rocket’s main body.
STEP 1
The rocket image is constructed by placing the nacelle image half the way across
the rocket’s main body and 30% from the bottom of the rocket’s main body.
STEP 2
Definitions for the value of sample expressions are

(define ROCKET-IMG (place-image

NACELLE
(/ (image-width ROCKET-MAIN) 2)
(* 0.7 (image-height ROCKET-MAIN))
ROCKET-MAIN))

(define ROCKET-IMG2 (place-image

NACELLE2
(/ (image-width ROCKET-MAIN2) 2)
(* 0.7 (image-height ROCKET-MAIN2))
ROCKET-MAIN2))

Notice that 30% from the bottom is the same as 70% from the top of the rocket’s
main body image. Since y values grow downward in the graphic plane, 70% of
the rocket’s main body image is computed.
STEP 3
There are two differences among the sample expressions: the rocket’s main body
image and the rocket’s nacelle image. The variables for these differences are
a-rocket-main-img and a-nacelle-img.
STEP 4
The signature and purpose statement are

;; image image → ci
;; Purpose: Create a rocket ci

Observe that the signature states that this function returns a ci. This is needed
because the returned image is intended for use in Aliens Attack.
STEP 5
The function header is
(define (mk-rocket-ci a-rocket-main-img a-nacelle-img)
STEP 6
Testing may be satisfied as follows:
25 Rocket Image 93

;; Tests using sample computations for mk-rocket-img

(check-expect (mk-rocket-ci ROCKET-MAIN NACELLE)
ROCKET-IMG)
(check-expect (mk-rocket-ci ROCKET-MAIN2 NACELLE2)
ROCKET-IMG2)
(check-expect (ci? ROCKET-IMG) #true)
(check-expect (ci? ROCKET-IMG2) #true)

;; Tests using sample values for mk-rocket-img

(check-expect (mk-rocket-ci
(mk-rocket-main-img
(mk-window-img
'blue)
(mk-fuselage-img
'green)
(mk-booster-img
(mk-single-booster-img 'yellow)))
(mk-nacelle-img
(mk-rocket-main-img
(mk-window-img
'blue)
(mk-fuselage-img 'green)
(mk-booster-img
(mk-single-booster-img 'yellow)))
'yellow))

Observe that the output of this function must be a character image and, therefore,
ci? is used to test its properties. Furthermore, observe that there is a single
sample value test and that this test contains a significant amount of repetition.
Both of these are considered poor style. Testing ought to be more thorough. The
repetitions make it difficult to read and understand the test.
STEP 7
The body of the function is

(place-image a-nacelle-img
(/ (image-width a-rocket-main-img) 2)
(* 0.7 (image-height a-rocket-main-img))
a-rocket-main-img)
94 4 Aliens Attack Version 0

STEP 8
All tests pass.

** Ex. 51 — Refine the tests using sample values for mk-rocket-ci. Make
sure to eliminate repeated expressions and include more than one test.

*** Ex. 52 — Personalize your game with cis for the shot, the alien, and the
rocket of your own creation.

26 Drawing Functions

In contrast to the development of mk-rocket-img, we shall use a top-down approach

to design the function to draw a ci in a scene. This computation requires a ci, an
image-x, an image-y, and a scene. Note that place-image cannot simply be used
to place a ci at the coordinate formed by the given image-x and image-y. Recall
Fig. 38 and the data definitions for image and pixel coordinates. There are many
more pixel coordinates than image coordinates given that one image point, (xi ,
yi ), covers many pixel points. We must now solve the problem of mapping image
coordinates to pixel coordinates. The design starts with the topmost function, which
draws a ci in a scene.
STEP 1
The given ci is placed in the given scene by translating the given image coordi-
nates to pixel coordinates.
STEP 2
Definitions for the value of sample expressions are

(define ROCKET-Y (sub1 MAX-CHARS-VERTICAL))

(define ROCKET-SCN (place-image

ROCKET-IMG
(image-x->pix-x
(/ MAX-CHARS-HORIZONTAL 2))
(image-y->pix-y ROCKET-Y)
E-SCENE))

(define ALIEN-SCN (place-image ALIEN-IMG

(image-x->pix-x 4)
(image-y->pix-y 7)
E-SCENE2))
26 Drawing Functions 95

(define SHOT-SCN (place-image

SHOT-IMG
(image-x->pix-x 17)
(image-y->pix-y 13)
E-SCENE2))

Observe that the three sample expressions evaluate to a scene because a ci

is placed in a scene. Further notice that two auxiliary functions are needed to
translate the image coordinates. The constant ROCKET-Y stems from thinking
ahead and observing that the rocket’s image-y coordinate is a constant in Aliens
Attack.
STEP 3
There are four differences among the sample expressions: the ci, the image-x and
-y coordinates, and the scene. The variables for these differences are char-img,
an-img-x, an-img-y, and scn.
STEP 4
The signature and purpose statement are

;; ci image-x image-y scene → scene

;; Purpose: Place the given ci in the given scene at the
;; given image coordinates

STEP 5
The function header is
(define (draw-ci char-img an-img-x an-img-y scn)
STEP 6
Testing may be satisfied as follows:

;; Tests using sample computations for draw-ci

(check-expect (draw-ci ROCKET-IMG
(/ MAX-CHARS-HORIZONTAL 2)
ROCKET-Y
E-SCENE)
ROCKET-SCN)
(check-expect (draw-ci ALIEN-IMG 4 7 E-SCENE) ALIEN-SCN)
(check-expect (draw-ci SHOT-IMG 17 13 E-SCENE) SHOT-SCN)

;; Tests using sample values for draw-ci

(check-expect (draw-ci
(square 10 'solid 'green) 8 8 E-SCENE)
96 4 Aliens Attack Version 0

Fig. 42 Top-left box of the scene image perspective

IMAGE-WIDTH/2

IMAGE-HEIGHT/2

(check-expect (draw-ci (ellipse 20 30 'outline 'black)

11 12 E-SCENE)

STEP 7
The body of the function is

(place-image char-img
(image-x->pix-x an-img-x)
(image-y->pix-y an-img-y)
scn)

STEP 8
All tests pass.
The design now focuses on the two auxiliary functions to map image coordinates
to pixel coordinates. Let us first consider mapping the image coordinate pair (0, 0) to
26 Drawing Functions 97

a pixel coordinate pair. This coordinate pair is the top-left box in Fig. 38b. The center
of the image ought to be placed at the center of the box. The dimensions of the box
are IMAGE-WIDTH × IMAGE-HEIGHT. As illustrated in Fig. 42, this means that the
center of this box is at pixel coordinates (IMAGE-WIDTH/2, IMAGE-HEIGHT/2).
What is the pixel coordinate of image coordinate (3, 2)? This pixel coordinate
(IMAGE-WIDTH/2, IMAGE-HEIGHT/2) must be translated 3 boxes to the right and
2 boxes down. Therefore, the pixel coordinate for the box at the image coordinate (3,
2) is:
(3 * IMAGE-WIDTH + IMAGE-WIDTH/2,
2 * IMAGE-HEIGHT + IMAGE-HEIGHT/2)
In general, the image coordinate (x, y) maps to the pixel coordinate:
(x * IMAGE-WIDTH + IMAGE-WIDTH/2,
y * IMAGE-HEIGHT + IMAGE-HEIGHT/2)
Observe that the mapping works for image coordinate pair (0, 0).
The above insight guides the design of the image-coordinate to pixel-coordinate
functions. The steps of the design recipe to transform an image-x may be satisfied
as follows:
STEP 1
For image-x, ix, the corresponding pixel-x is ix * IMAGE-WIDTH
+ IMAGE-WIDTH/2
STEP 2
Definitions for the value of sample expressions are

(define PIX-X5 (+ (* 5 IMAGE-WIDTH) (/ IMAGE-WIDTH 2)))

(define PIX-X12 (+ (* 12 IMAGE-WIDTH) (/ IMAGE-WIDTH 2)))

STEP 3
The only difference among the sample expressions is the image-x value. The
variable for this difference is ix.
STEP 4
The signature and purpose statement are

;; image-x → pixel-x
;; Purpose: To translate the given image-x to a pixel-x

STEP 5
The function header is
(define (image-x->pix-x ix)
98 4 Aliens Attack Version 0

STEP 6

;; Tests using sample computations for x->pix-x

(check-expect (image-x->pix-x 5) PIX-X5)
(check-expect (image-x->pix-x 12) PIX-X12)

;; Tests using sample values for x->pix-x

(check-expect (image-x->pix-x 0) (/ IMAGE-WIDTH 2))
(check-expect (image-x->pix-x (sub1 MAX-CHARS-HORIZONTAL))
585)

Observe that the tests using sample values are written for extrema values. That is,
the minimum and maximum possible values of an image-x are used.
STEP 7
The body of the function is

(+ (* ix IMAGE-WIDTH) (/ IMAGE-WIDTH 2)))

STEP 8
All tests pass.
The design of image-x->pix-x does not reveal the need for any new auxiliary
functions. The only task left is to design a function to translate an image-y to a
pixel-y. The steps of the design recipe may be satisfied as follows:
STEP 1
For image-y, iy, the corresponding pixel-y is iy * IMAGE-HEIGHT +
IMAGE-HEIGHT/2
STEP 2
Definitions for the value of sample expressions are

(define PIX-Y1 (+ (* 1 IMAGE-HEIGHT) (/ IMAGE-HEIGHT 2)))

(define PIX-Y6 (+ (* 6 IMAGE-HEIGHT) (/ IMAGE-HEIGHT 2)))

STEP 3
The only difference among the sample expressions is the image-y value. The
variable for this difference is iy.
STEP 4
The signature and purpose statement are

;; image-y → pixel-y
;; Purpose: To translate the given image-y to a pixel-y
26 Drawing Functions 99

STEP 5

The function header is

(define (image-y->pix-y iy)

STEP 6

;; Tests using sample computations for y->pix-y

(check-expect (image-y->pix-y 1) PIX-Y1)
(check-expect (image-y->pix-y 6) PIX-Y6)

;; Tests using sample values for y->pix-y

(check-expect (image-y->pix-y 0) (/ IMAGE-HEIGHT 2))
(check-expect (image-y->pix-y (sub1 MAX-CHARS-VERTICAL))
435)

Once again, observe that the tests using sample values are written for extrema
values.
STEP 7
The body of the function is

(+ (* iy IMAGE-HEIGHT) (/ IMAGE-HEIGHT 2)))

STEP 8
All tests pass.
The design of image-y->pix-y does not reveal the need for any new auxiliary
functions. This means that all the functions needed to draw a ci have been designed
and implemented.

** Ex. 53 — What happens if draw-ci does not translate image coordinates

to pixel coordinates?

* Ex. 54 — Use draw-ci and function composition to create a possible scene

for Aliens Attack with a rocket, multiple aliens, and multiple shots. Figure 37
is an example of such a scene.
100 4 Aliens Attack Version 0

27 What Have We Learned in This Chapter?

The important lessons of Chap. 6 are summarized as follows:

• The data representation chosen strongly influences how we think about and solve
problems.
• Exploring different data representations may lead to simpler programs.
• Data definitions are written to describe new data types.
• New data types may be used in function signatures.
• An API defines data formats, conventions, constructors, and selectors.
• Signatures and purpose statements make an API’s information hiding useful.
• Property-based testing may be useful in illustrating that the conventions of an
API are met.
Chapter 5
Making Decisions

Making decisions is part of life. For example, while driving on a highway, the driver
monitors the car’s speed. If the speed is too fast, then the driver decreases the speed.
If the speed is too slow, then the driver increases the speed. Otherwise, the driver
does not change the speed. It should not come as a surprise that making decisions is
also part of problem solving. Modern cars are equipped with a cruise control system.
The driver sets the cruise control at a certain speed. If the speed is too fast, then the
cruise control program decreases the speed. If the speed is too slow, then the cruise
control program increases the speed. Otherwise, the cruise control program does not
change the speed.
Regardless of whether it is a driver or a cruise control program, the decision
on how to adjust the speed depends on the car’s current speed. This decision
is may be represented by one of three possible expressions, say, 'increase,
'decrease, or 'steady. Think carefully about what this means. Multiple ex-
pressions for the value of a function means that determining how to adjust the
speed is a compound function. We can, in fact, write a mathematical func-
tion:
⎧
⎪
⎨'decrease if too-fast?(a-speed)
speed-change(a-speed) = 'increase if too-slow?(a-speed)
⎪
⎩
'steady otherwise

Two predicates, too-fast? and too-slow?, are used to evaluate the given speed.
In essence, the above function is stating that there is data variety in the domain of the
function. The input may be in one of three speed intervals indicating that the car is
either traveling too fast, too slow, or neither. In order to determine the speed change
a decision must be made as to which of the 3 expressions needs to be evaluated. This
decision is based on the speed range a-speed belongs to.
It turns out that speed ranges may be used to solve many problems. For example,
a police officer may use the following function to determine a course of action after

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 101
M. T. Morazán, Animated Problem Solving, Texts in Computer Science,
https://doi.org/10.1007/978-3-030-85091-3_5
102 5 Making Decisions

measuring the speed of a car:

⎧
⎪
⎨"Issue Ticket" if too-fast?(speed)
police-action(speed) = "Issue Warning" if too-slow?(speed)
⎪
⎩
"Take No Action" otherwise

Observe that this function must also decide which expression to use to provide an
answer for a given speed. Furthermore, observe that the two functions have the same
basic structure. This structural similarity stems from the fact that they process the
same type of data and suggests that a function template may be developed to process
a speed. A function template is like a skeleton for functions. It captures the expected
similarities among function that process the same type of data. For each data type a
problem solver defines a function template is developed.
For example, speed may be defined as follows:
A speed is a number such that either:
1. too-fast?(a-speed) is true
2. too-slow?(a-speed) is true
3. Neither too-fast?(a-speed) nor too-slow?(a-speed) is true
the template for functions that process speed is:
⎧
⎪
⎨. . . if too-fast?(a-speed)
f-on-speed(a-speed) = . . . if too-slow?(a-speed)
⎪
⎩
. . . otherwise
This template may be used to design any function that processes a speed. By providing
the missing expressions and a function name both functions above are obtained.
We can surmise from the examples above that compound functions arise from data
having variety. These functions need to decide which expression to evaluate based
on the variety the input belongs to. All functions that process the same data type with
variety have the same structure. This chapter introduces how to design functions that
make decisions. It introduces new BSL syntax to write conditional expressions which
are used to make decisions. A design recipe for compound functions is presented that
includes the development of data definitions for data with variety and of function
templates to capture structural similarities among functions. Several different types
of data with variety are explored.

28 Conditional Expressions in BSL

In BSL, there is syntax for expressions to make decisions. These new expressions are
called conditional expressions. The new expr production rule is:

expr ::= (cond [expr expr]+

[else expr])
28 Conditional Expressions in BSL 103

This rule states that a conditional expression starts with an opening parenthesis
and the keyword cond. Afterwards, there is 1 or more stanzas delimited by square
brackets. Each stanza contains two expressions. The first expression is a condition
and must evaluate to a Boolean. Think of this expression as a question whose
answer is either true or false. The second expression is called the consequence
and is only evaluated if the corresponding condition evaluates to #true. Think of
this expression as the value of the conditional expression when the answer to the
corresponding question is true. The last stanza in a conditional is called the default
stanza and contains in square brackets the keyword else and a single expression. This
expression represents the default value of the conditional. The default expression is
only evaluated if the conditions in all the other stanzas evaluate to #false.
The mathematical function speed-change(a-speed) is translated into BSL
syntax as follows:
(define (speed-change-bsl a-speed)
(cond [(too-fast? a-speed) 'decrease]
[(too-slow? a-speed) 'increase]
[else 'steady]))
The stanzas of a conditional are evaluated from top to bottom. For each stanza the
condition is evaluated first. If the condition evaluates to #true then the value of the
consequence is the value of the conditional. If the condition evaluates to #false
then evaluation continues with the next stanza. If the default stanza is reached, then
the default expression is evaluated to obtain the value of the conditional expression.
For example, for speed-change-bsl, if too-fast? evaluates to #true then the
value of the conditional is 'decrease. If it evaluates to #false then evaluation
continues with the next stanza. If too-slow? evaluates to #true then the value of
the conditional is 'decrease. If it evaluates to #false then evaluation continues
with the next stanza. The next stanza is the else stanza and, therefore, the default
expression is evaluated making the value of the conditional 'steady.
BSL provides shorthand when there are only two varieties of data. That is, short-
hand may be used when only the default stanza and one other stanza is needed. The
shorthand syntax is called an if-expression:

expr ::= (if expr expr expr)

This rule states that an if-expression is formed by an open parenthesis, three ex-
pressions, and a closing parenthesis. The first expression is called the condition and
must evaluate to a Boolean. If the condition is #true then the second expression is
evaluated to obtain the value of the if-expression. Otherwise, the third expression
is evaluated to obtain the value of the if-expression.
Consider, for example, translating the absolute value mathematical function into
BSL. In mathematical syntax the function is defined as follows:

a-num if a-num >= 0
abs-value(a-num) =
-a-num otherwise
104 5 Making Decisions

Observe that there are only two varieties of numbers for this function: nonnegative
and negative. Only two varieties means that an if-expression may be used to translate
it to BSL syntax:
(define (abs-value-bsl a-num)
(if (>= a-num 0)
a-num
(* -1 a-num)))
If the condition, (>= a-num 0), evaluates to #true then the value of the if-
expression is a-num. If the condition evaluates to #false then the value of the
if-expression is (* -1 a-num).

** Ex. 55 — Write a program to decide the speed change for a car if the
maximum speed is 120 kilometers per hour and the minimum speed is 50
kilometers per hour.

** Ex. 56 — Implement the following mathematical function in BSL:

⎧
⎪
⎨a-num
2 if a-num < 0
f(a-num) = a-num + 5 if anum < 5
⎪
⎩ a-num
e otherwise

29 Designing Functions to Process Data with Variety

When problem analysis reveals that data with variety needs to be processed then a
data definition must be created. Given that there is variety in the data, any function
that processes this type of data requires a conditional expression in its body. A
conditional expression is needed to determine which of the data varieties is being
processed. For each variety an expression for the value of the function needs to be
developed. A problem solver must also develop a function template to capture the
structural similarities among all functions that process the defined data type. Finally,
the number of tests must be greater than or equal to the number of varieties in the
data. In addition to at least one test per variety, it is also good practice to test border
cases that delimit varieties.
Let us examine poem verses as an example. First, we must choose how to rep-
resent a verse. We may, for example, represent a verse as a string. Some poets
feel that it is good practice to limit the length of each verse to 35 characters. A
verse with more than 35 characters is considered too long. A verse with 15 to 35
characters is considered fine. A verse with less than 15 characters is considered
too short. Based on this description, a data definition for a verse is formulated as
29 Designing Functions to Process Data with Variety 105

Fig. 43 Template for functions on a verse

;; verse ... ...
;; Purpose: ...
(define (f-on-verse a-verse ...)
(cond [(> (string-length a-verse) 35) ...]
[(<= 15 (string-length a-verse) 35) ...]
[else ...])

;; Expressions for sample computations

(define VARIETY1-CONSTANT ...)
(define VARIETY2-CONSTANT ...)
(define VARIETY3-CONSTANT ...)

;; Tests using sample computations

(check-expect (f-on-verse ...) VARIETY1-CONSTANT)
(check-expect (f-on-verse ...) VARIETY2-CONSTANT)
(check-expect (f-on-verse ...) VARIETY3-CONSTANT)
..
.

;; Test using sample values

(check-expect (f-on-verse ...) ...) ;; test for verse of length 35
(check-expect (f-on-verse ...) ...) ;; test for verse of length 15
..
.

follows:
;; A verse is either:
;; 1. A string of length greater than 35
;; 2. A string of length 15 to 35
;; 3. A string of length less than 15
The data definition for a verse informs problem solvers that any function that pro-
cesses a verse must have a conditional in its body that determines the variety of the
input verse. In addition, it informs problem solvers that at least 3 tests are needed:
one for each variety. Remember, we say at least because it is always good practice
to test boundary values. In this case, 5 tests are reasonable: one for each variety,
one for length 35, and one for length 15. It is important to note that verse vari-
eties are mutually exclusive. This means that no verse can be part of more than one
variety.
Based on the above data definition, observe that all functions that process a verse
share these features unique to verses:
1. A signature that has at least a verse as an input type.
2. A conditional in the body that determines if the given verse’s length is greater
than 35, between 15 and 34, or less than 15.
3. Sample expressions, if any, for each variety that is used to compute a value.
4. One test for each sample expression
5. Sample value tests including tests for borderline cases of length 35 and 15.
106 5 Making Decisions

This leads to the function template in Fig. 43. Observe that the signature includes a
verse as input and that each stanza in the skeleton’s body has a Boolean expression to
determine the variety of the verse. The default stanza, of course, does not explicitly
state that its Boolean expression is (< (string-length a-verse) 15). This
is unnecessary because verse varieties are mutually exclusive. Mutual exclusion
guarantees that if the first two Boolean expressions are false then the verse’s length
is less than 15. Stated differently:
¬(> (string-length a-verse) 35) ∧
¬(<= 15 (string-length a-verse) 35)
⇒
(< (string-length a-verse) 15)
This logical expression is stating that if the length of a verse is not greater than 35 and
it is not between 15 and 35 then it is less than 15. Finally, observe that the template
suggests 3 sample expressions and sample computation tests (one for each variety)
and 2 sample computation tests (one for each boundary value). The important point
to remember that there must be at least one test for each variety. More tests, of course,
may be added in the interest of clarity.
This function template can be specialized to express the solution to any problem
that requires processing a verse. Consider the problem of computing a string to
describe the length of a verse. To solve the problem, the function must decide
which constant string describing its variety to return: "Too Long", "Fine", or
"Too Short". Given that a verse is not processed to create any of these constant
strings, there are no sample expressions to describe how a verse is used to compute
these strings. If there are no sample expressions, there are also no tests using sample
computations.
The function’s signature, purpose, and header are:
;; verse → string
;; Purpose: Determine if the given verse is too long, fine,
;; or too short
(define (verse-type a-verse)
Observe that the signature uses the type, verse, defined by the data definition above.
Once a type is defined by a data definition it may be used in signatures. Signatures,
therefore, may contain types defined by BSL or by a data definition the programmer
creates. Further observe that the signature only has one input and, therefore, the
function only has one parameter. Finally, observe that the name of the function is
suggestive of its purpose and the name of the parameter is suggestive of the value it
represents.
Given that there are no sample expressions to test (as noted above), the tests using
sample values must be specialized to include a test for each variety of verse and for
each borderline length. Sample tests are:
;; Tests using sample values
(check-expect
(verse-type "Here is a sigh to those who love me,")
"Too Long")
29 Designing Functions to Process Data with Variety 107

Fig. 44 Program to determine verse type

;; A verse is either: 1. A string of length greater than 35
;; 2. A string of length 15 to 35
;; 3. A string of length less than 15
#| Funtion template for verse
;; verse ... --> ...
;; Purpose: ...
(define (f-on-verse a-verse ...)
(cond [(> (string-length a-verse) 35) ...]
[(<= 15 (string-length a-verse) 35) ...]
[else ...]))
;; Expressions for sample computations
(define VARIETY1-CONSTANT ...) (define VARIETY2-CONSTANT ...)
(define VARIETY3-CONSTANT ...)
;; Tests using sample computations
(check-expect (f-on-verse ...) VARIETY1-CONSTANT)
(check-expect (f-on-verse ...) VARIETY2-CONSTANT)
(check-expect (f-on-verse ...) VARIETY3-CONSTANT)
;; Test using sample values
(check-expect (f-on-verse ...) ...) ;; test for verse of length 35
(check-expect (f-on-verse ...) ...) ;; test for verse of length 15 |#

;; verse --> string

;; Purpose: Determine if the given verse is too long, fine, or too short
(define (verse-type a-verse)
(cond [(> (string-length a-verse) 35) "Too Long"]
[(<= 15 (string-length a-verse) 35) "Fine"]
[else "Too Short"]))
;; Tests using sample values
(check-expect (verse-type "Here is a sigh to those who love me,")
"Too Long")
(check-expect (verse-type "And a smile to those who hate") "Fine")
(check-expect (verse-type "Sorry for fate") "Too Short")
(check-expect (verse-type "The Battle of Culloden started now,") "Fine")
(check-expect (verse-type "Flaming heavens") "Fine")

(check-expect
(verse-type "And a smile to those who hate")
"Fine")
(check-expect (verse-type "Sorry for fate") "Too Short")
(check-expect
(verse-type "The Battle of Culloden started now,")
"Fine") ;; verse length 35
(check-expect
(verse-type? "Flaming heavens")
"Fine") ;; verse length 15
108 5 Making Decisions

Fig. 45 The design recipe for decision-making functions

1. Create data definitions with at least one having mutually exclusive varieties.
2. Develop a function template for each data definition.
3. Outline the computation.
4. Define constants for the value of sample expressions for each variety and name the differences.
5. Write the function’s signature and purpose.
6. Write the function’s header.
7. Write tests for each variety.
8. Write the function’s body.
9. Run the tests and, if necessary, redesign.

The next step is to specialize the body of the function in the template. For each
verse variety the proper string needs to be returned. The specialized body is:
(cond [(> (string-length a-verse) 35) "Too Long"]
[(<= 15 (string-length a-verse) 35) "Fine"]
[else "Too Short"]))
The complete program is displayed in Fig. 44. To make Fig. 44 easier to read the
comments including the data definition and the function template are displayed using
an italic font.
The development of verse-type suggests a series of steps that guide the de-
velopment of a decision-making function. Figure 45 displays the design recipe
for decision-making functions. This design recipe is a refinement of the general
design recipe for functions in Fig. 28. Therefore, the steps ought to feel familiar.
Step 1 requires the development of data definitions. Of these, at least one must
have varieties that are mutually exclusive. Step 2 requires the development of a
function template. The function’s body must be a conditional that has a stanza for
each variety. The Boolean expressions to determine the variety of the input are
part of the function’s body in the template. In addition, there must be at least one
test outlined for each data variety. Commonly (but not always), sample expression
tests are used to test each variety and sample value tests are used to test boundary
values. These first two steps are part of what is called data analysis and do not
have to be repeated for every problem being solved using the data type defined.
That is, these steps are performed once and the template may be used multiple
times.
The rest of the steps are taken to solve a specific problem. These are the steps
that need to be repeated for each problem solved. These steps specialize the func-
tion template. Step 3 requires an outline of how the function’s value is computed.
Step 4 requires the specialization of constant definitions for the values of sample
expressions. The sample expression illustrates how an answer is computed by pro-
cessing a given variety. The differences are named and serve as parameters to a
function definition. If a variety is not processed to formulate an answer, then no
sample expression is written. Steps 5 and 6 specialize the signature, the purpose
statement, and the function header. Step 7 requires the specialization of tests. There
must be at least one test for each data variety. Tests ought to also be thorough and,
29 Designing Functions to Process Data with Variety 109

Fig. 46 Function templates for data with variety

;; A vdata is either:
;; 1. Variety 1 ;; A vdata is either:
.. ;; 1. Variety 1
;; . ;; 2. Variety 2
;; N. Variety N
;; vdata . . . ... ;; vdata . . . ...
;; Purpose: . . . ;; Purpose: . . .
(define (f-on-vdata a-vdata . . .) (define (f-on-vdata a-vdata . . .)
(cond [<Variety 1 Test> ...] (if <Variety 1 Test>
.. ...
. ...))
[<Variety N-1 Test> ...]
[else ...])) ;; Sample expressions
;; Sample expressions (define VARIETY1-CONST . . .)
(define VARIETY1-CONST . . .) (define VARIETY2-CONST . . .)
..
.
(define VARIETYN-CONST . . .) ;; Tests using sample computations
;; Tests using sample computations (check-expect
(check-expect (f-on-vdata <variety 1> . . .)
(f-on-vdata <variety 1> . . .) VARIETY1-CONST)
VARIETY1-CONST) (check-expect
.. (f-on-vdata <variety 2> . . .)
. VARIETY2-CONST)
(check-expect
(f-on-vdata <variety N> . . .) ;; Tests using sample values
VARIETYN-CONST) (check-expect
;; Tests using sample values (f-on-vdata <variety i> . . .) . . .)
(check-expect ...
(f-on-vdata <variety i> . . .) . . .)
...

(a) Function Template Using cond (b) Function Template Using if

therefore, test boundary values between varieties. Write tests that use the constants
defined for sample computations and tests that use sample values. Step 8 requires the
specialization of the function’s body. Fill in the expressions for the answer of each
variety one at a time. Step 9, as before, requires running the tests and redesigning if
necessary.
The design recipe for decision-making functions informs us that data variety plays
a central role in problem solving. For example, the number of sample expressions,
of stanzas in a conditional, and of tests is proportional to the number of varieties in
the data. That is, the shape of the data influences the shape of a solution/program.
This suggests the function templates displayed in Fig. 46. Figure 46a displays the
function template for data with more than two varieties. Figure 46b displays the
function template for data with two varieties. In both, the number of constants for
sample expressions, of tests, and of needed expressions in the body of the function
are equal to the number of varieties. It is, of course, reasonable to increase the
110 5 Making Decisions

number of constants and tests for more thorough testing or to add clarity to the
program.
To illustrate the steps of the design recipe we will explore several different types
of data that have variety. It is impossible to outline all imaginable types of data with
variety. The goal here is to illustrate how to solve problems with several common
data type patterns that have variety.

* Ex. 57 — Write a program to compute a verse whose length is in [15..35]

from a given verse. If the given verse is too long, create a new verse by dropping
the letters in excess of 35 at the end of the given verse. If the verse is too short,
create a new verse by padding it with blanks at the end to make its length 15.
Otherwise, return the given verse.

* Ex. 58 — Implement a program to determine if a verse is printable. A verse

is printable if it is not too long. Carefully consider if a conditional is needed.
Do you need a conditional to compute a Boolean function?

30 Enumeration Types

A data definition that lists all possible values is called an enumeration type. To a
programmer this means that the conditional in a function that processes this type of
data only needs to distinguish the cases listed in the data definition. This is done by
checking for equality with the values listed in the data definition. Although this data
type represents the simplest form of variety, it may be cumbersome to use as the
number of varieties grows.
To illustrate the design process using an enumerated type consider creating
an animation for of a traffic light. We design our program for a simplified traf-
fic light that only changes colors from green to yellow to red to green and so
on. That is, the possibilities of a flashing red or a flashing yellow are not in-
cluded the design. Based on this there are only three possible traffic light images,
which may be defined as constants. Each image is created by overlaying the three
lights over a background. For example, the three images may be defined as fol-
lows:
(define BACKGROUND (rectangle 60 180 'solid 'black))

(define R-ON (overlay (above (circle 25 'solid 'red)

(square 10 'solid 'transparent)
(circle 25 'outline 'yellow)
(square 10 'solid 'transparent)
(circle 25 'outline 'green))
BACKGROUND))
30 Enumeration Types 111

(define Y-ON (overlay (above (circle 25 'outline 'red)

(square 10 'solid 'transparent)
(circle 25 'solid 'yellow)
(square 10 'solid 'transparent)
(circle 25 'outline 'green))
BACKGROUND))

(define G-ON (overlay (above (circle 25 'outline 'red)

(square 10 'solid 'transparent)
(circle 25 'outline 'yellow)
(square 10 'solid 'transparent)
(circle 25 'solid 'green))
BACKGROUND))
Each constant for a traffic light has a single light on denoted by a solid circle and
two lights off denoted by outline circles. There is a small space between the lights
obtained by drawing a transparent square.
To create an animation, the animate syntax defined in the universe teachpack
is used. When an animate expression is evaluated a clock is started that ticks 28
times per second. The programmer must write a function that takes as input a tick:
the value of the clock (the number of ticks since the simulation started) and that
returns the image to display. To stop a simulation, click Stop in DrRacket or close
the simulation window. The value of an animate expression is the value of the clock
when the animation is stopped.
Let us use a top-down design strategy given that the needed functions are not
known. What is known, however, is that clock ticks and a traffic light (which is
yet to be defined) must be processed. This is the starting point for the top-down
design strategy. To solve this problem, a programmer may decide that a given clock
tick is first transformed to a number that represents a traffic light. Observe that a
traffic light is not defined as an image. The number representing a traffic light is
then used to select the traffic light image. It is necessary to carefully think about
how a clock tick may be transformed to a traffic light. Given that there are only three
images, we may use three numbers. We need a mechanism to map a tick to one of
the three traffic light numbers. If we use 0, 1, and 2 to represent a traffic light, then
modular arithmetic may be used to map a tick to a traffic light. Observe that the
remainder of any tick by 3 is always 0, 1, or 2. Now, all that is left is developing data
definitions that attach meaning to each of these numbers chosen to represent a traffic
light.
The data analysis steps of the design recipe may be satisfied as follows:
STEP 1
We may define a tick and traffic light as follows:
A tick is an integer greater than or equal to 0.

;; A traffic light (tl) is either

;; 1. 0 --means the green light is on
112 5 Making Decisions

;; 2. 1 --means the yellow light is on

;; 3. 2 --means the red light is on

Observe that if the traffic light is increased by 1 (remainder 3) the next light
is the correct one. For example, increasing from 0 to 1 (remainder 3) means
that the traffic light changes from green to yellow. Similarly, incrementing
2 (remainder 3) results in 0 meaning that the traffic changes from red to
green.
STEP 2
The function template for tick is:
;; tick ... → ...
;; Purpose: ...
(define (f-on-tick a-tick) ...)

;; Sample expressions for f-on-tick

(define TICK-V0 ...)
..
.
(define TICK-VN ...)

;; Tests using sample computations for f-on-tick

(check-expect (f-on-tick ...) TICK-V0)
..
.
(check-expect (f-on-tick ...) TICK-V0)

;; Tests using sample values for f-on-tick

(check-expect (f-on-tick ...) ...)

The function template for tl is:

;; tl ... --> ...
;; Purpose: ...
(define (f-on-tl a-tl)
(cond [(= a-tl 0) ...]
[(= a-tl 1) ...]
[else ...]))

;; Sample expressions for f-on-tl

(define TL-0 ...)
(define TL-1 ...)
(define TL-2 ...)

;; Tests using sample computations for f-on-tl

(check-expect (f-on-tl ...) TL-0)
30 Enumeration Types 113

(check-expect (f-on-tl ...) TL-1)

(check-expect (f-on-tl ...) TL-2)

;; Tests using sample values for f-on-tl

(check-expect (f-on-tl ...) ...)

There is no variety in the data definition for tick. Therefore, f-on-tick does not
have a conditional. No decision needs to be made. The number of tests is also
arbitrary as there are no specific values that must be tested.
On the other hand, the data definition for tl has variety. Observe that there
is a conditional in the body of f-on-tl. There is a stanza for each vari-
ety of tl. The tests for each variety are defined here. The template also sug-
gests writing a sample expression and a corresponding test for each variety of
tl.
Now that the steps for data analysis have been completed, we can start solving
the problem of creating functions for the traffic light animation. We know that a
function to process ticks is needed. This function must return an image of a traffic
light. Starting with Step 3, the following the design recipe yields:
STEP 3
Given a tick, two values need to be computed. From a tick a tl value must be
computed. From a tl value an image must be computed. The computation of these
values is deferred to yet unwritten functions. The function tick->tl converts a
tick to a tl. The function image-of-tl converts a tl into a traffic light image.
STEP 4
To specialize the sample expressions, pick three different tick values such that
each maps to a different traffic light image. Specializing tests in this manner
guarantees to illustrate how a clock tick is mapped to each of the possible images.
Here are some sample expressions:
;; Sample expressions for draw-tick
(define TICK12-IMG (image-of-tl (tick->tl 12)))
(define TICK25-IMG (image-of-tl (tick->tl 25)))
(define TICK32-IMG (image-of-tl (tick->tl 32)))

The only difference among the three sample expressions is the value for tick. The
name for this difference is a-tick.
STEP 5
The signature and purpose statement are specialized as:

;; tick → image
;; Purpose: Compute the traffic light image for the
;; given tick
114 5 Making Decisions

STEP 6
The function header is specialized to:
(define (draw-tick a-tick)

STEP 7
The tests are specialized as follows:

;; Tests using sample expressions for ticks to images

(check-expect (draw-tick 12) TICK12-IMG)
(check-expect (draw-tick 25) TICK25-IMG)
(check-expect (draw-tick 32) TICK32-IMG)

;; Tests using sample values for ticks to images

(check-expect (draw-tick 77) R-ON)
(check-expect (draw-tick 99) G-ON)
(check-expect (draw-tick 241) Y-ON)

The sample value tests illustrate that the function works for more than three
predefined computations.
STEP 8
The body of the function is specialized as follows:
(image-of-tl (tick->tl a-tick))

STEP 9
All tests pass.
The result of Step 9 is expected for now. For the design to be complete all tests
must pass after the auxiliary functions are designed and implemented.
Top-down design has revealed that two auxiliary functions, tick->tl and
image-of-tl, are needed. Each is designed independently of the other. It does
not matter which is designed first. Let us start with tick->tl. Here is an example
of how to satisfy the steps of the design recipe:
STEP 3
Given a tick, the corresponding tl is given by the remainder of the tick and 3.
STEP 4
These are specialized sample expressions to illustrate how each variety of tl is
computed from a tick value:

;; Sample expressions for tick->tl

(define TL-0 (remainder 33 3))
(define TL-1 (remainder 77 3))
(define TL-2 (remainder 152 3))
30 Enumeration Types 115

The only difference among the three sample expressions is the value for tick. The
variable for this difference is a-tick.
STEP 5
The specialized signature and purpose statement are:

;; tick → tl
;; Purpose: Convert the given tick to a tl

STEP 6
The specialized function header is:
(define (tick->tl a-tick)

STEP 7
The specialized tests are:

;; Tests for tick->tl using sample expressions

(check-expect (tick->tl 33) TL-0)
(check-expect (tick->tl 77) TL-1)
(check-expect (tick->tl 152) TL-2)

;; Tests for tick->tl using sample values

(check-expect (tick->tl 0) 0)
(check-expect (tick->tl 451) 1)
(check-expect (tick->tl 182) 2)

STEP 8
The body of the function is:
(remainder a-tick 3)

STEP 9
All tests pass.
The design of tick->tl did not reveal the need for any new auxiliary functions.
Therefore, the only remaining task is the design of image-of-tl. For this function,
the presentation of the design is changed. Instead of outlining the result each step of
the design recipe one at a time, the changes made to the template for functions on tl
is directly presented.
This function converts a given tl to an image. We also observe that a tl is only
used to decide which image to return. Since a tl is not used to compute a value there
are no sample expressions or tests using sample expressions to write. Based on these
observations, the initial template specialization looks as follows:
116 5 Making Decisions

;; tl → image
;; Purpose: Return the traffic light image for the given tl
(define (image-of-tl a-tl . . .)
(cond [(= a-tl 0) . . .]
[(= a-tl 1) . . .]
[else ...]))

;; Tests for image-of-tl using sample values

(check-expect (image-of-tl 0) . . .)
(check-expect (image-of-tl 1) . . .)
(check-expect (image-of-tl 2) . . .)
Observe that despite not having tests using sample computations, there is still at least
one test for every variety of tl.
This rest of the specialization is fairly straight-forward. Our design has already
specified what image corresponds to each instance of tl. The final specialization
yields:
;; tl → image
;; Purpose: Return the traffic light image for the given tl
(define (image-of-tl a-tl)
(cond [(= a-tl 0) G-ON]
[(= a-tl 1) Y-ON]
[else R-ON]))

;; Tests using sample values

(check-expect (image-of-tl 0) G-ON)
(check-expect (image-of-tl 1) Y-ON)
(check-expect (image-of-tl 3) R-ON)

The final step to complete the traffic light simulation program is to require the
universe teachpack and to write the animate expression. Figure 47 displays the
structure of the program. Elements in angle brackets have been omitted from the
figure due to space limitations. It should be clear, however, what needs to be filled in
from all the design steps illustrated above. The second line in the program requires
the universe teachpack. The last line in the program is telling DrRacket to animate
the program using the draw-tick function. Congratulations! You have written your
first animation. Run the program. What do you think?

* Ex. 59 — A domesticated mammal is either a dog, a cat, or a horse. Design

and implement a program to compute the number of mammary glands of a
domesticated mammal.
* Ex. 60 — Design and implement a program that squares even digits and that
doubles odd digits.
30 Enumeration Types 117

Fig. 47 Traffic light animation program

(require 2htdp/image)
(require 2htdp/universe)

<Traffic Light Image Constant Definitions>

<Data Definition and function template for tick>
<Data Definition and function template for tl>

;; tick tl
;; Purpose: Convert the given tick to a tl
(define (tick->tl a-tick) (remainder a-tick 3))

<Sample expressions for ticks->tl>

<Tests for tick->tl>

;; tl image
;; Purpose: To return the traffic light image for the given tl
(define (image-of-tl a-tl)
(cond [(= a-tl 0) G-ON]
[(= a-tl 1) Y-ON]
[else R-ON]))

<Sample expressions for image-of-tl>

<Tests for image-of-tl>

;; tick image
;; Purpose: Compute the traffic light image for the given tick
(define (draw-tick a-tick)
(image-of-tl (tick->tl a-tick)))

<Sample expressions for draw-tick>

<Tests for draw-tick>

(animate draw-tick)

* Ex. 61 — A sequential circuit known as a flip-flop can be on (set) or off

(reset). When the flip-flop is on the next state is off. When the flip-flop is off the
next state is on. Design and implement a program to compute the next state of
a flip-flop.

** Ex. 62 — Design and implement an animation for a pedestrian traffic light.

A pedestrian traffic light displays one of two messages: Walk or Don’t Walk.
118 5 Making Decisions

31 Interval Types

Running the traffic light simulation reveals that the light changes too fast–28 times
per second. The program is well-designed and works from a certain perspective but
needs to be refined. That is, it needs to be improved. This is our first dive into the
process of iterative refinement. It is fairly clear that we need the light to change at a
slower pace so that the human eye can appreciate the changes. Instead of changing
the image 28 times per second, the animation may change the traffic light image, for
example, twice per second. This means that, instead of changing every clock tick,
the image needs to change every 14 clock ticks.
This refinement means that the data definition of tl must be updated. Instead of
defining tl as 0, 1, or 2, tl could define as follows:
;; A traffic light (tl) is either
;; 1. 0 --means the green light is on
;; 2. 1 --means the green light is on
..
.
;; 13. 13 --means the green light is on
;; 14. 14 --means the yellow light is on
..
.
;; 27. 27 --means the yellow light is on
;; 28. 41 --means the red light is on
..
.
;; 41. 41 --means the red light is on
This data definition leads to a function definition whose conditional has 42 stan-
zas. That is quite large and error-prone during development. Sometimes such data
definitions are unavoidable. Observe, however, that there is something special about
this (long) data definition. There is a lot of repetition. This suggests that there is an
abstraction that needs to be defined.
Many consecutive numbers have the same meaning. For example, the values from
0 through 13 all mean that the green light is on. We can borrow an idea from high
school mathematics called an interval. We can say, for example, that when tl is a
member of [0..13] it means that the green light is on. Equivalently, we can say that tl
is a member of [0..14), (-1..13], or (-1..14). Remember that a square bracket means
included and a parenthesis means not included. This leads to the following data
definition:
;; A traffic light (tl) is a member of either
;; 1. [0..13] --means the green light is on
;; 2. [14..27] --means the yellow light is on
;; 3. [28..41] --means the red light is on
This is called an interval type. An interval type defines orderable data by a set of
categories. For example, tl is defined using three intervals. The intervals must be
31 Interval Types 119

Fig. 48 Refined tick->tl for the traffic light animation

#|
;; STEP 3
;; A tl is the remainder of a tick by 42.

STEP 5:
;; tick --> tl
;; Purpose: Convert the given tick to a tl |#

;; STEPS 6 and 8
(define (tick->tl a-tick)
(remainder a-tick 42))

;; STEP 4: Sample expressions for tick->tl

(define TL-0 (remainder 11 42))
(define TL-1 (remainder 60 42))
(define TL-2 (remainder 123 42))

;; STEP 7: TESTS
;;Tests using sample expressions
(check-expect (tick->tl 11) TL-0)
(check-expect (tick->tl 60) TL-1)
(check-expect (tick->tl 123) TL-2)

;; STEP 7: Tests using sample values

(check-expect (tick->tl 0) 0)
(check-expect (tick->tl 325) 31)
(check-expect (tick->tl 650) 20)

mutually exclusive. That is, an instance of the data type must be a member of only
one interval.
Refining a data definition means that the program must also be refined. Any
function that manipulates or creates an instance of the refined data type and its
tests must be updated. To determine which functions to update, a programmer must
look for expressions that manipulate or create an instance of the updated data type.
Functions that are good candidates to be updated are those with signatures that refer
to the refined data type. Examining the signatures in the traffic light simulation
program reveals:
tick->tl: tick → tl Needs to be updated, because it creates an instance of tl.
image-of-tl: tl → image Needs to be updated, because it makes a decision based
on an instance of tl.
draw-tick: tick → image Doest not need to be updated. This function does not
manipulate nor creates an instance of tl.
The refinements necessary involve updating expressions and tests. The tick->tl
function must convert a tick into a tl. Finding the remainder of the given tick by 3
is no longer correct because a tick must be converted to an integer in [0..41].
Modular arithmetic, however, can still be used. Now, the remainder of a tick by 42
must be computed. Figure 48 displays the updated tick->tl function. The sample
120 5 Making Decisions

Fig. 49 Refined image-of-tl for the traffic light animation

#|
STEP 3:
Determine the interval and return the corresponding image

STEP 5:
;; tl --> image
;; Purpose: To return the traffic light image for the given tl
|#

;; STEPS 6 and 8
(define (image-of-tl a-tl)
(cond [(<= 0 a-tl 13) G-ON]
[(<= 14 a-tl 27) Y-ON]
[else R-ON]))

;; STEP 4: There are no sample expressions as a tl is not used to

;; compute a new value. It is only used to make a decision.

;; STEP 7: Tests using sample values

(check-expect (image-of-tl 10) G-ON)
(check-expect (image-of-tl 22) Y-ON)
(check-expect (image-of-tl 37) R-ON)

expressions and the body of the function now use 42 instead of 3. The tests have
been updated in order to have at least one test for each variety of tl using the new
data definition.
Figure 49 displays the refined version of image-of-tl. The image-of-tl func-
tion processes an instance of tl. This means that the conditional expression must
be refined to correspond to the new data definition of tl. Observe that the tests in-
side the conditional now determine what interval contains a-tl. The unit tests for
image-of-tl are also refined to include at least one test for each interval of tl.
An important lesson to absorb is that data in the real or an imaginary world may
be represented in different ways (like tl). Part of the design process is to select a
representation. The representation chosen influences how a problem solver thinks
about a program. Exploring different representation may provide insight into the
problem. It is also important to keep in mind that how data is represented may have
a profound impact on performance.

* Ex. 63 — Design and implement a program to compute the interest rate of a

savings account. The interest rate of a savings account depends on its balance.
For a balance larger than $1,000,000 the interest rate is 1.11%. For a balance
larger than $100,000 the interest rate is 1.05%. For a balance larger than $50,000
the interest rate is 1%. For a balance less than $50,000 the interest rate is 0.5%.
32 Itemization Types 121

* Ex. 64 — Design and implement a program to compute a letter grade on an

exam. Less than 60 points is an F. 60-69 points is a D. 70-79 points is a C. 80-89
points is a B. 90-100 points is an A.

* Ex. 65 — For the traffic light simulation three images are defined: G-ON,
Y-ON, and R-ON. Rewrite the traffic light simulation using the following data
definition:
;; A traffic light (tl) is either
;; 1. G-ON --means the green light is on
;; 2. Y-ON --means the yellow light is on
;; 3. R-ON --means the red light is on

** Ex. 66 — A sequential circuit known as a flip-flop can be on (set) or off

(reset). When the flip-flop is on the next state is off. When the flip-flop is off the
next state is on. Design and implement an animation of a flip-flop that changes
state every second.

*** Ex. 67 — Design and implement an animation for a pedestrian traffic light.
A pedestrian traffic light displays one of two messages: Walk or Don’t Walk.
The Walk message should be displayed for 4 s before it changes. The Don’t Walk
message should be displayed for 2 s before it changes.

32 Itemization Types

Enumeration types capture variety by exhaustively listing all possible values. Interval
types capture variety using a set of intervals to classify orderable data. What is
done if both a listing of values and intervals are needed? As it turns out this is
sometimes needed. To define such data an itemization type is used. An instance of
an itemization type may be a specific value or a member of an interval type. Observe
that an itemization type is a generalization of enumeration and interval type. When
all varieties are specific values an itemization type is called an enumeration type.
When all varieties are intervals an itemization type is called an interval type.
To illustrate how to design a program using an itemization type consider the prob-
lem of doubling or rotating an image using single alphanumeric or arrow keystrokes.
The image is rotated 90, 180, or 270◦ using, respectively, the right, down, and left
arrow keys. The image is doubled using the up arrow key. Nothing is done with the
image on an alphanumeric keystroke.
Before proceeding with the design of the solution it is necessary to know how keys
are represented and compared. In the universe teachpack, all keys are represented
with a string. For example, the z key is represented by "z". The right, down, left, and
up keys are represented, respectively, with the strings "right", "down, "left", and
"up". The key=? function is a predicate used to compare keys for equality. Do not
122 5 Making Decisions

Fig. 50 Data definitions for change image and alphanumeric keystrokes

STEP 1: DATA DEFINITIONS
;; An alphanumeric keystroke (aks) is a member of either:
;; 1. ["a".."z"] --means a letter keystroke
;; 2. ["0".."9"] --means a numeric keystroke

;; A change image keystroke (ciks) is either:

;; 1. "right" --means rotate image 90 degrees clockwise
;; 2. "down" --means rotate image 180 degrees clockwise
;; 3. "left" --means rotate image 270 degrees clockwise
;; 4. "up" --means double image size
;; 5. aks --means do nothing to image

mistake key=? as being the same as string=?. The former only compares strings
that represent keys while the latter compares any two strings.
To facilitate the design, assume that there is a constant ROCKET-IMG whose value
is a rocket image created using the functions in the image teachpack. Figure 50
displays Step 1 of the design recipe. An alphanumeric keystroke is an interval type.
There are two intervals: one for letters and one for numbers. Nothing else is an
alphanumeric keystroke. A change image keystroke is an itemization type. It has
four literal values: one for each arrow. It also has one interval type: an alphanumeric
keystroke. Observe that a data definition may refer to another data definition. The
meaning of each variety is also clearly stated for each variety within each data
definition.
Figures 51 and 52 display Step 2 of the design recipe. In Fig. 51, the template for
ciks has a function skeleton with a conditional that distinguishes the variety of the
value received as input. The template also suggests defining a constant and a test for
the value of a sample expression for each variety of ciks. In addition, it suggests the
definition of a sample value test for each variety of ciks. In Fig. 52, the template for
aks has a function skeleton with a conditional that distinguishes between the varieties
of aks. Two sample expressions along with corresponding test are suggested: one
for a letter keystroke and one for a digit keystroke. In the same manner, two sample
value tests are suggested by the template.
Figure 53 displays Steps 3 through 8 of the design recipe for the function that
processes a ciks. Step 3 explains that the rotate and scale functions are used
to compute new images. Step 4 illustrates how these two functions are used to
compute a new image using the ROCKET-IMG constant. Step 5 results in the spe-
cialized signature and purpose statement. The signature states the function takes
two inputs: an image and a ciks. The purpose statement indicates that the given
image is rotated, doubled, or left untouched. The result of step 6 reflects the re-
quirements stated by the signature. The two parameters are named in a manner that
suggest the type of their value. The name of the function is suggestive of its pur-
pose. Step 7 results in tests using sample computations and using sample values.
The tests with sample computations illustrate that the function computes the same
value as the sample expressions. The tests with sample values use images created
32 Itemization Types 123

Fig. 51 Function template for ciks

;; ciks ... --> ...
;; Purpose: ...
(define (f-on-ciks a-ciks ...)
(cond [(key=? a-ciks "right") ...]
[(key=? a-ciks "down") ...]
[(key=? a-ciks "left") ...]
[(key=? a-ciks "up") ...]
[else ...]))

;; Sample expressions for f-on-ciks

(define CIKS-RIGHT ...)
(define CIKS-DOWN ...)
(define CIKS-LEFT ...)
(define CIKS-UP ...))
(define CIKS-ALPHANUM ...))

;; Tests using sample computations for f-on-ciks

(check-expect (f-on-ciks "right" ...) CIKS-RIGHT)
(check-expect (f-on-ciks "down" ...) CIKS-DOWN)
(check-expect (f-on-ciks "left" ...) CIKS-LEFT)
(check-expect (f-on-ciks "up" ...) CIKS-UP)
(check-expect (f-on-ciks <aks> ...) CIKS-ALPHANUM)

;; Tests using sample values for f-on-ciks

(check-expect (f-on-ciks "left" ...) ...)
(check-expect (f-on-ciks "right" ...) ...)
(check-expect (f-on-ciks "down" ...) ...)
(check-expect (f-on-ciks "up" ...) ...)
(check-expect (f-on-ciks <aks> ...) ...)

Fig. 52 Function template for aks

;; aks ... --> ...
;; Purpose: ...
(define (f-on-aks an-aks ...)
(cond [(string<=? "a" an-aks "b") ...]
[(string<=? "0" an-aks "9") ...]))

;; Sample expressions for f-on-aks

(define AKS-LETTER ...)
(define AKS-NUM ...)

;; Tests using sample computations for f-on-aks

(check-expect (f-on-aks <aks1> ...) AKS-LETTER)
(check-expect (f-on-aks <aks2> ...) AKS-NUM)

;; Tests using sample values for f-on-aks

(check-expect (f-on-aks <aks3> ...) ...)
(check-expect (f-on-aks <aks4> ...) ...)

using functions from the image teachpack. The third test uses flip-vertical.
This test passes because vertically flipping an equilateral triangle results in the
124 5 Making Decisions

Fig. 53 The change image program

;; STEP 3: Use the functions rotate and scale to compute a new image
;; STEP 5: image ciks image
;; Purpose: Rotate 90/180/270 degrees, double or untouch image
;; STEP 6 and 8
(define (change-img an-img a-ciks)
(cond [(key=? a-ciks "right") (rotate 90 an-img)]
[(key=? a-ciks "down") (rotate 180 an-img)]
[(key=? a-ciks "left") (rotate 270 an-img)]
[(key=? a-ciks "up") (scale 2 an-img)]
[else an-img]))
;; STEP 4: Sample expressions for change-img
(define RRIGHT-RCKT (rotate 90 ROCKET-IMG))
(define RDOWN-RCKT (rotate 180 ROCKET-IMG))
(define RLEFT-RCKT (rotate 270 ROCKET-IMG))
(define DOUBLE-RCKT (scale 2 ROCKET-IMG))
;; STEP 7: TESTS
;; Tests using sample computations for change-img
(check-expect (change-img ROCKET-IMG "right") RRIGHT-RCKT)
(check-expect (change-img ROCKET-IMG "down") RDOWN-RCKT)
(check-expect (change-img ROCKET-IMG "left") RLEFT-RCKT)
(check-expect (change-img ROCKET-IMG "up") DOUBLE-RCKT)
(check-expect (change-img ROCKET-IMG "m") ROCKET-IMG)
;; Tests using sample values for change-img
(check-expect (change-img (rectangle 10 30 solid red) "left")
(rectangle 30 10 solid red))
(check-expect (change-img (ellipse 75 35 outline pink) "right")
(ellipse 35 75 outline pink))
(check-expect (change-img (triangle 25 solid gold) "down")
(flip-vertical (triangle 25 solid gold)))
(check-expect (change-img (square 15 outline green) "up")
(square 30 outline green))
(check-expect (change-img (star 10 solid yellow) "y")
(star 10 solid yellow))

same image as rotating the triangle 180◦ . In general, however, this is not true.
For example, vertically flipping a right triangle does not result in the same image
as rotating the triangle 180◦ . Finally, Step 8 specializes the body of the function.
The expressions for the consequence in each stanza of the conditional are based
on the sample expressions and on the last test using a sample value. For the for-
mer, the literal ROCKET-IMG is substituted with, an-img, the image parameter of
the function. For the latter, observe that the input image is the same as the output
image. This means that in the default stanza the value of the parameter must be
returned.
The final noteworthy observation is that the design of change-img did not reveal
the need for auxiliary functions. This means that the template for aks was never used
to design a function. That is fine. Not every problem requires every template to be
specialized. Other problems involving ciks may very well require the specialization
33 What Have We Learned in This Chapter? 125

of the aks function template. Think of the templates developed as different tools in a
toolbox. Sometimes you only need a hammer. Other times you only need a wrench.
Yet other times you need both.

** Ex. 68 — Refine the change image program to flip the given image vertically
for a letter keystroke and to flip the given image horizontally for a number
keystroke. Hint: There is no need to change any data definitions for this problem.

** Ex. 69 — Write a program to provide feedback to a student on their quiz

performance. If their average is 100%, then the feedback is "Excellent
Performance". If their average is greater than or equal to 90%, then the
feedback is "Very Good Performance". If their average is greater than or
equal to 80%, then the feedback is "Good Performance". If their average is
greater than or equal to 71%, then the feedback is "Just OK Performance".
If their average is 70%, then the feedback is "Borderline Performance".
If their average is greater than or equal to 60%, then the feedback is "Poor
Performance". If their average is less than or equal to 59%, then the feedback
is "Failing Performance".

33 What Have We Learned in This Chapter?

The important lessons of Chap. 5 are summarized as follows:

• Compound functions arise from data having variety and having to decide which
expression to evaluate depending on the variety of the input.
• Compound functions in BSL are written with a conditional expression in the body.
• The BSL provides two types of conditional expressions: cond- and if-expressions.
• The BSL syntax explored is displayed in Fig. 54.
– Syntax for data types defined by BSL is omitted.
• Data definitions are needed to describe data that has variety.
• The design recipe for functions that make decisions requires:
– A data definition to describe data that has variety
– A function template for every data definition
– At least one test for every variety of data
• Signatures may use types described by a data definition.
• Function templates are specialized to write functions.
126 5 Making Decisions

Fig. 54 Chapter 5 BSL grammar

program ::= {expr test def}*
def ::= (define <variable name> expr)
::= (define (<name> <name>+) expr)
test ::= (check-expect expr expr)
::= (check-within expr expr expr)
expr ::= number
::= string
::= \#<character constant>
::= <character+>
::= #true
::= #false
::= <image>
::= (<function> expr+)
::= (and expr expr expr* )
::= (or expr expr expr* )
::= (cond [expr expr]+
[else expr])
::= (if expr expr expr)

• There are 3 type categories to describe variety:

– Enumeration types
– Interval types
– Itemization types
Chapter 6
Aliens Attack Version 1

We all know that video games are fun to play. Designing them can also be fun despite
being complex programs. To manage this complexity, video games are developed
using the process of iterative refinement. Adding features to a video game in a
piecemeal manner allows the problem solver to conquer the complexity in small
steps. For example, for Aliens Attack we have already solved the problem of creating
cis (remember, ci is defined as a character image in 4).
Before proceeding, it is a good idea to get a handle on exactly what a video
game is. In many ways, a video game is similar to an animation. For instance,
scenes are flickered fast enough to create the illusion of movement. Video games,
however, allow players to interact with the animation. This interaction may change
the evolution of the game. The evolution of the game changes when a computer
event occurs. A computer event includes any action like pressing a key, the clock
ticking, or clicking the mouse. Players change the evolution of the game through, for
example, key pressing or mouse actions. For instance, in Aliens Attack a player may
use the left arrow and the right arrow keys to move the rocket. Like a simulation, the
evolution of a video game may also change by events not controlled by the player
like a clock tick. In Aliens Attack, for example, the aliens move every time the clock
ticks. It is now clear that functions called event handlers (or simply handlers) are
needed to process computer events.
The ability to design programs that make decisions allows us to implement video
games. Handlers need to make decisions. For example, the handler to process a
keystroke must distinguish what key has been pressed. If the right arrow key is
pressed, then the handler moves the rocket right. If the left arrow key is pressed, then
the handler moves the rocket left. Clearly, data definitions are needed to describe the
keystroke events that change the evolution of the game.
In addition to providing programmers with the ability to write animations, the
universe teachpack provides programmers with the ability to write video games in
BSL. This teachpack is an API. It provides the necessary syntax to associate handlers
with computer events and it defines the signature that these handlers must have.
Using the universe teachpack is an exercise in programming with (the abilities

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 127
M. T. Morazán, Animated Problem Solving, Texts in Computer Science,
https://doi.org/10.1007/978-3-030-85091-3_6
128 6 Aliens Attack Version 1

provided by and the restrictions imposed by) an API. Programming with an API is
a good skill to develop because programming with APIs is standard practice.

34 The Universe Teachpack

The universe teachpack requires the development of a data definition for a world.
A world is all the values that may change during the evolution of the game. Some of
the values may be de displayed when a game image is computed. Other values may
only be needed to advance the evolution of the game. In a fully implemented Aliens
Attack game, for example, the aliens and the direction the aliens travel in are part of
the world. The aliens are displayed in a game image as in Fig. 37. The direction the
aliens travel in, on the other hand, is not displayed in the image.
There is a special expression to run a video game called a big-bang expression
(bb-expr). This expression allows the programmer to provide the initial value of the
world and the handlers needed to advance the evolution of the game. The big-bang
expression syntax is:
bb-expr ::= (big-bang expr bb-clause+)
bb-clause ::= [to-draw function]
[on-key function]
[on-tick function]
[stop-when function]
[name expr]
..
.
A bb-expr has in parenthesis the keyword big-bang followed by an expression
that must evaluate to a world value followed by one or more bb-clauses (big-
bang clauses). A bb-clause associates an event with a handler function to pro-
cess the event. Each bb-clause has in square brackets an event describer (e.g.,
to-draw, on-tick, and stop-when) and a handler for the event. For example,
[to-draw draw-world] states to use draw-world to render the image of the
world, [on-key process-key] states to use process-key to process a key stroke,
[on-tick process-tick] states to use process-tick to process a clock tick,
and [stop-when game-over?] states to use game-over? to detect the end of
the game (or interactive simulation). The only required bb-clause is the to-draw
clause. The only clause that does not follow this scheme is the name clause. This
clause associates a name with the world. The expr in this clause must evaluate to a
string or symbol. The name appears in title of the scene. The vertical dots mean that
there are other varieties of bb-clauses. The big-bang documentation describes all
the possible bb-clauses. An important thing to remember is that a bb-clause may
not be repeated. For example, a bb-expr may only have one to-draw bb-clause.
The universe API also specifies the signature and purpose of the handlers. Each
handler serves a specific purpose like creating a scene, computing the next world
after a clock tick or keystroke, or detecting the end of the game/simulation. For
34 The Universe Teachpack 129

Fig. 55 Fish image

example, the following table displays the signature and purpose for four handlers:
on-draw: world → scene
Purpose: To create the world’s scene
on-key: world key → world
Purpose: To return the world after the given keystroke
on-tick: world → world
Purpose: To return the world after a clock tick
stop-when: world → Boolean
Purpose: To determine if the game/animation has ended
This means that we must write handlers that satisfy the above signatures. To a
programmer this suggest using a top-down design strategy to write a video game. Start
with the functions needed by the big-bang expression. Design auxiliary functions
as their need is revealed.
To illustrate the use of the universe teachpack consider developing an interactive
animation to rotate the image of the fish displayed in Fig. 55. The user may rotate
the image of the fish to point up, down, left, or right using the arrow keys. In Fig. 55
the fish is pointing up. If the user presses the up arrow key, then the image remains
unchanged. If the user presses the right arrow, then the image is rotated to make the
fish point right. A similar action takes place pressing the other arrow keys. If the
user presses any other key, then the fish image remains unchanged regardless of the
direction the fish is pointing in. For example, if the fish is pointing left and the m key
is pressed, then the fish image remains unchanged.
To start, first define the constants that are needed. A scene is needed. A scene is
defined as follows:
A scene is a WIDTH x HEIGHT image
Here WIDTH and HEIGHT are positive integer constants. Presumably, a scene is large
enough to contain the fish image. Otherwise, the fish image is truncated.
The fish image is also a constant. The following code defines all the scene and
fish image constants:
(define WIDTH 200)
(define HEIGHT 200)
(define E-SCENE-COLOR 'black)
(define E-SCENE (empty-scene WIDTH HEIGHT E-SCENE-COLOR))
130 6 Aliens Attack Version 1

(define FISH-W 46)

(define FISH-H 76)
(define FISH-BODY-IMG
(place-image
(ellipse 30 60 'solid 'green)
(/ FISH-W 2)
(/ FISH-H 2)
(rectangle FISH-W FISH-H 'solid E-SCENE-COLOR)))

(define FISH-EYE-IMG (overlay (circle 2 'solid 'lightred)

(circle 4 'solid 'white)))

(define FISH-TAIL-IMG (overlay/xy

(circle 20 'solid 'black)
0
-10
(circle 20 'solid 'green)))

(define FISH-IMG (place-image FISH-EYE-IMG

(- (/ FISH-W 2) 3)
(/ FISH-H 4)
(place-image
FISH-TAIL-IMG
(/ FISH-W 2)
(+ (/ FISH-H 4/3) 15)
FISH-BODY-IMG)))
Observe that the fish image is created using a divide and conquer strategy. First, the
body, eye, and tail images are defined. Second, these three images are combined to
create the fish image.
After defining the first set of constants (others may be needed as the solution
progresses), the design process starts with writing a run function that has a bb-expr
as its body. By convention in this textbook, the run function is used to start a game
or interactive animation and its bb-expr has a name clause. The central question to
start is what computer events need to be processed. For this interactive animation
only keystrokes need to be processed. Therefore, the run function may be written as
follows:
;; string → world
;; Purpose: To run the interactive fish simulation
(define (run a-name)
(big-bang
INIT-WORLD
[on-draw draw-world] ;; world → scene
[on-key process-key] ;; world key → world
[name a-name]))
34 The Universe Teachpack 131

The body of run gives us a starting point for the design process. The initial world,
INIT-WORLD, needs to be defined. For this, a world data definition is needed. The
function handlers draw-world and process-key and any needed auxiliary func-
tions need to be designed. Recall that the signatures of the handlers are defined by
the universe API (noted as comments in run). For the key-processing handler, a
data definition is needed for the keys that affect the evolution of the animation.
We start with the data definition for the keys that affect the evolution of the anima-
tion. Remember that every data definition means the development of a corresponding
function template. A fish keystroke is defined as follows:
A fish keystroke (fks) is either:
1. "up" --means fish in up direction
2. "right" --means fish in right direction
3. "down" --means fish in down direction
4. "left" --means fish in left direction

;; fks . . . → . . .
;; Purpose: . . .
(define (f-on-fks a-fks . . .)
(cond [(string=? a-key "up") . . .]
[(string=? a-key "right") . . .]
[(string=? a-key "down") . . .]
[else . . .]))

;; Sample Expressions for f-on-fks

(define UP-FKS-VAL . . .)
(define RIGHT-FKS-VAL . . .)
(define DOWN-FKS-VAL . . .)
(define LEFT-FKS-VAL . . .)

;; Tests using sample computations for f-on-fks

(check-expect (f-on-fks "up" . . .) UP-FKS-VAL)
(check-expect (f-on-fks "right" . . .) RIGHT-FKS-VAL)
(check-expect (f-on-fks "down" . . .) DOWN-FKS-VAL)
(check-expect (f-on-fks "left" . . .) LEFT-FKS-VAL)

;; Tests using sample values for f-on-fks

(check-expect (f-on-fish-ks . . .) . . .)
...
Observe that an fks is an enumeration type. Further observe that four sample ex-
pressions and four tests using sample computations are needed as required when
designing with data that has variety: one for each variety.
The above data definition now allows for a refinement of the key data definition.
Now, keys may be distinguished between keys that are an fks and those that are not.
The refined data definition is:
132 6 Aliens Attack Version 1

A key is either:
1. fks
2. not an fks

;; key . . . → . . .
;; Purpose: . . .
(define (f-on-key a-key . . .)
(if (fks? a-key)
...
. . .))

;; Sample Expressions for f-on-key

(define FKS-VAL . . .)
(define NON-FKS-VAL . . .)

;; Tests using sample computations for f-on-key

(check-expect (f-on-key . . .) FKS-VAL)
(check-expect (f-on-key . . .) NON-FKS-VAL)
...

;; Tests using sample values for f-on-key

(check-expect (f-on-key . . .) . . .)
...
Observe that key is an itemization type. There is something different, however, in the
formulation of this data definition. The second variety not an fks does not look
like an enumeration nor does it look like an interval. It is, however, an enumeration
type. Specifically, it is the complement of fks. That is, any key that is not an fks is
of the second variety. The use of not is simply shorthand for the large number of
non-fks keys. Given that keys are itemized in two varieties, the body of f-on-key is
an if-expression. This if-expression tests if a given key is an fks using an auxiliary
function that needs to be designed. Finally, observe that the number of sample
expressions and of tests using sample computations matches the number of varieties.
The final data definition we need is for the world. There are many ways to define
the world for this animation. Perhaps, the simplest way to define the world is to make
it an fks as follows:
A world is an fks

;; world . . . → . . .
;; Purpose: . . .
(define (f-on-world a-world)
. . .(f-on-fks a-world). . .)

;; Sample Expressions for f-on-world

(define WORLD-VAL . . .))
34 The Universe Teachpack 133

;; Tests using sample computations for f-on-world

(check-expect (f-on-world . . .) WORLD-VAL)
...

;; Tests using sample values for f-on-world

(check-expect (f-on-world . . .) . . .)
...
Observe that there is no variety in the world data definition. Therefore, the body
of f-on-world is not a conditional expression. This brings to the forefront an
important observation. The world and fks are different. They are different data
definitions. Therefore, at least two functions are needed to process a world: one to
process a world and one to process an fks. This is why there is a call to f-on-fks in
the body of f-on-world. You may ask yourself why two are needed when a world
is the same as fks. Indeed, this is true for this version of the simulation. However,
always design thinking that in the future you or someone else may need to refine
the program. For example, the data definition of the world may be refined. If your
solution already includes separate world-processing and fks-processing functions,
then the process of refinement is easier. A change in the world data definition does
not affect the functions that process an fks and only functions that process a world
need to be refined. On the other hand, if a single data definition is used, then a change
in the world data definition requires refinements to all functions that process a world
and functions for an fks need to be designed.
With the data definitions and function templates completed, the design of the
elements needed by the big-bang expression is next. The first is INIT-WORLD. Ask
yourself what value the initial world needs to be or you want it to be. In this interactive
animation, the initial direction the fish is pointing in is arbitrary. Therefore, the initial
world value may be defined as follows:
(define INIT-WORLD "up")
Next, focus turns to designing draw-world. This function processes a world.
Therefore, we specialize the template for functions on the world. Ask yourself what
this function needs to do. According the universe API, this function needs to
compute a scene. This means that the fish image must be placed in E-SCENE. This is
enough information to start specializing the template as follows:
;; world → scene
;; Purpose: To draw the world in the E-SCENE
(define (draw-world a-world)
. . .(f-on-fks a-world). . .)

;; Sample Expressions for draw-world

(define WORLD-VAL1 (place-image <an up fish image>
(/ WIDTH 2)
(/ HEIGHT 2)
E-SCENE))
134 6 Aliens Attack Version 1

(define WORLD-VAL2 (place-image <a left fish image>

(/ WIDTH 2)
(/ HEIGHT 2)
E-SCENE))

;; Tests using sample computations for draw-world

(check-expect (draw-world . . .) WORLD-VAL)
(check-expect (draw-world . . .) WORLD-VAL2)
...

;; Tests using sample values for draw-world

(check-expect (draw-world . . .) . . .)
...
The template is first specialized by substituting f-on-world with draw-world and
by creating sample expression to place a fish image in E-SCENE. Placing the fish
image in the middle of the empty scene is a reasonable choice. The fish image to
display depends on the fks value the world represents. This is a different type of
value and an auxiliary function is used to compute it. The template can now be
specialized to:
;; world → scene
;; Purpose: To draw the world in the E-SCENE
(define (draw-world a-world)
(place-image (compute-fish-img a-world)
(/ WIDTH 2)
(/ HEIGHT 2)
E-SCENE))

;; Sample Expressions for draw-world

(define WORLD-VAL1 (place-image
(compute-fish-img INIT-WORLD)
(/ WIDTH 2)
(/ HEIGHT 2)
E-SCENE))

(define WORLD-VAL2 (place-image

(compute-fish-img "left")
(/ WIDTH 2)
(/ HEIGHT 2)
E-SCENE))

;; Tests using sample computations for draw-world

(check-expect (draw-world "up") WORLD-VAL1)
(check-expect (draw-world "left") WORLD-VAL2)
34 The Universe Teachpack 135

;; Tests using sample values for draw-world

(check-expect (draw-world "down") )

(check-expect (draw-world "right") )

Observe that the only difference among the sample expressions is the value of the
world. This tells us that draw-world does not need any additional parameters. This
allows the sample expressions and the sample tests using sample computations to
be written. The final specialization step is for the tests using sample values. It is
reasonable to use the untested world values: "up" and "down". Now, all the steps
of the design recipe are satisfied except running the tests.
Focus now turns to the design of compute-fish-img. This function processes
an fks and, therefore, the template to process an fks value is specialized:
;; fks . . . → image
;; Purpose: Compute the fish image for the given fks
(define (compute-fish-img a-fks . . .)
(cond [(string=? a-key "up") . . .]
[(string=? a-key "right") . . .]
[(string=? a-key "down") . . .]
[else . . .]))

;; Sample Expressions for compute-fish-img

(define UP-FKS-VAL FISH-IMG)
(define RIGHT-FKS-VAL (rotate 270 FISH-IMG))
(define DOWN-FKS-VAL (rotate 180 FISH-IMG))
(define LEFT-FKS-VAL (flip-vertical (rotate 90 FISH-IMG)))

;; Tests using sample computations for compute-fish-img

(check-expect (compute-fish-img "up" . . .) UP-FKS-VAL)
(check-expect (compute-fish-img "right" . . .) RIGHT-FKS-VAL)
(check-expect (compute-fish-img "down" . . .) DOWN-FKS-VAL)
(check-expect (compute-fish-img "left" . . .) LEFT-FKS-VAL)

;; Tests using sample values for compute-fish-img

(check-expect (compute-fish-img . . .) . . .)
...
This function returns an image and this is reflected in the signature and the purpose
statement. The sample expressions compute a value for each variety. Observe that
each computes a value for a fish image using only constant values and that the
conditional only needs the value of the given fks. Thus, compute-fish-img does not
require more parameters. Finally, observe that all fks values have a sample expression
136 6 Aliens Attack Version 1

and that all possible returned fish images are produced by these. Therefore, there
is no need for tests using sample values. Based on these observations, the final
specialization step yields:
;; fks → image
;; Purpose: Compute the fish image for the given fks
(define (compute-fish-img a-fks)
(cond [(string=? a-key "up") FISH-IMG]
[(string=? a-key "right") (rotate 270 FISH-IMG)]
[(string=? a-key "down") (rotate 180 FISH-IMG)]
[else (flip-vertical (rotate 90 FISH-IMG))]))

;; Sample Expressions for f-on-fks

(define UP-FKS-VAL FISH-IMG)
(define RIGHT-FKS-VAL (rotate 270 FISH-IMG))
(define DOWN-FKS-VAL (rotate 180 FISH-IMG))
(define LEFT-FKS-VAL (flip-vertical (rotate 90 FISH-IMG)))

;; Tests using sample computations for compute-fish-img

(check-expect (compute-fish-img "up") UP-FKS-VAL)
(check-expect (compute-fish-img "right") RIGHT-FKS-VAL)
(check-expect (compute-fish-img "down") DOWN-FKS-VAL)
(check-expect (compute-fish-img "left") LEFT-FKS-VAL)

Once again, observe that template specialization completes all the steps of the
design recipe except running the tests.
The function process-key, as its name suggests, processes a key. This means
the template for functions on a key must be specialized. This function always takes
as input a world and a key and returns a world according to the universe API. If
the given key is an fks, then the next world is the given key. Otherwise, the world is
unchanged. The initial specialization yields:
;; world key → world
;; Purpose: To return the next world based on the
;; given key
(define (process-key a-world a-key . . .)
(if (fks? a-key)
...
. . .))

;; Sample Expressions for process-key

(define KEY-VAL "down")
(define KEY-VAL2 "left")
(define NON-KEY-VAL "up")

;; Tests using sample computations for process-key

(check-expect (process-key . . .) KEY-VAL)
(check-expect (process-key . . .) KEY-VAL2)
(check-expect (process-key . . .) NON-KEY-VAL)
...
34 The Universe Teachpack 137

;; Tests using sample values for process-key

(check-expect (process-key . . .) . . .)
...
The sample expressions by themselves have an obscured meaning in this case. The
expression for KEY-VAL is for inputs "right" and "down" to process-key. The
expression for KEY-VAL2 is for inputs "up" and "left" to process-key. The
expression for NON-KEY-VAL is for inputs "up" and "j" to process-key. From
these sample expressions we can surmise that if the given key is an fks, then the new
world is the given key. Otherwise, the next world is the value of the given world. The
final specialization is:
;; world key → world
;; Purpose: To return the next world based on the
;; given key
(define (process-key world a-key)
(if (fks? a-key)
a-key
a-world))

;; Sample Expressions for process-key

(define KEY-VAL "down")
(define KEY-VAL2 "left")
(define NON-KEY-VAL "up")

;; Tests using sample computations for process-key

(check-expect (process-key "right" "down") KEY-VAL)
(check-expect (process-key "up" "left") KEY-VAL2)
(check-expect (process-key "up" "j") NON-KEY-VAL)

;; Tests using sample values for process-key

(check-expect (process-key "left" "up") "up")
(check-expect (process-key "left" "right") "right")
(check-expect (process-key "left" "left") "left")
(check-expect (process-key "left" "v") "left")
(check-expect (process-key "right" "x") "right")
The tests using sample values aim to be thorough. The first two test the fks values
not present in the tests using sample computations. The third test shows that if the
given fks is the same as the given world, then the world is unchanged. The last two
tests use two different non-fks values.
The final task is to design fks? Ask yourself how to determine if a key is an fks.
If the key is any fks value, then the answer is true. Otherwise, the answer is false.
Following the steps of the design recipe yields:
138 6 Aliens Attack Version 1

;; key → Boolean
;; Purpose: Determine if the given key is an fks
(define (fks? a-key)
(if (or (key=? a-key "right")
(key=? a-key "down")
(key=? a-key "left")
(key=? a-key "up"))
#true
#false))

;; Sample Expressions for fks?

(define FKS-RVAL #true)
(define NON-FKS-VAL #false)

;; Tests using sample computations for fks?

(check-expect (fks? "right") FKS-RVAL)
(check-expect (fks? "m") NON-FKS-VAL)

;; Tests using sample values for fks?

(check-expect (fks? "up") #true)
(check-expect (fks? "down") #true)
(check-expect (fks? "left") #true)
(check-expect (fks? "q") #false)
(check-expect (fks? "t") #false)
The predicate has the structure suggested by the template. It has an if-expression
in its body. The tests using sample values add thoroughness by making sure all fks
values are tested and more non-fks values are tested. However, fks? is interesting. It
always returns the value of the test. This means that the if-expression is not needed
and fks? may be simplified to:
;; key → Boolean
;; Purpose: Determine if the given key is an fks
(define (fks? a-key)
(or (key=? a-key "right")
(key=? a-key "down")
(key=? a-key "left")
(key=? a-key "up")))
Certainly, this version of fks? is easier to understand and, therefore, a preferable
solution. The important lesson here is that when designing a predicate for data with
variety consider using a Boolean expression instead of a conditional. Remember,
the template is suggestive of the structure of functions. Sometimes the result of
specializing the template can be simplified.
The only step of the design recipe left is to run the tests. Make sure all the tests
pass.
35 A Video Game Design Recipe 139

* Ex. 70 — Why is flip-vertical used in compute-fish-img?

** Ex. 71 — Redesign the fish interactive simulation using the following data
definition for the world:
A world is an image, either:

1.

2.

3.

4.

* Ex. 72 — Consider the following data definition:

A signed integer is either in:
1. (-∞..-1]
2. [1..∞)

Write a predicate to determine if an integer is a signed integer. Simplify the

predicate as much as possible.

*** Ex. 73 — Write an interactive animation that echoes up to 80 lowercase

letter, space, or period keystrokes in a box. The simulation should stop with an
Enter keystroke.

35 A Video Game Design Recipe

This development suggests the design recipe for video game and interactive anima-
tions displayed in Fig. 56. Steps 1–3 are problem and data analysis that are performed
once. Step 1 requires a data definition for the elements that vary in a game. Step 2
140 6 Aliens Attack Version 1

Fig. 56 The design recipe for video game development

1. Create data definitions for the elements that vary.
2. Develop a function template and examples for each data definition.
3. Define the run function with no tests.

For a needed function:

4. Outline the computation.
5. Define constants for the value of sample expressions for each variety and name the differences.
6. Write the function’s signature and purpose.
7. Write the function’s header.
8. Write tests.
9. Write the function’s body.
10. Run the tests and, if necessary, redesign.

requires the development of a function template and of examples for each data def-
inition. If the defined data has variety, then at least one example for each variety is
needed. Step 3 requires the development of the run function. The body must be a
bb-expr. This function serves as a map to guide the top-down development of the
game. Ask yourself what events may affect the evolution of the game. There must be
a clause in the bb-expr for every event that may change the evolution of the game or
interactive simulation. This function is written without tests because it is only used
to initiate the game or simulation. It is not the solution to a problem. In fact, the run
function’s template may be defined as follows:
; string → world
; Purpose: To run the game
(define (run a-name)
(big-bang INIT-WORLD
[on-draw . . .]
[name a-name]
bb-clause∗))
The final clauses, of course, may not contain a repetition of any clause.
Steps 4–10 are performed for each function that is needed. This is done by
specializing the function template for the data being processed. These steps are the
same as those found in previous design recipes.

36 Adding the Rocket to Aliens Attack

The first refinement to Aliens Attack is to add the rocket. Figure 57 displays a scene
of the proposed new version of the video game. Before starting save your Aliens
Attack version 0 to a new file, say Aliens-Attack-v1. It is this new file that is to
be edited with refinements. The version 0 is kept as a safe back-up of work that has
been completed.
36 Adding the Rocket to Aliens Attack 141

Fig. 57 Aliens attack scene with a rocket

Recall that the rocket may move left to right without going off the edges of the
scene. Given that a rocket may exhibit changes as the game evolves, it needs to be
part of the world and, therefore, a data definition for a rocket is needed. Ask yourself
what changes about the rocket as the game advances. The characteristics that change
must be captured in the rocket’s data definition. At this point, it is important to
realize that a rocket is not the same as the image of the rocket. The image of the
rocket does not change as the rocket moves. When the rocket moves left or right the
rocket’s image-x changes. For example, if the image coordinates of the rocket are (5,
ROCKET-Y) and the rocket moves right, then the new image coordinates of the rocket
are (6, ROCKET-Y). That is, the rocket moves to the next image box to the right in
the scene. Observe, as noted in Chap. 6, that the rocket’s image-y is constant. Now
that the rocket’s changing characteristics are identified, the rocket’s data definition,
the template for functions on a rocket, and sample instances (rocket examples) may
be stated as follows:
#| A rocket is an image-x

;; Sample instances of rocket

(define ROCKET1 ...)
...
142 6 Aliens Attack Version 1

;; rocket . . . → . . .
;; Purpose: . . .
(define (f-on-rocket a-rocket . . .)
. . .(f-on-image-x a-rocket . . .) . . .)

;; Sample expressions for f-on-rocket

(define ROCKET1-VAL (f-on-rocket ROCKET1 . . .))
...

;; Tests using sample computations for f-on-rocket

(check-expect (f-on-rocket ROCKET1 . . .) ROCKET-VAL)
...

;; Tests using sample values for f-on-rocket

(check-expect (f-on-rocket . . .) . . .)
... |#

;; Sample instances of rocket

(define ROCKET1 7)
(define INIT-ROCKET (/ MAX-CHARS-HORIZONTAL 2))
Observe that to process a rocket the template suggests calling a function to process
an image-x. This design exhibits separation of concerns. Separation of concerns
reflects that solving a problem for one data type is different from solving a problem
for another data type. In this example, processing a rocket may require processing
an image-x. The processing of an image-x is done by a function to process an
image-x, not a function to process a rocket. Observe that the tests using sample
computations refer to the sample instances of a rocket and the values of sample
expressions. The tests using sample values do not refer to defined sample instances
or defined values for sample expressions. These tests are used to illustrate that a
function works for values not previously computed. Finally, when a data definition is
developed it is good practice to include at least one sample instance. In this example,
two sample instances of a rocket, ROCKET1 and INIT-ROCKET, are defined.
A similar piece of reasoning must be done to define the game’s world. Ask
yourself what may change as the game evolves. In this version of Aliens Attack, the
only characteristic that may change is the rocket. Therefore, the data definition of
the world, its function template, and sample instances may be stated as follows:
#| A world is a rocket

;; Sample instances of world

(define WORLD1 ...)

;; world . . . --> . . .
;; Purpose: . . .
(define (f-on-world a-world . . .)
. . .(f-on-rocket a-world . . .) . . .)
36 Adding the Rocket to Aliens Attack 143

;; Sample expressions for f-on-world

(define WORLD1-VAL . . .)

;; Tests using sample computations for f-on-world

(check-expect (f-on-world WORLD1 . . .) WORLD1-VAL)
...

;; Tests using sample values for f-on-world

(check-expect (f-on-world . . .) . . .)
... |#

;; Sample instances of world

(define WORLD1 ROCKET1)
(define INIT-WORLD INIT-ROCKET)
Observe that despite the world and the rocket being the same in this version of the
game, we use two different data definitions. A good designer always has a different
data definition for every data type. This makes it easier to make future refinements
to the game. We know that the data definition of the world will be expanded to
include, for example, the army of aliens. This expansion will be done without having
to revisit any code developed to manipulate a rocket. If we use a single definition,
then such an expansion would require writing new code for the world, the army of
aliens, and the rocket. Further observe that the template suggests that to process a
world the processing a rocket may be needed. Again, we see separation of concerns
in this design. Observe that, once again, the tests using sample computations refer to
previously defined constants. In contrast, once again, the tests using sample values
do not refer to previously defined constants. The sample worlds are defined using
the sample rockets.
Given that the player moves the rocket using keyboard it is necessary to define
what a key is. In this version of Aliens Attack the rocket is moved using the left and
right arrow keys. This suggests that a key may be defined as an itemization type. This
data definition highlights the keys that change the evolution of the game. The data
definition, function template, and sample instances for keys are defined as follows:
|# A key is either:
1. "right"
2. "left"
3. not "right" or "left"

;; Sample instances of key

(define KEY1 ...)
(define KEY2 ...)
(define KEY3 ...)
144 6 Aliens Attack Version 1

;; key . . . --> . . .
;; Purpose: . . .
(define (f-on-key a-key . . .)
(cond [(key=? a-key "right") . . .]
[(key=? a-key "left") . . .]
[else . . .]))

;; Sample expressions for f-on-key

(define KEY1-VAL . . .KEY1. . .)
(define KEY2-VAL . . .KEY2. . .)
(define KEY3-VAL . . .KEY3. . .)

;; Tests using sample computations for f-on-key

(check-expect (f-on-key KEY1 . . .) KEY1-VAL)
(check-expect (f-on-key KEY2 . . .) KEY2-VAL)
(check-expect (f-on-key KEY3 . . .) KEY3-VAL)
...

;; Tests using sample values for f-on-key

(check-expect (f-on-key . . .) . . .)
... |#

;; Sample instances of key

(define KEY1 "right")
(define KEY2 "left")
(define KEY3 "m")
Shorthand notation for the third key variety is used. The template suggests three
sample instances because there are three key varieties. In the body of f-on-key
there is a cond-expression to distinguish among the key varieties. As expected, the
tests using sample computations use the key sample instances and the values of the
sample expressions. The tests using sample values, on the other hand, do not used
defined values as expected.

* Ex. 74 — Develop function templates for ci, image-x, image-y, pixel-x, pixel-
y, and scene.

After completing steps 1 and 2 of the design recipe, we focus on the run template
specialization to complete step 3. The player moves the rocket using the left and
right arrows keys. This means that an on-key clause is also needed to process key
events. The specialization is as follows:
36 Adding the Rocket to Aliens Attack 145

; string → world
; Purpose: To run the game
(define (run a-name)
(big-bang INIT-WORLD
[on-draw draw-world]
[name a-name]
[on-key process-key]))
This states that the handler to render the game’s scene is draw-world and that the
handler to process key events is process-key.
Let us start by designing draw-world. To draw the world a rocket needs to be
drawn in the empty scene. The steps of the design recipe result in the following
specialization of the template for functions on a world:
;; world → scene
;; Purpose: To draw the world in E-SCENE
(define (draw-world a-world)
(draw-rocket a-world E-SCENE))

;; Sample expressions for draw-world

(define WORLD-SCN1 (draw-rocket INIT-WORLD E-SCENE))
(define WORLD-SCN2 (draw-rocket INIT-WORDLD2 E-SCENE))

;; Tests using sample computations for draw-world

(check-expect (draw-world INIT-WORLD) WORLD-SCN1)
(check-expect (draw-world INIT-WORLD2) WORLD-SCN2)

;; Tests using sample computations for draw-world

(check-expect (draw-world 0) )
(check-expect (draw-world (sub1 MAX-CHARS-HORIZONTAL))

)
146 6 Aliens Attack Version 1

The sample expressions maintain the separation of concerns by calling a rocket-

processing function to create the world’s scene. The only difference in the sample
expressions is the value of a world. Therefore, draw-world indeed only needs one
parameter. The requirement imposed by the universe API is met. The body of
draw-world uses the variable for the single difference instead of using a literal
value like the sample expressions. Observe that the tests using sample computations
use extrema values for the rocket data type. The design of draw-world awaits the
running of the tests.
The design of the auxiliary function draw-rocket is done by specializing the
template for functions on a rocket. This function draws the rocket’s ci in the given
scene using the given rocket as the image-x. Recall that we have already developed
the function draw-ci (see Chap. 6). The specialization of the template for functions
on a rocket yields:
;; rocket scene → scene
;; Purpose: To draw the rocket in the given scene
(define (draw-rocket a-rocket a-scene)
(draw-ci ROCKET-IMG a-rocket ROCKET-Y a-scene))

;; Sample expressions for draw-rocket

(define RSCN3 (draw-ci ROCKET-IMG 3 ROCKET-Y E-SCENE))

(define RSCN12 (draw-ci ROCKET-IMG 12 ROCKET-Y E-SCENE2))

;; Tests using sample computations for draw-rocket

(check-expect (draw-rocket 3 E-SCENE) RSCN3)
(check-expect (draw-rocket 12 E-SCENE2) RSCN12)

;; Tests using sample values for draw-rocket

(check-expect (draw-rocket 0 E-SCENE)

)
(check-expect (draw-rocket (sub1 MAX-CHARS-HORIZONTAL)
E-SCENE)

)
36 Adding the Rocket to Aliens Attack 147

The sample expressions use draw-ci to place the rocket’s ci at image coordinates
(3, ROCKET-Y) and (12, ROCKET-Y) in, respectively, E-SCENE and E-SCENE2.
The two differences among the sample expressions are abstracted and become the
parameters to draw-rocket. The tests using sample values, once again, test rocket
extrema values. There are no new auxiliary functions to design. Therefore, this
concludes the design of draw-world and its auxiliary functions.
The design of process-key is done by specializing the f-on-key template. The
universe API requires that the inputs be a world and a key. If the given key is
"right" or "left", then the next world is created by moving the rocket. Otherwise,
the next world is the given world. The specialization of the template yields:
;; world key → world
;; Purpose: Process a key event to return next world
(define (process-key a-world a-key)
(cond [(key=? a-key "right") (move-rckt-right a-world)]
[(key=? a-key "left") (move-rckt-left a-world)]
[else a-world]))

;; Sample expressions for process-key

(define KEY-RVAL (move-rckt-right INIT-WORLD))
(define KEY-LVAL (move-rckt-left INIT-WORLD))
(define KEY-OVAL INIT-WORLD2)

;; Tests using sample computations for process-key

(check-expect (process-key INIT-WORLD "right") KEY-RVAL)
(check-expect (process-key INIT-WORLD "left") KEY-LVAL)
(check-expect (process-key INIT-WORLD2 "m") KEY-OVAL)

;; Tests using sample values for process-key

(check-expect (process-key (sub1 MAX-CHARS-HORIZONTAL) "right")
(sub1 MAX-CHARS-HORIZONTAL))
(check-expect (process-key 0 "left") 0)
(check-expect (process-key 0 "o") 0)
(check-expect (process-key 11 ";") 11)
Observe that the need for two auxiliary functions is revealed: one to move the rocket
right and one to move the rocket left. Moving a rocket is done in a separate function
because it is a different data type. The sample expressions use the appropriate rocket-
moving function to compute the next world. The third sample expression captures
the idea that a rocket function is not called when the given key is not "right" or
"left" (like "m" in the third test using a sample computation). The first two tests
using sample values are written with extrema values to illustrate that the rocket does
not go off the scene. The last to tests are added for thoroughness.
The design of move-rckt-right requires clearly defining how to move a rocket
right. At a first glance, it may appear that adding 1 to the rocket suffices. Ask yourself,
however, can the rocket always be moved to the right. After designing process-key,
148 6 Aliens Attack Version 1

it is not difficult to see that when the rocket is MAX-CHARS-HORIZONTAL - 1 it

cannot move to the right. Adding 1 to MAX-CHARS-HORIZONTAL - 1 results in an
integer that is not a rocket. This means that a decision must be made to add or not to
add 1 to the given rocket. With this insight, the f-on-rocket template is specialized
as follows:
;; rocket → rocket
;; Purpose: Move the given rocket right
(define (move-rckt-right a-rocket)
(if (< a-rocket (sub1 MAX-CHARS-HORIZONTAL))
(add1 a-rocket)
a-rocket]))

;; Sample expressions for move-rckt-right

(define ROCKET-VAL3 (add1 3))
(define ROCKET-VALMAX (sub1 MAX-CHARS-HORIZONTAL))

;; Tests using sample computations for move-rckt-right

(check-expect (move-rckt-right 3) ROCKET-VAL3)
(check-expect (move-rckt-right (sub1 MAX-CHARS-HORIZONTAL))
ROCKET-VALMAX)

;; Tests using sample values for move-rckt-right

(check-expect (move-rckt-right 0) 1)
(check-expect (move-rckt-right 15) 16)
The sample expressions illustrate how rockets that may and may not move to the
right are computed. In the body of move-rckt-right a new rocket is created by
adding 1 only when the given rocket is not at the right edge of the scene. The tests
using sample values include an untested extrema value, 0, for the rocket type and
add an extra test for thoroughness.
A similar problem analysis leads to concluding that a conditional is needed to
decide when a rocket may be moved to the left. In this case, only when the rocket is
greater than or equal to 0. The template specialization results in:
;; rocket → rocket
;; Purpose: Move the given rocket left
(define (move-rckt-left a-rocket)
(if (> a-rocket 0)
(sub1 a-rocket)
a-rocket))

;; Sample expressions for move-rckt-left

(define ROCKET-VAL7 (sub1 7))
(define ROCKET-VAL0 0)
37 What Have We Learned in This Chapter? 149

;; Tests using sample computations for move-rckt-left

(check-expect (move-rckt-left 7) ROCKET-VAL7)
(check-expect (move-rckt-left 0) 0)

;; Tests using sample values for move-rckt-left

(check-expect (move-rckt-left (sub1 MAX-CHARS-HORIZONTAL))
(- MAX-CHARS-HORIZONTAL 2))
(check-expect (move-rckt-left 14) 13)
Once again, observe that the tests using sample values include an untested extrema
value, (sub1 MAX-CHARS-HORIZONTAL), for the rocket type and add an extra test
for thoroughness.
The final step of the design recipe is to run all the tests. Make sure that you do
not have any syntax errors and that all tests pass. Run the game and try moving the
rocket left and right.

** Ex. 75 — Redesign Aliens Attack version I using the following data defini-
tion for a rocket:
A rocket is an image-x such that it is either:
1. 0
2. (sub1 MAX-CHARS-HORIZONTAL)
3. (0..(sub1 MAX-CHARS-HORIZONTAL))

** Ex. 76 — Write an interactive simulation of a rocket flying down a scene

that stops when it reaches the bottom of the scene. If the user presses "u" the
rocket’s image points up. If the user presses "d" the rocket’s image points down.

*** Ex. 77 — Design an interactive traffic light simulation such that if the
player presses:
1."r" the traffic light goes to flashing red.
2."y" the traffic light goes to flashing yellow.
3."n" the traffic light goes to normal operation.

37 What Have We Learned in This Chapter?

The important lessons of Chap. 6 are summarized as follows:

• Video game and interactive animation programs require functions to process
computer events.
• The universe teachpack provides an API to program video games and interactive
animations.
• Programming with APIs is standard practice in Computer Science.
150 6 Aliens Attack Version 1

• The universe’s big-bang expression is used to run a video game and allows
a programmer to provide the world’s initial value and the handlers needed to
advance the game’s or animation’s evolution.
• Predicates for data types with varieties sometimes may be simplified by using a
Boolean expression instead of a conditional expression.
• The video game design recipe includes steps to:
– Develop data definitions and sample instances for every element that varies as
the game evolves.
– Develop a function template for every data definition.
– Develop a run function.
• Separation of concerns leads to functions that solve a problem for a single data
type and simplifies the process of iterative refinement.
• Developing video games is intellectually stimulating and exciting.
 Part II
Compound Data of Finite Size
Chapter 7
Structures

As we have seen, one of the important steps in problem solving is choosing a

representation for the data that exists in the real or in an imaginary world. In Chap. 6,
for example, a rocket is represented as an image-x value. This is an example of what
is called data of finite size. A single value is used to represent the single characteristic
that may change about the rocket as the game evolves.
Now consider the Cartesian plane and the integer-based points displayed in Fig.58.
An integer-based point on the Cartesian plain has two characteristics: an integer x
coordinate and an integer y coordinate. This is an example of compound data of
finite size. Compound data of finite size has a constant number of characteristics that
may vary from one instance of the data to another. For example, (0,2) and (-3,-1)
vary in both the x and y coordinates. The data is finite because a point is always
represented with 2 integer values (as opposed to an arbitrary number of values).
A natural question to ask is how are functions that process compound data of finite
size designed? Let us consider the problem of determining if an integer-based point
is in the first quadrant following the steps of the design recipe. The first step of the
design recipe requires that a representation be chosen for a point. You may very well
be thinking that you need two integers. Let us follow this instinct, but there is already
something that feels off. Why are we talking about two integers when the problem is
about integer-based points? We need to process a point but have no data definition
for a point. Instead, this design suggests thinking about x and y coordinates as two
separate values instead of a single value. Nonetheless, let us proceed. To determine
if a point is in the first quadrant, both the x and y coordinates must be greater than 0.
Based on this problem analysis, we can specialize the basic function template (from
Fig. 29) as follows:
;; integer integer → Boolean
;; Purpose: To determine if the given coordinates for a
;; point is in Q1
(define (in-Q1? x y)
(and (> x 0) (> y 0)))

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 153
M. T. Morazán, Animated Problem Solving, Texts in Computer Science,
https://doi.org/10.1007/978-3-030-85091-3_7
154 7 Structures

Fig. 58 The Cartesian plane with several points

(2,3)

(-1,1)

x
(0,0)

(-4,-3)

(4,-5)
y

;; Sample expression definitions for in-Q1?

(define CONSTANT-1 (and (> 3 0) (> 11 0)))
(define CONSTANT-2 (and (> -5 0) (> 23 0)))

;; Tests using sample computations for in-Q1?

(check-expect (in-Q1? 3 11) CONSTANT-1)
(check-expect (in-Q1? -5 23) CONSTANT-2)

;; Tests using sample values for in-Q1?

(check-expect (in-Q1? -3 -7) #false)
(check-expect (in-Q1? 83 -4) #false)
All the tests pass and this gives you guarded confidence that the program works.
However, think carefully about the above solution. There are certain incongruities.
The purpose statement for in-Q1? mentions a point, but in the signature there is
no point in the input. Instead, there are two integers. The sample expressions have
two integers, instead of a point, as differences. The body of in-Q1? processes two
38 The posn Structure 155

integers and not a single point. You may say that it is understood that the two integers
represent a point. Perhaps, that is clear to you now. Will it be clear to other people
who may read your code? How about to yourself 6 months from now? The truth is that
for a two-dimensional point these incongruities may be relatively easy to overcome.
What if you were solving a problem using Einstein’s Mathematics in which every
point has 4 dimensions or in string theory where a point has 10 dimensions? Should
every string theory point-processing function have 10 inputs? That certainly begins
to sound problematic.
The problem with the above solution is that it does not exhibit separation of
concerns. We need a function that takes as input a two-dimensional integer point.
We also need a function to process an integer x coordinate and another function to
process an integer y coordinate. Our immediate need, therefore, is a mechanism to
create a point value out of two coordinates. With such a mechanism, we may rewrite
in-Q1? as something similar to
;; 2D-ipoint → Boolean
;; Purpose: To determine if the given 2D-ipoint is in Q1
(define (in-Q1? a-2dipoint)
(and (x-in-Q1? (x-of a-2dipoint))
(y-in-Q1? (y-of a-2dipoint))))
Most will argue that the above function, if it were possible to write, is easier to
understand than the previous version. It is likely to be as clear 6 months from now
as it is today. To write such nifty functions, it is necessary to learn more about BSL.

38 The posn Structure

In BSL, compound data of finite size may be represented as a single piece of data
using a structure. A structure combines a fixed number of values into a single piece
of data. In addition, for every structure there is a constructor and there is a selector
for each characteristic. A constructor is a function to build instances of the compound
data. A selector retrieves the value of a characteristic.
For instance, BSL provides the posn structure. A posn is intended to represent a
position in a Cartesian plane. In other words, it is intended to represent what is called
a 2D-point (not to be confused with a 2D-ipoint). A posn has two characteristics
(or fields) called x and y. The constructor for a posn is make-posn. It requires two
inputs: one for x and one for y. There are two selector functions: posn-x to retrieve
the value of x and posn-y to retrieve the value of y. Run the following example in
DrRacket:
(define A-POSN (make-posn 3.2 9))
(define A-POSN2 (make-posn -8 9))

(check-expect (posn-x A-POSN) 3.2)

(check-expect (posn-x A-POSN2) -8)
156 7 Structures

(check-expect (posn-y A-POSN) 9)

(check-expect (posn-y A-POSN2) 9)
(check-expect (= (posn-x A-POSN) (posn-x A-POSN2)) #false)
(check-expect (= (posn-y A-POSN) (posn-y A-POSN2)) #true)
This code defines two posn constants that represent, respectively, the points (3.2,
9) and (-8, 9). The first two tests illustrate that posn-x correctly retrieves a posn’s
x value. Tests 3 and 4 illustrate that posn-y correctly retrieves a posn’s y value.
The last two tests illustrate that the x values of the two posns are different and that
the y values of the two posns are the same. After this example, it becomes easy to
understand that two (mathematical) laws govern posns:
(posn-x (make-posn x y)) = x
(posn-y (make-posn x y)) = y
Understanding the above laws, however, is not enough to properly use posns
to represent data from the real or an imaginary world. Consider, for example, the
following program:
(define A-POSN (make-posn "Animating" "Programs"))

(check-expect (posn-x A-POSN) "Animating")

(check-expect (posn-y A-POSN) "Programs")
(check-expect (string=? (string-append (posn-x A-POSN)
(posn-y A-POSN))
"Animating Programs")
#true)
The tests pass after running the program. What happened? A-POSN certainly does not
represent a 2D-point. The explanation is that the constructor make-posn creates
a structure that is able to store two values: one called x and another called y. The
constructor, however, does not enforce any type rules for x and y. That is, the rules
above state that there is an x value and a y value. It does not state anything about the
data types of x and y. Therefore, there is nothing from preventing a programmer to
use make-posn as done in the above program. The next pressing question, of course,
is how do we make clear what a posn represents?
The answer is that to process compound data of fixed size, a data definition is
needed. In turn, this means that we need to develop a function template and sample
instances of the data defined. An interesting consequence is that a structure sug-
gests how the data ought to be processed. Let us reconsider the representation of a
2D-point. The data definitions and function templates are displayed in Fig. 59. An
xcoord and a ycoord are defined as real numbers. Each needs a function template.
The most interesting part is the data definition for a 2D-point. A 2D-point is
defined as a posn structure whose x is an xcoord and whose y is a ycoord. This
suggest that solving a problem about a 2D-point is likely to require solving a prob-
lem about an xcoord, a ycoord, or both. Observe that the body of f-on-2D-point
captures this observation and exhibits separation of concerns by having references
38 The posn Structure 157

Fig. 59 The data definitions and function templates for the 2D-point program
#|;; An x-coordinate (xcoord) is a real number
;; A y-coordinate (ycoord) is a real number
;; A 2D-point is a structure (make-posn xcoord ycoord)

(define XCOORD1 . . .) . . .
(define YCOORD1 . . .) . . .
(define 2D-POINT1 . . .) . . .

;; xcoord . . . ... Purpose: . . .

(define (f-on-xcoord an-xcoord . . .) . . . an-xcoord. . .)
;; Sample expressions for f-on-xcoord
(define XCOORD1-VAL . . .) . . .
;; Tests using sample computations for f-on-xcoord
(check-expect (f-on-xcoord XCOORD1 . . .) XCOORD1-VAL) . . .
;; Tests using sample values for f-on-xcoord
(check-expect (f-on-xcoord . . .) . . .) . . .

;; ycoord . . . ... Purpose: . . .

(define (f-on-ycoord a-ycoord . . .) . . . a-ycoord. . .)
;; Sample expressions for f-on-ycoord
(define YCOORD1-VAL . . .) . . .
;; Tests using sample computations for f-on-ycoord
(check-expect (f-on-ycoord YCORD1 . . .) YCOORD1-VAL) . . .
;; Tests using sample values for f-on-ycoord
(check-expect (f-on-ycoord . . .) . . .) . . .

;; 2D-point . . . ... Purpose: . . .

(define (f-on-2D-point a-2Dpoint . . .)
(f-on-xcoord (posn-x a-2Dpoint)). . .(f-on-ycoord (posn-y a-2Dpoint)). . .)
;; Sample expressions for f-on-2D-point
(define 2D-POINT1-VAL . . .) ...
;; Tests using sample computations for f-on-2D-point
(check-expect (f-on-2D-point 2D-POINT1 . . .) 2D-POINT1-VAL) . . .
;; Tests using sample values for f-on-2D-2D-point
(check-expect (f-on-2D-point . . .) . . .) ... |#

to functions that process an xcoord and a ycoord. Based on this data analysis, the
2D-point program above may be refined to
;; Sample instances of xcoord
(define X1 3.2)
(define X2 -8)

;; Sample instances of ycoord

(define Y1 9)
(define Y2 9)

;; Sample instances of 2D-point

(define A-POSN (make-posn X1 Y1))
(define A-POSN2 (make-posn X2 Y2))
158 7 Structures

(check-expect (posn-x A-POSN) X1)

(check-expect (posn-x A-POSN2) X2)
(check-expect (posn-y A-POSN) Y1)
(check-expect (posn-y A-POSN2) Y2)
(check-expect (= (posn-x A-POSN) (posn-x A-POSN2)) #false)
(check-expect (= (posn-y A-POSN) (posn-y A-POSN2)) #true)
The results of data analysis may be reused multiple times. For example, design
a program to determine if a 2D-point is on the graph of f(x) = x3 . Problem
analysis tells us that (x, y) is on the graph of f(x) = x3 if y = x3 . Observe that
determining the answer requires processing x, but there is no need to process y.
This means that a function to process a ycoord is not needed. The initial template
specialization is
(define X1 3)
(define X2 -2.5)

(define Y1 27)
(define Y2 39)

(define 2D-POINT1 (make-posn X1 Y1))

(define 2D-POINT2 (make-posn X2 Y2))

;; 2D-point . . . → Boolean
;; Purpose: Determine if the given 2D-point is on the
;; graph of f(x) = x3
(define (on-x^3? a-2Dpoint)
. . .(f-on-xcoord (posn-x a-2Dpoint)). . .)

;; Sample expressions for on-x^3?

(define 2D-POINT1-VAL (= (cube (posn-x 2D-POINT1))
(posn-y 2D-POINT1)))
(define 2D-POINT2-VAL (= (cube (posn-x 2D-POINT2))
(posn-y 2D-POINT2)))

;; Tests using sample computations for on-x^3?

(check-expect (on-x^3? 2D-POINT1) 2D-POINT-VAL1)
(check-expect (on-x^3? 2D-POINT2) 2D-POINT-VAL2)

;; Tests using sample values for on-x^3?

(check-expect (on-x^3? . . .) . . .)
The code first includes sample instances of all the defined data types. The sample
instances chosen are a 2D-point that is on the graph of f(x) = x3 and a 2D-point
that is not on the graph of f(x) = x3 . Observe that the only difference among the
38 The posn Structure 159

sample expressions is the 2D-point processed. Therefore, on-x3 only needs one
parameter.
Tests using sample values are added for testing thoroughness. The complete
template specialization is
(define X1 3)
(define X2 -2.5)

(define Y1 27)
(define Y2 39)

(define 2D-POINT1 (make-posn X1 Y1))

(define 2D-POINT2 (make-posn X2 Y2))

;; 2D-point → Boolean
;; Purpose: Determine if the given 2D-point is on the
;; graph of f(x) = x3 3
(define (on-x^3? a-2Dpoint)
(= (cube (posn-x a2D-point)) (posn-y a-2Dpoint)))

;; Sample expressions for on-x^3?

(define 2D-POINT1-VAL (= (cube (posn-x 2D-POINT1))
(posn-y 2D-POINT1)))
(define 2D-POINT2-VAL (= (cube (posn-x 2D-POINT2))
(posn-y 2D-POINT1)))

;; Tests using sample computations for on-x^3?

(check-expect (on-x^3? 2D-POINT1) 2D-POINT1-VAL)
(check-expect (on-x^3? 2D-POINT2) 2D-POINT2-VAL)

;; Tests using sample values for on-x^3?

(check-expect (on-x^3? (make-posn 0 0)) #true)
(check-expect (on-x^3? (make-posn -3.3 -35.937)) #true)
(check-expect (on-x^3? (make-posn 5 25)) #false)
The function cube is not part of BSL and must be designed. This is done by
specializing the f-on-xcoord template. The answer is computed by raising the
given xcoord to the third power. The result of specializing the template is
;; xcoord → xcoord
;; Purpose: To cube the given xcoord
(define (cube an-xcoord) (expt an-xcoord 3))

;; Sample expressions for f-on-xcoord

(define XCOORD-VAL1 (expt -2 3))
(define XCOORD-VAL2 (expt 8 3))
(define XCOORD-VAL3 (expt 3 3))
160 7 Structures

Fig. 60 The refined general design recipe for functions

Data Design:

1. Outline data representation and create data definitions.

2. Create sample instances for each data definition.
3. Create a function template for each data definition.

For a needed function:

4. Outline the computation.
5. Define constants for the value of sample expressions.
6. Identify and name the differences among the sample expressions.
7. Write the function’s signature and purpose.
8. Write the function’s header.
9. Write tests.
10. Write the function’s body.
11. Run the tests and, if necessary, redesign.

;; Tests using sample computations for f-on-xcoord

(check-expect (cube -2) XCOORD-VAL1)
(check-expect (cube 8) XCOORD-VAL2)
(check-expect (cube X1) XCOORD-VAL3)

;; Tests using sample values for f-on-xcoord

(check-expect (cube 0) 0)
(check-within (cube X2) -15.62 0.01)
Observe that there is no need to define new instances of xcoords. Sample instances
have already been defined and may be freely used. Running the program reveals that
all the tests pass. This completes the design of on-x3?.
The steps taken suggest that the general design recipe for functions needs to be
refined. There are two parts: data design and function design. This division of labor
captures the important role that data representation plays in problem solving and in
programming. Figure 60 displays the refined general design recipe for functions. The
first three steps are part of data design and are performed once. These steps include
picking a data representation and creating data definitions, creating sample instances
of the data, and creating function templates. Steps 4–11 are the steps followed for
each function that needs to be written. These steps are fulfilled by specializing a
function template.

39 Going Beyond the Design Recipe

Before proceeding with another example, it is worth to take a pause to analyze the
result obtained for on-x3 . It is natural to ask if the solution may be simplified. This
39 Going Beyond the Design Recipe 161

step is not necessary but may be performed to aid code readability and/or to reduce
the amount of code.
The design steps for on-x3 revealed the need for and the subsequent design of
cube. The function cube, however, is rather simple and easy to understand. In fact,
defining cube may add little clarity to the program—this, of course, is a matter of
opinion. If you feel that function does not add clarity to a program, you may want
to consider inlining it. In this case, inlining is straightforward and results in the
following refined version of the program:
(define X1 3)
(define X2 -2.5)

(define Y1 27)
(define Y2 39)

(define 2D-POINT1 (make-posn X1 Y1))

(define 2D-POINT2 (make-posn X2 Y2))

;; 2D-point → Boolean
;; Purpose: Determine if the given point is on the graph
;; of f(x) = x^3
(define (on-x^3? a-2Dpoint)
(= (expt (posn-x a-2Dpoint) 3) (posn-y a-2Dpoint)))

;; Sample expressions for on-x^3?

(define 2D-POINT1-VAL (= (expt (posn-x 2D-POINT1) 3)
(posn-y 2D-POINT1)))
(define 2D-POINT2-VAL (= (expt (posn-x 2D-POINT2) 3)
(posn-y 2D-POINT1)))

;; Tests using sample computations for on-x^3?

(check-expect (on-x^3? 2D-POINT1) 2D-POINT1-VAL)
(check-expect (on-x^3? 2D-POINT2) 2D-POINT2-VAL)

;; Tests using sample values for on-x^3?

(check-expect (on-x^3? (make-posn 0 0)) #true)
(check-expect (on-x^3? (make-posn -3.3 -35.937)) #true)
(check-expect (on-x^3? (make-posn 5 25)) #false)
Observe that all calls to cube in on-x3 ? and its tests are eliminated, and therefore,
the need for defining cube is also eliminated. The program is certainly shorter. Is
this refined code clearer? It is certainly not significantly more difficult to understand
suggesting that it was fine to perform inlining. You should not, however, expect that
inlining code always results in shorter code and in just as clear code. On the contrary,
more often than not inlining results in code explosion. Code explosion means that the
length of the code becomes significantly larger. Instead of using cube once, imagine
162 7 Structures

it was used 100 times. Inlining in this scenario means that similar code is repeated
100 times in the program. Remember that in Chap. 3, this is precisely the situation we
wished to avoid. It is better to abstract over similar expressions to create a function
than to repeatedly have similar expressions throughout the code. Beyond eliminating
repetitions, abstraction over similar expressions also makes revisions easier. Imagine
that a better way to compute x3 is discovered. If cube is inlined 100 times, then the
revision to use the better way to compute x3 requires 100 edits. On the other hand, if
cube is not inlined, then only one edit is required. A good problem solver foresees
such possibilities and does not forsake separation of concerns in favor of having to
design fewer functions.
When should inlining be considered? If a function is only used once and its design
does not require any type of specialized knowledge, then you may consider inlining.
Keep in mind, however, that inlining is usually left to a compiler. As problem solvers,
the rule of thumb is to avoid inlining.

40 Revisiting in-Q1?

Developing the predicate to determine if an integer-based point is in the first quadrant

presents an excellent opportunity to put the revised design recipe in action. The
first three steps of the design recipe may be satisfied as displayed in Fig. 61. In this
program, both an ix and an iy are integers. Do not confuse these data definitions with
those developed for the program to determine if a 2D-point is on the graph of f(x)
= x3 . A 2D-ipoint is a posn that contains an ix and an iy. The function templates
to process an ix and an iy outline the expected structure for the function definition
and the tests. This structure is minimal because integers are not decomposable into
different pieces. The function template for 2D-ipoint, on the other hand, provides
more structure for problem solving because a 2D-ipoint is decomposable into
different pieces. To solve a problem about a 2D-ipoint, we can process an ix,
an iy, or both. Finally, Step 3 suggests the development of one or more sample
instances for each data definition. The minimum for each is one because none of the
data definitions has variety. In this case, two sample instances are defined for each
data type to exemplify a 2D-ipoint that is in the first quadrant and a 2D-ipoint
that is not in the first quadrant.
Function development starts with the specialization of the f-on-2D-ipoint
template for in-Q1?. For a given 2D-ipoint to be in the first quadrant, both its ix
and iy must be in the first quadrant interval for each dimension. Sample expressions
and tests using sample computations may be defined using the two sample instances
of 2D-ipoint (one in the first quadrant and one in the second quadrant). The
tests using sample values may test 2D-ipoints in quadrants three and four and an
additional point in quadrant one. These observations allow for the following template
specialization:
;; 2D-ipoint → Boolean
;; Purpose: To determine if the given 2D-ipoint is in the first
;; quadrant
40 Revisiting in-Q1? 163

Fig. 61 The first three steps of the design recipe defining 2D-ipoint
#| ;; An x-coordinate (ix) and a y-coordinate (iy) are integers
;; A 2D-ipoint is a structure (make-posn ix iy).
;; Sample instances of ix: (define IX1 . . .) . . .
;; Sample instances of iy: (define IY1 . . .)
;; Sample instances of 2D-ipoint: (define 2D-IPOINT1 (make-posn . . . . . .)). . .
;; ix . . . ... Purpose: . . .
(define (f-on-ix an-ix . . .) . . . an-ix. . .)
;; Sample expressions for f-on-ix
(define IX-VAL . . .)
;; Tests using sample computations for f-on-ix
(check-expect (f-on-ix . . .) IX-VAL) ...
;; Tests using sample values for f-on-ix
(check-expect (f-on-ix . . .) . . .) ...
;; iy . . . ... Purpose: . . .
(define (f-on-iy a-iy . . .) . . . a-iy. . .)
;; Sample expressions for f-on-iy
(define IY-VAL . . .) ...
;; Tests using sample computations for f-on-iy
(check-expect (f-on-iy . . .) IY-VAL) ...
;; Tests using sample values for f-on-iy
(check-expect (f-on-iy . . .) . . .) ...
;; 2D-ipoint . . . ... Purpose: . . .
(define (f-on-2D-ipoint a-2Dipoint . . .)
. . .(f-on-ix (posn-x a-2Dipoint)). . .(f-on-iy (posn-y a-2Dipoint)). . .)
;; Sample expressions for f-on-2D-ipoint
(define 2D-IPOINT-VAL . . .) ...
;; Tests using sample computations for f-on-2D-ipoint
(check-expect (f-on-2D-ipoint . . .) 2D-IPOINT-VAL) ...
;; Tests using sample values for f-on-2D-ipoint
(check-expect (f-on-2D-ipoint . . .) . . .) . . . |#
;; Sample instances of ix
(define IX1 3) (define IX2 -5)
;; Sample instances of iy
(define IY1 11) (define IY2 23)
;; Sample instances of 2D-ipoint
(define 2D-IPOINT1 (make-posn IX1 IY1))
(define 2D-IPOINT2 (make-posn IX2 IY2))

(define (in-Q1? a-2Dipoint)

(and (ix-in-Q1? (posn-x a-2Dipoint))
(iy-in-Q1? (posn-y a-2Dipoint))))

;; Sample expressions for in-Q1?

(define 2D-IPOINT1-VAL (and (ix-in-Q1? (posn-x 2D-IPOINT1))
(iy-in-Q1? (posn-y 2D-IPOINT1))))

(define 2D-IPOINT2-VAL (and (ix-in-Q1? (posn-x 2D-IPOINT2))

(iy-in-Q1? (posn-y 2D-IPOINT2))))
164 7 Structures

;; Tests using sample computations for in-Q1?

(check-expect (in-Q1? 2D-IPOINT1) 2D-IPOINT1-VAL)
(check-expect (in-Q1? 2D-IPOINT2) 2D-IPOINT2-VAL)

;; Tests using sample values for in-Q1?

(check-expect (in-Q1? (make-posn -3 -7)) #false)
(check-expect (in-Q1? (make-posn 83 -4)) #false)
(check-expect (in-Q1? (make-posn 100 864)) #true)
Observe that the steps of the design recipe, in essence, yield the function at the end
of the introduction to this chapter. Any reader of this function, including yourself 6
months from now, will be able to understand the solution to the problem.
The specialization of the template for functions on an ix and on an iy follows
the same pattern seen in the previous chapters. To be in the first quadrant interval,
an ix must be greater than 0. The same is true for an iy. The specialization process
yields
;; ix → Boolean
;; Purpose: To determine if the given ix is in the Q1 range
(define (ix-in-Q1? an-ix)
(> an-ix 0))

;; Sample expressions for ix-in-Q1?

(define IX1-VAL (> IX1 0))
(define IX2-VAL (> IX2 0))

;; Tests using sample computations for ix-in-Q1?

(check-expect (ix-in-Q1? IX1) IX1-VAL)
(check-expect (ix-in-Q1? IX2) IX2-VAL)

;; Tests using sample values for ix-in-Q1?

(check-expect (ix-in-Q1? 0) #false)
(check-expect (ix-in-Q1? 19380) #true)

;; iy → Boolean
;; Purpose: To determine if the given iy is in the Q1 range
(define (iy-in-Q1? an-iy)
(> an-iy 0))

;; Sample expressions for iy-in-Q1?

(define IY1-VAL (> IY1 0))
(define IY2-VAL (> IY2 0))

;; Tests using sample computations for iy-in-Q1?

(check-expect (iy-in-Q1? IY1) IY1-VAL)
(check-expect (iy-in-Q1? IY2) IY2-VAL)
40 Revisiting in-Q1? 165

;; Tests using sample values for iy-in-Q1?

(check-expect (iy-in-Q1? 0) #false)
(check-expect (iy-in-Q1? 19380) #true)
The only new element here is that we exploit the definition of sample instances to
write sample expressions and tests using sample computations.
After running the program, all the tests pass. Observe, however, that (iy-in-Q1?
(posn-y A-2D-IPOINT2)) in the definition of 2D-IPOINT-VAL2 is highlighted
in black. Such highlighting usually means that more tests are needed to reach the
untested code. In this case, it is impossible for the program to evaluate the highlighted
code. Remember how and works in BSL. The expressions after the first expression
that evaluates to #false are not evaluated. The x of A-2D-IPOINT2 is -5. Thus, the
value of its y is never tested.

* Ex. 78 — Design and implement predicates to

1. Determine if a 2D-ipoint is in the second quadrant.
2. Determine if a 2D-ipoint is in the third quadrant.
3. Determine if a 2D-ipoint is in the fourth quadrant.

* Ex. 79 — Design and implement predicates to

1. Determine if a 2D-ipoint is on the x-axis.
2. Determine if a 2D-ipoint is on the y-axis.

* Ex. 80 — Design and implement a function to determine if a 2D-point is

on the graph of f(x) = 2x3 - 4x2 + 8x - 1.

** Ex. 81 — Using 2D-points,

• Design and implement a function to compute the distance between two
points.
• Design and implement a function to compute the distance between a
2D-point and the origin.

** Ex. 82 — Design and implement a function to translate a 2D-point. The

translation is performed by creating a new 2D-point that moves a given
2D-point by some given amount on the x-axis and by some given amount
on the y-axis. For example, (make-posn -2 10) is the result of translating
(make-posn 5 -2) by moving it by -7 on the x-axis and by 12 on the y-axis.

** Ex. 83 — Design and implement a function to compute the length of a

person’s full name. A full name consists of a first name and a last name.
166 7 Structures

41 What Have We Learned in This Chapter?

The important lessons of Chap. 7 are summarized as follows:

• Compound data of finite size has a constant number of characteristics that may
vary from one instance of the data to another.
• Compound data of finite size is represented using structures.
• A structure combines a fixed number of values into a single piece of data.
• Every structure has a constructor and a selector for each characteristic.
• BSL provides the posn structure that has two fields: x and y.
• Processing structures requires a data definition that specifies the types of each
characteristic.
• Every data definition requires the development of a function template and sample
instances.
• The same structure may be used to represent different types of data.
• The results of the design recipe’s data design steps may be used to implement
more than one function.
• Inlining may result in code explosion caused by repetitions in the code.
• A good problem solver does not forsake separation of concerns in favor of having
to design fewer functions.
Chapter 8
Defining Structures

It turns out that rarely is the input to a function, for example, a single string or
number. It is far more common for there to be several pieces of related data that
need to be processed. A 2D-point, for instance, may be represented using a posn
structure. A posn structure works well, because a 2D-point only has two varying
characteristics. What needs to be done if data has more than two (but finite) varying
characteristics? Clearly, a structure with two fields, like a posn, cannot be used to
represent such data.
Consider representing a student that has a first name, a middle name, a last name,
and a grade point average. How can a student be represented? What is needed is a
structure that has four fields: one for each varying characteristic. You may imagine
that BSL provides a structure with four fields, but it does not. There is actually a
good reason for this. As you can imagine, the variety of compound data of fixed
size is infinite. Everybody can imagine data with 2, 3, 4, 5, 6, or more varying
characteristics. There is no way BSL can provide structures of all possible sizes.
Therefore, BSL gives programmers the ability to define their own structures. In this
manner, a programmer can define finite compound data of any size and provide
custom names to the structure and its fields. The first step is to learn the necessary
BSL syntax to define structures.

42 Defining Structures

To define structures the BSL grammar is expanded as follows:

program ::= {expr | test | def | defs}∗
defs ::= (define-struct <structure name> (<field name>∗ ))
A program may contain zero or more expressions, tests, definitions, and structure
definitions. A structure definition starts with an opening parenthesis, followed by the
keyword define-struct, followed by the structure’s name, followed in parentheses
by the name of each field. The number of fields a structure has may be zero or more,
but must be a finite number.

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 167
M. T. Morazán, Animated Problem Solving, Texts in Computer Science,
https://doi.org/10.1007/978-3-030-85091-3_8
168 8 Defining Structures

A structure definition creates much more than just a structure that can store
multiple values. It also creates the constructor and selector functions for the structure.
The naming conventions for these functions is straightforward. The constructor is
always named:
make-<structure name>
The number of arguments required by a constructor is equal to the number of fields
the structure has. The selectors are always named:
<structure name>-<field name>
The input to a selector must be an instance of <structure name>.
We can now define a structure for a 2D-point as follows:
(define-struct 2Dpoint (xval yval))
This structure definition creates the following functions to manipulate 2D-points:
;; X Y → 2Dpoint
;; Purpose: To create a 2Dpoint with the given values
(define (make-2Dpoint an-X an-Y) . . .)

;; 2Dpoint → X
;; Purpose: Return the xval of the given 2Dpoint
(define (2Dpoint-xval a-2Dpoint) . . .)

;; 2Dpoint → Y
;; Purpose: Return the yval of the given 2Dpoint
(define (2Dpoint-yval a-2Dpoint) . . .)

;; Any → Boolean
;; Purpose: To determine if the given value is a 2Dpoint
(define (2Dpoint? any-value) ...)
Do not worry about how these functions are implemented in BSL. There is no need to
know these details. It is important, however, to understand the structure API provided
by BSL. The first function is called a constructor. A constructor is a function that
is used to create data type instances. The rest of the functions are observers. An
observer is a function that tells us something about an instance of a data type. The
last function, 2Dpoint?, is a predicate observer that distinguishes between data that
is a 2Dpoint and data that is not a 2Dpoint.
Take note that the contracts for the constructor and the selectors refer to inde-
terminate types called X and Y. X and Y are type variables and represent a defined
type that exists in BSL or that has been defined by a programmer. The type variables
indicate that a structure is generic. This means that it works equally well for many
different data types. BSL assumes nothing about the type of data that the structure
stores. Genericity provides programmers with a great deal of flexibility, but must be
used with care. For example, a programmer may think that the following are valid
2D-points:
42 Defining Structures 169

(define A (make-2Dpoint 34 90))

(define B (make-2Dpoint "Hello" "World!"))

(define C (make-2Dpoint #true 'CS))

How does a programmer know that any of the above are correct uses of the
constructor? Without a data definition it is impossible to know, because X and Y may
represent any type. A data definition provides the context that defines what X and Y
are expected to be. For example, we can define a 2D-point as follows:
A 2Dpoint is a structure (make-2Dpoint R R)
This data definition states that X is a real number and that Y is a real number. In turn,
this tells us that only the definition for A above makes correct use of the 2Dpoint
constructor. In essence, the data definition tells you and others that the proper use of
a generic structure. In this example, a 2D-point is used properly only when it stores
two real numbers.
As you well know, the second step of data analysis in the design recipe tells us that
we ought to develop of function template for 2D-point. The template for 2D-point
is:
;; Sample Instances of 2Dpoint
(define P1 (make-2Dpoint . . .))

;; 2Dpoint . . . → . . .
;; Purpose: . . .
(define (f-on-2Dpoint a-2dpoint . . .)
. . .(2Dpoint-xval a-2dpoint). . .(2Dpoint-yval a-2dpoint). . .)

;; Sample expressions for f-on-2Dpoint

(define 2DP-VAL1 . . .(2Dpoint-xval P1). . .(2Dpoint-xval P1). . .)
...

;; Tests using sample computations for f-on-2Dpoint

(check-within (f-on-2Dpoint P1 . . .) 2DP-VAL1)
...

;; Tests using sample values for f-on-2Dpoint

(check-within (f-on-2Dpoint . . .) . . .)
...
Now, we can start solving problems with this representation of a 2D-point. For
example, write a function to compute the distance of a 2D-point from the origin.
The distance to the origin of a 2D-point may be computed using the distance
170 8 Defining Structures

formula. Given that one of the points, the origin, is always (0, 0) the distance formula
may be simplified as follows:

distance((x1, y1 ), (x2 , y2 )) = (x2 − x1 )2 +(y2 − y1 )2

distance((0, 0), (x2 , y2 )) = (x2 − 0)2 +(y2 − 0)2

= x22 + y22

The template specialization may be started as follows:

;; A 2Dpoint is a structure (make-2Dpoint R R)
(define-struct 2Dpoint (xval yval))

;; Sample instances of 2Dpoint

(define P1 (make-2Dpoint 5 10))
(define P2 (make-2Dpoint 25 11))

;; 2Dpoint → R
;; Purpose: To compute the distance to the origin for the
;; given 2Dpoint
(define (distance-to-origin a-2dpoint)
. . .(2Dpoint-xval a-2dpoint). . .(2Dpoint-yval a-2dpoint). . .)

;; Sample expressions for distance-to-origin

(define 2DP-VAL1 (sqrt (+ (sqr (2Dpoint-xval P1))
(sqr (2Dpoint-xval P1)))))
(define 2DP-VAL2 (sqrt (+ (sqr (2Dpoint-xval P2))
(sqr (2Dpoint-xval P2)))))

;; Tests using sample computations for distance-to-origin

(check-within (distance-to-origin P1) 2DP-VAL1 0.01)
(check-within (distance-to-origin P2) 2DP-VAL2 0.01)

;; Tests using sample values for distance-to-origin

(check-within (distance-to-origin . . .) . . .)
Observe that the sample expressions are written using the sample instances. There
is a single difference among the sample expressions: the 2Dpoint. This tells us that
distance-to-origin only needs a 2Dpoint as input. The tests using sample com-
putations use the constants defined for the sample instances and the value of the sam-
ple expressions. The final refinement completes the body of distance-to-origin
and the tests using sample values as follows:
;; A 2Dpoint is a structure (make-2Dpoint R R)
(define-struct 2Dpoint (xval yval))
43 Computing Structures 171

;; Sample instances of 2Dpoint

(define P1 (make-2Dpoint 5 10))
(define P2 (make-2Dpoint 25 11))

;; 2Dpoint → real-number
;; Purpose: To compute the distance to the origin for the
;; given 2Dpoint
(define (distance-to-origin a-2dpoint)
(sqrt (+ (sqr (2Dpoint-xval a-2dpoint))
(sqr (2Dpoint-yval a-2dpoint)))))

;; Sample expressions for distance-to-origin

(define 2DP-VAL1 (sqrt (+ (sqr (2Dpoint-xval P1))
(sqr (2Dpoint-xval P1)))))
(define 2DP-VAL2 (sqrt (+ (sqr (2Dpoint-xval P2))
(sqr (2Dpoint-xval P2)))))

;; Tests using sample computations for distance-to-origin

(check-within (distance-to-origin P1) 2DP-VAL1 0.01)
(check-within (distance-to-origin P2) 2DP-VAL2 0.01)

;; Tests using sample values for distance-to-origin

(check-within (distance-to-origin (make-2Dpoint 0 0))
0
0.01)
(check-within (distance-to-origin (make-2Dpoint -3 -4))
5
0.01)
The final refinement develops the body of distance-to-origin to be similar
to the sample expressions. Instead of using a concrete value, it is written using
the parameter that represents the difference. The tests using sample values test an
extrema value (i.e., the origin) and a 2Dpoint that is different from the sample
instances.

43 Computing Structures

Just like numbers, Booleans, strings, and posns, the structures a programmer defines
are first class is BSL. First class means that they may be passed as input to functions
and may be returned as a function value. In other words, functions may compute
instances of a structure.
To illustrate the computation of structures consider the problem of updating a
student’s grade point average. A student, as suggested by the description at the
172 8 Defining Structures

beginning of this chapter, has a first name, a middle name, a last name, and a grade
point average. This leads to the following data definitions, structure definition, and
function templates:
#| DATA DEFINITIONS
A GPA is a real number in [0..4].

;; Sample instances for GPA

(define GPA1 . . .)
...

A student is a structure
(make-student string string string R)
that contains a first name, a middle name, a last name,
and a grade point average.

;; Sample instances for student

(define STUD1 (make-student . . . . . . . . . . . .)
...
|#

(define-struct student (fn mn ln gpa))

#| FUNCTION TEMPLATES
;; GPA . . . → . . .
;; Purpose: . . .
(define (f-on-GPA a-gpa . . .) . . .)

;; Sample expressions for f-on-GPA

(define GPA1-VAL . . .)
...

;; Tests using sample computations for f-on-GPA

(check-expect (f-on-GPA GPA1 . . .) GPA1-VAL)
...

;; Tests using sample values for f-on-GPA

(check-expect (f-on-student . . .) . . .)
...

;; student . . . → . . .
;; Purpose: . . .
(define (f-on-student a-student . . .)
. . .(f-on-string (student-fn a-student)). . .
. . .(f-on-string (student-mn a-student)). . .
. . .(f-on-string (student-ln a-student)). . .
43 Computing Structures 173

. . .(f-on-GPA(student-gpa a-student)). . .)

;; Sample expressions for f-on-student

(define STUD1-VAL . . .)
...

;; Tests using sample computations for f-on-student

(check-expect (f-on-student STUD1 . . .) STUD1-VAL)
...

;; Tests using sample values for f-on-student

(check-expect (f-on-student (make-student . . . . . . . . . . . .)). . .)
...
|#
The GPA data definition and its function template are similar to those seen in previous
chapters. The template contains a skeleton for sample instances, for sample expres-
sions, for a function definition, and for tests. The student data definition is more
interesting. It specifies that a student structure has four fields: 3 strings and a GPA
representing, respectively, first, middle, and last names and a grade point average.
The structure definition names the structure student and the four fields fn for the
first name, mn for the middle name, ln for the last name, and gpa for the grade point
average. The template’s function definition states that there is at least one student
input named a-student. The body of the template’s definition contains the four
expressions that may be useful in solving problems involving a student: one for each
student field. Observe that the template suggests using a string function to process
a string and a GPA function to process a GPA. As with all templates, the function
template includes the basic skeleton for sample instances, sample expressions to ab-
stract over to obtain a function’s body, tests using sample computations and the value
of sample expressions, and tests using sample values that use different instances of
student.
To start, we can define several instances of GPA and of student. Given that the
program is intended to change an old GPA for a new GPA, it is likely a good idea to
define several old and new GPAs. In addition, it is a good idea to define a student for
each old-new GPA pair. The following are sample instances of GPA and student:
(define OLDGPA1 3.8)
(define OLDGPA2 3.7)
(define OLDGPA3 3.58)
(define NEWGPA1 3.93)
(define NEWGPA1 3.9)
(define NEWGPA1 3.8)

(define STUD1 (make-student "Barbara" "" "Mucha" OLDGPA1))

(define STUD2 (make-student "Joan" "Elizabeth"
"Feeney" OLDGPA2))
174 8 Defining Structures

(define STUD3 (make-student "Christopher" "Michael"

"Dutra" OLDGPA3))
Observe that the student values are defined using the old-GPA constants. These are
the GPAs that are to be changed for testing purposes.
To change a student’s grade point average, a student and the student’s new gpa are
needed. The change is done by creating a new student structure with the same names
and the new grade point average. This problem analysis leads to the following initial
specialization of the student function template:
;; student real . . . → student
;; Purpose: Update the given student’s gpa
(define (update-student-gpa a-student a-gpa . . .)
. . .(student-fn a-student). . .
. . .(student-mn a-student). . .
. . .(student-ln a-student). . .
. . .(student-gpa a-student). . .)

;; Sample expressions for update-student-gpa

(define STUD1-VAL (make-student (student-fn STUD1)
(student-mn STUD1)
(student-ln STUD1)
NEWGPA1))

(define STUD2-VAL (make-student (student-fn STUD2)

(student-mn STUD2)
(student-ln STUD2)
NEWGPA2))

(define STUD3-VAL (make-student (student-fn STUD3)

(student-mn STUD3)
(student-ln STUD3)
NEWGPA3))

;; Tests using sample computations for update-student-gpa

(check-expect (update-student-gpa STUD1 NEWGPA1 . . .)
STUD1-VAL)
(check-expect (update-student-gpa STUD2 NEWGPA2 . . .)
STUD2-VAL)
(check-expect (update-student-gpa STUD3 NEWGPA3 . . .)
STUD3-VAL)

;; Tests using sample values for update-student-gpa

(check-expect
(update-student-gpa (make-student . . . . . . . . . . . .))
. . .)
43 Computing Structures 175

Observe that the sample expressions create new instances of students that retain
the names of an existing student structure and that use one of the new GPA sam-
ple instances. There are two differences among the sample expressions: a student
and a GPA. This informs us that update-student-gpa only needs two parameters.
Further observe, that neither a string nor a GPA is processed. Therefore, there is no
need to call string or GPA processing functions. The tests using sample computa-
tions utilize the sample students, the sample (new) GPAs, and the values of the
evaluated sample expressions. This leads to the following final specializations for
update-student-gpa and the tests:
;; student real → student
;; Purpose: Update the given student’s gpa
(define (update-student-gpa a-student a-gpa)
(make-student (student-fn a-student)
(student-mn a-student)
(student-ln a-student)
a-gpa))

;; Tests using sample computations for update-student-gpa

(check-expect (update-student-gpa STUD1 NEWGPA1)
STUD1-VAL)
(check-expect (update-student-gpa STUD2 NEWGPA2)
STUD2-VAL)
(check-expect (update-student-gpa STUD3 NEWGPA3)
STUD3-VAL)

;; Tests using sample values for update-student-gpa

(check-expect (update-student-gpa
(make-student "Sandy" "" "Marinakys" 3.1)
3.4)
(make-student "Sandy" "" "Marinakys" 3.4))
(check-expect (update-student-gpa
(make-student "Lidia" "Carolina" "Vazquez" 3.5)
3.5)
(make-student "Lidia" "Carolina" "Vazquez" 3.5))
(check-expect (update-student-gpa
(make-student "Luis" "Manuel" "Diaz" 3.7)
3.6)
(make-student "Luis" "Manuel" "Diaz" 3.6))
As expected, no auxiliary functions to process a string or a gpa are needed. The new
tests demonstrate that update-student-gpa works for students other than those
defined. The second test illustrates that a new grade point average may be the same
as an existing grade point average resulting in an unchanged student structure. The
third test illustrates that update-student-gpa works when a grade point average
goes down.
176 8 Defining Structures

44 Structures for the Masses

Although BSL is exceptionally well-equipped and efficient in creating and manipu-

lating structures, it is not hard to imagine that most people are not as well-equipped.
Does the president of your university know what a structure is? Does your grand-
mother know what a structure is? Yes, it is possible that some university presidents
and some grandmothers know about structures. However, it is unlikely that the av-
erage university president or average grandmother actually knows about structures
in programming. Both, of course, may benefit from structures just like people in a
remote Amazon village benefit from vaccines without knowing anything about anti-
bodies. In other words, non-programmers may benefit from the work of programmers
without knowing anything about Computer Science.
This poses a challenge for problem solvers because the representation of data in a
program may not be easily understood by human beings. The chosen representation
commonly has to be transformed into something understood outside of the program.
Consider, for example, the values of STUD1 and STUD2 from the program above.
Entering these at the prompt displays:
(make-student "Barbara" "" "Mucha" 3.8)

and

(make-student "Joan" "Elizabeth" "Feeney" 3.7)

What would your university president or grandmother say if they saw that printed
on their computer screen? It is certain that even if they understand structures they
will be unable to understand what is being represented. The reason is simple: they
are not aware of the data definition of a student.
A common technique to make computer data understandable to people in the
real world is to convert it to a string. For example, we can design a function,
student2string, to transform a student structure to a string. At the prompt, your
university president and your grandmother may have the following interaction:
> (student2string STUD1)
"Barbara Mucha has a 3.8 grade point average."
> (student2string STUD2)
"Joan E. Feeney has a 3.7 grade point average."
We start by analyzing how the desired string is built. Observe that the returned
strings have the following components:
1. The student’s first name
2. The student’s middle name abbreviation if any
3. The string “has a”
4. The student’s grade point average
5. The string “grade point average.”
44 Structures for the Masses 177

The string for the student’s middle name abbreviation and the string for the student’s
grade point average need to be computed. Once computed, all the strings may be
appended to create the returned string. This analysis allows us to start specializing
the template for functions on a student. We start with developing sample expressions:
;; Sample expressions for student2string
(define STUD1-STR (string-append
(student-fn STUD1)
(middle-name-abbrev (student-mn STUD1))
(student-ln STUD1)
" has a "
(gpa->string (student-gpa STUD1))
" grade point average."))

(define STUD2-STR (string-append

(student-fn STUD2)
(middle-name-abbrev (student-mn STUD2))
(student-ln STUD2)
" has a "
(gpa->string (student-gpa STUD2))
" grade point average."))
Observe that computing the middle name abbreviation and grade point average
strings are relegated to auxiliary functions and are not the focus of a function to
process a student. The only difference among the sample expressions is the student
processed. Therefore, student2string only requires one parameter. This leads to
the following specialization for the function definition and the tests:
;; student → string
;; Purpose: Transform the given student to a string
(define (student2string a-student)
(string-append
(student-fn a-student)
(middle-name-abbrev (student-mn a-student))
(student-ln a-student)
" has a "
(gpa->string (student-gpa a-student))
" grade point average."))

;; Tests using sample computations for student2string

(check-expect (student2string STUD1) STUD1-STR)
(check-expect (student2string STUD2) STUD2-STR)

;; Tests using sample values for student2string

(check-expect
(student2string (make-student "Mercedes" "G." "Merayo" 3.97))
"Mercedes G. Merayo has a 3.97 grade point average.")
178 8 Defining Structures

(check-expect
(student2string (make-student "Manuel" "" "Núñez" 3.89))
"Manuel Núñez has a 3.89 grade point average.")
Let us continue with the task of designing the auxiliary function middle-name
-abbrev. We can observe that not every student has a last name. In this case, the
middle name string ought to be " " to provide a space in between the first and the last
names. If a student has a middle name, then the string ought to be " " followed by the
first letter of the middle name followed by ". ". Note that this scheme provides for
spaces before and after the middle name abbreviation. Our analysis clearly suggests
that there is variety in the data called middle name and, therefore, a data definition
is required:
A middle name (mn) is either:
1. ""
2. not ""
The corresponding function template is:
;; Sample instances for mn
(define MN1 "")
(define MN2 . . .)
...

;; mn ... --> ...

;; Purpose: ...
(define (f-on-nm an-nm ...)
(if (string=? an-nm "")
...
...))

;; Sample expressions for f-on-mn

(define MN1-VAL . . .)
(define MN2-VAL . . .)
...

;; Tests using sample computations for f-on-mn

(check-expect (f-on-mn MN1 . . .) MN1-VAL)
(check-expect (f-on-mn MN2 . . .) MN2-VAL)
...

;; Tests using sample values for f-on-mn

(check-expect (f-on-mn . . .) . . .)
...
Observe that the template suggests writing at least one sample instance for each
variety of mn. Of these, one must be "".
44 Structures for the Masses 179

Before proceeding with the design of middle-name-abbrev, it is important

to note that the above development means that the student data definition and
function template must be refined. The refinement must reflect that we now realize
that a middle name is not just any string. It is an nm. The refinements are:
A student is a structure, (make-student string mn string GPA),
that contains a first name, a middle name, a last name, and a
grade point average.

;; student . . . → . . .
;; Purpose: . . .
(define (f-on-student a-student . . .)
. . .(f-on-string (student-fn a-student)). . .
. . .(f-on-mn (student-mn a-student)). . .
. . .(f-on-string (student-ln a-student)). . .
. . .(f-on-GPA (student-gpa a-student)). . .)
These refinements make it clear to anyone reading your code what a student is and
how a student might to be processed. Observe that the middle name is now an nm
and that the middle name may be processed by calling a function that processes an
nm.
We can now proceed with the design of middle-name-abbrev. If the middle
name is "", then the middle name abbreviation is clearly " " according to our design.
Otherwise, the middle name abbreviation is a space followed by the first letter of the
given mn followed by a period and ending with a space. Sample instances of mn and
sample expressions for f-on-mn are:
;; Sample instances of mn
(define MN1 " ")
(define MN2 "Jose")
(define MN3 "Francisco")

;; Sample expressions for middle-name-abbrev

(define MN1-VAL " ")
(define MN2-VAL (string-append " " (substring MN2 0 1) ". "))
(define MN3-VAL (string-append " " (substring MN3 0 1) ". "))
First observe that computing the middle name abbreviation for "" does not pro-
cess the this nm. All that is required is returning the constant " ". Observe that
the sample expressions that process an mn only have one difference. This means
that middle-name-abbrev only needs one parameter. This leads to the following
specializations of the template:
;; mn → string
;; Purpose: To abbreviated the given middle name
(define (middle-name-abbrev a-mn)
(if (string=? a-mn "")
" "
(string-append " " (substring a-mn 0 1) ". ")))
180 8 Defining Structures

;; Tests using sample computations for middle-name-abbrev

(check-expect (middle-name-abbrev MN1) MN1-VAL)
(check-expect (middle-name-abbrev MN2) MN2-VAL)
(check-expect (middle-name-abbrev MN3) MN3-VAL)

;; Tests using sample values for middle-name-abbrev

(check-expect (middle-name-abbrev "Kaliman") " K. ")
The next task is to design gpa->string. To compute a string from a number
number->string may be used. This leads to the following specialization of the
function template for GPA:
;; GPA → string
;; Purpose: Transform the given GPA to a string
(define (gpa->string a-gpa)
(number->string a-gpa))

;; Sample expressions for gpa->string

(define GPA1-VAL "3.8")
(define GPA2-VAL "3.7")

;; Tests using sample computations for gpa->string

(check-expect (gpa->string OLDGPA1) GPA1-VAL)
(check-expect (gpa->string OLDGPA2) GPA2-VAL)

;; Tests using sample values for gpa->string

(check-expect (gpa->string 4.0) "4.0")
(check-expect (gpa->string 2.3) "2.3")
To complete the design run the tests. Alas! Six tests fail:
Actual value "19/5" differs from "3.8", the expected value.

Actual value "37/10" differs from "3.7", the expected value.

Actual value "4" differs from "4.0", the expected value.

Actual value "23/10" differs from "2.3", the expected value.

Actual value
"Mercedes G. Merayo has a 397/100 grade point average."
differs from
"Mercedes G. Merayo has a 3.97 grade point average.",
the expected value.
44 Structures for the Masses 181

Actual value
"Manuel Núñez has a 389/100 grade point average."
differs from
"Manuel Núñez has a 3.89 grade point average.",
the expected value.
The first four failures are from tests involving gpa->string and the last two failures
are from tests using sample values for student2string. What happened? Where
did those fractions in the actual values come from? Where do we start refining our
solution? To answer the first two questions we must recall the discussion of numbers
in Chap. 2 and better understand how BSL stores numbers. Whenever possible, BSL
stores a numerical value as exact (instead of inexact) number. This is done in an
effort to avoid errors involving computations with inexact numbers. Therefore, a real
number that may be exactly represented is stored as an integer or a fraction. For
example, 4.0 is stored as the integer 4 and 3.8 is stored as the fraction 19 5 . This
explains the test failures for gpa->string.
Observe that the same error is detected by the tests using sample values for
student2string. Why is the error not detected by tests using sample computations
for student2string? The answer lies in realizing that these tests check if a function
computes the same values as the sample expressions. Thus, these tests are unable to
detect errors in the sample expressions. Simply stated, the sample expressions using
concrete values evaluate to the same erroneous result as using student2string.
When abstracting over sample expressions, a logical error in the development of
the expressions is test-wise consistent with the use of a function created by the
abstraction. On the other hand, the tests using sample values break away from testing
that functions and expressions evaluate to the same value. Instead, they test that
functions evaluate to the right value. This explains why only the tests using sample
values detect the error. This exercise highlights the importance of always writing
tests using sample values.
To answer the third question, a good rule of thumb is to always start debugging
refinements with the auxiliary functions. An auxiliary function bug that manifests
itself in the tests for the function usually also manifests itself in the tests for functions
that use said auxiliary function. Therefore, fixing the auxiliary function may fix the
bugs in the functions that call the auxiliary function.
Using this principle, we start the debugging process with gpa->string. The
problem is that this function needs to return a string representing an inexact number,
not an exact number. Therefore, a function to transform an exact number to an inexact
number is needed. Exploring the BSL page in the Help Desk reveals the following
function:
number → number
Purpose: Converts an exact number to an inexact one.
(define (exact->inexact x) ...)
182 8 Defining Structures

This function looks like what is needed. The refined gpa->string is:
;; GPA → string
;; Purpose: Transform the given GPA to a string
(define (gpa->string a-gpa)
(number->string (exact->inexact a-gpa)))
Running the program reveals that all the tests pass. This completes the design of
student2string.

** Ex. 84 — Design a program to compute the distance between two points in

a 3D Cartesian coordinate system. A 3D Cartesian coordinate system has three
axes: x, y, and z. A point is an ordered triple, (x, y, z), with a coordinate on
each axis.

* Ex. 85 — 17 An alternate solution to fix gpa->string is to multiply the

given GPA and, #i1.0, inexact 1. The product is an inexact number. Redesign
gpa->string to use this approach.

** Ex. 86 — A DNA sequence consists of four nucleotide bases: adenine, cy-

tosine, guanine, thymine. These bases are denoted, respectively, A, C, G, and T.
For example, AATC is a DNA sequence of length 4. The complement of a DNA
sequence, S, is S reversed with As and Ts switched and Cs and Gs switched. For
example, the complement DNA for AATC is GATT. A DNA sequence of length
n is represented by n bases and a symbol for the species the DNA belongs to.
Design a program to compute the reverse of a DNA sequence of length 4.

*** Ex. 87 — All cars have five characteristics: a brand, a model, a color, a
manufacturing year, and a maximum speed in miles per hour. Design and write
a program to convert a car to a descriptive string. For example, a car with the
following characteristics:
Brand: Alpha Romeo
Model: Giulia
Color: Rosso competizione
Year: 2020
Speed: 160
is converted to:
The rosso competizione 2020 Alpha Romeo Giulia has a max speed
of 160 mph.

* Ex. 88 — The cost of running a business is described by fixed costs and by

variable costs. The total cost of running a business is given by the sum of fixed
and variable costs. Design a function that takes a single input for costs and that
returns the total cost.

17
This exercise is due to a discussion with Robby Findler.
45 What Have We Learned in This Chapter? 183

Fig. 62 Chapter 8 BSL grammar

program ::= {expr | test | def | defs}*
def ::= (define <variable name> expr)
::= (define (<name> <name>+) expr)
test ::= (check-expect expr expr)
::= (check-within expr expr expr)
defs ::= (define-struct <struct name> (<field name>*))
expr ::= number
::= string
::= \#<character constant>
::= <character>+
::= #true
::= #false
::= <image>
::= (<function> expr+ )
::= (and expr expr expr * )
::= (or expr expr expr* )
::= (cond [expr expr] +
[else expr])
::= (if expr expr expr)

45 What Have We Learned in This Chapter?

The important lessons of Chap. 8 are summarized as follows:

• Programmers may define structures.
• The explored BSL syntax is displayed in Fig. 62 and includes the required syntax
to define structures.
• Every structure definition provides a programmer with a constructor and a selector
for each field.
• Structures in BSL are generic and require a data definition to understand its correct
use and its correct processing.
• Structures in BSL are first class values.
• Internal representation of data inside a programming language (like BSL) some-
times needs to be transformed to make it understandable to people in the real
world.
• A common technique to make the internal representation of data understandable
to people in the real world is transforming it to a string.
• As a design advances it may become necessary to refine previously created data
definitions.
• Abstracting over sample expressions, may embed logical errors in the function
defined and these errors may not be detected by tests using sample computations.
• Writing tests using sample values is an effective way to detect logical errors not
caught by tests using sample computations.
Chapter 9
Aliens Attack Version 2

The ability to define and solve problems using structures endows us with the power
to refine Aliens Attack version 1 (from Chap. 6). This refinement adds an alien to
the game. The following is a sample scene for Aliens Attack version 2:

As in Aliens Attack version 1, there is a rocket. In addition, there is an alien. Two

things become immediately obvious. First, the world in this version of the game is
different. It contains, at least, a rocket and an alien. This means a refined world data
definition is needed. Second, a data definition for an alien is needed. These data
definitions are necessary to design a new program for Aliens Attack.
Before proceeding with our design efforts, it is important to note that a refinement
does not mean that we start writing code from scratch. On the contrary, well-designed
code with separation of concerns helps us avoid reinventing the wheel. That is, an
important reason to develop well-designed code is being able to reuse existing code
in a refinement. Inevitably, some functions will change and some new functions will

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 185
M. T. Morazán, Animated Problem Solving, Texts in Computer Science,
https://doi.org/10.1007/978-3-030-85091-3_9
186 9 Aliens Attack Version 2

need to be written. Some functions do not need to be refined and are used unchanged
in this new version of the game.
The plan is to follow the steps of the design recipe for video games in Fig. 56.
For every handler, a top-down design approach is followed. As the design advances,
functions that process instances of a refined data definition are updated. New auxiliary
functions are developed as their need arises.

46 Data Definitions

Sometimes it is easier to start defining the simplest data first. In this case, this means
starting with the data definition of an alien. To define an alien think about what may
change from one instance on an alien to another. An alien moves as follows:

Think carefully about what changes when the alien is moved right, left, or down.
Clearly, the image of the alien does not change. As we know, the image of the alien
is a constant. Therefore, an alien is not and does not contain an image. What changes
when the alien moves right? If you think about it for a while, it becomes clear that
the x coordinate of the alien changes. Given that this is an x coordinate in an Aliens
Attack scene, it has to be an image-x (defined in Fig. 4). What changes when the
alien moves left? Clearly, it is an image-x again. What changes when the alien
moves down? The alien’s image-x does not change. Instead, its image-y changes.
There is nothing more that changes about the alien. Therefore, we now know that an
alien has two characteristics that change: its image-x and its image-y. Now, think
about how an alien may be represented. Using a posn to represent an alien seems to
be a natural fit. Therefore, we may define an alien as follows:
#| An alien is a posn: (make-posn image-x image-y). |#
Next think about what changes in the world as the game evolves. Just as before,
the rocket may change. It is also fairly straightforward to see that the alien changes.
Does anything else change? Think about this carefully for a moment. Although not
rendered in a game scene, the direction the alien travels in changes. Sometimes the
alien is moving right. Sometimes it is moving left. Sometimes it is moving down.
Given that nothing else changes in the world, we have that the world has three
characteristics that may vary as the game evolves: the rocket, the alien, and the
direction that the alien travels in. It now becomes clear that a data definition for a
direction and a new data definition for a world are needed.
47 Function Templates and Sample Instances 187

Again, start by defining the simplest data. Given that a direction can only be one
of three values, we may define it as an enumerated type:
#| A direction (dir) is either:
1. 'right
2. 'left
3. 'down |#
This data definition is clearly stating that a dir may only be one of three symbols
at any point during the game’s evolution. Making a direction anything else is an
improper use of a direction.
We now turn our attention to defining a world. As noted above, a world has
three characteristics that may vary. Three is a finite number greater than one. This
immediately suggests using a structure with three fields. This means that the world
is no longer a rocket. Instead, the world is defined as follows:
#| A world is a structure: (make-world rocket alien dir). |#
This data definition clearly states that a world is composed of three fields such
that the first fields is a rocket, the second field is an alien, and the third field is a
dir. If any of these fields is made something different, then it is an improper use of
a world structure. Given that we have a data definition for a new type of structure,
the program must contain a structure definition for it. The world structure is defined
as follows:
(define-struct world (rocket alien dir))
Observe that the name of the structure and the name of each field is suggestive of
the data type they represent.

47 Function Templates and Sample Instances

There are two new data definitions and one refined data definition. This means we
need two new function templates and need to refine one function template. The
template for functions on an alien and instances of alien are:
#| ;; alien . . . → . . .
;; Purpose: . . .
(define (f-on-alien an-alien . . .)
. . .(f-on-image-x (posn-x an-alien). . .)
. . .(f-on-image-y (posn-y an-alien). . .))

;; Sample instances
(define ALIEN1 (make-posn . . . . . .))

;; Sample expressions for f-on-alien

(define ALIEN-VAL1 . . .)
...
188 9 Aliens Attack Version 2

;; Tests using sample computations for f-on-alien

(check-expect (f-on-alien . . .) ALIEN-VAL1)
...

;; Tests using sample values for f-on-alien

(check-expect (f-on-alien . . .) . . .)
... |#

;; Sample instances of alien

(define INIT-ALIEN (make-posn AN-IMG-X 0))
(define INIT-ALIEN2 (make-posn 3 MAX-IMG-Y))
Given that an alien is a posn containing an image-x and an image-y, f-on-alien
has function calls to process this type of data. Once again, take note that these
function calls make explicit the separation of concerns. Note that there is no variety
in alien and, therefore, the template suggests only one sample instance, one sample
expression, one test using sample computations, and one test using sample values.
The comments above each of these, however, remain plural to suggest thoroughness
and develop more than one for each. The first alien sample instance defines the
initial alien to be at the top of the scene with 0 as its y coordinate and in the
middle with AN-IMG-X as its x coordinate (recall that AN-IMG-X is defined as
(/ MAX-CHARS-HORIZONTAL 2)). The second alien sample instance defines an
alien at the bottom of the scene with MAX-IMG-Y as its y coordinate and toward the
left half with 3 and its x coordinate. Observe that both x values and both y values,
respectively, are valid instances of image-x and image-y. Therefore, both sample
alien instances are valid.
The template for functions on a dir and sample instances of dir are:
#| ;; dir . . . → . . .
;; Purpose: . . .
(define (f-on-dir a-dir . . .)
(cond [(eq? a-dir 'right) . . .]
[(eq? a-dir 'left) . . .]
[else . . .]))

;; Sample instances of dir

(define DIR1 . . .)
(define DIR2 . . .)
(define DIR3 . . .)

;; Sample expressions for f-on-dir

(define DIR-VAL1 . . .)
(define DIR-VAL2 . . .)
(define DIR-VAL3 . . .)
...
47 Function Templates and Sample Instances 189

;; Tests using sample computations for f-on-dir

(check-expect (f-on-dir . . .) DIR-VAL1)
(check-expect (f-on-dir . . .) DIR-VAL2)
(check-expect (f-on-dir . . .) DIR-VAL3)
...

;; Tests using sample values for f-on-dir

(check-expect (f-on-dir . . .) . . .)
... |#

;; Sample instances of dir

(define INIT-DIR 'right)
(define INIT-DIR2 'left)
(define INIT-DIR3 'down)
Observe that f-on-dir has a conditional expression in its body. This conditional
distinguishes among the 3 varieties of dir. Given that there are 3 varieties of dir
the template suggests defining 3 sample instances. The same is true for the number
of sample expressions and tests using sample computations: at least one for each
variety. It is straightforward to define sample instances for dir given that there are
only 3 varieties: one instance for each.
The refined template for functions on a world and sample instances of world
are:
;; world . . . → . . .
;; Purpose: . . .
(define (f-on-world a-world . . .)
. . .(f-on-rocket (world-rocket a-world) . . .)
. . .(f-on-alien (world-alien a-world) . . .)
. . .(f-on-dir (world-dir a-world) . . .))

;; Sample expressions for f-on-world

(define WORLD-VAL . . .)

;; Tests using sample computations for f-on-world

(check-expect (f-on-world . . .) WORLD-VAL)
...

;; Tests using sample values for f-on-world

(check-expect (f-on-world . . .) . . .)

;; Sample instances of world

(define INIT-WORLD (make-world INIT-ROCKET
INIT-ALIEN
INIT-DIR))
190 9 Aliens Attack Version 2

(define INIT-WORLD2 (make-world INIT-ROCKET2

INIT-ALIEN2
INIT-DIR2))
Observe the separation of concerns suggested by f-on-world a-world. It suggests
that solving a problem about a world may involve solving a problem about a rocket,
an alien, or a dir. This separation of concerns also leads to a divide and conquer
approach to problem solving with worlds. Problems for a rocket, an alien, and a dir
may be solved independently and their solutions combined to, for example, create a
new world. As there is not variety in world, the template suggests a minimum of one
sample expression, one test using sample computations, and one tests using sample
values. Observe that the sample world instances use the samples rocket, alien,
and dir instances previously defined. The number of world instances have been
expanded from those in Chap. 6 to easily write tests.

48 The run Function

We start code refinement with the run function because it informs us of the world-
processing handlers needed. This means we are following a top-down approach to
problem solving. It is necessary to determine what functions are needed in this new
version of Aliens Attack with its new world data definition. In Chap. 6, the run is
defined as follows:
; string → world
; Purpose: To run the game
(define (run a-name)
(big-bang INIT-WORLD
[on-draw draw-world]
[name a-name]
[on-key process-key]))
Ask yourself if draw-world and process-key are still needed. Ask yourself is
any other handlers are needed. Given that the world must still be rendered to the
screen and that the player must still be able to move the rocket, it is clear that both
draw-world and process-key are still needed. These functions, however, need to
be refined because they process a world whose data definition has been refined.
To determine if other handlers are needed it is necessary to think about how the
game ought to work. The world’s new element is the alien. what is the expected
behavior of the alien in the game? After some thought, it becomes clear that the
player does not control the movements of the alien. Instead, the alien must move
every time the clock ticks. This means that a clock tick handler is needed. Further
thought reveals that it is possible for this version of the game to end. It ends when the
alien reaches earth. This means that a handler to detect the end of the game is also
needed. We must associate these two new handlers with the game in the big-bang
expression. This leads to this refined version of the run function:
49 Drawing the World 191

(define TICK-RATE 1/4)

; string → world
; Purpose: To run the game
(define (run a-name)
(big-bang INIT-WORLD
[on-draw draw-world]
[name a-name]
[on-key process-key]
[on-tick process-tick TICK-RATE]
[stop-when game-over?]))
The handler to process clock ticks is process-tick and the handler to detect the
end of the game is game-over?. Observe that an optional argument is given to
on-tick. By limiting the number of clock ticks per second the speed at which the
alien moves is controlled. In this example, the clock ticks 4 times per second or,
equivalently, every 14 seconds. You may, of course, adjust the value of TICK-RATE
to your liking to make the alien move faster or slower.
The problems that we must solve are now clear. We must refine draw-world and
process-key and we must design process-tick and game-over?. In addition, it
may also be necessary to write new auxiliary functions.

49 Drawing the World

To start, we tackle the problem of drawing the world. It is almost certain that you have
already realized that a function to draw an alien is needed. This is good insight.
You should not, however, immediately jump to write such a function. Instead, be
disciplined about designing your solution to drawing the world. Follow a top-down
design strategy and start with the handler to draw the world. Allow the design process
to reveal all the auxiliary functions that are needed and design them when the need
arises.

49.1 The draw-world Refinement

To update draw-world think about which characteristics of the new definition of

world need to be rendered in the game’s scene. Like before, the rocket must be
rendered. In addition, the alien must be rendered. The direction, on the other hand,
is not rendered. Think about how to render the alien with what already exists in the
game’s code (from Chap. 6). A scene with the rocket is rendered by the current code.
The draw-world refinement needs to render the alien in the scene that contains the
rocket. This suggests updating the sample expressions for draw-world as follows:
192 9 Aliens Attack Version 2

;; Sample expressions for draw-world

(define WORLD-SCN1 (draw-alien
(world-alien INIT-WORLD)
(draw-rocket (world-rocket INIT-WORLD)
E-SCENE)))

(define WORLD-SCN2 (draw-alien

(world-alien INIT-WORLD2)
(draw-rocket (world-rocket INIT-WORLD2)
E-SCENE)))
Based on these sample expressions the definition of draw-world is refined to be:
;; world → scene
;; Purpose: To draw the world in E-SCENE
(define (draw-world a-world)
(draw-alien (world-alien a-world)
(draw-rocket (world-rocket a-world) E-SCENE)))
Observe that function composition is used to draw the alien in the scene that contains
the rocket. Further observe that a divide and conquer solution has been developed.
The problem of drawing a world has been broken into two smaller problems: drawing
an alien and drawing a rocket.
Let us carefully analyze the expressions in this function.
(draw-rocket (world-rocket a-world) E-SCENE)
evaluates to a scene. Specifically, the scene that renders the rocket in the empty scene.
This scene is input to (a new) auxiliary function to draw the alien. This expression:
(draw-alien (world-alien a-world)
(draw-rocket (world-rocket a-world) E-SCENE))
renders the alien in the scene that contains the rocket.
The tests using sample computations for draw-world remain unchanged because
they only refer to constants. Referring to constants is not affected by the changes
made to world. Thus, these tests remain as follows:
;; Tests using sample computations for draw-world
(check-expect (draw-world INIT-WORLD) WORLD-SCN1)
(check-expect (draw-world INIT-WORLD2) WORLD-SCN2)
The tests using sample values must now explicitly construct world values containing
a rocket, an alien, and a dir. Different world sample values may be constructed,
for example, by using different combinations of the defined sample instances for a
rocket, an alien, and a dir. For instance, these refined tests may be:
49 Drawing the World 193

(check-expect
(draw-world (make-world INIT-ROCKET2 INIT-ALIEN INIT-DIR3))

)
(check-expect
(draw-world (make-world INIT-ROCKET INIT-ALIEN2 INIT-DIR2))

49.2 Drawing Aliens

The design steps reveal the need for draw-alien. This function processes an alien.
Thus, we can specialize the template for functions on an alien to implement it. Think
about how to draw the alien. You have two inputs: an alien and a scene. To draw the
image of the alien character we may use draw-ci (just as is done to draw the rocket’s
character image). The image coordinates are obtained from the alien to draw. This
suggests the following initial template specialization:
;; alien scene → scene
;; Purpose: Draw the given alien in the given scene
(define (draw-alien an-alien scn)
. . .(f-on-image-x (posn-x an-alien). . .)
. . .(f-on-image-y (posn-y an-alien). . .))

;; Sample expressions for draw-alien

(define ALIEN-VAL1 (draw-ci ALIEN-IMG
(posn-x INIT-ALIEN)
(posn-y INIT-ALIEN)
E-SCENE))
194 9 Aliens Attack Version 2

(define ALIEN-VAL2 (draw-ci ALIEN-IMG

(posn-x INIT-ALIEN2)
(posn-y INIT-ALIEN2)
E-SCENE2))

;; Tests using sample computations for draw-alien

(check-expect (draw-alien . . . . . .) ALIEN-VAL1)
(check-expect (draw-alien . . . . . .) ALIEN-VAL2)

;; Tests using sample values for draw-alien

(check-expect (draw-alien . . . . . .) . . . )
The sample expressions do not process the alien’s image-x and image-y coordi-
nates. Therefore, there are no calls to functions that process these data types. The
only differences in the sample expressions are the alien and the scene. Therefore,
draw-alien does not need more parameters. For the body of draw-alien the dif-
ferences are substituted with the proper parameter. To write the tests using sample
computations the differences for each expression are used. Finally, a test using a
sample value is written to guard against bugs in our sample expressions. This leads
to the following code to draw an alien:
;; alien scene → scene
;; Purpose: Draw the given alien in the given scene
(define (draw-alien an-alien scn)
(draw-ci ALIEN-IMG
(posn-x an-alien)
(posn-y an-alien)
scn))

;; Sample expressions for draw-alien

(define ALIEN-VAL1 (draw-ci ALIEN-IMG
(posn-x INIT-ALIEN)
(posn-y INIT-ALIEN)
E-SCENE))

(define ALIEN-VAL2 (draw-ci ALIEN-IMG

(posn-x INIT-ALIEN2)
(posn-y INIT-ALIEN2)
E-SCENE2))

;; Tests using sample computations for draw-alien

(check-expect (draw-alien INIT-ALIEN E-SCENE) ALIEN-VAL1)
(check-expect (draw-alien INIT-ALIEN2 E-SCENE2) ALIEN-VAL2)

;; Tests using sample values for draw-alien

(check-expect (draw-alien INIT-ALIEN2 E-SCENE)
50 The process-key Refinement 195

)
There are no new auxiliary functions needed. Therefore, the required updates for
draw-world are done.

50 The process-key Refinement

The player does not control or affect the aliens nor the direction the aliens travel in.
Therefore, as before, key events only affect the rocket. This means that process-key
does not have to process an alien nor a dir. The refinement of process-key only
requires refining the worlds created in the sample expressions, in the function
definitions, and in the tests. Start with the sample expressions. These must now
evaluate to a world structure. Use the defined sample world instances to refine
them as follows:
;; Sample expressions for process-key
(define KEY-RVAL (make-world
(move-rckt-right (world-rocket INIT-WORLD))
(world-alien INIT-WORLD)
(world-dir INIT-WORLD)))
(define KEY-LVAL (make-world
(move-rckt-left (world-rocket INIT-WORLD))
(world-alien INIT-WORLD)
(world-dir INIT-WORLD)))
(define KEY-OVAL INIT-WORLD2)
Observe that in the sample expressions the alien and the dir are left unchanged in
the new world created. As expected, only the rocket may be changed. The rocket,
however, is not always changed and this is illustrated by the third test. The rocket
is not moved when, for example, m is pressed by the player. Therefore, a new world
is not created and an unchanged existing world is used to define the third constant.
Based on these new sample expressions, it is straightforward to refine the function
definition. The process-key body now creates worlds as follows:
;; world key → world
;; Purpose: Process a key event to return next world
(define (process-key a-world a-key)
(cond [(key=? a-key "right")
196 9 Aliens Attack Version 2

(make-world
(move-rckt-right (world-rocket a-world))
(world-alien a-world)
(world-dir a-world))]
[(key=? a-key "left")
(make-world
(move-rckt-left (world-rocket a-world))
(world-alien a-world)
(world-dir a-world))]
[else a-world]))
The tests using sample computations, as before, remain unchanged given that they
only reference constants:
;; Tests using sample computations for process-key
(check-expect (process-key INIT-WORLD "right") KEY-RVAL)
(check-expect (process-key INIT-WORLD "left") KEY-LVAL)
(check-expect (process-key INIT-WORLD2 "m") KEY-OVAL)
The tests using sample values are updated to use world structures. These tests
illustrate that the rocket may not move off the scene (the first two tests) and that keys
other than "left" and "right" beyond "m" leave the input world unchanged (the
last two tests). The updated tests are:
;; Tests using sample values for process-key
(check-expect (process-key
(make-world (sub1 MAX-CHARS-HORIZONTAL)
INIT-ALIEN
'right)
"right")
(make-world (sub1 MAX-CHARS-HORIZONTAL)
INIT-ALIEN
'right))
(check-expect (process-key (make-world 0
INIT-ALIEN
'left)
"left")
(make-world 0
INIT-ALIEN
'left))
(check-expect (process-key (make-world 0
INIT-ALIEN
'left)
"o")
(make-world 0
INIT-ALIEN
'left))
(check-expect (process-key INIT-WORLD2 ";") INIT-WORLD2)
51 Processing Ticks 197

The refinement process does not reveal the need for new auxiliary functions.
Therefore, the refinements for process-key are complete.

51 Processing Ticks

To design the process-tick handler think carefully about how the game must
evolve every time the clock ticks. What changes? What does not change? The rocket
is not affected by clock ticks given that the rocket only moves on keystrokes made
by the player. The alien, on the other hand, must move every clock tick. This means
that we need to design a function to move an alien.
A bit more subtle is the fact that when the alien moves the direction the new alien
must move in may be different. This means that a function to compute the direction
for the new alien is needed. This function must distinguish when not to change the
direction and between the cases to change the direction. Careful analysis reveals that
the direction only changes when the alien is at either the left or right edge of the
scene. We can specify how the direction changes as follows:
• New alien at right edge created by moving right means the new direction is down
• New alien at left edge created by moving left means the new direction is down
• New alien at right edge created by moving down means the new direction is left
• New alien at left edge created by moving down means the new direction is right
• Otherwise, the direction does not change.
This analysis suggests that to compute the direction of a new alien two pieces of data
are required: the new alien and the direction used to create the new alien.

51.1 The process-tick Handler

Based on the above problem analysis, the specialization of the template for a function
on a world to design process-click is started. The first step is to specialize based
on the analysis so far:
;; world → world
;; Purpose: Create a new world after a clock tick
(define (process-tick a-world)
. . .(f-on-rocket (world-rocket a-world) . . .)
. . .(f-on-alien (world-alien a-world) . . .)
. . .(f-on-dir (world-dir a-world) . . .))

;; Sample expressions for process-tick

(define AFTER-TICK-WORLD1
(make-world (world-rocket INIT-WORLD)
(move-alien (world-alien INIT-WORLD)
(world-dir INIT-WORLD))
198 9 Aliens Attack Version 2

(new-dir-after-tick
(move-alien (world-alien INIT-WORLD)
(world-dir INIT-WORLD))
(world-dir INIT-WORLD))))

(define AFTER-TICK-WORLD2
(make-world (world-rocket INIT-WORLD2)
(move-alien (world-alien INIT-WORLD2)
(world-dir INIT-WORLD2))
(new-dir-after-tick
(move-alien (world-alien INIT-WORLD2)
(world-dir INIT-WORLD2))
(world-dir INIT-WORLD2))))

;; Tests using sample computations for process-tick

(check-expect (process-tick INIT-WORLD) AFTER-TICK-WORLD1)
(check-expect (process-tick INIT-WORLD2) AFTER-TICK-WORLD2)

;; Tests using sample values for process-tick

(check-expect (process-tick (make-world INIT-ROCKET
(make-posn 1 5)
'left))
(make-world INIT-ROCKET
(make-posn MIN-IMG-X 5)
'down))

(check-expect (process-tick
(make-world
INIT-ROCKET2
(make-posn (- MAX-CHARS-HORIZONTAL 2) 10)
'right))
(make-world INIT-ROCKET2
(make-posn MAX-IMG-X 10)
'down))

(check-expect (process-tick (make-world INIT-ROCKET2

(make-posn MAX-IMG-X 2)
'down))
(make-world INIT-ROCKET2
(make-posn MAX-IMG-X 3)
'left))

(check-expect (process-tick (make-world INIT-ROCKET2

(make-posn MIN-IMG-X 2)
'down))
51 Processing Ticks 199

(make-world INIT-ROCKET2
(make-posn MIN-IMG-X 3)
'right))
The sample expressions are written using (to be designed) functions to move an
alien and to compute a new direction. The arguments to move an alien are fairly
straightforward to determine: the alien to move and the direction to move. If more
arguments are needed the design of move-alien will reveal them. The arguments
for new-dir-after-tick are based on the problem analysis above. The design of
new-dir-after-tick will reveal if more arguments are needed. The tests using
sample computations are written, as expected, using previously defined constants.
In this case, these tests illustrate when the alien does not change direction. The tests
using sample values are written building world instances. These tests illustrate that
the changes in direction occur as expected when an alien is at either edge of the
scene.
The next task is to specialize the body for process-tick. As done before, this
is achieved by abstracting over the sample expressions. This yields the following
function:
;; world → world
;; Purpose: Create a new world after a clock tick
(define (process-tick a-world)
(make-world
(world-rocket a-world)
(move-alien (world-alien a-world) (world-dir a-world))
(new-dir-after-tick (move-alien (world-alien a-world)
(world-dir a-world))
(world-dir a-world))))

51.2 The Design of new-dir-after-tick

The design of new-dir-after-tick is interesting because it takes as input two

different types of data. Should this function be designed by specializing the template
for a function on an alien or a function on a direction? Problem analysis may shed
some light on this question.
Every time the clock ticks the direction of the alien may change. When does it
change? When does it not change? We can identify the following three conditions:
1. If the direction is right, then the new direction may be left or down.
2. If the direction is left, then the new direction may be down or left.
3. If the direction is down, then the new direction may be right or left.
This analysis suggests that the function ought to be designed around the given
direction. Therefore, the template for a function on a direction is specialized.
200 9 Aliens Attack Version 2

Given that testing aliens on the left and right edges is needed, the following
instances of aliens are defined to facilitate the writing of sample expressions and
tests:
(define LEFT-EDGE-ALIEN (make-posn MIN-IMG-X 10))
(define RIGHT-EDGE-ALIEN (make-posn MAX-IMG-X 6))
Observe that there are two separate concerns. The first is to determine the given
direction instance. Second, is to compute the new direction. The computation of the
new direction ought to be split among different auxiliary functions depending on the
given direction:
1. new-dir-after-down is used when the given direction is down.
2. new-dir-after-left is used when the given direction is left.
3. new-dir-after-right is used when the given direction is right.
Based on this design idea we can specialize the template for a function on a direction.
Sample expressions may be written as follows:
;; Sample expressions for new-dir-after-tick
(define NEW-DIR-LEDGE-ALIEN-DOWN
(new-dir-after-down LEFT-EDGE-ALIEN))
(define NEW-DIR-REDGE-ALIEN-DOWN
(new-dir-after-down RIGHT-EDGE-ALIEN))
(define NEW-DIR-INIT-ALIEN-LEFT
(new-dir-after-left INIT-ALIEN))
(define NEW-DIR-LEDGE-ALIEN-LEFT
(new-dir-after-left LEFT-EDGE-ALIEN))
(define NEW-DIR-INIT-ALIEN-RIGHT
(new-dir-after-right INIT-ALIEN))
(define NEW-DIR-REDGE-ALIEN-RIGHT
(new-dir-after-right RIGHT-EDGE-ALIEN))
Observe that there are two sample expressions for each auxiliary function. This is
due to the fact that for each possible direction the new direction may be one of two
values. For example, if the given direction is down, then the new direction may be
right if the given alien is at the left edge and may be left if the given alien is at the
right edge. Similarly, if the given direction is left, then the new direction is down
if the given alien is at the left edge and remains left otherwise. Finally, if the given
direction is right, then the new direction is down if the given alien is at the right
edge and remains right otherwise. In summary, there is a sample expression for each
possible outcome for a given direction.
Based on the conditions and the sample expressions from above the definition
template and tests are specialized as follows:
;; alien dir → dir
;; Purpose: Return new alien direction
(define (new-dir-after-tick an-alien old-dir)
(cond [(eq? old-dir 'right)
(new-dir-after-right an-alien)]
51 Processing Ticks 201

[(eq? old-dir 'left)

(new-dir-after-left an-alien)]
[else (new-dir-after-down an-alien)])

;; Tests using sample computations for new-dir-after-tick

(check-expect (new-dir-after-tick LEFT-EDGE-ALIEN 'down)
NEW-DIR-LEDGE-ALIEN-DOWN)
(check-expect (new-dir-after-tick RIGHT-EDGE-ALIEN 'down)
NEW-DIR-REDGE-ALIEN-DOWN)
(check-expect (new-dir-after-tick INIT-ALIEN 'left)
NEW-DIR-INIT-ALIEN-LEFT)
(check-expect (new-dir-after-tick LEFT-EDGE-ALIEN 'left)
NEW-DIR-LEDGE-ALIEN-LEFT)
(check-expect (new-dir-after-tick INIT-ALIEN 'right)
NEW-DIR-INIT-ALIEN-RIGHT)
(check-expect (new-dir-after-tick RIGHT-EDGE-ALIEN 'right)
NEW-DIR-REDGE-ALIEN-RIGHT)

;; Tests using sample values for new-dir-after-tick

(check-expect (new-dir-after-tick (make-posn MIN-IMG-X 10)
'down)
'right)
(check-expect (new-dir-after-tick (make-posn MAX-IMG-X 12)
'down)
'left)
(check-expect (new-dir-after-tick (make-posn 10 10)
'left)
'left)
(check-expect (new-dir-after-tick (make-posn MIN-IMG-X 15)
'left)
'down)
(check-expect (new-dir-after-tick (make-posn 10 14)
'right)
'right)
(check-expect (new-dir-after-tick (make-posn MAX-IMG-X 3)
'right)
'down)
The specialization of the definition template uses the three conditions identified
above to call the appropriate auxiliary function. The tests using sample computations
illustrate that new-dir-after-tick computes the values of the sample expressions.
The tests using sample values further illustrate the possible outcomes for each
possible direction instance.
202 9 Aliens Attack Version 2

51.3 The Design of Auxiliary Functions for

new-dir-after-tick

The design of new-dir-after-tickrevealed the need for three auxiliary functions.

Our next task is to design these functions and any further auxiliary functions that are
needed. Let us start by designing new-dir-after-down, new-dir-after-left,
and new-dir-after-right. For each of these functions the returned direction
depends on whether or not the given alien is at one of the edges.

51.3.1 Design of new-dir-after-down

Carefully consider what the next direction ought to be if the previous direction is
down. When the previous direction is down we know that the given alien must be at
one of the edges. If the given alien is at the left edge, then the new direction is right.
If the given alien is at the right edge, then the new direction is left. Observe that
this function must determine which edge the alien is at. This is a different problem
from computing the new direction. In the spirit of separation of concerns, detecting
if an alien is at the left or the right edge is delegated to an auxiliary function. Further
observe that this means that new-dir-after-down does not process the given alien
and, therefore, the design of this function does not specialize the template for a
function on an alien. Instead, it is designed around the decision that must be made
based on the edge the alien is located at.
We are free to choose if the function detects the alien at the left edge or at the
right edge because if the detection for an edge fails, then the alien must be at the
other edge. We arbitrarily choose to detect if the alien is at the left edge. The function
using this strategy looks as follows:
;; alien → direction
;; Purpose: Compute the direction of the given alien
;; when previous direction is down
(define (new-dir-after-down an-alien)
(if (alien-at-left-edge? an-alien)
'right
'left))

;; Sample expressions for new-dir-after-down

(define AT-LEDGE-DOWN 'right)
(define AT-REDGE-DOWN 'left)

;; Tests using sample computations for new-dir-after-down

(check-expect (new-dir-after-down LEFT-EDGE-ALIEN)
AT-LEDGE-DOWN)
(check-expect (new-dir-after-down RIGHT-EDGE-ALIEN)
AT-REDGE-DOWN)
51 Processing Ticks 203

;; Tests using sample values for new-dir-after-down

(check-expect (new-dir-after-down (make-posn MIN-IMG-X 4))
'right)
(check-expect (new-dir-after-down (make-posn MAX-IMG-X 9))
'left)
The sample expressions are very simple given that once the alien’s edge is known all
that is needed is to return a direction. The function uses a conditional to decide which
direction to return. If the alien is at the left edge the returned direction is right. If the
alien is at the right edge, then the returned direction is left. The tests using sample
computations illustrate that the function computes the expected new direction. The
tests using sample values illustrate that the function works for additional arbitrary
aliens at one of the edges.

51.3.2 Design of new-dir-after-left

The design of new-dir-after-left is similar to that of new-dir-after-down.

It is designed around making a decision based on whether or not the given alien
is at the left edge. If the alien is at the left edge, then the new direction is down.
Otherwise, the direction is unchanged because the alien ought to continue moving
left. The function may be implemented as follows:
;; alien → direction
;; Purpose: Compute the direction of the given alien
;; when previous direction is left
(define (new-dir-after-left an-alien)
(if (alien-at-left-edge? an-alien)
'down
'left))

;; Sample expressions for new-dir-after-left

(define AT-LEDGE 'down)
(define NOT-AT-LEDGE 'left)

;; Tests using sample computations for new-dir-after-left

(check-expect (new-dir-after-left LEFT-EDGE-ALIEN) 'down)
(check-expect (new-dir-after-left INIT-ALIEN) 'left)

;; Tests using sample values for new-dir-after-left

(check-expect (new-dir-after-left RIGHT-EDGE-ALIEN) 'left)
There are two sample expressions: one for each possible outcome. The tests using
sample computations use an alien at the left edge and an alien that is not at the
left edge to illustrate that the function properly computes the value of the sample
204 9 Aliens Attack Version 2

expressions. The test using a sample value illustrates that the function works for an
extrema value: an alien at the right edge.

51.3.3 Design of new-dir-after-right

The design of new-dir-after-right is similar to that of new-dir-after-left.

It is arbitrarily designed around making a decision based on whether or not the given
alien is at the right edge. If the alien is at the right edge, then the new direction is
down. Otherwise, the direction is unchanged because the alien ought to continue
moving right. The function may be implemented as follows:
;; alien → direction
;; Purpose: Compute the direction of the given alien
;; when previous direction is right
(define (new-dir-after-right an-alien)
(if (alien-at-right-edge? an-alien)
'down
'right))

;; Sample expressions for new-dir-after-right

(define AT-REDGE 'down)
(define NOT-AT-REDGE 'right)

;; Tests using sample computations for new-dir-after-right

(check-expect (new-dir-after-right RIGHT-EDGE-ALIEN)
AT-REDGE)
(check-expect (new-dir-after-right INIT-ALIEN)
NOT-AT-REDGE)

;; Tests using sample values for new-dir-after-right

(check-expect (new-dir-after-right LEFT-EDGE-ALIEN)
'right)
Once again, we see that there are two sample expressions: one for each possible
outcome. The tests using sample computations use an alien at the right edge and an
alien that is not at the right edge to illustrate that the function properly computes the
value of the sample expressions. The test using a sample value illustrates that the
function works for an extrema value: an alien at the left edge.

51.3.4 Design of alien-at-left-edge?

The design of the three previous functions reveals the need for two more auxiliary
functions. Let us start with the design of alien-at-left-edge?. How do we know
if an alien is at the left edge? A given alien is at the left edge if its image-x coordinate
51 Processing Ticks 205

is MIN-IMG-X. This means that the given alien must be processed and, therefore,
this function is designed by specializing the template for a function on an alien. This
specialization yields:
;; alien → Boolean
;; Purpose: Determine if he given alien is at the left edge
(define (alien-at-left-edge? an-alien)
(= (posn-x an-alien) MIN-IMG-X))

;; Sample expressions for alien-at-left-edge?

(define LEDGE-VAL1 (= (posn-x INIT-ALIEN) MIN-IMG-X))
(define LEDGE-VAL2 (= (posn-x LEFT-EDGE-ALIEN) MIN-IMG-X))

;; Tests using sample computations for alien-at-left-edge?

(check-expect (alien-at-left-edge? INIT-ALIEN) LEDGE-VAL1)
(check-expect (alien-at-left-edge? LEFT-EDGE-ALIEN) LEDGE-VAL2)

;; Tests using sample values for alien-at-left-edge?

(check-expect (alien-at-left-edge? (make-posn 3 2)) #false)
(check-expect (alien-at-left-edge? (make-posn MIN-IMG-X 8))
#true)
The sample expressions illustrate that the desired Boolean is computed by testing an
alien’s image-x coordinate with MIN-IMG-X. The definition template specialization
uses a posn’s x coordinate selector to access the alien’s image-x coordinate. The
tests using sample values illustrate that the function works for aliens other than those
used in the tests using sample computations.

51.3.5 Design of alien-at-right-edge?

How do we know if an alien is at the right edge? A given alien is at the right edge
if its image-x coordinate is MAX-IMG-X. This means that the given alien must be
processed Once again, this is a function that is designed by specializing the template
for a function on an alien. This specialization yields:
;; alien → Boolean
;; Purpose: Determine if the given alien is at the
;; right edge
(define (alien-at-right-edge? an-alien)
(= (posn-x an-alien) MAX-IMG-X))

;; Sample expressions for alien-at-right-edge?

(define REDGE-VAL1 (= (posn-x INIT-ALIEN) MAX-IMG-X))
(define REDGE-VAL2 (= (posn-x (make-posn MAX-IMG-X 8))
MAX-IMG-X))
206 9 Aliens Attack Version 2

;; Tests using sample computations for alien-at-right-edge?

(check-expect (alien-at-right-edge? INIT-ALIEN)
REDGE-VAL1)
(check-expect (alien-at-right-edge? RIGHT-EDGE-ALIEN)
REDGE-VAL2)

;; Tests using sample values for alien-at-right-edge?

(check-expect (alien-at-right-edge? (make-posn 1 1)) #false)
(check-expect (alien-at-right-edge? RIGHT-EDGE-ALIEN) #true)
The sample expressions illustrate that the desired Boolean is computed by testing an
alien’s image-x coordinate with MAX-IMG-X. The definition template specialization
uses a posn’s x coordinate selector to access the alien’s image-x coordinate. The
tests using sample values illustrate that the function works for aliens other than those
used in the tests using sample computations.
Given that the last two functions do not reveal the need for more auxiliary func-
tions the design of new-dir-after-tick is complete. You may have noticed that
each of the auxiliary functions is small and easily understandable. This is a direct
consequence of respecting the separation of concerns during the design process.

* Ex. 89 — Redesign new-dir-after-down to test if the given alien is at the

right edge.

* Ex. 90 — Redesign new-dir-after-left to test if the given alien is at the

right edge.

* Ex. 91 — Redesign new-dir-after-right to test if the given alien is at

the left edge.

** Ex. 92 — Inline both alien-at-left-edge? and

alien-at-left-edge? to eliminate them as defined functions. Is this
a good idea?

51.4 The Design of move-alien

The next problem to solve is moving an alien. This is another interesting function
because we must decide how to design it. It takes two different types of data as input:
an alien and a dir. Should it be designed by specializing the template for functions
on an alien or the template for functions on a dir? To make this decision requires
further problem analysis.
Specializing the template for functions on a dir means that a new alien is created
by computing either a new image-x value or a new image-y value depending on
the given dir. A conditional is needed to determine which is computed. If the
51 Processing Ticks 207

direction is 'right or 'left, then the new alien is constructed with a new image-x
coordinate. If the direction is 'down, then the new alien is constructed with a new
image-y coordinate. In summary, for each possible direction either a new image-x
or a new image-y value must be computed.
Specializing the template for functions on a alien means that a new alien is
constructed by computing an image-x value using the given alien’s image-x coor-
dinate and the given direction and by computing an image-y value using the given
alien’s image-y coordinate and the given direction. A consequence of this design
path is that the functions to compute the new image-x and image-y coordinates
must have a conditional. The image-x value remains unchanged if the direction is
'down. The image-y value remains unchanged if the direction is either 'right or
'left.
As it turns out move-alien may be designed by specializing either template. In
such a case, you get to choose which design path to follow. Which design feels more
natural? Which design seems easier? Perhaps, specializing the template for functions
of a dir is easier. It only requires one conditional as opposed to two required by the
other design choice.
Let us choose to specialize the template for functions of a dir. This means
deciding how to compute a new alien for each variety of dir. We start by specializing
the template as follows:
;; alien dir → alien
;; Purpose: Move given alien in given direction
(define (move-alien an-alien a-dir)
(cond [(eq? a-dir 'right) . . .]
[(eq? a-dir 'left) . . .]
[else . . .]))

;; Sample expressions for move-alien

(define MALIEN-VAL1-1
(make-posn (move-right-image-x (posn-x INIT-ALIEN))
(posn-y INIT-ALIEN)))
(define MALIEN-VAL1-2
(make-posn (move-right-image-x (posn-x INIT-ALIEN2))
(posn-y INIT-ALIEN2)))
(define MALIEN-VAL2-1
(make-posn (move-left-image-x (posn-x INIT-ALIEN))
(posn-y INIT-ALIEN)))
(define MALIEN-VAL2-2
(make-posn (move-left-image-x (posn-x INIT-ALIEN2))
(posn-y INIT-ALIEN2)))
(define MALIEN-VAL3-1
(make-posn (posn-x INIT-ALIEN)
(move-down-image-y (posn-y INIT-ALIEN))))
(define MALIEN-VAL3-2
(make-posn (posn-x INIT-ALIEN2)
(move-down-image-y (posn-y INIT-ALIEN2))))
208 9 Aliens Attack Version 2

;; Tests using sample computations for move-alien

(check-expect (move-alien INIT-ALIEN 'right) MALIEN-VAL1-1)
(check-expect (move-alien INIT-ALIEN2 'right) MALIEN-VAL1-2)
(check-expect (move-alien INIT-ALIEN 'left) MALIEN-VAL2-1)
(check-expect (move-alien INIT-ALIEN2 'left) MALIEN-VAL2-2)
(check-expect (move-alien INIT-ALIEN 'down) MALIEN-VAL3-1)
(check-expect (move-alien INIT-ALIEN2 'down) MALIEN-VAL3-2)

;; Tests using sample values for move-alien

(check-expect (move-alien (make-posn MAX-IMG-X 3) 'down)
(make-posn MAX-IMG-X 4))
(check-expect (move-alien (make-posn MAX-IMG-X 3) 'left)
(make-posn (sub1 MAX-IMG-X) 3))
(check-expect (move-alien (make-posn 0 5) 'right)
(make-posn 1 5))
There are two sample expressions for each variety of dir. For each variety the only
difference is the alien value. This confirms that move-alien only needs an alien
and a dir as input. Observe that among the sample expressions for each dir the
new alien is constructed differently. Either the x coordinate is moved right, the x
coordinate is moved left, or the y coordinate is moved down. There is, as expected,
a test using sample computations for each evaluated sample expression. The tests
using sample values explicitly illustrate that a new alien is correctly computed.
The next task is to specialize the body of move-alien. Abstracting over the
sample expressions for each dir variety leads to:
;; alien dir → alien
;; Purpose: Move given alien in given direction
(define (move-alien an-alien a-dir)
(cond [(eq? a-dir 'right)
(make-posn (move-right-image-x (posn-x an-alien))
(posn-y an-alien))]
[(eq? a-dir 'left)
(make-posn (move-left-image-x (posn-x an-alien))
(posn-y an-alien))]
[else (make-posn
(posn-x an-alien)
(move-down-image-y (posn-y an-alien)))]))
52 Subtyping 209

52 Subtyping

The development of the auxiliary functions needed by move-alien is paused to

address a bug in our design. Have you picked up on it? Remember that problem solvers
must always be careful about logical mistakes. Consider carefully the following test:
(check-expect (move-alien INIT-ALIEN2 'down) MALIEN-VAL3-2)
Does this test make sense? Recall that INIT-ALIEN2’s image-y coordinate is
MAX-IMG-Y. That is the maximum image-y value. Can this alien be moved down?
Clearly, moving INIT-ALIEN2 down is impossible because the y coordinate of the
new alien would be larger than MAX-IMG-Y. We are facing a situation where the test
is suggesting that a function given a valid input value returns an invalid output value.
The problem is that the auxiliary functions should not process an image-x value
or an image-y value. Instead, they need to process a subset of the values defined by
these data types. This means that we need to define image-x and image-y subtypes.
A subtype defines a proper subset of the values of an existing type. If you have a
type A and B is a subtype of A, then any instance of B is also an instance of A. The
opposite is not true. Not every instance of A is also an instance of B. This means that
a function that processes data type A can also process data type B and that a function
that processes data type B cannot process data type A.
Consider, for example, the following data definition for nonnegative integers:
#| A nonnegative integer, nni, is an integer greater than
or equal to 0. #|

;; nni . . . → . . .
(define (f-on-nni an-nni) . . . an-nni . . .)

;; Sample instances of nni

(define NNI1 . . .)

;; Sample expression definitions for nni

(define NNI1-VAL . . . NNI1 . . .)

;; Tests using sample computations for f-on-nni

(check-expect (f-on-nni NNI1 . . .) NNI1-VAL)

;; Tests using sample values for f-on-nni

(check-expect (f-on-nni . . .) . . .)
Now consider writing a function to return the ratio of, x and y, two nnis. It is not
difficult to see that the ratio, xy , is not always defined. As you know, y cannot be 0.
This means that this ratio-computing function cannot process two nnis. Instead, it
ought to process an nni (x) and a positive integer (y). That is, we need to define an
nni subtype. This may be done as follows:
210 9 Aliens Attack Version 2

#| A positive integer, posint, is an nni greater than 0. |#

;; posint . . . → . . .
(define (f-on-posint a-posint) . . . a-posint . . .)

;; Sample instances of posint

(define POSINT1 . . .)

;; Sample expression definitions for posint

(define POSINT1-VAL . . . POSINT1 . . .)

;; Tests using sample computations for f-on-posint

(check-expect (f-on-posint POSINT1 . . .) POSINT1-VAL)

;; Tests using sample values for f-on-posint

(check-expect (f-on-nni . . .) . . .)
Observe that every posint is an nni, but not every nni, specifically 0, is a posint.
The definition of this subtype allows us to write the ratio-computing function using
the template for functions on a posint as follows:
;; nni posint → number
;; Purpose: Return x/y
(define (ratio-x-y x y) (/ x y))

;; Sample instances of posint

(define POSINT1 1)
(define POSINT2 20)
(define POSINT3 3)
(define POSINT4 77)

;; Sample expression definitions for ratio-x-y

(define POSINT1-VAL (/ POSINT1 POSINT2))
(define POSINT2-VAL (/ POSINT3 POSINT4))

;; Tests using sample computations for ratio-x-y

(check-expect (ratio-x-y POSINT1 POSINT2) POSINT1-VAL)
(check-expect (ratio-x-y POSINT3 POSINT4) POSINT2-VAL)

;; Tests using sample values for ratio-x-y

(check-expect (ratio-x-y 2 3) 2/3)
(check-expect (ratio-x-y 3 4) 3/4)
Observe that a new data definition, posint, is used in the contract. This communi-
cates to any programmer that to properly use this function y may not be 0. Using this
subtype definition (/ x y) is always defined. Any programmer using this function
must guarantee that the argument for y always satisfies the signature. Further observe
52 Subtyping 211

that instances of posint, POSINT1, and POSINT3 are used as nni arguments for x.
This can only be done because posint is a subtype of nni.
Subtypes may also be used for the image-x moving functions needed by
move-alien. For instance, the image-x values greater than MIN-IMG-X may be
defined as:
#|
An image-x>min is an image-x in [(add1 MIN-IMG-X)..MAX-IMG-X]

;; image-x>min . . . → . . .
;; Purpose: . . .
(define (f-on-image-x an-img-x>min . . .)
. . . an-img-x. . .)

;; Sample instances of image-x>min

(define IMG-X>MIN1 . . .)
...

;; Sample expressions for f-on-image-x>min

(define IMGX>MIN-VAL1 . . .)
...

;; Sample tests using sample computations for f-on-image-x>min

(check-expect (f-on-image-x>min . . .) IMGX>MIN-VAL1)
...

;; Sample tests using sample computations for f-on-image-x>min

(check-expect (f-on-image-x>min . . .) . . .) |#
Armed with this data definition and function template, for example, move-left-
image-x may be refined as follows:
;; image-x>min → image-x
;; Purpose: Move the given image-x>min left
(define (move-left-image-x an-img-x>min)
(sub1 an-img-x>min))

;; Sample expressions for move-left-image-x

(define IMGX>MIN-VALL1 (sub1 AN-IMG-X))
(define IMGX>MIN-VALL2 (sub1 MAX-IMG-X))

;; Tests using sample computations for move-left-image-x

(check-expect (move-left-image-x AN-IMG-X) IMGX>MIN-VALL1)
(check-expect (move-left-image-x MAX-IMG-X) IMGX>MIN-VALL2)

;; Tests using sample values for move-left-image-x

(check-expect (move-left-image-x 9) 8)
212 9 Aliens Attack Version 2

It is important to note that move-left-image-x always returns an image-x. That

is, the value returned may be MIN-IMG-X and may never be less than MIN-IMG-X.
Using the subtype, image-x>min, guarantees this.
The same design strategy results in the following data definitions and function
templates for image-x<max and image-y<max:
#|
An image-x<max, is an image-x in [MIN-IMG-X..(sub1 MAX-IMG-X)]

;; image-x<max . . . → . . .
;; Purpose: . . .
(define (f-on-image-x<max an-img-x<max . . .)
. . . an-img-x. . .)

;; Sample instances of image-x<max

(define IMG-X<MAX1 . . .)
...

;; Sample expressions for f-on-image-x<max

(define IMGX<MAX-VAL1 . . .)
...

;; Tests using sample computations for f-on-image-x<max

(check-expect (f-on-image-x<max . . .) IMGX<MAX-VAL1)
...

;; Tests using sample computations for f-on-image-x<max

(check-expect (f-on-image-x<max . . .) . . .)

An image-y<max is an image-y in [MIN-IMG-Y..(sub1 MAX-IMG-Y)]

;; image-y<max . . . → . . .
;; Purpose: . . .
(define (f-on-image-y<max an-img-y<max . . .)
. . .˜an-img-y<max. . .)

;; Sample instances of image-y

(define IMG-Y<MAX1 . . .)
...

;; Sample expressions for f-on-image-y<max

(define IMGY<MAX-VAL1 . . .)
...
52 Subtyping 213

;; Tests using sample computations for f-on-image-y<max

(check-expect (f-on-image-y<max . . .) IMGY<MAX-VAL1)
...

;; Tests using sample values for f-on-image-y<max

(check-expect (f-on-image-y<max . . .) . . .) |#
Using these subtypes and function templates the move-right-image-x and
move-down-image-y code is:
;; image-x<max → image-x
;; Purpose: Move the given image-x<max right
(define (move-right-image-x an-img-x<max)
(add1 an-img-x<max))

;; Sample expressions for move-right-image-x<max

(define IMGX<MAX-VALR1 (add1 MIN-IMG-X))
(define IMGX<MAX-VALR2 (add1 11))

;; Tests using sample computations for move-right-image-x

(check-expect (move-right-image-x MIN-IMG-X) IMGX<MAX-VALR1)
(check-expect (move-right-image-x 11) IMGX<MAX-VALR2)

;; Tests using sample computations for move-right-image-x

(check-expect (move-right-image-x 12) 13)

;; image-y<max → image-y
;; Purpose: To move the given image-y<max down
(define (move-down-image-y an-img-y<max) (add1 an-img-y<max))

;; Sample expressions for move-down-image-y

(define IMG-Y<MAX1 (add1 MIN-IMG-Y))
(define IMG-Y<MAX2 (add1 AN-IMG-Y))

;; Tests using sample computations for move-down-image-y

(check-expect (move-down-image-y MIN-IMG-Y) IMG-Y<MAX1)
(check-expect (move-down-image-y AN-IMG-Y) IMG-Y<MAX2)

;; Tests using sample values for move-down-image-y

(check-expect (move-down-image-y 2) 3)
There is no doubt that creating refined data definitions for subtypes feels like a
lot of extra work now. It is, however, time well spent. It makes solutions to problems
easier to understand. This is important, because it makes it easier to refine and for
others to work with your code. Think about this carefully. If you had to maintain
or refine another person’s code, would you not prefer to have code that is easier to
214 9 Aliens Attack Version 2

understand? This is a matter of ethics. We need to always strive to develop the easiest
code to understand and refine. Otherwise, ourselves and those that maintain our code
may have a harder job in the future.

** Ex. 93 — Design move-alien by specializing the template for functions

on an alien. Is this a better design? Why or why not?

** Ex. 94 — The move-alien design separates the moving on the x-axis in

two: moving right and moving left. Redesign move-alien to only use two
auxiliary functions: move-image-x and move-image-y. Is this a better design?
Why or why not?

52.1 Checking Errors

Most modern programming languages allow a programmer to define subtypes and

functions to process subtypes. There are some programming languages, however,
that do not provide programmers with these abilities. There are also cases where the
programmer is unable to guarantee correct input (e.g., when the input to a function
depends on the user of a program). What do you do in such cases? One option is to
write guarded functions. A guarded function uses a conditional expression to first
check if the input is valid. If so, it computes the result. Otherwise, it throws an error
and provides an error message.
BSL provides the necessary syntax to generate and check errors. An error always
returns a string that ought to help understand error. The syntax for generating an
error is:
(error expr)
Use error whenever an error must be thrown. The expr must evaluate to a string
and is returned as the error message. The syntax to test for an error is similar to
check-expect:
(check-error expr expr)
This test states that the first given expr throws an error that returns the second given
expr. The second given expression must evaluate to a string.
As an example, consider again the ratio-computing function. If the subtype
posint is not defined, the guarded function may look as follows:
;; nni nni → number throws error
;; Purpose: Return x/y
(define (x-out-of-y x y)
(if (> y 0)
(/ x y)
(error "x-out-of-y: y cannot be zero.")))
52 Subtyping 215

(check-expect (x-out-of-y 2 3) 2/3)

(check-expect (x-out-of-y 3 4) 3/4)
(check-error (x-out-of-y 10 0) "x-out-of-y: y cannot be zero.")
Observe that the purpose of this function is to return a number. However, this is not
always possible because division by zero is undefined. The signature reflects this by
indicating that the function can throw an error. The function’s body has a conditional
to test the value of y. Whenever y is zero an error is thrown. The third test checks
that the correct error message is returned.
The error message generated contains the name of the function. This is included
to help the programmer debug the code. It is important for error messages to be
informative without being prescriptive. That is, the error message should not suggest
how to fix the bug. Suggesting how to fix a bug is a poor approach because we have
no way of knowing what design decisions made the programmer led to the bug.
In addition to the function name, an error message may contain values that led
to the error. If values are included in an error message a mechanism to include such
information in a string is required. BSL provides the format function to generate
strings with customized information. The syntax is:
(format string expr∗ )
The given string may contain zero or more times the special character, ∼s, that is
substituted with the value of one of the expressions provided after the string. The
ith ∼s is substituted with the value of the ith expression.
Consider the following move-right-image-x implementation designed without
defining the posint subtype:
;; image-x → image-x throws error
;; Purpose: Move the given image-x right
(define (move-right-image-x an-img-x)
(if (< an-img-x MAX-IMG-X)
(add1 an-img-x)
(error (format "move-right-image-x: The character
at x=∼s cannot move right."
MAX-IMG-X))))

;; Sample expressions for move-right-image-x

(define IMGX-VALR1 (add1 MIN-IMG-X))
(define IMGX-VALR2 (add1 11))
(define IMGX-VALRE (format "move-right-image-x: The character
at x=∼s cannot move right."
MAX-IMG-X))

;; Tests using sample computations for move-right-image-x

(check-expect (move-right-image-x MIN-IMG-X) IMGX-VALR1)
(check-expect (move-right-image-x 11) IMGX-VALR2)
(check-error (move-right-image-x MAX-IMG-X) IMGX-VALRE)
216 9 Aliens Attack Version 2

;; Tests using sample values for move-right-image-x

(check-expect (move-right-image-x 12) 13)
First observe that move-right-image-x is now a guarded function that may throw
as error. The body first checks, in essence, if the given image-x is a posint.
Second, observe that error message generated includes the function name and the
image-x value that causes the error. Both are likely to be useful to a programmer
while debugging their code. Third, observe that the error message generated does
not suggest to the programmer how to fix the error. Fourth, observe that a sample
expression is added to illustrate the computation of an error message. Fifth, observe
that the tests include one error-checking test. There is a single test because there is a
single image-x value that generates an error. If there were multiple possible values
that generate an error, then there would be more than one error-checking test.
A natural question to ask is how does not defining the posint subtype affect
the design of move-alien. Even though the function move-alien itself does not
include code to raise an error it may still happen. It happens when one of its auxiliary
functions raises an error. This must be reflected in the signature for move-alien.
Otherwise, it is impossible to discern that it may throw an error. The error returned by
an auxiliary function is returned by move-alien. This means sample expressions and
tests must be more carefully designed. Sample expressions may not use values, like
INIT-ALIEN2, that generate an error. Tests must check expected error messages from
auxiliary functions using values that force throwing an error (like INIT-ALIEN2).
The function move-alien may be written as follows:
;; alien dir → alien throws error
;; Purpose: Move given alien in given direction
(define (move-alien an-alien a-dir)
(cond [(eq? a-dir 'right)
(make-posn (move-right-image-x (posn-x an-alien))
(posn-y an-alien))]
[(eq? a-dir 'left)
(make-posn (move-left-image-x (posn-x an-alien))
(posn-y an-alien))]
[else
(make-posn (posn-x an-alien)
(move-down-image-y (posn-y an-alien)))]))

;; Sample expressions for move-alien

(define MALIEN-VAL1-1
(make-posn (move-right-image-x (posn-x INIT-ALIEN))
(posn-y INIT-ALIEN)))
(define MALIEN-VAL1-2
(make-posn (move-right-image-x (posn-x INIT-ALIEN2))
(posn-y INIT-ALIEN2)))
52 Subtyping 217

(define MALIEN-VAL2-1
(make-posn (move-left-image-x (posn-x INIT-ALIEN))
(posn-y INIT-ALIEN)))
(define MALIEN-VAL2-2
(make-posn (move-left-image-x (posn-x INIT-ALIEN2))
(posn-y INIT-ALIEN2)))
(define MALIEN-VAL3-1
(make-posn (posn-x INIT-ALIEN)
(move-down-image-y (posn-y INIT-ALIEN))))
(define MALIEN-VAL3-2
(make-posn (posn-x (make-posn 1 8))
(move-down-image-y
(posn-y (make-posn 1 8)))))

;; Tests using sample computations for move-alien

(check-expect (move-alien INIT-ALIEN 'right) MALIEN-VAL1-1)
(check-expect (move-alien INIT-ALIEN2 'right) MALIEN-VAL1-2)
(check-expect (move-alien INIT-ALIEN 'left) MALIEN-VAL2-1)
(check-expect (move-alien INIT-ALIEN2 'left) MALIEN-VAL2-2)
(check-expect (move-alien INIT-ALIEN 'down) MALIEN-VAL3-1)
(check-expect (move-alien (make-posn 1 8) 'down)
MALIEN-VAL3-2)

;; Tests using sample values for move-alien

(check-expect (move-alien (make-posn MAX-IMG-X 3) 'down)
(make-posn MAX-IMG-X 4))
(check-expect (move-alien (make-posn MAX-IMG-X 3) 'left)
(make-posn (sub1 MAX-IMG-X) 3))
(check-expect (move-alien (make-posn 0 5) 'right)
(make-posn 1 5))
(check-error
(move-alien INIT-ALIEN2 'down)
(format "move-down-image-y: The character at y=~s cannot
move down."
MAX-IMG-Y))
(check-error
(move-alien (make-posn 0 5) 'left)
(format "move-left-image-x: The character at x=~s cannot
move left."
MIN-IMG-X))
(check-error
(move-alien (make-posn MAX-IMG-X 15) 'right)
(format "move-right-image-x: The character at x=~s cannot
move right."
MAX-IMG-X))
218 9 Aliens Attack Version 2

(check-error
(move-alien (make-posn 7 MAX-IMG-Y) 'down)
(format "move-down-image-y: The character at y=~s cannot
move down."
MAX-IMG-Y))
The refinements needed by not defining the subtype posint are in italics to make
them easier to appreciate. They will not be in italics when typed into DrRacket.
The contract now clearly indicate that this function may throw an error. Observe that
there is a new sample expression to move an alien down that replaces the previous
test that used INIT-ALIEN2. The tests using sample computations are also refined
in the same manner. The most interesting refinements are to the tests using sample
values. Error tests have been added. The first error-test is written to make sure an
error is thrown and the correct error message is returned when an attempt is made to
move INIT-ALIEN2 down. Error-checks for inappropriately trying to move an alien
left or right are also added. An alien cannot be moved left if its image-x is 0 nor
can an alien be moved right if its alien-x is MAX-IMG-X.
Overall, it now becomes clear that defining subtypes simplifies the design and the
code that is developed. Functions designed using subtypes are shorter, because they
do not have a guard. Tests are simpler because throwing errors does not need to be
checked. Signatures are simpler because errors are not thrown. All this means that a
function is easier to understand, to refine, and to explain to others. Whenever possible
design using subtypes. In this textbook the version of move-alien developed using
the subtype posint is adopted.

* Ex. 95 — Throughout the semester the Program Correctness course has 5

quizzes: Q0–Q4. A lower numbered quiz is always given and graded before a
higher numbered quiz. A quiz grade may be an integer in [0..100] or, if the
quiz has not been given, 'NA. At the beginning of the semester all the quiz
grades are 'NA. Write a function to compute the quiz average to date, which
may throw an error.

** Ex. 96 — The lengths of the sides of a triangle, a, b, and c, define a right

triangle if a2 + b2 = c2 . Write a function that given the three lengths of a
right triangle returns the image of the right triangle. This function may throw
an error with one of two messages communicating that either the hypothenuse
is too short or the lengths do not define a right triangle.

53 The game-over? Handler

To design the game-over? handler it is necessary to think about when this version
of the game comes to an end. The only way the game comes to an end is when the
alien reaches earth. In other words, the problem of detecting the end of the world
53 The game-over? Handler 219

is reduced to solving a problem about an alien. Think carefully about how can an
alien reaching earth be determined. Visually, the alien has reached earth when it is
at the bottom of scene. Now, think it terms of the alien data definition. How can it
be determined that an alien has reached the bottom of the scene? The alien can only
be at the bottom of the scene if its y coordinate is MAX-IMG-Y.
The above problem analysis allows specializing the template for functions on a
world as follows:
;; world → Boolean
;; Purpose: Detect if the game is over
(define (game-over? a-world)
(alien-reached-earth? (world-alien a-world)))

;; Sample expressions for game-over?

(define GAME-OVER (alien-reached-earth?
(world-alien INIT-WORLD2)))
(define GAME-NOT-OVER (alien-reached-earth?
(world-alien INIT-WORLD)))

;; Tests using sample computations for game-over?

(check-expect (game-over? INIT-WORLD2) GAME-OVER)
(check-expect (game-over? INIT-WORLD) GAME-NOT-OVER)

;; Tests using sample values for game-over?

(check-expect (game-over? (make-world 8
(make-posn 0 3)
'right))
#false)
(check-expect (game-over? (make-world 8
(make-posn 0 MAX-IMG-Y)
'right))
#true)
An auxiliary function, alien-reached-earth?, is used to process the world’s
alien. Two sample expressions are defined to illustrate each of the possible Boolean
outcomes. Observe that INIT-WORLD2 is used to illustrate when the game is over
because its alien has a y coordinate equal to MAX-IMG-Y. The tests using sample
values illustrate each of the possible Boolean outcomes using explicitly constructed
worlds for these tests.
The design of alien-reached-earth? is based on specializing the template
for functions on an alien. The y value of the alien must be tested for equality with
MAX-IMG-Y. The template specialization yields:
;; alien → Boolean
;; Purpose: Determine if the given alien reached earth
(define (alien-reached-earth? an-alien)
(= (posn-y an-alien) MAX-IMG-Y))
220 9 Aliens Attack Version 2

Fig. 63 Alternatives for the final scene

Alien Above the Bottom in Last Scene. Alien at Bottom in Last Scene.

;; Sample expressions for alien-reached-earth?

(define ALIEN-EARTH1 (= (posn-y INIT-ALIEN) MAX-IMG-Y))
(define ALIEN-EARTH2 (= (posn-y INIT-ALIEN2) MAX-IMG-Y))

;; Tests using sample computations for alien-reached-earth?

(check-expect (alien-reached-earth? INIT-ALIEN) ALIEN-EARTH1)
(check-expect (alien-reached-earth? INIT-ALIEN2) ALIEN-EARTH2)

;; Tests using sample values for alien-reached-earth?

(check-expect (alien-reached-earth? (make-posn 14 0)) #false)
(check-expect (alien-reached-earth? (make-posn 9 MAX-IMG-Y))
#true)
In the sample expressions INIT-ALIEN and INIT-ALIEN2 are used to illustrate,
respectively, an alien that has not reached earth and an alien that has reached earth.
The tests using sample values illustrate the same using explicitly constructed aliens.

54 Computing the Last Scene

Running the code reveals that all the tests pass. Hooray! Allowing the game to run
until the alien reaches earth, however, reveals a small unexpected problem. The final
scene is displayed in Fig. 63a. The alien is not at the bottom of the scene. Frankly, it
does not look like it quite reached earth. It would be nicer if the final scene displayed
were the one in Fig. 63b.
This problem occurs because the universe teachpack stops drawing the world
when game-over? evaluates to #true. In the world displayed in Fig. 63a the alien
is about to move down and the direction is about to change to right. When this
move occurs, however, the universe teachpack detects the end of the game and
never draws that final world. To remedy this situation the universe API allows the
programmer to specify a function to draw the last world. This function is specified
in the stop-when stanza of the big-bang expression as follows:
54 Computing the Last Scene 221

; string → world
; Purpose: To run the game
(define (run a-name)
(big-bang INIT-WORLD
[on-draw draw-world]
[name a-name]
[on-key process-key]
[on-tick process-tick TICK-RATE]
[stop-when game-over? draw-last-world]))
The function draw-last-world must take as input a world and must return a scene.
To implement draw-last-world the template for functions on a world is special-
ized. The goal is to draw the given final world with a message stating that earth was
conquered. This message may be placed, for example, half away across and a quarter
of the way down in the scene using a font size of 36 and red letters. This analysis
yields the following specialization of the template for functions on a world:18
;; world → scene throws error
;; Purpose: To draw the game’s final scene
(define (draw-last-world a-world)
(if (= (posn-y (world-alien a-world)) MAX-IMG-Y)
(place-image (text "EARTH WAS CONQUERED!" 36 'red)
(/ E-SCENE-W 2)
(/ E-SCENE-H 4)
(draw-world a-world))
(error
(format "draw-last-world: Invalid world with ∼s as the
alien’s y coordinate. The alien’s y coordinate
must be in [∼s..∼s]."
(posn-y (world-alien a-world))
MIN-IMG-Y
MAX-IMG-Y))))

;; Sample Instance of (final) world

(define FWORLD1 (make-world 13 (make-posn 0 14) 'right))
(define FWORLD2 (make-world 7 (make-posn 19 14) 'left))

;; Sample expressions for f-on-world

(define FWORLD1-VAL (place-image
(text "EARTH WAS CONQUERED!" 36 'red)
(/ E-SCENE-W 2)
(/ E-SCENE-H 4)
(draw-world FWORLD1)))

18
The string inputs to format should be typed in one line in DrRacket.
222 9 Aliens Attack Version 2

Fig. 64 Chapter 9 BSL grammar

program ::= {expr | test | def | defs}*
def ::= (define <variable name> expr)
::= (define (<name> <name>+) expr)
test ::= (check-expect expr expr) | (check-error expr expr)
::= (check-within expr expr expr)
defs ::= (define-struct <struct name> (<field name>*))
expr ::= number
::= string
::= \#<character constant>
::= <character+>
::= #true
::= #false
::= <image>
::= (<function> expr+ )
::= (and expr expr expr*)
::= (or expr expr expr* )
::= (cond [expr expr] +
[else expr])
::= (if expr expr expr)

(define FWORLD2-VAL (place-image

(text "EARTH WAS CONQUERED!" 36 'red)
(/ E-SCENE-W 2)
(/ E-SCENE-H 4)
(draw-world FWORLD2)))

;; Tests using sample computations for f-on-world

(check-expect (draw-last-world FWORLD1) FWORLD1-VAL)
(check-expect (draw-last-world FWORLD2) FWORLD2-VAL)

;; Tests using sample values for f-on-world

(check-error
(draw-last-world (make-world 10 (make-posn 7 20) 'right))
"draw-last-world: Invalid world with 20 as the alien’s y
coordinate. The alien’s y coordinate must be in [0..14].")
Observe that draw-world is used to draw the given world. When draw-last-world
is used by the universe teachpack the given world is the world that stops the game.
To make test development easier two sample instances of a (final) world are defined
and used both in the sample expressions and in the sample computation tests. The
tests using a sample value illustrates that this function throws an error and returns
the right error message for a world in which the y coordinate of the alien is not equal
to MAX-IMG-Y.

*** Ex. 97 — Carefully explain why the alien never goes off the scene.
55 What Have We Learned in This Chapter? 223

** Ex. 98 — An alternative ending for the game is the player saving earth if the
alien crashes into the rocket. Redesign game-over? and draw-last-world to
implement this alternative ending.

** Ex. 99 — Redesign the game such that move-rocket uses

move-right-image-x and move-right-image-y.

* Ex. 100 — The draw-last-world design did not develop a world subtype.
As a consequence, draw-last-world is a guarded function that may throw an
error. Redesign draw-last-world by developing a world subtype for worlds
that end the game.

55 What Have We Learned in This Chapter?

The important lessons of Chap. 9 are summarized as follows:

• The explored BSL syntax is displayed in Fig. 64 and includes the required syntax
to test error messages.
• Refining a data definition means that sample instances and functions (sample
expressions, defined functions, and tests) must be refined.
• Well-designed code allows for the reuse existing code when a refinement is
required.
• Subtypes are defined when a proper subset of a given type needs to be processed.
• If T2 is a subtype of T1, then functions on T1 may process T2 instances. Functions
on T2 may not process T1 instances.
• A function may throw an error.
• Error throwing is indicated by having throws error in the signature.
• The function error is used to throw an error.
• A guarded function returns a value if the input is valid. Otherwise, it throws an
error.
• The function format is used to create strings customized with values.
• A function to draw a final scene may be specified in the stop-when stanza of a
big-bang expression.
Chapter 10
Structures and Variety

Chapter 9 revealed that functions may consume instances of different data types. This
is the case of functions like move-alien and hit?. It turns out that it is common
for functions to consume different data types. Functions like move-alien and hit?
have a separate parameter for each data type. Now consider designing a function
that processes motorized vehicles like cars, motorcycles, and personal transporters
or a function that processes geometric shapes like squares, rectangles, triangles, and
ellipses. This is a situation that is quite different from designing move-alien and
hit?. Now, there is a single input: a motor vehicle. This input, however, may be an
instance of several different data types.
This chapter further explores how to design functions that consume data that may
vary. Given that there is variety in the data conditional expressions play a central
role. Remember that a conditional is needed to distinguish the different varieties.
The design may follow a bottom-up or a top-down approach. So far, a top-down
approach has been emphasized. In this chapter, a bottom-up approach is taken. Why
would you ever take a bottom-up approach? Sometimes it is easier to formulate the
simplest forms of data and then use them to define more complex data.
We shall explore the design of a function to process motor vehicles. The charac-
teristics of the different motor vehicles are:
Car Has a tank capacity in gallons, a miles per gallon, a maximum speed, and a
mode. The mode indicates if the car is running in economic mode. When the car
is running in economic mode, it may travel 20% further than in normal mode, but
the accelerator is less responsive.
Motorcycle Has a tank capacity in gallons, a miles per gallon, and a maximum
speed.
Personal Transporter Has a maximum miles per hour and maximum hours per
charge.

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 225
M. T. Morazán, Animated Problem Solving, Texts in Computer Science,
https://doi.org/10.1007/978-3-030-85091-3_10
226 10 Structures and Variety

56 A Bottom-Up Design

A bottom-up design starts with defining the different motor vehicles first. A car has
four characteristics and this suggests using a structure. A car and sample instances
of a car may be defined as follows:
#|
A car is a structure
(make-carr integer>=0 integer>=0 integer>=0 integer>=0 Boolean)
containing tank capacity in gallons, miles per gallon, maximum
speed, and economy mode flag.

;; Sample instances of carr

(define CARR1 . . .)

;; carr . . . → . . .
;; Purpose: . . .
(define (f-on-carr a-carr . . .)
. . .(carr-gallons a-carr). . .(carr-mpg a-carr)
. . .(carr-maxspeed a-carr). . .(carr-mode a-carr))

;; Sample expressions for f-on-carr

(define CARR1-VAL . . . CARR1 . . .)
...

;; Tests using sample computations for f-on-carr

(check-expect (f-on-carr CARR1 . . .) CARR1-VAL)
...

;; Tests using sample values for f-on-carr

(check-expect (f-on-carr . . .) . . .)
... |#

;; Structure Definition
(define-struct carr (gallons mpg maxspeed mode))

;; Sample Instances of carr

(define CARR1 (make-carr 20 35 140 #true))
(define CARR2 (make-carr 22 15 160 #false))
You may think that there is a spelling mistake because carr, not car, is defined.
This is not a spelling mistake. There is a function named car in BSL. Therefore,
the structure name carr is chosen. Nonetheless, carr represents a car. Given that
there is no variety the template suggests at least one sample instance. In this case,
two sample instances have been defined: a carr in economic mode and a carr not
in economic mode. The definition template lists the four expressions, using carr’s
56 A Bottom-Up Design 227

selector functions, that may be useful to process a carr. Also given that there is no
variety in carr the template suggests at least one test using sample computations
and one test using sample values.
We now turn our attention to data analysis for motorcycles. An mc and sample
instances of an mc may be defined as follows:
#| A motorcycle, mc, is a structure
(make-mc integer>=0 integer>=0 integer>=0)
with a tank capacity in gallons, miles per gallon, and a
maximum speed.

;; Sample instances of mc
(define MC1 . . .)

;; mc . . . → . . .
;; Purpose: . . .
(define (f-on-mc an-mc . . .)
. . .(mc-gallons an-mc). . .(mc-mpg an-mc)
. . .(mc-maxspeed an-mc). . .)

;; Sample expressions for f-on-mc

(define MC1-VAL . . . MC1 . . .)
...

;; Tests using sample computations for f-on-mc

(check-expect (f-on-mc MC1 . . .) MC1-VAL)
...

;; Tests using sample values for f-on-mc

(check-expect (f-on-mc . . .) . . .)
... |#

;; Structure Definition
(define-struct mc (gallons mpg maxspeed))

;; Sample instances of mc
(define MC1 (make-mc 8 22 135))
(define MC2 (make-mc 10 20 145))
Observe that the defining an mc is similar to defining a carr. Useful expressions to
extract data from the given mc instance are found in the function definition template.
Given that there is no variety the function template suggests defining at least one
sample instance, one sample expression, one test using sample computations, and
one test using sample values. Two sample instances are defined for thorough testing
and to parallel the work done for carr.
228 10 Structures and Variety

The last motor vehicle to define is the personal transporter. An pt and sample
instances of an pt may be defined as follows:
#| A personal transporter, pt, is a structure
(make-pt integer>=0 integer>=0)
with a maximum miles per hour and a maximum hours per
charge.

;; Sample instances of pt
(define PT1 . . .)

;; pt . . . → . . .
;; Purpose: . . .
(define (f-on-pt a-pt . . .)
. . .(pt-mph a-pt). . .(pt-hpc a-pt))

;; Sample expressions for f-on-pt

(define PT1-VAL . . . PT1 . . .)
...

;; Tests using sample computations for f-on-pt

(check-expect (f-on-pt PT1 . . .) PT1-VAL)
...

;; Tests using sample values for f-on-pt

(check-expect (f-on-pt . . .) . . .)
... |#

;; Structure Definition
(define-struct pt (mph hpc))

;; Sample instances of pt
(define PT1 (make-pt 7 7))
(define PT2 (make-pt 6 10))
Observe that the data design steps are essentially the same as those taken for mc and
carr.
It is important to take stock of what has been achieved so far by following a
bottom-up approach. There are three data definitions with a function template and
sample instances for each. These data definitions correspond to the motor vehicle
subtypes. A bottom-up approach, therefore, first defines subtypes.
Once subtypes are defined, the next step is to define a union type. A union type
enumerates the type varieties that may be used to build instances. If A, B, and C are
types, then D may be defined as the union of A, B, and C. A, B, and C are subtypes of
D. We say that D is A’s, B’s, and C’s supertype. Functions written for a supertype must
be able to process instances of each of its subtypes. That is, a function on a supertype
56 A Bottom-Up Design 229

must be polymorphic. Polymorphic means that a function can process different types
of data.
In our example, a supertype for motor vehicles is needed. The supertype mv and
instances of mv may be defined as a union type:
#| A motor vehicle, mv, is either:
1. carr
2. mc
3. pt

;; Sample instances of mv
(define MVCARR1 . . .)
(define MVMC1 . . .)
(define MVPT1 . . .)

;; mv . . . → . . .
;; Purpose: . . .
(define (f-on-mv an-mv . . .)
(cond [(car? an-mv) (f-on-car an-mv . . .)]
[(car? an-mv) (f-on-mc an-mv . . .)]
[else (f-on-mv an-mv . . .)]))

;; Sample expressions for f-on-mv

(define MVCARR1-VAL . . . MVCARR1 . . .)
(define MVMC1-VAL . . . MVMC1 . . .)
(define MVPT1-VAL . . . MVPT1 . . .)
...

;; Tests using sample computations for f-on-mv

(check-expect (f-on-mv MVCARR1 . . .) MVCARR1-VAL)
(check-expect (f-on-mv MVMC1 . . .) MVMC1-VAL)
(check-expect (f-on-mv MVPT1 . . .) MVPT1-VAL)
...

;; Tests using sample values for f-on-mv

(check-expect (f-on-mv . . .) . . .)
... |#

;; Sample instances of mv
(define MVCARR1 CARR1)
(define MVCARR2 CARR2)
(define MVMC1 MC1)
(define MVMC2 MC2)
(define MVPT1 PT1)
(define MVPT2 PT2)
230 10 Structures and Variety

Fig. 65 Maximum distance for mc and pt

;; mc integer>=0
;; Purpose: Return max distance the
;; given mc may travel on full
;; pt integer>=0
(define (mc-maxdist an-mc)
;; Purpose: Return max distance
(* (mc-gallons an-mc)
;; the given pt may travel on full
(mc-mpg an-mc)))
(define (pt-maxdist a-pt)
(* (pt-mph a-pt) (pt-hpc a-pt)))
;; Sample expressions for
;; mc-maxdist
;; Sample expressions for f-on-pt
(define MC1-VAL (* (mc-gallons MC1)
(define PT1-VAL (* (pt-mph PT1)
(mc-mpg MC1)))
(pt-hpc PT1)))
(define MC2-VAL (* (mc-gallons MC2)
(define PT2-VAL (* (pt-mph PT2)
(mc-mpg MC2)))
(pt-hpc PT2)))
;; Tests using sample computations
;; Tests using sample computations
;; for mc-maxdist
;; for pt-maxdist
(check-expect (mc-maxdist MC1)
(check-expect (pt-maxdist PT1)
MC1-VAL)
PT1-VAL)
(check-expect (mc-maxdist MC1)
(check-expect (pt-maxdist PT2)
MC1-VAL)
PT2-VAL)
;; Tests using sample values for
;; Tests using sample values
;; mc-maxdist
;; for pt-maxdist
(check-expect (mc-maxdist
(check-expect
(make-mc 10 10 120))
(pt-maxdist (make-pt 4 5))
100)
20)
(check-expect (mc-maxdist
(make-mc 8 15 105)) (b) Maximum Distance for pt .
120)

(a) Maximum Distance for mc .

Observe that a union type is nothing more than a data type that has variety. You
have already learned about union types in previous chapters (e.g., speed in Chap. 5
and mn in Chap. 8) although they were not called union types. Therefore, it follows
that the definition template has a conditional to distinguish among the subtypes. As
expected, the function template also suggests defining at least one sample expression
and one test using sample computations for each subtype. Observe that the sample
instances are defined using the sample instances of mv’s subtypes. This can only be
done because an instance of a subtype is also an instance of the supertype. That is,
instances of carr, mc, and pt are instances of mv.
With data analysis, finished problems involving motor vehicles may be solved.
Consider writing a function to compute the maximum distance a motor vehicle
may travel. Following a bottom-up approach means that functions to compute the
maximum distance for each subtype (i.e., carr, mc, and pt) are written first. The
result of specializing the function templates for mc, pt, and carr is displayed,
respectively, in Figs. 65a and 66a. The specialization of the three templates varies in
one significant way. In Fig. 65a,b the functions return an integer greater than or equal
56 A Bottom-Up Design 231

Fig. 66 Maximum distance for carr and mv

;; carr number>=0 ;; mv number>=0
;; Purpose: Return max distance the ;; Purpose: Return max distance
;; given carr may travel on full ;; the given mv may travel on full
(define (carr-maxdist a-carr) (define (mv-maxdist amv)
(if (carr-mode a-carr) (cond
(* 1.2 [(carr? amv) (carr-maxdist amv)]
(carr-gallons a-carr) [(mc? amv) (mc-maxdist amv)]
(carr-mpg a-carr)) [else (pt-maxdist amv)]))
(* (carr-gallons a-carr) ;; Sample expressions for mv-maxdist
(carr-mpg a-carr)))) (define MVCARR1-VAL
;; Sample expressions for f-on-carr (carr-maxdist MVCARR1))
(define CARR1-VAL (define MVCARR2-VAL
(* 1.2 (carr-maxdist MVCARR2))
(carr-gallons CARR1) (define MVMC1-VAL
(carr-mpg CARR1))) (mc-maxdist MVMC1))
(define CARR2-VAL (define MVPT1-VAL
(* (carr-gallons CARR2) (pt-maxdist MVPT1))
(carr-mpg CARR2))) ;; Tests using sample computations
(check-expect (mv-maxdist MVCARR1)
;; Tests using sample computations MVCARR1-VAL)
;; for carr-maxdist (check-expect (mv-maxdist MVCARR2)
(check-expect (carr-maxdist CARR1) MVCARR2-VAL)
CARR1-VAL) (check-expect (mv-maxdist MVMC1)
(check-expect (carr-maxdist CARR2) MVMC1-VAL)
CARR2-VAL) (check-expect (mv-maxdist MVPT1)
;; Tests using sample values MVPT1-VAL)
;; for carr-maxdist ;; Tests using sample values
(check-expect (check-expect
(carr-maxdist (mv-maxdist
(make-carr 9 21 10 #true)) (make-carr 25 40 140 #true))
226.8) 1200)
(check-expect (check-expect
(carr-maxdist (mv-maxdist
(make-carr 9 21 10 #false)) (make-carr 25 40 140 #false))
189) 1000)

(a) Maximum Distance for carr. (b) Maximum Distance for mv.

to zero. In Fig. 66a the function returns a number greater than or equal to zero. This
difference in the signatures arises from the observation that carr-maxdist may not
always return an integer. For example, (* 1.2 25 40) evaluates to 1131.6. All
three functions, however, are tested using check-expect. Using check-within
for carr-maxdist is not necessary, because (* 1.2 integer>=0 integer>=0)
is always an exact integer. A careful reader of the code may wonder about not
using check-within. To mitigate this type of concern an assumption statement
may accompany the function definition as follows:
;; carr → number>=0
;; Purpose: Return max distance the given carr may travel
;; on full
232 10 Structures and Variety

;; Assumption: (* 1.2 integer>=0 integer>=0) is an exact

;; number>=0
(define (carr-maxdist a-carr)
(if (carr-mode a-carr)
(* 1.2 (carr-gallons a-carr) (carr-mpg a-carr))
(* (carr-gallons a-carr) (carr-mpg a-carr))))
After defining the functions for the mv subtypes, the bottom-up design process
continues with the design of the function for the supertype: mv-maxdist. The
template for functions on an mv suggests using the three auxiliary functions in
Figs. 65a and 66a. What is mv-maxdist’s return type? Processing an mc or a pt
returns an integer>=0. Processing a carr returns a number>=0. These are two
different types. An obvious solution is to create a new union type data definition:
A distance (dist) is either:
1. integer>=0
2. number>=0
Before writing the function template and the sample instances, it is worthwhile to
carefully think about this data definition. Observe that integer>=0 is a subtype
of number>=0. This means that the value returned by mv-maxdist is always a
number>=0. Given that in this problem there is no need to distinguish between
the varieties, the data definition for dist is not needed. We can safely claim that
mv-maxdist returns a number>=0. With this analysis, the steps of the design recipe
may be completed by specializing the template for functions on an mv. The special-
ization is displayed in Fig. 66b.

*** Ex. 101 — Change the data definition of a motor vehicle to include a boat:
A boat is either:
1. Cruise Boat that has two gas tanks in gallons, a
miles per gallon, and a mode. The mode indicates
if the boat is empty. When empty, a cruise boat
may travel 50 more miles per gallon than when not
empty.
2. Speed Boat that has a gas tank per gallon, a
miles per gallon, and a number of passengers: 1
or 2. When the boat has 2 passengers it travels
10 fewer miles per gallon than when it only has 1
passenger.
Refine mv-maxdist for the new motor vehicle definition.
*** Ex. 102 — Consider the following data definition:
A geometric shape is either:
1. square: Has a length
2. rectangle: Has a length and a width
3. circle: Has a radius
4. ellipse: Has a major axis length and a minor axis length
57 Code Refactoring 233

Design and implement a function to compute the area of a geometric shape.

*** Ex. 103 — Design and implement a function to compute the perimeter of
a geometric shape as defined in the previous problem.

* Ex. 104 — A flag is a Boolean value used to indicate if a property is present

or not. Consider the following data definition:
A university student is either:
1. freshman: Has a name, last name, enrollment year, and gpa
2. sophomore: Has a name, last name, enrollment year, gpa,
and Dean’s List flag
3. junior: Has a name, last name, enrollment year, gpa,
and Dean’s List flag
4. senior: Has a name, last name, enrollment year, gpa,
and Dean’s List flag
Write a function that converts a university student to a string.

57 Code Refactoring

Although we have a well-designed and working program, it is possible to do better.

Take a close look at the expressions used in the auxiliary functions:
(* 1.2 (carr-gallons a-carr) (carr-mpg a-carr)) ;;in mv-maxdist
(* (carr-gallons a-carr) (carr-mpg a-carr)) ;;in mv-maxdist
(* (mc-gallons an-mc) (mc-mpg an-mc)) ;;in mc-maxdist
(* (pt-hpc a-pt) (pt-mph a-pt)) ;;in pt-maxdist
These expressions all look very similar. The last three expressions are structurally
the same. They compute the product of energy units and miles per energy unit. This
suggests that abstraction over expressions ought to be used to create a single function
to compute these values.
We cannot, however, directly use abstraction over expressions to include the first
expression. Although similar, it is structurally different from the other expressions.
The first expression multiplies three, not two, quantities. In such a case, consider
using a technique called code refactoring. Code refactoring restructures existing
expressions (or functions) without changing their external behavior. Expressions
may be restructured to look the same and, therefore, allow for abstraction over similar
expressions. These changes, however, must be done without changing the purpose
of the expressions. The first expression above has, 1.2, a constant of proportionality
not present in the other three expressions. We can refactor the last three expressions
to contain a constant of proportionality as follows:
(* 1.2 (carr-gallons a-carr) (carr-mpg a-carr))
(* 1 (carr-gallons a-carr) (carr-mpg a-carr))
234 10 Structures and Variety

Fig. 67 The mv-maxdist helper function after code refactoring

;; number>=0 integer>=0 integer>=0 --> number>=0
;; Purpose: Compute the maximum distance
(define (refactored-helper k energy-units miles/energy-unit)
(* k energy-units miles/energy-unit))

;; Sample expressions for refactored-helper

(define DIST1 (* 1.2 (carr-gallons CARR1) (carr-mpg CARR1)))
(define DIST2 (* 1 (carr-gallons CARR2) (carr-mpg CARR2)))
(define DIST3 (* 1 (mc-gallons MC1) (mc-mpg MC1)))
(define DIST4 (* 1 (pt-hpc PT1) (pt-mph PT1)))

;; Tests using sample computations for refactored-helper

(check-expect
(refactored-helper 1.2 (carr-gallons CARR1) (carr-mpg CARR1))
DIST1)
(check-expect
(refactored-helper 1 (carr-gallons CARR2) (carr-mpg CARR2))
DIST2)
(check-expect
(refactored-helper 1 (mc-gallons MC1) (mc-mpg MC1))
DIST3)
(check-expect
(refactored-helper 1 (pt-hpc PT1) (pt-mph PT1))
DIST4)

;; Tests using sample values for refactored-helper

(check-expect (refactored-helper 1.2 11 30) 396)

(* 1 (mc-gallons an-mc) (mc-mpg an-mc))

(* 1 (pt-hpc a-pt) (pt-mph a-pt))
With this new insight, the function in Fig. 67 is designed and written. Observe that
for each of the three differences in the sample expressions, there is a parameter in
refactored-helper. This one function is able to do the work of the three functions
in Figs. 65a and 66a. Therefore, we can substitute all the code in Figs. 65a and 66a
with the code in Fig. 67.
Eliminating the code to compute the maximum distance for each subtype of motor
vehicle means that mv-maxdist needs to be refined to use a single helper function.
Only the sample expressions and the function definition need to be updated. The tests
remain unchanged because mv-maxdist’s external behavior remains unchanged. For
an mc and a pt the constant of proportionality is always 1. The energy and energy
units and the miles per energy unit are extracted from the given structure. The
constant of proportionality may vary from one carr to another. Therefore, it is
necessary to determine which constant of proportionality to use. This is achieved
using a conditional to distinguish if the given carr is running in economic mode.
The refined code is displayed in Fig. 68. Observe that in the conditional’s carr
stanza there is an if-expression to determine the constant of proportionality to use.
Further observe that the tests are the same as the tests in Fig. 66b.
57 Code Refactoring 235

Fig. 68 Refined mv-maxdist using refactored helper function

;; mv number>=0
;; Purpose: Return max distance the given mv may travel on full
(define (mv-maxdistance an-mv)
(cond
[(carr? an-mv)
(if (carr-mode an-mv)
(refactored-helper 1.2 (carr-gallons an-mv) (carr-mpg an-mv))
(refactored-helper 1 (carr-gallons an-mv) (carr-mpg an-mv)))]
[(mc? an-mv)
(refactored-helper 1 (mc-gallons an-mv) (mc-mpg an-mv))]
[else (refactored-helper 1 (pt-hpc an-mv) (pt-mph an-mv))]))

;; Sample expressions for mv-maxdistance

(define MVCARR1-VAL
(refactored-helper 1.2 (carr-gallons MVCARR1) (carr-mpg MVCARR1)))
(define MVCARR2-VAL
(refactored-helper 1 (carr-gallons MVCARR2) (carr-mpg MVCARR2)))
(define MVMC2-VAL
(refactored-helper 1 (mc-gallons MVMC2) (mc-mpg MVMC2)))
(define MVPT1-VAL
(refactored-helper 1 (pt-hpc MVPT1) (pt-mph MVPT1)))

;; Tests using sample computations for mv-maxdistance

(check-expect (mv-maxdist MVCARR1) MVCARR1-VAL)
(check-expect (mv-maxdist MVCARR2) MVCARR2-VAL)
(check-expect (mv-maxdist MVMC2) MVMC2-VAL)
(check-expect (mv-maxdist MVPT1) MVPT1-VAL)

;; Tests using sample values for mv-maxdistance

(check-expect (mv-maxdist (make-carr 25 40 140 #true)) 1200 )
(check-expect (mv-maxdist (make-carr 25 40 140 #false)) 1000 )
(check-expect (mv-maxdist (make-mc 12 25 140)) 300 )
(check-expect (mv-maxdist (make-pt 8 5)) 40 )

Code refactoring has significantly reduced the size of the code. The original
code’s four functions are reduced to two. Code refactoring plays a central role in
software development. It is used to improve design, structure, and efficiency without
changing software functionality.

** Ex. 105 — The expressions used to compute the maximum distance may
instead be refactored as follows:
(* (* 1.2 (carr-gallons a-carr)) (carr-mpg a-carr))
(* (carr-gallons a-carr) (carr-mpg a-carr))
(* (mc-gallons an-mc) (mc-mpg an-mc))
(* (pt-hpc a-pt) (pt-mph a-pt))
236 10 Structures and Variety

The computation of the maximum distances is always the product of two num-
bers. Use this code refactoring idea to refine the implementation of mv-maxdist
in Fig. 66b.

** Ex. 106 — Horner’s Method. Implement functions for:

f(x) = 4x3 + 10x2 + 5x + 8
g(x) = 3x4 + x3 + 17x2 + 7x + 2
Now that you have implemented the two functions, observe that many of the
same values are repeatedly computed. For example, in f the value of x2 is
computed for 10x2 and 4x3 and in g the value of x2 is computed for 17x2 , x3 ,
and 3x4 . Repeatedly computing the same value is inefficient.
In 1819, William George Horner observed that a function’s expression may be
refactored to avoid repeating computations. For example, consider the following
h:
h(x) = 87x4 + 31x2 + 17x + 99
Computing the value of h(x) naively requires 9 multiplications and 3 additions.
Horner observed that h(x) may be refactored to:
h(x) = 99 + x(17 + x(31 + x(0 + x(87))))
Observe that computing the value of h(x) only requires 4 multiplications and
4 additions now. Refactor your code for f(x) and g(x) to take advantage of
Horner’s method. Make sure you only refine the sample expressions and the
definition bodies.

58 What Have We Learned in This Chapter?

The important lessons of Chap. 10 are summarized as follows:

• Use a bottom-up design approach when it is easier to define the simplest forms
of data first.
• A supertype is a collection of one or more subtypes that define a data type.
• A union type is an enumeration of subtypes that may be used to define a supertype.
• A bottom-up design of a function that processes a supertype instance starts with
the auxiliary functions, one for each subtype, and then designs the function for
the supertype.
• If a supertype has variety, has a subtype, A, that contains all the other subtypes,
and had no need to distinguish among the varieties of the supertype, then A may
be used instead of defining subtypes.
• Code refactoring restructures existing expressions or functions without changing
their external behavior.
• Abstracting over refactored expressions may be an effective means to simplify
the function design.
58 What Have We Learned in This Chapter? 237

• When the subtype code is refactored, supertype sample expressions and functions
may have to be refined.
• Code refactoring is used to improve design, structure, and efficiency without
changing software functionality.
Chapter 11
Aliens Attack Version 3

Writing union types that contain structure-based subtypes endows us with the power
to refine Aliens Attack version 2 (from Chap. 9). This chapter discusses adding a
shot to the game. As in Aliens Attack version 2, the world in version 3 has a rocket,
an alien, and a direction. The world data definition is now refined to contain a shot.
We emphasize the singular here. There can be at most one shot in the game at any
time. A player may shoot only when there is no shot in the game. At the beginning
of the game, for example, the player may shoot. A player may not shoot again until
the last shot created goes off the top of the scene.
Now that a shot is part of the world it becomes possible for the player to win the
game. When does this happen? It is fairly straightforward to see that the player ought
to win when the alien is hit by the shot. It is less straightforward to precisely define
when an alien has been hit. Clearly, the alien has been hit when the coordinates of
the alien and the shot are the same. For example, if the alien and the shot are both (4
2) then the alien has been hit and the player wins the game. Now consider an alien
at (0 7), a shot at (0 6), and the direction being 'down. This puts the alien and
the shot at the left edge of the scene. The alien moves to coordinate (0 6) and the
shot moves to coordinates (0 7). Has the alien been hit? On the one hand, the alien
and the shot are never the same. On the other hand, the alien and the shot cross each
other. This is a situation that forces game designer to choose feature. If it is not a hit,
it makes the game a bit more challenging for the player to win. If it is a hit, it makes
the game a bit easier to win. The presented design chooses the former feature. An
alien moving down “sees” the shot coming and skittles beneath it to avoid the hit.
Adding a shot to the world means that world’s function template and world
sample instances must also be refined. In addition, any functions and tests that build
a world must be updated. This is the same process followed to develop Aliens Attack
2. The refinement presented follows a top-down design. The first step is to refine
the world data definition. The second step is to define a shot. The third step is
the refinement of the program and the design of new functions. Observe that data
analysis is performed first. Remember that if you do not have a clear understanding
of the data being processed, then it is impossible to design a solution to a problem.

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 239
M. T. Morazán, Animated Problem Solving, Texts in Computer Science,
https://doi.org/10.1007/978-3-030-85091-3_11
240 11 Aliens Attack Version 3

Fig. 69 A sample rendering of Aliens attack with a shot

59 Data Definitions

Figure 69 displays a rendering of the game that now includes a shot. The only new
component is the shot. Therefore, the world data definition is refined as follows:
#| A world is a structure: (make-world rocket alien dir shot).

;; world . . . → . . .
;; Purpose: . . .
(define (f-on-world a-world . . .)
. . .(f-on-rocket (world-rocket a-world) . . .)
. . .(f-on-alien (world-alien a-world) . . .)
. . .(f-on-dir (world-dir a-world) . . .)
. . .(f-on-shot (world-shot a-world) . . .))

;; Sample instances for world

(define WORLD1 (make-world . . .))

;; Sample expressions for f-on-world

(define WORLD-VAL1 . . .)

;; Tests using sample computations for f-on-world

(check-expect (f-on-world WORLD1 . . .) WORLD-VAL1)
...
59 Data Definitions 241

;; Tests using sample values for f-on-world

(check-expect (f-on-world . . .) . . .) #|
The data definition includes a new field of type shot. This new type is not yet defined.
Therefore, the world data definition is written assuming that the shot data definition
will be developed. Observe that the sample instances of the world are not yet refined
for this reason. Their refinement requires instances of shot and, therefore, must be
postponed until shot is defined. Nonetheless, the definition template contains an
expression, (f-on-shot (world-shot a-world) . . .), to process a shot. Once
again, we see a disciplined separation of concerns. A function on a world does not
process a shot. Instead, a function on a shot is used to process the given world’s shot.
Now, let us think about what a shot is. Remember that a shot is not an image of a
shot. A player uses a key, say the space bar key (or simply the space key), to shoot.
Think about what may change from one instance on a shot to another. A shot moves
as follows:

A shot always moves up. What changes when a shot moves up? Clearly, an image-y
coordinate is needed to represent a shot that moves up. Is an image-x coordinate
needed? As the shot moves up the image-x coordinate never changes. However, the
image-x coordinate may vary among shots. This coordinate is determined by the
position of the rocket when the shot is created. If the rocket is 4 when the player
shoots, then the new shot’s image-x coordinate is 4. If the rocket is 10 when the
player shoots, then the new shot’s image-x coordinate is 10. This means that an
image-x coordinate is needed. Using a posn to represent a shot is a good option.
What posn should be used for the shot at the beginning of the game? This is
an interesting question because the player has not created a shot. If a shot has not
been created, how can it have image-x and image-y coordinates? This situation is
suggesting that there is variety in shots: non-existing and existing shots. We have
already determined to represent an existing shot using a posn. The representation
choice for a non-existing shot is arbitrary. We shall define, NO-SHOT, a symbol
constant to represent such a shot. Based on this data analysis the data definition for
a shot and the template for a function on a shot are:
(define NO-SHOT 'no-shot)

#| A shot is either:
1. NO-SHOT
2. A structure (make-posn image-x image-y)
242 11 Aliens Attack Version 3

;; Sample instances of shot

(define SHOT1 . . .)
(define SHOT2 . . .)

;; shot . . . → . . .
;; Purpose: . . .
(define (f-on-shot a-shot . . .)
(if (eq? a-shot NO-SHOT)
...
. . .(f-on-image-x (posn-x a-shot))
. . .(f-on-image-y (posn-y a-shot)). . .)))

;; Sample expressions for f-on-shot

(define SHOT1-VAL . . .)
(define SHOT2-VAL . . .)
...

;; Tests using sample computations for f-on-shot

(check-expect (f-on-shot SHOT1 . . .) SHOT1-VAL)
(check-expect (f-on-shot SHOT2 . . .) SHOT2-VAL)
...

;; Tests using sample values for f-on-shot

(check-expect (f-on-shot . . .) . . .)
... |#

Observe that a posn is used to represent both an alien and a shot. Although
the same representation is used, the expected behavior is quite different. A posn
representing a shot may only be moved up. On the other hand, a shot representing
an alien may be may be moved right, left, and down. This further highlights why
data definitions are important. They clearly outline that not all data represented using
a posn is the same.
Now that a shot is defined it is possible to define sample instances of a shot and
refined the sample instances of a world. These are:
;; Sample instances of shot
(define INIT-SHOT NO-SHOT)
(define SHOT2 (make-posn (/ MAX-CHARS-HORIZONTAL 2)
(/ (sub1 MAX-CHARS-VERTICAL) 2)))
(define SHOT3 (make-posn 4 MAX-IMG-Y))
(define SHOT4 (make-posn 14 MIN-IMG-Y))

;; Sample instances of world

(define INIT-WORLD (make-world INIT-ROCKET INIT-ALIEN
INIT-DIR INIT-SHOT))
(define INIT-WORLD2 (make-world INIT-ROCKET2 INIT-ALIEN2
DIR2 SHOT2))
60 The draw-world Refinement 243

(define WORLD3 (make-world 7 (make-posn 3 3)

'right (make-posn 3 3)))
As suggested by the template for a function on a shot, there is at least one shot
instance for each variety. In addition, shots are defined for extrema image-y values.
The world instances now contain a shot. The first instance is a world that does
not contain a shot. The second instance is a world that contains a shot. The third
world instance illustrates that the same posn instance may be an alien or a shot.

60 The draw-world Refinement

In addition to rendering the rocket and the alien, the shot must be rendered. The
only decision that needs to be made is whether the alien or the shot is rendered
on top when a hit occurs. This is mostly a matter of personal preferences. In order
to clearly see that a hit has occurred, the chosen design renders the shot on top of
the alien. This means that the shot must be rendered in a scene that contains the
alien. Given that we have already developed code to render a scene that has the
alien and the rocket, all we need is function composition to draw the shot in
such a scene. First, let us refine the sample expressions for draw-world:
;; Sample expressions for draw-world
(define WORLD-SCN1
(draw-shot (world-shot INIT-WORLD)
(draw-alien (world-alien INIT-WORLD)
(draw-rocket
(world-rocket INIT-WORLD)
E-SCENE))))

(define WORLD-SCN2
(draw-shot (world-shot INIT-WORLD2)
(draw-alien (world-alien INIT-WORLD2)
(draw-rocket
(world-rocket INIT-WORLD2)
E-SCENE))))
Observe that the shot is rendered in a scene that contains the alien (and the rocket)
using function composition. In addition to such a scene, the new function to draw
the shot takes as input the world’s shot.
The above sample expressions lead to the following refinement of the function
definition for draw-world:
;; world → scene
;; Purpose: To draw the world in E-SCENE
244 11 Aliens Attack Version 3

(define (draw-world a-world)

(draw-shot (world-shot a-world)
(draw-alien (world-alien a-world)
(draw-rocket (world-rocket a-world)
E-SCENE))))
The final step for draw-world is to refine the tests. The tests using sample compu-
tations do not need to be refined because they do not build a world. The tests using
sample values, on the other hand, need to be refined given that they build worlds.
These are refined to construct a world in which the player has not created a shot
and a world in which the player has not created a shot:
(check-expect
(draw-world (make-world INIT-ROCKET2 INIT-ALIEN
INIT-DIR3 NO-SHOT))

)
(check-expect
(draw-world (make-world INIT-ROCKET INIT-ALIEN2
INIT-DIR2 SHOT2))

)
The next step is to design the auxiliary function to draw a shot. To do so, the
template for a function on a shot is specialized. Think carefully about how to draw a
shot. There are two varieties of shot. Think about each variety independently. If the
given shot is 'NO-SHOT, then there is nothing to draw in the given scene. If the given
shot is a posn, then the shot must be drawn at the image coordinates contained in
the shot. This problem analysis leads to the following template specialization:
;; shot scene → scene
;; Purpose: To draw the shot in the given scene
(define (draw-shot a-shot scn)
(if (eq? a-shot NO-SHOT)
scn
(draw-ci SHOT-IMG (posn-x a-shot) (posn-y a-shot) scn)))
60 The draw-world Refinement 245

;; Sample expressions for draw-shot

(define INIT-SHOT-VAL E-SCENE)
(define SHOT2-VAL (draw-ci SHOT-IMG (posn-x SHOT2)
(posn-y SHOT2) E-SCENE))
(define SHOT3-VAL (draw-ci SHOT-IMG (posn-x SHOT3)
(posn-y SHOT3) E-SCENE2))

;; Tests using sample computations for draw-shot

(check-expect (draw-shot INIT-SHOT E-SCENE) INIT-SHOT-VAL)
(check-expect (draw-shot SHOT2 E-SCENE) SHOT2-VAL)
(check-expect (draw-shot SHOT3 E-SCENE2) SHOT3-VAL)

;; Tests using sample values for draw-shot

(check-expect (draw-shot INIT-SHOT E-SCENE2) E-SCENE2)
(check-expect (draw-shot SHOT2 E-SCENE2)

)
The signature identifies a shot and a scene as input values and a scene as the output
value. The purpose statement concisely describes what the function does. The sample
expressions illustrate the computations that are performed for the different shot
varieties. The first sample expression illustrates that nothing needs to be computed for
INIT-SHOT (which is 'NO-SHOT). The second and third sample expressions illustrate
how to compute a scene that contains a shot using draw-ci (the same function used
to draw an alien a rocket). The conditional expression in the body of the function
definition first determines if the given shot is 'NO-SHOT. If so, the given scene
is returned without any further computation as done in the first sample expression.
Otherwise, the shot ci is rendered in the given scene at the image coordinates
contained in the given shot as done in the other sample expressions. The tests using
sample computations employ the differences among the sample expressions and the
defined constants, as before, to illustrate that the defined function returns the same
values as those obtained from the sample expressions. Finally, the tests using sample
values further illustrate that the function works for each variety of shot using a
combination of input values not used in the sample expressions.
246 11 Aliens Attack Version 3

61 The process-key Refinement

The key event handler must be updated to construct worlds that contain a shot. In
addition, this function must also process the key event associated with shooting. The
key choice is arbitrary, but as indicated before a popular choice is to use the space
key to allow a player to shoot.19 This means that the key data definition and the
template for a function on a key must be refined. This is done as follows:
A key is either:
1. "right"
2. "left"
3. " "
4. not "right", "left", or " "

;; Sample instances of key

(define KEY1 ...)
(define KEY2 ...)
(define KEY3 ...)
(define KEY4 ...)

;; key . . . → . . .
;; Purpose: . . .
(define (f-on-key a-key . . .)
(cond [(key=? a-key "right") . . .]
[(key=? a-key "left") . . .]
[(key=? a-key " ") . . .]
[else . . .]))

;; Sample expressions for f-on-key

(define KEY1-VAL . . . KEY1 . . .)
(define KEY2-VAL . . . KEY2 . . .)
(define KEY3-VAL . . . KEY3 . . .)
(define KEY4-VAL . . . KEY4 . . .)

;; Tests using sample computations for f-on-key

(check-expect (f-on-key KEY1 . . .) KEY1-VAL)
(check-expect (f-on-key KEY2 . . .) KEY2-VAL)
(check-expect (f-on-key KEY3 . . .) KEY3-VAL)
(check-expect (f-on-key KEY4 . . .) KEY4-VAL)
...

;; Tests using sample values for f-on-key

(check-expect (f-on-key . . .) . . .)
...
19
Another popular choice is to use the up-arrow key.
61 The process-key Refinement 247

The data definition now contains four key varieties. Therefore, there are now four
sample instances, four sample expressions, four stanzas in the definition-template’s
conditional, and four tests using sample computations. The refined template clearly
informs us that there must be at least 4, not 3, of each of these items in the definition
of process-key.
Next, analyze the problem of processing the space key. When should a shot be
created? There can only be at most one shot in a world. This means a new shot
cannot be created every time the player tries to shoot. A decision must be made as
to whether to construct a new shot or not. The details of how this is done are left to
an auxiliary function, but there are two cases. If there is no shot in the game, then
a new shot is created and added to the world. Otherwise, a new shot is not created
and the shot in the world is left unchanged.
Start by refining the sample instances. It is logical to start with this refinement
because these constants may be used to simplify sample expression refinement and
development. The four needed sample instances may be defined as follows:
;; Sample instances of key
(define KEY1 "right")
(define KEY2 "left")
(define KEY3 "m")
(define KEY4 " ")
Next, refine the sample expressions for process-key. Recall that the new world
data definition tells us that worlds must be constructed with a shot and that the
template for a function on a key tells us there must be at least four sample expressions.
The sample expressions may be refined as follows:
;; Sample expressions for process-key
(define KEY-RVAL (make-world (move-rckt-right
(world-rocket INIT-WORLD))
(world-alien INIT-WORLD)
(world-dir INIT-WORLD)
(world-shot INIT-WORLD)))
(define KEY-LVAL (make-world (move-rckt-left
(world-rocket INIT-WORLD))
(world-alien INIT-WORLD)
(world-dir INIT-WORLD)
(world-shot INIT-WORLD)))
(define KEY-SVAL (make-world (world-rocket INIT-WORLD)
(world-alien INIT-WORLD)
(world-dir INIT-WORLD)
(process-shooting
(world-shot INIT-WORLD)
(world-rocket INIT-WORLD))))
(define KEY-SVAL2 (make-world (world-rocket INIT-WORLD2)
(world-alien INIT-WORLD2)
(world-dir INIT-WORLD2)
248 11 Aliens Attack Version 3

(process-shooting
(world-shot INIT-WORLD2)
(world-rocket INIT-WORLD2))))
(define KEY-OVAL INIT-WORLD2)
The sample expressions for "right" and "left", KEY-RVAL and KEY-LVAL, are
refined to build a world with an unchanged shot. The sample expression for the fourth
key variety, KEY-OVAL, remains unchanged because a world is not constructed. The
remaining sample expressions, KEY-SVAL and KEY-SVAL2, illustrate how a new
world is computed, respectively, when a world does not contain a shot and when
it contains a shot. In either case, as per the problem analysis above, the processing
of the shooting key is left to an auxiliary function. This function requires at least two
inputs: the world’s shot and the world’s rocket. The shot is needed to determine
if a new shot may be created. The rocket is needed to provide the image-x
coordinate if a new shot is created. If any further inputs are needed by this function
its design ought to reveal them.
Abstracting over the sample expressions confirms that process-key must be
refined to have a conditional with four stanzas. The refined function is:
;; world key → world
;; Purpose: Process a key event to return next world
(define (process-key a-world a-key)
(cond [(key=? a-key "right")
(make-world (move-rckt-right (world-rocket a-world))
(world-alien a-world)
(world-dir a-world)
(world-shot a-world))]
[(key=? a-key "left")
(make-world (move-rckt-left (world-rocket a-world))
(world-alien a-world)
(world-dir a-world)
(world-shot a-world))]
[(key=? a-key " ")
(make-world (world-rocket a-world)
(world-alien a-world)
(world-dir a-world)
(process-shooting
(world-shot a-world)
(world-rocket a-world)))]
[else a-world]))
Observe that there is now a stanza for the new key variety (i.e., " ") and that world
construction always includes a shot.
Tests using sample computations are refined by adding tests using KEY-SVAL and
KEY-SVAL2:
(check-expect (process-key INIT-WORLD " ") KEY-SVAL)
(check-expect (process-key INIT-WORLD2 " ") KEY-SVAL2)
61 The process-key Refinement 249

The other tests using sample computations remain unchanged. Tests using sample
values are updated to correctly construct worlds and to further illustrate how the
space key is processed. For example, the following are refined tests using sample
values:
;; Tests using sample values for process-key
(check-expect (process-key (make-world
(sub1 MAX-CHARS-HORIZONTAL)
INIT-ALIEN
'right
INIT-SHOT)
"right")
(make-world (sub1 MAX-CHARS-HORIZONTAL)
INIT-ALIEN
'right
INIT-SHOT))

(check-expect (process-key (make-world 0

INIT-ALIEN
'left
INIT-SHOT)
" ")
(make-world 0
INIT-ALIEN
'left
(make-posn 0 MAX-IMG-Y)))

* Ex. 107 — Refine all the process-key tests using sample values that are in
Aliens Attack 2.
* Ex. 108 — Add tests using sample values to further illustrate how the space
key is processed.

* Ex. 109 — There are repetitions in the process-key tests using sample
values. For example, the first test above creates the same world twice. Rewrite
the tests without repetitions. Make sure not to change what is being tested.

With the refinement of process-key completed focus turns to the design of the
auxiliary function to process the space key. This function takes as input two different
types of data: a shot and a rocket. Which template ought to be used to design
this function? To answer this question it is necessary to do careful problem analysis.
This function must create a new shot only when the game does not already have a
posn shot. When ought a new shot be created? A new shot is created only when
the game’s shot is 'NO-SHOT. This immediately suggests designing this function
by specializing the template for a function on a shot because its body distinguishes
among shot varieties.
250 11 Aliens Attack Version 3

Can this function be designed by specializing the template for a function on a

rocket? If you think carefully about this the answer is no. The given rocket does
not have enough information to decide whether or not a shot ought to be created.
This is a situation in which one input dominates another input. When facing such a
problem design around the dominating input.
Given that the dominating input is the given shot, the function template for a
function on a shot is specialized as follows:
;; shot rocket → shot
;; Purpose: To process a shooting attempt
(define (process-shooting a-shot a-rocket)
(if (eq? a-shot NO-SHOT)
(make-posn a-rocket MAX-IMG-Y)
a-shot))

;; Sample expressions for process-shooting

(define PS-SHOT-VAL1 (make-posn INIT-ROCKET MAX-IMG-Y))
(define PS-SHOT2-VAL SHOT2)
(define PS-SHOT3-VAL SHOT3)

;; Tests using sample computations for process-shooting

(check-expect (process-shooting INIT-SHOT INIT-ROCKET)
PS-SHOT-VAL1)
(check-expect (process-shooting SHOT2 INIT-ROCKET)
PS-SHOT2-VAL)
(check-expect (process-shooting SHOT3 INIT-ROCKET2)
PS-SHOT3-VAL)

;; Tests using sample values for process-shooting

(check-expect (process-shooting NO-SHOT 8)
(make-posn 8 MAX-IMG-Y))
(check-expect (process-shooting (make-posn 17 9) 8)
(make-posn 17 9))
The sample expressions illustrate the returned value for each shot variety. When it
is 'NO-SHOT a new posn shot is created using the rocket as the shot’s image-x
coordinate. Otherwise, a new shot is not created. The body of the definition returns
a new posn shot when the game does not have a shot. Otherwise, it returns the
given shot. The tests using sample computations illustrate that process-shooting
computes the same values as the sample expressions. The tests using sample val-
ues further illustrate that process-shooting works for both shot varieties. This
completes the design of process-key.
62 The process-tick Refinement 251

62 The process-tick Refinement

To refine process-tick, like any other problem, first perform problem analysis.
Outline what needs to happen when the clock ticks. In Aliens Attack 2 tick processing
created a new world by moving the alien and computing a direction. These tasks
must still be performed in this new Aliens Attack version. To construct a new world,
however, a shot instance is needed. Does a shot change after a clock tick? Once
again, reason about each shot variety. If the shot is 'NO-SHOT, then it does not
change. If the shot is a posn, then it either must move up or disappear because it has
reached the top of the scene. The details of how the new shot is computed is left
to an auxiliary function that processes a shot.

62.1 The Refinement

The sample expressions must now illustrate that the alien is moved, a new direction
is computed, and the shot is moved (if possible). The sample expressions are refined
as follows:
;; Sample expressions for process-tick
(define AFTER-TICK-WORLD1
(make-world
(world-rocket INIT-WORLD)
(move-alien (world-alien INIT-WORLD)
(world-dir INIT-WORLD))
(new-dir-after-tick (move-alien
(world-alien INIT-WORLD)
(world-dir INIT-WORLD))
(world-dir INIT-WORLD))
(move-shot (world-shot INIT-WORLD))))

(define AFTER-TICK-WORLD2
(make-world
(world-rocket INIT-WORLD2)
(move-alien (world-alien INIT-WORLD2)
(world-dir INIT-WORLD2))
(new-dir-after-tick (move-alien
(world-alien INIT-WORLD2)
(world-dir INIT-WORLD2))
(world-dir INIT-WORLD2))
(move-shot (world-shot INIT-WORLD2))))
Observe that, just as in Aliens Attack 2, the rocket is unchanged, the alien is moved,
and a new direction is computed. Moving the shot is left to an auxiliary function,
move-shot, that is to be designed after completing the process-tick refinement.
252 11 Aliens Attack Version 3

The refinement of process-tick’s definition is accomplished by adding the call

to move-shot:
;; world → world
;; Purpose: Create a new world after a clock tick
(define (process-tick a-world)
(make-world (world-rocket a-world)
(move-alien (world-alien a-world)
(world-dir a-world))
(new-dir-after-tick (move-alien
(world-alien a-world)
(world-dir a-world))
(world-dir a-world))
(move-shot (world-shot a-world))))
Observe that the specialization is similar to code refactoring. The function body
is restructured without changing the behavior of the function. When a function is
refined or refactored without changing its external behavior, the function’s signature
remains unchanged.
Given that the external behavior of the function is unchanged, the tests using
sample computations remain unchanged. The tests using sample values, on the other
hand, must be refined because worlds are built. The worlds built for the call to
process-tick must now contain a shot. Think carefully about the expressions
needed for the expected values. If there is no shot or the shot is at the top of the
scene in the given world, then the expected world must be built using 'NO-SHOT.
Otherwise, the shot is moved up. The refined tests are:
;; Tests using sample values for process-tick
(check-expect (process-tick (make-world INIT-ROCKET
(make-posn 1 5)
'left
INIT-SHOT))
(make-world INIT-ROCKET
(make-posn MIN-IMG-X 5)
'down
INIT-SHOT))

(check-expect (process-tick
(make-world
INIT-ROCKET2
(make-posn (- MAX-CHARS-HORIZONTAL 2) 10)
'right
SHOT2))
(make-world INIT-ROCKET2
(make-posn MAX-IMG-X 10)
'down
(move-shot SHOT2)))
62 The process-tick Refinement 253

(check-expect (process-tick (make-world INIT-ROCKET2

(make-posn MAX-IMG-X 2)
'down
(make-posn 15 6)))
(make-world INIT-ROCKET2
(make-posn MAX-IMG-X 3)
'left
(make-posn 15 5)))

(check-expect (process-tick (make-world

INIT-ROCKET2
(make-posn MIN-IMG-X 2)
'down
(make-posn 2 MIN-IMG-Y)))
(make-world INIT-ROCKET2
(make-posn MIN-IMG-X 3)
'right
NO-SHOT))
Remember that image-y coordinates grow from top to bottom. Thus, moving the
shot is accomplished by decreasing its image-y coordinate.

62.2 The move-shot Design

Writing tests for process-tick has provided an insight into how to move a shot: a
posn shot is moved by decreasing its image-y coordinate. It is still necessary to
perform problem analysis. There are two shot varieties. How is a 'NO-SHOT shot
moved? If there is no shot in the game, then there is nothing to move. Therefore, the
shot remains unchanged. Now, carefully think about how a posn shot is moved.
Does a shot always move up? If it did, then the player would only be able to
shoot once during the game. This, however, is not what is needed. We wish to
allow the player to shoot again when a shot goes off the top of the scene. How
does this affect how a shot is moved? In order to allow the player to shoot again
a posn shot must become a 'NO-SHOT shot when it moves off the top of the
scene. This analysis suggests that there are three conditions that must be tested:
(eq? a-shot NO-SHOT) → shot remains unchanged
(= (posn-y a-shot) MIN-IMG-Y) → shot changes to 'NO-SHOT
(not (= (posn-y a-shot) MIN-IMG-Y)) → shot’s image-y is decreased
Armed with the insights provided by problem analysis, the template for functions
on a shot is specialized. We start with sample expressions. Given that there are three
conditions at least 3 sample expressions need to be defined:
254 11 Aliens Attack Version 3

;; Sample expressions for move-shot

(define MSHOT1-VAL INIT-SHOT)
(define MSHOT2-VAL (make-posn
(posn-x SHOT2)
(move-up-image-y (posn-y SHOT2))))
(define MSHOT4-VAL NO-SHOT)
The first sample expression is for a 'NO-SHOT shot. The second sample expression
is for a posn shot that may move up. The third sample expression is for a posn
shot that may not be moved up. Observe that separation of concerns is maintained
by requiring a new auxiliary function to move an image-y instance up. The design
of this new function is delayed until the design of move-shot is completed.
The tests for move-shot are as follows:
;; Tests using sample computations for move-shot
(check-expect (move-shot INIT-SHOT) MSHOT1-VAL)
(check-expect (move-shot SHOT2) MSHOT2-VAL)
(check-expect (move-shot SHOT4) MSHOT4-VAL)

;; Tests using sample values for move-shot

(check-expect (move-shot (make-posn 5 5)) (make-posn 5 4))
(check-expect (move-shot (make-posn 5 MIN-IMG-Y)) NO-SHOT)
The tests using sample computations use, as expected, the values of the sample
expressions. The tests using sample values further illustrate how posn shots are
moved. They clearly show when the image-y coordinate is decreased and when a
posn shot becomes a 'NO-SHOT shot.
The definition template may now be specialized. The conditional must have three
stanzas (not two as suggested by the template):
;; shot → shot
;; Purpose: To move the given shot
(define (move-shot a-shot)
(cond [(eq? a-shot NO-SHOT) a-shot]
[(= (posn-y a-shot) MIN-IMG-Y) NO-SHOT]
[else (make-posn (posn-x a-shot)
(move-up-image-y (posn-y a-shot)))]))
Observe that the first and second stanzas always return the same value: 'NO-SHOT.
This means that the function may be refactored in an attempt to simplify it. The first
two stanzas may be combined into one stanza. The resulting function is:
;; shot → shot
;; Purpose: To move the given shot
(define (move-shot a-shot)
(cond [(or (eq? a-shot NO-SHOT)
(= (posn-y a-shot) MIN-IMG-Y))
NO-SHOT]
62 The process-tick Refinement 255

[else (make-posn (posn-x a-shot)

(move-up-image-y (posn-y a-shot)))]))
This simplification step is not necessary. Such simplification steps ought to be
performed only if you feel it makes the program easier to understand.
We move onto solving the problem of moving an image-y value up. Moving up
simply means decreasing the value of the given image-y. Specializing the template
for a function on an image-y is fairly straightforward. The result of the specialization
is:
;; image-y>min → image-y
;; Purpose: To move the given image-y>min up
(define (move-up-image-y an-img-y>min)
(sub1 an-img-y>min))

;; Sample expressions for move-up-image-y

(define IMGY>MIN-VAL1 (sub1 IMG-Y>MIN1))
(define IMGY>MIN-VAL2 (sub1 IMG-Y>MIN2))

;; Tests using sample computations for move-up-image-y

(check-expect (move-up-image-y AN-IMG-Y) IMGY>MIN-VAL1)
(check-expect (move-up-image-y MAX-IMG-Y) IMGY>MIN-VAL2)

;; Tests using sample values for f-on-image-y>min

(check-expect (move-up-image-y 6) 5)
(check-expect (move-up-image-y 11) 10)
At this point, you have hopefully developed enough expertise with template special-
ization that a detailed explanation for the design of move-up-image-y is no longer
necessary. Nonetheless, make sure you understand how each step of the design recipe
is satisfied. It is important to note, however, that the signature of the function is care-
fully crafted. Observe that the input is not an image-y value. Instead, it is a subtype
of image-y: the image-y instances greater than MIN-IMG-Y. This is necessary to
guarantee that the returned value is always an image-y instance.

* Ex. 110 — The design of move-shot may more strictly adhere to the tem-
plate for a function on a shot by not changing the number of conditions from
2 to 3. The following is an alternative design using the conditional to strictly
discriminate among the shot varieties:
;; shot → shot
;; Purpose: To move the given shot
(define (move-shot a-shot)
(cond [(eq? a-shot NO-SHOT) a-shot]
[else
(cond [(= (posn-y a-shot) MIN-IMG-Y) NO-SHOT]
[else (make-posn
256 11 Aliens Attack Version 3

(posn-x a-shot)
(move-up-image-y (posn-y a-shot)))]))
This design clearly communicates how each shot variety is processed. The
processing of a posn shot requires a conditional expression. Is this design
easier to understand? Do you prefer this design or the design that does not have
a nested conditional expression?

63 The game-over? Refinement

By now, it is likely clear that the game may end in two ways. The first, as before,
occurs when the alien reaches earth and the player loses. The second occurs when
the alien is hit by a shot and the player wins.
The above problem analysis suggests that at least three sample expressions are
needed. One is needed to illustrate when the game is not over. Two are needed to
illustrate when the game ends. One for when the player loses and one for when the
player wins. For example, the following are sample expressions that satisfy these
constraints:
;; Sample expressions for game-over?
(define GAME-OVER1 (or (alien-reached-earth?
(world-alien INIT-WORLD2))
(hit? (world-shot INIT-WORLD2)
(world-alien INIT-WORLD2))))

(define GAME-OVER2 (or (alien-reached-earth?

(world-alien WORLD3))
(hit? (world-shot WORLD3)
(world-alien WORLD3))))

(define GAME-NOT-OVER (or (alien-reached-earth?

(world-alien INIT-WORLD))
(hit? (world-shot INIT-WORLD)
(world-alien INIT-WORLD))
The first sample expression is written using a world in which the alien has reached
earth. The second sample expression is written using a world in which the alien
has been hit by a shot. The third sample expression is written using a world for
which the game is not over.
The tests are written as follows:
;; Tests using sample computations for game-over?
(check-expect (game-over? INIT-WORLD2) GAME-OVER1)
(check-expect (game-over? WORLD3) GAME-OVER2)
(check-expect (game-over? INIT-WORLD) GAME-NOT-OVER)
63 The game-over? Refinement 257

;; Tests using sample values for game-over?

(check-expect (game-over? (make-world 8
(make-posn 0 3)
'right
NO-SHOT))
#false)
(check-expect (game-over? (make-world
8
(make-posn 0 MAX-IMG-Y)
'right
(make-posn 12 11)))
#true)
(check-expect (game-over? (make-world 8
(make-posn 0 5)
'right
(make-posn 0 5)))
#true)
As expected, there is a test using each value computed from the sample expressions.
The tests using sample values further illustrate when a world has ended and when
it has not ended. The first builds a world that has not ended. The second builds a
world that has ended because the alien has reached earth. The third builds a world
that has ended because the alien has been hit by the shot.
The function definition is refined using or, as in the sample expressions, to detect
each of the possible ways the game is over. The refined function is:
;; world → Boolean
;; Purpose: Detect if the game is over
(define (game-over? a-world)
(or (alien-reached-earth? (world-alien a-world))
(hit? (world-shot a-world) (world-alien a-world))))
Observe that the problems of determining either way the world may end are delegated
to auxiliary functions. As before, alien-reached-earth? is used to detect when
the player has lost the game. On the other hand, hit? needs to be designed and
written to detect when the player has won the game. It is straightforward to observe
that determining a hit requires at least an alien and a shot.

63.1 The hit? Design

We must now analyze how to determine if a given alien is hit by a given shot. A
'NO-SHOT shot can never hit an alien. This means that the given shot must be
a posn shot to have a hit. If the given shot is a posn shot, then its image-x
and image-y coordinates must be the same as those of the given alien. Observe
258 11 Aliens Attack Version 3

that nothing in this problem analysis suggests that a decision must be made. Thus,
a conditional is not needed. Given this observation, we proceed by specializing the
template for a function on an alien (and not the template for a function on a shot
that contains a conditional).
We start by writing sample expressions for hit?. The idea is to write at least
two sample expressions: one that returns #false and one that returns #true. The
following are two such sample expressions:
(define ALIEN3 (make-posn 4 MAX-IMG-Y))

;; Sample expressions for hit?

(define NHIT-VAL (and (posn? INIT-SHOT)
(= (posn-x INIT-SHOT)
(posn-x INIT-ALIEN))
(= (posn-y INIT-SHOT)
(posn-y INIT-ALIEN))))
(define HIT-VAL (and (posn? SHOT3)
(= (posn-x SHOT3)
(posn-x ALIEN3))
(= (posn-y SHOT3)
(posn-y ALIEN3))))
The first sample expression illustrates an alien and a shot that have not hit. For the
sample expression that illustrates a hit an alien and a shot with same coordinates
are needed. None of the sample aliens and shots has this characteristic. Therefore,
a new sample alien is defined with the same coordinates as SHOT3. Observe that
for NHIT-VAL no error is thrown by trying to extract coordinates from INIT-SHOT.
This occurs because INIT-SHOT is not a posn. Given that (posn? INIT-SHOT) is
#false the other inputs to and are never evaluated as explained in Chap. 2. Thus,
no error is thrown.
The following are the tests for hit?:
;; Tests using sample computations for hit?
(check-expect (hit? INIT-SHOT INIT-ALIEN) NHIT-VAL)
(check-expect (hit? SHOT3 ALIEN3) HIT-VAL)

;; Tests using sample values for hit?

(check-expect (hit? (make-posn 0 0) (make-posn 0 0)) #true)
(check-expect (hit? (make-posn 15 6) (make-posn 1 2)) #false)
Both sets of tests illustrate when there is and there is not a hit. The tests using sample
values make it clear to any reader of the code that the coordinates must be equal to
have a hit.
The function definition is developed by abstracting over the sample expressions.
This abstraction yields:
;; shot → Boolean
;; Purpose: To determine if the given shot has hit the
;; given alien
63 The game-over? Refinement 259

(define (hit? a-shot an-alien)

(and (posn? a-shot)
(= (posn-x a-shot) (posn-x an-alien))
(= (posn-y a-shot) (posn-y an-alien))))

63.2 The draw-last-world Refinement

The computation of the last scene must also be refined. Now, the player may win
and, if so, a winning message ought to be displayed. To start, it is best to define a
sample final world representing a player win such as:
(define FWORLD3 (make-world 7 (make-posn 19 3)
'left (make-posn 19 3)))
Now, think carefully. How do we decide to display a winning or a losing message?
How do we know that FWORLD1 and FWORLD2 represent a loss for the player and
FWORLD3 represents a win for the player? As before, it is a loss if the image-y of the
alien is MAX-IMG-Y. It is a win if the alien is hit.
The sample expression using FWORLD3 is:
(define FWORLD3-VAL (place-image
(text "EARTH WAS SAVED!" 36 'green)
(/ E-SCENE-W 2)
(/ E-SCENE-H 4)
(draw-world FWORLD3)))
Observe that the message is different from the other two sample expressions in Aliens
Attack 2. The development of a third sample expression means that a test using this
new sample value must be added:
(check-expect (draw-last-world FWORLD3) FWORLD3-VAL)
The other tests using sample values remain unchanged. The test using a sample value
is changed to build a world that contains a shot. It is still used to test the error
thrown. The refined test is:
;; Tests using sample values for draw-last-world
(check-error
(draw-last-world (make-world 10
(make-posn 7 20)
'right
(make-posn 2 3)))
"draw-last-world: Invalid world with #(struct:posn 7 20)
as the alien value and #(struct:posn 2 3) as the shot value.")
Observe that the world constructed does not represent a game-ending world and,
thus, is expected to force draw-last-world to throw an error.
260 11 Aliens Attack Version 3

The function definition refinement adds a stanza to the conditional to determine

if the player has won. Clearly, this is when the alien is hit. If so, an EARTH WAS
SAVED! message is displayed. The refined function is:
;; world → scene throws error
;; Purpose: To draw the game’s final scene
(define (draw-last-world a-world)
(cond [(= (posn-y (world-alien a-world)) MAX-IMG-Y)
(place-image (text "EARTH WAS CONQUERED!" 36 'red)
(/ E-SCENE-W 2)
(/ E-SCENE-H 4)
(draw-world a-world))]
[(hit? (world-shot a-world) (world-alien a-world))
(place-image (text "EARTH WAS SAVED!" 36 'green)
(/ E-SCENE-W 2)
(/ E-SCENE-H 4)
(draw-world a-world))]
[else (error
(format
"draw-last-world: Invalid world with ~s as the
alien value and ~s as the shot value."
(world-alien a-world)
(world-shot a-world)))]))
This completes the refinement for Aliens Attack 3. Good job! Run and enjoy the
game!

* Ex. 111 — Change the data definition of a key to:

A key is either:
1. "right"
2. "left"
3. "up"
4. not "right", "left", or "up"
Refine the code to use this new data definition.
**** Ex. 112 — Change the data definition of an alien to be:
An alien is a structure: (make-alien posn integer>=0).
The nonnegative integer represents the health of the alien. That is, the number
of times the alien must be hit before it is destroyed. Carefully analyze how the
handlers must be refined before editing the code.
64 What Have We Learned in This Chapter? 261

64 What Have We Learned in This Chapter?

The important lessons of Chap. 11 are summarized as follows:

• When a representation choice is arbitrary you may choose any (simple) rep-
resentation. For example, a non-existing shot is represented by a symbol (i.e.,
'NO-SHOT).
• Two different types may have the same representation like a posn is used to
represent both an alien and a shot.
• The operations valid on an instance of a type do not depend on the representation.
For example, an alien may move left, right, and down while a shot may only
move up.
• The implementation of an operation may vary for each variety. For example, a
'NO-SHOT shot is not drawn, but a posn shot is drawn.
• If a function has an input of type A that dominates the other inputs, design the
function by specializing the template for a function on A.
• Sometimes a function to process a type with variety may need to distinguish more
conditions beyond the number of varieties. For instance, move-shot has three,
not two, conditions.
• Refactoring may simplify well-designed code.
• Sometimes defining new instances of a type makes writing tests easier.
 Part III
Compound Data of Arbitrary Size
Chapter 12
Lists

Congratulations! You are no longer a beginning student. Change your language

in DrRacket to Intermediate Student with lambda. We shall refer to this
language as ISL+. All your programs written in BSL will still work in ISL+. In fact,
BSL is a subset of ISL+.
In this chapter, we begin to study data of arbitrary size. By data of arbitrary size,
we mean data whose size is not constant. In a practical sense, this means that we
shall begin to learn how to solve problems involving compound data that may have
a varying number of elements. This does not mean that data may be of infinite size.
Every instance of our data will be of finite size. Unlike structures, however, not every
instance shall be of the same size.
Data of arbitrary size is a natural part of human life. For example, it is typical for
persons to make a grocery list before going to the supermarket. My grocery list, of
course, may very well be different than your grocery list. This is no surprise given
that we may have different tastes. For example, a vegetarian’s grocery list will not
contain animal proteins while a non-vegetarian will contain animal proteins. Beside
content, there is another important difference between these lists. They may contain
a different number of items. For example, one grocery list may contain milk, walnuts,
and bread while another grocery list may contain sliced turkey, milk, bread, onions,
and salmon. Yet, someone else’s grocery list may be empty. Regardless of the size,
we all agree that they are valid grocery lists. That is a grocery list is data of arbitrary
size.
Lists may be used to represent much more than just the groceries that we need to
buy. For example, in Aliens Attack we would like to have more than one alien and
we would like to allow the player to shoot more than one shot at a time. Consider
carefully what it means to have multiple aliens and multiple shots in Aliens Attack. Is
the number of aliens constant? Is the number of shots constant? Clearly, the answer
to both questions is no. As the game advances the number of aliens decreases as they
are hit. That is, the number of aliens is not constant. Similarly, the number of shots
varies. At the beginning of the game, the number of shots is 0. Every time the player
shoots the number of shots increases. Every time a shot hits an alien or goes off the
scene, the number of shots decreases. This tells us that aliens and shots are data of
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 265
M. T. Morazán, Animated Problem Solving, Texts in Computer Science,
https://doi.org/10.1007/978-3-030-85091-3_12
266 12 Lists

arbitrary size just as a grocery list. Thus, we may consider using a list to represent
the aliens and the shots.

65 Creating and Accessing Lists in ISL+

Given that many different types of data are naturally represented as lists, ISL+
provides support for lists. The smallest possible list is the empty list. In ISL+, the
empty list is represented as '(). We can use the empty list to define many different
types of list. For example, an empty grocery list, an empty list of aliens, and an
empty list of shots may be defined as follows:
(define E-GLIST '())
(define E-LOA '())
(define E-LOS '())

(check-expect (empty? E-GLIST) #true)

(check-expect (empty? E-LOA) #true)
(check-expect (empty? E-LOS) #true)
In essence, '() is an ISL+ constant that represents any type of empty list. It is the
basis of all lists. The ISL+ predicate empty? returns true if its input is '() and
#false otherwise.
Non-empty lists, may be constructed by adding an element to the front of an
existing list. Just like constructors are needed to create structures, a constructor is
needed to build a new list. In ISL+ this function is cons (for construct). The signature
and purpose statement for cons are the following:
any list → list
Purpose: To construct a new list by adding the given value
to the front of the given list
The contract is stating that the first argument may be of any type, that the second
argument must be a list (of any type), and that it returns a list (of any type). This
means that cons is a generic function that may be used to construct any type of list.
We can now use cons to build a grocery list, a list of aliens, and a list of shots of
any (finite) size. For example, a grocery list with 2 elements, a list of 5 aliens, and
list of 3 shots may be defined as follows:
(define A-GLIST (cons "milk"
(cons "apples"
E-GLIST)))
(define A-LOA (cons (make-posn 10 2)
(cons (make-posn 5 12)
(cons (make-posn 4 8)
(cons (make-posn 15 7)
(cons (make-posn 6 6)
E-LOA))))))
65 Creating and Accessing Lists in ISL+ 267

(define A-LOS (cons (make-posn 4 2)

(cons NO-SHOT
(cons (make-posn 17 3)
E-LOS))))

(check-expect (cons? E-GLIST) #false)

(check-expect (cons? E-LOA) #false)
(check-expect (cons? E-LOS) #false)
(check-expect (cons? A-GLIST) #true)
(check-expect (cons? A-LOA) #true)
(check-expect (cons? A-LOS) #true)
The arguments to cons are lined up for you to easily confirm that its second argument
is always a list. That is, the second argument is always a list constructed using cons
or is the empty list (i.e., E-GLIST, E-LOA, or E-LOS). The first argument varies
depending on the type of list constructed. For the grocery list, the first argument is
always a string. For a list of aliens, the first argument is always an alien. For the list
of shots, the first argument is always a shot. Now, it should be even clearer that lists,
just as structures, are generic. The tests use the ISL+ predicate cons? that returns
true when its input is a non-empty list. Otherwise, it returns false.
As with structures, selectors are needed to extract the elements used to construct
a list. Given that a non-empty list is always constructed using cons, two selectors are
needed to extract the components of a list. The selector to extract the first element of
a given list is first. The selector to extract the list containing the rest of the elements
(not including the first element) of a given list is rest. Care must be taken when
using these selector functions because the empty list does not have a first element
nor does it have a value for the rest of the elements. If either list-selector function is
applied to the empty list, an error is thrown. The following tests illustrate how the
list-selector functions work:
(check-error (first E-GLIST))
(check-error (first E-LOA))
(check-error (first E-LOS))

(check-expect (first A-GLIST) "milk")

(check-expect (first A-LOA) (make-posn 10 2))
(check-expect (first A-LOS) (make-posn 4 2))

(check-error (rest E-GLIST))

(check-error (rest E-LOA))
(check-error (rest E-LOS))

(check-expect (rest A-GLIST) (cons "apples" E-GLIST))

268 12 Lists

(check-expect (rest A-LOA)

(cons (make-posn 5 12)
(cons (make-posn 4 8)
(cons (make-posn 15 7)
(cons (make-posn 6 6)
E-LOA)))))
(check-expect (rest A-LOS)
(cons NO-SHOT (cons (make-posn 17 3) E-LOS)))
Here a special form of check-error is used that only checks if an error is thrown. It
does not require an error message to check. This form of check-error is particularly
useful when testing functions whose error messages we did not design.
How is the second element of a list extracted? The third element? To achieve
extractions beyond the first list element you need to compose calls to first and
rest. For example, the second element of A-LOS, NO-SHOT, is given by (first
(rest A-LOS)). Let us examine how this works:
(first (rest A-LOA)) = (first
(rest (cons (make-posn 4 2)
(cons NO-SHOT
(cons
(make-posn 17 3)
E-LOS)))
= (first (cons NO-SHOT
(cons (make-posn 17 3)
E-LOS)))
= NO-SHOT
The first step substitutes A-LOA for its value. In the second step, the inner most
function application (using rest) is evaluated and is replaced with its value. The
third step applies the outermost function application to obtain NO-SHOT.

* Ex. 113 — Design and implement a function, third-of-list, to extract

the third element of a list. Use compositions of first and rest.

* Ex. 114 — Design and implement a function, fourth-of-list, to extract

the fourth element of a list. Use compositions of first and rest.

* Ex. 115 — Design and implement a function, fifth-of-list, to extract

the fifth element of a list. Use compositions of first and rest.

** Ex. 116 — Design and implement a function, tenth-of-list, to extract

the tenth element of a list. Use compositions of first and rest.

Extracting the second, third, fourth, and so on up to the eighth element of a list
is common enough that ISL+ provides functions to do so directly. Not surprisingly,
these functions are called second, third, fourth, and so on up to eighth. Be
mindful when using these functions because if given a list that is too short they
throw an error. These function provide a useful shorthand notation to access list
elements up to the eighth. From the ninth on you must write your own function.
66 Shorthand for Building Lists 269

66 Shorthand for Building Lists

Consider constructing a list of first 7 digits. Your code may look something like this:
(define SEVEN-DIGITS
(cons
0
(cons 1
(cons 2
(cons 3
(cons 4
(cons 5
(cons 6 empty))))))))
(check-expect (first SEVEN-DIGITS) 0)
(check-expect (second SEVEN-DIGITS) 1)
(check-expect (third SEVEN-DIGITS) 2)
(check-expect (fourth SEVEN-DIGITS) 3)
(check-expect (fifth SEVEN-DIGITS) 4)
(check-expect (sixth SEVEN-DIGITS) 5)
(check-expect (seventh SEVEN-DIGITS) 6)
Observe that there is a lot of repetition in the code to construct the list. Specifically,
there is an application of cons for every element of the list. This may not be too
cumbersome for a list with seven elements, but what if you are now asked to define
a list with the integers in [0..19]? You would need to write cons twenty times. To
avoid all this repetition ISL+ provides three shorthand constructors for lists.
The first is used when the elements of the list are known in advance. A quoted
list has the elements listed inside parenthesis preceded by a '. For example, we may
refactor the SEVEN-DIGITS definition as follows:
(define SEVEN-DIGITS '(0 1 2 3 4 5 6))

(check-expect (first SEVEN-DIGITS) 0)

(check-expect (second SEVEN-DIGITS) 1)
(check-expect (third SEVEN-DIGITS) 2)
(check-expect (fourth SEVEN-DIGITS) 3)
(check-expect (fifth SEVEN-DIGITS) 4)
(check-expect (sixth SEVEN-DIGITS) 5)
(check-expect (seventh SEVEN-DIGITS) 6)
Using a quoted list eliminates the need to repeatedly write cons to construct the
list. It is important to note that nothing after the quote (inside the parenthesis) is
evaluated. That means all the listed elements are literal values in the list. This is
important because there can be no expressions that need to be evaluated inside the
parenthesis. For example, consider the following code:
270 12 Lists

(define SOME-SQRS1 (cons (sqr 0)

(cons (sqr 1)
(cons (sqr 2)
empty))))

(define SOME-SQRS2 '((sqr 0) (sqr 1) (sqr 2)))

(check-expect SOME-SQRS1 '(0 1 4))

(check-expect (not (equal? SOME-SQRS1 SOME-SQRS2)) #true)
Both tests pass. For SOME-SQRS1 the expressions using sqr are evaluated. For
SOME-SQRS2 they are not evaluated because they are inside a quoted list. Instead of
being evaluated, they are considered literal sublists. Thus, the value of SOME-SQRS2
is:
(cons (cons 'sqr (cons 0 empty))
(cons (cons 'sqr (cons 1 empty))
(cons (cons 'sqr (cons 2 empty))
empty)))
Naturally, you must be asking yourself if there is shorthand notation to create a
list from evaluated expressions. After all, it would be nice not to repeatedly write
cons to define SOME-SQRS1. The shorthand provided by ISL+ is list. This is a
list-constructor function that evaluates all of its arguments and creates a list of the
results. The number of arguments is arbitrary. If no arguments are provided list
returns '(). We can refactor the SOME-SQRS2 definition and update the tests as
follows:
(define SOME-SQRS1 (cons (sqr 0)
(cons (sqr 1)
(cons (sqr 2)
empty))))

(define SOME-SQRS2 (list (sqr 0) (sqr 1) (sqr 2)))

(check-expect SOME-SQRS1 '(0 1 4))

(check-expect (equal? SOME-SQRS1 SOME-SQRS2) #true)
Use list when you want all the expressions to be evaluated to construct a list.
The third shorthand is a quasiquoted list. A quasiquoted list is a combination of '
and list. Instead of preceding the opening parentheses with a ', it is preceded with
`. The ` indicates that some subexpression inside the parenthesis may have to be
evaluated. To indicate that an expression needs to be evaluated, it must be preceded
by a ,. If there are no expressions preceded by a comma inside the parenthesis, then
` and ' yield the same value when evaluated. To illustrate the use of quasiquote
consider the following code:
(define X 2)
(define Y 3)
(define Z 4)
67 Recursive Data Definitions 271

(define A-LIST '((add1 X) J (* Z Y)))

(define A-LIST2 `(,(add1 X) J ,(* Z Y)))

(check-expect (first A-LIST) '(add1 X))

(check-expect (first A-LIST2) 3)

(check-expect (second A-LIST) 'J)

(check-expect (second A-LIST2) 'J)

(check-expect (third A-LIST) '(* Z Y))

(check-expect (third A-LIST2) 12)
The expressions (add1 X) and (* Z Y) are evaluated to construct A-LIST2 be-
cause they are preceded by a comma in a quasiquoted list. The same expressions are
treated as literal values for A-LIST because they are inside a quoted list.

67 Recursive Data Definitions

We have explored how to create and access lists. As part of our exploration, we have
discovered that lists may be used to store different types of data. We have not yet
explored how to process a list. To do so we first need to be able to develop a data
definition and a function template for a given type of list. Consider defining a list of
numbers. Let us look at some examples based on our knowledge of lists so far:
'()
(cons 87 '())
(cons 24 (cons 87 '()))
(cons 16 (cons 24 (cons 87 '())))
(cons 31 (cons 16 (cons 24 (cons 87 '()))))
If you think about it carefully, you can see a distinct pattern. There are two list-of-
numbers varieties. The first is the empty list of numbers. The second is a non-empty
list of numbers constructed using cons. This immediately suggests that to define a
list of numbers, we need a data definition with two varieties:
;; A list of numbers (lon) is either
;; 1. '()
;; 2. (cons ??? ???)
Now, we must be precise about the types of the arguments to cons. Given that we
are defining a list of numbers, it is clear that the first argument must be a number.
This takes us a step closer to the needed data definition:
272 12 Lists

;; A list of numbers (lon) is either

;; 1. '()
;; 2. (cons number ???)
What is the type of the second argument to cons? To answer this question, let
us analyze one of our examples: '(cons 31 (cons 16 (cons 24 (cons 87
'())))). In this example, 31 is the first number of the list. Clearly, it fulfills
what is required by the data definition we have so far. What is '(cons 16 (cons
24 (cons 87 '())))? If you think about it carefully, the inescapable conclusion
is that it is a lon. Is this really the case? Does this make sense? Let us look at another
example: '(cons 87 '()). Here the argument types to cons are number (the 87)
and, once again, a lon ('()). You can verify that all other examples satisfy this
pattern. This analysis is telling us that the type of the second argument to cons is
lon. Thus, the lon data definition is:
;; A list of number (lon) is either
;; 1. '()
;; 2. (cons number lon)
A list of numbers is defined in terms of itself. That is, it is a circular data definition.
A type defined in terms of itself is a recursive data type.
Does this make any sense? Can this data definition be used to build any list of
numbers? It turns out that you can build any list of numbers using this data definition.
To build the empty list of numbers, we simply use the first rule and write: '(). How
about a list that is not empty? Here is how you build '(cons 24 (cons 87 '())):
lon → (cons 24 lon) using rule 2 to substitute lon
→ (cons 24
(cons 87 lon)) using rule 2 to substitute lon
→ (cons 24
(cons 87 empty)) using rule 1 to substitute lon
This is an example of a derivation. A derivation is the series of steps used to show
how to build an instance of a data type using the varieties of a data definition. It is
not difficult to see that any valid list of numbers can be derived using the above data
definition. Can anything that is not a list of numbers be derived? Let us try to derive
'(cons 42 (cons -6 (cons "Hi" (cons 87 '())))):
lon → (cons 42 lon) using rule 2 to substitute lon
→ (cons 42
(cons -6 lon)) using rule 2 to substitute lon
The derivation fails because there is no lon variety whose first element is a string.
Therefore, '(cons 31 (cons 16 (cons "Hi" (cons 87 '())))) is not a lon.
It is also not difficult to see that according to our data definition all the elements of
a lon must be numbers. If any element of a given list is not a number, then it is not
a lon.
It is very likely that you have been taught to steer away from recursive data
definitions. This is unfortunate because as we have seen recursive data definitions
67 Recursive Data Definitions 273

are useful to define data of arbitrary size like lists. In fact, any data of arbitrary size
requires a recursive data definition. So, why are students steered away from recursive
data definitions? This is likely rooted in the fact that recursive data definitions may be
nonsense. For example, someone may attempt to define a list of numbers as follows:
; A list of numbers (lon2) is a (cons number lon2)
Is this a useful data definition? Let us try to derive '(cons 24 (cons 87 '())):
lon2 → (cons 24 lon2) substitute lon2 using rule 2
→ (cons 24
(cons 87 lon2)) substitute lon2 using rule 2
The derivation fails because we are unable to instantiate the lon2 in (cons 87
lon2). You may say that clearly it should be '(). In fact, this would be wrong.
Nowhere in this data definition does it say that '() is a lon2. It is important to be
precise with our data definitions. Otherwise, we will be unable to solve problems.
This leads to asking ourselves what constitutes a useful recursive data definition. In
order to be useful, a recursive data definition must have the following characteristics:
1. At least two subtypes (varieties)
2. At least one subtype that does not contain a selfreference
3. At least one subtype that contains a selfreference
The subtypes that do not contain a selfreference are known as base subtypes.
These are concrete values that are known to be instances of the data type defined.
These concrete values break the circularity in a derivation. The subtypes that do
contain a selfreference are known as recursive subtypes. These are the varieties
that introduce circularity which endows us with the power to define data of arbitrary
size. In the recursive data definition for lon, we have '() as a base subtype and
(cons number lon) as a recursive subtype. The recursive data definition for lon2
is not useful because it does not have a base subtype.
With our newly acquired knowledge, we can now define a representation for
multiple aliens and multiple shots in Aliens Attack. Given that both are data of
arbitrary size, we need a recursive data definition for each. At this point, the only
recursive data type we know is list. Thus, let us try to use a list to define an arbitrary
number of aliens. Remember that it must have the three characteristic above required
for a useful recursive data definition. A list of aliens may be defined as follows:
;; A list of alien (loa) is either:
;; 1. '()
;; 2. (cons alien loa)

;; Sample instances of loa

(define E-LOA '())
(define INIT-LOA
(list
(make-posn 8 0) (make-posn 9 0) (make-posn 10 0)
(make-posn 11 0) (make-posn 12 0) (make-posn 13 0)
274 12 Lists

(make-posn 8 1) (make-posn 9 1) (make-posn 10 1)

(make-posn 11 1) (make-posn 12 1) (make-posn 13 1)
(make-posn 8 2) (make-posn 9 2) (make-posn 10 2)
(make-posn 11 2) (make-posn 12 2) (make-posn 13 2)))
Observe that the loa data definition has two subtypes: one base subtype and one
recursive subtype. In addition, following the steps of the design recipe, an instance
of each variety of loa is defined. The initial list of aliens, INIT-LOA, contains 18
aliens. This ought to make the game more interesting. You are encouraged to make
the initial list of aliens shorter or longer to personalize the game to your liking.
A list of shots is defined as follows:
;; A list of shot (los) is either:
;; 1. '()
;; 2. (cons shot los)

;; Sample instances of los

(define INIT-LOS '())
(define LOS2 (list (make-posn 8 0) (make-posn 10 5)))
Observe that the requirements for a useful data definition are satisfied and that sample
instances are also defined.
* Ex. 117 — Define sample instances for lon.

* Ex. 118 — Create a data definition and sample instances for a grocery list.

* Ex. 119 — In Chap. 7 a data definition for a 2D-point is developed. Why or

why not is this a useful recursive data definition?

*** Ex. 120 — Create a data definition for a composed image. A composed
image may contain one or more rectangles and circles that are above, next, or
overlaid in relation to each other.

68 Generic Data Definitions

The next natural step is to develop a function template for lon, loa, and los.
However, notice that the three data definitions are almost identical. In other words,
there is a lot of repetition among them. When we have repetition among expressions,
an abstraction step introduces a variable to obtain a function. Can we apply an
abstraction step to avoid repetitions among data definitions?
The situation we face is a bit different from abstraction over expressions. When
you abstract over expressions, a difference is always a value and a variable is used
to represent the value. This felt quite natural because you are familiar with variables
from your Mathematics courses. The difference among the data definitions for lon,
loa, and los is a type. For lon the type used is number. For loa the type used
69 Function Templates for Lists 275

is alien. For los the type used is shot. When abstracting over data definitions
the variables used to capture the differences are type variables. Type variables
represent types not instances of types (i.e., values). A type variable is needed for
each difference.
Let X be the single difference among lon, loa, and los. Instead of using a
concrete type (like number, alien, or shot), we write a data definition using X:
;; A list of X ((listof X)) is either:
;; 1. '()
;; 2. (cons X (listof X))
The data definition is recursive just like the data definitions for lon, loa, and los.
Observe that if we plug in number for X we obtain the data definition for lon.
Similarly, plugging in alien and shot for X yields, respectively, the data definitions
for loa and los. The notation (listof X) is used to emphasize that a type must be
plugged in to obtain a concrete data definition. Clearly, this data definition works for
many different types. A data definition that works for many different types is called
a generic (or parameterized) data definition.
As we shall see, a generic data definition may be used in two ways in our signatures.
We may substitute X with a concrete type to specify a concrete data definition or we
may use (listof X) directly. We shall start with examples of the former. The use
of the latter will follow after we learn how to abstract over functions.

69 Function Templates for Lists

We can write a function template for a (listof X) and use it to define the templates
for lon, loa, and los. Before that, however, we must learn to write a function
template for a recursive data definition. Using the knowledge you have accumulated
so far, we can begin to develop the function template for (listof X):
;; Sample instances of (listof X)
(define LOX1 . . .)
(define LOX2 . . .)

;; (listof X) . . . → . . .
;; Purpose: . . .
(define (f-on-loX a-loX . . .)
(if (empty? a-loX)
...
. . .))

;; Sample expressions for f-on-loX

(define LOX1-VAL . . .)
(define LOX2-VAL . . .)
...
276 12 Lists

;; Tests using sample computations for f-on-loX

(check-expect (f-on-loX LOX1 . . .) LOX1-VAL)
(check-expect (f-on-loX LOX2 . . .) LOX2-VAL)
...

;; Tests using sample values for f-on-los

(check-expect (f-on-loX . . .) . . .)
...
A function template for a generic type contains all the same elements as a concrete
function template: sample instances templates, a function definition template, sample
expression templates, and test templates. The definition template has a conditional
expression because there is variety in the data. So far, there is nothing really new
here except not knowing what X is at this time. That is fine because when a type is
plugged in for X, the yield is a concrete function template. The question we must
answer is how do we deal with selfreferences in the data definition. Specifically, how
is . . . substituted in the definition template’s else clause. This clause tells us how
to process a list constructed using cons. Let us examine carefully what needs to be
processed:
(cons X (listof X))
There are two different types that need to be processed: X and (listof X). Process-
ing an X is done by calling a function on an X. How do we process a (listof X)?
The natural answer is to call a function that processes a (listof X). In other words,
the function must call itself. Let us finalize the template using this new insight:
;; Sample instances of (listof X)
(define LOX1 . . .)
(define LOX2 . . .)
...

;; (listof X) . . . → . . .
;; Purpose: . . .
(define (f-on-loX a-loX . . .)
(if (empty? a-loX)
...
. . .(f-on-X (first a-loX)). . .
. . .(f-on-loX (rest a-loX) . . .). . .))

;; Sample expressions for f-on-loX

(define LOX1-VAL . . .)
(define LOX2-VAL . . .)
...
70 Designing List-Processing Functions 277

;; Tests using sample computations for f-on-loX

(check-expect (f-on-loX LOX1 . . .) LOX1-VAL)
(check-expect (f-on-loX LOX2 . . .) LOX2-VAL)
...

;; Tests using sample values for f-on-loX

(check-expect (f-on-loX . . .) . . .)
...
A new principle now becomes clear. Every selfreference in a data definition becomes
a selfreference in the definition template and, eventually, a selfreference in a func-
tion definition. That is, data of arbitrary size is processed by a recursive function.
Recursion that is based on the structure of your data is called structural recursion
(or natural recursion). As before with structures, the structure of your data suggests
the structure of your functions. Observe that a divide and conquer strategy naturally
arises from the structure of a (listof X). To solve a problem for a (listof X) a
subproblem for an X and a subproblem for a (smaller) (listof X) must be solved.
The fascinating part is that the subproblem for the smaller (listof X) is solved the
same way as the larger (listof X). When using structural recursion, it is always the
case that the subproblem is smaller and, therefore, closer to a base case. This means
that when structural recursion is correctly used, your code is guaranteed to eventually
stop. It becomes impossible for a function to implement an infinite recursion.

70 Designing List-Processing Functions

We can use the generic data definition for a (listof X) to define a list of numbers:
A lon is a (listof number)
Let us take a close look at what this means. Plugging in number for X in the generic
data definition for (listof X) yields:
;; A list of number (lon) is either
;; 1. '()
;; 2. (cons number lon)
This is exactly the data definition derived above. Plugging in number for X in the
function template for a function on a (listof X) yields a concrete function template
for a lon:
;; Sample instances of lon
(define LON1 . . .)
(define LON2 . . .)
...
278 12 Lists

;; lon . . . → . . .
;; Purpose: . . .
(define (f-on-lon a-lon . . .)
(if (empty? a-lon)
...
. . .(f-on-number (first a-lon)). . .
. . .(f-on-lon (rest a-lon) . . .). . .))

;; Sample expressions for f-on-lon

(define LON1-VAL . . .)
(define LON2-VAL . . .)
...

;; Tests using sample computations for f-on-lon

(check-expect (f-on-lon LON1 . . .) LON1-VAL)
(check-expect (f-on-lon LON2 . . .) LON2-VAL)
...

;; Tests using sample values for f-on-lon

(check-expect (f-on-lon . . .) . . .)
...
We can now use the template to write functions to process a list of numbers.
For example, write a function that squares a list of numbers. The design idea is to
traverse the given list and construct a new list with the square of each number in the
list. Specializing the template yields:
;; Sample instances of lon
(define LON1 '())
(define LON2 '(1 2 3 4))

;; lon → lon
;; Purpose: Return a list of the squares of the given lon
(define (square-lon a-lon)
(if (empty? a-lon)
'()
(cons (sqr (first a-lon))
(square-lon (rest a-lon)))))

;; Sample expressions for square-lon

(define LON1-VAL '())
(define LON2-VAL (cons (sqr (first '(1 2 3 4)))
(square-lon (rest '(1 2 3 4)))))

;; Tests using sample computations for square-lon

(check-expect (square-lon LON1) LON1-VAL)
(check-expect (square-lon LON2) LON2-VAL)
71 What Have We Learned in This Chapter? 279

;; Tests using sample values for square-lon

(check-expect (square-lon '(10 8 4)) '(100 64 16))
It is worth carefully analyzing the expressions used in the function definition:
(sqr (first a-lon)) square of the first number in a-lon
(square-lon (rest a-lon)) list of squares obtained from the rest of a-lon
In the function definition cons is used to construct the new list. A new square is
added to the front of the list of squares obtained from the rest of the given list.
There is a sample expression and corresponding test for each lon variety. Finally, as
always, there is one or more tests using sample values.
To close this chapter, it is worth noting that there is a popular legend among
students and professionals that recursion is hard. If you understand the design of
square-lon, then you know better. Recursion is not hard nor should it be feared.
Understanding how to design solutions to problems based on data definitions makes
recursion quite natural. In the next chapter, you will gain more experience designing
recursive functions. Enjoy the journey!

* Ex. 121 — Design a function to cube a given lon.

*** Ex. 122 — Design a function that returns a list of the lengths of all the
strings in a list of strings. Make sure you have a data definition for all the list
data types needed.

**** Ex. 123 — A student’s backpack may be modeled using a (listof

symbol). For example, a backpack containing a notebook, a laptop, and a
pen is represented as '(notebook laptop pen). Design a function to determine if
a given item is contained in a given backpack.

*** Ex. 124 — For the previous exercise, do you need a conditional to de-
fine the predicate as suggested by the data definition for (listof X)? If you
used a conditional for the previous exercise refactor your solution to not use a
conditional.

71 What Have We Learned in This Chapter?

The important lessons of Chap. 12 are summarized as follows:

• Data of arbitrary size occurs naturally in real life and in virtual worlds.
• A list is data of arbitrary size with two varieties:
– '()
– A list built using cons
280 12 Lists

• The constructor for a non-empty list, cons, builds a new list by adding an element
to the front of an existing list.
• The selectors for lists are first and rest.
• When using an error-throwing function, we did not write use check-error to
test that an error is thrown.
• A quoted list may be used when all the list values are known and can be enumer-
ated.
• The function list may be used when the expressions for each list value are
known and can be enumerated.
• A quasiquoted list may be used when for the list elements either a value or an
expression for the value is known. Expressions that need to be evaluated must be
preceded by a comma.
• A recursive data definition is needed to define data of arbitrary size like lists.
• A recursive data definition always has a variety of subtypes of which at least one
is a base subtype and at least one is a recursive subtype.
• A recursive subtype contains a selfreference to its supertype.
• A generic data definition is an abstraction over types and may be used to define
many different concrete data types.
• For (listof X) there are two varieties: '() is the base subtype and (cons X
(listof X)) is the recursive subtype.
• A generic data definition leads to a generic function template.
• In a function template for data of arbitrary size, the definition template has a
recursive call for every selfreference in the data it is designed to process.
• Data of arbitrary size is processed using a recursive function.
• Recursion based on the structure of the data is called structural recursion.
• Structural recursion implements a divide and conquer design.
• Proper use of structural recursion guarantees that functions are not infinite recur-
sions.
• A concrete type and a concrete function template is obtained from a generic type
and its generic template by plugging in a type for each type variable.
• A concrete type and its concrete template are used to design functions. For
example, (listof number) and its function template are used to design any
function that processes a list of numbers.
Chapter 13
List Processing

In this chapter, we explore common list operations. These include summarizing,

searching, mapping, filtering, and sorting. A summarizing operation is one that
computes an aggregate value from the list elements. For example, you may need to
compute the length of a list or the sum of a list of numbers. A search operation is
used when list membership needs to be determined. You may want to know if a list
contains a value or if any list element satisfies a property. Mapping is used when you
want to apply a function to every list element and return a new list with the results.
Filtering is used when you need to extract all the list elements that satisfy a property.
Finally, list sorting is used when the list elements need to be put in order. In order
to sort a list, the list elements must be numerical or ordinal data. That is, it must be
possible to naturally put list elements in order. For example, a list of numbers may be
ordered in non-decreasing or non-increasing order. On the other hand, a list of colors
cannot be ordered. This is because colors are nominal data. There is no natural way
to order colors. Does brown come before green or does green come before brown?
Given that lists must be processed, the functions we design will be recursive. We
shall bring all our accumulated program design knowledge to bear on the functions
developed. Chief among these is careful problem analysis and separation of concerns.
Remember that for every value that needs to be computed, a separate function needs
to be designed.

72 List Summarizing

As a student, you are likely to keep track of your quiz grades in a class throughout
the semester. Why would you want to do this? Perhaps, your quiz average is 15% of
your final grade. Therefore, you would like to compute your quiz average every time
you earn a new quiz grade. How can you represent your quiz grades? Ask yourself if
you know the number of quizzes in advance. The most likely answer is that you do
not. After all, professors are likely to have pop quizzes or not announce the number
of quizzes in advance. Therefore, they may be, for example, 0, 19, or 7 quizzes.
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 281
M. T. Morazán, Animated Problem Solving, Texts in Computer Science,
https://doi.org/10.1007/978-3-030-85091-3_13
282 13 List Processing

The point is that the number of quizzes is arbitrary and must be represented using
compound data of arbitrary size. Thus, a structure is a poor choice to represent the
quiz grades, and a list of numbers is a much better representation.
A first attempt to define a list of quiz grades may be:
A list of quiz grades (loq) is a (listof number).
Is this a reasonable representation of quiz grades? Unfortunately, it is not. Given this
data definition, the following is a valid list of quiz grades:
'(87 65 92 -45 88 -7 98)
Clearly, this is a list of numbers. Is it a list of quiz grades? Clearly, it is not because
a quiz grade cannot be negative. We need to refine our data definition to be more
specific about what a quiz grade is. Assuming a quiz grade is based on a 100-point
scale, we may say:
;; A quiz grade (qg) is a number in [0..100]

;; A list of quiz grades (loq) is a (listof qg)

;; Sample values of loq

(define LOQ1 '())
(define LOQ2 '(87 65 92 88 98))
Now we have a data definition that is specific for quiz grades. A loq is a subtype of
a list of numbers. This means that a function on a list of numbers can process a loq,
but a function on a loq may not be able to process a list of numbers.
As we learned in Chap. 12, we can use qg to derive a concrete function template
for a loq:
;; Sample instances of loq
(define LOX1 . . .)
(define LOX2 . . .)

;; loq . . . → . . .
;; Purpose: . . .
(define (f-on-loq a-loq . . .)
(if (empty? loq)
'()
. . .(f-on-qg (first a-loq)). . .
. . .(f-on-loq (rest a-loq) . . .). . .))))

;; Sample expressions for f-on-loq

(define LOQ1-VAL . . .)
(define LOQ2-VAL . . .)
...
72 List Summarizing 283

;; Tests using sample computations for f-on-loq

(check-expect (f-on-loq LOQ1 . . .) LOQ1-VAL)
(check-expect (f-on-loq LOQ2 . . .) LOQ2-VAL)
...

;; Tests using sample values for f-on-loq

(check-expect (f-on-loq . . .) . . .)
...
Consider writing a function to compute the average of a loq. With data analysis
done, the next step of the design recipe is problem analysis to outline the computation.
Ask yourself how a quiz average is computed. You need to know the sum of all the
quiz grades. You need to know the number of quiz grades. You need to divide the
sum of quiz grades by the number of quiz grades. We can now specify what needs
to be computed:
1. Compute the length of the given loq
2. Compute the sum of the given loq
3. Divide the sum by the length
Given that three different values are needed, how many different functions are
needed? The most natural answer is three: one for each needed value. This is under-
lined by the principle of separation of concerns.
Let us start with the function that computes the average of a loq. This function
needs to summarize a property of the given loq by dividing its sum by its length.
Take note of two facts. First, this function does not process the given loq. Instead,
it manipulates two list-summarizing values. This means that using the template for
a function on a loq is incorrect. This template is needed only when a function
needs to traverse a list. Second, the given loq may not be empty. It is impossible to
compute the average of an empty loq because division by 0 is undefined. Therefore,
this function throws an error if the given loq is empty. The design of this function
assumes that functions to compute the sum and the length exist. Following the steps
of the design recipe results in:
;; loq → number throws error
;; Purpose: To compute the given loq’s average
(define (avg-loq a-loq)
(if (empty? a-loq)
(error (avg-loq expects an non-empty loq))
(/ (sum-loq a-loq) (loq-len a-loq))))

;; Sample expressions for avg-loq

(define LOQ2-VAL (avg-loq LOQ2))

;; Tests using sample computations for avg-loq

(check-expect (avg-loq LOQ2) LOQ2-VAL)
284 13 List Processing

;; Tests using sample values for avg-loq

(check-error (avg-loq LOQ1) "avg-loq expects an non-empty.")
Observe that both subtypes of loq are tested, albeit, not as tests using sample
expressions.

* Ex. 125 — Define more sample instances of loq and more tests using sample
expressions for avg-log.

*** Ex. 126 — Develop a data definition for a non-empty (listof X). Re-
design avg-loq around processing a non-empty (listof qg).

We now proceed to design the auxiliary functions for the sum and the length of
a loq. The sum of an empty loq is 0. The sum of a non-empty loq is the sum of
the rest of the given loq and the first element of the given loq. Observe that we
statically reason about the loq. That is, we reason about what to do with the needed
value obtained from the first quiz grade and the needed value obtained from the rest
of the list. We do not dynamically reason about the loq. That is, we do not say that
we add the first element of the list, the second element of the list, the third element
of the list, and so on. We do not because it is not always clear what “and so on”
means. Furthermore, static reasoning is easily translated to a solution expressed as
a program. That is, it is clear that two things are needed to be done: process the
first element and process the rest of the given list. This problem analysis leads to the
following specialization of the template for a function on a loq:
;; Sample values of loq
(define LOQ3 '(50 90 80 60 70))

;; loq → number
;; Purpose: To compute the sum of the given loq
(define (sum-loq a-loq)
(if (empty? a-loq)
0
(+ (first a-loq) (sum-loq (rest a-loq)))))

;; Sample expressions for sum-loq

(define SUM-LOQ1-VAL 0)
(define SUM-LOQ2-VAL (+ (first LOQ2) (sum-loq (rest LOQ2))))
(define SUM-LOQ3-VAL (+ (first LOQ3) (sum-loq (rest LOQ3))))

;; Tests using sample computations for sum-loq

(check-expect (sum-loq LOQ1) SUM-LOQ1-VAL)
(check-expect (sum-loq LOQ2) SUM-LOQ2-VAL)

;; Tests using sample values for sum-loq

(check-expect (sum-loq '(96 97 99 100)) 392)
72 List Summarizing 285

An extra sample loq instance is defined to improve the illustration of the abstraction
that leads to the function definition. Observe that it is easy to see that the code
implements the strategy suggested by static reasoning.
The final task is to design a function to compute the length of a loq. The length
is 0 if the given loq is empty. What is the answer if the loq is not empty? Use static
reasoning to determine the solution. According to the template for a function on a
loq, you need a value obtained from the first qg, a value obtained from the rest of the
qgs, and a function to combine these two values. Ask yourself, what does the first qg
contribute to the length? It contributes 1 to the length of the given loq. What does
the rest of the quiz grades contribute to the length? It contributes its length. How do
we combine these two numbers to obtain the length of the given loq? The length
of the given loq is obtained by adding these two numbers. This problem analysis
yields the following template specialization:
;; loq → number
;; Purpose: To compute the length of the given loq
(define (length-loq a-loq)
(if (empty? a-loq)
0
(+ 1 (length-loq (rest a-loq)))))

;; Sample expressions for length-loq

(define LEN-LOQ1-VAL 0)
(define LEN-LOQ2-VAL (+ 1 (length-loq (rest LOQ2))))
(define LEN-LOQ3-VAL (+ 1 (length-loq (rest LOQ3))))

;; Tests using sample computations for length-loq

(check-expect (length-loq LOQ1) LEN-LOQ1-VAL)
(check-expect (length-loq LOQ2) LEN-LOQ2-VAL)
(check-expect (length-loq LOQ3) LEN-LOQ3-VAL)

;; Tests using sample values for length-loq

(check-expect (length-loq '(96 97 99 100 89 94)) 6)
Once again observe that the code implements the strategy suggested by static rea-
soning. Developing static reasoning skills is important because, as we will discover,
it makes proving the correctness of a program easier.

* Ex. 127 — Design and implement a function to compute the average of a list
of numbers.
** Ex. 128 — Design and implement a function to compute the product of a
list of numbers.
286 13 List Processing

* Ex. 129 — Design and implement a function to count the number of aliens
in a (listof alien).
** Ex. 130 — Design and implement a function to append the strings in a list
of strings.

73 List Searching

Another common operation is searching a given list for a value. For example, given
a list of numbers, a-lon, and a number, x, you need the sublist of a-lon that starts
with the first occurrence of x. We start with problem analysis. If a-lon is empty
the answer is '() because a-lon does not contain x. What if the list is not empty?
Think about where x may be in terms of the structure of a-lon. We observe that
x may be the first element of the list. In this case, the answer is a-lon given that
(first a-lon) is the first occurrence of x. It is also possible that (first a-lon)
is not equal to x. In this case, (rest a-lon) must be searched. This, of course, is
done recursively as suggested by the template for a function on an lon. Will this
design idea work if x is not in a-lon? If you think carefully about it you realize
that it does. When x is in (rest a-lon), the recursive search stops when x equals
(first a-lon). When x is not in (rest a-lon), the recursive search stops when
a-lon equals '() and returns '()—the correct answer. Finally, observe that there
are three conditions that need to be detected (not two as suggested by the template
for a function on a lon).
We now need to develop sample expressions using sample lon instances to
illustrate how the result is computed for each of the three conditions. We define the
following sample lon instances:
;; Sample instances of lon
(define LON1 '())
(define LON2 '(1 2 3 4))
Using these sample instances, we define the following sample expressions:
;; Sample expressions for xsublist-lon
(define XSUBLIST-LON1-VAL '())
(define XSUBLIST-LON2-VAL1 LON2)
(define XSUBLIST-LON2-VAL2 (xsublist-lon 3 (rest LON2)))
(define XSUBLIST-LON2-VAL3 (xsublist-lon 0 (rest LON2)))
The first sample expression illustrates the answer when searching for, say, 7 in the
empty lon. The second sample expression illustrates the answer when searching for
1 in an lon that has 1 as its first number. The third and fourth sample expressions
illustrate how to compute the answer when the given number (3 and 0, respectively)
is not the first number in the given lon. Observe that the third sample expression
73 List Searching 287

illustrates searching for a number that is in the rest of the given list while the fourth
sample expression searches for a value not contained in the rest of the list. These
last two sample expressions illustrate the expression needed to make a recursive call
to search the rest of the given lon–just as suggested by the function template for a
function on a lon.
We can now finalize the specialization of the function template as follows:
;; number lon → lon
;; Purpose: To return the sublist of the given lon that
;; starts with the first instance of the given
;; number.
(define (xsublist-lon x a-lon)
(cond [(empty? a-lon) '()]
[(equal? x (first a-lon)) a-lon]
[else (xsublist-lon x (rest a-lon))]))

;; Tests using sample computations for xsublist-lon

(check-expect (xsublist-lon 7 LON1) XSUBLIST-LON1-VAL)
(check-expect (xsublist-lon 1 LON2) XSUBLIST-LON2-VAL1)
(check-expect (xsublist-lon 3 LON2) XSUBLIST-LON2-VAL2)
(check-expect (xsublist-lon 0 LON2) XSUBLIST-LON2-VAL3)

;; Tests using sample values for xsublist-lon

(check-expect (xsublist-lon 99 '(96 97 99 100)) '(99 100))
(check-expect (xsublist-lon 87 '(86 47 10)) '())
Observe that the conditional in the function’s body has three stanzas: one for each
condition that must be detected. The first is used when the given lon is empty. The
second is used when the given lon has x as its first element. The third is used when
the rest of the list must be searched. The tests using sample values further illustrate
that the function works for searches that are successful and that are unsuccessful.

* Ex. 131 — Design and implement a function that returns the sublist of a
given (listof string) that starts with the first instance of a string that has a
length less than 5.

*** Ex. 132 — Design and implement a function that returns the first instance
of an even number in a given (listof number). If there are no even numbers
in the given (listof number), the function returns #false. For this problem,
you need a data definition for the function’s result type.

**** Ex. 133 — A list of alphabet symbols is a list that only contains the
symbols either in ['a..'z] or in ['A..'Z]. Design and implement a function
that returns the first instance of a symbol in ['a..'z] in a given list of alphabet
symbols. Make sure to follow all the steps of the design recipe. You need to
develop two data definitions to solve this problem.
288 13 List Processing

74 List ORing

A variant of list searching is list ORing. ORing determines if there exists a list value
in a given list, L, that satisfies some property P. That is, we seek to determine if any
element in L satisfies P. Think about the varieties of L. What is the answer if L is
empty? The answer may not be clear to you. That is fine. Let us delay the answer for
now and think about a list of length 1 whose only element, p, does not satisfy P. The
list looks as follows:
L = (cons p '())
Is there an element in L that satisfies P? Clearly the answer is no. Now, think
about how L is processed. You need to apply P to p to obtain #false. This result
must be ored with the result obtained from the rest of L. The rest of L is always
recursively processed. This means that '() must be recursively processed. The only
way processing (cons p '()) returns #false is if processing '() returns #false.
This means that the answer is #false when L is empty. If L is not empty, then you
must apply or to the result of applying P to p and of recursively processing the rest
of L.
For example, in Aliens Attack with multiple aliens all the aliens move in the same
direction. The direction only changes when there exists an alien that is either at the
left or the right edge of the scene. This means that we need predicates to determine
if there exists an alien located at either the right or left edge in a list of aliens.

74.1 Determining If an Alien Is at the Left Edge

Carefully think about what it means for there to exist an alien that is at the left edge
of the scene. One condition is that the given list of aliens cannot be empty. This is
a necessary condition for such an alien to exist, but it is not sufficient. A necessary
condition is one that must be true for a predicate to be true, but is not enough to
conclude that the predicate is true. A sufficient condition is one that allows us to
conclude that the predicate is true. A list of aliens must not be empty for there to
be an alien at the left edge of the scene. A list of aliens that is not empty is not
sufficient, because it may not have an alien at the left edge. This means we need
more conditions to hold in order to conclude that there exists an alien at the left edge.
What other condition(s) must hold? Once again, think in terms of the structure of
the list you need to process. Either the first alien is at the left edge or an alien in the
rest of the list of aliens is at the left edge. Observe that no decision must be made to
compute this value. That is, a conditional as suggested by the template for a function
on a list of aliens is not needed. Instead, we can use and to detect if both conditions
hold. We can use or to detect if the first alien is at the left edge or if any alien in the
rest of the list is at the left edge.
74 List ORing 289

Based on the above problem analysis, we can specialize the template for a function
on a loa as follows:
(define ALIEN-8-0 (make-posn 8 0))

;; Sample instances of loa

(define EDGE-LOA (list ALIEN-8-0
(make-posn MIN-IMG-X 11)
(make-posn 5 5)))
(define EDGE-LOA2 (list (make-posn 1 11)
(make-posn MAX-IMG-X 5)))

;; loa → Boolean
;; Purpose: To determine if any alien is at scene’s left edge
(define (any-alien-at-left-edge? a-loa)
(and (not (empty? a-loa))
(or (alien-at-left-edge? (first a-loa))
(any-alien-at-left-edge? (rest a-loa)))))

;; Sample expressions for any-alien-at-left-edge?

(define LEDGE-E-LOA-VAL
(and (not (empty? E-LOA))
(or (alien-at-left-edge? (first E-LOA))
(any-alien-at-left-edge? (rest E-LOA)))))

(define LEDGE-INIT-LOA-VAL
(and (not (empty? INIT-LOA))
(or (alien-at-left-edge? (first INIT-LOA))
(any-alien-at-left-edge? (rest INIT-LOA)))))

(define LEDGE-LOA-VAL
(and (not (empty? EDGE-LOA))
(or (alien-at-left-edge? (first EDGE-LOA))
(any-alien-at-left-edge? (rest EDGE-LOA)))))

(define LEDGE-LOA2-VAL
(and (not (empty? EDGE-LOA2))
(or (alien-at-left-edge? (first EDGE-LOA2))
(any-alien-at-left-edge? (rest EDGE-LOA2)))))

;; Tests using sample computations any-alien-at-left-edge?

(check-expect (any-alien-at-left-edge? E-LOA)
LEDGE-E-LOA-VAL)
(check-expect (any-alien-at-left-edge? INIT-LOA)
LEDGE-INIT-LOA-VAL)
290 13 List Processing

(check-expect (any-alien-at-left-edge? EDGE-LOA)

LEDGE-LOA-VAL)
(check-expect (any-alien-at-left-edge? EDGE-LOA2)
LEDGE-LOA2-VAL)

;; Tests using sample values for any-alien-at-left-edge?

(check-expect (any-alien-at-left-edge?
(list (make-posn MIN-IMG-X 8)
(make-posn 6 3)
(make-posn MAX-IMG-X 10)))
#true)
(check-expect (any-alien-at-left-edge?
(list (make-posn 3 8)
(make-posn MIN-IMG-X 3)
(make-posn 5 2)))
#true)
(check-expect (any-alien-at-left-edge?
(list (make-posn MAX-IMG-Y 8)))
#false)
(check-expect (any-alien-at-left-edge?
(list (make-posn 3 8)
(make-posn 5 2)))
#false)
The constant ALIEN-8-0 is defined to help write loa instances and future tests.
Two new sample loa instances are defined to facilitate the development of tests.
The first new sample instance has an alien at the left edge. The second new sample
instance has an alien at the right edge. The sample expressions use and to determine
that the list is not empty and that at least one of the aliens is at the left edge.
To determine if there is at least one alien at the left side or is used. The job of
determining if a single alien is at the left edge is delegated to an auxiliary function
because an alien is not the same type as the one that the function processes (i.e.,
alien  loa). Observe that the only difference among the sample expressions is
the list of aliens used. Thus, any-alien-at-left-edge? has only one parameter.
As expected, any-alien-at-left-edge? uses its parameter, instead of a concrete
value, in its body. The tests using sample computations illustrate the function works
for four interesting loas: empty, does not contain an alien at either edge, contains
an alien at the left edge, and contains an alien at the right edge. The tests using
sample values further illustrate how the function works when the given loa is not
empty.
The auxiliary function needed, alien-at-left-edge?, was already for Aliens
Attack version 2 (in Chap. 9). Therefore, any-alien-at-left-edge?’s design and
implementation is complete.
74 List ORing 291

74.2 Determining If an Alien Is at the Right Edge

Naturally, the next task is to design a function to determine if any alien is at the
right edge. Its design is similar to any-alien-at-left-edge?’s design. The
sufficient condition tests if an alien is at the right edge instead of the left edge.
That is, a given loa has an alien at the right edge if the loa is not empty
and either the first alien is at the left edge or an alien in the rest of the loa
is at the left edge. The specialization of the template for functions on a loa
yields:
;; loa → Boolean
;; Purpose: To determine if any alien is at scene’s right edge
(define (any-alien-at-right-edge? a-loa)
(and (not (empty? a-loa))
(or (alien-at-right-edge? (first a-loa))
(any-alien-at-right-edge? (rest a-loa)))))

;; Sample expressions for any-alien-at-right-edge?

(define EDGE-E-LOA-VAL
(and (not (empty? E-LOA))
(or (alien-at-right-edge? (first E-LOA))
(any-alien-at-right-edge? (rest E-LOA)))))

(define EDGE-INIT-LOA-VAL
(and (not (empty? INIT-LOA))
(or (alien-at-right-edge? (first INIT-LOA))
(any-alien-at-right-edge? (rest INIT-LOA)))))

(define EDGE-LOA-VAL
(and (not (empty? EDGE-LOA))
(or (alien-at-right-edge? (first EDGE-LOA))
(any-alien-at-right-edge? (rest EDGE-LOA)))))

(define EDGE-LOA2-VAL
(and (not (empty? EDGE-LOA2))
(or (alien-at-right-edge? (first EDGE-LOA2))
(any-alien-at-right-edge? (rest EDGE-LOA2)))))

;; Tests using sample computations any-alien-at-right-edge?

(check-expect (any-alien-at-right-edge? E-LOA)
EDGE-E-LOA-VAL)
(check-expect (any-alien-at-right-edge? INIT-LOA)
EDGE-INIT-LOA-VAL)
(check-expect (any-alien-at-right-edge? EDGE-LOA)
EDGE-LOA-VAL)
292 13 List Processing

(check-expect (any-alien-at-right-edge? EDGE-LOA2)

EDGE-LOA2-VAL)

;; Tests using sample values for any-alien-at-right-edge?

(check-expect (any-alien-at-right-edge?
(list (make-posn MIN-IMG-X 8)
(make-posn 6 3)
(make-posn MAX-IMG-X 10)))
#true)
(check-expect (any-alien-at-right-edge?
(list (make-posn 3 8)
(make-posn MAX-IMG-X 3)
(make-posn 5 2)))
#true)
(check-expect (any-alien-at-right-edge?
(list (make-posn MIN-IMG-Y 8)))
#false)
(check-expect (any-alien-at-right-edge?
(list (make-posn 3 8)
(make-posn 5 2)))
#false)
Test coverage is similar to that of any-alien-at-left-edge?. The auxiliary func-
tion needed, alien-at-right-edge?, has already been written for Aliens Attack
version 2 (in Chap. 9). Therefore, any-alien-at-right-edge?’s design and im-
plementation is complete.

74.3 Determining If an Alien Has Reached Earth

As a final example consider determining if any alien in a given loa has reached
earth. This is another problem that requires oring a list. Such an alien exists in a
given loa if it is not empty and either the first alien has reached earth or one of
the rest of the aliens has reached earth. Once again we specialize the template for a
function on a loa and obtain:
(define EARTH-REACHED-LOA
(list (make-posn 1 11)
(make-posn MAX-IMG-X MAX-IMG-Y)))

;; loa → Boolean
;; Purpose: Determine if any alien has reached earth
(define (any-alien-reached-earth? a-loa)
(and (not (empty? a-loa))
(or (alien-reached-earth? (first a-loa))
(any-alien-reached-earth? (rest a-loa)))))
74 List ORing 293

;; Sample expressions for any-alien-reached-earth?

(define ANY-REACHED-VAL1
(and (not (empty? E-LOA))
(or (alien-reached-earth? (first E-LOA))
(any-alien-reached-earth? (rest E-LOA)))))
(define ANY-REACHED-VAL2
(and (not (empty? INIT-LOA))
(or (alien-reached-earth? (first INIT-LOA))
(any-alien-reached-earth? (rest INIT-LOA)))))
(define ANY-REACHED-VAL3
(and (not (empty? EARTH-REACHED-LOA))
(or (alien-reached-earth?
(first EARTH-REACHED-LOA))
(any-alien-reached-earth?
(rest EARTH-REACHED-LOA)))))

;; Tests using sample computations for any-alien-reached-earth?

(check-expect (any-alien-reached-earth? E-LOA)
ANY-REACHED-VAL1)
(check-expect (any-alien-reached-earth? INIT-LOA)
ANY-REACHED-VAL2)
(check-expect (any-alien-reached-earth? EARTH-REACHED-LOA)
ANY-REACHED-VAL3)

;; Tests using sample values for any-alien-reached-earth?

(check-expect (any-alien-reached-earth?
(list (make-posn 3 MAX-IMG-Y)
(make-posn 7 5)
(make-posn 8 2)))
#true)
(check-expect (any-alien-reached-earth? (list (make-posn 3 12)
(make-posn 7 5)
(make-posn 8 2)))
#false)
To make the development of tests easier a new loa instance with an alien that
has reached earth is defined. The auxiliary function, alien-reached-earth?, was
developed in Chap. 9. The tests cover both loa varieties and lists that contain and do
not contain an alien that reached earth.
Observe that any-alien-at-left-edge?, any-alien-at-right-edge?, and
any-alien-reached-earth? are all very similar. What does this suggest to you?
How do we avoid so much repetition? If you are thinking that an abstraction step is
needed, then you are on the right track. We shall soon learn how to avoid repetitions
among functions.
294 13 List Processing

* Ex. 134 — Design and write a function to determine if a given list of numbers
contains an even number.
* Ex. 135 — Design and write a function to determine if a given list of numbers
contains a multiple of 10.

* Ex. 136 — Design and write a function to determine if a given list of strings
contains "Ekaterina Ermilkina".
** Ex. 137 — A car is defined by a brand, a model, and a price. Design and
write a function to determine if a given list of cars contains a car that costs less
than $20,000.

* Ex. 138 — Design and write a function to determine if a given list of images
contains an image with a width greater than 100 pixels.

75 List ANDing

Another variant of list searching determines if all elements of some list, L, satisfy
some property. That is, we seek to determine if all of L’s elements satisfy a predicate
P. How is this determined? Think about the list varieties. What is the answer if L
is empty? The answer may not be obvious to you. That is fine. Let us once again
consider a list of length 1 whose only element, p, satisfies P. In this case, do all
the list elements satisfy P? Clearly, the answer is yes. Now, think about how this is
computed. The list looks as follows:
L = (cons p '())
Let us name the function that processes L: all-satisfy-P?. To process p you
apply P to p to obtain #true. The rest of the list is processed recursively by calling
all-satisfy-P?. All the list elements, if any, in the rest of L must also satisfy P.
This means that the following expression must evaluate to #true:
(and (P p) (all-satisfy-P? (rest L)))
Given that the rest of L is '(), (all-satisfy-P? (rest L)) must evaluate to
#true in order for (all-satisfy-P? L) to evaluate to #true. Thus, we have that
the answer must be #true when L is empty. When the list is not empty you and
the result of applying P to the first element of the list and the result obtained from
recursively processing the rest of the list.
75 List ANDing 295

75.1 All Even in a lon

Consider determining if all the numbers in a list of numbers are even. If the given
lon is empty then, as suggested above, the answer is true. It is certainly true that all
the numbers (i.e., none) in the empty list are even. If the given lon is not empty, then
the first number must be even and the rest of the numbers must be even. Observe that
no decision must be made to solve this problem. Therefore, the use of a condition
expression as suggested by the template for a function on a lon is not needed. Instead,
the result for the different varieties of lon is ored.
To start specializing the template for a function on a lon as suggested by our
problem analysis let us start by defining useful sample lon instances and sample
expressions for our function all-even?:
;; Sample instances of lon
(define E-LON '())
(define AE-LON '(88 98 22 78 506))
(define NAE-LON '(8 561 683 788))

;; Sample expressions for all-even-lon?

(define E-LON-VAL (or (empty? E-LON)
(and (even? (first E-LON))
(all-even-lon? (rest E-LON)))))
(define AE-LON-VAL (or (empty? AE-LON)
(and (even? (first AE-LON))
(all-even-lon? (rest AE-LON)))))
(define NAE-LON-VAL (or (empty? NAE-LON)
(and (even? (first NAE-LON))
(all-even-lon? (rest NAE-LON)))))
At least one example of each lon variety and sample lons that are and are not all
even are defined. The sample expressions state that the elements of a given lon are
all even if the list is empty or if all the elements satisfy Racket’s even? predicate.
The process of abstracting over the sample expressions identifies that there is
only one difference: the lon being processed. This step leads to the following
specialization of the definition template:
;; lon → Boolean
;; Purpose: To determine if the given lon only has even numbers
(define (all-even-lon? a-lon)
(or (empty? a-lon)
(and (even? (first a-lon))
(all-even-lon? (rest a-lon)))))
The steps of the design recipe are completed by specializing and running the tests.
The result of specializing the tests are:
296 13 List Processing

;; Tests using sample computations for all-even-lon?

(check-expect (all-even-lon? E-LON) E-LON-VAL)
(check-expect (all-even-lon? AE-LON) AE-LON-VAL)
(check-expect (all-even-lon? NAE-LON) NAE-LON-VAL)

;; Tests using sample values for all-even-lon?

(check-expect (all-even-lon? '(9 1 7)) #false)
(check-expect (all-even-lon? '(4 6 8)) #true)
The tests using sample values employ the sample lon instances and the constants
defined for the values of the sample expressions. The tests using sample values
further illustrate how the function works using lists that have either all even or all
odd numbers.

75.2 Determining if a lon Is Sorted

Consider the problem of determining of a list of numbers is sorted in non-decreasing

order. For example, '(73 87 87 99 104) is in non-decreasing order and '(73 87
87 17 205) is not in non-decreasing order. How do you know if a lon is or is not
in non-decreasing order? As you traverse the list from left to right the first number
must be less than or equal to the second number. What if the lon does not have two
numbers? In this case, the lon must be either '() or (cons number '()). In both
cases the lon is sorted in non-decreasing order.
The data analysis above suggests that for this problem we need a different data
definition for a list of numbers. Observe that there are two base cases: '() and
(cons number '()). If the length of the list is greater than 1, then there must be
at least two numbers in the list. This means that we need to define a list of numbers
as follows:
;; A lon2 is either:
;; 1. '()
;; 2. (list number)
;; 3. (list number number lon2)
A list of length greater than one is defined recursively. Notice that this is a valid data
definition. There are at least two varieties. Two varieties are nonrecursive and one
variety is recursive.
The corresponding template for a function on a lon2 is:
;; Sample instances of lon2
;; (define E-LON2 . . .)
;; (define A-LON2 . . .)
;; (define B-LON2 . . .)
;; ...
75 List ANDing 297

;; lon2 . . . --> . . .
;; Purpose: . . .
;; (define (f-on-lon2 a-lon2 . . .)
;; (cond [(empty? a-lon2) . . .)
;; [(= (length a-lon2) 1) . . .]
;; [else . . .(f-on-number (first a-lon2))
;; . . .(f-on-number (second a-lon2))
;; . . .(f-on-lon2 (rest a-lon2))
;; . . .(f-on-lon2 (rest (rest a-lon2)))]))

;; Sample expressions for sorted-lon2

;; (define E-LON2-VAL . . .)
;; (define A-LON2-VAL . . .)
;; (define B-LON2-VAL . . .)
;; ...

;; Tests using sample computations for sorted-LON2

;; (check-expect (sorted-lon2? E-LON2 . . .) E-LON2-VAL)
;; (check-expect (sorted-lon2? A-LON2 . . .) A-LON2-VAL)
;; (check-expect (sorted-lon2? B-LON2 . . .) B-LON2-VAL)
;; ...

;; Tests using sample values for sorted-lon2

;; (check-expect (sorted-lon2? . . .) . . .)
;; ...
The template suggests at least three sample instances need to be defined given that
there are three varieties of lon2. The definition template suggests using a conditional
with three clauses to capture each lon2 variety. Of these, the else clause is the most
interesting. It states that the first and second numbers may be processed by calling
a function on a number. Furthermore, the rest of the numbers may be processed in
one of two ways. The first eliminates the number in the first position of the list for
the recursive call. The second eliminates the first two numbers for the recursive call.
Why is there two ways to process the rest of the numbers? Observe that both (rest
a-lon2) and (rest (rest a-lon2)) are a lon2. The first of these may be used
when the first number no longer needs to be processed. The second is used when the
first two numbers no longer need to be processed. In the template there are also at
least three sample expressions to illustrate how a computation is performed for each
lon2 variety. The suggested number of tests correspond to the number of sample
instances and sample expressions for a lon2. Finally, the tests using sample values
will be used to illustrate that the function works for concrete values other than those
defined as sample instances.
Armed with a data definition and a function template, we can proceed with
problem solving. If the given lon2 is empty, then the numbers are sorted in non-
decreasing order. The same is true if the given lon2 only has one number. This
suggests that the first two varieties may be combined in our function definition
298 13 List Processing

because they yield the same answer. If the given lon2 has two or more elements,
then the first two elements must be in non-decreasing order. How is the recursive
call made? It must not eliminate the second number in the given lon2 because it is
needed to make sure it is in non-decreasing order relative to the third number if it
exists. That is, the recursive call is made with the rest of the given lon2. Observe
that a decision must not be made to compute the needed answer. We can or the result
of determining if the list is of at most length 1 and the result of anding the result
of comparing the first two numbers and the result of recursively processing the rest
of the given lon2. This analysis leads to the following specialization of the function
template:
;; Sample instances of lon2
(define E-LON2 '())
(define A-LON2 '(88))
(define SORTED-LON2 '(8 56 68 788))
(define UNSORTED-LON2 '(5 7 8 4 9))

;; lon2 → Boolean
;; Purpose: To determine if the given lon2 is sorted in
;; nondecreasing order
(define (sorted-lon2? a-lon2)
(or (<= 0 (length a-lon2) 1)
(and (<= (first a-lon2) (second a-lon2))
(sorted-lon2? (rest a-lon2)))))

;; Sample expressions for sorted-lon2

(define E-LON2-VAL
(or (< (length E-LON2) 2)
(and (<= (first E-LON2) (second E-LON2))
(sorted-lon2? (rest E-LON2)))))
(define A-LON2-VAL
(or (< (length A-LON2) 2)
(and (<= (first A-LON2) (second A-LON2))
(sorted-lon2? (rest A-LON2)))))
(define SORTED-LON2-VAL
(or (< (length SORTED-LON2) 2)
(and (<= (first SORTED-LON2) (second SORTED-LON2))
(sorted-lon2? (rest SORTED-LON2)))))
(define UNSORTED-LON2-VAL
(or (< (length UNSORTED-LON2) 2)
(and (<= (first UNSORTED-LON2)
(second UNSORTED-LON2))
(sorted-lon2? (rest UNSORTED-LON2)))))

;; Tests using sample computations for sorted-LON2

(check-expect (sorted-lon2? E-LON2) E-LON2-VAL)
75 List ANDing 299

(check-expect (sorted-lon2? A-LON2) A-LON2-VAL)

(check-expect (sorted-lon2? SORTED-LON2) SORTED-LON2-VAL)
(check-expect (sorted-lon2? UNSORTED-LON2) UNSORTED-LON2-VAL)

;; Tests using sample values for sorted-lon2

(check-expect (sorted-lon2? '(8 7 6 5)) #false)
(check-expect (sorted-lon2? '(5 6 7 8)) #true)
The sample instances have an instance of each lon2 variety. In addition, there are
two instances of the third lon2 variety: one that is sorted in non-decreasing order
and one that is not. These instances make illustrating how the function works through
tests easier. The sample expressions illustrate that a lon2 sorted in non-decreasing
order is either of length less than 2 or has the first two numbers in non-decreasing
order and the rest of the numbers (not including the first one) in non-decreasing
order. Abstracting over the sample expressions yields a function definition. Finally,
the tests illustrate that the function works for the sample lon2 instances and for other
concrete lon2 instances.

** Ex. 139 — Design and write a function to determine if a list of posns only
contains aliens as defined for Aliens Attack.
** Ex. 140 — Design and write a function to determine if a list of images only
contains images that have an area less than 200 squared pixels.

* Ex. 141 — Design and write a function to determine if a list of Booleans

only contains #trues.

** Ex. 142 — Design and write a function to determine if a list of Booleans

only contains #falses.

* Ex. 143 — Design and write a function to determine if a list of numbers is

in non-decreasing order using the conditional suggested by the template for a
function on a lon2.
* Ex. 144 — Design and write a function to determine if a list of numbers is
in non-increasing order.

** Ex. 145 — Design and write a function to determine if a list of strings is in

alphabetical order. You may use Racket’s string<=? predicate.

*** Ex. 146 — Design and write a function to determine if in a list of numbers,
L, for every even i in [0..(sub (length L))] the i + 1 number, if it exists,
has a different parity than the ith number.
300 13 List Processing

76 List Mapping

List mapping refers to the process of applying a function to every element of a list
and returning a list of the results. This type of operation is very common when
solving problems involving lists. For example, in Aliens Attack multiple aliens and
multiple shots may be represented using, respectively, a list of aliens and a list of
shots. Every time the clock ticks all the aliens and all the shots must be moved. How
is this accomplished?

76.1 Moving a List of Aliens

First consider how to move every alien. Imagine the world in Aliens Attack contains
a list of aliens and a direction. All the aliens must move in the same direction every
time the clock ticks. This means that the function move-alien must be mapped over
the given loa to create the list of the moved aliens. This list of moved aliens becomes
the world’s list of aliens after the clock tick. The list of aliens must be processed,
thus, suggesting specializing the template for a function on a list of aliens. Think
about how to process the list of aliens based on its structure. If the list of aliens
is empty, then there are no aliens to move and the list of moved aliens is empty.
Otherwise, the first alien is moved using the function move-alien and the rest of
the aliens are moved. The list of moved aliens is created by consing these two values.
Based on this problem analysis, the template for a function on a list of aliens may be
specialized starting with the sample expressions as follows:
;; Sample expressions for move-loa
(define MELOA-VAL E-LOA)
(define MILOA-VAL (cons (move-alien (first INIT-LOA) 'left)
(move-loa (rest INIT-LOA) 'left)))
(define MILOA-VAL2 (cons (move-alien (first INIT-LOA) 'right)
(move-loa (rest INIT-LOA) 'right)))
(define MILOA-VAL3 (cons (move-alien (first INIT-LOA) 'down)
(move-loa (rest INIT-LOA) 'down)))
The first sample expression states that when the empty list of aliens is moved the
resulting list of moved aliens is the empty list of aliens. The other three sample
expressions state that a non-empty list of aliens is moved by creating a list containing
the first alien moved in a given direction and the rest of the aliens moved in the same
given direction. There is such a sample expression for each possible instance of dir.
Abstracting over the sample expressions yields the specialization of the definition
template:
;; loa dir → loa
;; Purpose: To move the given loa in the given dir
(define (move-loa a-loa dir)
76 List Mapping 301

(if (empty? a-loa)

E-LOA
(cons (move-alien (first a-loa) dir)
(move-loa (rest a-loa) dir))))
To complete the steps of the design recipe the tests must be specialized for move-loa.
The tests using sample computations used the defined instances of loa and the
defined constants for the value of sample expressions. The tests using sample com-
putations utilize concrete lists of aliens and each of the possible dir instances. We
may specialize the testing suite as follows:
;; Tests using sample computations for move-loa
(check-expect (move-loa E-LOA 'right) MELOA-VAL)
(check-expect (move-loa INIT-LOA 'left) MILOA-VAL)
(check-expect (move-loa INIT-LOA 'right) MILOA-VAL2)
(check-expect (move-loa INIT-LOA 'down) MILOA-VAL3)

;; Tests using sample values for move-loa

(check-expect (move-loa (cons (make-posn 1 1)
(cons (make-posn 1 2) '()))
'right)
(cons (make-posn 2 1)
(cons (make-posn 2 2) '())))

(check-expect (move-loa (cons (make-posn 1 1)

(cons (make-posn 1 2) '()))
'left)
(cons (make-posn 0 1)
(cons (make-posn 0 2) '())))
(check-expect (move-loa (cons (make-posn 1 1)
(cons (make-posn 1 2) '()))
'down)
(cons (make-posn 1 2)
(cons (make-posn 1 3) '())))

76.2 Moving a List of Shots

Designing a function to move a list of shots is also a mapping operation. The function
move-shot must be mapped over a given list of shots. Given that a list of shots must
be processed, the template for a function on a list of shots must be specialized. The
result of this specialization is:
;; los → los
;; Purpose: To move the given list of shots
(define (move-los a-los)
302 13 List Processing

(if (empty? a-los)

E-LOS
(cons (move-shot (first a-los))
(move-los (rest a-los)))))

;; Sample expressions for move-los

(define MOVED-E-LOS-VAL E-LOS)
(define MOVED-A-LOS-VAL (cons (move-shot (first A-LOS))
(move-los (rest A-LOS))))

;; Tests using sample computations for move-los

(check-expect (move-los E-LOS) MOVED-E-LOS-VAL)
(check-expect (move-los A-LOS) MOVED-A-LOS-VAL)

;; Tests using sample values for move-los

(check-expect (move-los (cons (make-posn 12 7)
(cons NO-SHOT E-LOS)))
(cons (make-posn 12 6)
(cons NO-SHOT E-LOS)))
Observe the remarkable similarity between move-loa and move-los. This is
not terribly surprising because both map a function over a list. It does, however,
cement our idea that there must be an abstraction step to eliminate repetitions among
functions.

76.3 Returning a Different List Type

Both move-loa and move-los return the same type of list that they get as input.
This occurs for move-loa because the mapped function move-alien returns an
alien. Similarly, this occurs for move-los because the mapped function move-shot
returns a shot. This is not a property of mapping a function over a list. Instead, it is a
property of the function mapped. A mapping function may take as input a (listof
X) and return a (listof Y) if the signature of the mapped function is: X → Y.
Consider, for example, computing the lengths of each string in a (listof
string). The length of a string is always a number. More specifically, it is al-
ways a, natnum, natural number (i.e., a nonnegative integer). Therefore, the result
type for a function that solves this problem is: (listof string) → (listof
natnum). The function must process the given (listof string) and compute the
length of each string.
We start the process of specializing the template for a function on a (listof
string) by creating sample instances of (listof string):20

20
Verses courtesy of Lord Byron.
76 List Mapping 303

;; Sample instances of a (listof string)

(define E-LOSTR '())
(define NE-LOSTR '("Here’s" "a" "sigh" "to" "those" "who"
"love" "me."))
(define NE-LOSTR2 ’("And" "a" "smile" "for" "to" "those"
"who" "hate")
These sample (listof string) are used to define variables for the values of
sample expressions to illustrate how the answer is computed:
;; Sample expressions for lengths-lostr
(define E-LOSTR-VAL '())
(define NE-LOSTR-VAL (cons (string-length (first NE-LOSTR))
(lengths-lostr (rest NE-LOSTR))))
(define NE-LOSTR2-VAL (cons (string-length (first NE-LOSTR))
(lengths-lostr (rest NE-LOSTR))))
The sample expressions state that when the given (listof string) is empty,
then the answer is the empty (listof natnum). When the (listof string) is
not empty, then the answer is obtained by consing the length of the first string
in the given (listof string) and the lengths of the rest of the strings in the
given (listof string). Observe that the rest of the given (listof string)
is processed recursively by the function that computes the lengths of a (listof
string).
To specialize the definition template, the empty string is returned if the given
(listof string) is empty. The expression for when the given (listof string)
is not empty is obtained by abstracting over the second and third sample expressions.
This yields the following function definition:
;; (listof string) → lon
;; Purpose: Return a list of the string lengths in the
;; given (listof string)
(define (lengths-lostr a-lostr)
(if (empty? a-lostr)
'()
(cons (string-length (first a-lostr))
(lengths-lostr (rest a-lostr)))))
The tests using sample computations illustrate that the function properly computes
the value of the sample expressions. The tests using sample values illustrate the
function works for different sample values. The tests are:
;; Tests using sample computations for lengths-lostr
(check-expect (lengths-lostr E-LOSTR) E-LOSTR-VAL)
(check-expect (lengths-lostr NE-LOSTR) NE-LOSTR-VAL)
304 13 List Processing

;; Tests using sample values for lengths-lostr

(check-expect (lengths-lostr
(list "And" "whatever" "sky’s" "above" "me"))
(list 3 8 5 5 2))
(check-expect (lengths-lostr
(list "Here’s" "a" "heart" "for" "every"
"fate"))
(list 6 1 5 3 5 4))

* Ex. 147 — Design and write a function to cube the numbers in a list of
numbers.
** Ex. 148 — Design and write a function to return a list of x values from a
list of posns.

* Ex. 149 — Display the template for a function on a list of strings.

** Ex. 150 — Display the data definition and the template for a function on a
list of images. Design and write a function to compute the perimeters of a list
of images.

*** Ex. 151 — Design and write a function to double the even numbers and
to triple the odd numbers in a list of numbers.

**** Ex. 152 — A numstr is either a number or a string representing a number.

Design and write a function to change the numbers into strings (representing
the numbers) and change the strings into numbers. For example, the following
test ought to pass:
(check-expect (transform-lonumstr (list 10 "10" "-5" -8))
(list "10" 10 -5 "-8"))

Search the documentation for functions that transform a number into a string
and vice versa.

77 List Filtering

List filtering refers to the process of removing/filtering elements from a given list
that do not satisfy a condition. Alternatively, you may think of list filtering as
returning/extracting a list of the elements that satisfy a given condition. This is
achieved by testing every element of the list as it is traversed.
Three examples are presented. The first extracts the even numbers out of a given
list of numbers. The second filters out the aliens hit by a shot for a given list of aliens
and a given list of shots. The third extracts shots that have not hit an alien from a
given list of shots and a given list of aliens.
77 List Filtering 305

77.1 Extracting Even numbers

Consider the problem of returning the even numbers found in a list of numbers. The
list must be traversed. This means that the list must be processed and you ought to
think in terms of the structure of a list of numbers. If the given list is empty, then
it contains no even numbers and the empty list is the list of all even numbers in the
given list. If the given list is not empty, then we think in terms of the first number
and the rest of the numbers. The first number is included in the resulting list if it is
even and it is not included if it is odd. The rest of the list is processed recursively as
suggested by the template for a function on a lon. This analysis suggests that there
are three conditions that need to be determined: the list is empty, the first element of
the list is even, and the first element of the list is not even.
Given our problem analysis, we can proceed to specialize the template for a
function on a lon. First we define sample instances of a lon to use in our tests:
;; Sample instances of lon
(define LON1 '())
(define LON2 '(0 1 2 -3 4 5 -6 7 8 9))
(define LON3 '(11 75 -31 49))
Observe that each list satisfies one of the conditions. LON1 is empty. LON2 has an
even number as its first element. LON3 has an odd number as its first element.
We can now write sample expressions to illustrate how each of the different lists
is processed:
;; Sample expressions for extract-evens
(define LON1-VAL '())
(define LON2-VAL (cons (first LON2)
(extract-evens (rest LON2))))
(define LON3-VAL (extract-evens (rest LON3)))
When the input list is empty, the answer is the empty list. When the first number is
even this number is consed to the result of extracting the even numbers from the
rest of the list. When the first number is not even, then it is not consed to the result
of extracting the even numbers from the rest of the list.
Abstracting over the sample expressions helps us specialize the definition tem-
plate:
;; lon → lon
;; Purpose: To return list of the even numbers in the
;; given list
(define (extract-evens a-lon)
(cond [(empty? a-lon) '()]
[(even? (first a-lon))
(cons (first a-lon) (extract-evens (rest a-lon)))]
[else (extract-evens (rest a-lon))]))
306 13 List Processing

Instead of two stanzas, the conditional has three stanzas as per our problem analysis.
When the given lon is not empty, even? is used to determine if the first number is
even and if so it is added to the resulting list. If it is not even, then number must be
odd and is not added to the resulting list.
The tests using sample computations, as always, are specialized using the sample
lon instances and the defined constants for the values of the sample expressions:
;; Tests using sample computations for extract-evens
(check-expect (extract-evens LON1) LON1-VAL)
(check-expect (extract-evens LON2) LON2-VAL)
(check-expect (extract-evens LON3) LON3-VAL)
The tests using sample computations are specialized using extrema lon instances. In
this case, we use a list that only contains even numbers and a list that only contains
odd numbers:
;; Tests using sample values for extract-evens
(check-expect (extract-evens '(2 4 6 8)) '(2 4 6 8))
(check-expect (extract-evens '(-71 -9 -909 -55)) '())

77.2 Removing Hit Aliens

In Aliens Attack, we wish to have multiple shots and multiple aliens. Among other
things, this means that aliens hit by a shot must be removed from the world’s list of
aliens. This is accomplished by traversing the loa and testing each alien. If the first
alien has been hit by any shot, then it is not added to the resulting loa. Otherwise
the first alien is added to the resulting loa. Consider carefully how to design such
a function. There are two inputs: a list of aliens and a list of shots. We must decide
which template to specialize to write the function. Our problem analysis suggests
that for each alien we must traverse the list of shots to determine if it has been hit by
any of the shots. This means that the function ought to be designed to traverse the
aliens using the template for a function on a loa.
Observe that our problem analysis reveals that there are three conditions and,
therefore, a conditional expression with three stanzas is needed to specialize the
template. This means we need at least three sample expressions to illustrate how
each condition is processed. The following are sample expressions for this purpose:
;; Sample expressions for remove-hit-aliens
(define EMP-LOA-VAL '())
(define INIT-LOA-VAL (cons (first INIT-LOA)
(remove-hit-aliens (rest INIT-LOA)
INIT-LOS)))
(define INIT-LOA-VAL2 (remove-hit-aliens (rest INIT-LOA) LOS2))
These sample expressions use previously defined loa and los instances. The first
expression states that when the given loa is empty the answer is empty. There are
77 List Filtering 307

no aliens that have been hit in the empty loa. The second expression corresponds to
the case when the first alien in the given loa is not hit by any shot. In this case, the
first alien is added to the result list. The third sample expression corresponds to the
case when the first alien is hit. In this case, the first alien is not added to the result
list.
Using a variable instead is concrete values for the loa and the los allows us to
specialize the template definition:
;; loa los → loa
;; Purpose: To remove the aliens from the given loa hit by any
;; shot in the given los
(define (remove-hit-aliens a-loa a-los)
(cond [(empty? a-loa) '()]
[(hit-by-any-shot? (first a-loa) a-los)
(remove-hit-aliens (rest a-loa) a-los)]
[else (cons (first a-loa)
(remove-hit-aliens (rest a-loa) a-los))]))
Observe that there is a stanza for each of the conditions identified. The function
hit-by-any-shot? is an auxiliary predicate to determine if any shot has hit the
first alien. It needs to be designed and written. For now, we assume that the function
exists to fulfill the separation of concerns principle.
The tests using sample computations use the previously defined loa and los
instances and the defined constants for the value of the sample expressions:
;; Tests using sample computations for remove-hit-aliens
(check-expect (remove-hit-aliens E-LOA LOS2) EMP-LOA-VAL)
(check-expect (remove-hit-aliens INIT-LOA INIT-LOS)
INIT-LOA-VAL)
(check-expect (remove-hit-aliens INIT-LOA LOS2) INIT-LOA-VAL2)
The first test using sample values illustrates that the function works when more than
one alien must be removed. Observe that the los includes both shot varieties. The
second test illustrates that the function works when none of the aliens need to be
removed given a non-empty loa and a non-empty los. The tests are the following:
(check-expect (remove-hit-aliens
(cons (make-posn 1 1)
(cons (make-posn 2 2)
(cons (make-posn 3 3)
(cons (make-posn 4 4) '()))))
(cons (make-posn 3 3)
(cons NO-SHOT
(cons (make-posn 1 1) '()))))
(cons (make-posn 2 2)
(cons (make-posn 4 4) '())))
308 13 List Processing

(check-expect
(remove-hit-aliens (cons (make-posn 1 1)
(cons (make-posn 2 2)
(cons (make-posn 3 3) '())))
(cons (make-posn 7 4)
(cons NO-SHOT
(cons (make-posn 3 5)
'()))))
(cons (make-posn 1 1)
(cons (make-posn 2 2)
(cons (make-posn 3 3) '()))))
Now, we must design the auxiliary function hit-by-any-shot?. Given an alien,
how do we determine if the alien has been hit by any shot in a given los? The los
must be traversed to determine if hit?21 holds for any of the shots. Determining
if a predicate holds for any member of a list is a list oring function. Based on our
previous experience in this chapter, a conditional is not required for such a function.
Instead, an and-expression is used to determine that the given list of shots is not
empty and an or-expression is used to determine if the first shot has hit the alien or
some other shot has hit the alien.
Based on our problem analysis, we may write the following sample expressions:
;; Sample expressions for hit-by-any-shot?
(define HIT-LOS1-VAL (and (not (empty? LOS2))
(or (hit? (first LOS2) ALIEN-8-0)
(hit-by-any-shot? ALIEN-8-0
(rest LOS2)))))
(define HIT-LOS2-VAL (and (not (empty? INIT-LOS))
(or (hit? (first INIT-LOS) ALIEN-8-0)
(hit-by-any-shot?
ALIEN-8-0
(rest INIT-LOS)))))
Each sample expression tests if the a los is not empty. If the list of shots is not
empty, then both expressions test if the first shot has hit the alien and test if any other
shot has hit the alien. The first expression uses a non-empty los that contains a shot
that has hit the given alien. The second uses an empty los that, of course, does not
contain any shot that has hit the given alien.
Abstraction over the concrete expressions informs us how to specialize the defi-
nition template for a function on a los as follows:
;; alien los → Boolean
;; Purpose: To determine if the given alien is hit by any shot
;; in the given los
(define (hit-by-any-shot? an-alien a-los)
(and (not (empty? a-los))
(or (hit? (first a-los) an-alien)
(hit-by-any-shot? an-alien (rest a-los)))))

21
The predicate hit? was designed in Sect. 63.1.
77 List Filtering 309

Finally, we can write tests as follows:

;; Tests using sample computations for hit-by-any-shot?
(check-expect (hit-by-any-shot? ALIEN-8-0 LOS2)
HIT-LOS1-VAL)
(check-expect (hit-by-any-shot? ALIEN-8-0 INIT-LOS)
HIT-LOS2-VAL)

;; Tests using sample values for hit-by-any-shot?

(check-expect (hit-by-any-shot?
(make-posn 8 3)
(list (make-posn 1 1) (make-posn 10 7)))
#false)
(check-expect (hit-by-any-shot?
(make-posn 10 7)
(list (make-posn 1 1) (make-posn 10 7)))
#true)
The tests using sample values illustrate that the function returns the correct answer
when the given alien is and is not hit. Given that no more auxiliary functions are
needed, this completes the design of remove-hit-aliens.

* Ex. 153 — Design and write a function to return the odd numbers in a given
list of numbers.
* Ex. 154 — Design and write a function to extract the strings of length greater
than 5 from a given list of strings.

*** Ex. 155 — Consider the following data definition:

;; A TV character (character) is a structure,
;; (make-character string image), with a name
;; and an image of the character.
(define-struct character (name img))

Design and write a function to extract the images that have an area less than
10,000 pixels2 from a given list of characters.

** Ex. 156 — Design and write a function that extracts the multiples of 3 and
7 from a given list of numbers.

* Ex. 157 — Design and write a function to remove the strings that start with
“M” from a given list of strings.
310 13 List Processing

77.3 Removing Shots

In Aliens Attack, we need to filter the list of aliens. We must now decide if we must
also filter the list of shots. Should a single shot be able to hit multiple aliens or
should it only be able to hit one alien? We adopt the second posture. This means
that the list of shots must be filtered to eliminate shots that have hit an alien. As we
now know, filtering requires traversing a given list to create a new list with elements
that satisfy a property. In this case, we need to extract the shots that have not hit an
alien. Think carefully about what it means to have a shot that has not hit an alien.
Which shots are these? Clearly, a posn shot that does not have the same image-x
and image-y coordinates as any alien is a shot that must be preserved. What shots
need to be filtered out? A posn shot that has the same coordinates as any alien
must be filtered out. In addition, all NO-SHOT will also be filtered out because such
a shot no longer can affect the evolution of the game. To filter the shots we need as
input the list of shots and the list of aliens. As the list of shots is traversed any shot
that is NO-SHOT or a posn shot that has hit an alien is not added to the resulting
list of shots. Otherwise, the shot is added to the resulting list of shots.
The sample expressions in the template for a function on a list of shots may be
specialized as follows:
;; Sample loa instances
(define LOA3 (list (make-posn 1 9) (make-posn 8 0)))
(define LOA4 (list (make-posn 1 9) (make-posn 8 5)))

;; Sample expressions for remove-shots

(define RM-INIT-LOS-VAL INIT-LOS)
(define RM-LOS2-VAL (remove-shots (rest LOS2) LOA3))
(define RM-LOS2-VAL2 (cons (make-posn 8 0)
(remove-shots (rest LOS2)
LOA4)))
Two new loa sample instances are defined to easily define variables for the values
of the sample expressions. There is a sample expression for each possible condition.
Either the given los is empty, the first shot is removed, or the first shot is not
removed.
Based on the sample expressions, the definition template is specialized as follows:
;; los loa → los
;; Purpose: To remove hit and NO-SHOTs from the given loa
(define (remove-shots a-los a-loa)
(cond [(empty? a-los) a-los]
[(or (eq? (first a-los) NO-SHOT)
(hit-any-alien? (first a-los) a-loa))
(remove-shots (rest a-los) a-loa)]
[else (cons (first a-los)
(remove-shots (rest a-los) a-loa))]))
77 List Filtering 311

Observe that a shot is filtered out if it is either NO-SHOT or it has hit an alien. That is,
the shot is not consed into the resulting list of shots. Determining if the first shot has
hit an alien requires the given loa to be traversed. This means processing a different
kind of data and, therefore, this task is relegated to an auxiliary function that needs
as input a shot and an loa. Once again, we see the importance of separation of
concerns in program design.
The tests are specialized as follows:
;; Tests using sample computations for remove-shots
(check-expect (remove-shots INIT-LOS INIT-LOA)
RM-INIT-LOS-VAL)
(check-expect (remove-shots LOS2 (list (make-posn 1 9)
(make-posn 8 0)))
RM-LOS2-VAL)
(check-expect (remove-shots LOS2 (list (make-posn 1 9)
(make-posn 8 5)))
RM-LOS2-VAL2)

;; Tests using sample values for remove-shots

(check-expect (remove-shots (list (make-posn 2 9)
(make-posn 10 10))
INIT-LOA)
(list (make-posn 2 9) (make-posn 10 10)))
(check-expect (remove-shots (list (make-posn 8 2)
(make-posn 13 1))
INIT-LOA)
'())
The tests using sample computations, as always, illustrate that the function correctly
computes the values of the sample expressions. The tests using sample values illus-
trate that the function works when none of the shots must be filtered out and when
all of the shots must be filtered out.
Now our focus turns to the design of the needed auxiliary function. To de-
termine if a shot has hit an alien for a given loa, the given loa must be tra-
versed. We may use a conditional expression as suggested by the template for a
function on an loa, but careful problem analysis reveals that it is not needed. A
given shot has hit any alien if the given loa is not empty and either the shot has
hit the first alien or the shot has hit any other alien. This is a list ORing opera-
tion.
Based on this problem analysis, sample expressions may be written as follows:
;; Sample instance of a shot
(define SHOT5 (make-posn 11 1))

;; Sample expressions for hit-any-alien?

(define HIT-ANY-ALIEN0
312 13 List Processing

(and (not (empty? E-LOA))

(or (hit? NO-SHOT (first E-LOA))
(hit-any-alien? NO-SHOT (rest E-LOA)))))
(define HIT-ANY-ALIEN1
(and (not (empty? E-LOA))
(or (hit? SHOT3 (first INIT-LOA))
(hit-any-alien? NO-SHOT (rest INIT-LOA)))))
(define HIT-ANY-ALIEN2
(and (not (empty? INIT-LOA))
(or (hit? NO-SHOT (first INIT-LOA))
(hit-any-alien? NO-SHOT (rest INIT-LOA)))))

(define HIT-ANY-ALIEN3
(and (not (empty? INIT-LOA))
(or (hit? SHOT5 (first INIT-LOA))
(hit-any-alien? SHOT5 (rest INIT-LOA)))))
A new shot instance is defined to facilitate the writing of the sample expressions.
There are four sample expressions in order to have expressions for both varieties
of aliens and shots. There are four possible combinations as input. The computa-
tion for an empty loa and a NO-SHOT is illustrated by the first sample expression.
The computation for an empty loa and a posn shot is illustrated by the second
sample expression. The computation for a non-empty loa and a NO-SHOT is il-
lustrated by the third sample expression. The computation for a non-empty loa
and a posn shot is illustrated by the fourth sample expression. The previously
defined hit? (in Chap. 11) is used to detect if a shot has hit an alien. This is
consistent with the goal to recycle as much code as possible during a refinement
step.
Abstracting over the sample expressions leads to the following function definition:
;; shot loa → Boolean
;; Purpose: Determine if the given shot has hit any alien
;; in the given loa
(define (hit-any-alien? a-shot a-loa)
(and (not (empty? a-loa))
(or (hit? a-shot (first a-loa))
(hit-any-alien? a-shot (rest a-loa)))))
Observe that this is, indeed, very similar to other list ORing functions developed in
this chapter.
The function is tested as follows:
;; Tests using sample computations for hit-any-alien?
(check-expect (hit-any-alien? NO-SHOT E-LOA)
HIT-ANY-ALIEN0)
(check-expect (hit-any-alien? SHOT3 E-LOA)
HIT-ANY-ALIEN1)
77 List Filtering 313

(check-expect (hit-any-alien? NO-SHOT INIT-LOA)

HIT-ANY-ALIEN2)
(check-expect (hit-any-alien? SHOT5 INIT-LOA)
HIT-ANY-ALIEN3)

;; Tests using sample values for hit-any-alien?

(check-expect (hit-any-alien? (make-posn 9 3)
(list (make-posn 6 2)
(make-posn 9 3)
(make-posn 9 11)))
#true)
(check-expect (hit-any-alien? (make-posn 11 3)
(list (make-posn 6 2)
(make-posn 9 3)
(make-posn 9 11)))
#false)
(check-expect (hit-any-alien? NO-SHOT
(list (make-posn 6 2)
(make-posn 9 3)
(make-posn 9 11)))
#false)
The tests using sample computations test every possible combination of the varieties
of the input. The first test using sample values illustrates that the function works
when a posn shot has hit an alien other than the first. The second test using sample
values illustrates that the function works when a posn shot has not hit an alien. The
third test using sample values illustrates that the function works when a NO-SHOT is
given as input using a new loa.

* Ex. 158 — Design and write a function to extract the odd numbers from a
given list of numbers.

** Ex. 159 — Design and write a function to filter out numbers that are not a
multiple of 9 from a given list of numbers.

** Ex. 160 — Design and write a function to filter out the NO-SHOTs from a
given list of shots.

*** Ex. 161 — Design and write a function to filter out from a list of numbers
any number that appears in another given list of numbers.

*** Ex. 162 — Consider the following data definition:

314 13 List Processing

#| A store item (item) is a structure,

(make-item string number number),
that has the item’s name, price, and quantity in stock.
|#
(define-struct item (name price quantity))

Design and write a function to extract from a list of items those that are below
a given price.

78 List Sorting

The solution to many problems requires sorting a list. In order for a list to be sortable,
it must contain numerical or ordinal data. That is, any two elements are comparable
to determine which goes first and which goes second. For example, a list of numbers
and a list of strings may be sorted. A list that contains nominal data is not sortable.
For example, a list of symbols or a list of religions is not sortable.22 For sortable data,
we need a predicate that determines which of two elements goes first. For example,
≤ and ≥ are two such predicates for numbers depending on whether the list must be
sorted in non-decreasing order or in non-increasing order.
Let us take as an example sorting a list of numbers in non-decreasing order. For
example, '(87 65 90 21) after sorted in non-decreasing order is '(21 65 87
90). How is this accomplished? Clearly, the given list must be traversed and a new
list must be constructed. Therefore, think about the varieties of a list of numbers.
If the given lon is empty, then the sorted lon is also empty. If the given list is not
empty, think about its structure. The template for a function on an lon tells us to
process the rest of the list recursively. Assuming the function is named sort-lon
and the given list is a-lon, think about what this expression evaluates to:
(sort-lon (rest a-lon))
This expression ought to evaluate to a sorted list containing all the elements of
a-lon except a-lon’s first element. This means that we must combine a-lon’s first
element with a sorted list. How is this done? The sorted list must be traversed to
find the right place to insert a-lon’s first element. If this is done, then we have a
sorted list that contains all the elements of the given list. Take note that inserting
a number into a sorted list of numbers is a different problem from sorting a list of
numbers. A different problem means we need an auxiliary function to perform this
computation. Keeping consistent with the principle of separation of concerns, we
assume this inserting function exists and works.

22
The words (i.e., strings) that represent the names of different religions are sortable, but this is a
property of strings and not a property of religions.
78 List Sorting 315

Based on our problem analysis, template specialization begins with a-lon in-
stances and sample expressions as follows:
;; Sample instances of a lon
(define E-LON '())
(define SORTED-LON '(17 18 29 37 41 52))
(define UNSORTED-LON '(89 21 1 77 23))

;; Sample expressions for sort-lon

(define E-LON-VAL '())
(define SORTED-LON-VAL
(insert (first SORTED-LON)
(sort-lon (rest SORTED-LON))))
(define UNSORTED-LON-VAL
(insert (first UNSORTED-LON)
(sort-lon (rest UNSORTED-LON))))
Two instances of a non-empty lon are defined to illustrate that the function works
for a sorted and unsorted list. The first sample expression illustrates the answer when
the given lon is empty. The second and third sample expressions illustrate that for a
given non-empty lon, the sorted list is obtained by inserting the first number of the
given list in the list obtained by sorting the rest of the list.
Abstracting over the sample expressions helps use specialize the template defini-
tion:
;; sort-lon: lon '() lon
;; Purpose: To sort the given lon in non-decreasing order
(define (sort-lon a-lon)
(cond [(empty? a-lon) '()]
[else (insert (first a-lon) (sort-lon (rest a-lon)))]))
Observe that insert still needs to be designed and implemented.
The tests are may be written as follows:
;; Tests using sample computations for sort-lon
(check-expect (sort-lon E-LON) E-LON-VAL)
(check-expect (sort-lon SORTED-LON) SORTED-LON-VAL)
(check-expect (sort-lon UNSORTED-LON) UNSORTED-LON-VAL)

;; Tests using sample values for sort-lon

(check-expect (sort-lon (list 5 4 3 2 1))
(list 1 2 3 4 5))
(check-expect (sort-lon (list 63 12 76 99 0))
(list 0 12 63 76 99))
The tests using sample values illustrate that the function works for a list that is in
non-increasing order and a different unsorted list.
With the design and implementation of sort-lon completed, we turn to the
design and implementation of the insert function. We have a number and a lon
316 13 List Processing

sorted in non-decreasing order. We need a lon sorted in non-decreasing order that

contains the given number and all the numbers in the given sorted lon. Let us analyze
the problem based in the lon varieties. If the given sorted list is empty, then the
resulting list must only contain the given number. What if the given sorted list is not
empty? Then we must determine if the given number is less than or equal to the first
number in the given sorted list. If so, we obtain the needed sorted list by consing
the two. For example, given
9 and '(10 43 88 100)
the complete sorted list is obtained by the following expression:
(cons 9 '(10 43 88 100))
What if the given number is greater than the first number in the given sorted list? In
this case, the first number of the resulting sorted list is the first number of the given
sorted list. The given number must be inserted in the rest of the given sorted list.
Processing the rest of the list, as we know, is done recursively. For example, given
53 and '(10 43 88 100)
the complete sorted list is obtained by the following expression:
(cons 10 (insert 53 '(43 88 100)))
Here the result of the call to insert is:
'(43 53 88 100)
Adding the 10 to the front of this list is the sorted list that is needed. Based on this
analysis we may conclude that the function must determine which one of 3 possible
is faced: the list is empty, the given number is less than or equal to the first number
of the list, and the given number is greater than the first number of the list.
To develop sample expressions, a sorted lon and two numbers to insert are
defined:
(define SORTED-LON2 '(31 87 95 102))
(define INSERT-NUM1 20)
(define INSERT-NUM2 90)
We display a sample expression for each possible condition:
;; Sample expressions for insert
(define ELON-VAL (list INSERT-NUM1))
(define SORTEDLON-VAL (cons INSERT-NUM1 SORTED-LON2))
(define UNSORTEDLON-VAL (cons (first SORTED-LON2)
(insert INSERT-NUM2
(rest SORTED-LON2))))
The first sample expression illustrates how a number is inserted into the empty list
of numbers. The second illustrates how a number is inserted when it is less than or
equal than the first number in the sorted list. The third expression illustrates how a
number is inserted when it is greater than the first number in the sorted list.
78 List Sorting 317

Building on the sample expressions, the definition template is specialized as

follows:
;; insert: number lon → lon
;; Purpose: To insert the given number in the given lon to
;; create an lon in non-decreasing order
;; ASSUMPTION: Given lon is sorted in nondecreasing order
(define (insert a-num a-lon)
(cond [(empty? a-lon) (list a-num)]
[(<= a-num (first a-lon)) (cons a-num a-lon)]
[else (cons (first a-lon)
(insert a-num (rest a-lon)))]))
Observe, the assumption that the given lon is sorted in non-decreasing is made
explicit. This communicates clearly to the readers and users of our code how the
solution was designed.
The tests for insert are the following:
;; Tests using sample computations for insert
(check-expect (insert INSERT-NUM1 E-LON) ELON-VAL)
(check-expect (insert INSERT-NUM1 SORTED-LON2)
SORTEDLON-VAL)
(check-expect (cons (first SORTED-LON2)
(insert INSERT-NUM2 (rest SORTED-LON2)))
UNSORTEDLON-VAL)

;; Tests using sample values for insert

(check-expect (insert 3 '(1 2 4 5))
'(1 2 3 4 5))
(check-expect (insert 101 '(-7 -5 0 2 232))
'(-7 -5 0 2 101 232))
The tests using sample values explicitly illustrate with concrete values how a number
is inserted into a sorted list in non-decreasing order.

* Ex. 163 — Design and write a function that sorts a list of numbers in non-
increasing order.

*** Ex. 164 — Consider the following data definition:

#| A store item (item) is a structure,
(make-item string number number),
that has the item’s name, price, and quantity in stock. |#
(define-struct item (name price quantity))

Design and write a function that sorts a list of items in non-decreasing order by
price.
318 13 List Processing

*** Ex. 165 — Stores usually need space for new inventory and put on sale
the items they have the most of in stock. Using the item data definition from
the previous problem, design and write a function that sorts a list of items in
non-increasing order by quantity.

79 What Have We Learned in This Chapter?

The important lessons of Chap. 13 are summarized as follows:

• List summarizing computes an aggregate value from the values in a given list.
• List searching determines if a given value is contained in a list. Two common
variants of list searching are:
– List ORing determines if there is any value in a given list that satisfies a property.
– List ANDing determines if all the values in a given list satisfy a property.
• List mapping applies a function to every element of the list and returns a list with
the results.
• List filtering extracts the elements of a list that satisfy a given predicate.
• List sorting orders lists of ordinal values based on a predicate that can be used to
compare any two list values.
• Think statically about the structure of a list during problem analysis to design a
list-processing function.
• A necessary condition is one that must be true for a predicate to be true, but is
not enough to conclude that the predicate is true.
• A sufficient condition is one that allows us to conclude that the predicate is true.
• Careful problem analysis reveals which functions that take a list as input processes
the list. These are the functions that are developed using a template for a function
on a list.
• List processing may require a conditional expression with more than two condi-
tions.
• Functions may take as input a (listof X) and may return a (listof Y).
• There is a rumor among beginner programmers that recursion is hard. Now you
know better!
Chapter 14
Natural Numbers

In this chapter we explore a different data type of arbitrary size. This is a data type
that is a subtype of the number type. It is likely that you have seen this data type in
your Mathematics textbooks. In a Mathematics textbook you may see the following
definition for the natural numbers:
The natural numbers are 0, 1, 2, . . .
You probably interpret the . . . as meaning all the way to infinity. Thus, you are
confident that you understand what the natural numbers are. A natural question that
you may be asked is: What is a natural number? You may answer that a natural
number is an element of the set of the natural numbers. This may be true, but is
it useful in problem solving? That is, can you solve a problem if the given data is
a natural number? How do you process a natural number? You may say that you
process a natural number the same way you process a number. Yes, of course, this is
possible for some problems because the natural numbers are a subset of the numbers.
Is it always the case? If n is a natural number how you compute n!? In a Mathematics
textbook you may find the following:
n! = 1 * 2 * * 3 * . . . * n-1 * n
Do you understand how to compute n!? You can probable compute 3! and 5!.
Can you compute 1000!? If the answer is no then you need to write a function
to perform the computation. Can you do this? There are several problems with the
approach taken in some Mathematics textbooks for our purposes. A value needs to
be computed from a natural number, but we do not have a data definition for a natural
number. In addition, we do not know how to program . . . .
Consider more carefully developing a data definition for the set of natural numbers.
A natural number may be small like 0. It may be large like 10000. It may be of medium
size like 1587. It may be super large like 100000000. What is all this telling us?
It is telling us that a natural number is data of arbitrary size. What do we need to
design functions that process a natural number? You have probably already realized
the answer. We need a recursive data definition and a function template.

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 319
M. T. Morazán, Animated Problem Solving, Texts in Computer Science,
https://doi.org/10.1007/978-3-030-85091-3_14
320 14 Natural Numbers

80 Data Definition for a Natural Number

Recall that for a recursive data definition we need at least two varieties. At least
one variety must not have a selfreference and at least one variety must have a
selfreference. Let us start with the variety that does not have a selfreference. When
we studied lists, the nonrecursive variety was the smallest list possible: '(). What
is the smallest natural number? Clearly, it is 0. Therefore, for our data definition we
can say that a natural number is 0. Now, ask yourself how can other natural numbers
be created? To make a larger list we use cons. What can we use to make a bigger
natural number? Let us start with 0, which we know is a natural number. Can we
create the next natural number, 1, using 0? Sure, we increment 0 by 1. Can we build,
2, the next natural number? Well we can increment the natural number 1 by 1. We
can build 3 by incrementing 2 by 1. There is a pattern here. If n is a natural number,
the next natural is obtained by incrementing it by 1. We can now write the needed
data definition:
A natural number (natnum) is either:
1. 0
2. (add1 natnum)
Observe that there are two natnum varieties of which one does not have a selfrefer-
ence and other does.
We can use add1 to create a new natnum. What is the selector function to extract
the natnum used to build a given nonzero natnum? What natnum is used to build 5?
It is 4. What natnum is used to build 10? It is 9. Again we see a pattern. The selector
function is sub1.
Observe that there is one selfreference in the data definition. This means that
processing a nonzero natnum requires one recursive call. Based on this observation
the template for a function on a natnum is the following:
#| ;; Sample instances of natnum
(define ZERO 0)
(define NATNUMA . . .)
...

;; natnum . . . → . . .
;; Purpose: . . .
(define (f-on-natnum a-natnum . . .)
(if (= a-natnum 0)
...
(. . .a-natnum. . .(f-on-natnum (sub1 a-natnum) . . .))))

;; Sample expressions for f-on-natnum

(define ZERO-VAL . . .)
(define NATNUMA-VAL . . .)
...
81 Computing Factorial 321

;; Tests using sample computations for f-on-natnum

(check-expect (f-on-natnum ZERO . . .) ZERO-VAL)
(check-expect (f-on-natnum NATNUMA . . .) NATNUMA-VAL)
...

;; Tests using sample values for f-on-natnum

(check-expect (f-on-natnum . . .) . . .)
(check-expect (f-on-natnum . . .) . . .)
...
|#
The template has all the expected components: sample instances, a function defini-
tion template, sample expressions, tests using sample computations, and tests using
sample values. Observe that one of the sample instances is 0. This is because it is the
only instance of the first variety. You may also observe that in the definition template
the expression for the recursive call contains sub1 applied to the input natnum.

81 Computing Factorial

Armed with the data definition and the template for a function on a natnum we are
ready to tackle problem solving. We start with designing a function to compute the
factorial of a natural number. First we define a few sample natnums as follows:
;; Sample instances of natnum
(define ZERO 0)
(define TEN 10)
(define FIFTY 50)
Two nonzero natnums are defined to facilitate the abstraction process over the sample
expressions. The values chosen are ones for which anyone is unlikely to compute its
factorial by hand. Thus, a function is really needed.
Next, we must analyze how to solve the problem. For this we reason statically
about the structure of a natnum (just as we reason statically about the structure of a
list). Start with the nonrecursive variety. What is the factorial of 0? The mathematical
definition for factorial above does not specify this. It is not clear what the value of
0! is. Let us proceed with the second variety and come back to the first variety later.
If the given natnum, n, is not 0, the template suggests combining the given number
and the factorial of n - 1. For example, for the factorial of 4 the template suggests
combining 4 and 3!. We know from the mathematical definition above that:
3! = 1 * 2 * 3
Given that value of 3! and the value 4, how do we get the value of 4!? Let us use
the mathematical definition again to examine what 4! is
4! = 1 * 2 * 3 * 4
322 14 Natural Numbers

It now becomes clear that 4 and the value of 3! must be multiplied. This informs us
that when the given natnum, n, is not 0, then n and (n - 1)! must be multiplied
to obtain n!. We can now turn back to think about what the value of 0! must be. A
value must be returned when n is 0. Let us look at how 4! is computed:
4! = 4 * 3 * 2 * 1 * ??
This sample computation captures the design idea of multiplying 4 by 3!. At each
step n is decremented by 1. Eventually, n becomes 0. The ?? represents the value of
0!. Observe that the product of 4 through 1 is 4!. It now becomes clear that 0! is 1.
Any other value for 0! would make the product not equal to 4!. Finally, observe that
the product of natural numbers is a natural number. For example, 4! is 24, which
is a natnum. This means that any function that computes the factorial of a natnum
must return a natnum.
With a clear idea of how to compute the factorial of a natnum, we write sample
expressions:
;; Sample expressions for factorial
(define ZERO-VAL 1)
(define TEN-VAL (* 10 (factorial (sub1 10))))
(define FIFTY-VAL (* 50 (factorial (sub1 50))))
The first sample expression illustrates the computation when the given natnum is 0.
The other two sample expressions illustrate the computation when the given natnum
is not 0. As per our problem analysis the given natnum is multiplied by the factorial
of said number decremented by 1.
Abstraction over the sample expressions yields the specialization of the definition
template
;; natnum → natnum
;; Purpose: Compute the factorial of the given natnum
(define (factorial a-natnum)
(if (= a-natnum 0)
1
(* a-natnum (factorial (sub1 a-natnum)))))
Observe that the return type is natnum as suggested by our problem analysis. In the
body of the function 1 is returned if the given natnum is 0. Otherwise, the product
of the given natnum, a-natnum, and the factorial of a-natnum - 1 is returned. In
other words, the values returned are consistent with our problem analysis.
The final template specialization step is to develop the tests. The tests using
sample computations are
;; Tests using sample computations for factorial
(check-expect (factorial ZERO) ZERO-VAL)
(check-expect (factorial TEN) TEN-VAL)
(check-expect (factorial FIFTY) FIFTY-VAL)
82 Computing Tetrahedral Numbers 323

These tests, as before, illustrate that the function correctly computes the values of
the sample expressions. They validate the abstraction over the sample expressions.
The tests using sample values are
;; Tests using sample values for f-on-natnum
(check-expect (factorial 3) 6)
(check-expect (factorial 5) 120)
The sample values chosen are small enough that any reader of the code can easily
verify that the expected values are correct. To finalize the steps of the design recipe
run the program to make sure all the tests pass.

* Ex. 166 — Design and implement a function to compute the sum of the first
n squares, where n is a natural number. For example, for n = 3 the sum of the
first 3 nonzero squares is: 32 + 22 + 12 .

*** Ex. 167 — Design and implement a function to compute n2 , where n is

a natural number, using only addition. Hint: n2 = the sum of the first n odd
natural numbers. For example, 42 = 7 + 5 + 3 + 1.

** Ex. 168 — Design and implement a function to compute the product of two
natural numbers only using addition. For example, 3 * 4 = 3 + 3 + 3 + 3.

82 Computing Tetrahedral Numbers

A tetrahedral is a triangular pyramid. At each layer of a tetrahedral we find a set

of objects, all the same, that form an equilateral triangle. Figure 70 displays three
tetrahedra of different heights made of disks. In Fig. 70a the tetrahedral is of height
one and is built using a single disk. In Fig. 70b the tetrahedral is of height two and
is built using 4 disks: 3 for the bottom level and one for the top level. In Fig. 70c
the tetrahedral is of height three and is built using 10 disks: 6 for the bottommost
layer, 3 for the middle layer, and 1 for the top layer. In Fig. 70c it is easy to see the
triangles that form each layer of the tetrahedral. The numbers 1, 4, and 10 are the
first three numbers in a sequence of numbers called the tetrahedral numbers. The
tetrahedral number in position n of the sequence is the number of objects needed to
build a tetrahedral of height n.
Imagine that you wish to build a tetrahedral of disks (all the same) of height 20.
How many disks do you need? In other words, what is the 20th tetrahedral number?
Given that the answer is not immediately obvious, we need to write a function to
compute tetrahedral numbers. Observe that the position, n, of the desired tetrahedral
number is always a natnum: the first, the fifth, the twentieth, etc. This gives us
insight into how we may design a function to compute tetrahedral numbers. We need
to process a natnum. Thus, we ought to specialize the template for a function on a
natnum.
324 14 Natural Numbers

Fig. 70 Three tetrahedra

(c) Height 3.
(b) Height 2.
(a) Height 1.

Fig. 71 Division of tetrahedral of height 3

(b) Bottom Layer.
(a) Top Layers.

Consider the structure of a natural number. How many disks are needed to build
a tetrahedral of height 0? If the height is 0, it means that it contains no disks.
Therefore, the 0th tetrahedral number is 0. How many disks are needed if the height
is not 0? Once again, consider the tetrahedral in Fig. 70c. We can use a divide and
conquer strategy to determine how to compute its tetrahedral number. Divide the
tetrahedral into two parts: the bottommost layer and all the other layers. This division
is displayed in Fig. 71. Figure 71a is a tetrahedral of height 2. The template suggests
that this value be computed recursively. How do we compute the number of disks in
the bottom layer? Recall that each layer is a triangle. Observe that the bottommost
layer is a triangle of height 3 (i.e., three layers). This means that we need to compute
the number of coins needed for a triangle of height 3. The number of disks needed
to build a triangle is a triangular number. In general, when the height, h, is not zero,
we need to compute hth triangular number and the (h - 1)th tetrahedral. How
are these two numbers combined to obtain the hth tetrahedral number? Once again
looking at Fig. 71, it becomes clear that these two numbers must be added.
Based on the problem analysis we define natnum instances and sample expres-
sions as follows:
;; Sample instances of natnum
(define ZERO 0)
(define NINE 9)
(define SIXTY 60)

;; Sample expressions for nth-tetra

(define ZERO-VAL 0)
(define NINE-VAL (+ (nth-tri NINE) (nth-tetra (sub1 NINE))))
(define SIXTY-VAL (+ (nth-tri SIXTY) (nth-tetra (sub1 SIXTY))))
The chosen names for the functions, nth-tetra and nth-tri, reflect their purpose.
The first sample expression illustrates that for height 0 the number of needed disks
82 Computing Tetrahedral Numbers 325

Fig. 72 Three triangles of disks

(c) Triangle of height 3.
(b) Triangle of height 2.
(a) Triangle of height 1.

is simply 0 as per our problem analysis. The second and third sample expressions
illustrate that for a given nonzero natnum you add the triangular number for the given
natnum and recursively compute the tetrahedral number for the input’s previous
natnum.
To specialize the template abstraction over the sample expressions is performed
to obtain:
;; natnum → natnum
;; Purpose: Compute the nth tetrahedral number
(define (nth-tetra a-natnum)
(if (= a-natnum 0)
0
(+ (nth-tri a-natnum) (nth-tetra (sub1 a-natnum)))))
Once again, our design assumes that the auxiliary function exists and works correctly.
The tests are the following:
;; Tests using sample computations for nth-tetra
(check-expect (nth-tetra ZERO) ZERO-VAL)
(check-expect (nth-tetra NINE) NINE-VAL)

;; Tests using sample values for nth-tetra

(check-expect (nth-tetra 3) 10)
(check-expect (nth-tetra 2) 4)
Once again, observe that small values have been chosen to make it easy for the reader
of our code to verify that the expected value is correct.
We now proceed to design and implement the auxiliary function nth-tri. For
every natural number there is a different disk of triangles. Each of these triangles has
a different height. For example, Fig. 72 displays three triangles of disks of different
heights. Our goal is to compute the number of disks for a given height. Given that
the height is a natnum, reason about the structure of a natnum. How many disks
are used for a triangle of height 0? Height zero means that 0 disks are used. How
many disks are used for a triangle whose height is not 0? The disks at each level of
the triangle must be added. Let us use the triangles in Fig. 72 to reason about this.
For height 1, 1 disk is needed for the bottommost (only) level and the disks needed
for the rest of the levels (0 levels) ought to be computed recursively. This means we
have 1 and 0. These numbers are combined by addition to obtain the total number
of disks needed (i.e., 1 + 0 = 1). For height 2, we need two disks for the bottommost
326 14 Natural Numbers

level and recursively compute the disks needed for a triangle of height 1. Observe
that 2 + 1 = 3, the number of disks needed for a triangle of height 2. For height 3
the number of disks is 3 + the number of disks needed for height 2 = 6. The pattern
holds. The given height is added to the triangular number obtained from height −1.
For example, for height 3:
triangular number for 3
= 3 + 2 + 1
= 3 + 3
= 3 + triangular number for 2
= height + triangular number for (height - 1)
Based on our problem analysis, we can write sample expressions for nth-tri:
;; Sample expressions for nth-tri
(define ZERO-VAL-TRI 0)
(define NINE-VAL-TRI (+ NINE (nth-tri (sub1 NINE))))
(define SIXTY-VAL-TRI (+ SIXTY (nth-tri (sub1 SIXTY))))
The same natnum instances defined for nth-tetra are used. The first expression
illustrates that for 0 the answer is simply 0 as per our problem analysis. The second
and third sample expressions illustrate that for a nonzero natnum the given natnum
is added to the triangular number obtained from the decremented given natnum.
Abstraction over the sample expressions is used to specialize the function defini-
tion template:
;; natnum → natnum
;; Purpose: Compute the triangular number for the given natnum
(define (nth-tri a-natnum)
(if (= a-natnum 0)
0
(+ a-natnum (nth-tri (sub1 a-natnum)))))
Observe that the return value is a natnum because 0 is a natnum and adding two
natnums results in a natnum. The body of the function implements the design
obtained from our problem analysis.
The tests may be written as follows:
;; Tests using sample computations for nth-tri
(check-expect (nth-tri ZERO) ZERO-VAL-TRI)
(check-expect (nth-tri NINE) NINE-VAL-TRI)
(check-expect (nth-tri SIXTY) SIXTY-VAL-TRI)

;; Tests using sample values for f-on-natnum

(check-expect (nth-tri 4) 10)
(check-expect (nth-tri 8) 36)
Once again, observe that small natnums are used in the tests using sample values.
To complete the implementation run the program and make sure all the tests pass.
83 Making Copies 327

* Ex. 169 — The number of disks used for a triangle is equal to the sum of
the natural numbers from the given height, h, down to 0. This sum is equal to
h(h+1)
2 . Refactor nth-tri to take advantage of this formula.

83 Making Copies

Consider making photocopies of a one-page document. You go to a copy machine,

punch in the number of copies needed, and hit the copy button. There is a program
that controls the copy machine that stops when the desired number of copies have
been made. How does that program work?
Instead of making photocopies of a one-page document, we will focus on making
copies of a given BSL+ value. It is straightforward to see that two input values are
needed: the number of copies to make and the value to copy. A user of this function
may need 20 copies. Another user may need 100. Yet another user may need 0. The
first thing to notice is that the number of copies needed is always a natural number.
This gives us insight into how to solve the problem. The function that makes the
copies must process a natural number. What type of value does this function need to
return? How can you represent 20 copies of a value? 100 copies? 0 copies? Observe
that the size of the output must be proportional to the number of copies needed. This
means that the value returned must also be of arbitrary size. That is, it must be a
recursive data type. Given that multiple values must be stored (all the same) using
a (listof X) is natural. Now we know the signature, purpose, and header of the
function:
;; natnum X → (listof bsl-value)
;; Purpose: To make the given number of copies of the
;; given value
(define (make-n-copies n val)
To determine how to build the result list reason about the structure of the natural
number being processed. If the given natural number is 0, then the resulting list is
empty. The empty list contains 0 copies of the given value. What if the given natural
number is not 0? If it is not 0, then at least once copy must be added to the result
list and the rest of the copies are added recursively. The rest of the copies, of course,
may be 0 or more.
The specialization of the template for a function on a natnum starts with the
following value constants and sample natnums:
(define ORIGIN (make-posn 0 0))
(define CSPROF "Marco T. Morazán")

;; Sample instances of natnum

(define ZERO 0)
328 14 Natural Numbers

(define TWENTY 20)

(define FIFTY 50)
The sample expressions are based on the problem analysis above:
;; Sample expressions for make-n-copies
(define ZERO-VAL '())
(define TWENTY-VAL
(cons ORIGIN (make-n-copies (sub1 TWENTY)
ORIGIN)))
(define FIFTY-VAL
(cons CSPROF (make-n-copies (sub1 FIFTY)
CSPROF)))
The first sample expression illustrates how to make 0 copies of, say #true, a value.
The second and third sample expressions illustrate how to make 1 or more copies of
a value. For example, FIFTY copies of CSPROF are needed. Observe that
(make-n-copies (sub1 FIFTY) CSPROF)
is a list containing 49 copies of CSPROF. The 50 needed copies are made by adding
one copy to the front of this list using cons.
Abstraction over the sample expressions leads to the following definition of tem-
plate specialization:
;; natnum X → (listof X)
;; Purpose: To make the given number of copies of the
;; given value
(define (make-n-copies n val)
(if (= n 0)
'()
(cons val (make-n-copies (sub1 n) val))))
The tests are the following:
;; Tests using sample computations for make-n-copies
(check-expect (make-n-copies ZERO #true)
ZERO-VAL)
(check-expect (make-n-copies TWENTY (make-posn 0 0))
TWENTY-VAL)
(check-expect (make-n-copies FIFTY CSPROF)
FIFTY-VAL)

;; Tests using sample values for make-n-copies

(check-expect (make-n-copies 3 100) '(100 100 100))
(check-expect (make-n-copies 5 'x) '(x x x x x))
84 What Have We Learned in This Chapter? 329

* Ex. 170 — Design and implement a function that returns the first n squares,
where n is a natural number.
** Ex. 171 — Let f be
f(x) = 5x5 - 10x3 + 5x + 87

Design and implement a function that takes a natural number, n, as input and
that returns a list of the values of f(x), for natural number x in [n..0].

**** Ex. 172 — The natural numbers greater than 100, natnum>100, are a
subtype of the natural numbers. Provide a data definition for natnum>100.
Design and write a function that takes as input a natnum>100, k, and that
returns the sum of the natural numbers greater than 100 and less than or equal
to k.

84 What Have We Learned in This Chapter?

The important lessons of Chap. 14 are summarized as follows:

• Natural numbers are of arbitrary size and are a subtype of the number type.
• Natural numbers are processed recursively.
• The constructor for natural numbers is add1.
• The selector for natural numbers is sub1.
• Reasoning statically about the structure of a natnum is a powerful technique to
solve problems involving natnums.
• Functions that process natural numbers may return any type of data.
• Structural recursion is a form of divide-and-conquer problem solving.
Chapter 15
Interval Processing

Chapter 5 introduced interval types to describe a value and used to design functions
that make decisions. This chapter returns to intervals, but not as a mechanism solely
used to describe a value. Instead, intervals are considered compound data that may
be processed. Chapter 14 hinted at the concept of an interval as data that may be
processed. If you think about it carefully, processing a natnum, n, means that all the
natural numbers in [0..n] must be processed. The first call to the function does
something for n, the second call does something for n - 1, and so forth until the last
call does something for 0. Processing natural numbers fixes the lower end of the
interval to 0, while the higher end of the interval is of arbitrary size.
This chapter generalizes the concept of an interval such that both ends may
vary. For example, in your Mathematics textbooks you may have seen an interval of
integers described as
[i..j], where i < j
This is intended to represent all the integers between i and j. The square brackets
indicate that i and j are included in the interval. In other words, [i..j] represents
all the integers greater than or equal to i and less than or equal to j. If a parenthesis
is used, it means that the value is not included. For example, the following means
that i is not included in the interval:
(i..j], where i < j
From this you can surmise the intended meaning of [i..j) and (i..j).
The question now is how do you process an interval of integers. The description
found in many Mathematics textbooks hides the structure of an interval making it hard
to reason about how an interval is processed. Clearly, an interval is data of arbitrary
size and, therefore, we need a recursive data definition. Can you discern from the
mathematical description above what the base case or cases are? For example, can
an interval be empty?
Understanding to how to process intervals is important. In future courses you will
study compound data called a vector (or array). Usually, the part of the vector
that needs to be processed is defined by an interval of natural numbers (a subtype
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 331
M. T. Morazán, Animated Problem Solving, Texts in Computer Science,
https://doi.org/10.1007/978-3-030-85091-3_15
332 15 Interval Processing

of the integers). Before studying vector intervals, however, this chapter focuses on
integer intervals.

85 Interval Data Definition

Without loss of generality, we shall focus on intervals that include their end points,
that is, on intervals such as [i..j]. It is straightforward for you to extend the work
presented to an interval that does not include one or both end points.
Consider what the interval [-2..4] represents
[-2 -1 0 1 2 3 4]
It represents 7 consecutive integers. This fact, however, tells us nothing about the
structure of an interval that we need to determine in order to process an interval.
A natural question to ask what is the smallest interval possible. For example, is
[23..23] the smallest an interval can be? It only contains 1 number: 23. Can an
interval contain 0 numbers? Thinking carefully about this ought to lead you to the
conclusion that an interval can contain 0 numbers. In other words, an interval may
be empty—a fact well-hidden in the mathematical description above. For example,
[0..-1] is an empty interval. There are no integers greater than or equal to 0 and
less than or equal to -1. Given an interval [low..high], how do we know that the
interval is empty? If low is greater than high, then the interval is empty.
How do we define a non-empty interval? Once again consider [-2..4]:
[-2 -1 0 1 2 3 4]
Here low = -2 and high = 4. Is there any natural way to decompose this interval
into two or more parts? Given that the interval is not empty it must have at least one
number. Let us take that number to be, 4, the high-end of the interval. We have left:
[-2 -1 0 1 2 3]
This is an interval. Specifically, it is [-2..3]. We can rewrite [-2..4] as follows:
[[-2..3] 4]
This means that a non-empty interval has a structure: the high number and the
subinterval with the rest of the numbers. Observe that the high number of the
subinterval is 3. That is, the high end of the given interval is decremented by 1 (i.e.,
(sub1 4) = 3). We can now write a data definition for an interval as follows:
An interval ([low..high]) is two integers such that it is
either the:
1. empty interval (interpretation: low > high)
2. non-empty interval (interpretation: low ≤ high)
[[low..(sub1 high)] high]
85 Interval Data Definition 333

Fig. 73 Template for a function on an interval

|# ;; Sample instances of interval
(define LOW1 . . .) (define HIGH1 . . .)
(define LOW2 . . .) (define HIGH2 . . .)
...

;; [int..int] . . . ...
;; Purpose: . . .
(define (f-on-interval low high . . .)
(if (> low high)
...
(. . .high. . .(f-on-interval low (sub1 high). . .
. . .low. . ..(f-on-interval (add1 low) high). . .)))

;; Sample expressions for f-on-interval

(define LOW1-HIGH1-VAL . . .)
(define LOW2-HIGH2-VAL . . .)
...
;; Tests using sample computations for f-on-interval
(check-expect (f-on-interval LOW1 HIGH1 . . .) LOW1-HIGH1-VAL)
(check-expect (f-on-interval LOW2 HIGH2 . . .) LOW2-HIGH2-VAL)
...
;; Tests using sample values for f-on-interval
(check-expect (f-on-interval . . .) . . .)
(check-expect (f-on-interval . . .) . . .)
...
#|

Observe that an interval is compound data. It is always represented using two integers.
At first, this might suggest it is data of finite size. However, remember that an interval
represents an arbitrary number of consecutive integers. Further observe that the data
definition contains information about how to interpret each variety of interval. This
is important because it explains how the two varieties are distinguished. For an
empty interval low is greater than high. For a non-empty interval, we have high
and the subinterval [low..(sub1 high)]. This definition suggests that an interval
is processed from high down to low until the lower subinterval is empty.
The choice of dividing an interval into high and [low..(sub1 high)] seems
rather arbitrary. Just as easily we can define a non-empty interval as
[low [(add1 low)..high]
The subinterval may be on the high end. This definition suggests that an interval
is processed from low to high until the higher subinterval is empty. Which is
correct? They are both correct. Observe that in both cases the subinterval is smaller,
which tells us that the circularity in the data definition is not infinite. Eventually, the
subinterval is empty and the circularity stops. This means we can refine the interval
data definition to be:
An interval ([low..high]) is two integers such that it
is either:
334 15 Interval Processing

1. empty interval (interpretation: low > high)

2. non-empty interval (interpretation: low ≤ high)
[[low..(sub1 high)] high] or [low [(add1 low)..high]
This data definition is different from any other we have seen before. It states that
we can choose how to decompose a non-empty interval. In terms of computing this
means that an interval may be processed from high down to low or from low to
high. In fact, a computation may combine both ways of decomposing an interval. It
is important to emphasize that there are only two varieties of intervals, not three.
Different manners of decomposing a variety do not add varieties to a data definition.
Figure 73 displays the template for a function on an interval. The template suggests
defining at least two intervals (one for each variety). Observe that each interval
instance requires defining two integer constants: a low and a high value. The signature
for the definition template uses our interval notation to indicate that the two integers
represent an interval. The fact that an interval is represented by two integers is
captured in the function header that has least two parameters: low and high. The
body of the definition template distinguishes between the interval varieties by testing
(> low high) as per our interpretation of an empty interval. The template also
clearly suggests that a non-empty interval may be processed in two different ways:
process the low or the high element and recursively process the rest of the elements.
The rest of the template is what ought to be expected. There is a sample expression for
each defined interval instance. The values of these expressions, the sample instances,
and the function are used to define the tests using sample computations. Finally, as
before, tests using sample values are also required to validate that the function works
beyond correctly computing the value of sample expressions.

86 Revisiting Factorial

In Chap. 14, factorial was designed around processing a natural number. We

now consider a different design around processing an interval. We know that n!
is the product of all the natural numbers in [1..n]. This means that we may
design factorial around processing this interval. The refactored code is dis-
played in Fig. 74. Observe that in the sample expressions the auxiliary function
interval-product is called to compute the product of the integers in the interval
from 1 to a given natural number. Given that the natural numbers are a subset of the
integers, the intervals are valid. Abstraction over the sample expressions yields the
refactored function definition. The tests remain unchanged as expected because refac-
toring does not change the purpose of the function. Further observe that factorial
is no longer recursive because it does not process the given natural number. Instead,
it calls the auxiliary interval-processing function on [1..a-natnum].
Now we must design the auxiliary function to compute the product of a given
interval. We know from high school algebra that multiplication is commutative. This
means that it does not matter if the given interval is processed from high down to low
or from low to high. Arbitrarily, we choose to process the interval from high down
86 Revisiting Factorial 335

Fig. 74 Factorial refactored using interval processing

;; Sample intances of natnum
(define ZERO 0) (define TEN 10) (define FIFTY 50)

;; natnum natnum
;; Purpose: Compute the factorial of the given natnum
(define (factorial a-natnum) (interval-product 1 a-natnum))

;; Sample expressions for factorial

(define ZERO-VAL (interval-product 1 ZERO))
(define TEN-VAL (interval-product 1 TEN))
(define FIFTY-VAL (interval-product 1 FIFTY))

;; Tests using sample computations for factorial

(check-expect (factorial ZERO) ZERO-VAL)
(check-expect (factorial TEN) TEN-VAL)
(check-expect (factorial FIFTY) FIFTY-VAL)

;; Tests using sample values for f-on-natnum

(check-expect (factorial 3) 6)
(check-expect (factorial 5) 120)

to low. Now, think about the structure of an interval. If the given interval is empty,
a number that does not change a product must be returned. The only such value is
1. Recall that any number, x, times 1 is x. If the interval is not empty, the function
template suggests creating the final product using high and the product obtained
from [low..high-1]. All that is needed is to multiply these two values.
Based on this problem analysis, we may write interval instances and sample
expressions as follows:
;; Sample instances of interval
(define LOW1 1)
(define HIGH1 0)
(define LOW2 -3)
(define HIGH2 5)
(define LOW3 4)
(define HIGH3 7)

;; Sample expressions for interval-product

(define LOW1-HIGH1-VAL 1)
(define LOW2-HIGH2-VAL
(* HIGH2 (interval-product LOW2 (sub1 HIGH2))))
(define LOW3-HIGH3-VAL
(* HIGH3 (interval-product LOW3 (sub1 HIGH3))))
The first interval is empty. Therefore, the sample expression to compute its product
is simply 1 as suggested by our problem analysis. The second and third intervals
are not empty. The sample expressions to compute their product, as suggested by
336 15 Interval Processing

our problem analysis, multiply the high-end value with the product of the rest of the
interval.
Abstracting over the sample expressions yields the following specialization of the
definition template:
;; [int..int] → int
;; Purpose: Compute the product of the integers in the
;; given interval
(define (interval-product low high)
(if (> low high)
1
(* high (interval-product low (sub1 high)))))
Observe that the function implements our design idea by distinguishing between an
empty and a non-empty interval.
Finally, tests may be written as follows:
;; Tests using sample computations for interval-product
(check-expect (interval-product LOW1 HIGH1) LOW1-HIGH1-VAL)
(check-expect (interval-product LOW2 HIGH2) LOW2-HIGH2-VAL)
(check-expect (interval-product LOW3 HIGH3) LOW3-HIGH3-VAL)

;; Tests using sample values for interval-product

(check-expect (interval-product 3 5) 60)
(check-expect (interval-product 6 9) 3024)
Each test provides a valid interval to the function as input. The tests using sample
values make use of intervals for which any reader of the code may easily verify that
the expected value is correct. Run the program and make sure all the tests pass.
We now have two different function implementations to compute the factorial of
a natural number. Which one is preferable? They both compute the correct value.
Therefore, in that regard there is no difference. Is there a difference in the number of
multiplications performed? They both perform the same number of multiplications
and, therefore, no difference there. Is one design more versatile? Here there is
a difference. The factorial design in Chap. 14 is limited to only computing a
factorial value. In contrast, the design in this chapter can be used to do more than just
compute a factorial value. The tests for interval-product illustrate computations
that cannot be performed by the factorial function from Chap. 14. More versatile
designs allow programmers to solve different problems with the same function.

87 Creating an Army of Aliens

For the next Aliens Attack version an initial army of aliens must be created. This
army of aliens is represented as a list of aliens. Consider the problem of creating this
list. Figure 75 displays Aliens Attack with an initial list of aliens. The list of aliens
87 Creating an Army of Aliens 337

Fig. 75 Aliens attack with an army of aliens

contains aliens in three different lines. Each line has the same number of aliens lined
up to form columns. Therefore, a decision must be made on the number of alien
lines and the number of aliens per line. For the army of aliens displayed in Fig. 75
the following constants are defined:
(define ALIEN-LINES 3)
(define ALIENS-PER-LINE 12)
The aliens start at the top row. That is, each alien in the top row of the scene has an
image-y of 0. The alien line is placed toward the scene’s horizontal middle and the
aliens span the same range of image-x values in every line.
These observations give us insight into how to create the initial list of aliens. The
number of alien lines that need to be computed, num-lines, is a natural number.
We can process this natural number to create the needed list of aliens. At each step,
a new alien line is added to the list of aliens using the number of lines that still need
to be computed to determine the image-y value for the aliens in the new line. Each
line of aliens may be of arbitrary size and, therefore, is represented as a list of aliens.
Now, think about the structure of a natural number. If num-lines is 0, then no
more alien lines need to be added and the empty list of aliens is returned. Otherwise,
a line of aliens is created for the target range of image-x values and this new
line is added to the list of aliens obtained by recursively processing num-lines’s
predecessor. Observe that the range of image-x values are consecutive values. That
is, it may be represented using an interval.
This design idea is still incomplete, but it allows us to start specializing the sample
instances and the sample expressions required by the template for a function on a
natural number as follows:
338 15 Interval Processing

Fig. 76 Number of alien lines left and image-y values for alien line
num-lines image-y Value
num-lines image-y Value 5 4
3 2 4 3
2 1 3 2
1 0 2 1
1 0

;; Sample instances of natnum

(define ZERO 0)
(define THREE 3)
(define FIVE 5)

;; Sample expressions for create-alien-army

(define ZERO-VAL '())
(define THREE-VAL (. . . (make-alien-line image-y low high )
(create-alien-army (sub1 THREE))))
(define FIVE-VAL (. . . (make-alien-line image-y low high )
(create-alien-army (sub1 FIVE))))
When the number of lines is 0, as per our design idea, the empty list is returned. When
the number of lines is not 0, the previous natural number is recursively processed to
obtain a list of aliens that is only missing the first line. An auxiliary function is used
to compute a single alien line. This auxiliary function needs the image-y value for
the aliens in the line and the interval for image-x values.
What should the image-y value be? To answer this question consider the tables
in Fig. 76. The tables present the values for, num-lines, the number of lines still
not computed and the corresponding image-y value for the next alien line. Do
you see the pattern? The image-y value is always one less than num-lines. This
means that for each call to make-alien-line the needed image-y value is (sub1
num-lines).
What is the needed interval for image-x values? Given that the aliens ought to
be relatively centered on the x-axis the aliens can be placed around the midpoint of
possible image-x values: (/ MAX-CHARS-HORIZONTAL 2). Half the aliens ought
to be placed before the midpoint and half after the midpoint. This leads to the
following interval definition for image-x values:
(define ALIEN-LINE-XLOW (- (/ MAX-CHARS-HORIZONTAL 2)
(/ ALIENS-PER-LINE 2)))
(define ALIEN-LINE-XHIGH (sub1 (+ (/ MAX-CHARS-HORIZONTAL 2)
(/ ALIENS-PER-LINE 2))))
Half the number of aliens per line is added and subtracted from the midpoint. This
makes the interval one number too big. Therefore, 1 is subtracted from the high
end of the interval. The choice of subtracting 1 from the high end of the interval is
arbitrary. Instead, 1 could be subtracted from the low end of the interval.
Finally, we need to decide how to combine an alien line (i.e., a (listof alien))
and a (listof alien). Consider this example using cons:
87 Creating an Army of Aliens 339

(cons (list (make-posn 9 0) (make-posn 10 0) (make-posn 11 0))

(list (make-posn 9 1) (make-posn 10 1) (make-posn 11 1)
(make-posn 9 2) (make-posn 10 2) (make-posn 11 2)))
The resulting list is
(list
(list (make-posn 9 0) (make-posn 10 0) (make-posn 11 0))
(make-posn 9 1)
(make-posn 10 1)
(make-posn 11 1)
(make-posn 9 2)
(make-posn 10 2)
(make-posn 11 2))
Observe that this is not a list of aliens. The first element of the list is not an alien.
Instead, it is a list of aliens. Remember that cons adds its first argument to the
front of its second (list) argument. This is the wrong thing to do in this case. We
need a function that combines two lists into one list. Such a function exists in BSL+:
append. The append function takes as input zero or more lists and creates a new list
that contains all the elements in the given lists starting with the first given list. For
example, consider:
(append
(list (make-posn 9 0) (make-posn 10 0) (make-posn 11 0))
(list (make-posn 9 1) (make-posn 10 1) (make-posn 11 1)
(make-posn 9 2) (make-posn 10 2) (make-posn 11 2)))
The resulting list is:
(list
(make-posn 9 0) (make-posn 10 0) (make-posn 11 0)
(make-posn 9 1) (make-posn 10 1) (make-posn 11 1)
(make-posn 9 2) (make-posn 10 2) (make-posn 11 2))
Observe that the result is a list of aliens. That is, every list element is an alien.
We can now finish the specialization of the create-alien-army sample expres-
sions as follows:
;; Sample expressions for create-alien-army
(define ZERO-VAL '())
(define THREE-VAL (append (make-alien-line (sub1 THREE)
ALIEN-LINE-XLOW
ALIEN-LINE-XHIGH)
(create-alien-army (sub1 THREE))))
(define FIVE-VAL (append (make-alien-line (sub1 FIVE)
ALIEN-LINE-XLOW
ALIEN-LINE-XHIGH)
(create-alien-army (sub1 FIVE))))
340 15 Interval Processing

To specialize the definition template we use abstraction over the sample expressions
to obtain
;; natnum → loa
;; Purpose: Create initial alien army with the given number
;; of lines
(define (create-alien-army num-lines)
(if (= num-lines 0)
'()
(append (make-alien-line (sub1 num-lines)
ALIEN-LINE-XLOW
ALIEN-LINE-XHIGH)
(create-alien-army (sub1 num-lines)))))
The tests are written as follows:
;; Tests using sample computations for create-alien-army
(check-expect (create-alien-army ZERO) ZERO-VAL)
(check-expect (create-alien-army THREE) THREE-VAL)
(check-expect (create-alien-army FIVE) FIVE-VAL)

;; Tests using sample values for create-alien-army

(check-expect
(create-alien-army 1)
(list (make-posn 4 0) (make-posn 5 0) (make-posn 6 0)
(make-posn 7 0) (make-posn 8 0) (make-posn 9 0)
(make-posn 10 0) (make-posn 11 0) (make-posn 12 0)
(make-posn 13 0) (make-posn 14 0) (make-posn 15 0)))
A small natural number, 1, is used in the test using a sample value to make it easy
for any code reviewer to verify that the expected value is correct.
We may now proceed to design and implement make-alien-line. This function
must traverse the given interval and create a list of aliens. Think in terms of the
structure of an interval. If the interval is empty, there are no aliens to create and
the empty list of aliens is returned. Otherwise, a new alien is created using low-end
value of the given interval and the given image-y value. This new alien is added to
the front of the list of aliens obtained from processing the rest of the interval.
The function is implemented by specializing the template for a function on an
interval. The following sample interval instances and sample expressions are defined:
;; Sample instances of interval
(define LOW1 THREE)
(define HIGH1 ZERO)
(define LOW2 ZERO)
(define HIGH2 THREE)
(define LOW3 THREE)
(define HIGH3 FIVE)
87 Creating an Army of Aliens 341

;; Sample expressions for make-alien-line

(define LOW1-HIGH1-VAL '())
(define LOW2-HIGH2-VAL (cons (make-posn LOW2 3)
(make-alien-line 3
(add1 LOW2)
HIGH2)))
(define LOW3-HIGH3-VAL (cons (make-posn LOW3 2)
(make-alien-line 2
(add1 LOW3)
HIGH3)))
The first interval is empty and the remaining two are not empty. As per our problem
analysis, when the interval is empty, the line of aliens returned is empty. Otherwise,
a new alien is added to the front of the list containing the rest of the aliens in the line.
Observe that the new alien is constructed using the low-end value of an interval and
a given image-y value (i.e., 3 and 2 in the tests).
Specializing the definition template is done by abstracting over the sample ex-
pressions:
;; image-y [image-x..image-x] → loa
;; Purpose: Create an loa with an alien for each value in
;; the given interval using the given image-y
(define (make-alien-line an-image-y low high)
(if (> low high)
'()
(cons (make-posn low an-image-y)
(make-alien-line an-image-y (add1 low) high))))
As per our problem analysis, each alien constructed contains the low value of the
given interval and the given image-y value.
The tests may be written as follows:
;; Tests using sample computations for make-alien-line
(check-expect (make-alien-line 3 LOW2 HIGH2) LOW2-HIGH2-VAL)
(check-expect (make-alien-line 2 LOW3 HIGH3) LOW3-HIGH3-VAL)

;; Tests using sample values for make-alien-line

(check-expect (make-alien-line 1 17 19)
(list (make-posn 17 1)
(make-posn 18 1)
(make-posn 19 1)))
Once again, observe that the interval used in the sample value test is small enough
that any code reviewer may verify that the expected value is correct. Run the program
and make sure all the tests pass.
342 15 Interval Processing

88 Largest Prime in an Interval

You may recall from high school Mathematics that a prime number is a natural
number greater than 1 that is not the product of smaller natural numbers. For example,
7 is a prime number because it is not divisible by any natural number in [2..6].
On the other hand, 51 is not a prime number because it is divisible by 3 and 17. To
determine if a given natural number, n, greater than 1 is prime, we must establish
that it is not divisible by any natural number in [2..n]. It is not necessary to
check all the natural numbers in this interval. Observe that it suffices to check if any
natural in [2..(quotient n 2)] divides n because none of the natural numbers
in [(quotient n 2)..n-1] divide n.
Consider the problem of finding the largest prime number in a given interval. As
you know, you have two options: you can process the interval from low to high or
from high down to low. Does it matter which one you choose? Consider finding
the largest prime in [9..22]. If this interval is processed from low to high, you
discover that 9 and 10 are not prime. This means that may think of the interval as
divided into two pieces:
[9..10]: interval has no primes
[11..22]: interval not explored and may have a largest prime
In the next step you determine that 11 is prime. Now, you may think of the interval
in three pieces:
[9..10]: interval has no primes
[11..11]: 11 is the smallest prime
[12..22]: interval not explored and may have a larger prime
How do you determine if 11 is the largest prime? You need to find, if it exists, the
largest prime in [12..22]. If it exists, then that number is the largest prime. If it
does not exist, then 11 is the largest prime. In this example, it is clear that 11 is not
the largest prime but you do not know that until the entire unexplored interval is
processed. This uncertainty requires the function to somehow remember 11 as the
previous largest prime in case the unexplored interval does not have a prime.
Now consider processing the interval from high down to low. This means that
you first discover that 22, 21, and 20 are not primes. You may think of the interval
divided as follows:
[20..22]: interval has no primes
[9..19]: interval not explored and may have a largest prime
In the next step you determine that 19 is a prime and you may think of the interval
divided as follows:
[20..22]: interval has no primes
[19..19]: 19 is the largest prime
[9..18]: interval not explored and may have a largest prime
88 Largest Prime in an Interval 343

You immediately know that the first prime found is the largest one. There is no need
to process the unexplored interval. Clearly, for this problem it is easier to process the
interval from high down to low. If the given interval is empty, the function returns
-1 to indicate that the interval does not contain a prime number. If the interval is not
empty, the largest number, high, in the interval is tested to see if it is not divisible by
any natural number in [2..(quotient high 2]. If so, high is returned as it is the
largest prime. Otherwise, the rest of the interval is recursively processed. Observe
that there are three, not two, conditions that need to be distinguished.
Based on the problem analysis, the following three interval instances and three
sample expressions are developed:
;; Sample instances of interval
(define LOW1 2)
(define HIGH1 1)
(define LOW2 2)
(define HIGH2 23)
(define LOW3 13)
(define HIGH3 16)

;; Sample expressions for largest-prime

(define LOW1-HIGH1-VAL -1)
(define LOW2-HIGH2-VAL HIGH2)
(define LOW3-HIGH3-VAL (largest-prime LOW3 (sub1 HIGH3)))
The first interval is empty. The second interval has a high-end number that is prime.
The third interval does not have a high-end number that is prime. The sample
expressions illustrate how the answer is computed for each different type of interval.
For the empty interval, as per our design idea, the answer is -1. If the interval is
not empty and the high-end number is prime, then the high-end number is returned.
Finally, if the high-end number is not prime, then the rest of the interval is processed.
We can now proceed with the specialization of the definition template. A condi-
tional with three stanzas is required as follows:
;; [int..int] → int
;; Purpose: Return largest prime in given interval.
;; If none, return -1
;; ASSUMPTION: low > 1
(define (largest-prime low high)
(cond [(> low high) -1]
[(not (is-divisible? high 2 (quotient high 2))) high]
[else (largest-prime low (sub1 high))]))
The first thing to observe is that the function returns an int and not a prime natural
number. This is because the function may return −1. The second observation is
the explicit assumption that low is greater than 1. This is because a prime number
must be greater than 1. The body of the function implements the design idea. The
auxiliary function is-divisible? is used to determine if high is divisible by any
344 15 Interval Processing

natural number in [2..(quotient high 2)]. If it is not, high is the largest prime
as outlined in our problem analysis. Otherwise, the rest of the interval is recursively
searched for a prime.
We may specialize the tests as follows:
;; Tests using sample computations for largest-prime
(check-expect (largest-prime LOW1 HIGH1) LOW1-HIGH1-VAL)
(check-expect (largest-prime LOW2 HIGH2) LOW2-HIGH2-VAL)
(check-expect (largest-prime LOW3 HIGH3) LOW3-HIGH3-VAL)

;; Tests using sample values for largest-prime

(check-expect (largest-prime 2 2) 2)
(check-expect (largest-prime 24 28) -1)
(check-expect (largest-prime 29 35) 31)
The tests using sample values illustrate that the function correctly works for non-
empty intervals when there is and there is not a prime number in the given interval.
An important lesson to derive from this problem so far is that it is worth exploring
different designs. In this case, one design led to a simpler solution to the problem. A
good problem solver explores different designs, whenever possible, before commit-
ting to one.
We turn our attention to the design of is-divisible?. This function must
also process an interval. Once again, we must decide in what direction to process the
given interval. If the given natural number is not divisible by any number in the given
interval, then the entire interval must be processed to determine this. The direction
of the processing uses makes no difference. If the given number is divisible, the
search for a factor may stop when any factor is found. It also does not matter what
direction the search is done in. We arbitrarily choose from low to high. Observe
that this is an ORing operation on an interval (not a list). We need to determine if
any natural number in the given interval is a factor of the given natural number. The
given natural number is divisible if the given interval is not empty and the interval’s
low-end number divides the given natural number or a number in the rest of the
interval divides the given natural number. Observe that, just as with lists, an ORing
operation does not require a conditional expression as suggested by the template for
a function on an interval.
Based on our problem analysis we can write sample intervals and sample expres-
sions as follows:
;; Sample instances of interval
(define LOW4 21)
(define HIGH4 20)
(define LOW5 2)
(define HIGH5 13)
(define LOW6 2)
(define HIGH6 8)
88 Largest Prime in an Interval 345

;; Sample expressions for is-divisible?

(define LOW4-HIGH4-VAL
(and (not (> LOW4 HIGH4))
(or (= (remainder 11 LOW4) 0)
(is-divisible? 11 (add1 LOW4) HIGH4))))
(define LOW5-HIGH5-VAL
(and (not (> LOW5 HIGH5))
(or (= (remainder 27 LOW5) 0)
(is-divisible? 27 (add1 LOW5) HIGH5))))
(define LOW6-HIGH6-VAL
(and (not (> LOW6 HIGH6))
(or (= (remainder 17 LOW6) 0)
(is-divisible? 17 (add1 LOW6) HIGH6))))
The sample expressions illustrate the computation for empty and non-empty intervals
using both numbers that are and that are not divisible by the any member of an
interval. As expected all the expressions are similar because there is no need for a
conditional expression.
Abstraction over the sample expressions yields the specialized definition template:
;; natnum [natnum..natnum] → Boolean
;; Purpose: Determine if the given natnum is divisible by
;; any natnum in the given interval
(define (is-divisible? target low high)
(and (not (> low high))
(or (= (remainder target low) 0)
(is-divisible? target (add1 low) high))))
Finally, the test templates are specialized as follows:
;; Tests using sample computations for is-divisible?
(check-expect (is-divisible? 11 LOW4 HIGH4) LOW4-HIGH4-VAL)
(check-expect (is-divisible? 27 LOW5 HIGH5) LOW5-HIGH5-VAL)
(check-expect (is-divisible? 17 LOW6 HIGH6) LOW6-HIGH6-VAL)

;; Tests using sample values for largest-prime

(check-expect (is-divisible? 2 2 2) #true)
(check-expect (is-divisible? 51 2 25) #true)
(check-expect (is-divisible? 53 2 26) #false)
The tests using sample computations illustrate that the abstraction is correctly done.
The first test using a sample value illustrates that the function works when the interval
only contains the smallest possible value. The second such test illustrated that the
function works when the number is divisible by some number in the interval. Finally,
the third such test illustrates that the function works when the number is not divisible
by some number in the interval. Run the program and make sure all the tests pass.
Another important lesson from this example is that a problem may require pro-
cessing several intervals. Frequently, although not always, each interval that needs
346 15 Interval Processing

to be processed requires a different function to be designed. Such is the case for this
example. Furthermore, the different intervals do not have to be processed in the same
direction.

* Ex. 173 — Design and implement a function to sum the integers in a given
interval.
** Ex. 174 — Design and implement a function to sum the even positive inte-
gers in a given interval.

** Ex. 175 — Design and implement a function to determine if an interval

contains a square. For example, [2..5] contains a square because 4 = 22 .

*** Ex. 176 — BSL+ includes the function list-ref that has the following
signature:
(listof X) natnum → X throws error

This function returns the ith element of the list (counting from 0). If the list is
too short, then it does not have an ith element and the function throws an error.
Design and implement a function to determine if a given (listof String)
contains a given string in a given interval. For example, the following tests
should pass:
(check-expect
(contains-in? "Basia"
(list "Walter" "Skyler" "Basia" "Madrid")
0
2)
#true)
(check-expect
(contains-in? "Bob"
(list "Walter" "Skyler" "Basia" "Madrid")
0
1)
#false)

** Ex. 177 — Design and implement a function to determine if an interval

does not contain any prime numbers.

** Ex. 178 — Carefully explain why the following test passes:

(check-expect (is-divisible? 2 2 2) #true)
89 What Have We Learned in This Chapter? 347

89 What Have We Learned in This Chapter?

The important lessons of Chap. 15 are summarized as follows:

• An interval is compound data of arbitrary size consisting of 0 or more consecutive
integers.
• Four different interval types may be defined:
– [low..high] includes low and high.
– (low..high] does not include low and includes high.
– [low..high) includes low and does not include high.
– (low..high) does not include both low and high.
• An interval may be processed from the low-end value to the high-end value or it
may be processed from the high-end value down to the low-end value.
• We may reason about the structure of an interval to solve problems.
• Different manners of decomposing a variety of data do not add varieties to the
data definition.
• Functions that can be used to solve more than one problem provide powerful
versatility to programmers that make problem solving easier.
• Problem solving using intervals may require more than one interval to be pro-
cessed.
• Different intervals processed by a program can be processed in different directions.
• Exploring different designs may lead to a simpler solution to a problem.
Chapter 16
Aliens Attack Version 4

This chapter presents the next refinement for Aliens Attack that adds multiple aliens
and shots. In the past few chapters some of the needed auxiliary functions have been
designed. This chapter brings together the design of the refinement with the already
designed auxiliary functions. The first step is to refine the data definition for a world
and, consequently, the template for functions on a world. This will lead us to the
refinements needed to update the game.
This refinement is the largest one so far. This is why auxiliary functions were
defined throughout the previous chapters (Chaps. 13–15). In this manner, the refine-
ment is not overwhelming. Such an approach is common in software engineering.
With a data representation idea in place, auxiliary functions may be designed first.
In other words, a bottom-up approach may be used. A primary lesson is that in the
software development process both top-down and bottom-up designs may coexist.

90 New world Data Definition and Function Template

The world refinement changes the game from having a single alien and a single shot
to having multiple aliens and multiple shots. The data definitions for a list of aliens
(loa) and a list of shots (los) were defined in Sect. 67. We integrate these to the
world data definition as follows:
;; A world is a structure: (make-world rocket loa dir los)
(define-struct world (rocket aliens dir shots))
Observe that the names of the selectors change from world-alien and world-shot
to, respectively, world-aliens and world-shots. We do so to make sure the
structure’s field names convey to any reader of our code what they represent. This
refinement also means that the template for a function on a world must also be
refined. Specifically, a function on a world should no longer call a function on an
alien or a function on a shot. Instead, it must call functions on a loa and on a los.

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 349
M. T. Morazán, Animated Problem Solving, Texts in Computer Science,
https://doi.org/10.1007/978-3-030-85091-3_16
350 16 Aliens Attack Version 4

Based on these observations the template for a function on a world is refined to

be
;; world ... → ...
;; Purpose: ...
(define (f-on-world a-world ...)
...(f-on-rocket (world-rocket a-world) ...)
...(f-on-loa (world-aliens a-world) ...)
...(f-on-dir (world-dir a-world) ...)
...(f-on-los (world-shots a-world) ...))

;; Sample instances for world

(define WORLD1 (make-world ...))

;; Sample expressions for f-on-world

(define WORLD-VAL1 ...)

;; Tests using sample computations for f-on-world

(check-expect (f-on-world WORLD1 ...) WORLD-VAL1)
...

;; Tests using sample values for f-on-world

(check-expect (f-on-world ...) ...)
The changes are all in the definition template. The changes clearly inform us that
solving a world problem now may involve solving loa and los problems. A world-
processing function should not call functions that solve an alien or a shot problem.
This is important to realize because it helps organize your thoughts as a problem
solver.
The sample worlds may now be updated as follows:
(define INIT-LOA (create-alien-army ALIEN-LINES))

(define INIT-WORLD (make-world INIT-ROCKET

INIT-LOA
INIT-DIR
E-LOS))

(define INIT-WORLD2 (make-world INIT-ROCKET2

(list INIT-ALIEN2)
DIR2
(list SHOT2)))

(define WORLD3 (make-world 7

(list (make-posn 3 3))
'right
(list (make-posn 3 3))))
91 The draw-world Refinement 351

The initial list of aliens, INIT-LOA, is defined using the create-alien-army

developed in Sect. 87. This list of aliens and the initial list of shots (defined in
Sect. 67) are used to define, INIT-WORLD, the initial world. The remaining two
worlds are updated to contain the single alien and the single shot used in Aliens
Attack 3.
Now, our job will follow a top-down approach to the refinement. The run function
and its big-bang expression remain unchanged. The refinements needed, as before,
start with the world-processing functions in the big-bang expression.

91 The draw-world Refinement

Instead of drawing an alien and a shot, drawing the world now means drawing a
loa and a los. The sample expressions for draw-world must be refined. To do so
we must realize that two new problems must be solved: drawing a list of aliens and
drawing a list of shots. We assume these functions exist and that they need as input
the proper list to draw and the scene to draw in. The refined sample expressions are
;; Sample expressions for draw-world
(define WORLD-SCN1
(draw-los (world-shots INIT-WORLD)
(draw-loa (world-aliens INIT-WORLD)
(draw-rocket
(world-rocket INIT-WORLD)
E-SCENE))))

(define WORLD-SCN2
(draw-los (world-shots INIT-WORLD2)
(draw-loa (world-aliens INIT-WORLD2)
(draw-rocket
(world-rocket INIT-WORLD2)
E-SCENE))))
Abstracting over the sample expressions yields the refined world drawing function:
;; world → scene
;; Purpose: To draw the world in E-SCENE
(define (draw-world a-world)
(draw-los (world-shots a-world)
(draw-loa (world-aliens a-world)
(draw-rocket (world-rocket a-world)
E-SCENE))))
352 16 Aliens Attack Version 4

It is important not to forget to refine the tests. The tests using sample computations
do not need to be refined because they do not explicitly refer to a world. The tests
using sample values, on the other hand, do explicitly refer to world instances. Each
test is refined as follows to use a loa and a los:
;; Tests using sample computations for draw-world
(check-expect (draw-world (make-world INIT-ROCKET2
(list INIT-ALIEN)
DIR3
empty))

(check-expect (draw-world (make-world INIT-ROCKET

(list INIT-ALIEN2)
DIR2
(list SHOT2)))

)
Next, it is necessary to design the two auxiliary functions. To draw the aliens a
list of aliens must be processed. This suggests reasoning about the structure of a
loa. If the given list of aliens is empty, there are no aliens to draw and the result is
the given scene. If the given list of aliens is not empty, then the rest of the aliens are
recursively processed to obtain a scene that contains all the aliens except the first. In
this scene the first alien is drawn to obtain a scene that contains all the aliens.
Based on this design idea we can write the following sample expressions:
;; Sample expressions for draw-loa
(define ELOA-VAL E-SCENE)
(define ILOA-VAL (draw-alien (first INIT-LOA)
(draw-loa (rest INIT-LOA)
E-SCENE)))
91 The draw-world Refinement 353

There is a sample expression for each loa variety. The sample expression for a
non-empty loa uses the previously designed draw-alien function from Sect. 49.2.
Based on the sample expressions the definition template is specialized as follows:
;; loa scene → scene
;; Purpose: To draw the given loa in the given scene
(define (draw-loa a-loa scn)
(if (empty? a-loa)
scn
(draw-alien (first a-loa)
(draw-loa (rest a-loa) scn))))
The test templates are specialized as follows:
;; Tests using sample computations for draw-loa
(check-expect (draw-loa E-LOA E-SCENE) ELOA-VAL)
(check-expect (draw-loa INIT-LOA E-SCENE) ILOA-VAL)

;; Tests using sample values for draw-loa

(check-expect (draw-loa (cons (make-posn 3 4)
(cons (make-posn 15 8) empty))
E-SCENE2)

)
Given that there are no new problems to solve for drawing a list of aliens the design
of draw-aliens is complete.
To draw the shots a list of shots must be processed. This suggests reasoning about
the structure of a los. If the given list of shots is empty, there are no shots to draw
and the result is the given scene. If the given los is not empty, then the rest of the
shots are recursively processed to obtain a scene that contains all the shots except
the first. In this scene the first shot is drawn to obtain a scene that contains all the
shots.
This problem analysis leads to the following sample expressions:
;; Sample expressions for draw-los
(define ELOS-VAL E-SCENE)
(define ALOS-VAL (draw-shot (first A-LOS)
(draw-los (rest A-LOS) E-SCENE)))
354 16 Aliens Attack Version 4

Observe that the draw-shot function, developed in Sect. 60, is used when the given
los is not empty. As with the sample expressions for draw-aliens, abstraction over
the sample expressions for draw-los leads to the definition template specialization:
;; los scene → scene
;; Purpose: To draw the given los in the given scene
(define (draw-los a-los scn)
(if (empty? a-los)
scn
(draw-shot (first a-los)
(draw-los (rest a-los) scn))))
Observe that draw-los is remarkably similar to draw-loa. This is clearly suggesting
that an abstraction ought to be possible to avoid all the repetition. We shall explore
this soon in the next part of this book.
The tests may be written as follows:
;; Tests using sample computations for draw-los
(check-expect (draw-los E-LOS E-SCENE) ELOS-VAL)
(check-expect (draw-los A-LOS E-SCENE) ALOS-VAL)
;; Tests using sample values for draw-los
(check-expect (draw-los (cons (make-posn 14 8)
(cons (make-posn 3 2) empty))
E-SCENE2)

)
The tests complete the design of draw-los. Given that there are no more problems
to solve the refinement of draw-world is completed.

92 The process-key Refinement

The process-key function must be refined, because it builds worlds. In addition,

it must now always add new shots to the list of shots when the player presses the
space bar. When the space bar is pressed, a posn shot is always created and added
to the world’s list of shots. This can be accomplished using cons given that it does
not matter where the shot is added.
92 The process-key Refinement 355

Based on this problem analysis the refined sample expressions are

;; Sample expressions for process-key
(define KEY-RVAL
(make-world
(move-rckt-right (world-rocket INIT-WORLD))
(world-aliens INIT-WORLD)
(world-dir INIT-WORLD)
(world-shots INIT-WORLD)))
(define KEY-LVAL
(make-world
(move-rckt-left (world-rocket INIT-WORLD))
(world-aliens INIT-WORLD)
(world-dir INIT-WORLD)
(world-shots INIT-WORLD)))
(define KEY-SVAL
(make-world
(world-rocket INIT-WORLD)
(world-aliens INIT-WORLD)
(world-dir INIT-WORLD)
(cons (process-shooting (world-rocket INIT-WORLD))
(world-shots INIT-WORLD))))
(define KEY-SVAL2
(make-world
(world-rocket INIT-WORLD2)
(world-aliens INIT-WORLD2)
(world-dir INIT-WORLD2)
(cons (process-shooting (world-rocket INIT-WORLD2))
(world-shots INIT-WORLD2))))
(define KEY-OVAL INIT-WORLD2)
Observe that every world created has a list of aliens and a list of shots. In addition,
processing the space bar key adds the result of process-shooting to the front of
the given world’s list of shots.
Based on the sample expressions, the refined function definition is
;; world key → world
;; Purpose: Process a key event to return next world
(define (process-key a-world a-key)
(cond [(key=? a-key "right")
(make-world (move-rckt-right (world-rocket a-world))
(world-aliens a-world)
(world-dir a-world)
(world-shots a-world))]
356 16 Aliens Attack Version 4

[(key=? a-key "left")

(make-world (move-rckt-left (world-rocket a-world))
(world-aliens a-world)
(world-dir a-world)
(world-shots a-world))]
[(key=? a-key " ")
(make-world (world-rocket a-world)
(world-aliens a-world)
(world-dir a-world)
(cons (process-shooting
(world-rocket a-world))
(world-shots a-world)))]
[else a-world]))
The tests using sample computations do not change because they do not contain
any world references. The tests using sample values, on the other hand, do need to
be refined to have world references updated. These all involve using a loa and los
as follows:
;; Tests using sample values for process-key
(check-expect
(process-key
(make-world (sub1 MAX-CHARS-HORIZONTAL) INIT-LOA
'right E-LOS)
"right")
(make-world (sub1 MAX-CHARS-HORIZONTAL) INIT-LOA
'right E-LOS))

(check-expect (process-key (make-world 0 INIT-LOA 'left E-LOS)

"left")
(make-world 0 INIT-LOA 'left E-LOS))

(check-expect (process-key (make-world 0 INIT-LOA 'left E-LOS)

"o")
(make-world 0 INIT-LOA 'left E-LOS))

(check-expect (process-key INIT-WORLD2 ";") INIT-WORLD2)

(check-expect
(process-key (make-world 0 INIT-LOA
'left E-LOS)
" ")
(make-world 0 INIT-LOA 'left (list (make-posn 0 MAX-IMG-Y))))
92 The process-key Refinement 357

(check-expect
(process-key (make-world 0 INIT-LOA 'left (list SHOT2))
"left")
(make-world 0 INIT-LOA 'left (list SHOT2)))
As before, these tests still explicitly illustrate how a key is processed.
Having refined process-key to integrate the world refinements it is fairly easy
to mistakenly believe the design and implementation are done. It is important to
implement all the changes made to the game. Recall that a shot must be added every
time the space bar is pressed by the player. This means that process-shooting
must always return a posn shot. In Aliens Attack 3, this function makes a decision
to return either 'NO-SHOT or a posn shot. Therefore, it must be refined to always
return an alien.
Given that a posn shot must always be returned by process-shooting, the
image-x value must be the given rocket value and theimage-y value must be
MAX-IMG-Y to put the shot at the bottom at the scene. This problem analysis leads
to these new sample expressions:
;; Sample expressions for process-shooting
(define PS-SHOT-VAL1 (make-posn INIT-ROCKET MAX-IMG-Y))
(define PS-SHOT2-VAL (make-posn INIT-ROCKET2 MAX-IMG-Y))
Abstraction over the sample expression yields the new function definition:
;; shot rocket → shot
;; Purpose: To create a new shot
(define (process-shooting a-rocket)
(make-posn a-rocket MAX-IMG-Y))
The new tests may be written as follows:
;; Tests using sample computations for process-shooting
(check-expect (process-shooting INIT-ROCKET) PS-SHOT-VAL1)
(check-expect (process-shooting INIT-ROCKET2) PS-SHOT2-VAL)

;; Tests using sample values for process-shooting

(check-expect (process-shooting 8) (make-posn 8 MAX-IMG-Y))
(check-expect (process-shooting 16) (make-posn 16 MAX-IMG-Y))
Observe that all the tests clearly illustrate that a posn shot is always returned.
This completes the design and implementation of process-key. The primary
lesson from this example is that refinements do not only involve changes in data
definition. They also include refinements in functionality.
358 16 Aliens Attack Version 4

93 The process-tick Refinement

In Aliens Attack 3, process-tick moves the alien, computes a new direction,

and moves the shot. What does process-tick have to do in Aliens Attack 4?
This function must perform the same tasks for multiple aliens and shots. That is,
process-tick must move a list of aliens, compute a new direction, and move a list
of shots. The function to move a list of aliens is designed in Sect. 76.1. The function
to move a list of shots is designed in Sect. 76.2. Moving aliens and shots, however, is
not enough. As aliens and shots move, they may hit each other and must be removed.
Therefore, functions to remove hit aliens and hit shots are required. The function
to remove hit aliens is designed in Sect. 77.2. The function to remove hit shots is
designed in Sect. 77.3.
What is required to compute a new direction? In Aliens Attack 3, computing new
direction uses the next (moved) alien and the current direction. If the moved alien is
at the left or right edge, then a new direction is returned. With multiple aliens the
direction changes if any of the aliens is either at the left or right edge. Does it make
sense to follow the same strategy as in Aliens Attack 3? That is, does it make sense
to compute the new direction using the (moved) list of aliens for the next world and
the current direction? It seems perfectly reasonable because the new direction is for
the next list of aliens.
Based on this problem analysis, we begin to specialize the template for a function
on a world. We use the defined world instances to write the following refined sample
expressions:
;; Sample expressions for process-tick
(define AFTER-TICK-WORLD1
(make-world
(world-rocket INIT-WORLD)
(remove-hit-aliens
(move-loa (world-aliens INIT-WORLD)
(world-dir INIT-WORLD))
(move-los (world-shots INIT-WORLD)))
(new-dir-after-tick
(remove-hit-aliens
(move-loa (world-aliens INIT-WORLD)
(world-dir INIT-WORLD))
(move-los (world-shots INIT-WORLD)))
(world-dir INIT-WORLD))
(remove-shots (move-los (world-shots INIT-WORLD))
(move-loa (world-aliens INIT-WORLD)
(world-dir INIT-WORLD)))))
93 The process-tick Refinement 359

(define AFTER-TICK-WORLD2
(make-world
(world-rocket INIT-WORLD2)
(remove-hit-aliens
(move-loa (world-aliens INIT-WORLD2)
(world-dir INIT-WORLD2))
(move-los (world-shots INIT-WORLD2)))
(new-dir-after-tick
(remove-hit-aliens
(move-loa (world-aliens INIT-WORLD2)
(world-dir INIT-WORLD2))
(move-los (world-shots INIT-WORLD2)))
(world-dir INIT-WORLD2))
(remove-shots (move-los (world-shots INIT-WORLD2))
(move-loa (world-aliens INIT-WORLD2)
(world-dir INIT-WORLD2)))))
Observe that the only function not yet designed is new-dir-after-tick. For
now, we assume this function exists and we will design it later. The input to
new-dir-after-tick is the given list of aliens for the world being built as in-
put. That is, the moved list of aliens with the hit aliens removed. This is following
the design, as mentioned above, of Aliens Attack 3.
Abstracting over the sample expressions yields the specialized definition template:
;; world → world
;; Purpose: Create a new world after a clock tick
(define (process-tick a-world)
(make-world (world-rocket a-world)
(remove-hit-aliens
(move-loa (world-aliens a-world)
(world-dir a-world))
(move-los (world-shots a-world)))
(new-dir-after-tick
(remove-hit-aliens
(move-loa (world-aliens a-world)
(world-dir a-world))
(move-los (world-shots a-world))))
(world-dir a-world))
(remove-shots (move-los (world-shots a-world))
(move-loa (world-aliens a-world)
(world-dir a-world)))))
360 16 Aliens Attack Version 4

The final step is to specialize the tests. The tests using sample computations do not
change as they do not directly refer to a world. We use the same tests using sample
values as in Aliens Attack 3, but these tests are specialized to use the new world
data definition:
;; Tests using sample values for process-tick
(check-expect
(process-tick (make-world INIT-ROCKET
(cons (make-posn 1 5) '())
'left
E-LOS))
(make-world INIT-ROCKET
(cons (make-posn MIN-IMG-X 5) '())
'down
E-LOS))

(check-expect
(process-tick (make-world
INIT-ROCKET2
(list (make-posn (- MAX-CHARS-HORIZONTAL 2)
10))
'right
(cons SHOT2 '())))
(make-world INIT-ROCKET2
(list (make-posn MAX-IMG-X 10))
'down
(cons (move-shot SHOT2) '())))

(check-expect (process-tick (make-world INIT-ROCKET2

(cons (make-posn MAX-IMG-X 2) '())
'down
(cons (make-posn 15 6) '())))
(make-world INIT-ROCKET2
(cons (make-posn MAX-IMG-X 3) '())
'left
(cons (make-posn 15 5) '())))

(check-expect (process-tick (make-world INIT-ROCKET2

(list (make-posn MIN-IMG-X 2))
'down
(list (make-posn 2 MIN-IMG-Y))))
(make-world INIT-ROCKET2
(list (make-posn MIN-IMG-X 3))
'right
'()))
93 The process-tick Refinement 361

This completes the design and implementation of process-tick. Remember that

all auxiliary functions except new-dir-after-tick are designed in the previous
chapters.

93.1 The new-dir-after-tick Design

We now turn our focus to the design and implementation of new-dir-after-tick.

Recall from the design of new-dir-after-tick in Sect. 51.2 that the new direction
computed depends on the given (current) direction. If the given direction is down,
then the new direction may be right or left. If the given direction is right, then the new
direction may be right or down. If the given direction is left, then the new direction
may be left or down. This design is still valid for Aliens Attack 4 albeit it for a loa,
instead of an alien, now. The implementation proceeds by specializing the template
for a function on a direction.
The sample expressions ought to illustrate how the given direction is processed.
To this end we write two sample expressions for each direction. Each expression
illustrates the computation of one of the two possible next directions. The sample
expressions are as follows:
;; Sample expressions for new-dir-after-tick
(define NEW-DIR-EDGE-LOA-DOWN (new-dir-after-down EDGE-LOA))
(define NEW-DIR-EDGE-LOA2-DOWN (new-dir-after-down EDGE-LOA2))
(define NEW-DIR-INIT-LOA-LEFT (new-dir-after-left INIT-LOA))
(define NEW-DIR-EDGE-LOA-LEFT (new-dir-after-left EDGE-LOA))
(define NEW-DIR-INIT-LOA-RIGHT (new-dir-after-right INIT-LOA))
(define NEW-DIR-EDGE-LOA2-RIGHT
(new-dir-after-right EDGE-LOA2))
Recall from Sect. 74.1 that EDGE-LOA has an alien at the left and EDGE-LOA2 has an
alien at the right edge. INIT-LOA does not have an alien at either edge. The first two
tests illustrate how to compute the new direction when the given direction is down.
The next two tests do the same when the given direction is left. The final two tests
do so when the given direction is right.
Abstracting over the sample expressions yields the definition template special-
ization. The abstraction over each pair of sample expressions for a given direction
yields the expression needed in each stanza of the conditional:
;; loa dir → dir
;; Purpose: Return new aliens direction
(define (new-dir-after-tick a-loa old-dir)
(cond [(eq? old-dir 'down)
(new-dir-after-down a-loa)]
[(eq? old-dir 'left)
(new-dir-after-left a-loa)]
[else (new-dir-after-right a-loa)]))
362 16 Aliens Attack Version 4

The tests are written as follows:

;; Tests using sample computations for new-dir-after-tick
(check-expect (new-dir-after-tick EDGE-LOA 'down)
NEW-DIR-EDGE-LOA-DOWN)
(check-expect (new-dir-after-tick EDGE-LOA2 'down)
NEW-DIR-EDGE-LOA2-DOWN)
(check-expect (new-dir-after-tick INIT-LOA 'left)
NEW-DIR-INIT-LOA-LEFT)
(check-expect (new-dir-after-tick EDGE-LOA 'left)
NEW-DIR-EDGE-LOA-LEFT)
(check-expect (new-dir-after-tick INIT-LOA 'right)
NEW-DIR-INIT-LOA-RIGHT)
(check-expect (new-dir-after-tick EDGE-LOA2 'right)
NEW-DIR-EDGE-LOA2-RIGHT)

;; Tests using sample values for new-dir-after-tick

(check-expect
(new-dir-after-tick (list (make-posn MIN-IMG-X 10))
'down)
'right)
(check-expect
(new-dir-after-tick (list (make-posn MAX-IMG-X 12))
'down)
'left)
(check-expect
(new-dir-after-tick (list (make-posn 10 10))
'left)
'left)
(check-expect
(new-dir-after-tick (list (make-posn MIN-IMG-X 15))
'left)
'down)
(check-expect
(new-dir-after-tick (list (make-posn 10 14))
'right)
'right)
(check-expect
(new-dir-after-tick (list (make-posn MAX-IMG-X 3))
'right)
'down)
The tests using sample values also illustrate the computation of the next direction
using explicit values. Recall that MIN-IMG-X and MAX-IMG-X are the minimum and
maximum image-x values. We now proceed with the design of the needed auxiliary
functions.
93 The process-tick Refinement 363

93.1.1 The new-dir-after-down Design

This function must decide for the given list of aliens if the next direction is right or
left. The needed conditional must determine if the given loa has an alien at the right
edge or at the left edge. The assumption is made that the given loa does not have
aliens at both edges. Without loss of generality, we choose to determine if the given
loa contains an alien at the left edge. To do so we use any-alien-at-left-edge?
designed in Sect. 74.1.
The design of the function is straightforward and its implementation is
;; loa → direction
;; Purpose: Compute the direction of the given alien
;; when previous direction is down
;; ASSUMPTION: The given loa does not have aliens at
;; both edges
(define (new-dir-after-down a-loa)
(if (any-alien-at-left-edge? a-loa) 'right 'left))

;; Sample expressions for new-dir-after-down

(define AT-LEDGE-DOWN 'right)
(define AT-REDGE-DOWN 'left)

;; Tests using sample computations for new-dir-after-down

(check-expect (new-dir-after-down EDGE-LOA2) AT-REDGE-DOWN)
(check-expect (new-dir-after-down EDGE-LOA) AT-LEDGE-DOWN)

;; Tests using sample values for new-dir-after-down

(check-expect (new-dir-after-down
(list (make-posn MIN-IMG-X 4)))
'right)
(check-expect (new-dir-after-down
(list (make-posn MAX-IMG-X 9)))
'left)
The tests using sample values use an explicit loa and illustrate that the function
works correctly for a variety of loas.

93.1.2 The new-dir-after-left Design

This function must decide for the given list of aliens if the next direction is down or
left. The needed conditional must determine if the given loa has an alien at the left
edge. If so, the next direction is down. Otherwise, it is left. To do so we, once again,
use any-alien-at-left-edge? designed in Sect. 74.1.
364 16 Aliens Attack Version 4

The design of the function is straightforward and its implementation is

;; alien → direction
;; Purpose: Compute the direction of the given alien
;; when previous direction is left
(define (new-dir-after-left a-loa)
(if (any-alien-at-left-edge? a-loa) 'down 'left))

;; Sample expressions for new-dir-after-left

(define AT-LEDGE 'down)
(define NOT-AT-LEDGE 'left)

;; Tests using sample computations for new-dir-after-left

(check-expect (new-dir-after-left EDGE-LOA) AT-LEDGE)
(check-expect (new-dir-after-left INIT-LOA) NOT-AT-LEDGE)

;; Tests using sample values for new-dir-after-left

(check-expect (new-dir-after-left (list RIGHT-EDGE-ALIEN))
'left)
Observe that the test using a sample value illustrates that the function works when the
given loa only has an alien on the right edge of the scene. It uses an alien constant
defined in Sect. 51.2.

93.1.3 The new-dir-after-right Design

This function must decide for the given list of aliens if the next direction is down
or right. The needed conditional must determine if the given loa has an alien at the
right edge. If so, the next direction is down. Otherwise, it is right. To do so we use
any-alien-at-right-edge? designed in Sect. 74.2.
The design of the function is straightforward and its implementation is
;; loa → direction
;; Purpose: Compute the direction of the given loa
;; when previous direction is right
(define (new-dir-after-right a-loa)
(if (any-alien-at-right-edge? a-loa)
'down
'right))

;; Sample expressions for new-dir-after-right

(define AT-REDGE 'down)
(define NOT-AT-REDGE 'right)
94 The game-over? Refinement 365

;; Tests using sample computations for new-dir-after-right

(check-expect (new-dir-after-right EDGE-LOA2)
AT-REDGE)
(check-expect (new-dir-after-right INIT-LOA)
NOT-AT-REDGE)

;; Tests using sample values for new-dir-after-right

(check-expect (new-dir-after-right (list LEFT-EDGE-ALIEN))
'right)
Observe that the test using a sample value illustrates that the function works when
the given loa only has an alien on the left edge of the scene. It uses an alien constant
defined in Sect. 51.2.

94 The game-over? Refinement

Ask yourself when should the game be over in Aliens Attack 4. When does the player
lose the game? The player loses when any alien reaches earth. A predicate to detect
this condition, any-alien-reached-earth?, is designed in Sect. 74.3. When does
the player win the game? The player wins when there are no more aliens left.
Based on our problem analysis we can define a world instance and write sample
expressions as follows:
(define WORLD4 (make-world 4 '() 'left '()))

;; Sample expressions for game-over?

(define GAME-OVER1
(or (any-alien-reached-earth?
(world-aliens INIT-WORLD2))
(not (any-aliens-alive?
(world-aliens INIT-WORLD2)))))

(define GAME-OVER2
(or (any-alien-reached-earth?
(world-aliens WORLD4))
(not (any-aliens-alive?
(world-aliens WORLD4)))))

(define GAME-NOT-OVER
(or (any-alien-reached-earth?
(world-aliens INIT-WORLD))
(not (any-aliens-alive?
(world-aliens INIT-WORLD))))))
366 16 Aliens Attack Version 4

The sample world is defined to facilitate writing the sample expressions. The first
sample expression computes the answer for a world having an alien that has reached
earth. The second sample expression computes the answer for a world that has no
aliens. The third sample expression computes the answer using a world for which
the game is not over. Determining whether there are aliens left in the world, just like
any-alien-reached-earth?, is a loa problem. In the interest of separation of
concerns, a loa-processing function is needed.
Abstraction over the sample expressions yields
;; world → Boolean
;; Purpose: Detect if the game is over
(define (game-over? a-world)
(or (any-alien-reached-earth? (world-aliens a-world))
(not (any-aliens-alive? (world-aliens a-world)))))
The tests for the function are
;; Tests using sample computations for game-over?
(check-expect (game-over? INIT-WORLD2) GAME-OVER1)
(check-expect (game-over? WORLD4) GAME-OVER2)
(check-expect (game-over? INIT-WORLD) GAME-NOT-OVER)

;; Tests using sample values for game-over?

(check-expect (game-over?
(make-world
8 (list (make-posn 0 3)) 'right NO-SHOT))
#false)
(check-expect (game-over?
(make-world 8
(list (make-posn 0 MAX-IMG-Y))
'right
(list (make-posn 12 11))))
#true)
(check-expect (game-over?
(make-world 8
(list (make-posn 0 5))
'right
(list (make-posn 0 5))))
#false)
The auxiliary function any-aliens-alive? does not have to process the given
loa. Instead, it needs to test if the given loa is not empty. This may be achieved
using cons?. The function is implemented as follows:
;; loa Boolean → Boolean
;; Purpose: Determine if there is a posn alien in the given loa
(define (any-aliens-alive? a-loa) (cons? a-loa))
94 The game-over? Refinement 367

;; Sample expressions for any-aliens-alive?

(define NOT-ALIVE-VAL (cons? E-LOA))
(define ALIVE-VAL (cons? INIT-LOA))

;; Tests using sample computations any-aliens-alive?

(check-expect (any-aliens-alive? E-LOA) NOT-ALIVE-VAL)
(check-expect (any-aliens-alive? INIT-LOA) ALIVE-VAL)
;; Tests using sample values for any-aliens-alive?
(check-expect (any-aliens-alive? '()) #false)
(check-expect (any-aliens-alive? (list (make-posn 9 2)
(make-posn 6 4)))
#true)

94.1 The draw-last-world Refinement

The final refinement involves the function used to draw the last world. Recall that
this function is used only when game-over? returns #true. How do we determine
if the player has won or lost? We may use any-alien-reached-earth? and
any-aliens-alive? to determine this. Based on this problem analysis we may
write sample worlds and sample expressions as follows:
;; Sample Instance of (final) world
(define FWORLD1 (make-world 13
(list (make-posn 0 MAX-IMG-Y))
'right
E-LOS))
(define FWORLD2 (make-world 7
(list (make-posn 19 MAX-IMG-Y))
'left
(list (make-posn 2 4))))
(define FWORLD3
(make-world 7 '() 'left (list (make-posn 19 3))))

;; Sample expressions for draw-last-world

(define FWORLD1-VAL (place-image
(text "EARTH WAS CONQUERED!" 36 'red)
(/ E-SCENE-W 2)
(/ E-SCENE-H 4)
(draw-world FWORLD1)))
(define FWORLD2-VAL (place-image
(text "EARTH WAS CONQUERED!" 36 'red)
(/ E-SCENE-W 2)
(/ E-SCENE-H 4)
(draw-world FWORLD2)))
368 16 Aliens Attack Version 4

(define FWORLD3-VAL (place-image

(text "EARTH WAS SAVED!" 36 'green)
(/ E-SCENE-W 2)
(/ E-SCENE-H 4)
(draw-world FWORLD3)))
The first and second sample worlds represent a loss for the player given that they
contain an alien with MAX-IMG-Y as their y coordinate. The first world has no shots
and the second has a non-empty los. The third sample world represents a win for the
player given that it contains no aliens. The first two sample expressions illustrate how
to draw a final losing world, while the third illustrates how to draw a final winning
world.
The function’s signature and purpose are the same as for Aliens Attack 3. The
only difference in the function are the predicates used in the conditional expression:
;; world → scene throws error
;; Purpose: To draw the game’s final scene
(define (draw-last-world a-world)
(cond [(any-alien-reached-earth? (world-aliens a-world))
(place-image (text "EARTH WAS CONQUERED!" 36 'red)
(/ E-SCENE-W 2)
(/ E-SCENE-H 4)
(draw-world a-world))]
[(not (any-aliens-alive? (world-aliens a-world)))
(place-image (text "EARTH WAS SAVED!" 36 'green)
(/ E-SCENE-W 2)
(/ E-SCENE-H 4)
(draw-world a-world))]
[else
(error (format
"draw-last-world: Given world has ~s aliens and
none have reached earth."
(length (world-aliens a-world))))]))
Recall that an error is thrown if a non-final world is given as input.
The tests are written as follows:
;; Tests using sample computations for draw-last-world
(check-expect (draw-last-world FWORLD1) FWORLD1-VAL)
(check-expect (draw-last-world FWORLD2) FWORLD2-VAL)
(check-expect (draw-last-world FWORLD3) FWORLD3-VAL)
95 A Bug Despite Hundreds of Tests Passing 369

;; Tests using sample values for draw-last-world

(check-expect (draw-last-world (make-world 10 '() 'left '()))

)
(check-error
(draw-last-world
(make-world 10
(list (make-posn 3 3) (make-posn 7 8))
'right
(list (make-posn 2 3))))
"draw-last-world: Given world has 2 aliens and none have
reached earth.")
The tests using sample values test a final and a non-final world. For the non-final
world, as in Aliens Attack 3, check-error is used to check the error thrown.
To complete the design run the program. Make sure all the tests pass and play the
game several games. Does it all work as expected? Make sure that you play the game
several times before proceeding.

95 A Bug Despite Hundreds of Tests Passing

After playing the game several times you may or may not have noticed that there is a
bug. If you did not notice it, do not be concerned. This is a bug that does not always
manifests itself. A special set of conditions must occur to be able to see the bug. To
illustrate the bug define the following world:
(define BUG-WORLD (make-world
10
(list
(make-posn 4 6)
(make-posn 10 6)
(make-posn 11 6)
(make-posn 12 6)
(make-posn 10 7)
(make-posn 11 7)
(make-posn 13 7)
(make-posn 15 8))
'left
(list (make-posn 0 12))))
370 16 Aliens Attack Version 4

Observe that there is an alien that is farther left than all the other aliens and that the
aliens are moving left. Now, edit the run function to use BUG-WORLD as follows:
; string → world
; Purpose: To run the game
(define (run a-name)
(big-bang BUG-WORLD
[on-draw draw-world]
[name a-name]
[on-key process-key]
[on-tick process-tick TICK-RATE]
[stop-when game-over? draw-last-world]))
Run the game several times and see if you can notice the bug. What do you notice
about how the aliens move?
Let us closely analyze what happens. In particular, we shall focus on the first
alien, the only shot, and the direction. The following table captures how these values
change as the clock ticks:
TICK ALIEN SHOT DIRECTION
0 (make-posn 4 6) (make-posn 0 12) 'left
1 (make-posn 3 6) (make-posn 0 11) 'left
2 (make-posn 2 6) (make-posn 0 10) 'left
3 (make-posn 1 6) (make-posn 0 9) 'left
4 (make-posn 0 6) (make-posn 0 8) 'left
5 (make-posn 0 7) (make-posn 0 7) 'down
The alien starts at image coordinates (4 6) and the shot starts at image coordinates
(0 12). The alien moves left and the shot moves up until tick 4. At tick 4 the alien
is at the left edge and the direction changes to down. At tick 5 the alien has moved
down and the shot has moved up. Observe that the shot hits the alien. Therefore,
process-tick removes both from the world. The new direction is computed by
new-dir-after-tick using the list of aliens that does not contain the hit alien and
'down. This function, in turn, calls new-dir-after-down. Pay close attention to
what new-dir-after-down does. It tests if any alien is at the left edge. This test
returns #false and the function returns 'left. The aliens go back to moving left
when they should move right. This means that we were not careful enough in our
original problem analysis. It is not perfectly reasonable to compute the new direction
using the moved and filtered list of aliens for the next world being built. Take note of
the fact that despite hundreds of tests passing there is still a bug in the program. This
ought to drive home an important lesson. Tests do not guarantee that a program is
correct. They only give us cautious optimism that it is correct. As problem solvers it
is our responsibility to make sure a program is correct. In a future course in Discrete
Mathematics, Program Correctness, or Formal Verification you will learn how to
prove programs correct.
95 A Bug Despite Hundreds of Tests Passing 371

** Ex. 179 — Define a world that manifests the bug when the aliens are moving
right.

For now, we must revisit our problem analysis to fix the bug. The problem in our
example is that the hit alien is removed before computing the new direction. If the
hit alien is not removed, then new-dir-after-down’s test returns #true and the
function returns 'right. That is, the returned direction is correct. Therefore, we
must refine process-tick to be
;; world → world
;; Purpose: Create a new world after a clock tick
(define (process-tick a-world)
(make-world (world-rocket a-world)
(remove-hit-aliens
(move-loa (world-aliens a-world)
(world-dir a-world))
(move-los (world-shots a-world)))
(new-dir-after-tick
(move-loa (world-aliens a-world)
(world-dir a-world))
(world-dir a-world))
(remove-shots
(move-los (world-shots a-world))
(move-loa (world-aliens a-world)
(world-dir a-world)))))
Observe that the new direction is computed without removing hit aliens. Run the
program and make sure all the tests pass. Enjoy the new version of the game!

***** Ex. 180 — PROJECT: Change the world and alien data definitions
to:
;; A world is a structure
;; (make-world rocket loa dir los natnum)
(define-struct world (rocket aliens dir shots ticks2shoot))

;; An alien is a structure: (make-alien posn natnum)

(define-struct alien (pos health))

The player is limited to shoot every ticks2shoot tick. When a player shoots a
world with ticks2shoot equal to some natural number, a constant is created.
After every tick ticks2shoot is decreased by 1. The player may shoot again
when ticks2shoot is 0.
Aliens have a level of health representing how many times they must be hit
before destroyed. Every time an alien is hit its health is decreased by 1. When
an alien’s health is 0, it is removed. Pick an alien color scheme to indicate the
372 16 Aliens Attack Version 4

health of the alien. For example, if an alien starts with a health of 3, then at 3 it
is drawn black, at two it is drawn orange, and at 1 it is drawn red.

96 What Have We Learned in This Chapter?

The important lessons of Chap. 15 are summarized as follows:

• Top-down and bottom-up designs may both be used when developing a large
piece of software.
• Refinements reflect changes in both data definitions and in functionality.
• Designs may vary between refinements. A design idea implemented in one re-
finement may not work in the next.
• Decision-making functions may not need to make a decision in a new refinement.
• All tests passing does not guarantee that there are no bugs in a program.
• Some bugs do not always manifest themselves and as programmers we must be
aware of this.
• Careful problem analysis helps protect us from subtle bugs whose existence is
difficult to realize.
Chapter 17
Binary Trees

We have explored data of arbitrary size whose definitions only contain one selfref-
erence. For example, (listof X) and natnum only have one selfreference:
A list of X is either: A natural number is either:
1. '() 1. 0
2. (cons X (listof X)) 2. (add1 natnum)
There are, however, many data types that contain more than one selfreference.
Consider, for example, representing the results of an arbitrary number of coin flips.
Each time you flip a coin the result is either heads or tail. How can we represent all
the possible outcomes of flipping a coin an arbitrary number of times? Figure 77
displays a graphical representation of all possible outcomes for 3 coin flips. At the
top level, 0 heads and 0 tails have been flipped. To the left at the next level, after a
head is flipped, we have 1 head and 0 tails. To the right at the next level, after a tail
is flipped, we have 0 heads and 1 tail. You may confirm that going left represents
that a head is flipped and going right represents that a tail is flipped as you traverse
down to the next level.
The data representation in Fig. 77 is called a binary tree. A binary tree may be
empty. A non-empty binary tree is made up of 1 or more nodes. A node has data,
like the number of heads and tails flipped, and at most two children. The parent node
is connected to its children by an edge. If an edge goes left, it connects the parent
with its left child. If an edge goes right, it connects the parent with its right
child. The top node of a binary tree is called the root. A node that does not have
any children is called a leaf. In Fig. 77, the root is the node that contains H:0 T:0
and all the nodes in the bottommost level are leaves. Binary trees are commonly used
to efficiently access data contained in nodes and to represent data with a bifurcating
structure—a structure where placement of a node (e.g., left or right) is part of the
information being represented.

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 373
M. T. Morazán, Animated Problem Solving, Texts in Computer Science,
https://doi.org/10.1007/978-3-030-85091-3_17
374 17 Binary Trees

Fig. 77 A tree of coin flips

H:0 T:0

H:1 T:0 H:0 T:1

H:2 T:0 H:1 T:1 H:1 T:1 H:0 T:2

H:3 T:0 H:2 T:1 H:2 T:1 H:1 T:2 H:2 T:1 H:1 T:2 H:1 T:2 H:0 T:3

Fig. 78 A tree of natural numbers

0

1 2

3 4 5 6

7 8 9 10 11 12 13 14

97 Binary Tree Data Definition

Like a list, a binary tree may contain any kind of data. As problem solvers we are free
to define that type of data each node contains. For example, Fig. 77 depicts a binary
tree of coin-tally, where a coin-tally is a structure with two numbers, and
Fig. 78 depicts a binary tree of natnum. You may have already observed that binary
trees are generic. The question now, of course, is how are binary trees defined. Let
us focus closely on Fig. 78. Every node contains a natural number and two edges.
What do the edges represent? Let us closely examine the root. The root is connected
to its left child that contains 1. In addition to 1, the left child also has two children
(containing 3 and 4). The left child is a binary tree. Let us examine the right child of
the root. It contains 2 and two children. It is also a binary tree. How about the node
containing 12? It has 12 and two children (both are empty). We can now define a
binary tree of natural numbers:
;; A binary tree of natural numbers, (btof natnum), is either:
;; 1. '()
;; 2. (list natnum (btof natnum) (btof natnum))
Observe that we arbitrarily chose to represent the empty (btof natnum) as the empty
list and the non-empty btnatnum as a non-empty list. This may seem awkward for
the non-empty (btof natnum) given that there are always three elements: a natnum
and two (btof natnum). A structure would seem to be a better fit. We shall revisit
this issue in Chap. 18. Also observe that the circularity in the recursive definition
is useful. There are two varieties. One is not recursive and the other is recursive.
97 Binary Tree Data Definition 375

The recursive variety, however, has two selfreferences not one like a list or a natural
number. What does this tells us? It tells us that solving a problem involving a (btof
natnum) may require recursive calls to solve the same problem for the left subtree
and the right subtree.
Let us define a binary tree of coin tallies. In Fig. 77 we see the same structure as
in Fig. 78. This tells us that the data definitions will be similar. We define a binary
tree for coin tallies as follows:
;; A cointally is a structure: (make-cointally natnum natnum)
(define-struct cointally (heads tails))

;; A binary tree of coin tallies, (btof cointally), is either:

;; 1. '()
;; 2. (list cointally (btof cointally) (btof cointally))
The cointally data definition is used to capture the two counts that every node
has. Observe that, indeed, the (btof cointally) data definition is very similar to
the (btof natnum) data definition. This informs us that an abstraction is needed
to obtain a generic data definition. Otherwise, we will need to write a new data
definition for every type of binary tree we need to process.
Abstracting over the data definitions above yields the following generic data
definition:
;; A binary tree of X, (btof X), is either:
;; 1. '()
;; 2. (list X (btof X) (btof X))
Recall that X is a type variable. Plugging in natnum for X yields the data definition for
(btof natnum). Plugging in cointally for X yields the data definition for (btof
cointally). We must also develop a template for a function on a (btof X). There
are a couple of things that the definition template needs. The first is a conditional
expression because there is a variety in the data. The second is that the conditional’s
stanza for the non-empty (btof X) needs two recursive expressions: one for each
subtree. The function template is
;; Sample instances of btX
(define E-BTX '())
(define BTX2 . . .)

;; (btof X) . . . → . . .
;; Purpose: . . .
(define (f-on-btx a-btx . . .)
(if (empty? a-btx)
...
. . .(f-on-X (first a-btx))
(f-on-btx (second a-btx))
(f-on-btx (thrid a-btx)). . .)))
376 17 Binary Trees

;; Sample expressions for f-on-btx

(define E-BTX-VAL . . .)
(define BTX2-VAL . . .)
...

;; Tests using sample computations for f-on-btx

(check-expect (f-on-btx E-BTX . . .) E-BTX-VAL)
(check-expect (f-on-btx BTX2 . . .) BTX2-VAL)
...

;; Tests using sample values for f-on-btx

(check-expect (f-on-btx . . .) . . .)
...
The template for a function on a (btof X) ought to look familiar. It contains all the
elements needed to complete the steps of the design recipe. One of the (btof X)
sample instances is directly defined as the empty list given that this is found in the
data definition and never varies. Observe that the definition template has expressions
to recursively process the left and right subtrees of the given tree.

98 Traversing a Binary Tree

The structure of a list must be traversed from its first element to the last element.
Unlike a list, there are different ways a binary tree may be traversed. In this regard,
binary trees are similar to intervals. An interval may be traversed from its low-end
value to its high-end value and vice versa. Variety in the traversal possibilities offers
problem solvers flexibility in problem solving.
If we examine the definition template for a function on a (btof X), we see that
there is a call to a function on a X. This call processes the value at the root. There
are two recursive calls. These traverse the left and right subtrees to process all their
values. The template does not prescribe the order in which these calls are made.
That is, we are free to make these calls in any order depending on the problem
being solved and the design we choose to follow. Three of the basic strategies
are preorder traversal, inorder traversal, and postorder traversal.
A preorder traversal first processes the value at the root, then processes the left
subtree, and ends by processing the right subtree. An inorder traversal first processes
the right subtree, then processes the value at the root, and ends by processing the
left subtree. A postorder traversal first processes the left subtree, then processes the
right subtree, and ends by processing the value at the root.
98 Traversing a Binary Tree 377

Consider extracting all the strings from a (btof string). Given that binary
trees are data of arbitrary size, the returned value must also be data of arbitrary size.
A natural fit is to return a (listof string). Now reason in terms of the structure
of a binary tree. If the tree is empty, it contains no strings and the returned value
is the empty list. Otherwise, we must decide how to traverse the tree. The problem
statement does not specify the order in which the strings must be returned. Therefore,
we arbitrarily choose to return the list of strings following a preorder traversal of
the given binary tree. This means that the function adds the root string to the list
of strings that first contains the strings in the left subtree and then the strings in the
right subtree.
We define sample (btof string) as follows:
;; Sample instances of (btof X)
(define E-BTSTRING '())
(define BTSTRING2 (list
"Pokemon"
(list "Attack on Titan"
(list "Dragon Ball" '() '())
(list "One Piece" '() '()))
(list "Naruto"
(list "Death Note" '() '())
(list "Detective Conan" '() '()))))
The non-empty binary tree of strings represents a set of popular anime. The root of
the tree contains "Pokemon". Its left subtree is rooted at "Attack on Titan" and
its right tree is rooted at "Naruto". The leaves of the binary tree are the children of
the nodes containing "Attack on Titan" and "Naruto". Observe that the leaves
all have empty subtrees.
We formulate sample expressions as follows:
;; Sample expressions for max-btnatnum
(define EBTSTRING-VAL '())
(define BTSTRING2-VAL
(cons (first BTSTRING2)
(append (btstring-extract (second BTSTRING2))
(btstring-extract (third BTSTRING2)))))
As per our problem analysis, processing the empty (btof string) results in an
empty list of strings. If the given (btof string) is not empty, like BTSTRING2, the
right subtree’s list of strings in preorder is recursively computed. The same is done
for the left subtree. These two lists are made into one using append. To the front of
this list the string at the root is added. Observe that this is precisely a list of all the
strings in BTSTRING2 in preorder.
378 17 Binary Trees

The sample expressions provide us with the structure needed to specialize the
definition template as follows:
;; (btof string) → (listof string)
;; Purpose: Extract the strings in the given (btof string)
;; in preorder
(define (btstring-extract a-btstring)
(if (empty? a-btstring)
'()
(cons (first a-btstring)
(append (btstring-extract (second a-btstring))
(btstring-extract (third a-btstring))))))
The final step is to specialize the tests. As with all data types, we have tests using
sample computations and tests using sample values. The tests developed are:
;; Tests using sample computations for btstring-extract
(check-expect (btstring-extract E-BTSTRING) EBTSTRING-VAL)
(check-expect (btstring-extract BTSTRING2) BTSTRING2-VAL)

;; Tests using sample values for btstring-extract

(check-expect (btstring-extract
(list "Attack on Titan"
'()
(list "Sailor Moon"
(list "Fullmetal Alchemist"
'()
'())
'())))
(list "Attack on Titan"
"Sailor Moon"
"Fullmetal Alchemist"))

(check-expect (btstring-extract
(list "This"
(list "BT"
(list "is"
(list "like"
(list
"a"
(list "list." '() '())
'())
'())
'())
'())
'()))
(list "This" "BT" "is" "like" "a" "list."))
99 The Maximum of a (btof int) 379

The tests using sample values illustrate that the function works for binary trees that
are not full. That is, for binary trees containing leaf nodes at different levels or
containing nodes that have one subtree. Observe that the second of the tests using
sample values illustrates that a binary tree may have a structure similar to a list. This
occurs when either all left subtrees are empty or all right subtrees are empty.

99 The Maximum of a (btof int)

Consider the problem of finding the maximum in a (btof int). For example, the
maximum of the (btof int) in Fig. 78 is 14. How did you determine this? You
probably scanned all the numbers in the tree. That is, you traversed the tree. Think
about what is needed to determine the maximum value and how the (btof int)
may be traversed to find the maximum. What does the root value need to be compared
to determine the maximum? It is not difficult to see that the root value needs to be
compared to the maximum of the left subtree and the maximum of the right subtree.
That means that a postorder traversal is a natural fit to this problem.
Now reason in terms of the structure of a (btof int). What is the answer if
the (btof int) is empty? If the given (btof int) is empty, then it contains no
maximum number. This informs us that we either need to define a subtype for a
non-empty (btof int) or the function must throw an error if it is given a empty
(btof int) as input. On this occasion, we design using the second option and have
the function throw an error. If the given (btof int) is not empty, then the function
needs to determine the maximum among the left subtree maximum, the right subtree
maximum, and the root value. Care must be taken because any of the subtrees or
both may be empty. A recursive call must not be made with an empty subtree to
avoid having the function throw and error. This means that there are 5 conditions that
must be tested for. For each, the value returned depends on the non-empty subtrees:
(btof int) Return
Empty Throw error
Leaf Root value
Left subtree empty Max of the root value and the right subtree max
Right subtree empty Max of the root value and the left subtree max
Neither subtree empty Max of root value, left subtree max, and right subtree max
380 17 Binary Trees

Based on this problem analysis, we define 5 sample (btof int) instances as

follows:
;; Sample instances of (btof int)
(define E-BTINT '()) ;; empty tree
(define BTINT1 (list 200 '() '())) ;; leaf
(define BTINT2 (list 100 ;; neither subtree is empty
(list 1
(list 3 '() '())
(list 4 '() '()))
(list 2
(list 5 '() '())
(list 6 '() '()))))
(define BTINT3 (list -10 ;; left subtree is empty
'()
(list 78 '() '())))
(define BTINT4 (list -8 ;; right subtree is empty
(list -5
(list 47 '() '())
(list -1 '() '()))
'()))
There is a sample instance for each condition above as indicated by the comments.
In this manner, we can easily write a sample expression for each condition.
Assuming the function name is max-btint, we may write sample expressions as
follows:
;; Sample expressions for max-btint
(define BTINT1-VAL (first BTINT1))
(define BTINT2-VAL (max (first BTINT2)
(max-btint (second BTINT2))
(max-btint (third BTINT2))))
(define BTINT3-VAL (max (first BTINT3)
(max-btint (third BTINT3))))
(define BTINT4-VAL (max (first BTINT4)
(max-btint (second BTINT4))))
First, there is no sample expression for the empty (btof int) because a value is not
computed. The error message generated is tested as part of the tests using a sample
value. Second, observe that for each non-empty (btof int) there are no calls to
max-btint with an empty tree.
99 The Maximum of a (btof int) 381

We can now specialize the definition template as follows:

;; (btof int) → int throws error
;; Purpose: Find the maximum in the given (btof int)
(define (max-btint a-btint)
(cond [(empty? a-ntint)
(error "An empty (btof int) has no maximum value.")]
[(and (empty? (second a-btint))
(empty? (third a-btint)))
(first a-btint)]
[(empty? (second a-btint))
(max (first a-btint) (max-btint (third a-btint)))]
[(empty? (third a-btint))
(max (first a-btint) (max-btint (second a-btint)))]
[else (max (first a-btint)
(max-btint (second a-btint))
(max-btint (third a-btint)))]))
Observe that the conditional has 5 stanzas. That is, one stanza for each condition
identified. As with all conditionals, the first two stanzas are for nonrecursive cases.
The last three are recursive cases. Observe that only the fifth stanza (the else stanza)
has two recursive calls, two traverse both subtrees. These recursive calls appear in
an argument position to max. This means that the subtrees are traversed first before
examining the root value. The same is true for the non-empty subtree in stanzas 3 and
4. That is, this function implements a postorder traversal as expected by the problem
analysis.
The tests are the following:
;; Tests using sample computations for max-btint
(check-expect (max-btint BTINT1) BTINT1-VAL)
(check-expect (max-btint BTINT2) BTINT2-VAL)
(check-expect (max-btint BTINT3) BTINT3-VAL)
(check-expect (max-btint BTINT4) BTINT4-VAL)

;; Tests using sample values for max-btint

(check-error
(max-btint E-BTINT)
"An empty (btof natnum) has no maximum value.")
(check-expect (max-btint (list -67
(list -50 '() '())
(list -8 '() '())))
-8)
Observe that the error message when an empty binary tree is provided as input is
tested. The second test using a sample value illustrates that the function works when
all the integers in the binary tree are negative. Run the program and make sure all
the tests pass.
382 17 Binary Trees

*** Ex. 181 — Provide a data definition for a non-empty (btof int). Design
and implement max-btint using your new data definition.

* Ex. 182 — Design and implement a function to compute the minimum of a

(btof int).
** Ex. 183 — Design and implement a function to return the longest string
(btof string).

*** Ex. 184 — Design and implement a function to return the subtree in a
(btof int) that has the tree’s maximum as its root. If the root value is the
maximum value in the tree, then the entire tree is the subtree that has the tree’s
maximum as its root.

100 Binary Search Trees

A database stores organized information in a computer program. Consider represent-

ing a database of criminals for a law enforcement agency. We may define a criminal
record as follows:
;; A criminal record (cr) is a structure
;; (make-cr natnum string string)
;; with an id number, a nickname, and a name.
(define-struct cr (id nickname name))
The following are sample cr instances that may be used to populate a database:
(define TEFLONDON (make-cr 241
"The Teflon Don"
"John Gotti Jr."))

(define BABYFACE (make-cr 77

"Babyface Nelson"
"Lester Joseph Gillis"))

(define SCARFACE (make-cr 23

"Scarface"
"Alphonse Gabriel Capone"))

(define DILLINGER (make-cr 675

"Gentleman John"
"John Herbert Dillinger"))

(define BUGSY (make-cr 874 "Bugsy" "Benjamin Siegel"))

100 Binary Search Trees 383

(define PAULIE (make-cr 1

"Big Pauli"
"Paul Castellano"))

(define VITO (make-cr 55 "Don Vitone" "Vito Genovese"))

The next question we must think about is how will the database be represented.
Given that the number of criminals is not known in advance, a database is of arbitrary
size. This means we need a recursive data structure to represent it. We shall explore
three possible representations: a list of cr, a binary tree of cr, and a binary search
tree of cr.

100.1 A (listof cr) Representation

A natural way to represent the criminal database is to use a list. We may define a
criminal database and sample instances as follows:
;; A criminal database is a (listof cr).

;; Sample instances
(define CR-DB0 '())
(define CR-DB1 (list TEFLONDON BABYFACE SCARFACE
DILLINGER BUGSY))
Is this a good and efficient representation? Answering such a question is at the heart
of many problems. That is, it is not only important to solve a problem. It is equally
important to have a fast and efficient solution. Here is where representation plays
a significant role. A representation influences how a problem solver thinks about a
problem and how efficient a solution is.
Consider the problem of returning a cr in a database given an id number. To
solve this problem, as before, we think in terms of the structure of the database.
What is the answer if the database is empty? In this case, the wanted cr does not
exist. Therefore, we define the following cr to signal that the a cr with the given id
number does not exist:
(define DNE-CR (make-cr 0 "DNE" "CR DOES NOT EXIST"))
When the database is empty, the answer is DNE-CR. What is the answer if the database
is not empty? This means that there is at least a cr in the database. The id of the
first cr is compared with the given id. If they match, then the first cr is returned.
Otherwise, the rest of the database is recursively searched. The function to retrieve
a cr from a database represented using a (listof cr) is displayed in Fig. 79. The
function is designed using the template for a function on a (listof cr).
384 17 Binary Trees

Fig. 79 The function to retrieve a cr from a (listof cr)

;; number (listof cr) cr
;; Purpose: To return list of the even numbers in the given list
(define (get-record-locr id a-locr)
(cond [(empty? a-locr) DNE-CR]
[(= (cr-id (first a-locr)) id) (first a-locr)]
[else (get-record-locr id (rest a-locr))]))

;; Sample expressions for get-record-locr

(define GR1-VAL DNE-CR)
(define GR2-VAL TEFLONDON)
(define GR3-VAL (get-record-locr 675 (rest CR-DB1)))

;; Tests using sample computations for get-record-locr

(check-expect (get-record-locr 899 CR-DB1) GR1-VAL)
(check-expect (get-record-locr 241 CR-DB1) GR2-VAL)
(check-expect (get-record-locr 675 CR-DB1) GR3-VAL)

;; Tests using sample values for get-record-locr

(check-expect (get-record-locr 55 (list PAULIE VITO)) VITO)
(check-expect (get-record-locr 1 (list PAULIE VITO)) PAULIE)

100.2 A (btof cr) Representation

Another representation option for the criminal database is to use a binary tree. The
idea here is to have a cr at the root and distribute the remaining crs among the
two subtrees. For example, we may define the same database from the previous
subsection as
(define CR-DB3 (list
DILLINGER
(list BABYFACE '() (list SCARFACE '() '()))
(list TEFLONDON (list BUGSY '() '()) '())))
DILLINGER is at the root of the tree and the rest of the criminal records are evenly
distributed among the subtrees.
Once again, consider how to retrieve a criminal record given an id number. Reason
in terms of the structure of a binary tree. If the given binary tree is empty, then the
answer is DNE-CR. What is the answer if the binary tree is not empty? In this case,
the tree must be traversed. It is reasonable to first check the root’s cr. Therefore, a
preorder traversal is appropriate. There are several cases that must be distinguished.
We start by checking if the root’s cr’s id matches the given id. If they match, then
the answer is the root cr and there is no need to traverse the subtrees. If they do not
match, then we traverse the left subtree first. The cr being searched for may or may
not be in the left subtree. How do we know if it is or it is not? Thinking about this
carefully reveals that we may determine the answer by examining the result obtained
by traversing the left subtree. If the result is DNE-CR, then the right subtree must be
100 Binary Search Trees 385

Fig. 80 The function to retrieve a cr from a (btof cr)

;; number (btof cr) cr
;; Purpose: To return list of the even numbers in the given list
(define (get-record-btocr id a-btocr)
(cond [(empty? a-btocr) DNE-CR]
[(= (cr-id (first a-btocr)) id) (first a-btocr)]
[(equal? (get-record-btocr id (second a-btocr)) DNE-CR)
(get-record-btocr id (third a-btocr))]
[else (get-record-btocr id (second a-btocr))]))

;; Sample expressions for get-record-btocr

(define GRBT1-VAL DNE-CR)
(define GRBT2-VAL DILLINGER)
(define GRBT3-VAL (get-record-btocr 874 (third CR-DB3)))
(define GRBT4-VAL (get-record-btocr 77 (second CR-DB3)))

;; Tests using sample computations for get-record-btocr

(check-expect (get-record-btocr 899 CR-DB0) GRBT1-VAL)
(check-expect (get-record-btocr 675 CR-DB3) GRBT2-VAL)
(check-expect (get-record-btocr 874 CR-DB3) GRBT3-VAL)
(check-expect (get-record-btocr 77 CR-DB3) GRBT4-VAL)

;; Tests using sample values for get-record-locr

(check-expect (get-record-btocr 55 (list PAULIE () (list VITO () ())))
VITO)
(check-expect (get-record-btocr 1 (list PAULIE () (list VITO () ())))
PAULIE)
(check-expect (get-record-btocr 16 (list BABYFACE
(list SCARFACE () ())
(list BUGSY () ())))
DNE-CR)

traversed for the answer. Otherwise, the answer is the result obtained from traversing
the left subtree.
The specialized template is displayed in Fig. 80. Observe that problem analysis
reveals 4 conditions that must be distinguished. There is a sample expression for
each. The first two sample expressions are for the nonrecursive cases, respectively,
the empty binary tree and CR-DB3’s root id matching the given id. The third sample
expression illustrates how to traverse CR-DB3’s right subtree when the left subtree
traversal returns DNE-CR. The fourth sample expression illustrates how to compute
the answer if CR-DB3’s left subtree contains the searched for cr.
Based on the problem analysis and the sample expressions the definition template
is specialized to have a conditional expression with 4 stanzas. The nonrecursive cases
are straightforward to understand. The recursive cases, on the other hand, need to
be scrutinized. As per our problem analysis, the left subtree is traversed first. It is
traversed, however, to determine if the result is DNE-CR and not to actually return a
result. As per our problem analysis, the right subtree is traversed if the traversal of the
left subtree is DNE-CR. If the result of the left subtree is not DNE-CR, then this result
is recomputed to return the answer. In other words, the function may traverse the
386 17 Binary Trees

left subtree twice. This does not seem very efficient. Why do the same work twice?
For small binary trees this may not matter. However, the repetition of a computation
may be costly if the given binary tree contains millions of criminal records.
Finally, observe that the first test using a sample computation searches the empty
database. The tests using sample values show that successful and unsuccessful
searches return the correct value.

100.3 A (bstof cr) Representation

We have implemented two different ways to search the criminal-record database

using two different representations. The first was relatively straightforward. The
second is inefficient because it repeats work done. The question now is can searching
the database be made faster. A possible optimization is to represent the database
using a (listof cr) sorted by id number. In this manner, a search may sometimes
be able to determine that a cr does not exist before traversing the entire list. For
example, consider CR-DB1 sorted by id number. Searching for a record with an id
of 10 ends with the first record (given that SCARFACE’s id is 23) and searching for
an id record with an id of 300 ends after four comparisons (given that DILLINGER’s
id is 675). In both cases, the entire (listof cr) does not have to be searched to
determine that the record with the given id number does not exist.
Can we improve searching when using a binary tree representation? This is an
interesting and nontrivial question. There is potential for significant improvements if
we can guarantee certain binary tree properties. For example, we can avoid traversing
the whole tree searching for a cr if, for example, all the ids less than the root are in
the left subtree and all the ids greater than the root are in the right subtree. In this
manner if the given id does not match the root, then only the left or only the right
subtree needs to be traversed.
A (btof X) that satisfies these invariant properties is called a binary search
tree. We may define a binary search tree of X as follows:
;; A (bstof X) is either
;; 1. '()
;; 2. (list X (bstof X) (btsof X))
;; SUCH THAT
;; A. All Xs in the left subtree are less than the root X
;; B. All Xs in the right subtree are greater than the root X

Observe that the data definition lists the invariant properties. In this manner, it is
clear to all readers and problem solvers what a (bstof X) is. That is, (bstof X)
is a subtype of (btof X). Take note of the fact that the invariants define content
properties of a binary tree and do not affect the structure of a binary tree. Therefore,
the template of a function on a (btof X) is specialized when a binary search tree
problem is solved.
100 Binary Search Trees 387

The sample database used in the previous subsections may now be represented as
a binary search tree:
(define CR-DB4 (list TEFLONDON
(list BABYFACE
(list SCARFACE '() '())
'())
(list DILLINGER
'()
(list BUGSY '() '()))))
Observe that all the crs with ids less than the root’s id are in the left subtree and all
the crs with an id larger than the root’s id are in the right subtree. The same is true
for the subtrees rooted at BABYFACE, SCARFACE, DILLINGER, and BUGSY. In other
words, the invariant properties always hold and, therefore, CR-DB4 is a (bstof cr).
Given an id and a (bstof X), how do we search for the needed cr? We reason
about the structure of a (bstof X). That is, we reason about the structure of a
(btof X). If the (bstof X) is empty, once again, the answer is DNE-CR. If the
given id matches the id at the root, then the answer is the root cr. If the given id is
less than the root’s id, then the left subtree is recursively searched. Otherwise, the
right subtree is recursively searched.
Following the steps of the design recipe, we define a constant for the value of an
expression for each of the four conditions:
;; Sample expressions for get-record-bstocr
(define GRBST1-VAL DNE-CR)
(define GRBST2-VAL (first CR-DB4))
(define GRBST3-VAL (get-record-bstocr 23 (second CR-DB4)))
(define GRBST4-VAL (get-record-bstocr 675 (third CR-DB4)))
The sample expressions correspond, respectively, to searching a an empty (bstof
X), matching the root id number, needing to search the left subtree for an id number
less than the root id number, and needing to search the right subtree for an id number
greater than the root id number.
Based on the sample expressions we can specialize the definition template to be
;; number (btof cr) → cr
;; Purpose: To return the cr with the given id if it
;; exists in the given (bstof cr)
(define (get-record-bstocr id a-bstocr)
(cond [(empty? a-bstocr) DNE-CR]
[(= (cr-id (first a-bstocr)) id) (first a-bstocr)]
[(< id (cr-id (first a-bstocr)))
(get-record-bstocr id (second a-bstocr))]
[else (get-record-bstocr id (third a-bstocr))]))
388 17 Binary Trees

The stanzas for the nonrecursive cases appear first in the conditional. Observe that
both subtrees are never traversed. Based on the relation between the given id number
and the root’s id number only the left or the right subtree is traversed. Furthermore,
observe that neither subtree is ever traversed twice. Clearly, searching a (bstof X)
is more efficient than searching a (btof X) when the needed cr is in the left subtree.
The tests are written following the same motivating guidelines as the tests using
a (btof cr). The tests are
;; Tests using sample computations for get-record-bstocr
(check-expect (get-record-bstocr 899 CR-DB0) GRBST1-VAL)
(check-expect (get-record-bstocr 241 CR-DB4) GRBST2-VAL)
(check-expect (get-record-bstocr 23 CR-DB4) GRBST3-VAL)
(check-expect (get-record-bstocr 675 CR-DB4) GRBST4-VAL)

;; Tests using sample values for get-record-bstocr

(check-expect (get-record-bstocr 55
(list PAULIE
'()
(list VITO '() '())))
VITO)
(check-expect (get-record-bstocr 1
(list PAULIE
'()
(list VITO '() '())))
PAULIE)
(check-expect (get-record-bstocr 14
(list PAULIE
'()
(list VITO '() '())))
DNE-CR)
As with the tests using a (btof cr), the first test using a sample computation
searches the empty database. The tests using sample values show that successful and
unsuccessful searches return the correct value.

101 Abstract Running Time

We have three different ways to search a criminal database based on three different
representations. Which one should we choose? The answer depends on the criteria
that must be met. If the most important criteria is code simplicity, then it is likely
that the representation using a (listof cr) is the best. Most of the time, however,
it is more important for a program to be efficient in terms of execution time, memory
usage, or both.
101 Abstract Running Time 389

Fig. 81 The function to build a list of random natural numbers less than 10,000,000
;; natnum (listof natnum<10000000)
;; Purpose: Create a list on n random natnums < 10000000
(define (build-random-list n)
(if (= n 0)
()
(cons (random 10000000) (build-random-list (sub1 n)))))

;; Sample expressions for build-random-list

(define L1 ())
(define L2 (cons (random 10000000) (build-random-list (sub1 1000))))

;; Tests using sample computations for build-random-list

(check-expect (build-random-list 0) ())
(check-expect (length (build-random-list 1000)) 1000)
(check-expect (all<10000000 (build-random-list 1000)) #t)

;; Tests using sample values for build-random-list

(check-expect (build-random-list 0) ())
(check-expect (length (build-random-list 1000)) 1000)
(check-expect (all<10000000 (build-random-list 1000)) #t)

Let us focus on execution time. How can we describe the efficiency of a program?
In other words, how can we compare two programs to decide which one is more
efficient in terms of execution time? Clearly, we all dislike slow programs and want
all our programs to be fast. One way to measure efficiency is by timing the programs,
given the same input, that are being compared. In ISL+ we can time programs using
the time function. The time function takes as input an expression to evaluate and
returns its value. Before returning the value of the expression it prints information
about execution time in milliseconds including CPU time. For example consider
timing the sorting of a (listof natnum) using insertion sorting (from Sect. 78)
and ISL+’s built-in sorting function23 as follows:
(define LST1 (build-random-list 5000))

(define SORTED1 (time (sort-lon LST1)))

(define SORTED2 (time (sort LST1 <)))
The same list is sorted by two different functions. The function build-random-list
is displayed in Fig. 81. The expression (random 10000000) evaluates to a random
number in [0..10000000). Property-based testing is used given that we do not have
a way of knowing which random number is generated. The properties tested are
the length of the list and proper membership in the list. The auxiliary predicate,
all<10000000, used to write property-based tests is displayed in Fig. 82.

23
The second argument to sort tells the function to use < to compare numbers.
390 17 Binary Trees

Fig. 82 The predicate to determine if (listof natnum) members are all less than
10,000,000
;; (listof natnum) Boolean
;; Purpose: Determine if all natnums in the given list are in
;; [0..10000000)
(define (all<10000000 L)
(or (empty? L)
(and (<= 0 (first L) 9999999)
(all<10000000 (rest L)))))

;; Sample expressions for all<10000000

(define B1 (or (empty? ())
(and (<= 0 (first ()) 9999999)
(all<10000000 (rest ())))))
(define B2 (or (empty? (5000 34 893499))
(and (<= 0 (first (5000 34 893499)) 9999999)
(all<10000000 (rest (5000 34 893499))))))

;; Tests using sample computations for all<10000000

(check-expect (all<10000000 ()) B1)
(check-expect (all<10000000 (5000 34 893499)) B2)

;; Tests using sample values for all<10000000

(check-expect (all<10000000 (500 -9 38928)) #f)

Running the program 5 times yields the following CPU-time measurements:

Run Insort Sort

1 26406 15
2 25922 0
3 25719 16
4 25109 16
5 26344 47

At first glance, the collected data suggests that ISL+’s sort function is faster than our
sort-lon function. The most salient feature from the above table, however, is that
execution timing is an unreliable measure. For neither sorting function do we always
get the same result given the same input. Furthermore, the variation between different
runs is noticeable. In fact, if you run the same experiments on your computer, you
are likely to get different results also. The lesson here is that execution times may
vary on a computer and may vary among different computers. There must be a better
way to compare the expected execution time of programs.
In fact there is a better way to compare the expected execution time of programs.
Instead of counting milliseconds, we can count the number of operations performed
in relation to the size of the input. For example, we may count the number of
(recursive) function calls or the number of comparisons made. For a given input,
these do not vary among different runs nor among different computers. This means
that we expect a program that performs fewer of the counted operations to be faster
102 The Complexity of Searching the Criminal Database 391

regardless of the computer used. This is what is called abstract running time
or the complexity of a program. It is important to note that we do not count every
single operation done by the computer. Instead, we count abstract operations, like
recursive calls or comparisons, that may involve many computer operations. How
abstract operations are implemented may vary from one computer to another. Some
computer systems may use more or may use less computer operations to implement
an abstract operation. The number of abstract operations themselves does not vary
and that is what makes abstract running time a good basis for comparison when
evaluating the efficiency of a program. If a program performs n abstract operations
and on a given machine each operation performs k computer operations, then the
number of computer operations is k * n. That is, the number of computer operations
is proportional to some constant k. This constant of proportionality is different for
different computers and is ignored when determining abstract running time.
Typically, we are concerned with the worst-case scenario to establish the abstract
running time of a program. The number of abstract operations is described in terms
of the size of the input by a mathematical function that we must derive from the
code. If the size of the input is n, then we say that the complexity is the highest
power of n in this function. Big O notation is used to describe the complexity
of a program in terms of the input size. If a program performs 5n2 +2n+1 abstract
operations, we say the complexity of the program is O(n2 ), where n is the size
of the input. Observe that the constant of proportionality, the lower powers of n,
and constant factors are ignored. For example, consider the build-random-list
function in Fig. 81. If the input is 5, then there are five recursive calls performed for:
4, 3, 2, 1, 0. When n is 0, the recursion stops. Similarly, if the input is 3, then there
are three recursive calls performed for: 2, 1, 0. In general, the function is recursively
called n times. Thus, we say that the complexity of the function is O(n).

102 The Complexity of Searching the Criminal Database

To determine if any of the criminal-record database implementations is superior

we compare their abstract running time. To start, consider the database represented
as a (listof cr) and the searching function from Sect. 100.1. In the best case
get-record-locr is called only once. This occurs when the function is called with
the empty list or when the record searched for is the first one in the list. Therefore,
in the best case the function performs O(1), or simply O(k), calls. What is the
worst-case scenario? Searching for an id number not in the database means that the
function is called once for every criminal record in the database to determine that
a criminal record with the given id number does not exist. If the given list has n
records, then the complexity is O(n). This is a linear function. If the size of the
input is doubled, then the number of operations performed is also doubled.
Is searching the database any better if it is represented as a sorted (listof cr)?
Clearly, in the best case the function performs O(k) calls for the same reasons above.
What is the worst-case scenario? In the worst case the given id number is larger than
392 17 Binary Trees

any id number in the database. This means that the entire database must be searched
to determine that a criminal record with the given id number does not exist, thus
making the complexity O(n). This tells us that the best and worst cases for both
representations (unsorted and sorted (listof cr)) are expected to have the same
performance. Neither is superior to the other.
Let us now consider the search complexity for the representation using a (btof
cr) from Sect. 100.2. In the best case get-record-btocr is called only once. That
is, in the best case the complexity is O(k). The worst-case scenario is searching for
an id number that is not in the tree. The left subtree is traversed to determine that
the left subtree does not contain the given id number and, subsequently, the right
subtree is traversed. This means that the number of calls to get-record-btocr is
equal to the number of nodes in the binary tree. Therefore, the search complexity is
O(n). This is the same complexity as using a (listof cr) and, therefore, using a
binary tree provides no improvement.
We now turn our attention to representing the database using a (bstof cr).
In the best case, as with the other representations, get-record-bstocr is only
called once and the search complexity is O(k). Analyzing the worst-case scenario
is more subtle. It occurs when a search is performed for an id number that is larger
(or smaller) than any id number in the binary search tree. A careless glance at the
code may suggest that at each step half of the criminal records are eliminated from
the search because only the left or the right subtree is traversed. Is it guaranteed that
half of the criminal records are eliminated from the search every time the function
is called? Consider, for example, the following database represented using a binary
search tree:
(define BST-LST
(list
PAULIE
'()
(list SCARFACE
'()
(list VITO
'()
(list BABYFACE
'()
(list TEFLONDON
'()
(list DILLINGER
'()
(list BUGSY '() '()))))))))
Observe that every left subtree is empty. This means that the structure of this binary
search tree is similar to the structure of a list. In an arbitrary binary search tree with
n nodes and the same type of structure searching for a cr with an id number greater
than anything in the database requires n calls to get-record-bstocr. Therefore,
in the worst case the complexity of this function is O(n). This means that all three
103 Balanced (bstof cr) 393

of our database representations have a searching function with the same complexity.
None are expected to always be better than any of the others.

103 Balanced (bstof cr)

Our efforts to date, regardless of the database representation, have yielded a linear
time searching algorithm. When using binary search trees, the problem is that in
the worst case only the record at the root is eliminated from the search, thus forcing
the examination of all records in the database. To overcome this problem we can
represent the database using a balanced binary search tree. A balanced binary tree
is one in which the height of the left and right subtrees of any node differs by no
more than 1. In other words, the binary tree is not lopsided and the descendants of
a node are roughly evenly distributed in its subtrees. We define a balanced binary
search tree as follows:
;; A (bbstof X) is either
;; 1. '()
;; 2. (list X (bbstof X) (bbtsof X))
;; SUCH THAT
;; A. All Xs in the left subtree are less than the root X
;; B. All Xs in the right subtree are greater than the root X
;; C. (<= (- (height left-subtree) (height right-subtree)) 1)
The third invariant guarantees that the nodes are evenly distributed and, therefore, at
each step of the search roughly half of the nodes left to explore are eliminated.

103.1 Creating a Balanced Binary Search Tree

However, can a balanced binary search tree be created in the first place? Without loss
of generality, we assume that we have a (listof cr) sorted in non-decreasing order
by id number. The middle element of the list must be the root of the balanced binary
search tree. All the crs before the middle element must be placed in the left (balanced)
subtree and all the crs after the middle element must be placed in the right (balanced)
subtree. Accessing the middle element of the given list is fairly straightforward. If
the input list, a-locr, is not empty, then we know that its first element is indexed
by 0 and its last element is indexed by (sub1 (length a-locr)). Therefore,
the middle element is indexed by (quotient (length a-locr) 2). How then
are the elements before and after the middle element referenced? Observe that the
valid indexes into a-locr may be represented by an interval: [0..(sub1 (length
a-locr))]. This means that the elements before the middle element are indexed
by the interval elements in [0..(sub1 (quotient (length a-locr) 2))] and
that the elements after the middle element are indexed by the interval elements
394 17 Binary Trees

in [(add1 (quotient (length a-locr) 2))..(sub1 (length a-locr))].

This analysis suggests that we can, indeed, create a balanced binary search tree using
a function that processes an interval. If the given (listof cr) is empty, then the
interval to process must be empty given that there are no crs in the list. Otherwise,
the interval to process is defined by the minimum and maximum indexes into the
list. We may write sample expressions as follows:
(define SORTED-LOCR (list PAULIE SCARFACE VITO BABYFACE
TEFLONDON DILLINGER BUGSY))

;; Sample expressions for locr->bstocr

(define LOCR->BBST1 (create-bstofcr CR-DB0 0 -1))
(define LOCR->BBST2 (create-bstofcr SORTED-LOCR 0 6))
A sorted (listof cr) is defined to facilitate writing sample expressions and tests.
The first sample expression is for the empty (listof cr) and the second is for the
non-empty SORTED-LOCR.
Based on the sample expressions we can write the function as follows:
;; (listof cr) → (bbstof cr)
;; Purpose: Create (bbst cr) from the given sorted (listof cr)
;; ASSUMPTION: The given (listof cr) is sorted in nondecreasing
;; order by id number
(define (locr->bbstocr a-locr)
(create-bbstofcr a-locr 0 (sub1 (length a-locr))))
Observe that locr->bbstocr only calls a function to process an interval. When the
given list is empty, the interval is [0..-1], and when it is not empty, it is [0..(sub1
(length a-locr))], which are the intervals suggested by our problem analysis.
Finally, the function is tested as follows:
;; Tests using sample computations for locr->bstocr
(check-expect (locr->bbstocr CR-DB0) LOCR->BBST1)
(check-expect (locr->bbstocr SORTED-LOCR) LOCR->BBST2)

;; Tests using sample values for locr->bstocr

(check-expect (locr->bbstocr (list PAULIE VITO BUGSY))
(list VITO
(list PAULIE '() '())
(list BUGSY '() '())))
Recall that CR-DB0 is defined as '(). The single test using a sample value illustrates
the structure of the generated balanced binary tree.
103 Balanced (bstof cr) 395

We now focus on the design of create-bbstofcr. This function has as input

a (listof cr) and an interval of indexes, [low..high], into the list. We reason
about the structure of the interval. If the interval is empty, then the needed tree is
'(). If the tree is not empty, then the root of the needed tree is the middle list element
in the interval. That element is indexed by low plus half the length of the interval:
(+ low (quotient (- high low) 2))
The elements for the left and right subtrees, respectively, are indexed by the elements
in the following intervals:
[low..(sub1 (+ low (quotient (- high low) 2)))]

[(add1 (+ low (quotient (- high low) 2)))..high]

Observe that the high-end for the left subtree elements is exactly one less than the
middle element index and that low-end for the right subtree elements is exactly one
more than the middle element index. To build the desired balanced binary search
tree all that is needed is to create a list with the middle element, the tree obtained
by processing the first subinterval, and the tree obtained by processing the second
subinterval.
Based on our problem analysis above we write the following sample expressions:
;; Sample expressions for create-bbstofcr
(define CBST1 '())
(define CBST2
(list (list-ref SORTED-LOCR (+ 0 (quotient (- 6 0) 2)))
(create-bbstofcr
SORTED-LOCR
0
(sub1 (+ 0 (quotient (- 6 0) 2))))
(create-bbstofcr
SORTED-LOCR
(add1 (+ 0 (quotient (- 6 0) 2)))
6)))
As per the problem analysis, the empty binary search tree is returned when the given
(listof cr) is empty. When the given (listof cr) is not empty, a non-empty
binary search tree is built using the middle cr and the trees obtained by processing
the subintervals before and after the middle element.
Based on the sample expressions we specialize the template for a function on an
interval as follows:
;; (listof cr) [natnum..natnum] → (bbstof cr)
;; Purpose: To create a bbst from the sorted list elements in
;; the given interval
(define (create-bbstofcr a-locr low high)
(if (< high low)
'()
396 17 Binary Trees

(list (list-ref a-locr (+ low (quotient (- high low) 2)))

(create-bbstofcr
a-locr
low
(sub1 (+ low (quotient (- high low) 2))))
(create-bbstofcr
a-locr
(add1 (+ low (quotient (- high low) 2)))
high))))
Observe that instead of using concrete values, like 0 and 6, the parameters low and
high are used in their place. Finally, the tests are written as follows:
;; Tests using sample computations for create-bstofcr
(check-expect (create-bbstofcr CR-DB0 0 -1) CBST1)
(check-expect (create-bbstofcr SORTED-LOCR 0 6) CBST2)

;; Tests using sample values for create-bstofcr

(check-expect
(create-bbstofcr
(list PAULIE SCARFACE VITO DILLINGER BUGSY)
0
4)
(list VITO
(list PAULIE
'()
(list SCARFACE '() '()))
(list DILLINGER
'()
(list BUGSY '() '()))))
Observe that the test using a sample value is another example that illustrates that the
resulting binary search tree is balanced.

103.2 Analysis

Does representing a database as a balanced binary search tree improve the complexity
of searching? Consider the balanced binary search tree in Fig. 83 (that only depicts
id numbers). Consider searching for a cr record with an id of 87. After comparing
87 with 70 (the root id number) all the elements in the left subtree are discarded
from the search. That is, roughly half of the remaining elements are discarded. After
the comparison with 110 half of the remaining elements are discarded again from
the search. This happens again after the comparison with 90 and with 80. Finally,
after the comparison with 85 get-record-bbstocr is called with the empty tree
and the function halts. Notice that 6 calls are made to get-record-bbstocr. The
103 Balanced (bstof cr) 397

Fig. 83 A balanced binary search tree

70

30 110

10 50 90 130

0 20 40 60 80 100 120 140

75 85 135 150

same number of calls occurs if a cr with an id of 200 is searched for. If a cr with

an id of 45 is searched for, 5 calls to the function occur. Observe that in the worst
case the number of calls is equal to the height of the tree plus 2. That is a call for
every node on the longest path plus one call with the empty tree. Such is the case,
for example, for this path:
70 → 110 → 130 → 140 → 135
What is the length of the longest path in a balanced binary search tree with n nodes?
After each step down the tree about half of the elements are discarded. How many
times can we divide n by 2? If n is a power of 2, then the answer is lg n (i.e., log2 n).
If n is not a power of 2, then the answer is lg n , the largest integer less than or equal
to lg n. This means that in the worst case lg n + 2 calls are made. For example,
the balanced binary search tree in Fig. 83 has 19 nodes and in the worst case the
maximum number of calls is
lg n + 2 = lg 19 + 2
= 6
This means that the complexity of searching a balanced binary search tree is O( lg n).
This is significantly better than an O(n). For instance, in the worst case, the number
of function calls to search a database with 1024 crs represented as a (listof cr)
is proportional to 1024. The number of calls is proportional to 10 when the database
is represented as a balanced binary search tree.
Equally noteworthy is that create-bstofcr is a recursive function that is not
based on structural recursion. That is, the recursive calls are not made with the
subinterval used to build the given interval. Instead, new intervals are generated for
the recursive calls. This is called generative recursion. We shall study generative
recursion more in-depth in the next volume of this textbook. For now, you need to
be aware that there are other useful forms of recursion that we have not studied yet.
398 17 Binary Trees

** Ex. 185 — Design and implement a function that takes as input a (bbstof
cr) and that returns the criminal record with the largest id number. Make sure
to do careful problem analysis to make the function as simple as possible.

*** Ex. 186 — Design and implement a function that takes as input a (bbstof
cr) and that returns its height.

*** Ex. 187 — Design and implement a function that takes as input a (bbstof
cr) and that returns a list of criminal names in order by id number.

*** Ex. 188 — Design and implement a function that takes as input a (bbstof
number) and that returns the tree’s second largest number.

**** Ex. 189 — Design and implement a predicate that takes as input a number
and a (bbstof number) and that determines if the given number is a member
of the tree. Also design and implement a predicate that takes as input a number
and a (listof number) and that determines if the given number is a member
of the list. Which do you expect to be faster? Justify your answer using abstract
running time.

**** Ex. 190 — Design and implement a predicate that takes as input a
(bbstof number) and that determines if all tree elements are even. Also
design and implement a predicate that takes as input a (btof number) and
that determines if all tree elements are even. Which do you expect to be faster?
Justify your answer using abstract running time.

** Ex. 191 — What is the complexity of sum-loq from Sect. 72?

**** Ex. 192 — What is the complexity of remove-hit-aliens from

Sect. 77.2?

104 What Have We Learned in This Chapter?

The important lessons of Chap. 17 are summarized as follows:

• A binary tree, (btof X), is used to represent data of arbitrary size.
• The data definition for binary trees has two selfreferences.
• The definition template for a function on a (btof X) has two expressions for
recursive calls: one for each subtree.
• Three common binary tree traversal strategies visit nodes in the following order:
– Preorder: root, left subtree, right subtree
– Inorder: left subtree, root, right subtree
– Postorder: left subtree, right subtree, root
104 What Have We Learned in This Chapter? 399

• A binary search tree, (bstof X), is a subtype of (btof X).

• In a (bstof X) anything in the left subtree is less than that the root value and
anything in the right subtree is greater than the root value. This is an invariant
property for all nodes in the binary search tree.
• Timing programs is an unreliable metric to compare expected execution time.
• Abstract running time counts the number of major operations based on the size
of the input.
• Abstract running time is a more reliable measure to compare expected running
time.
• Typically programmers are interested in the complexity of the worst-case scenario.
• Big O notation is used to describe the complexity of a program in terms of the
input size.
• A balanced binary search tree, (bbstof X), is a binary search tree subtype in
which the height of the left and right subtrees of any node differs by no more than
1.
• The use of a (bbstof X) reduces the complexity of a search from O(n) to
O( lg n).
• Unlike structural recursion that uses the structure of an input to make recur-
sive calls, generative recursion creates new instances of the input type to make
recursive calls.
Chapter 18
Mutually Recursive Data

So far, we have only considered data types that only refer to themselves. For example,
look at the data definitions for a (listof X) and a natnum:
A list of X is either: A natural number (natnum) is either:
1. '() 1. 0
2. (cons X (listof X)) 2. (add1 natnum)
These recursive data definitions only refer to themselves. Now, take a look at the data
definition for a binary tree:
;; A binary tree of X, (btof X), is either:
;; 1. '()
;; 2. (list X (btof X) (btof X))
This data definition also only refers to itself. However, as discussed in Chap. 17,
representing a non-empty binary tree as a list is a bit awkward given that it is not of
arbitrary size. A binary tree is of arbitrary size, but a non-empty binary tree always
has 3 elements: something of type X and two subtrees of type X. That is, the number
of elements needed to represent a non-empty binary tree is finite. This suggests that
a non-empty binary tree ought to be represented using a structure with three fields.
A first attempt to refine the binary tree data definition may be
;; A binary tree of X, (btof X), is either:
;; 1. '()
;; 2. A structure, (make-node X (btof X) (btof X)), with
;; an X value and two subtrees of X values
This data definition ought to feel more natural. An item of finite size is represented
using a structure and not a list.
This revised data definition, however, is still a bit awkward. We have a data
definition within a data definition. That is, we define a node inside the definition of

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 401
M. T. Morazán, Animated Problem Solving, Texts in Computer Science,
https://doi.org/10.1007/978-3-030-85091-3_18
402 18 Mutually Recursive Data

Fig. 84 Templates for node and (btof X)

#| ;; Sample instances of (nodeof X)
(define NODE0 (make-node . . . . . . . . .)) ...
;; node . . . ... Purpose:
(define (f-on-node a-node)
. . .(f-on-X (node-val a-node))
. . .(f-on-btx (node-ltree a-node))
. . .(f-on-btx (node-rtree a-node)))
;; Sample expressions for f-on-node
(define NODE0-VAL . . .)
...
;; Tests using sample computations for f-on-node
(check-expect (f-on-node NODE0 . . .) NODE0-VAL) . . .
;; Tests using sample values for f-on-node
(check-expect (f-on-node (make-node . . . . . . . . .) . . .) . . .) ...

;; Sample instances of (btof X)

(define BTX0 ()) (define BTX1 (make-node . . .)) . . .
;; btx . . . ... Purpose:
(define (f-on-btx a-btx . . .)
(if (empty? a-node)
...
(f-on-node a-btx . . .))))
;; Sample expressions for f-on-btx
(define BTX0-VAL (f-on-btx BTX0 . . .))
(define BTX1-VAL (f-on-node BTX1 . . .))
...
;; Tests using sample computations for f-on-btx
(check-expect (f-on-btx BTX0 . . .) BTX0-VAL)
(check-expect (f-on-btx BTX1 . . .) BTX1-VAL)
...
;; Tests using sample values for f-on-btx
(check-expect (f-on-btx . . .) . . .)
... |#
(define-struct node (val ltree rtree))

a binary tree. It is much more natural to have distinct and separate data definitions,
one for a binary tree and one for a node, as follows:
;; A (nodeof X) is a structure, (make-node X (btof X) (btof X)),
;; with an X value and two (btof X) values.

;; A binary tree of X, (btof X), is either:

;; 1. '()
;; 2. node

Observe that there is something new about these data definitions. These data
definitions are mutually recursive. This means that they define intertwined data. The
data definitions refer to each other. A binary tree cannot be defined without knowing
what a node is and a node cannot be defined without knowing what a binary tree is.
105 Designing with Mutually Recursive Data 403

105 Designing with Mutually Recursive Data

What impact does mutually recursive data have on problem solving? In a very real
sense we already know what references to a data definition mean for our code. For
example, in the data definition for a (list of X) there is a reference to a (listof
X). This means that we must call a function to process a (list of X). Given that
it is a selfreference it turns into a recursive call in our code.
The same principle applies to mutually recursive data definitions. The reference
to a node in the data definition for a (btof X) means that processing a non-empty
binary tree requires calling a function on a node. Similarly, the references to a (btof
X) in the data definition of a node require calls to a function that processes a (btof
X). Observe that the data definition for a (btof X) is still circular. A (btof X)
refers to node and a node refers (back) to (btof X). That is, it defines data of
arbitrary size. The definition is useful because the circularity ends when a (btof
X) is '(). These observations lead to the function templates and structure definition
displayed in Fig. 84. Immediately you can observe that the templates embody the
principle of separation of concerns. A node is processed by a node-processing
function and a (btof X) is processed by a (btof X)-processing function.

105.1 Revisiting the Maximum of a (btof int)

In Sect. 99 we designed a function to find the maximum of a (btof int). Now,

we revisit this problem and solve it based on our refined data definition. To start,
we define sample instances of node and (btof int). Care must be taken writing
these in our program because the definitions are mutually recursive. Remember that
we may not use a value before it is defined. In other words, we may not use a node
or a (btof int) to build another before defining them. For example, we may not
define a binary tree that only contains 177 as follows:
(define NODE177 (make-node 177 BTI0 BTI0))

(define BTI0 '())

We cannot define NODE177 in terms of BTI0 before defining BTI0. The proper way
to build this tree is
(define BTI0 '())

(define NODE177 (make-node 177 BTI0 BTI0))

Observe that anything needed to build NODE177 is defined before its definition. This
indicates the general strategy that we must follow to build instances of mutually
recursive data. Build nonrecursive instances first and then build instances that only
depend on existing instances. For binary trees this means that the empty tree must
be defined first. Once the empty tree is defined, all the leaves may by defined. Once
404 18 Mutually Recursive Data

Fig. 85 A binary tree of integers

100

456 564

177 231 970 487

Fig. 86 Mutually recursive sample instances of (btof int) and (nodeof X)

(define BTI0 ())

(define NODE177 (make-node 177 BTI0 BTI0))

(define NODE231 (make-node 231 BTI0 BTI0))
(define NODE970 (make-node 970 BTI0 BTI0))
(define NODE487 (make-node 487 BTI0 BTI0))

(define BTI177 NODE177)

(define BTI231 NODE231)
(define BTI970 NODE970)
(define BTI487 NODE487)

(define NODE456 (make-node 456 BTI177 BTI231))

(define NODE564 (make-node 564 BTI970 BTI487))

(define BTI456 NODE456)

(define BTI564 NODE564)

(define NODE100 (make-node 100 BTI456 BTI564))

(define BTI100 NODE100)

all the leaves are defined, all nodes that only have leaves as children may be defined.
This process continues until the root node is defined. In other words, we build the
instances needed from the bottom of the binary tree to the root of the tree level by
level. Consider defining the (btof int) displayed in Fig. 85. First, we define the
empty tree. Second, we define all the leaf nodes using the defined empty tree. Third,
we define binary trees rooted at the leaves. Fourth, we define nodes for the parents
of the leaves. Fifth, we define binary trees rooted at the parents of the leaves. Finally,
we define the root node and the corresponding binary tree. The definitions for the
binary tree in Fig. 85 are displayed in Fig. 86. The tree in Fig. 85 is BTI100.
Based on our sample instances we may write sample expressions for the maximum
of a (btof int) as follows:
;; Sample expressions for max-btint
(define BTI100-VAL (max-node BTI100))
(define BTI456-VAL (max-node BTI456))
105 Designing with Mutually Recursive Data 405

Observe that there is no sample expression for the empty (btof int) as this would
throw an error. For a non-empty binary tree a function to find the maximum of a
node is called. As usual, we assume that auxiliary functions exist and work.
These expressions lead to the following specialization of the definition template:
;; (btof int) → int throws error
;; Purpose: Return the maximum of the given (btof int)
(define (max-btint a-btint)
(if (empty? a-btint)
(error "An empty (btof int) has no maximum value.")
(max-node a-btint)))
Observe that this function is now much simpler and easier to read than the function
developed in Sect. 99. It is clear how the answer is computed. If the given binary tree
is empty, then there is no maximum integer. Otherwise, the maximum of the given
(btof int) is the maximum of a node.
The tests may be written as follows:
;; Tests using sample computations for max-btint
(check-expect (max-btint BTI100) BTI100-VAL)
(check-expect (max-btint BTI456) BTI456-VAL)

; Tests using sample values for f-on-btx

(check-error
(max-btint BTI0)
"An empty (btof int) has no maximum value.")
(check-expect (max-btint (make-node -97
(make-node -89 BTI0 BTI0)
(make-node -99 BTI0 BTI0)))
-89)
Observe that once again error checking is part of the tests using sample values.
The second of these tests illustrates that the function works when the tree contains
negative numbers.
We now turn to the problem of finding the maximum integer in a node. As in
Sect. 99, we must take care not to call max-btint with an empty tree. If a node is a
leaf, then the node’s value is the maximum. Otherwise, the maximum of the node’s
value and the maximum of any non-empty subtree are computed. This leads to the
following new node instances and four sample expressions:
(define NODE200 (make-node 200 BTI0 (make-node 311 BTI0 BTI0)))
(define NODE650 (make-node 650 (make-node 912 BTI0 BTI0) BTI0))

;; Sample expressions for max-node

(define NODE487-VAL (node-val NODE487))
(define NODE200-VAL (max (node-val NODE200)
(max-node (node-rtree NODE200))))
406 18 Mutually Recursive Data

(define NODE650-VAL (max (node-val NODE650)

(max-node (node-ltree NODE650))))
(define NODE456-VAL (max (node-val NODE456)
(max-node (node-ltree NODE456))
(max-node (node-rtree NODE456))))
The two new node instances are written to make the development of sample expres-
sions easier. The first has an empty left tree and the second has an empty right tree.
Observe that there is an expression for each node condition possible: leaf, only left
subtree is empty, only right subtree is empty, and no subtrees are empty.
Based on the sample expressions, we may specialize the definition template for a
function on a node as follows:
;; node → int
;; Purpose: Return the max int in the given node
(define (max-node a-node)
(cond [(and (empty? (node-ltree a-node))
(empty? (node-rtree a-node)))
(node-val a-node)]
[(empty? (node-ltree a-node))
(max (node-val a-node)
(max-btint (node-rtree a-node)))]
[(empty? (node-rtree a-node))
(max (node-val a-node)
(max-btint (node-ltree a-node)))]
[else (max (node-val a-node)
(max-btint (node-ltree a-node))
(max-btint (node-rtree a-node)))]))
Observe the absence of first, second, and third to access the components of a
node. These list functions needed to be used in Sect. 99 because a non-empty tree
was represented as a list. Now the interface is cleaner and more elegant. For instance,
any reader of our code easily understands that (node-rtree a-node) evaluates to
a-node’s right subtree. In Sect. 99, (third a-btint) was used to access the right
subtree. Does this expression in any way suggest that it ought to evaluate to the right
subtree? You are now beginning to understand the importance of writing elegant
code by using separation of concerns.
The tests may be written as follows:
;; Tests using sample computations for max-node
(check-expect (max-node NODE487) NODE487-VAL)
(check-expect (max-node NODE200) NODE200-VAL)
(check-expect (max-node NODE650) NODE650-VAL)
(check-expect (max-node NODE456) NODE456-VAL)
106 Evaluating Arithmetic Expressions 407

;; Tests using sample values for max-node

(check-expect (max-node (make-node
789
(make-node -1000 BTI0 BTI0)
(make-node 3000 BTI0 BTI0)))
3000)
Observe that any reader can determine that the function properly computes the value
of the sample expressions. In addition, they can see that max-node works for an
arbitrary node that contains both positive and negative integers.

106 Evaluating Arithmetic Expressions

In a seminal 1960 research article, Recursive Functions of Symbolic Expressions and

Their Computation by Machine, Part I, Prof. John McCarthy from The Massachusetts
Institute of Technology defined symbolic expressions to represent the expressions
that we use today to program in ISL+.24 McCarthy developed a set of mutually
recursive data definitions for symbolic expressions that may be used to represent a
computation that, subsequently, may be evaluated by a computer.
In this section, we shall not tackle the problem of evaluating ISL+ programs.
Instead, we tackle the problem of evaluating the subset of symbolic expressions
representing arithmetic expressions using +, -, and *. To start, let us look at a few
sample expressions:
45 87
(+ 67 54) (* (- 56 43) 44 (+ 7 4))
Observe that there are two varieties of symbolic expressions. The simplest symbolic
expression is a number. The second variety is a list that contains a symbol repre-
senting a function followed by a list of symbolic expressions. This list of symbolic
expressions represents the arguments to the function.
The above data analysis naturally leads to the following set of mutually recursive
data definitions:
A symbolic expression (sexpr) is either:
1. number
2. slist

A symbolic list (slist) is: (cons function losexpr)

A function is either: A list of sexpr (losexpr) is either:

1. '+ 1. '()
2. '- 2. (cons sexpr losexpr)
3. '*

24
In fact, many other programming languages represent programs as symbolic expressions such as
LISP, Scheme, Racket, and BSL.
408 18 Mutually Recursive Data

Fig. 87 The function templates for sexpr and slist

#| ;; Samples of sexpr
(define SEXPR1 ...) (define SEXPR2 ...)

;; sexpr ... --> ... Purpose:

(define (f-on-sexpr a-sexpr)
(if (number? a-sexpr)
(f-on-number a-sexpr ...)
(f-on-slist a-sexpr ...)))
;; Sample expressions for f-on-sexpr
(define SEXPR1-VAL ...) (define SEXPR2-VAL ...)

;; Tests using sample computations for f-on-sexpr

(check-expect (f-on-sexpr SEXPR1) SEXPR1-VAL)
(check-expect (f-on-sexpr SEXPR2) SEXPR2-VAL)
;; Tests using sample values for f-on-sexpr
(check-expect (f-on-sexpr ... ...) ...)

;; Samples of slist
(define SLIST1 ’()) (define SLIST2 ...)

;; slist ... --> ... Purpose:

(define (f-on-slist an-slist)
(if (empty? an-slist)
...
(cons (f-on-sexpr (first an-slist) ...)
(f-on-slist (rest an-slist) ...))))
;; Sample expressions for f-on-slist
(define SLIST1-VAL ...) (define SLIST2-VAL ...)

;; Tests using sample computations for f-on-slist

(check-expect (f-on-slist SLIST1 ...) SLIST1-VAL)
(check-expect (f-on-slist SLIST2 ...) SLIST2-VAL)
;; Tests using sample values for f-on-slist
(check-expect (f-on-slist ... ...) ...)
|#

Observe that processing an sexpr requires processing either a number or an slist.

Processing an slist requires processing a function and a losexpr. Processing
a function requires processing one of the three symbols. Finally, processing a
losexpr requires either processing the empty list or processing a sexpr and a
losexpr. It is clear that sexpr, slist, and losexpr are mutually recursive. With
this understanding the function templates in Figs. 87 and 88 are developed. It is
worth noting now that below functions are developed following the steps of the
design recipe. Given that the functions are mutually recursive, however, in DrRacket
all the functions must appear before the sample expressions and tests. Otherwise,
DrRacket will throw an error telling you that a function is used before its definition.
106 Evaluating Arithmetic Expressions 409

Fig. 88 The template for function

#| ;; function ... --> ... Purpose:
(define (f-on-function a-function)
(cond [(eq? a-function ’+) ...]
[(eq? a-function ’-) ...]
[else ...]))

;; Sample expressions for f-on-function

(define PLUS-VAL ...)
(define SUBT-VAL ...)
(define MULT-VAL ...)

;; Tests using sample computations for f-on-function

(check-expect (f-on-function PLUS ...) PLUS-VAL)
(check-expect (f-on-function SUBT ...) SUBT-VAL)
(check-expect (f-on-function MULT ...) SUBT-VAL)

;; Tests using sample values for f-on-function

(check-expect (f-on-function ... ...) ...)
|#

We first start by defining the following sample instances for the different data
types:
;; Sample instances of slist
(define SLIST1 '(* (+ 44 -44) (- 20 10)))
(define SLIST2 '(* 3 (+ 6 4)))

;; Sample instances of sexpr

(define SEXPR1 67)
(define SEXPR2 SLIST2)

;; Sample instances of (listof sexpr)

(define LOSEXPR1 '())
(define LOSEXPR2 '(80 (+ 70 20) (* 4 (- 30 5))))

;; Sample instances of (listof number)

(define LON1 '())
(define LON2 '(1 2 3 4 5))
(define LON3 '(-10 -20 -30))
These sample instances are used to develop the steps of the design recipe for the
program to evaluate arithmetic expressions. The function sample instances are
omitted because it is simple enough to directly use the three possible symbols. The
(listof number) instances are explicitly written to make it easier to develop the
function that applies a given function to a list of arguments.
Now that we know what is required to process the different data types, we may
proceed with problem analysis for evaluating arithmetic expressions. We start with
410 18 Mutually Recursive Data

analyzing how to evaluate a sexpr. The evaluation of a number sexpr is the number
itself. The evaluation of an slist sexpr requires calling a function to evaluate an
slist. This analysis leads to the following sample expressions:
;; Sample expressions for eval-sexpr
(define SEXPR1-VAL SEXPR1) ;; for a number
(define SEXPR2-VAL (eval-slist SEXPR2)) ;; for a slist
Observe that the sample expressions illustrate how to implement our problem anal-
ysis.
The next step is to specialize the definition template and tests. This is accomplished
as follows:
;; sexpr → number throws error
;; Purpose: To evaluate the given sexpr
(define (eval-sexpr a-sexpr)
(if (number? a-sexpr)
a-sexpr
(eval-slist a-sexpr)))

;; Tests using sample computations for eval-sexpr

(check-expect (eval-sexpr SEXPR1) SEXPR1-VAL)
(check-expect (eval-sexpr SEXPR2) SEXPR2-VAL)

;; Tests using sample values for eval-sexpr

(check-expect (eval-sexpr 42) 42)
(check-expect (eval-sexpr '(+ -1 (* 6 (+ 8 2)))) 59)
At this point, the function and tests ought to feel quite straightforward to you. They
are the direct results of our problem analysis and of following the steps of the design
recipe. The only detail that remains unclear is why does the signature state that this
function can throw an error. Bear with this for now. The answer will become clear
as our design progresses.
As expected, our function design has revealed the need for an auxiliary function:
eval-slist. This function must evaluate an slist sexpr. Recall that an slist
represents the application of a function to its arguments. To do so, the arguments to
the function must be evaluated; then the function must be applied to them. Observe
that these are two different tasks and, therefore, two auxiliary functions are needed:
one to evaluate the arguments and one to apply the function. The template for a
function of an slist is specialized as follows:
;; Sample expressions for eval-slist
(define SLIST1-VAL (apply-f (first SLIST1)
(eval-args (rest SLIST1))))
(define SLIST2-VAL (apply-f (first SLIST2)
(eval-args (rest SLIST2))))
106 Evaluating Arithmetic Expressions 411

;; slist → number
;; Purpose: To evaluate the given slist
(define (eval-slist an-slist)
(apply-f (first an-slist) (eval-args (rest an-slist))))

;; Tests using sample computations for eval-slist

(check-expect (eval-slist SLIST1) SLIST1-VAL)
(check-expect (eval-slist SLIST2) SLIST2-VAL)

;; Tests using sample values for eval-slist

(check-expect (eval-slist
'(- (* 10 10) (+ 5 (* 2 3) 7))) 82)
Observe that the sample expressions clearly illustrate that the arguments are evaluated
first and then the function is applied to them. Given that there is only one difference
among the sample expressions eval-slist only needs one parameter. The test using
a sample value illustrates that the function works for all functions with multiple nested
levels.
To evaluate the list of arguments each sexpr in the given list must be evaluated
and a list of the results must be returned. This is a mapping operation (as we
studied in Sect. 76). Therefore, the template for a function on a (listof sexpr)
is specialized. If the given list is empty, the empty list is returned. Otherwise, cons
is used to create a new list containing the result of evaluating the first sexpr and
recursively evaluating the rest of the sexprs. The template specialization is
;; Sample expressions for eval-args
(define LOSEXPR1-VAL '())
(define LOSEXPR2-VAL (cons (eval-sexpr (first LOSEXPR2))
(eval-args (rest LOSEXPR2))))

;; (listof sexpr) arrow (listof number) throws error

;; Purpose: To evaluate the sexprs in the given list
(define (eval-args a-losexpr)
(if (empty? a-losexpr)
'()
(cons (eval-sexpr (first a-losexpr))
(eval-args (rest a-losexpr)))))

;; Tests using sample computations for eval-args

(check-expect (eval-args LOSEXPR1) LOSEXPR1-VAL)
(check-expect (eval-args LOSEXPR2) LOSEXPR2-VAL)

;; Tests using sample values for eval-args

(check-expect (eval-args '((- 1 1) 89)) '(0 89))
As before, the reason for the signature stating that the function may throw an error
will become clear as our design advances.
412 18 Mutually Recursive Data

We may now proceed to design apply-f. This function takes as input a function
and a (listof number). It is necessary to determine the value of the given
function to properly combine the given numbers. The combination of the given
numbers involves one of three different tasks. Each of these tasks is performed by a
different auxiliary function. Therefore, the template for a function on a function is
specialized as follows:
;; Sample expressions for apply-f
(define APPLY1-VAL (sum-lon LON1))
(define APPLY2-VAL (subt-lon LON2))
(define APPLY3-VAL (mult-lon LON3))

;; function (listof number) → number throws error

;; Purpose: Apply the given function to the given numbers
(define (apply-f a-function a-lon)
(cond [(eq? a-function '+) (sum-lon a-lon)]
[(eq? a-function '-) (subt-lon a-lon)]
[else (mult-lon a-lon)]))

;; Tests using sample computations for apply-f

(check-expect (apply-f '+ LON1) APPLY1-VAL)
(check-expect (apply-f '- LON2) APPLY2-VAL)
(check-expect (apply-f '* LON3) APPLY3-VAL)

;; Tests using sample values for apply-f

(check-expect (apply-f '+ '(10 10 10)) 30)
(check-expect (apply-f '- '(20 0)) 20)
(check-expect (apply-f '* '(9 -2)) -18)
The design of sum-lon is exactly the same as the design of sum-loq in Sect. 72.
The design of mult-lon is very similar and the results of following the steps of the
design recipe are displayed in Fig. 89.
Of the three auxiliary functions the most interesting one is subt-lon. We must
carefully analyze how the arithmetic minus function is applied to an arbitrary number
of arguments. Let us look at a few interactions with DrRacket:
> (- 1)
-1
> (- 2 -4)
6
> (- 10 20 -5)
-5
How is - applied to its arguments? According the ISL+ documentation this is how
subtraction works:
Subtracts the second (and following) number(s) from the first;
negates the number if there is only one argument.
106 Evaluating Arithmetic Expressions 413

Fig. 89 The function to multiply the numbers in a list of numbers

;; Sample instances of (listof number)
(define LON1 ())
(define LON2 (1 2 3 4 5))
(define LON3 (-10 -20 -30))

;; lon number
;; Purpose: To multiply the numbers in the given lon
(define (mult-lon a-lon)
(if (empty? a-lon)
1
(* (first a-lon) (mult-lon (rest a-lon)))))

;; Sample expressions for mult-lon

(define LON1-MULT 1)
(define LON2-MULT (* (first LON2) (mult-lon (rest LON2))))

;; Tests using sample computations for mult-lon

(check-expect (mult-lon LON1) LON1-MULT)
(check-expect (mult-lon LON2) LON2-MULT)

;; Tests using sample computations for mult-lon

(check-expect (mult-lon (-10 0 10)) 0)

Therefore, we have
(- 1) = -1
(- 2 -4) = 2 - -4
(- 10 20 -5) = 10 - 20 - -5
What happens if - is provided no arguments? Try it in DrRacket. You can see that an
error is thrown. This is why eval-sexpr, eval-slist, eval-args, and apply-f
may throw an error. The function subt-lon is called by apply-f and, therefore,
calling apply-f may generate an error. In turn, apply-f is called by eval-slist,
which itself is called by eval-sexpr. Thus, these two functions may also throw
an error. Finally, eval-args may throw an error because it calls eval-sexpr. In
general, if a function f may lead to a call to function g that throws an error, then f
may also throw an error.
Now that we understand how to apply -, we can determine how to implement a
function to perform this task. If the given list of numbers is empty, then an error is
thrown. Otherwise, we need to subtract the rest of the numbers from the first number.
How can this be done? Observe that a - is placed before each number in the rest of
the list. By using a little Algebra we can factor out the -. For example, the last two
expressions above may be rewritten as follows:
(- 2 -4) = 2 - -4 = 2 - (-4)
(- 10 20 -5) = 10 - 20 - -5 = 10 - (20 + -5)
We can subtract from the first number the sum of the rest of the numbers.
414 18 Mutually Recursive Data

We can now specialize the template for a function on a (listof number) as

follows:
;; lon → number throws error
;; Purpose: To subtract the given lon
(define (subt-lon a-lon)
(if (empty? a-lon)
(error "No numbers provided to -.")
(- (first a-lon) (sum-lon (rest a-lon)))))

;; Sample expressions for subt-lon

(define LON2-SUBT (- (first LON2) (sum-lon (rest LON2))))

;; Tests using sample computations for subt-lon

(check-error (subt-lon LON1) "No numbers provided to -.")
(check-expect (subt-lon LON2) LON2-SUBT)

;; Tests using sample computations for subt-lon

(check-expect (subt-lon ’(-10 10 10)) -30)
Observe that we make use of sum-lon in our implementation. In this manner, there
is no need for us to write a function that does repeated subtractions for us.

107 Trees

The concept of a binary tree may be generalized. Instead of each node having at
most two children, in a tree a node may have an arbitrary number of children. Figure
90 displays a tree representing moves on a Tic Tac Toe board.25 At level 0, the root
has 3 children. At level 1, the children of the root each have two children. At level 2,
the grandchildren of the root each have 1 child. Finally, at level 3 the nodes do not
have any children. As you can observe the number of children a node has is arbitrary.
How many children does the node representing the empty board at the beginning
of a game has? Clearly, such a node would have 9 children—one for each possible
move the first player may make.
Trees are versatile and may be used to represent many real or imaginary objects.
One use of trees is to represent a search space. A search space defines all possible
solutions to a problem and is searched to find a solution. For example, Fig. 90
represents all possible paths in a Tic Tac Toe gaming starting at the root node. Such a
tree, for instance, may be taken as input by a function that determines the next move
the computer ought to make.

25
Tic Tac Toe is played on a 3 × 3 board formed by using two vertical and two horizontal lines.
There are two players. One is designated 'X and the other 'O. The player’s goal is to get three
symbols in a row.
107 Trees 415

Fig. 90 A tree of Tic Tac Toe paths

o
x x o

o x

o x o x o

x x o x x o x x o
o x o x o x x

o x o o x o o x o x o o o o

x x o x x o x x o x x o x x o x x o

o x o o x o x o o x o x x o x x

o x o o x o o o x o x x o o x o x o

x x o x x o x x o x x o x x o x x o

o x x o x x o x x o o x o x x o x x

We must now decide how to represent a search tree for Tic Tac Toe. Observe that
a search tree is made of nodes. Stop and think carefully what this means to a problem
solver. It suggests that two data definitions are needed: one for a node and one for a
search tree. Let us think carefully about a node. A node has a board (i.e., the current
state of the game) and a list of children. We use of list because the number of children
is arbitrary. Observe two more things. The first is the need for a third data definition:
a board. The second is that a node is of finite size and always contains two elements.
Now, let us think carefully about what a search tree is. What should the search tree
be if there is no game? We are free to choose any representation. Without loss of
generality, we shall say it is '(). If there is a game, then the search tree must be a
node. This data analysis suggests the following data definitions:
;; A node is a structure: (make-node board (listof st))
(define-struct node (board children))

;; A solution tree (st) is either:

;; 1. '()
;; 2. node

;; A board value (bval) is either:

;; 1. 'X
;; 2. 'O
;; 3. 'B

;; A board is a structure:
;; (make-board bval bval bval bval bval bval bval bval bval)
(define-struct board (p0 p1 p2 p3 p4 p5 p6 p7 p8))
416 18 Mutually Recursive Data

;; Sample board values

(define INIT-BOARD (make-board 'B 'B 'B
'B 'B 'B
'B 'B 'B))
(define BOARD1 (make-board 'X 'O 'X
'O 'X 'O
'X 'O 'X))
(define BOARD2 (make-board 'X 'B 'B
'O 'B 'B
'B 'B 'B))
(define BOARD3 (make-board 'X 'B 'B
'X 'O 'B
'B 'B 'B))
(define BOARD4 (make-board 'X 'B 'B
'B 'B 'B
'B 'B 'B))
(define BOARD5 (make-board 'X 'X 'O
'B 'X 'X
'O 'O 'O))
(define BOARD6 (make-board 'X 'B 'B
'B 'X 'X
'O 'B 'O))
A board is a structure with nine bvals and a bval is one of three symbols. Observe
that node, (listof st), and st are mutually recursive. That is, node refers to
(listof st), (listof st) refers to st, and st refers to node. As problem solvers
this immediately tells us that writing a function to process a node requires writing
functions to process a (listof st) and a function to process an st.
Given out insights we can now write the required function templates as follows:
#| TEMPLATES
;; Sample instances of (listof st)
(define LOST1 '())
(define LOST2 . . .)

;; (listof st) . . . → . . .
;; Purpose:
(define (f-on-lost an-lost)
(if (empty? an-lost)
...
. . .(f-on-st (first an-lost)). . .(f-on-lost (rest an-lost))))

;; Sample expressions for f-on-st

(define ST1-VAL . . .)
(define ST2-VAL . . .)
107 Trees 417

;; Tests using sample computations f-on-st

(check-expect (f-on-st ST1 . . .) ST1-VAL)
(check-expect (f-on-st ST2 . . .) ST2-VAL)

;; Tests using sample values f-on-st

(check-expect (f-on-st . . . . . .) . . .)

;; Sample instances of node

(define NODE1 . . .)

;; node . . . → . . .
;; Purpose:
(define (f-on-node a-node)
. . .(f-on-board (node-board a-node)). . .
. . .(f-on-lost (node-children a-node)). . .)

;; Sample expressions for f-on-node

(define NODE1-VAL . . .)

;; Tests using sample computations f-on-node

(check-expect (f-on-node NODE1 . . .) NODE1-VAL)

;; Tests using sample values f-on-node

(check-expect (f-on-node . . . . . .) . . .)

;; Sample instances of ST
(define ST1 '())
(define ST2 . . .)

;; st . . . → . . .
;; Purpose:
(define (f-on-st an-st)
(if (empty? an-st)
...
(f-on-node an-st)))

;; Sample expressions for f-on-st

(define ST1-VAL . . .)
(define ST2-VAL . . .)

;; Tests using sample computations f-on-st

(check-expect (f-on-node ST1 . . .) ST1-VAL)
(check-expect (f-on-node ST2 . . .) ST2-VAL)
418 18 Mutually Recursive Data

;; Tests using sample values f-on-st

(check-expect (f-on-st . . . . . .) . . .)
|#
As expected, f-on-node has an expression to call f-on-lost, f-on-lost has an
expression to call f-on-st, and f-on-st has an expression to call f-on-node. The
function templates for board and bval are left as exercises.

* Ex. 193 — Develop a template for a function on a board.

* Ex. 194 — Develop a template for a function on a bval.

107.1 Creating a Search Tree for Tic Tac Toe

A natural question that you may have is how to create a tree representing, for example,
the complete st for Tic Tac Toe starting from the blank board. To create an st we
need to know whose turn it is. Therefore, we create the following data definition and
sample instances:
;; A turn is either 'X or 'O

(define INIT-TURN 'X)

(define O-TURN 'O)
By convention the first move is made by 'X and this is the convention we shall follow.
Given a board and a turn the blanks in the board must be identified. A child for the
given board must be created for each blank. Observe that this data analysis suggests
that the function to create an st does not process neither the given board nor the
given turn. Instead, a function to find the blanks in the given board and a function
to process the blanks (by creating an st for each possible successive board) are
needed. A node, therefore, is created using the given board and the sts created by
processing the blanks of the given board. Sample expressions may be written as
follows:
;; Sample expressions for create-st
(define BOARD1-VAL (make-node BOARD1
(process-blanks
BOARD1
(find-blanks BOARD1)
'O)))
(define BOARD2-VAL (make-node BOARD2
(process-blanks
BOARD2
(find-blanks BOARD2)
'X)))
107 Trees 419

The function process-blanks returns a (listof st) containing one st for each
blank in the given board.
Abstracting over the sample expressions yields the following function:
;; board turn → st
;; Purpose: Create the st for the given board
(define (create-st a-board a-turn)
(make-node a-board
(process-blanks a-board
(find-blanks a-board)
a-turn)))
This function may be tested as follows:
;; Tests using sample computations for create-st
(check-expect (create-st BOARD1 'O) BOARD1-VAL)
(check-expect (create-st BOARD2 'X) BOARD2-VAL)

;; Tests using sample values for create-st

(check-expect (create-st (make-board 'X 'O 'X
'O 'X 'O
'X 'B 'B)
'O)
(make-node
(make-board 'X 'O 'X
'O 'X 'O
'X 'B 'B)
(list
(make-node
(make-board 'X 'O 'X
'O 'X 'O
'X 'O 'B)
(list
(make-node
(make-board 'X 'O 'X
'O 'X 'O
'X 'O 'X)
'())))
(make-node
(make-board 'X 'O 'X
'O 'X 'O
'X 'B 'O)
(list
(make-node
(make-board 'X 'O 'X
'O 'X 'O
'X 'X 'O)
'()))))))
420 18 Mutually Recursive Data

The test using a sample value is written using a board that only has a couple of moves
left to make it feasible to write the expected st.

*** Ex. 195 — Write the function find-blanks by specializing the template
for a function on a board. To do so, you may find the following data definition
useful:
;; A board position (bpos) is a natnum in [0..8]

The board positions may be numbered starting at the top left corner at 0 and
ending at the bottom right corner at 8. The function, find-blanks, must return
a (listof bpos).

*** Ex. 196 — Write model checking tests using boards that have a large num-
ber of moves left. What properties can you test about an st?

We can now turn our attention to the design of process-blanks. We assume

that you have implemented a solution to the first exercise above. This function must
process the given (listof bpos). For each blank position in the given list a new
board is created by placing the given bval at that position in the given board. This
board is then used to create an st for the (listof st) returned. We may write
sample expressions as follows:
;; Sample expressions for process-blanks
(define PROC-0BLANKS '())
(define PROC-6BLANKS
(cons (create-st (place-on-board BOARD3
(first '(1 2 5 6 7 8))
'O)
(if (eq? 'O 'X) 'O 'X))
(process-blanks
BOARD3 (rest '(1 2 5 6 7 8)) 'O)))
If the given (listof bpos) is empty, then there are no sts to create and the empty
(listof st) is the answer. Otherwise, a non-empty (listof st) is returned. For
BOARD3, the first st is created using the board obtained by placing an O in position
1 and the next turn (i.e., X). The rest of the sts are recursively created by processing
'(2 5 6 7 8) along with the same board (i.e., BOARD3) and the same turn (i.e., O).
107 Trees 421

Based on the insights obtained from our problem analysis and sample expressions
we may write the function as follows:
;; board (listof bpos) turn → (listof st)
;; Purpose: Build the st for the given board and its given list
;; of blank positions
(define (process-blanks a-board a-lobpos a-turn)
(if (empty? a-lobpos)
'()
(cons (create-st
(place-on-board a-board (first a-lobpos) a-turn)
(if (eq? a-turn 'X) 'O 'X))
(process-blanks a-board (rest a-lobpos) a-turn))))
Observe that the if-expression is used to swap X to O and O to X for the call to
create-st. Finally, tests are written as follows:
;; Tests using sample computations for process-blanks
(check-expect (process-blanks BOARD1 '() 'X)
PROC-0BLANKS)
(check-expect (process-blanks BOARD3 '(1 2 5 6 7 8) 'O)
PROC-6BLANKS)

;; Tests using sample values for process-blanks

(check-expect (process-blanks
(make-board 'B 'X 'O
'O 'X 'X
'O 'X 'O)
'(0)
'X)
(list (make-node
(make-board 'X 'X 'O
'O 'X 'X
'O 'X 'O)
'())))
The test using a sample value is written using a board that only has one move left to
make it feasible to write the expected (listof st). Finally, given that the functions
are mutually recursive, remember to place the functions first in your file and after
them the results of the other steps of the design recipe.

** Ex. 197 — Write the function place-on-board by specializing the tem-

plate for a function on a bpos.

*** Ex. 198 — Write model checking tests for process-blanks. What prop-
erties can you test for a returned (listof st)?
422 18 Mutually Recursive Data

*** Ex. 199 — The design of create-st continues to place moves on boards
where one of the players has already won. Redesign create-st so that it does
not create children for such boards.

107.2 Can Win Tic Tac Toe?

We now turn to an example of processing an st. Given an st consider determining

if it is possible for a given player to win by any number of moves. We must clearly
define what a move may be. It is not difficult to see that a move is a whole number
in [0..8]. This means that the root board in the given st and all the boards under
it must be checked for a win by the given turn. If the given st is empty, then the
answer is #false. Otherwise, according to the template for a function on an st, a
function to determine if the given turn may win in any number of moves for a given
node must be called. Based on this problem analysis, we may right sample instances
and expressions as follows:
;; Sample st instances
(define INIT-ST (create-st INIT-BOARD 'X))
(define BRD7-ST (create-st BOARD7 'O))

;; Sample expressions for st-can-win?

(define WIN-ST1-VAL #false)
(define WIN-ST2-VAL (node-can-win? INIT-ST 'X))
(define WIN-ST3-VAL (node-can-win? BRD7-ST 'O))
The first st sample instance is created using the initial board and the second from
BOARD7 that has two moves left. The first sample expression is for the empty st. The
second sample expression ought to evaluate to #true because X can still win the
game. The third sample expression ought to evaluate to #false because O cannot
win the game.
Abstracting over the sample expressions leads to the following function definition:
;; st turn → Boolean
;; Purpose: Determine if there is a sequence of moves for
;; the given turn to win
(define (st-can-win? an-st a-turn)
(if (empty? an-st)
#false
(node-can-win? an-st a-turn)))
Observe that separation of concerns renders a function that is straightforward to
understand. The answer is false if the given st is empty. Otherwise, the answer is
given by processing a node.
107 Trees 423

The tests may be written as follows:

;; Tests using sample computations st-can-win?
(check-expect (st-can-win? WIN-ST1 'X) WIN-ST1-VAL)
(check-expect (st-can-win? WIN-ST2 'X) WIN-ST2-VAL)
(check-expect (st-can-win? WIN-ST3 'O) WIN-ST3-VAL)

;; Tests using sample values can-win-st?

(check-expect (st-can-win? (create-st (make-board 'O 'X 'B
'X 'O 'X
'O 'X 'O)
'X)
'X)
#false)

(check-expect (st-can-win? (create-st (make-board 'B 'B 'B

'X 'O 'X
'B 'B 'B)
'O)
'O)
#true)
As always, the tests using sample computations illustrate that the abstraction over
the sample expressions computes the expected values. The tests using sample val-
ues illustrate using explicit values for which both possible outcomes are properly
computed.
With st-can-win? designed and implemented, we can now turn our attention
to designing node-can-win?. Recall that a node has a board and a list of children
sts. What does it mean that the given turn can win? The given turn can win if it
has won in the node’s board. This means two things. First, the given turn has three
of its values in any row. Second, the given turn’s opponent has not won. The given
turn can also win if it can win through any of the given node’s children. Based on
this problem analysis we may write sample instances and expressions as follows:
;; Sample instances of node
(define WIN-NODE2 WIN-ST2)
(define WIN-NODE3 WIN-ST3)

;; Sample expressions for node-can-win?

(define WIN-NODE2-VAL
(or
(and (not (has-win? (node-board WIN-NODE2)
(flip 'X)))
(has-win? (node-board WIN-NODE2) 'X))
(lost-can-win? (node-children WIN-NODE2) 'X)))
424 18 Mutually Recursive Data

(define WIN-NODE3-VAL
(or
(and (not (has-win? (node-board WIN-NODE3)
(flip 'O)))
(has-win? (node-board WIN-NODE3) 'O))
(lost-can-win? (node-children WIN-NODE3) 'O)))
The sample expressions test if the given node’s board is a win for the given turn or
if any of the children lead to a win. The auxiliary function, flip, is used to change
an 'X to an 'O and vice versa. As expected, the children are processed by calling a
function on a lost.
Abstracting over the sample expressions leads to the following function definition:
;; node turn → Boolean
;; Purpose: To determine if the given turn can reach a win
(define (node-can-win? a-node a-turn)
(or (and (not (has-win? (node-board a-node) (flip a-turn)))
(has-win? (node-board a-node) a-turn))
(lost-can-win? (node-children a-node) a-turn)))
This function may be tested as follows:
;; Tests using sample computations node-can-win?
(check-expect (node-can-win? WIN-NODE2 'X) WIN-NODE2-VAL)
(check-expect (node-can-win? WIN-NODE3 'O) WIN-NODE3-VAL)

;; Tests using sample values node-can-win?

(check-expect (node-can-win?
(create-st (make-board 'B 'B 'B
'B 'X 'B
'B 'B 'B)
'O)
'O)
#true)

(check-expect (node-can-win?
(create-st (make-board 'O 'O 'O
'X 'X 'B
'X 'B 'B)
'X)
'X)
#false)
The tests using sample computations, once again, illustrate that the function computes
the values of the sample expressions. The tests using sample values provide examples
of nodes for which a win is possible and a win is not possible for the given turn.
107 Trees 425

* Ex. 200 — Design and implement the function flip.

The function lost-can-win? must determine if for a given turn any of its sts
lead to a win. If the given list is empty, then the answer is #false. Otherwise, a win
is possible if the first st leads to a win or any of the rest of the sts leads to a win.
Observe that a Boolean value is computed. This means we may opt to use a Boolean
function without using a conditional expression as suggested by the template for a
function on a lost. We may write sample instances and expressions as follows:
;; Sample instances of lost
(define WIN-LOST1 '())
(define WIN-LOST2 (node-children WIN-NODE2))
(define WIN-LOST3 (node-children WIN-NODE3))

;; Sample expressions for lost-can-win?

(define WIN-LOST1-VAL (and (not (empty? WIN-LOST1))
(or (st-can-win? (first WIN-LOST1)
'X)
(lost-can-win? (rest WIN-LOST1)
'X))))
(define WIN-LOST2-VAL (and (not (empty? WIN-LOST2))
(or (st-can-win? (first WIN-LOST2)
'X)
(lost-can-win? (rest WIN-LOST2)
'X))))
(define WIN-LOST3-VAL (and (not (empty? WIN-LOST3))
(or (st-can-win? (first WIN-LOST3)
'O)
(lost-can-win? (rest WIN-LOST3)
'O))))
Observe that the sample expressions are communicating the ideas in our problem
analysis. A win is possible only if the given lost is not empty and either the first st
leads to a win or any of the other sts lead to a win.
Abstracting over the sample expressions allows us to continue specializing the
template for a function on a lost as follows:
;; (listof st) turn → Boolean
;; Purpose: Determine if the given turn can win through any st
(define (lost-can-win? an-lost a-turn)
(and (not (empty? an-lost))
(or (st-can-win? (first an-lost) a-turn)
(lost-can-win? (rest an-lost) a-turn))))
426 18 Mutually Recursive Data

Finally, the function may be tested as follows:

;; Tests using sample computations lost-can-win?
(check-expect (lost-can-win? WIN-LOST1 'X) WIN-LOST1-VAL)
(check-expect (lost-can-win? WIN-LOST2 'O) WIN-LOST2-VAL)
(check-expect (lost-can-win? WIN-LOST3 'O) WIN-LOST3-VAL)

;; Tests using sample values lost-can-win?

(check-expect
(lost-can-win?
(node-children (create-st (make-board 'X 'B 'O
'O 'X 'O
'B 'X 'X)
'O))
'O)
#false)

(check-expect (lost-can-win?
(node-children
(create-st (make-board 'B 'B 'O
'O 'X 'O
'B 'X 'X)
'X))
'X)
#true)
The tests using sample values clearly communicate that both possible expected values
are properly computed. For the tests using sample values, any reader of our code
can easily observe that for the first test O cannot win and for the second test X can
win. As with all mutually recursive functions, place all function definitions before
the results of the other steps of the design recipe in order to run the program.

108 Project: Tic Tac Toe

The goal of this project is for you to design and implement a Tic Tac Toe game. In
this game a player and the computer are opponents. The player is X and plays first.
The game initially displays an initial board similar to the one displayed in Fig. 91.
Feel free to embellish the game with graphics that you like. The player uses the
mouse to click on a board box to make a move. If it is the player’s turn and the
clicked box is blank, then an X is placed in the clicked box. After this the computer
makes a move by placing an O in any of the remaining boxes. The game cycles in
this manner until there is a winner or when there is a draw when all the boxes in the
board are not blank.
108 Project: Tic Tac Toe 427

Fig. 91 The initial board image for Tic Tac Toe

To get you started here is the run function for the game:
(define (run a-name)
(big-bang INIT-WORLD
(on-draw draw-world)
(on-mouse process-mouse)
(on-tick process-tick)
(stop-when game-over? draw-last-world)
(name A-NAME)))
You need to follow a top-down design strategy starting with the event-handling
functions in the big-bang expression. The rest of the section will help you with
each of these.

108.1 Data Analysis

There are two basic components that must be represented: the world and the image
for representing the world. Of these the easiest is the image of the world. This image
ought to include the game’s current board. For example, at the beginning of the
game the image rendered must include the initial board displayed in Fig. 91. If you
examine this image, you can see that the scene contains the board that has 9 tiles of
the same size and same color. You may find the following constants useful:
(define TILE-LEN 100)
(define TILE-CLR 'green)
(define WIDTH (* 3 TILE-LEN))
(define HEIGHT (* 3 TILE-LEN))
(define E-SCENE (empty-scene WIDTH HEIGHT))
428 18 Mutually Recursive Data

How can we represent the world? We can take advantage of the work we have
done in Sect. 107.1. As the game advances, we are stepping through the st that
has the initial board at the root. For a given st, the next board must be one of the
children. This child becomes the world’s st after each move. In addition, we need
to track the current turn to determine if the player is allowed to make a move. This
analysis leads the following data definition:
;; A world is a structure: (make-world st turn).
(define-struct world (st turn))

* Ex. 201 — Develop the template for a function on a world.

* Ex. 202 — Define, INIT-WORLD, the initial world.

108.2 Design draw-world

Drawing the world requires drawing a board and any embellishments that you like
to personalize the game to your liking (e.g., whose turn it is). What board ought to
be rendered? In accordance with our data representation, this board ought to be at
the root in the world’s st.

* Ex. 203 — Design and implement a function to draw the world.

108.3 Design process-mouse

In the universe API, a mouse event handler needs four inputs: a world, an x
coordinate, a y coordinate, and a mouse event. The given x and y coordinates
correspond to the position of the mouse in the scene when a mouse event happens.
There are several mouse events defined and you may read about them through the
Help Desk. Of interest for us is the "button-down" event. This occurs when the
player clicks. When the player clicks on a tile, an X is placed in the clicked tile if:
1. It is X’s turn.
2. The clicked box is empty (i.e., blank).
3. The mouse event is "button-down".
To do so, a new world must be created by choosing the appropriate child from the
world’s st. If any of the conditions above do not hold, then handler returns the given
world.
The above problem analysis suggests that at least two auxiliary functions are
needed. One is needed to convert the x and y coordinates into a bpos. Another is
108 Project: Tic Tac Toe 429

needed to search an st’s children for the child st that is rooted at a given board. In
all likelihood you may find it useful to use the place-on-board function left as an
exercise above.

* Ex. 204 — Create data definitions and function templates for, ttt-x and
ttt-y, valid x and y coordinates for the Tic Tac Toe game.

*** Ex. 205 — Design and implement process-mouse, a function to process

mouse events.
** Ex. 206 — Design and implement mouse-tile that takes as input ttt-x
and ttt-y values and returns the corresponding bpos corresponding to the
clicked tile.
*** Ex. 207 — Design and implement get-st that takes as input a board and
a lost and that returns the st in the given lost that has the given board at its
root.

108.4 Design process-tick

The computer ought to make a move when the clock ticks and it is O’s turn. If it is X’s
turn, then the given world ought to be returned. To make a move the program must
decide the move to make and then extract the right child st from the given world.
The natural question that arises is how does the computer make a move. Two
popular approaches to have a program make a move in a game are rule-based
moves and heuristic-based moves. In the rule-based approach there is a rule
that, given a board, decides what the next move ought to be. In the heuristic-based
approach there is a function that assigns a score to each possible board after a move
and the move with the best score is chosen. We shall pursue a rule-based approach for
our version of Tic Tac Toe. In follow-up courses, you shall study the heuristic-based
approach that is commonly used in Artificial Intelligence.
Think carefully about how you play Tic Tac Toe. It is fairly clear that if you can
win you make the winning move. Otherwise, you determine if you can block and do
so if possible. What if you cannot win or block? We need to decide what move the
program ought to make. A reasonable strategy may be to play the center box if it is
empty. Otherwise, play a blank corner. If all corners are taken, then play a middle
space. This analysis means that we have a compound function with 5 conditions:
(cond [(can-o-win? a-board) (make-o-win a-board)]
[(can-o-block? a-board) (make-o-block a-board)]
[(center-empty? a-board) (make-o-center a-board)]
[(any-corner-empty? a-aboard) (make-o-corner a-board)]
[else (make-o-mid a-board)])
430 18 Mutually Recursive Data

The moves with the highest priorities must be tested for first (i.e., win then block) to
guarantee they take precedence.

*** Ex. 208 — Design and implement a function to make a computer (i.e., O)
move based on the rule outlined above.
*** Ex. 209 — Design and implement a function to process a clock tick.

***** Ex. 210 — The rule to make a computer move outlined above does not
consider creating or preventing the player from creating a fork. A fork is a
board in which there are two winning moves. Redesign your function to make
a computer move to take into account forking.

108.5 Design game-over?

The game is over if either player has won or if the board is full. A player has won if
there are three of the same symbols on any row, column, or diagonal. A board is full
if there are no 'Bs left in the board.
The function to draw the last world may simply be draw-world. A better design
indicates the winning player or a draw.

** Ex. 211 — Design and implement a function to detect when the game is
over.
* Ex. 212 — Design and implement a function to draw the last world by indi-
cating the winner or a draw.

109 What Have We Learned in This Chapter?

The important lessons of Chap. 18 are summarized as follows:

• Data definitions that refer to each other are said to be mutually recursive.
• In a data definition, references to a different type of data, X, mean that in our code
there needs to be a call to an X-processing function.
• For mutually recursive data, build nonrecursive instances first and then build
instances that only depend on existing instances.
• To write a function for mutually recursive data, write a function for each data type
assuming the functions to process the other data types exist.
• In a tree each node has an arbitrary number of children.
• Two popular ways to have a program make a move in a game are the rule-based
approach and the heuristic-based approach.
109 What Have We Learned in This Chapter? 431

• In the rule-based approach there is a rule that, given a board, decides what the
next move ought to be.
• In the heuristic-based approach there is a function that assigns a score to each
possible board after a move and the move with the best score is chosen.
• The use of heuristic functions is common in Artificial Intelligence.
Chapter 19
Processing Multiple Inputs of Arbitrary
Size

Our focus up to know has mostly been on the design of functions that consume
a single instance of data of arbitrary size. This chapter discusses how functions
that consume multiple inputs of arbitrary size are designed. To simplify this initial
discussion we shall focus on designing functions that consume two inputs of arbitrary
size. How do you design such a function?
We outline three basic properties that you may establish once problem analysis is
performed. These are:
• One input has the dominant role.
• The two inputs must be processed simultaneously.
• There is no clear relationship between the inputs.
If one of the inputs, say of type X, has a dominant role, then the function may be
designed using the template for a function on X. You were exposed to this idea in
Sect. 61 albeit to process data of finite size. If the two inputs must be processed at
the same time, then pick one of the inputs and design the function using the template
for a function on the type of the given input. If there is not a clear relationship
between the inputs, then you need to analyze all possible combinations of the inputs’
subtypes.

110 One Input Has a Dominant Role

Imagine that ISL+ did not include a function, append, to append two given lists. If
the programming language you are using does not include a function, you need then
you must design your own. How you design and implement a function to append two
lists? How did the creators of ISL+ design the append function? We shall follow
the steps of the design recipe to determine the answer.

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 433
M. T. Morazán, Animated Problem Solving, Texts in Computer Science,
https://doi.org/10.1007/978-3-030-85091-3_19
434 19 Processing Multiple Inputs of Arbitrary Size

We always start with problem analysis. We are given, say, L1 and L2. Ask yourself
what the result of appending the two lists ought to look like. It is not difficult to see
that a new list must be created that has all the elements of L1 followed by all the
elements of L2. To add all the elements of L1 to the resulting list L1 must be traversed.
Do we need to also traverse L2? As L1 is traversed cons is used to created a list
that contains the first element of L1 and the result of appending the rest of L1 and
L2. Consider what occurs with the last element of L1. This element must be consed
with the result of appending the rest of L1, which is empty, and L2. What ought to
be the result of appending empty with L2? Since L1 (i.e., empty) has no elements,
nothing ought to appear before the elements of L2. In other words, L2 is the answer.
Observe that L2 does not need to be traversed and, therefore, L1 plays a dominant
role. This informs us that we ought to design this function around processing L1.
We shall be careful with the next step of the design recipe, data analysis, to make
sure the design is sound. The types of the two inputs may be defined as (listof
X) and (listof Y). Observe that the two inputs may or may not be of the same
type. They are the same type (not value) when X = Y. Otherwise, they are of different
types. The natural question to ask, therefore, is what is the type of the returned list?
If X  Y, then the result is neither of type (listof X) nor (listof Y). We know,
however, that every element of the result list is either of type X or Y. This means
there is variety among the elements of the result list. To capture this variety we need
the following data definition:
An XY is either:
1. X
2. Y
Armed with this data definition, we can now write the returned type as (listof
XY). Note that this means that every element of the returned list is either an X or a Y.
It does not specify that all the X elements must appear before the Y elements. Such a
specification is done by the purpose statement of the function.
We can now specialize the template for a function on X to write a function to
append two lists. Start by writing sample instances for a (listof X) and (listof
Y). For example, we may write:
;; Sample instances of (listof X), where X = number
(define LON1 '())
(define LON2 '(10 20 30 40))

;; Sample instances of (listof Y), where Y = string

(define LOS1 '())
(define LOS2 '("" "A" "BB"))
110 One Input Has a Dominant Role 435

Following the design recipe, we next develop sample expressions for myappend
based on the varieties of (listof X):
;; Sample expressions for myappend
(define LON1-LOS1-VAL LOS1)
(define LON1-LON2-VAL LON2)
(define LOS2-LOS1-VAL (cons (first LOS2)
(myappend (rest LOS2) LOS1)))
(define LON2-LOS2-VAL (cons (first LON2)
(myappend (rest LON2) LOS2)))
Observe that we test every possible combination of inputs’ subtypes. As per our
problem analysis if the first given list is empty then the answer is the second given
list. Otherwise, a new list is created by consing the first element of the first given
list to the result of appending the rest of the first given list and the second given list.
Abstracting over the sample expressions yields the following function:
;; (listof X) (listof Y) → (listof XY)
;; Purpose: Create a new list with members of the first given
;; list followed by members of the second given list
(define (myappend L1 L2)
(if (empty? L1)
L2
(cons (first L1) (myappend (rest L1) L2))))
Observe that the signature states the return type as the one we developed as part of
data analysis. Further observe that the purpose statement specifies that a new list is
created with the elements of L1 followed by the elements of L2.
Finally, test may be written as follows:
;; Tests using sample computations for myappend
(check-expect (myappend LON1 LOS1) LON1-LOS1-VAL)
(check-expect (myappend LON1 LON2) LON1-LON2-VAL)
(check-expect (myappend LOS2 LOS1) LOS2-LOS1-VAL)
(check-expect (myappend LON2 LOS2) LON2-LOS2-VAL)

;; Tests using sample values for myappend

(check-expect (myappend '(#true #false) '(a b c))
'(#true #false a b c))
(check-expect (myappend '(M) '(T M)) '(M T M))
Observe that the tests using sample values illustrate that the function works for lists
of different sizes that both have and do not have the same types of elements. As
always, the tests using sample computations illustrate that the function correctly
computes the values of the sample expressions.
436 19 Processing Multiple Inputs of Arbitrary Size

*** Ex. 213 — Design and implement a function that takes as input a natural
number, n, and a (listof X), L, and that returns a list with n copies of L.

*** Ex. 214 — Design and implement a predicate to determine if all the ele-
ments of a given list are members of a second given list. You may find the ISL+
function member? useful.
**** Ex. 215 — Design and implement a function that takes as input a (btof
X), btx, and a (listof X), L, of length 2 and that returns btx with all
occurrences of the first element of L substituted with the second element of L.
*** Ex. 216 — Design and implement a function that takes as input two natural
numbers and that returns the product of the two numbers. Your function may
not use * and must use +. Recall, for example, that 4 * 3 = 4 + 4 + 4.

111 Inputs Must Be Processed Simultaneously

Consider the problem of multiplying corresponding numbers in two lists of numbers

and returning a list of the products. This operation is useful to solve the following
problems:
1. Compute employees’ gross pay from a list of hours worked and a list of hourly
rates. The ith number of hours worked is paid using the ith hourly rate.
2. Compute the total inventory value of the different items in a store from a list of
item quantities and a list of item prices. Each item in the ith quantity has the ith
price.
Observe that for both problems the ith elements of each list must be multiplied to
compute a list of products. Further observe that both lists must be of the same size.
For this type of problem, we cannot process one of the lists and simply return the
other list when the first list is processed. In other words, neither list is dominant. Note
that both lists must be processed at the same time. As the lists are traversed, the first
element of both lists is multiplied and consed to the result of processing the rest of
both lists. That is, the processing of the lists is synchronized. When the processing of
two instances of data of arbitrary size is synchronized, the function may be designed
using the template for functions on the type of either input. There is, however, a
significant difference. The recursive call must be made using the selector for the
substructure of both inputs. For example, the recursive call for the synchronized
processing of two list of numbers, L1 and L2, is made with (rest L1) and (rest
L2).
111 Inputs Must Be Processed Simultaneously 437

Given that to multiply corresponding numbers the two lists must be of the same
size, the lists are either both empty or they are both not empty. What is the answer
if the lists are empty? There are no corresponding numbers to multiply. Therefore,
the result must be the empty list. If the two lists are not empty, then this is when the
product of first element of each list is consed to the result of multiplying the rest of
both lists. The function may be designed by specializing the template for a function
on a (listof number).
Based on our problem analysis we can write sample instances and sample expres-
sions as follows:
;; Sample instances of lon
(define ELON '())
(define HOURS-WORKED '(1 23 39 27))
(define HOURLY-RATES '(13 22 18 34))
(define QUANTITIES '(10 3 86 27 8))
(define PRICES '(80 50 5 10 20))

;; Sample expressions for mlist

(define NOTHING '())
(define GROSSPAY
(cons (* (first HOURS-WORKED) (first HOURLY-RATES))
(mlist (rest HOURS-WORKED) (rest HOURLY-RATES))))

(define TOTALS
(cons (* (first QUANTITIES) (first PRICES))
(mlist (rest QUANTITIES) (rest PRICES))))
The first sample expression illustrates that the result is empty when the given lists
are empty. The two other sample expressions illustrate that for non-empty lists a new
list is constructed using the product of the first element of each list and recursively
processing the rest of both lists.
Abstracting over the sample expressions yields the following function:
;; (listof num) (listof num) → (listof num)
;; Purpose: Return a list with the products of corresponding
;; elements in the given lists
;; Assumption: The given lists are of the same length
(define (mlist L1 L2)
(if (empty? L1)
'()
(cons (* (first L1) (first L2))
(mlist (rest L1) (rest L2)))))
Observe that the assumption that the two lists are of the same size is clearly indicated.
This informs any reader of the code that the function is specifically designed for two
lists of the same size.
438 19 Processing Multiple Inputs of Arbitrary Size

The function may be tested as follows:

;; Tests using sample computations for mlist
(check-expect (mlist ELON '()) NOTHING)
(check-expect (mlist HOURS-WORKED HOURLY-RATES) GROSSPAY)
(check-expect (mlist QUANTITIES PRICES) TOTALS)

;; Tests using sample values for mlist

(check-expect (mlist '(1 2 3) '(1 2 3)) '(1 4 9))
In addition to illustrating that the function computes the expected values, the second
and third tests using sample computations illustrate some of the practical purposes
of mlist: computing gross pay and inventory value. The test using a sample value
illustrates how to compute the squares of a list of numbers.

* Ex. 217 — Design and implement a function that takes as input two (listof
number) and that returns a (listof number) containing the sums of corre-
sponding numbers in the given list.

** Ex. 218 — Redesign and re-implement mlist by removing the assumption

that both lists are of the same size. The function ought to terminate when either
list is empty.

** Ex. 219 — Design and implement a function that takes as input a list of first
names and a list of last names and that returns a list of full names.
**** Ex. 220 — Design and implement a function that takes as input two
(listof Boolean) and that returns a (listof Boolean) indicating if the
corresponding elements in the given lists are the same.

**** Ex. 221 — Design and implement a function that takes as input two
(listof number) and that returns an intertwined list of numbers. For example,
given '(1 3 5) and '(2 4 6) the function returns '(1 2 3 4 5 6).

112 No Clear Relationship Between the Inputs

It is not uncommon to have more than one complex input and not be able to identify
a relationship between the inputs. That is, neither input dominates the other and
the processing of the inputs is not synchronized. In such cases, it is necessary to
analyze each possible combination of the inputs’ subtypes. A conditional expression
is needed in which the number of conditions is equal to the number of possible input
subtype combinations. Each stanza in the conditional expression tests for a specific
combination of subtypes. For instance, consider a function that processes a (listof
X) and a natural number. Each of these types has two subtypes. Thus, there are
112 No Clear Relationship Between the Inputs 439

four possible subtype combinations. The needed conditional expression identifies

the combination that the inputs represent. Assuming the list is a-lox and the natural
number is a-natnum, the needed conditional expression for the function called f
may be outlined as follows:
(cond [(and (empty? a-lox) (zero? a-natnum)) . . .]
[(and (empty? a-lox) (not (zero? a-natnum))) . . .]
[(and (cons? a-lox) (zero? a-natnum)) . . .]
[else . . .(f a-lox (sub1 n)). . .
. . .(f (rest a-lox) n). . .
. . .(f (rest a-lox) (sub1 n)). . .])
Observe that this is a divide and conquer strategy. The problem of processing
(listof X) and a natural number is divided into four smaller subproblems. The
solution for each subproblem is developed independently. In the case when a-lox
is not empty and a-natnum is not zero the recursive call to f may be done with the
substructure of either input or the substructure of both inputs. This is why the else
clause of the conditional has three different expressions that may be used for the
recursive call. As a designer you must decide which of these expressions to use (for
any recursive call in f). It is important to note that this design strategy may always
be used when processing more than one input of arbitrary size.
Consider the problem of merging two lists of numbers sorted in non-decreasing
order. The goal is to create a sorted list that only contains all the elements of the
given lists. For example, given '(15 31 67) and '(22 44 87 100) the resulting list is
'(15 22 31 44 67 87 100). Ask yourself if one list dominates the other. Clearly, this is
not the case because the next element to add to the result list may come from either
of the given lists. That is, the first element of the result list may be the first element
of the first given list or the first element of the second given list. Now, we know that
both lists must be processed. Ask you yourself if the processing of the lists must be
synchronized. Once again, the answer is no because the first element of the result
may come from either given list.
We are left to conclude that there is no clear relationship between the inputs. This
means processing the inputs by determining the subtype combination of the given
inputs. If the function is called merge and the inputs are called sl1 and sl2, then
the needed conditional expression is outlined as follows:
(cond [(and (empty? sl1) (empty? sl2)) . . .]
[(and (empty? sl1) (cons? sl2)) . . .]
[(and (cons? sl1) (empty? sl2)) . . .]
[else . . .(merge sl1 (rest sl2)). . .
. . .(merge (rest sl1) sl2). . .
. . .(merge (rest sl1) (rest sl2)) . . .])])
We can now reason about each case in the conditional independently. If both
given lists are empty, then there is nothing to add to the result and the answer is
'(). If only one of the lists is empty, then only the elements in the other list must
be added. Therefore, the answer is the other list. If both lists are not empty, then a
440 19 Processing Multiple Inputs of Arbitrary Size

decision must be made to either create a new list using sl1’s or sl2’s first element.
The element chosen must be the smaller of the two. The recursive call, therefore, is
made with the rest of the list whose first element is added to the result and the other
list.
Let us now use the following instances of a (listof number) to develop our
design and implementation:
;; Sample instances of (listof number)
(define ELON '())
(define SL1 '(1 22 30))
(define SL2 '(13 22 108 346))
(define SL3 '(-89 -50 0 6 90))
(define SL4 '(-240))
We start by writing sample expressions for each of the possible subtype combinations.
If both given lists are empty, the answer is the empty list. We define a sample
expression as follows:
;; Sample expressions for merge
(define ELON-ELON-VAL '())
If only one of the given lists is empty, then the answer is the other list. We define
sample expressions as follows:
(define ELON-SL1-VAL SL1)
(define ELON-SL2-VAL SL2)
(define SL3-ELON-VAL SL3)
(define SL4-ELON-VAL SL4)
The first two sample expressions illustrate the answer when the first given list is
empty. The second two sample expressions illustrate the answer when the second
given list is empty. For the final combination (i.e., when both lists are not empty) we
need examples that make the first element of the result the first element of the first
given list and examples that make the first element of the result the first element of
the second given list. We may write the following sample expressions:
(define SL1-SL2-VAL (cons (first SL1) (merge (rest SL1) SL2)))
(define SL3-SL2-VAL (cons (first SL3) (merge (rest SL3) SL2)))
(define SL3-SL4-VAL (cons (first SL4) (merge SL3 (rest SL4))))
(define SL1-SL4-VAL (cons (first SL4) (merge SL1 (rest SL4))))
The first two sample expressions take the first element of the first given list and
the second two sample expressions take the first element of the second given list.
Observe that the recursive call is always made with the rest of the list whose first
element is added to the result and with the other list untouched. It ought to be clear
that the processing of the lists is not synchronized.
112 No Clear Relationship Between the Inputs 441

To write the function definition we abstract over the sample expressions for each
subtype combination. This process yields the following function:
;; (listof num) (listof num) → (listof num)
;; Purpose: Return a list sorted in nondecreasing order that
;; only contains all the elements of the given lists
;; Assumption: Given lists are sorted in nondecreasing order
(define (merge sl1 sl2)
(cond [(and (empty? sl1) (empty? sl2)) '()]
[(and (empty? sl1) (cons? sl2)) sl2]
[(and (cons? sl1) (empty? sl2)) sl1]
[else (if (<= (first sl1) (first sl2))
(cons (first sl1) (merge (rest sl1) sl2))
(cons (first sl2) (merge sl1 (rest sl2))))]))
Observe that when both given lists are not empty a conditional expression is used
to determine from which list to take the first of the result. Further observe that the
recursive call is always done with the rest of the list from which the first element is
selected and the other list.
All conditions must be tested including the two in the else stanza of the cond-
expression. The tests may be written as follows:
;; Tests using sample computations for merge
(check-expect (merge ELON ELON) ELON-ELON-VAL)
(check-expect (merge ELON SL1) ELON-SL1-VAL)
(check-expect (merge ELON SL2) ELON-SL2-VAL)
(check-expect (merge SL3 ELON) SL3-ELON-VAL)
(check-expect (merge SL4 ELON) SL4-ELON-VAL)
(check-expect (merge SL1 SL2) SL1-SL2-VAL)
(check-expect (merge SL3 SL2) SL3-SL2-VAL)
(check-expect (merge SL3 SL4) SL3-SL4-VAL)
(check-expect (merge SL1 SL4) SL1-SL4-VAL)

;; Tests using sample values for merge

(check-expect (merge '() '()) '())
(check-expect (merge '() '(8 77)) '(8 77))
(check-expect (merge '(43 67 91) '()) '(43 67 91))
(check-expect (merge '(15 31 67) '(22 44 87 100))
'(15 22 31 44 67 87 100))
(check-expect (merge '(66 99) '(1 56 83)) '(1 56 66 83 99))

The tests using sample computations are organized in the same manner as the
conditional expression in merge. The first test is for when both given lists are empty.
The second and third tests are for when only the first given list is empty. The next
two tests are for when only the second given list is empty. The final four tests are
for when both given lists are not empty. Of these, the first two are for when the first
given list’s first element is used to construct the result. The last two are for when
the second given list’s first element is used to construct the result. The tests using
sample values are organized in the same order, but there is only a single test for each
possible condition.
442 19 Processing Multiple Inputs of Arbitrary Size

*** Ex. 222 — Design and implement a function that extracts the nth element
of a list. The elements of the list are numbered 0 to n-1. If there is no nth
element, then the function ought to throw an error. You may not use list-ref
in your solution.

* Ex. 223 — Design and implement a function that merges two lists of numbers
sorted in non-decreasing order.

*** Ex. 224 — Design and implement a function that consumes two (listof
Boolean) and intertwines them based on the first element of each list. If the
first elements are the same, then put the first element of the first list in the result.
Otherwise, put the first element of the second list in the result. If either given
input is empty, nothing more is added to the result. For example, given '(#true
#true #true #false #false) and '(#true #false #false) the result
is '(#true #true #true #true #false #false).
*** Ex. 225 — Design and implement a function that takes as input two
(listof number) and a Boolean and that returns an intertwined list of num-
bers. For example, given '(1 3 5) and '(2 4 6) the function returns '(1 2 3 4 5
6). The Boolean input is used to decide from which given list to add its first
element to the result. If the Boolean is true, take the first element from the first
list. Otherwise, take the first element from the second list. Do you prefer this
version of the function over the version that synchronizes the processing of the
lists? Why or why not?

113 What Have We Learned in This Chapter?

The important lessons of Chap. 19 are summarized as follows:

• Three common scenarios that may be identified when processing multiple in-
stances of data of arbitrary size are: one input dominates the other, the processing
of the inputs must be synchronized, and there is no clear relationship between the
inputs.
• If one input of type X dominates the other inputs, then specialize the template for
a function on X.
• If processing inputs of type X and Y must be synchronized, specialize either the
template for a function on X or the template for a function on Y, but make recursive
calls using the substructure of both inputs.
• If there is no clear relationship between the inputs, use a conditional to determine
which combination of input subtypes is faced and independently formulate the
answer for each possible subtype combination.
• Subtype combination analysis may always be used when processing more than
one input of arbitrary size.
• Sometimes a data definition for the type returned by a function must be defined.
 Part IV
Abstraction
Chapter 20
Functional Abstraction

Chapter 3 discusses how to abstract over similar expressions to eliminate repetitions,

which leads to the development of functions. Section 68 discusses how to eliminate
repetitions among data definitions, which leads to the development of generic data
definitions. This chapter discusses functional abstraction. In other words, it discusses
how to eliminate repetitions among functions. This leads to powerful abstract func-
tions (or generic functions) that can perform many different jobs for us. It will also
conceptually change the meaning of data in your mind. As we shall see functions are
data just like a number, a list, and an image are data. Thus, you will soon realize that
the conceptual line between data and functions is imaginary.
If you already have code that works, why would you want to perform abstraction
to eliminate differences? The answer is twofold. On the one side there are practical
programming reasons. By eliminating repetition among functions you make your
code easier to understand and maintain, shorter, and more elegant. Collectively this
means that it becomes easier to make improvements or necessary changes in the
future. On the other side it allows you to think about solutions to problems more
abstractly, thus making problem solving and programming easier.
To illustrate the process consider the functions displayed in Fig. 92. These are
functions to determine, correspondingly, if 'laptop and 'pen are contained in a
(listof symbol). Observe that the two functions are nearly identical. They only
vary in the use of the symbol 'laptop or 'pen. Recall that the body of each
function is an expression and a variable is used to abstract away differences among
expressions to create functions. This suggest the same ought to happen here. That is,
use a variable to represent the symbol that varies and create a new abstract function
that has a parameter for this varying symbol. The other parameter in the original
functions is still needed and all the parameters (including the new one for the varying

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 445
M. T. Morazán, Animated Problem Solving, Texts in Computer Science,
https://doi.org/10.1007/978-3-030-85091-3_20
446 20 Functional Abstraction

Fig. 92 Two very similar functions

;; contains-laptop?: (listof symbol) Boolean
;; Purpose: Determine if the given list contains laptop
(define (contains-laptop? a-los)
(and (not (empty a-los))
(or (symbol=? (first a-los) laptop)
(contains-laptop? (rest a-los)))))

;; contains-pen?: (listof symbol) Boolean

;; Purpose: Determine if the given list contains pen
(define (contains-pen? a-los)
(and (not (empty a-los))
(or (symbol=? (first a-los) pen)
(contains-pen? (rest a-los)))))

symbol) are used in the body of the abstract function. This abstraction step yields
the following function:
;; contains?: symbol (listof symbol) → Boolean
;; Purpose: Determine if the given list contains the
;; given symbol
(define (contains? a-symbol a-los)
(and (not (empty a-los))
(or (symbol=? (first a-los) a-symbol)
(contains? a-symbol (rest a-los)))))
We can now use the abstract function to refactor contains-laptop? and
contains-pen? as follows:
;; contains-laptop?: (listof symbol) → Boolean
;; Purpose: Determine if the given list contains 'laptop
(define (contains-laptop? a-los) (contains? 'laptop a-los))

;; contains-pen?: (listof symbol) → Boolean

;; Purpose: Determine if the given list contains 'pen
(define (contains-pen? a-los) (contains? 'pen a-los))

Observe how much shorter and readable these functions are now. Furthermore,
observe that now there is only one recursive function instead of two. The benefit of
the abstract function goes even further. It can be used to write functions to determine
if a (listof symbol) contains 'book, 'keys, and 'money as follows:
;; contains-book?: (listof symbol) → Boolean
;; Purpose: Determine if the given list contains 'book
(define (contains-book? a-los) (contains? 'book a-los))

;; contains-keys?: (listof symbol) → Boolean

;; Purpose: Determine if the given list contains 'keys
(define (contains-keys? a-los) (contains? 'keys a-los))
115 Functions as Values 447

;; contains-money?: (listof symbol) → Boolean

;; Purpose: Determine if the given list contains 'money
(define (contains-money? a-los) (contains? 'money a-los))
This illustrates the power of abstract functions. They can perform many different tasks
that allows us to simplify code design and development. Using abstract functions
means that we need to write less code and may have fewer bugs.

** Ex. 226 — Write tests for contains?.

* Ex. 227 — Write tests for contains-laptop?, contains-pen?,
contains-book?, contains-keys?, and contains-money?.

114 A Design Recipe for Abstraction

The goal for functional abstraction is to abstract over similar expressions in the body
of different functions with a similar number of parameters. The process followed to
abstract over contains-laptop? and contains-pen? may be generalized into the
following design recipe for functional abstraction:
Mark Differences Compare and mark the differences in the bodies of the functions.
Create the Abstraction Define an abstract function that has the same number of
parameters as the functions abstracted over plus a parameter for every difference
identified in the previous step. The body of the abstract function references the
parameters for the differences instead of the values found in the functions ab-
stracted over. Make sure the signature takes into account that the new parameters’
types may vary making the function generic.
Refactor Refactor the functions abstracted over to use the abstract function.
The first step asks you to identify the significant differences in the expressions
that constitute the bodies of the functions. This means you ought to identify different
values. It does not mean differences in function or variable names. The second step
asks you to define the abstract function. This requires you to pick a descriptive name
for the function. The parameter list must include a parameter for each difference and
a set of parameters representing the parameters in the functions abstracted over. The
third step asks you to refactor the bodies of the functions abstracted over to use the
new abstract function. The tests for these refactored functions unchanged.

115 Functions as Values

To illustrate the design recipe for abstraction in practice consider computing the
slope of a secant line. A secant line crosses the graph of a function, f(x), at exactly
two points: (x1 , f(x1 )) and (x2 , f(x2 )). When the distance between x1 and x2
448 20 Functional Abstraction

Fig. 93 Functions for the slope of the secant line

;; number number number
;; Purpose: Compute slope of secant line for given x-values for f(x) = x^2
(define (x^2-sec-slope x1 x2) (/ (- (sqr x2) (sqr x1)) (- x2 x1)))

;; Sample expressions for x^2-sec-slope

(define MSEC1 (/ (- (sqr 4) (sqr 2)) (- 4 2)))
(define MSEC2 (/ (- (sqr 3.75) (sqr -2.03)) (- 3.75 -2.03)))

;; Tests using sample computations for x^2-sec-slope

(check-within (x^2-sec-slope 2 4) MSEC1 0.01)
(check-within (x^2-sec-slope -2.03 3.75) MSEC2 0.01)

;; Tests using sample values for x^2-sec-slope

(check-within (x^2-sec-slope 1.5 3.2) 4.7 0.01)

;; number number number

;; Purpose: Compute slope of the secant line for given x-values
;; for f(x)= x^3
(define (x^3-sec-slope x1 x2) (/ (- (cube x2) (cube x1)) (- x2 x1)))

;; Sample expressions for x^2-sec-slope

(define MSEC3 (/ (- (cube 5) (cube 1)) (- 5 1)))
(define MSEC4 (/ (- (cube 10.1) (cube -4.6)) (- 10.1 -4.6)))

;; Tests using sample computations for x^3-sec-slope

(check-within (x^3-sec-slope 1 5) MSEC3 0.01)
(check-within (x^3-sec-slope -4.6 10.1) MSEC4 0.01)

;; Tests using sample values for x^3-sec-slope

(check-within (x^3-sec-slope 1.5 3.2) 17.29 0.01)

is very small, the slope of the secant line may be used as an approximation for the
slope of f(x) at the midpoint between the two points. The slope of a (secant) line is
fairly easy to compute as you learned in your high school algebra course:
f x2 − f x1
m = x2 − x1
This formula informs us that to compute the slope of the secant line we need two
x-values. Figure 93 displays two functions, xˆ2-sec-slope and xˆ3-sec-slope,
to compute the slope of the secant line, respectively, for f(x) = x2 and f(x) =
x3 .26
Observe that xˆ2-sec-slope and xˆ3-sec-slope are very similar and, there-
fore, good candidates for abstraction. Following the steps of the design recipe for
abstraction we find that the only significant difference is the function that is ap-
plied to an x-value. In xˆ2-sec-slope sqr is used and in xˆ3-sec-slope cube
is used. There is something new to learn here. The differences are functions. This
means that functions must be data just like a natural number and a posn are data.

26
The design of a function to cube its input is found in Sect. 38.
115 Functions as Values 449

In programming languages that provide support for functions as data we say that
the language has first-class functions. Among other things this means that functions
may be passed as arguments to a function. ISL+ is one such language and, therefore,
we may proceed with the process of functional abstraction.
The next step of the design recipe is to create an abstract function. For our current
problem this means that the abstract function needs a parameter for the single
difference, say f, and two parameters for the two x-values (as xˆ2-sec-slope and
xˆ3-sec-slope have). To write this function, however, we need to denote a function
in the signature. This is accomplished by writing a signature for the type of function
the parameter denotes. In this case both f(x) = x2 and f(x) = x3 have the same
signature: number number → number. The result of this step of the design recipe
for abstraction is
;; (number → number) number number → number
;; Purpose: Approximate the slope of the secant line between
;; the given x-values for the given function
(define (f-sec-slope f x1 x2) (/ (- (f x2) (f x1)) (- x2 x1)))
Observe that parentheses are placed around function values in the signature to aid
readability. It makes it clear that f-sec-slope returns a number. The body of the
function looks the same as the bodies of xˆ2-sec-slope and xˆ3-sec-slope
except that references to sqr and cube are now references to the parameter f.
The third step of the design recipe has refactor xˆ2-sec-slope and
xˆ3-sec-slope to use f-sec-slope. This means, in essence, that each of these
functions must call f-sec-slope with the right arguments. In practical terms, this
means xˆ2-sec-slope must make the call with sqr and the x-values it gets as input
and xˆ3-sec-slope must make the call with cube and the x-values it gets as input.
The result of this step is
;; number number → number
;; Purpose: Approximate the slope of the secant line between
;; the given x-values for f(x) = x^2
(define (x^2-secant-slope x1 x2) (f-sec-slope sqr x1 x2))

;; number number → number

;; Purpose: Approximate the slope of the secant line between
;; the given x-values for f(x) = x^3
(define (x^3-secant-slope x1 x2) (f-sec-slope cube x1 x2))
Observe that nothing changes except the bodies of the functions. This includes the
tests for both functions. Go ahead and run the tests to make sure they all pass.
By now you may find it strange that this discussion has ignored writing tests for
abstract functions. There is a good reason for this. When an abstract function returns
a value that is not a function we can write tests for it. However, we do not for two
reasons. The first is the refactoring performed on the functions that are abstracted.
These refactored functions use the abstract function and have tests. If the testing is
thorough, bugs in the abstract function will be caught by the existing tests. More
450 20 Functional Abstraction

importantly, however, is the fact that first-class functions push programmers against
the limits of what is possible. Chapter 22 discusses how a function may be returned
as the value of a function just like a string or a number may be returned. To directly
test such a function the returned function and another function must be proven
equivalent. As you will learn in a Formal Languages and Automata Theory course
or a Computability course, testing the equivalence of functions is an unsolvable
problem. That is, there is no solution to the problem of determining if two functions
are equivalent. It is for this reason that abstract functions are tested through the
functions that make use of them.

116 Abstraction Over List-Processing Functions

Chapter 13 discusses several common operations over lists. If you think about it for
a minute, it is not unreasonable to expect functions that perform the same operation
for different types of lists to be almost identical. If this is the case, then they are
good candidates for abstraction. Sometimes, however, the expressions in the bodies
of these functions may not be the same. In such cases code refactoring may be used
to make the functions almost identical before performing functional abstraction.
This section presents the abstraction over functions that perform common list-
processing operations. Specifically, this section looks at the operations outlined in
Chap. 13: list summarizing, list searching, list ORing, list ANDing, list mapping, list
filtering, and list sorting.

116.1 List Summarizing

In Sect. 72 the following functions to sum and to compute the length of a list of quiz
grades are developed:
;; (listof quizgrade) → number
;; Purpose: To compute the sum of the given
;; (listof quizgrade)
(define (sum-loq a-loq)
(if (empty? a-loq)
0
(+ (first a-loq) (sum-loq (rest a-loq)))))

;; (listof quizgrade) → number

;; Purpose: To compute the length of the given
;; (listof quizgrade)
(define (length-loq a-loq)
(if (empty? a-loq)
0
(+ 1 (length-loq (rest a-loq)))))
116 Abstraction Over List-Processing Functions 451

In the same section two different exercises ask you to design and implement a
function to compute the product of a list of numbers and a function to append the
strings in a list of strings. Your solutions likely look like this:
;; (listof number) → number
;; Purpose: Compute the product of the given (listof number)
(define (product-lon a-lon)
(if (empty? a-lon)
1
(* (first a-lon) (product-lon (rest a-lon)))))

;; (listof string) → string

;; Purpose: Append the strings in the given (listof string)
(define (append-los a-lostr)
(if (empty? a-lostr)
""
(string-append (first a-lostr)
(append-los (rest a-lostr)))))
We can immediately observe that all four functions are very similar, but not struc-
turally identical. Observe that only length-loq does not explicitly use the first
element of the list. Instead of having an application expression involving the first list
element it has the constant expression 1.
This is a situation where code refactoring may be useful to create similarities.
How can length-loq use its first element? At first glance it may seem that it
cannot. If you think about it carefully, however, the first element of the list given to
length-loq always contributes 1 to its length. This represents a constant function
like f(x) = 1. Therefore, length-loq may be refactored to
;; X → 1
;; Purpose: Return 1
(define (constant-f1 an-x) 1)

;; loq → number
;; Purpose: To compute the length of the given loq
(define (length-loq a-loq)
(if (empty? a-loq)
0
(+ (constant-f1 (first a-loq))
(length-loq (rest a-loq)))))
Now, length-loq uses its first element. It is still, however, different from the other
3 functions. It applies a function to its first element and the other 3 functions use the
first element directly.
To perform functional abstraction we need all 4 functions abstracted over to apply
a function to the first element of the given list. That is, sum-loq, product-lon, and
append-los need to apply a function to their first list element. This function must
452 20 Functional Abstraction

return the value it gets as input. In Mathematics, such a function is known as the
identity function: f(x) = x. We can implement the identity function and refactor
the three functions as follows:
;; X → X
;; Purpose: Return the given input
(define (id an-x) an-x)

;; (listof quizgrade) → number

;; Purpose: Compute the sum of the given
;; (listof quizgrade)
(define (sum-loq a-loq)
(if (empty? a-loq)
0
(+ (id (first a-loq)) (sum-loq (rest a-loq)))))

;; (listof number) → number

;; Purpose: Compute the product of the given
;; (listof number)
(define (product-lon a-lon)
(if (empty? a-lon)
1
(* (id (first a-lon)) (product-lon (rest a-lon)))))

;; (listof string) → string

;; Purpose: Append the strings in the given
;; (listof string)
(define (append-los a-lostr)
(if (empty? a-lostr)
""
(string-append (id (first a-lostr))
(append-los (rest a-lostr)))))
All four functions now look almost identical. That is, their structure is the same
because they use the same types of expressions. This means they are good candidates
for abstraction.
The first step is to mark the differences. There are three differences among the four
functions: the base value returned when the given list is empty, the function applied to
the first element of the list, and the function used to combine the value obtained from
the first element of the given list and the value obtained from recursively processing
the rest of the given list.
116 Abstraction Over List-Processing Functions 453

Armed with a clear understanding of the differences we may sketch the abstract
function as follows:
;; ? (? → ?) (? ? → ?) (listof ?) → ?
;; Purpose: Summarize given list
(define (accum base-val ffirst comb a-lox)
(if (empty? a-lox)
base-val
(comb (ffirst (first a-lox))
(accum base-val ffirst comb (rest a-lox)))))
Observe that the abstract function preserves the structure of the functions abstracted
over. That is, it uses the same type of expressions in its body. The question that
remains is the signature of the abstract function. The types of lists processed by the
functions abstracted over vary. Let us denote the type of list processed by the abstract
as (listof X). This yields the following partial signature:
? (? → ?) (? ? → ?) (listof X) → ?
If the elements of the given list are of type X, then the input to the function applied
to the first element of the list must be also of type X. Therefore, the partial signature
now is
? (X → ?) (? ? → ?) (listof X) → ?
Observe that the returned type of the functions abstracted over varies. Let us denoted
this type as Z making the partial signature:
? (X → ?) (? ? → ?) (listof X) → Z
Whenever the abstract returns a value, it must be of type Z. This means that the
base value and the value returned by the combining function must also be Z. These
observations allow us to make the partial signature:
Z (X → ?) (? ? → Z) (listof X) → Z
Observe that the result of the recursive call in the abstract function is the second
input to the combining function. Thus, the partial signature now is
Z (X → ?) (? Z → Z) (listof X) → Z
Observe that the type of value returned by the function applied to the first element
of the list varies in the functions abstracted over. Furthermore, this value is the first
input to the combining function. If we denote its type as Y, the signature for accum
is
Z (X → Y) (Y Z → Z) (listof X) → Z
Finally, note that the signature defines a generic function. That is, the types X, Y,
and Z are not known until the function receives its input. Just like functions have
parameters signatures also have parameters. For functions the parameters must be
declared. We shall adopt the same policy for signatures. To declare type parameters
454 20 Functional Abstraction

for signatures we shall write the type variables inside angled brackets before the
signature. The signature for accum is
<X Y Z> Z (X → Y) (Y Z → Z) (listof X) → Z
This signature informs any reader of our code that the types represented by X, Y, and
Z only become known when arguments are provided to accum.
The final step is to refactor the functions abstracted over to use the abstract
function. The result of this step is
;; (listof quizgrade) → number
;; Purpose: To compute the length of the given
;; (listof quizgrade)
(define (length-loq a-loq) (accum 0 constant-f1 + a-loq))

;; (listof quizgrade) → number

;; Purpose: To compute the sum of the given
;; (listof quizgrade)
(define (sum-loq a-loq) (accum 0 id + a-loq))

;; (listof number) → number

;; Purpose: To compute the product of the given
;; (listof number)
(define (product-lon a-lon) (accum 1 id * a-lon))

;; (listof string) → string

;; Purpose: To append the strings in the given
;; (listof string)
(define (append-los a-lostr)
(accum "" id string-append a-lostr))
Observe that the abstract function is always given as input the values used in the
original functions. Is the abstract function properly used? At the very least the
arguments provided to accum must satisfy accum’s signature. Let us consider the
call to accum in product-lon. These are the types of the arguments:
base-val: number ffirst: (number → number)
comb: (number number → number) a-lox: (listof number)
If we let X = Y = Z = number, we may rewrite the types of the arguments as follows:
base-val: Z ffirst: (X → Y)
comb: (Y Z → Z) a-lox: (listof X)
Observe that the types of the arguments satisfy the types expected by accum’s
signature. In accordance with this new convention, the proper way to write id is
;; <X> X → X
;; Purpose: Return the given input
(define (id an-x) an-x)
116 Abstraction Over List-Processing Functions 455

Fig. 94 A function to extract the X-values from a (listof posn)

;; Sample instances of (listof posn)
(define ELOP ())
(define LOP1 (list (make-posn 1 9) (make-posn 2 8)
(make-posn 3 7) (make-posn 4 6)))

;; (listof posn) lon

;; Purpose: To extract the x-values in the given list of posns
(define (get-xs-lop a-lop) (accum () posn-x cons a-lop))

;; Sample expressions for get-xs-lop

(define ELOP-VAL (accum () posn-x cons ELOP))
(define LOP1-VAL (accum () posn-x cons LOP1))

;; Tests using sample computations for get-xs-lop

(check-expect (get-xs-lop ELOP) ELOP-VAL)
(check-expect (get-xs-lop LOP1) LOP1-VAL)

;; Tests using sample values for get-xs-lop

(check-expect (get-xs-lop (list (make-posn -10 2) (make-posn -5 7)))
(-10 -5))

Take time to appreciate what has been achieved by using abstraction. The most
obvious achievement is that instead of having 4 recursive functions there is only
one recursive function (i.e., the abstract function). This makes the code shorter and
more elegant. Elegance in this case refers to using a common function to accumulate
(thus the name accum) the summarizing value being computed. It is also noteworthy
that tests do not have to be rewritten for the functions abstracted over. Go ahead
and run your tests to verify that they all pass. After refactoring, however, the sample
expressions no longer reflect the expressions abstracted over to write the bodies of
the functions. We are now using a new way (i.e., the abstract function) to compute
the same value.
The most important achievement, however, is that we now have a powerful
function to perform list-summarizing computations that may be directly used to
solve other problems. Figure 94 displays the solution to extracting the x-values
from a (listof posn). Observe that the sample expressions use accum. That is,
get-xs-lop is directly designed using our abstract function. There is no need to
design a recursive function because the needed recursion is done by accum.

* Ex. 228 — Design and implement a nonrecursive function to extract the y-

values of a (listof posn).

*** Ex. 229 — Design and implement a nonrecursive function to make a copy
of a given (listof X).

*** Ex. 230 — Design and implement a nonrecursive function to find the max-
imum of a non-empty list of natural numbers.
456 20 Functional Abstraction

**** Ex. 231 — Design and implement a nonrecursive function to count the
number of strings that have a length less than 5 in a given list of strings.

**** Ex. 232 — Design and implement a nonrecursive function that only uses
accum as an auxiliary function to compute the average of a given non-empty
list of numbers.
*** Ex. 233 — Argue that the calls to accum in length-loq, sum-loq, and
append-lostr satisfy accum’s signature.

116.2 List Searching

Section 73 discusses the design of the following function:

;; number (listof number) → (listof number)
;; Purpose: To return the sublist of the given lon that starts
;; with the first instance of given number.
(define (xsublist-lon x a-lon)
(cond [(empty? a-lon) '()]
[(equal? x (first a-lon)) a-lon]
[else (xsublist-lon x (rest a-lon))]))
It is not difficult to now design a function that returns the sublist of a given list of
symbols that starts with the first instance of a symbol that is not equal to a given
symbol. The resulting function definition may be implemented as
;; symbol (listof symbol) → (listof symbol)
;; Purpose: To return the sublist of the given list that starts
;; with the first symbol not equal to given symbol.
(define (nonxsublist-los x a-los)
(cond [(empty? a-los) '()]
[(not (equal? x (first a-los))) a-los]
[else (notxsublist-los x (rest a-los))]))
Structurally these functions are almost the same. The only structural difference is the
function composition in nonxsublist-los’ssecond condition. This means we need
to refactor the code, if possible, to make it a two-argument application expression as
in xsublist-lon.
Recall from Mathematics that (g ◦ h)(x) = g(h(x)). This means that any
expression for a function composition may be refactored into an expression that
applies a new function (namely (g ◦ h)(x)). For this problem we need to design a
function to compute the negation of the equal? function. The result of this design is
116 Abstraction Over List-Processing Functions 457

Fig. 95 The function for the complement of equals?

;; <A B> A B Boolean
;; Purpose: To determine is the given values are not equal?
(define (not-equal? v1 v2) (not (eq? v1 v2)))

;; Sample expressions for not-equal?

(define VAL1 (not (equal? 5 3)))
(define VAL2 (not (equal? (1 2) (1 2))))

;; Tests using sample computations for not-equal?

(check-expect (not-equal? 5 3) VAL1)
(check-expect (not-equal? (1 2) (1 2)) VAL2)

;; Tests using sample values for not-equal?

(check-expect (not-equal? "MATTHIAS" "SHRIRAM") #true)
(check-expect (not-equal? (make-posn (sub1 1) 10) (make-posn 0 (+ 5 5)))
#false)

displayed in Fig. 95. Armed with this function nonxsublist-lon may be refactored
to
;; symbol (listof symbol) → (listof symbol)
;; Purpose: To return the sublist of the given list that starts
;; with the first symbol not equal to given symbol.
(define (nonxsublist-los x a-los)
(cond [(empty? a-los) '()]
[(not-equal? x (first a-los)) a-los]
[else (notxsublist-los x (rest a-los))]))
Observe that the structures of xsublist-lon and nonxsublist-los are now
the same and, therefore, are good candidates for abstraction. The only difference is
the predicate used to compare the first element of the list. The abstraction step results
in
;; (? → Boolean) ? (listof ?) → (listof ?)
;; Purpose: To return the sublist of the given list that starts
;; with first element that satisfies given predicate.
(define (pred-sublist pred x a-lox)
(cond [(empty? a-lox) '()]
[(pred x (first a-lox)) a-lox]
[else (pred-sublist pred x (rest a-lox))]))
Once again, care must be taken to write the signature of the abstract function. Observe
that the list type processed varies among the functions abstracted over. If we denote
this type as (listof X), the partial signature is
<X> (? ? → Boolean) ? (listof X) → (listof ?)
458 20 Functional Abstraction

Observe that the second argument to the given predicate is a list element. This means
the partial signature becomes
<X> (? X → Boolean) ? (listof X) → (listof ?)
Now observe that in the second line of the conditional the given list is returned. This
suggests that the return type for accum is (listof X) making the partial signature:
<X> (? X → Boolean) ? (listof X) → (listof X)
Finally, observe that the type of the first argument to the functions abstracted over
varies but is always the same as the list elements’ type. Furthermore, this argument
is also the first input to the given predicate. Thus, we may conclude that the signature
of pred-sublist is
<X> (X X → Boolean) X (listof X) → (listof X)
It is now possible to refactor xsublist-lon and nonxsublist-lon to
;; number (listof number) → (listof number)
;; Purpose: To return the sublist of the given lon that starts
;; with the first instance of the given number.
(define (xsublist-lon x a-lon) (pred-sublist equal? x a-lon))

;; symbol (listof symbol) → (listof symbol)

;; Purpose: To return the sublist of the given los that starts
;; with first symbol not equal to given symbol.
(define (nonxsublist-los x a-los)
(pred-sublist not-equal? x a-los))
Once again, observe that the difference among the functions is part of the input to
the abstract function.
Always ask yourself if the arguments provided to an abstract function satisfy
the abstract function’s signature. The arguments in xsublist-lon’s call have the
following types:
pred: A B → Boolean x: number a-lox: (listof number)
This means that in the contract for pred-sublist X = number. Finally, observe that
this also means that A = B = X = number, which satisfies equal?’s signature. Based
on these observations we may conclude that the arguments provided to the abstract
function satisfy the abstract function’s signature.
Once again, take time to think about what abstraction has accomplished. We
now have a powerful function that can return the sublist that starts with the first ele-
ment that satisfies any predicate we can write. Figure 96 displays the implementation
of a function that returns the sublist of a list of strings that starts with the given string.
116 Abstraction Over List-Processing Functions 459

Fig. 96 Function to return the sublist that starts with a given string
;; string (listof string) (listof string)
;; Purpose: To return the sublist of the given (listof string) that starts
;; with the first instance of the given string.
(define (strsublist-lostr str a-lostr) (pred-sublist string=? str a-lostr))

;; Sample expressions for strsublist-lostr

(define STRSUBLIST-LOS1-VAL (pred-sublist string=? "Janice" LOS1))
(define STRSUBLIST-LOS2-VAL (pred-sublist string=? "Matthias" LOS2))

;; Tests using sample computations for strsublist-lostr

(check-expect (strsublist-lostr "Janice" LOS1) STRSUBLIST-LOS1-VAL)
(check-expect (strsublist-lostr "Matthias" LOS2) STRSUBLIST-LOS2-VAL)

;; Tests using sample values for strsublist-lostr

(check-expect (strsublist-los "Eladio" ("Rob" "Eladio" "Juan" "Eladio"))
("Eladio" "Juan" "Eladio"))
(check-expect (strsublist-los "zed" ("u" "v" "x" "y" "z" "ZED")) ())

*** Ex. 234 — Design and implement a nonrecursive function that takes as
input a number and a (listof number) sorted in non-decreasing order and
that returns the list of numbers greater than the given number.

*** Ex. 235 — Design and implement a nonrecursive function that consumes
an x-value and a (listof posn) and that returns the sublist starting with the
first posn that has an x-value greater than the given x-value.

* Ex. 236 — Design and implement a nonrecursive function that takes as input
a negative number and a (listof natnum) sorted in nonincreasing order and
that returns the (listof natnum) greater than the given negative number.

** Ex. 237 — Argue that the call to pred-sublist in nonxsublist-los

satisfies pred-sublist’s signature.

116.3 List ORing

Section 74.1 discusses the development of the following function to determine if any
alien in a (listof alien) is at the left edge:
;; (listof alien) → Boolean
;; Purpose: To determine if any alien is at scene’s left edge
(define (any-alien-at-left-edge? a-loa)
(and (not (empty? a-loa))
(or (alien-at-left-edge? (first a-loa))
(any-alien-at-left-edge? (rest a-loa)))))
460 20 Functional Abstraction

As an exercise you were also asked to write a function to determine if a given list of
numbers contains an even number. Your solution likely looked as follows:
;; (listof number) → Boolean
;; Purpose: To determine if the list of numbers contains an
;; even number
(define (has-even-lon? a-lon)
(and (not (empty? a-lon))
(or (even? (first a-lon))
(has-even-lon? (rest a-lon)))))
Observe that both functions are structurally the same and, therefore, candidates for
abstraction.
The only significant difference between the functions is the predicate that is
applied to the first element of the list. Performing the abstraction step leads to the
following function:
;; <X> (X → Boolean) (listof X) → Boolean
;; Purpose: Determine if any list element satisfies the
;; given predicate
(define (ormap-pred pred? a-lox)
(and (not (empty? a-lox))
(or (pred? (first a-lox))
(ormap-pred pred? (rest a-lox)))))
Observe the list input types for any-alien-at-left-edge? and has-even-lon?
are different. Therefore, the type of ormap-pred’s second parameter is (listof
X). Its first parameter must be a predicate that takes as input a list element of type
X and returns a Boolean. Thus, the type of ormap-pred’s first parameter is (X →
Boolean).
Refactoring any-alien-at-left-edge? and has-even-lon? yields
;; alien → Boolean
;; Purpose: Determine if he given alien is at the left edge
(define (alien-at-left-edge? an-alien)
(= (posn-x an-alien) MIN-IMG-X))

;; (listof alien) → Boolean

;; Purpose: To determine if any alien is at scene’s left edge
(define (any-alien-at-left-edge? a-loa)
(ormap-pred alien-at-left-edge? a-loa))

;; (listof number) → Boolean

;; Purpose: Determine if the list of numbers contains an even
;; number
(define (has-even-lon? a-lon)
(ormap-pred even? a-lon))
116 Abstraction Over List-Processing Functions 461

Observe that calling ormap-pred with alien-at-left-edge? and a-loa makes

its signature:
;; (alien → Boolean) (listof alien) → Boolean
That is, X = alien. The abstract function applies the given alien-predicate to all the
elements of the given list of aliens and returns the oring of all the results. This is
exactly what any-alien-at-left-edge? must do. Similarly, calling ormap-pred
with even? and a-lon makes its signature:
;; (number → Boolean) (listof number) → Boolean
That is, X = number and it applies the given number-predicate to all the elements
of the given list of numbers and returns the oring of all the results. This is exactly
what has-even-lon? must do.
Take time again to reflect on what has been achieved. We now have a powerful
function to perform list oring operations. For example, any-alien-reached-
earth? from Sect. 74.3 and any-alien-at-right-edge? from Sect. 74.2 may be
refactored to
;; (listof alien) → Boolean
;; Purpose: Determine if any alien has reached earth
(define (any-alien-reached-earth? a-loa)
(ormap-pred alien-reached-earth? a-loa))

;; (listof alien) → Boolean

;; Purpose: To determine if any alien is at scene’s right edge
(define (any-alien-at-right-edge? a-loa)
(ormap-pred alien-at-right-edge? a-loa))
The abstract function we named ormap-pred is so powerful and useful that ISL+
provides it and is called ormap. Now that you understand the abstraction use ormap
in your programs.

* Ex. 238 — Design and implement a nonrecursive predicate to determine if a

given list of numbers contains an odd number.

*** Ex. 239 — Section 43 defines a student. A student makes the Dean’s list
if their grade point average is 3.5 or above. Design and implement a nonrecursive
predicate to determine if there are any students on the Dean’s list for a given
(listof student).
*** Ex. 240 — Design and implement a nonrecursive predicate to determine
if there are any images with an area of more than 500 pixels in a given (listof
image).

**** Ex. 241 — Design and implement a nonrecursive predicate to determine

if there are any prime numbers in a given (listof number).
462 20 Functional Abstraction

*** Ex. 242 — Argue that the calls to ormap-pred in

any-alien-reached-earth? and any-alien-at-right-edge? satisfy
ormap-pred’s signature.

116.4 List ANDing

In Sect. 75.1 the following predicate to determine if all the elements in a list of
numbers are even is designed:
;; (listof number) → Boolean
;; Purpose: Determine if the given list of numbers only has
;; even numbers
(define (all-even-lon? a-lon)
(or (empty? a-lon)
(and (even? (first a-lon))
(all-even-lon? (rest a-lon)))))
If you were asked to write a predicate to determine if all strings in a list of strings
had a length less than or equal to 5, your function definitions would look as follows:
;; string → Boolean
;; Purpose: Determine if the length of the given string is ≤ 5
(define (string-len<5? a-str) (<= (string-length a-str) 5))

;; (listof string) → Boolean

;; Purpose: Determine if the given (listof string) only has
;; strings of length ≤ 5
(define (string-all-len<5? a-lostr)
(or (empty? a-lostr)
(and (string-len<5? (first a-lostr))
(string-all-len<5? (rest a-lostr)))))
You can immediately observe that all-even-lon? and string-all-len<5? are
good candidates for abstraction. The structure of their bodies is exactly the same and
no refactoring is required.
The only significant difference is the predicate applied to the first element of
the given list. Therefore, the abstract function only needs one extra parameter. The
abstraction step yields
;; <X> (X → Boolean) (listof X) → Boolean
;; Purpose: Determine if the elements of the list satisfy
;; the given predicate
(define (andmap-pred pred a-lox)
(or (empty? a-lox)
116 Abstraction Over List-Processing Functions 463

(and (pred (first a-lox))

(andmap-pred pred (rest a-lox)))))
Observe that the functions abstracted from process different types of lists. Therefore,
the list processed by the abstract function is denoted by (listof X). In the functions
abstracted from, the predicate is applied to the first element of the list and returns a
Boolean. In the abstract function the elements of the list are of type X. Therefore,
the type of the predicate is (X → Boolean).
The functions abstracted from can now be refactored to
;; lon → Boolean
;; Purpose: Determine if the given lon only has even numbers
(define (all-even-lon? a-lon) (andmap-pred even? a-lon))

;; (listof string) → Boolean

;; Purpose: Determine if the given (listof string) only
;; has strings of length <= 5
(define (string-all-len<5? a-lostr)
(andmap-pred string-len<5? a-lostr))
Once again you can see that the elegance and clarity provided by the abstract function
are truly remarkable. Furthermore, we have a powerful abstract function to perform
list anding operations on a (listof X). This list operation is so useful that ISL
provides it and it is called andmap. Now that you understand the abstraction, use
andmap in your programs.

* Ex. 243 — Design and implement a nonrecursive predicate to determine if a

given list of numbers contains only odd numbers.

*** Ex. 244 — Design and implement a nonrecursive predicate to determine

if all the elements in a (listof posn) are in the first quadrant of the Cartesian
plane.

* Ex. 245 — Design and implement a nonrecursive predicate to determine if

all the elements in a (listof Boolean) are #true.
***** Ex. 246 — Section 75.2 discusses the design of a predicate,
sorted-lon2?, to determine if the given (listof number) is sorted in
non-decreasing order. Design and implement a predicate to determine if the
given (listof number) is sorted in nonincreasing order. Perform functional
abstraction over the two functions and refactor them to use the abstract function.
*** Ex. 247 — Argue that the calls to andmap-pred in all-even-lon? and
string-all-len<5? satisfy andmap-pred’s signature.
464 20 Functional Abstraction

116.5 List Mapping

Section 76 discusses the design of the following functions to apply a function to all
the elements of a list and return the list of obtained results:
;; (listof shot) → (listof shot)
;; Purpose: To move the given list of shots
(define (move-los a-los)
(if (empty? a-los)
E-LOS
(cons (move-shot (first a-los))
(move-los (rest a-los)))))

;; (listof string) → (listof number)

;; Purpose: Return the string lengths in the given
;; (listof string)
(define (lengths-lostr a-lostr)
(if (empty? a-lostr)
'()
(cons (string-length (first a-lostr))
(lengths-lostr (rest a-lostr)))))
By now you have probably noticed that these functions process different types of
lists and are good candidates for abstraction. The only significant difference between
them is the function applied to the given list’s first element.
Abstracting over the functions leads to the following abstract function:
;; <X Y> (X → Y) (listof X) → (listof Y)
;; Purpose: Return the values from applying the given function
;; to the given list’s elements
(define (map-f f a-lox)
(if (empty? a-lox)
'()
(cons (f (first a-lox))
(map-f f (rest a-lox)))))
Given that functions abstracted over process different types of lists, the list type for
map-f’s list parameter is (listof X). The given function is applied to the elements
of this list. Thus, the input to the given function must be of type X. The result of
this given function varies among the functions abstracted from. Therefore, the return
type of the given function is Y. This means that the type of the given function is
(X →Y). Note that X may or may not be the same as Y. The results of applying the
given function are placed in a list and this list is the returned value. The result list,
therefore, is of type (listof Y).
116 Abstraction Over List-Processing Functions 465

Refactoring the functions abstracted over to use the abstract function yields
;; los → los
;; Purpose: To move the given list of shots
(define (move-los a-los) (map-f move-shot a-los))

;; (listof string) → lon

;; Purpose: Return the string lengths in the given
;; (listof string)
(define (lengths-lostr a-lostr)
(map-f string-length a-lostr))
Now that we have a function to perform list mapping operations, observe how short
and elegant the above functions have become. Many experienced programmers argue
that the use of abstract functions is the poetry of programming. The abstract function
we named map-f is named map in ISL+. Now that you understand the abstraction
go ahead and use map in your programs.

*** Ex. 248 — Section 76 discusses the design of move-alien. Can this func-
tion be implemented using map? Justify your answer.

* Ex. 249 — Design and implement a nonrecursive to add 1 to the elements of

a list of numbers.
**** Ex. 250 — Design and implement a nonrecursive function to extract the
full names from a list of students. Section 43 defines the type student. Define
your function twice: once using map and once using accum. Which version do
you prefer? Why?

*** Ex. 251 — Implement map-f using accum.

**** Ex. 252 — Design and implement a nonrecursive function to compute

the second largest factor of each natural number in a list of natural numbers.
Keep in mind that the largest factor of a natural number, n, is n. Any auxiliary
functions you need to write may be recursive.

** Ex. 253 — Design and implement a nonrecursive function to make a copy

of a given list.

*** Ex. 254 — Argue that the calls to map-f in move-los and
lengths-lostr satisfy map-f’s signature.
466 20 Functional Abstraction

116.6 List Filtering

Section 77 discusses the design of a function to extract the even numbers from a list
of numbers and has an exercise to extract the posn shots from a list of shots. Both
functions are
;; (listof number) → (listof number)
;; Purpose: Return a list of the even numbers in the given list
(define (extract-evens a-lon)
(cond [(empty? a-lon) '()]
[(even? (first a-lon))
(cons (first a-lon) (extract-evens (rest a-lon)))]
[else (extract-evens (rest a-lon))]))

;; (listof shot) → (listof shot)

;; Purpose: To return a list of posn shots in the given list
(define (extract-posn-shots a-los)
(cond [(empty? a-los) '()]
[(posn? (first a-los))
(cons (first a-los) (extract-posn-shots (rest a-los)))]
[else (extract-posn-shots (rest a-los))]))
A visual inspection reveals that both functions have the same type of expression in
their bodies and, therefore, are good candidates for abstraction.
The only significant difference is the predicate applied to the first element of the
list. The abstraction step yields the following function:
;; <X> (X → Boolean) (listof X) → (listof X)
;; Purpose: Return the elements in the given list that
;; satisfy the given predicate
(define (filter-pred pred a-lox)
(cond [(empty? a-lox) '()]
[(pred (first a-lox))
(cons (first a-lox)
(filter-pred pred (rest a-lox)))]
[else (filter-pred pred (rest a-lox))]))
The functions abstracted over process different types. Therefore, filter-pred’s list
argument is of type (listof X). The predicates in functions abstracted over take
as input a list element and return a Boolean. Thus, the filter-pred’s predicate
type is (X → Boolean). Finally, the functions abstracted over always return a list
of the same type they get as input. This means that filter-pred’s return type must
be (listof X).
116 Abstraction Over List-Processing Functions 467

The functions abstracted over are refactored to use filter-pred as follows:

;; lon → lon
;; Purpose: Return list of the even numbers in the given list
(define (extract-evens a-lon) (filter-pred even? a-lon))

;; los → los
;; Purpose: Return list of posn shots in the given list
(define (extract-posn-shots a-los) (filter-pred posn? a-los))
The elegance of the refactored functions is appreciated by those that understand the
abstract function. This abstraction is so powerful and useful that ISL+ provides this
abstract function and it is named filter. Now that you understand the abstraction,
go ahead and use filter in your programs.

* Ex. 255 — Design and implement a nonrecursive function to extract the odd
numbers from a list of numbers.
** Ex. 256 — Design and implement a nonrecursive function to extract the
multiples of 10 from a list of numbers.

*** Ex. 257 — Design and implement a nonrecursive function to extract the
prime numbers from a list of numbers. Your auxiliary function to determine if
a given number is prime may be recursive.

*** Ex. 258 — Section 77 discusses the design of remove-hit-aliens. Can

this function be refactored to use filter? Justify your answer.

*** Ex. 259 — Design and implement a nonrecursive function to extract the
posns from a (listof posn) that are either on the x-axis or the y-axis.

*** Ex. 260 — Argue that the calls to filter-pred in extract-evens and
extract-posn-shots satisfy filter-pred’s signature.

116.7 List Sorting

Section 78 discusses insertion sorting. The following functions are designed to sort
a list in non-decreasing order:
;; number (listof number) → (listof number)
;; Purpose: Insert the given number in the given list to
;; create an list of numbers in non-decreasing order
;; ASSUMPTION: The given lon is sorted in nondecreasing order
(define (insert a-num a-lon)
(cond [(empty? a-lon) (list a-num)]
468 20 Functional Abstraction

[(<= a-num (first a-lon)) (cons a-num a-lon)]

[else (cons (first a-lon)
(insert a-num (rest a-lon)))]))
;; (listof number) → (listof number)
;; Purpose: Sort the given list of numbers in non-decreasing
;; order
(define (sort-lon a-lon)
(cond [(empty? a-lon) '()]
[else
(insert (first a-lon) (sort-lon (rest a-lon)))]))
If you needed to write code to sort a list of numbers in nonincreasing order, the
functions you develop would likely look like this:
;; number (listof number) → (listof number)
;; Purpose: Insert the given number in the given list to create
;; a list of numbers in nonincreasing order
;; ASSUMPTION: The given lon is sorted in nonincreasing order
(define (insert>= a-num a-lon)
(cond [(empty? a-lon) (list a-num)]
[(>= a-num (first a-lon)) (cons a-num a-lon)]
[else (cons (first a-lon)
(insert>= a-num (rest a-lon)))]))
;; (listof number) '() (listof number)
;; Purpose: Sort the given list of numbers in nonincreasing
;; order
(define (sort-lon>= a-lon)
(cond [(empty? a-lon) '()]
[else
(insert>= (first a-lon) (sort-lon>= (rest a-lon)))]))
Finally, consider the following code to sort a (listof item) in non-decreasing
order by price:
;; An item (item) is a structure
;; (make-item string number number)
;; that has the item’s name, price, and quantity in stock.
(define-struct item (name price quantity))

;; item item → Boolean

;; Purpose: Determine if the first given item’s price is <=
;; to the second given item’s price
(define (leq-item item1 item2)
(<= (item-price item1) (item-price item2)))
;; item (listof item) → (listof item)
;; Purpose: Insert the given item in the given loi to create an
;; loi in nondecreasing order by price
116 Abstraction Over List-Processing Functions 469

;; ASSUMPTION: The given loi is sorted in nondecreasing order

;; by price
(define (insert-loi an-item a-loi)
(cond [(empty? a-loi) (list an-item)]
[(leq-item an-item (first a-loi)) (cons an-item a-loi)]
[else (cons (first a-loi)
(insert-loi an-item (rest a-loi)))]))
;; loi → loi
;; Purpose: Sort the given loi in nonincreasing order by price
(define (sort-loi a-loi)
(cond [(empty? a-loi) '()]
[else (insert-loi (first a-loi)
(sort-loi (rest a-loi)))]))
Immediately you may notice that these functions are good candidates for abstraction.
Observe that the only significant difference between the inserting functions is the
predicate used to compare two values. The sorting functions are essentially the same
also. They call an inserting function and make a recursive call. Their only significant
difference is the inserting function called.
Abstracting over the inserting functions yields the following abstract function:
;; <X> (X → Boolean) X (listof X) → (listof X)
;; Purpose: Insert the given number in the given lox to
;; create sorted lox using the given predicate
;; ASSUMPTION: Given lon is sorted as defined by the given
;; predicate
(define (insert-pred pred an-x a-lox)
(cond [(empty? a-lox) (list an-x)]
[(pred an-x (first a-lox)) (cons an-x a-lox)]
[else (cons (first a-lox)
(insert-pred pred an-x (rest a-lox)))]))
The list types processed by the inserting functions vary. Therefore, the list type for
insert-pred is (listof X). The value to insert is always of the same type as the
values in the given list. This means that for insert-pred the value to insert must
be of type X. Finally, the predicate always takes as input a list value and returns a
Boolean. Thus, for insert-pred, the predicate type is (X →Boolean).
Refactoring the inserting functions yields
;; insert: number (listof number) → (listof number)
;; Purpose: To insert the given number in the given lon to
;; create a list of numbers in non-decreasing order
;; ASSUMPTION: The given list is sorted in non-decreasing order
(define (insert a-num a-lon) (insert-pred <= a-num a-lon))

;; number (listof number) → (listof number)

;; Purpose: To insert the given number in the given lon to
470 20 Functional Abstraction

;; create a list of numbers in nonincreasing order

;; ASSUMPTION: The given list is sorted in nonincreasing order
(define (insert>= a-num a-lon) (insert-pred >= a-num a-lon))

;; item (listof item) → (listof item)

;; Purpose: To insert the given item in the given loi to
;; create a list of numbers in nondecreasing order
;; by price
;; ASSUMPTION: The given list is sorted in nondecreasing order
;; by price
(define (insert-loi an-item a-loi)
(insert-pred leq-item an-item a-loi))
The simplification of these three functions is, indeed, very impressive.
The only difference in the functions sort-lon, sort-lon>=, and sort-loi
needs to also be abstracted away. Instinctively, we may create an abstract function that
takes as input the proper inserting function. Such an approach yields the following
abstract sorting function:
;; <X> (X (listof X) → (listof X)) (listof X) → (listof X)
;; Purpose: Sort given lox using the given inserting function
(define (sort-lox insert-f a-lox)
(cond [(empty? a-lox) '()]
[else (insert-f (first a-lox)
(sort-lox (rest a-lox)))]))
Although technically correct this is the wrong abstraction for anyone using our
sorting function. It forces the user of the function to provide an inserting function.
The average user may not find this abstraction useful. The average user is likely to
only want to sort a list using a predicate. They are not interested in nor inclined to
develop the proper inserting function even if they can use insert-pred to do so.
The proper abstraction for sorting only requires the user to specify the predicate
they wish to use to sort the given list. In other words, the user provides a predicate and
a list to sort. How can we develop an abstract sorting function that provides the right
abstraction? In essence, we need to eliminate the inserting function as the difference
and make the sorting predicate the difference. This can be achieved by refactoring
the sorting functions before performing the abstraction step. More concretely, we
can inline the proper use of insert-pred into the bodies of the sorting functions.
This yields the following refactored sorting functions:
;; (listof number) → (listof number)
;; Purpose: Sort the given list of numbers in nondecreasing
;; order
(define (sort-lon a-lon)
(cond [(empty? a-lon) '()]
[else (insert-pred <=
(first a-lon)
(sort-lon (rest a-lon)))]))
116 Abstraction Over List-Processing Functions 471

;; (listof number) '() (listof number)

;; Purpose: Sort the given list of numbers in nonincreasing
;; order
(define (sort-lon>= a-lon)
(cond [(empty? a-lon) '()]
[else (insert-pred >=
(first a-lon)
(sort-lon>= (rest a-lon)))]))

;; (listof item) → (listof item)

;; Purpose: To sort the given list of items in nonincreasing
;; order by price
(define (sort-loi a-loi)
(cond [(empty? a-loi) '()]
[else
(insert-pred leq-item
(first a-loi)
(sort-loi (rest a-loi)))]))
There are two important consequences of inlining the use of insert-pred. The first
is the only difference among the sorting functions is the predicate used to perform
the sorting. The second is that insert, insert>=, and insert-loi are no longer
needed and may safely be deleted from our program.
Performing the abstraction step on the sorting functions now yields the following
abstract sorting function:
;; <X> (X → Boolean) (listof X) → (listof X)
;; Purpose: To sort the given lox using the given predicate
(define (sort-pred pred a-lox)
(cond [(empty? a-lox) '()]
[else (insert-pred pred
(first a-lox)
(sort-pred (rest a-lox)))]))
In the functions abstracted from, the list type processed varies. For sort-pred this
means that the given list’s type is (listof X). Observe that the given predicate’s
input must be the same type as the elements in the given list. Therefore, its type is
(X → Boolean).
Refactoring the sorting functions to use sort-pred yields
;; (listof number) → (listof number)
;; Purpose: To sort the given list of numbers in
;; non-decreasing order
(define (sort-lon a-lon) (sort-pred <= a-lon))
472 20 Functional Abstraction

;; (listof number) → (listof number)

;; Purpose: To sort the given list of numbers in
;; nonincreasing order
(define (sort-lon>= a-lon) (sort-pred >= a-lon))

;; (listof item) → (listof item)

;; Purpose: To sort the given list of items in
;; nonincreasing order
;; by price
(define (sort-loi a-loi) (sort-pred leq-item a-loi))
What do you think about how simple it is now to write sorting functions? Be aware,
however, that the elimination of insert, insert>=, and insert-loi requires that
you update the sample expressions for the sorting functions. This powerful abstract
function to write sorting functions is called sort in ISL+.
The efforts to perform functional abstraction using the sorting functions had left
us an important lesson. Not all possible abstractions are useful or make code easier to
understand. The first attempt to abstract the sorting functions yielded an abstraction
that is unlikely to be useful. It is important to be mindful about the usefulness of the
abstractions we create. Experience will teach you that the wrong abstraction makes
programs harder to understand or harder to understand.

** Ex. 261 — Section 43 defines the type student. Design and implement a
nonrecursive function to sort a list of students in non-decreasing order by grade
point average.

** Ex. 262 — Design and implement a nonrecursive function to sort a list of

images in nonincreasing order by area.

*** Ex. 263 — Design and implement a nonrecursive function to sort a list of
aliens in non-decreasing order by how high they are in a scene.

*** Ex. 264 — Argue that the calls to sort-pred in sort-lon, sort-lon>=,
and sort-loi satisfy sort-pred’s signature.

117 Abstraction over Interval-Processing Functions

Consider the following functions to process an interval. The first computes the sum
of the squares of the integers in the given interval. The second computes the list of
strings representing the integers in the interval.
;; [int..int] → int
;; Purpose: Compute the product of the squares of the ints
;; in the given interval
117 Abstraction over Interval-Processing Functions 473

(define (interval-product-sqrs low high)

(if (> low high)
1
(* (sqr low) (interval-product-sqrs (add1 low) high))))

;; [int..int] → (listof string)

;; Purpose: Construct the list of strings for the ints in
;; the given interval
(define (interval->lostr low high)
(if (> low high)
'()
(cons (number->string low)
(interval->lostr (add1 low) high))))
A visual examination reveals that they are good candidates for abstraction. Both
functions have the same number of parameters and process the interval from the low
value to the high value. They both have the same type of expression in their body:
an if-expression that tests if the given interval is empty, a consequent expression that
is a concrete value, and alternative expression that applies a two-input function to
the values obtained from applying a one-input function to the first element of the list
and the result of recursively processing the rest of the list.
There are three significant differences: the base value returned, the function
applied to the first element of the interval, and the combining function. This is
reminiscent of the differences among list-summarizing functions in Sect. 116.1 sug-
gesting that the above functions are interval-summarizing functions. The result of
the abstraction step yields
;; <X Y> X (int → Y) (Y X → X) [int..int] → X
;; Purpose: Compute a value by traversing the given interval
;; from low to high using the given base value,
;; function to apply to low, and combining function.
(define (interval-accum-l2h base-val ffirst comb low high)
if (> low high)
base-val
(comb (ffirst low)
(interval-accum-l2h base-val ffirst
comb (add1 low) high))))
The type of the interval ought to be clear. The base value, the second input and the
return type of the combining function, and the return type of the abstract function
must all be the same and it is denoted by X. The type applied of the function to the
low element of the interval must be (int → Y) given that the low element is an
integer and it may return any type of data. Finally, the first input to the combining
function must be of type Y making the type of the combining function (Y X → X).
474 20 Functional Abstraction

The final step refactoring the functions abstracted over yields

;; [int..int] → (listof string)
;; Purpose: Construct the list of strings for the ints in
;; the given interval
(define (interval->lostr low high)
(interval-accum-l2h '() number->string cons low high))

;; [int..int] → int
;; Purpose: Compute the product of the squares of the ints
;; in the given interval
(define (interval-product-sqrs low high)
(interval-accum-l2h 1 sqr * low high))
Once again, observe that the functions abstracted over have been greatly simplified.
This abstract function, like the others in this chapter, allows the programmers to
spend more time on problem solving and less time on mundane typing of repetitions,
which usually means fewer bugs in the software developed.

** Ex. 265 — Design and implement a nonrecursive function to sum the inte-
gers in a given interval.

*** Ex. 266 — Design and implement a nonrecursive predicate to determine

if a given interval contains a multiple of 7.

*** Ex. 267 — ISL+’s random function takes as input, n, a nonzero natural
number and returns a random natural number in [0..n-1]. Design and imple-
ment a nonrecursive function that takes as input an interval, [i..j], and that
returns a list of random numbers less than 1000, one for each integer in the
given interval. For example, (generate-randoms 10 12) may return '(786 31 87).

*** Ex. 268 — Argue that the calls to interval-accum-l2h in

interval->lostr and interval-product-sqrs satisfy
interval-accum-l2h’s signature.

***** Ex. 269 — Design and implement an abstract function to process an

interval from right to left (i.e., from the high end to the low end of the inter-
val). Refactor interval-product in Sect. 86 to use the abstract function you
develop.
118 What Have We Learned in This Chapter? 475

118 What Have We Learned in This Chapter?

The important lessons of Chap. 20 are summarized as follows:

• Functional abstraction allows us to eliminate repetitions in our programs and to
write a code that is more elegant, shorter, easier to maintain, and with fewer bugs.
• Abstract (or generic) functions can perform many different tasks.
• Thinking abstractly makes problem solving and programming easier.
• The design recipe for abstraction has three steps: mark the differences, create the
abstract function, and refactor the functions abstracted over.
• Functions are a type of value. Programming languages that provide support for
functions as values are said to have first-class functions.
• Care must be taken to develop the signature of an abstract function and signatures
need to be explicitly parameterized (i.e., type variables in angled brackets before
the signature).
• Care must be taken to make sure the arguments provided to an abstract function
satisfy the abstract function’s signature.
• ISL+ provides the following abstract functions: ormap, andmap, map, filter,
and sort.
• Not all possible abstractions are useful or make code easier to understand.
• Inlining can help to create useful abstractions.
• Two useful abstract functions may be developed to process an interval: one that
processes from the low end to the high end and one that processes from the high
end to the low end.
• Many experienced programmers consider abstract functions the poetry of pro-
gramming.
Chapter 21
Encapsulation

As the use of auxiliary functions increases, the size of programs grows. You may
have already noticed in Aliens Attack, for example, that you have related functions
scattered all over your program’s file. This makes it difficult to understand a design
and to make refinements. It is therefore desirable to encapsulate the definitions in
our programs. Encapsulation means that we package together all related definitions
as a single piece of software. This allows any reader of our code to more easily
understand the design and makes the process of program refinement easier.
New syntax is required to encapsulate definitions. This new syntax is a local-
expression. In essence, a local-expression allows us to have definitions inside of
functions. In this manner auxiliary functions are defined inside the functions that
need them instead of being defined elsewhere (possibly far away) in our program.
As this chapter and Chap. 25 illustrate, the power of encapsulation has consequences
for design and for performance.

119 Local-Expressions

A local-expression is used to group together an arbitrary number of related defini-

tions. The local-expression syntax is
expr ::= (local [def∗ ] expr)
A local-expression has inside parenthesis the keyword local, then inside square
brackets zero or more definitions, and finally an expression. The expression after
the local definitions is called the body of the local-expression. This body may
refer to any of the definitions locally defined. The value of a local-expression is
the value of its body. We say that the local definitions are encapsulated inside the
local-expression.

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 477
M. T. Morazán, Animated Problem Solving, Texts in Computer Science,
https://doi.org/10.1007/978-3-030-85091-3_21
478 21 Encapsulation

Consider, for example, the function to move a (listof shot) designed for the
Aliens Attack game in Sect. 76.2:
;; los → los
;; Purpose: To move the given list of shots
(define (move-los a-los)
(if (empty? a-los)
E-LOS
(cons (move-shot (first a-los))
(move-los (rest a-los)))))
Revisiting this function makes us realize that the same one-input function,
move-shot, is applied to every shot in the given list. This suggest that the function
may be refactored and simplified using map(similar to the design in Sect. 116.5:
;; los → los
;; Purpose: To move the given list of shots
(define (move-los a-los) (map move-shot a-los))
Can any reader fully understand and can any programmer easily make refinements
to this code? To a certain degree yes, but not fully. Any refinement to move-los,
for example, is likely to involve changes to move-shot. Therefore, it is desirable to
encapsulate both of these functions into a single package. Using a local-expression
allows us to refactor the dispersed code to
;; los → los
;; Purpose: To move the given list of shots
(define (move-los a-los)
(local [;; shot → shot
;; Purpose: To move the given shot
(define (local-move-shot a-shot)
(cond [(eq? a-shot NO-SHOT) a-shot]
[else
(cond [(= (posn-y a-shot) MIN-IMG-Y)
NO-SHOT]
[else
(make-posn
(posn-x a-shot)
(move-up-image-y
(posn-y a-shot)))]))
(map local-move-shot a-los))
The body of move-los is now a local-expression. This expression locally defines
the auxiliary function local-move-los. The name of move-los has been prefixed
by local to highlight that it is now a locally defined function. Such prefixing is not
necessary in practice. The body of the local-expression is the body of the original
move-los. Now, move-los is a complete package containing all the functions
needed to move a list of shots.
119 Local-Expressions 479

No new computational power is offered by local-expressions. Anything com-

puted using a local-expression may be computed without it. In other words, they are
simply syntactic sugar to make program development easier. Any local-expression
may be eliminated as follows:
1. Give every local definition a unique name not found elsewhere in the program and
change all references using the old name to be references using the new names.
2. Move the renamed definitions out of the local-expression.
3. Replace the local-expression with its body.
In essence, these steps are a refactoring recipe to eliminate local-expressions. This
refactoring recipe is not presented to develop a practical skill. It is presented to
clarify the role of local-expressions. In addition, be aware that this refactoring
recipe is incomplete as we shall see later in this chapter. Nonetheless, let us consider
the following concrete example:
(define DELTA 10)
(+ DELTA (local [(define DELTA 100)
(define VAL (+ DELTA DELTA))]
VAL))
What is the value of the sum? 30? 210? 300? Some other value? Let us refactor the
program to eliminate the local-expression. After Step 1 we have
(define DELTA 10)
(+ DELTA (local [(define DLT 100)
(define VAL (+ DLT DLT))]
VAL))
Observe that inside the original local-expression all references to DELTA are to the
locally defined variable. Therefore, all of them are changed to DLT. Now that all the
defined names do no clash with names defined outside the local-expression Step 2
may be completed to yield
(define DELTA 10)
(define DLT 100)
(define VAL (+ DLT DLT))
(+ DELTA (local [] VAL))
Given that there are no local definitions Step 3 replaces the local-expression with
its body to yield
(define DELTA 10)
(define DLT 100)
(define VAL (+ DLT DLT))
(+ DELTA VAL)
In this form it becomes clear that the value of the sum is 210.
480 21 Encapsulation

120 Lexical Scoping

Why is it necessary to rename definitions before lifting them outside a local-

expression? How do we know which references need to be updated to use the new
name? The answer is lexical scoping (or static scoping). The lexical scope of a
definition or a variable (as the parameters of a function) is the part of the program
where the definition or variable is valid. Consider the following program fragments
that contain a function definition to multiply two natural numbers:
..
.
;; natnum natnum → natnum
;; Purpose: Multiply the given natural numbers
(define (mult x y)
(if (= x 0)
0
(+ y (mult (sub1 x) y))) )
..
.
There are three names declared in this program fragment: mult is the name of a
function that takes two inputs, x is the name of the first parameter, and y is the
name of the second parameter. These three declarations do not have the same lexical
scope. That is, the program defines where these variables are valid. We say that
mult has global scope (or simply is global) because it is not locally defined (inside
a local-expression) nor is it a parameter. The lexical scope of x and y, on the other
hand, is the body of the function. That is, the parameters x and y are only valid in
the body of mult (emphasized in slanted fonts) and are invalid elsewhere. We say
that x and y are bound variables in the body of mult. That is, there is a declaration
for these variables in mult (specifically the parameters in the function header).
Now, consider this larger fragment of the program:
..
.
(define x 2)

;; natnum → natnum
;; Purpose: Add x to the given number
(define (add-x-y-and-z y) (+ x y z))

;; natnum → natnum
;; Purpose: Compute the factorial of the given number
(define (fact x)
(if (= x 0)
1
(* x (fact (sub1 x)))))
120 Lexical Scoping 481

;; natnum natnum → natnum

;; Purpose: Multiply the given natural numbers
(define (mult x y)
(if (= x 0)
0
(+ y (mult (sub1 x) y))))
..
.
In this program fragment there are 4 declarations that have global scope: x (the one
defined as 2), add-x, fact, and mult. There are four declarations that do not have
global scope. The y parameter of add-x-y-and-z is only valid in its body. The x
parameter of fact is only valid in its body. The x and y parameters of mult are
only valid in its body. Observe that x is declared and referenced multiple times. In
(* x (fact (sub1 x))), for example, the references to x are bound to fact’s
x parameter. Similarly, in (+ x y z, for example, the x referenced is bound to
addd-x-y-z’s x parameter. Lexical scoping allows programmers to use the same
variable names multiple times. The scoping rules make it possible to understand
where a variable is bound or if it is free. In the body of addd-x-y-z we say that y
and z are free variables. A free variable is one that is not bound to a declaration. In
this case, y and z are free because they are not declared in addd-x-y-z.
Most programming languages use lexical scoping. Lexical scoping determines
where a variable is bound based on the syntax of the programming language. Global
definitions are valid everywhere unless they are shadowed by a local declaration. If
the same variable name is reused in a program, the local declaration is the bounding
declaration. All other declarations of this variable are shadowed and unaccessible in
the scope of the local declaration. In other words, the closest declaration is always
the bounding declaration. For example, in the program fragment above the definition
that binds x to 2 is shadowed by the declarations of x (as parameters) in fact and
mult. In these functions, references to x are always to the parameter of the function
and not to the global x. In contrast, the reference to the free variable x in add-x-y-z
is bound to the global x. The same is true for the reference to z that must be defined
in the part of the program not displayed. A free variable that is not bound to a
declaration is said to be unbound and is an error.
The lexical scoping rules are:
1. The definitions not typed inside a local-expression have global lexical scope
and are valid everywhere unless shadowed.
2. The lexical scope of parameters is the function that declares them unless shad-
owed.
3. The lexical scope of a local definition is the declarations and the body of the
local-expression unless shadowed.
We can now revisit the example used to illustrate the refactoring recipe to eliminate
local-expressions:
(define DELTA 10)
(+ DELTA (local [(define DELTA 100)
(define VAL (+ DELTA DELTA))]
VAL))
482 21 Encapsulation

We do not need to refactor this code to understand it. The declaration of DELTA as
10 has global scope. This declaration is shadowed by the local declaration of DELTA
inside the local-expression. According to Rule 3 above, inside this expression
DELTA is 100. This makes VAL 200. Given that VAL is returned in the body of the
local, the value of the local-expression is 200. The local declaration of DELTA is not
valid outside of the local-expression. Therefore, the reference to DELTA as the first
argument to + must be to a different declaration of DELTA. In this case, the closest
declaration of DELTA is the global one making +’s first argument 10. It is now also
clear, without refactoring, that the value of the sum is 210.
As a final example, consider the following program:
(define X 2)
(define Y 3)

(+ X Y (local [(define X 10)]

(+ X (local [(define Y 10)]
(+ X Y)))))
What is the value of the sum? The X and Y defined before the summing expression
have global scope. Inside first local-expression the X bound to 5 shadows the global
X. Inside the innermost local-expression X is a free variable and the Y bound to
10 shadows the global Y. Where does the value of X come from? It comes from the
closest declaration in scope: the declaration of X in the containing local-expression.
This makes the value of the sum 35. If this is not clear, you can always refactor to
remove the local-expressions. Start with the innermost local-expression to get:
(define X 2)
(define Y 3)
(define Y-INNERMOST 10)
(+ X Y (local [(define X 10)]
(+ X (+ X Y-INNERMOST))))
The Y declaration in the innermost local-expression has been renamed to
Y-INNERMOST and moved to the global level. The innermost local-expression is
replaced with its body. Observe that the reference to X in (+ X Y-INNERMOST) is
not free. That is, the previously free variable X is now a bound variable. The same
steps are taken to remove the only local-expression left. This process yields
(define X 2)
(define Y 3)
(define Y-INNERMOST 10)
(define X-MIDDLE 10)
(+ X Y (+ X-MIDDLE (+ X-MIDDLE Y-INNERMOST)))
The local X bound to 10 is renamed to X-MIDDLE and the local-expression is
substituted with its body. It this form it becomes clear that the value of the sum is 35.
121 Using Local-Expressions 483

121 Using Local-Expressions

As with any new type of expression it is important to understand when and how
to use a local-expression. There are four reasons to use a local-expression. The
first, as we begun to discuss above, is encapsulation. Use to a local-expression to
organize your code to package together related functions. The second is to make
code more readable. The third is to make the use of abstract functions possible. The
fourth is to eliminate multiple evaluations of the same expression.
The following subsections discuss each of these reasons in more detail. Keep
in mind that sometimes the use of a local-expression is subjective. For example,
what does it mean for code to be more readable? For some people the use of a
local-expression may make understanding the design of a function easier and for
others the same function is perfectly understandable without it. If you feel that a
local-expression makes your code more readable, then do not hesitate to use it.

121.1 Encapsulation

Encapsulation is a good practice to develop because it makes refinements easier.

When a refinement is needed, all the code is located in one place (i.e., one package).
In addition, it also makes it easier to move code from one file to another given that
all the needed code is packaged together. To encapsulate code use these steps:
1. Develop your program using any of the design recipes.
2. Test your functions thoroughly.
3. Identify related definitions to package together.
4. Identify the main function among the functions to be packaged.
5. Place all the functions inside a local-expression.
6. Place the local-expression inside a new function that has the same signature,
purpose, and header as the main function.
7. In the body of the local-expression call the main function with the given argu-
ments.
8. Comment out the sample expressions and the tests for all functions except the
main function and comment out the sample expressions and the tests using sample
computations for the main function.
It is important to remember not to encapsulate functions before thorough testing has
been performed and you are relatively sure that the functions are bug free. Otherwise,
you risk making untested buggy functions local. This is important because localized
functions cannot be tested due to scoping rules. Localized functions do not exist
outside the local-expression and, therefore, tests cannot be written for them. In
essence, test before encapsulation.
484 21 Encapsulation

Fig. 97 Draft encapsulated draw-world function definition

;; world → scene Purpose: To draw the world in E-SCENE
(define (draw-world a-world)
(local
[(define E-SCENE-COLOR pink)
(define E-SCENE (empty-scene E-SCENE-W E-SCENE-H E-SCENE-COLOR))
;; los scene → scene Purpose: Draw given los in given scene
(define (draw-los a-los scn)
(if (empty? a-los) scn
(draw-shot (first a-los)
(draw-los (rest a-los) scn))))
;; loa scene → scene Purpose: Draw given loa in given scene
(define (draw-loa a-loa scn)
(if (empty? a-loa) scn
(draw-alien (first a-loa) (draw-loa (rest a-loa) scn))))
;; rocket scene → scene Purpose: Draw rocket in given scene
(define (draw-rocket a-rocket a-scene)
(draw-ci ROCKET-IMG a-rocket ROCKET-Y a-scene))
;; world → scene Purpose: To draw the world in E-SCENE
(define (draw-world a-world)
(draw-los (world-shots a-world)
(draw-loa
(world-aliens a-world)
(draw-rocket (world-rocket a-world) E-SCENE))))]
(draw-world a-world)))

Consider the function draw-world from Aliens Attack (its design is discussed
in Sect. 91):
;; world → scene
;; Purpose: To draw the world in E-SCENE
(define (draw-world a-world)
(draw-los (world-shots a-world)
(draw-loa (world-aliens a-world)
(draw-rocket (world-rocket a-world)
E-SCENE))))
Observe that 3 auxiliary functions and a constant are used. Where are these functions
and constant defined? Clearly, they are defined elsewhere in the program that forces
any reader of the code to search for them. Further observe that the auxiliary functions
and the constant are not needed elsewhere in the program. Given that the functions
have been thoroughly tested, they are ideal for encapsulation along with the constant.
The definitions to encapsulate are E-SCENE, E-SCENE-COLOR (only used to define
E-SCENE in the program), draw-world, draw-loa, and draw-rocket. The main
function is draw-world.
The draft new code after encapsulation for draw-world is displayed in Figs. 97
and 98. Observe that in Fig. 97 all the related definitions identified are localized.
The interface with the rest of the program (e.g., using draw-world in the big-bang
expression inside run) is unchanged because the signature, purpose, and function
121 Using Local-Expressions 485

Fig. 98 Tests for draft encapsulated draw-world function

;; Tests using sample computations for draw-world
(check-expect (draw-world (make-world INIT-ROCKET2 (list INIT-ALIEN)
DIR3 empty))

)
(check-expect (draw-world (make-world INIT-ROCKET (list INIT-ALIEN2)
DIR2 (list SHOT2)))

header of the new (global) draw-world implementation remain unchanged. Take some
time to think about the call to draw-world inside the local-expression. Is this a
recursive call? Why or why not? Remember the scoping rules. It is not a recursive call
because it is calling the local draw-world function. There are no sample expressions
or tests using sample computations in Fig. 98 because the local functions are not in
scope and, therefore, may not be used to write sample expressions. This highlights
the importance of testing before encapsulation.
Observe that draw-shot is an auxiliary function that is only needed by draw-los.
The same observation may be made for draw-alien and draw-loa. Respectively
encapsulating these functions yield these new implementations for draw-los and
draw-loa:
;; los scene → scene
;; Purpose: To draw the given los in the given scene
(define (draw-los a-los scn)
(local [;; shot scene → scene
;; Purpose: To draw the shot in the given scene
(define (draw-shot a-shot scn)
(if (eq? a-shot NO-SHOT)
scn
(draw-ci SHOT-IMG
(posn-x a-shot)
(posn-y a-shot)
scn)))
486 21 Encapsulation

Fig. 99 Encapsulated draw-world function, part I

;; world scene Purpose: To draw the world in E-SCENE
(define (draw-world a-world)
(local
[(define E-SCENE-COLOR pink)
(define E-SCENE (empty-scene E-SCENE-W E-SCENE-H E-SCENE-COLOR))
;; image image-x image-y image image
;; Purpose: Place first image in second image at given coordinates
(define (draw-ci char-img an-img-x an-img-y scn)
(local [;; image-x pixel-x Purpose: Translate image-x to pixel-x
(define (image-x->pix-x ix)
(+ (* ix IMAGE-WIDTH) (/ IMAGE-WIDTH 2)))
;; image-y pixel-y Purpose: Translate image-y to pixel-y
(define (image-y->pix-y iy)
(+ (* iy IMAGE-HEIGHT) (/ IMAGE-HEIGHT 2)))
;; image image-x image-y image image
;; Purpose: Place first image in second image at given
;; coordinates
(define (draw-ci char-img an-img-x an-img-y scn)
(place-image char-img
(image-x->pix-x an-img-x)
(image-y->pix-y an-img-y)
scn))]
(draw-ci char-img an-img-x an-img-y scn)))
;; los scene scene
;; Purpose: To draw the given los in the given scene
(define (draw-los a-los scn)
(local [;; shot scene scene Purpose: Draw shot in given scene
(define (draw-shot a-shot scn)
(if (eq? a-shot NO-SHOT) scn
(draw-ci SHOT-IMG (posn-x a-shot) (posn-y a-shot) scn)))
;; los scene scene
;; Purpose: To draw the given los in the given scene
(define (draw-los a-los scn)
(if (empty? a-los) scn
(draw-shot (first a-los) (draw-los (rest a-los) scn))))]
(draw-los a-los scn)))

;; los scene → scene

;; Purpose: To draw the given los in the given scene
(define (draw-los a-los scn)
(if (empty? a-los)
scn
(draw-shot (first a-los)
(draw-los (rest a-los) scn))))]
(draw-los a-los scn)))
121 Using Local-Expressions 487

Fig. 100 Encapsulated draw-world function, part II

;; loa scene scene
;; Purpose: Draw given loa in given scene
(define (draw-loa a-loa scn)
(local
[;; alien scene scene Purpose: Draw alien in given scene
(define (draw-alien an-alien scn)
(draw-ci ALIEN-IMG (posn-x an-alien) (posn-y an-alien) scn))
;; loa scene scene Purpose: Draw given loa in given scene
(define (draw-loa a-loa scn)
(if (empty? a-loa) scn
(draw-alien (first a-loa) (draw-loa (rest a-loa) scn))))]
(draw-loa a-loa scn)))

;; rocket scene scene Purpose: Draw given rocket in given scene

(define (draw-rocket a-rocket a-scene)
(draw-ci ROCKET-IMG a-rocket ROCKET-Y a-scene))

;; world scene Purpose: To draw the world in E-SCENE

(define (draw-world a-world)
(draw-los (world-shots a-world)
(draw-loa (world-aliens a-world)
(draw-rocket (world-rocket a-world)
E-SCENE))))]
(draw-world a-world)))

;; loa scene → scene

;; Purpose: To draw the given loa in the given scene
(define (draw-loa a-loa scn)
(local [;; alien scene → scene
;; Purpose: Draw the given alien in the given scene
(define (draw-alien an-alien scn)
(draw-ci ALIEN-IMG
(posn-x an-alien)
(posn-y an-alien)
scn))
;; loa scene → scene
;; Purpose: To draw the given loa in the given scene
(define (draw-loa a-loa scn)
(if (empty? a-loa)
scn
(draw-alien (first a-loa)
(draw-loa (rest a-loa) scn))))]
(draw-loa a-loa scn)))
Observe that draw-ci is an auxiliary function used by draw-rocket, draw-shot,
and draw-alien. If draw-ci is made local to these functions, there would be
repeated code and this ought to be avoided. Instead, what is needed is to place
draw-ci in a location where it is in scope of all three of these functions. This means
488 21 Encapsulation

that draw-ci cannot be local to, for example, draw-shot because it would be out
of scope for the remaining two functions. If left at the global level draw-ci would
be in scope for all three functions. This, however, leaves draw-ci outside of the
only package that needs it. Therefore, the best solution is to make draw-ci local
to draw-world. Figures 99 and 100 display the completed encapsulated version of
draw-world. Observe that the auxiliary functions only needed by draw-ci are also
locally encapsulated.
Take a moment to appreciate what has been accomplished. The code is now
organized in a manner that allows any reader (including yourself 6 months from now)
to easily find related functions. It also means that when a refinement is necessary
we know where the new tested functions must be placed. For example, if a score is
added to the world, then after testing you know that the draw-score function ought
to be local to draw-world.
*** Ex. 270 — Section 106 discusses the design of eval-sexpr. Encapsulate
this function and its auxiliary functions.

**** Ex. 271 — Section 88 discusses the design of largest-prime. Encap-

sulate this function and its auxiliary functions.

121.2 Readability

Section 106 discusses the design of the following functions to evaluate arithmetic
expressions using +, -, and *:
;; (listof sexpr) arrow (listof number) throws error
;; Purpose: To evaluate the sexprs in the given list
(define (eval-args a-losexpr)
(if (empty? a-losexpr)
'()
(cons (eval-sexpr (first a-losexpr))
(eval-args (rest a-losexpr)))))

;; function (listof number) → number throws error

;; Purpose: To apply the given function to the given numbers
(define (apply-f a-function a-lon)
(cond [(eq? a-function '+) (sum-lon a-lon)]
[(eq? a-function '-) (subt-lon a-lon)]
[else (mult-lon a-lon)]))

;; slist → number
;; Purpose: To evaluate the given slist
(define (eval-slist an-slist)
(apply-f (first an-slist) (eval-args (rest an-slist))))
121 Using Local-Expressions 489

;; sexpr → number throws error

;; Purpose: To evaluate the given sexpr
(define (eval-sexpr a-sexpr)
(if (number? a-sexpr)
a-sexpr
(eval-slist a-sexpr)))
What does (first a-losexpr) represents? Do you remember? Is it reasonable to
expect others reading this code for the first time to know or easily figure it out? It is
unlikely that you remember what all the expressions in the program represent and
it is even more unlikely that someone reading your code will be able to discern the
meaning of every expression. To mitigate this problem a local-expression may be
used to define intermediate values using descriptive variable names.
To start consider the function eval-slist. In its body there is an application
expression to apply apply-f to the first element of the given slist and the result
obtained from applying eval-args to the rest of the given slist. Anyone that is
familiar with ISL+ can make these observations, but such a verbatim description
of the code does not elucidate the reader about why the code does this. Anyone,
of course, can go read Sect. 106 to understand what this function does, but this is
time consuming. Instead of placing this burden on the readers of the code we can
define local variables with names that elicit the meaning of the expressions. The
idea is that we do not rely on list selector applications to showcase the meaning of
expressions to evaluate arithmetic expressions. Instead, local variables may be used
to easily discern that (first an-slist) represents the arithmetic operator and
that (eval-args (rest an-slist)) the arguments for the arithmetic operator.
This idea leads to the following refactoring:
;; slist → number
;; Purpose: To evaluate the given slist
(define (eval-slist an-slist)
(local [(define operator (first an-slist))
(define arg-exprs (rest an-slist))]
(apply-f operator (eval-args arg-exprs))))
Observe that eval-slist is more readable now. Any reader can now immediately
understand that the slist’s operator is applied to the arguments obtained from
evaluating the argument-expressions in the given slist.
A similar refactoring can be done to make it easier to understand what the
function eval-args does. Observe that the expressions (first a-losexpr) and
(rest a-losexpr) give the reader no indication of what they represent. The former
490 21 Encapsulation

represents the first argument and the latter represents the rest of the arguments. With
this in mind, a local-expression may be used to refactor eval-args as follows:
;; (listof sexpr) arrow (listof number) throws error
;; Purpose: To evaluate the sexprs in the given list
(define (eval-args a-losexpr)
(if (empty? a-losexpr)
'()
(local [(define first-arg (first a-losexpr))
(define rest-args (rest a-losexpr))]
(cons (eval-sexpr first-arg)
(eval-args rest-args)))))
As you solve a problem remember that the goal of a program is not only to
produce a solution. An equally important goal is to clearly communicate how a
problem is solved. The use of local-expressions can be very useful to make clear
what expressions represent.

*** Ex. 272 — Section 107.1 discusses the design of process-blanks.

Refactor this function to make it more readable.
**** Ex. 273 — Section 107.2 discusses the design of node-can-win?.
Refactor this function to make it more readable. Make sure not to unneces-
sarily evaluate any arguments for and and or. Your refactored code should not
do more work than the original function.

*** Ex. 274 — Section 98 discusses the design of btstring-extract.

Refactor this function to make it more readable.

121.3 Furthering Functional Abstraction

Consider the function to move a (listof alien):

;; (listof alien) dir → (listof alien)
;; Purpose: To move the given list of aliens in the
;; given direction
(define (move-loa a-loa dir)
(if (empty? a-loa)
E-LOA
(cons (move-alien (first a-loa) dir)
(move-loa (rest a-loa) dir))))
121 Using Local-Expressions 491

The same function, move-alien, is applied to every alien in the given list and a
list of the results is returned. This suggests using map to re-implement this function.
Recall, however, the signature for map:
(X → Y) (listof X) → (listof Y)
It expects a one-input function as input. Do you see the problem? Look at the
signature for move-alien:
alien direction → alien
It is a two-input function and, therefore, cannot be given as input to map.
Observe, however, that to move an alien the only thing that varies is the alien to
be moved. The direction is the same for all aliens. Given that only one thing varies,
this suggests that inside move-loa the moving of an alien may be implemented as a
one-input function. This function may be defined locally and given to map as input:
;; (listof alien) dir → (listof alien)
;; Purpose: To move the given list of aliens in the
;; given direction
(define (move-loa a-loa dir)
(local [(define (alien-mover an-alien)
(move-alien an-alien dir))]
(map alien-mover a-loa)))
Why is such refactoring possible? Note that the value of dir does not change in the
recursive call. Therefore, inside move-loa dir may be considered a constant value
and a simpler function may be defined that does not have dir as input. Instead, dir
becomes a free variable inside the new local function. In general, within a function
any variables that do not change from one recursive call to the next may be treated
as constants to define a local function in which these variables are free variables. In
this case, it makes it possible to refactor a two-input function to move an alien into
a one-input function in which the direction to move the alien in is a free variable.
A note of caution is in order at this time. The local-expression in move-loa can-
not be eliminated using the draft refactoring recipe presented in Sect. 119. The prob-
lem is that alien-mover has a free variable, dir, that only exists inside move-loa.
Therefore, alien-mover cannot simply be moved outside the scope of dir. We
shall discuss how to solve this problem in Chap. 22.

** Ex. 275 — Refactor move-posns to use map. Write and run tests to validate
your transformation.
;; number number (listof posn) → (listof posn)
;; Purpose: Move the posns by the given amounts
(define (move-posns deltax deltay a-loposn)
(if (empty? a-loposn)
'()
(cons (move-posn deltax deltay (first a-loposn))
(move-posns deltax deltay (rest a-loposn)))))
492 21 Encapsulation

**** Ex. 276 — Refactor get-all-in-interval to use filter. Write and

run tests to validate your transformation.
;; [int..int] (listof number) → (listof number)
;; Purpose: Extract the list numbers that are in the
;; given interval
(define (get-all-in-interval low high a-lon)
(cond [(empty? a-lon) '()]
[(<= low (first a-lon) high)
(cons (first a-lon)
(get-all-in-interval low high (rest a-lon)))]
[else (get-all-in-interval low high (rest a-lon))]))

**** Ex. 277 — Refactor all-outside-interval? to use andmap. Write

and run tests to validate your transformation.
;; [int..int] (listof number) → Boolean
;; Purpose: Determine if all list numbers are outside
;; given interval
(define (all-outside-interval? low high a-lon)
(or (empty? a-lon)
(and (not (<= low (first a-lon) high))
(all-outside-interval? low high (rest a-lon)))))

121.4 One-Time Expression Evaluation

It may be surprising to you that local-expressions may be used to improve program

performance. That is, the use of local-expressions may make a program faster.
Programs may be noticeably slowed down by having to compute the value of an
expression more than once. In fact the abstract running time of a program may be
worsened by programs that evaluate the same expression multiple times.
Consider, for example, the program to return the last item in a given (listof
item) that costs less than a given price displayed in Fig. 101. Observe that the
expression (get-last a-price (rest a-loi))) may be evaluated twice for
every call to get-last. This expression is evaluated once to test if the result is
empty and then again if the result is not empty. At a glance you can immediately tell
this is inefficient. It is simply silly to evaluate the same expression to get the same
value twice.
121 Using Local-Expressions 493

Fig. 101 A program to get last item under a given price

;; An item is a structure, (make-item string number), with a name & price.
(define-struct item (name price))

;; A result is either: 1. () or 2. item

;; Sample items
(define IT0 (make-item "ball" 10)) (define IT1 (make-item "doll" 8))
(define IT2 (make-item "bat" 12)) (define IT3 (make-item "glove" 15))
(define IT4 (make-item "cap" 7)) (define IT5 (make-item "bike" 100))

;; Sample (listof item)

(define LOI0 ())
(define LOI1 (list IT0 IT1 IT2 IT3 IT4 IT5))
(define LOI2 (list IT1 IT2 IT3 IT5))
(define LOI3 (list IT1 IT2 IT4 IT5))

;; number (listof item) result Purpose: Get last item < given price
(define (get-last a-price a-loi)
(cond [(empty? a-loi) ()]
[(< (item-price (first a-loi)) a-price)
(if (empty? (get-last a-price (rest a-loi))) (first a-loi)
(get-last a-price (rest a-loi)))]
[else (get-last a-price (rest a-loi))]))

;; Sample expressions for get-last

(define L0-V ()) (define L1-V (get-last 10 (rest LOI1)))
(define L2-V (first L2)) (define L3-V (get-last 11 (rest LOI3)))

;; Tests using sample computations for get-last

(check-expect (get-last 120 L0) L0-V)
(check-expect (get-last 10 L1) L1-V)
(check-expect (get-last 9 L2) L2-V)
(check-expect (get-last 11 L3) L3-V)

;; Tests using sample values for get-last

(check-expect (get-last 150
(list (make-item "pen" 100)
(make-item "ring" 600)
(make-item "knife" 125)))
(make-item "knife" 125))
(check-expect (get-last 120
(list (make-item "pen" 100)
(make-item "ring" 600)
(make-item "knife" 125)))
(make-item "pen" 100))
494 21 Encapsulation

Instead of evaluating the same expression multiple times to obtain the same value,
locally define a variable and only evaluate the expression once. In the remainder of
the code reference the local variable instead of evaluating the expression again.
Following this principle get-last is refactored to
;; number (listof item) → result
;; Purpose: Get last item < given price
(define (get-last a-price a-loi)
(cond [(empty? a-loi) '()]
[(< (item-price (first a-loi)) a-price)
(local [(define rest-result (get-last a-price
(rest a-loi)))]
(if (empty? rest-result)
(first a-loi)
rest-result))]
[else (get-last a-price (rest a-loi))]))
Observe that rest-result captures the value obtained from processing the rest of
the list of items. Therefore, this local variable may be used instead of processing the
rest of the list of items multiple times.
Is this type of refactoring nitpicking or can it have a significant impact on per-
formance? To answer this question let us identify the input that exhibits the worst
possible performance. This occurs when the price given to get-last is larger than
the price of all the items in the given list. In such a case there is always an item with
a smaller price in the rest of a non-empty list. Let us define a list of with all identical
items using interval-accum-l2h (from Sect. 117) as follows:
;; int → item
;; Purpose: Returns a doll item that costs 30
(define (make-doll-item n) (make-item "doll" 30))

;; Sample expressions for make-doll-item

(define DOLL1 (make-item "doll" 30))
(define DOLL2 (make-item "doll" 30))

;; Tests using sample computations for make-doll-item

(check-expect (make-doll-item 4) DOLL1)
(check-expect (make-doll-item 8) DOLL2)

;; Tests using sample values for make-doll-item

(check-expect (make-doll-item -5) (make-item "doll" 30))
(check-expect (make-doll-item 0) (make-item "doll" 30))

(define LOIL
(interval-accum-l2h '() make-doll-item cons 0 19))
121 Using Local-Expressions 495

Fig. 102 A tree of natural numbers

(get-last 1200 LOIL)

(get-last 1200 (rest LOIL)) (get-last 1200 (rest LOIL))

(get-last 1200 (get-last 1200 (get-last 1200 (get-last 1200

(rest (rest LOIL))) (rest (rest LOIL))) (rest (rest LOIL))) (rest (rest LOIL)))

LOIL is a (short) list of 20 identical items. If we rename the second version of the
function as get-last2, we can time the two versions as follows:
(time (get-last 1200 LOIL))
(time (get-last2 1200 LOIL))
The results obtained for CPU time are
Function CPU Time
get-last 1547
get-last2 0

On your computer the measured numbers might vary, but it is undeniable that there
is a significant difference in actual running time in favor of the version that uses a
local-expression.
How is this explained? Why is the get-last2 much faster? To answer this let us
look at the abstract running time of both functions. To ascertain the abstract running
time for get-last let us visualize the beginning of the function calls made. These
calls are displayed in Fig. 102. The initial call to get-last generates two calls using
the rest of the given list. Each of these two calls generates two calls with the rest
of the given list. We can extrapolate that every call to get-last will generate two
more recursive calls until the list is empty. How many calls are made for a list of
length n? The binary tree of calls in Fig. 102 would have height n (remember that
the root of the binary tree is at height 0). Each node in the binary tree represents a
function call. The binary tree would be a full binary tree (i.e., all leaves are at level
n). Therefore, the number of calls is equal to the number of nodes in a full binary
tree. The following table illustrates the number of nodes in a full binary tree of height
h:
h Number of Nodes
0 0
1 3
2 7
3 15
4 31
.. ..
. .
496 21 Encapsulation

From the table we can extrapolate that a full binary tree of height h has 2h+1 − 1
nodes. This means that the number of function calls for a list of length n is 2n+1 − 1
making get-last O(2n ). That is, the amount of work done by get-last grows
exponentially as the length of the input list grows.
What is the abstract running time of get-last2? Observe that for every function
call there is at most only one recursive call generated. This means that the total
number of function calls for a list of length n is n. This makes get-last2 O(n). In
other words, the work done is proportional to the size of the list. This explains why
get-last2 is much faster than get-last.
The lesson you should walk away with is that evaluating the same expression mul-
tiple times may have a profound impact on abstract and actual running time. It is not
the case, however, that eliminating the repetitive evaluation of the same expression
always leads to better abstract running time. In fact, it may not even have a noticeable
impact on actual running time. Another thing worth noting is that a program that has
an exponential abstract running time is not a practical solution to a problem. The
amount of work done grows too quickly and makes the computation of a solution
unfeasible even for input of moderate size. In future courses in Computer Science
you will study techniques to approximate a solution using, for example, heuristics
that have a polynomial (not exponential) abstract running time.

*** Ex. 278 — The powerset of a set, S, is the set of all subsets of S, including
the empty set and S itself. A colleague has written the following function to
compute the powerset of a set (foolishly not all the steps of the design recipe
have been followed):
;; A set is a (listof X)

;; set → (listof set)

;; Purpose: Return the power set of the given set
(define (powerSet a-set)
(if (empty? a-set)
(list '())
(local [(define (add2subset a-subset)
(cons (first a-set) a-subset))]
(append (powerSet (rest a-set))
(map add2subset (powerSet (rest a-set)))))))

(check-expect (powerSet '()) (list '()))

(check-expect (powerSet '(1 2))
(list '() (list 2) (list 1) (list 1 2)))
(check-expect (powerSet '(a b c))
'(() (c) (b)
(b c) (a) (a c)
(a b) (a b c)))
122 What Have We Learned in This Chapter? 497

If you run the code the tests pass. However, your colleague believes there is
a bug in the code. The following use of the function seems to never return an
answer or simply takes too long to be of practical use:
(powerSet '(a b c d e f g h i j k l m n o p q r s
t u v w x y z))

Help your colleague to make the function faster.

** Ex. 279 — Design and write a function that takes as input an x-value and a
list of posns and that returns the last posn in the given list that has the given
x-value. If no such posn exists, then your function ought to return #false.

**** Ex. 280 — Your friend is studying Computer Science at a university that
does not teach students about program design. Your friend wrote the following
function that takes as input a (btof number) and then returns the maximum
between the sum of the left and right subtrees:
(define (bt-sum t)
(if (empty? t)
0
(+ (first t)
(first (rest t))
(first (rest (rest t))))))

(define (leftsum>rightsum? a-numbt)

(if (> (numbt-sum (second a-numbt))
(numbt-sum (third a-numbt)))
(numbt-sum (second a-numbt))
(numbt-sum (third a-numbt))))

You are horrified, of course, by such unreadable code that computes the value
of the same expression more than once. Redesign the code to make it more
readable and avoid evaluating expressions more than once by following the
steps of the design recipe.

122 What Have We Learned in This Chapter?

The important lessons of Chap. 21 are summarized as follows:

• Encapsulation means that we package together all related functions as a single
piece of software.
• A local-expression allows us to define new variables and functions inside of
functions.
• The local-expression syntax is expr ::= (local [def∗ ] expr)
498 21 Encapsulation

• No new computational power is provided by local-expressions. They are syn-

tactic sugar to help programmers better organize and design programs.
• The lexical scope of a variable is the section of the program where it is valid.
• Within its lexical scope a variable may be shadowed by a declaration of the same
variable.
• In lexical scoping a reference is always to the nearest declaration.
• Encapsulate functions only after thoroughly testing them.
• Local definitions can improve communicating how a problem was solved.
• Local definitions can aid in furthering the use of abstract functions by defining
functions where only values that change are parameters and that have one or
more free variables whose values do not change for a given function call (like in
move-loa).
• A repeated expression inside a function may be eliminated by defining a lo-
cal variable for its value and referencing the variable instead of evaluation the
expression multiple times.
• Repeated expressions inside a function may make it have an exponential abstract
running time and, therefore, be an impractical solution to a problem.
Chapter 22
Lambda Expressions

Section 115 introduces the idea that functions are a type of value just like numbers,
strings, and images are types of values. It presents how functions may be passed
as arguments to other functions. This ability allows us to develop powerful abstract
functions like map, filter, and ormap. This chapter continues the exploration of
functions as data. Specifically, it explores how to create a function value and how
to return a function as the value of a function application. This ability has profound
implications for problem solving and program design. For example, it leads to
functions that compute functions and to object-oriented programming (as Chap. 25
discusses).
Why would you want to compute a function in the first place? The short answer is
that it is a common operation in human life. Consider, for example, the composition
of two functions as you have studied in high school Mathematics. You may recall
how function composition is typically written:
(f ◦ g)(x) = f(g(x))
This equation is stating that the composition of f and g provides the output of g
as input to f. Although it is more subtle and not very emphasized in a high school
Mathematics course, the composition of f and g is a function. If we have two
functions f and g, the above equation tells us how to build a new function that we
call the composition of f and g. For example, Fig. 103 displays two functions. One
adds four to its input and the other doubles its input. If you were asked to compose
add4 and double (i.e., write a function that adds 4 to the doubling of its input) and
to compose double and add4 (i.e., write a function that doubles the sum of its input
and 4), your code may look as displayed in Fig. 104.
Observe that in Fig. 104 add4-o-double and double-o-add4 are good can-
didates for abstraction. As you expect the differences are the composed functions.

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 499
M. T. Morazán, Animated Problem Solving, Texts in Computer Science,
https://doi.org/10.1007/978-3-030-85091-3_22
500 22 Lambda Expressions

Fig. 103 Functions to add 4 and double a given number

;; number number Purpose: Add 4 to the given number
(define (add4 x) (+ x 4))

;; Sample expressions for add4

(define ADD4-VAL1 (+ 5 4)) (define ADD4-VAL2 (+ -7 4))

;; Tests using sample computations for add4

(check-expect (add4 5) ADD4-VAL1) (check-expect (add4 -7) ADD4-VAL2)
;; Tests using sample values for add4
(check-expect (add4 0) 4) (check-expect (add4 -10) -6)

;; number number Purpose: Double the given number

(define (double x) (* 2 x))

;; Sample expressions for double

(define DBL-VAL1 (* 2 10)) (define DBL-VAL2 (* 2 -3))

;; Tests using sample computations for double

(check-expect (double 10) DBL-VAL1) (check-expect (double -3) DBL-VAL2)
;; Tests using sample values for double
(check-expect (double -7) -14) (check-expect (double 65) 130)

Using the design recipe for functional abstraction yields the following abstract func-
tion and refactored functions:
;; (number → number) (number → number) number → number
;; Purpose: Return f(g(x))
(define (f-o-g f g x) (f (g x)))

;; number → number
;; Purpose: Add 4 to the double of the given number
(define (add4-o-double x) (f-o-g add4 double x))

;; number → number
;; Purpose: Double the sum of the given number and 4
(define (double-o-add4 x) (f-o-g double add4 x))
All tests pass after running them. The abstract function, however, is not quite what
we want. Observe that x is a parameter in the abstract function, but not a difference in
the original functions. We want f-o-g to return a function and not the value of said
function for some value x. This means x should not be a parameter in the abstract
function. In other words, we want an abstract function with the following signature
and purpose:
;; (number → number) (number → number) → (number → number)
;; Purpose: Return the composition of the given functions
(define (f-o-g f g) . . .)
123 Anonymous Functions 501

Fig. 104 Two ways of composing add2 and double

;; number number Purpose: Add 4 to the double of the given number
(define (add4-o-double x) (add4 (double x)))

;; Sample expressions for add4-o-double

(define ADD4-DBL-VAL1 (add4 (double 0)))
(define ADD4-DBL-VAL2 (add4 (double -1)))

;; Tests using sample computations for add4-o-double

(check-expect (add4-o-double 0) ADD4-DBL-VAL1)
(check-expect (add4-o-double -1) ADD4-DBL-VAL2)
;; Tests using sample values for add4-o-double
(check-expect (add4-o-double -8) -12)
(check-expect (add4-o-double 50) 104)

;; number number Purpose: Double the sum of the given number and 4
(define (double-o-add4 x) (double (add4 x)))

;; Sample expressions for double-o-add4

(define DBL-ADD4-VAL1 (double (add4 10)))
(define DBL-ADD4-VAL2 (double (add4 -2)))

;; Tests using sample computations for double-o-add4

(check-expect (double-o-add4 10) DBL-ADD4-VAL1)
(check-expect (double-o-add4 -2) DBL-ADD4-VAL2)
;; Tests using sample values for double-o-add4
(check-expect (double-o-add4 40) 88)
(check-expect (double-o-add4 -6) -4)

123 Anonymous Functions

You probably suspect already that returning a function is easy. You simply need to
return the function you want just like you return a number or a string of your choice.
This is partially correct. Consider, for example, the following program:
;; number → (number → number)
;; Purpose: Return + if the given number is even.
;; Otherwise, return -.
(define (choose-f x) (if (even? x) + -))

(check-expect ((choose-f 4) 1 5) 6)
(check-expect ((choose-f 3) 1 5) -4)
The function choose-f returns a function. Specifically, it returns + if its input is
even and it returns - if its input is odd. Therefore, ((choose-f 4) 1 5) is (+
1 5) and ((choose-f 3) 1 5) is (- 1 5). The application of choose-f to a
number returns a function that is then applied to the given arguments.
502 22 Lambda Expressions

The above example works because it handles functions that exist and have a
name. This is why it is easy to return a function. If a function has a name, it may
be returned as the value of a function. In other words, a function name is simply a
variable whose value is a function. In contrast, we want f-o-g to return a function
that does not exist in the program or in the programming language and, of course,
does not have a name. Therefore, we need an expression that returns a function when
it is evaluated. Such an expression is a lambda-expression (or λ-expression). The
syntax for a λ-expression in ISL+ is
expr ::= (lambda (symbol+ ) expr)
::= (λ (symbol+ ) expr)
These two expressions mean exactly the same thing. To type a "λ" in DrRacket use
Ctrl-\ (hold down the control key and press the backslash key). In parenthesis, a
λ-expression has the keyword lambda or the symbol λ followed by a list of symbols
and an expression. The list of symbols are the parameters of the function and the
expression is the body of the function. For example, the following expression returns
a function that adds 10 to its input:
(λ (a-number) (+ a-number 10))
We may apply this function to a number just like any other function that takes as
input a number. Consider, for example, the following expression:
((λ (a-number) (+ a-number 10)) 25)
When this application expression is evaluated, the λ-expression returns a function
that has a parameter called a-number and that has a body that adds 10 to a-number.
This function is applied to 25 to obtain 35 (the value of the application expression).
In essence, a λ-expression allows us to create an anonymous function. That is a
function without a name. If we wish to give such a function a name, then we need to
use define. For example, the function above to add 10 to its input may be bound to
a name as follows:
(define add10 (λ (a-number) (+ a-number 10)))
It is now possible for you to understand that
(define (<name> <name>$^+$) expr)
is syntactic sugar for
(define <name> (lambda (<name>$^+$) expr))
In other words, define binds a variable to value. The value may be any valid value
in ISL+ including functions.
You should consider using a λ-expression on its own (i.e., without binding its
value to a variable) when:
• A new function is needed.
• The needed function is not recursive.
• The needed function is only used once.
124 Revisiting Function Composition 503

Keep in mind that a λ-expression evaluates to an anonymous function. Therefore, it

cannot recursively call itself. In order for a function to be recursive, it must have a
name. Commonly, a λ-expression is used to provide an argument to a function or to
return a function as the value of a function call.

** Ex. 281 — Write a λ-expression to create a function that computes f(x) =

5x3 + 10x2 + 2x + 31. Use it to compute f(0) and f(10).

*** Ex. 282 — Write a λ-expression to create a function that returns the ab-
solute value of its input. Use it to compute the absolute value of 0, 22, and
-40.
**** Ex. 283 — Write a λ-expression to create a function that returns the
second largest value of three given numbers. Apply it to 87, 27, and 100.

*** Ex. 284 — Write a λ-expression to create a function that moves a shot for
Aliens Attack. Apply it to (make-posn 4 5) and to NO-SHOT.

*** Ex. 285 — Write a λ-expression to create a predicate to determine if a

given number is in a given interval. Apply it to 10 and [89..100] and to 67
and [50..85].
*** Ex. 286 — An item has a name and a price. Write a λ-expression to create
a predicate to determine if a given item’s price is less than a given price. Write
application expressions that illustrate that the function works.

124 Revisiting Function Composition

Given add4 and double, we need two functions: (add4 ◦ double)(x) and
(double ◦ add4)(x). We can define these functions using a λ-expression as fol-
lows:
;; number → number
;; Purpose: Add 4 to the double of the given number
(define add4-o-double (λ (x) (add4 (double x))))

;; number → number
;; Purpose: Double the sum of the given number and 4
(define double-o-add4 (λ (x) (double (add4 x))))
Observe that using a λ-expression does not change the steps of the design recipe.
You can substitute these new definitions for the definitions in Fig. 104 and all the
tests will pass. Try it out!
504 22 Lambda Expressions

Observe that the two expressions are nearly identical. This strongly suggests
abstracting over the expressions. This abstraction step yields the following function:
;; (number → number) (number → number) → (number → number)
;; Purpose: Return f(g(x))
(define (f-o-g f g) (λ (x) (f (g x))))
We now have the correct abstraction. We have a function that given two functions
returns a function for the composition of the two functions. We can now refactor the
original definitions to use the abstraction to get
;; number → number
;; Purpose: Add 4 to the double of the given number
(define add4-o-double (f-o-g add4 double))

;; number → number
;; Purpose: Double the sum of the given number and 4
(define double-o-add4 (f-o-g double add4))
You can now add f-o-g to and substitute these two functions in Fig. 104. All the
test will pass. Try it out!
Take some time to ponder what has been accomplished. We can design functions
that compute functions. This means we can now use functions that we did not write.
We use functions that were computed for us. This is a powerful tool in your design
arsenal. A program can create specialized functions as needed. In fact, the abstraction
provided by f-o-g is so useful that it is provided by ISL+ as compose.

** Ex. 287 — In high school Mathematics you learned that an invertible func-
tion, f, maps a value from its domain to a value in its range. The inverse of f,
f-1 , maps the range value back to the domain value. This suggests that the com-
position of f and f-1 is the identity function (implemented as id in Sect. 116.1).
Write a program to test this hypothesis. Your different definitions for the identity
function and corresponding tests ought to look like this:
;; number → number
;; Purpose: Return the given number
(define id-f1 (make-id f1 f1−1 ))

;; number → number
;; Purpose: Return the given number
(define id-f2 (make-id f2 f2−1 ))

(check-expect (id-f1 100) 100)

(check-expect (id-f2 4) 4)
125 Curried Functions 505

*** Ex. 288 — You have a software company that supports manufacturing
companies. The total costs of any manufacturing company is given by the sum
of its variable costs and its fixed costs. The variable cost is proportional to the
number of units produced each month, while fixed costs are the same every
month. For example, assume that for company C1 the cost per unit is $5 and
the fixed costs are $100 and that for company C2 the cost per unit is $3 and the
fixed costs are $75. The respective functions for the total costs of companies C1
and C2 may be written as follows:
;; number → number
;; Purpose: Return the total costs of C1
(define (total-costs-C1 units-produced)
(+ (* units-produced 5) 100))

(check-expect (total-costs-C1 10) 150)

;; number → number
;; Purpose: Return the total costs of C1
(define (total-costs-C2 units-produced)
(+ (* units-produced 3) 75))

(check-expect (total-costs-C2 20) 135)

As the number of companies you develop total costs functions for grows, you
discover that typing similar functions over and over is a waste of your time.
Abstract over the above functions to develop an abstract function that computes
a total cost function. Refactor the above functions to make use of your abstract
function.

125 Curried Functions

Creating specialized recursive functions as a program is evaluated may be imple-

mented using both local- and λ-expressions. Consider the following functions from
Aliens Attack to draw a list of aliens and a list of shots:
;; loa scene → scene
;; Purpose: To draw the given loa in the given scene
(define (draw-loa a-loa scn)
(if (empty? a-loa)
scn
(draw-alien (first a-loa)
(draw-loa (rest a-loa) scn))))
506 22 Lambda Expressions

;; los scene → scene

;; Purpose: To draw the given los in the given scene
(define (draw-los a-los scn)
(if (empty? a-los)
scn
(draw-shot (first a-los)
(draw-los (rest a-los) scn))))
These functions are clear candidates for abstraction. The only significant difference
between them is the function used to draw the first list element. Performing the
abstraction step yields the following function:
;; (listof X) image → image
;; Purpose: To draw the given list in the given scene
(define (draw-lox draw-x a-lox scn)
(if (empty? a-lox)
scn
(draw-x (first a-lox)
(draw-lox draw-x (rest a-lox) scn))))
Given an X-drawing function draw-lox draws a given (listof X) in the given
scene. Observe that draw-lox has a structure very similar to accum from Sect.116.1.
It can easily be refactored to have the same structure using id (also from Sect. 116.1)
as follows:
;; (listof X) image → image
;; Purpose: To draw the given list in the given scene
(define (draw-lox draw-x a-lox scn)
(if (empty? a-lox)
scn
(draw-x (id (first a-lox))
(draw-lox draw-x (rest a-lox) scn))))
This means draw-lox may be refactored using accum as follows:
;; (listof X) image → image
;; Purpose: To draw the given list in the given scene
(define (draw-lox draw-x a-lox scn)
(accum scn id draw-x a-lox))
Refactoring the functions abstracted over yields
;; (listof alien) scene → scene
;; Purpose: To draw the given loa in the given scene
(define (draw-loa a-loa scn) (draw-lox draw-alien a-loa scn))

;; (listof shot) scene → scene

;; Purpose: To draw the given los in the given scene
(define (draw-los a-los scn) (draw-lox draw-shot a-los scn))
125 Curried Functions 507

Observe that there is still a lot of repetition among these two functions. The only
significant difference is the function used to draw either an alien or a shot. Ab-
stracting over these functions suggests creating a function that may be outlined as
follows:
;; (X → image) → ???
;; Purpose: ???
(define (draw-lox-maker draw-x) (. . .))
What should such a function return? This proposed abstract function only takes as
input a drawing function. This drawing function by itself is of little use without the
list of elements to draw and the scene to draw them into. This means that this list
and scene are still needed as input. If they are needed as input, then draw-lox must
return a function that consumes these inputs. The returned function must capture
the similarities in draw-loa and draw-los. With this understanding the abstraction
step yields
;; (X image → image) → ((listof X) image → image)
;; Purpose: Return a function to draw a (listof X)
(define (draw-lox-maker draw-x)
(λ (a-lox scn) (draw-lox draw-x a-lox scn)))
The contract states that this function takes as input a function that consumes an
X and an image and that returns an image (i.e., a drawing function). It returns a
function that consumes a (listof X) and an image and that returns an image (i.e.,
a function to draw the given list). The resulting function is obtained by evaluating
a λ-expression.27 Observe that draw-lox-maker returns a specialized list-drawing
function.
The abstract function draw-lox-maker is an example of a curried function.
A curried function is one that consumes part of its input and returns a function
that consumes the rest of the input. In this example, draw-lox-maker consumes a
drawing function and returns a function that consumes a list of elements to draw and
an image. The returned function is specialized using the value given as input. For
example, consider the result of refactoring the functions abstracted over:
;; loa scene → scene
;; Purpose: To draw the given loa in the given scene
(define draw-loa (draw-lox-maker draw-alien))

;; los scene → scene

;; Purpose: To draw the given los in the given scene
(define draw-los (draw-lox-maker draw-shot))

27
The body of draw-lox-maker could also be implemented using a local-expression that defines
the returned function.
508 22 Lambda Expressions

What are draw-loa and draw-los? Like any other function, we may plug-in the
given arguments to draw-lox-maker and substitute the result into the above defi-
nitions:
;; loa scene → scene
;; Purpose: To draw the given loa in the given scene
(define draw-loa (λ (a-lox scn)
(draw-lox draw-alien a-lox scn)))

;; los scene → scene

;; Purpose: To draw the given los in the given scene
(define draw-los (λ (a-lox scn)
(draw-lox draw-shot a-lox scn)))
Observe that if we use the define syntactic sugar and change the name of a-lox,
we obtain exactly the functions that were abstracted over.
Curried functions allow programmers to stage providing the inputs to a function.
This allows for the partial evaluation of functions to produce specialized functions
(as done by draw-lox-maker). Do you feel this is strange? Consider that without
knowing it you have been using curried functions. The big-bang expression used
to create video games and simulations is an example of currying in action. As a
programmer you provide the functions, for example, to process keystrokes and to
process clock ticks. Later a player and the computer provide the rest of the input
(i.e., the keystrokes and the world).
The power of curried functions allows us to revisit move-loa from Sect. 121.3:
;; (listof alien) dir → (listof alien)
;; Purpose: To move the given list of aliens in the
;; given direction
(define (move-loa a-loa dir)
(local [(define (alien-mover an-alien)
(move-alien an-alien dir))]
(map alien-mover a-loa)))
Recall that we were unable to remove the local-expression because alien-mover
has a free variable (i.e., dir). To remove the local-expression an alien-mover
function specialized with the value of dir is needed. This can be done using a curried
function. When a local function has free variables, a function that takes as input the
free variables and that returns a specialized function is needed. In this case, the
function needs to consume dir and return a specialized function for alien-mover.
The refactored code without the local-expression is
;; dir → (alien → alien)
;; Purpose: Return a function to move an alien in the
;; given direction
(define (alien-mover-maker a-dir)
(λ (an-alien) (move-alien an-alien a-dir)))
125 Curried Functions 509

;; (listof alien) dir → (listof alien)

;; Purpose: To move the given list of aliens in the
;; given direction
(define (move-loa a-loa dir)
(map (alien-mover-maker dir) a-loa))
Observe that for the first input to map a specialized alien-mover function is com-
puted using the value of dir. The function alien-mover-maker contains no free
variables and, therefore, may be moved outside the local-expression allowing for
the elimination of said expression.
A cautionary note is warranted at this point. Computing a function’s complete
set of free variables and lifting it out of its lexical scope is a nontrivial process.
Such lifting becomes harder when local functions are mutually recursive. The free
variables of a function f are the free variables referenced in its body and the free
variables in f’s scope that are referenced by any function reachable from f by one or
more function calls. The process to lift functions outside of their scope is called λ-
lifting and you may study it in a future course on the implementation of programming
languages or on your own.

*** Ex. 289 — You have a physicist friend who routinely writes functions to
scale a list of numbers as the following (pardon the physicist who does not know
about the steps of the design recipe):
(define (scale-by-2 L) (map (lambda (x) (* 2 x)) L))

(define (scale-by-5 L) (map (lambda (x) (* 5 x)) L))

(define (scale-by-3 L) (map (lambda (x) (* 3 x)) L))

..
.

She complains that programming is too tedious because she must write almost
the same code over and over again. Help your friend by writing a curried function
to scale a list of numbers. Refactor the above functions to show your friend how
all the list-scaling functions can be computed for her.

*** Ex. 290 — In Sect. 117 the following functions that accumulate a value by
processing an interval from low to high are developed:
;; [int..int] → (listof string)
;; Purpose: Construct the list of strings for the ints in
;; the given interval
(define (interval->lostr low high)
(interval-accum-l2h '() number->string cons low high))

;; [int..int] → int
;; Purpose: Compute the product of the squares of the ints
;; in the given interval
(define (interval-product-sqrs low high)
(interval-accum-l2h 1 sqr * low high))
510 22 Lambda Expressions

Observe the similarities between the two functions. Develop a curried function
that produces specialized functions to accumulate a value by processing an
interval from low to high. The specialized functions produced should only take
an interval as input.

*** Ex. 291 — The following functions combine three images:

;; image image image → image
;; Purpose: Create an image by putting the images
;; above each other
(define (above-3imgs img1 img2 img3)
(above img1 img2 img3))

(check-expect (above-3imgs (circle 10 'outline 'red)

(circle 20 'outline 'red)
(circle 40 'outline 'red))

;; image image image → image

;; Purpose: Create an image by putting the images beside
;; each other
(define (beside-3imgs img1 img2 img3)
(beside img1 img2 img3))

(check-expect (beside-3imgs (circle 10 'outline 'red)

(circle 20 'outline 'red)
(circle 40 'outline 'red))

Develop a curried function that produces specialized functions to combine

three images. Refactor the above functions to use your curried function. Use
your curried function to define a function that puts the first image above the
image containing the other two images besides each other.
126 Designing Using Existing Abstractions 511

126 Designing Using Existing Abstractions

You may have the impression that designing abstract functions is a lot of work. You
may feel that it always requires abstracting over similar functions and refactoring
the functions abstracted over. Without a doubt this is an effective way to write
more elegant and easier to maintain code. An abstract function, however, may be
directly designed just like any other function. This section explores two examples
that illustrate how this can be done. If an abstraction already exists in another
domain, it can be directly designed and implemented. The two examples explored
are computing a series and computing an approximation for π.

126.1 Computing the Value of a Series

In Mathematics, a set of ordered terms is called a sequence. For example, the

sequence of the first 5 odd natural numbers is
1 3 5 7 9
This sequence is finite. A sequence may also be infinite. For example, you have
probably seen the sequence of odd natural numbers written as follows:
1 3 5 7 9 ...
Each term in a sequence may be computed using its index in the sequence. Indexing
starts at 0. This means, for example, that the 0th odd natural number is 1 and the 4th
odd natural number is 9. The following function computes the ith element of the
sequence of odd natural numbers:
;; natunum → natnum Purpose: Return ith odd number
(define (ith-odd i) (add1 (* 2 i)))

;; Sample expressions for ith-odd

(define 0TH-ODD (add1 (* 2 0)))
(define 9TH-ODD (add1 (* 2 9)))

;; Tests using sample computations for ith-odd

(check-expect (ith-odd 0) 0TH-ODD)
(check-expect (ith-odd 9) 9TH-ODD)

;; Tests using sample values for ith-odd

(check-expect (ith-odd 2) 5)
(check-expect (ith-odd 4) 9)
The sum of the elements of a sequence is called a series. Summing up the elements
of a sequence may provide the means by which to compute a number. For example,
512 22 Lambda Expressions

the sum of the first n elements of the sequence of natural odd numbers is equal to
n2 . We will not prove this here, but we will illustrate it with the following examples:
02 = 0

32 = 1 + 3 + 5

52 = 1 + 3 + 5 + 7 + 9
Mathematicians would never explicitly write out all the elements of the sum for two
reasons. The first is that it is too cumbersome. The second is that it is impossible to
do so when the sequence is infinite. Instead, they have created syntax to abstractly
capture and compactly write the idea of summing the terms of a sequence:
02 = 0

32 = Σi=0
2 (ith-odd i)

52 = Σi=0
4 (ith-odd i)

Look closely at what the mathematical notation is communicating. Adding zero

terms is 02 . For all other natural numbers a square is obtained by adding n terms
produced using ith-odd. It is not difficult to see that a function for the terms for
any sequence may be used. In fact, we may write
n−1
seriesf (n) = Σi=0 (f i)
Observe that this defines a curried function. It first takes as input the term function
and then to actually compute a value it takes the number of terms to add. We can
implement a function to return a series-computing function as follows:
;; (natnum → number) → (natnum → number)
;; Purpose: Return a function to compute the series for
;; the sequence defined by the given function
(define (series term-f)
(lambda (n)
(local [;; natnum → number
;; Purpose: Compute the series using the given
;; number of terms
(define (add-terms n)
(if (= n 0)
(term-f 0)
(+ (term-f n) (add-terms (sub1 n)))))]
(if (= n 0)
(term-f 0)
(add-terms (sub1 n))))))
This function returns a function to compute the value of a series that is specialized for
the given term function. The returned function expects as input, n, a natural number
126 Designing Using Existing Abstractions 513

(i.e., the number of terms to add). If n is 0, it returns the 0th term. Otherwise, it
adds the first n terms using structural recursion. Observe that 1 is subtracted from
n to add the terms. This is done because the number of integers in [0..n-1] is n.
Given that recursion is needed to process a given natural number greater than 0, the
function that does this work must be locally defined.
It now becomes straightforward to define a function to compute n2 . All that is
needed is to give series the function ith-odd as input. The computed function
may be defined and tested as follows:
;; natnum → natnum
;; Purpose: Compute the square of the given number
(define n-square (series ith-odd))

;; Sample expressions for nth-square

(define ZERO-SQ (n-square 0))
(define THREE-SQ (n-square 3))

;; Tests using sample computations for nth-square

(check-expect (n-square 0) ZERO-SQ)
(check-expect (n-square 3) THREE-SQ)

;; Tests using sample values for nth-square

(check-expect (n-square 10) 100)
(check-expect (n-square 4) 16)
(check-expect (n-square 325) (sqr 325))
(check-expect (n-square 57) (sqr 57))
Observe that the tests using sample values use ISL+’s sqr function. This clearly
communicates that the functions are expected to be equivalent. It would be nice to
simply write:
(check-expect n-square sqr)
Remember, however, that establishing the equivalence of two functions is an unsolv-
able problem. Therefore, it is impossible to write such a test.

*** Ex. 292 — Redesign series using the following contract for add-terms:
;; natnum>=1 → number

Admittedly, computing a series is hardly impressive if all we do is compute the

square of a number. After all, we can use sqr or * to compute n2 . Let us continue to
explore the power of the series function.
Your Discrete Mathematics Professor proposes the following hypothesis: the
series of the first n cubes is always a square. Do you know if this is true or false? You
may or may not know the answer already. Regardless of that you can (very) quickly
write a program to test the hypothesis. The idea is to define a function to compute
514 22 Lambda Expressions

the series for cubes and test if the result is always a square. Here is a sample program
to do so:
;; natnum → natnum
;; Purpose: Return the sum of the first n cubes
(define sum-cubes (series (lambda (n) (* n n n))))

(check-expect (integer? (sqrt (sum-cubes 0))) #true)

(check-expect (integer? (sqrt (sum-cubes 5))) #true)
(check-expect (integer? (sqrt (sum-cubes 209))) #true)
Observe that a λ-expression is used to define a function that cubes its input because
it is not recursive and is only needed once. After running the tests you discover that
they all pass and you begin to suspect that the Professor’s hypothesis is true. In your
Discrete Mathematics course you will learn how to actually prove that the Professor’s
hypothesis is true. The lesson to walk away with is that abstractions offer you the
opportunity to easily experiment and test a hypothesis.

126.2 Approximating π

An irrational number, like π, cannot be exactly represented by a fraction. How then

can the value of π be represented? The answer is that it cannot. All that we can do
is approximate the value of π. In the fourteenth century the Indian mathematician
Madhava of Sangamagrama discovered an infinite series to approximate the value
of π (known today as the Madhava–Leibniz series, also named after the German
mathematician Gottfried Wilhelm (von) Leibniz). The infinite series is
∞ (4 ∗ −1i
π = Σi=0 (ith-odd i) )
This formula states that if we add an infinite number of terms the value of π is
obtained. It is, of course, impossible to add an infinite number of terms. Therefore,
the value of π is approximated by adding the first n terms of the series.
Using the abstract series function we can create a function to approximate π as
follows:
;; natnum → number
;; Purpose: Approximate pi using the given number of terms
(define pi-series (series (lambda (i)
(* 4 (/ (expt -1 i) (ith-odd i))))))
Observe that a λ-expression is used to define the term function given as input to
series. Once again, this is done because the term function is not recursive and is
only needed once.
The question now is to determine how to test this function. Given that an approx-
imation is being computed our tests need to illustrate that as the number of terms
126 Designing Using Existing Abstractions 515

used to approximate π increases the approximation gets better. This requires experi-
menting with the function to make sure it converges. That is, the approximation error
gets smaller as the number of terms grows. The following tests achieve this goal:
(check-within (pi-series 500) 3.1415 0.01) ;; 3.139592
(check-within (pi-series 3000) 3.1415 0.001) ;; 3.141259
(check-within (pi-series 12000) 3.1415 0.0001) ;; 3.141509
The tests clearly illustrate that as the number of terms increases the accuracy of the
results gets better. The values returned by pi-series are written as comments so
that any reader can easily see how the approximation of π is converging.
In closing, developing your design and abstraction skills using λ-expressions and
first-class functions is important. They are becoming more prevalent across all pro-
gramming languages. Virtually all functional programming languages (e.g., ISL+,
Racket, Haskell, Clean, and F#) have λ-expressions and support first-class func-
tions. This support is now also found in object-oriented programming languages
(e.g., Java, Kotlin, and Scala). In fact, now you may even use λ-expressions in
Excel.

**** Ex. 293 — The product of the terms of a sequence from a to b whose
elements are produced by f is defined as follows:
b f (i) = f(a) * f(a+1) * f(a+2) * . . . * f(b-1) * f(b)
Πi=a

b f (i). Use your curried

Design a curried function, pseries, to implement Πi=a
function to define a function to compute n!.

***** Ex. 294 — The number of ways to choose k elements out of a finite set
with n elements
n
is k n+1−i
k = Πi=1 i
n

k is read as n chooses k. Use your pseries function

from the previous
exercise to define a function to compute the value of nk . HINT: n is a free
variable in your term function.

*** Ex. 295 — Your Discrete Mathematics Professor postulates that

b k 4 = 1 n(n+1)(2n+1)(3n2 +3n − 1)
Σi=1 30

Write a program to test this hypothesis. Do you suspect the Professor is right or
wrong?

*** Ex. 296 — The Mathematics Professor claims that the constant e, an irra-
tional number, may be approximated using the first n values of the following
infinite series:
e = Σi=0∞ 1
i!

Write a program to illustrate that the series converges as n grows.

516 22 Lambda Expressions

127 What Have We Learned in This Chapter?

The important lessons of Chap. 22 are summarized as follows:

• Functions are a type of value.
• Computing functions are common in human life.
• A function name is simply a variable whose value is a function.
• A λ-expression evaluates to an anonymous function.
• The syntax for a λ-expression in ISL+ is
expr ::= (lambda (symbol+ ) expr)
::= (λ (symbol+ ) expr)
• Use a λ-expression on its own when a nonrecursive function that is only used
once is needed.
• Functions can compute functions.
• A curried function is one that consumes part of its input and returns a function
that consumes the rest of the input.
• Curried functions allow programmers to partially evaluate a function to produce
a specialized function.
• Abstractions offer you the opportunity to easily experiment and test a hypothesis.
• Tests for functions that approximate a value must illustrate that the approximation
error gets smaller as the number of terms grows.
Chapter 23
Aliens Attack Version 5

This chapter illustrates the use of encapsulation, abstract functions, and λ-expressions
in the context of a larger piece of software. In essence, this chapter refactors Aliens
Attack from Chap. 16. The goal is to group together related functions and to stream-
line the implementation of functions. The refactored code for the video game will
have a set of constants needed by multiple functions or for testing, 6 global func-
tions, and tests for the global functions. The global functions needed are run and the
functions used in the big-bang expression in run’s body.
Technically, the functions used in the big-bang expression can be encapsulated
inside of run. This may be a perfectly reasonable choice for thoroughly tested
code. We abstain from doing so for two reasons. First, run becomes too large and
obfuscated. Second, as discussed in Part V, different components in a multiplayer
game may need a different subset of the global functions. Therefore, it is best not to
encapsulate them for now.
An outline of the desired refactored code for Aliens Attack is displayed in Fig. 105.
The code is organized in 9 sections: data definitions, constants that are needed by
more than a single function or test, structure definitions and sample instances, and
6 functions. No new data definitions nor changes to existing data definitions are
required for this new version of the video game. Therefore, these remain unchanged
and are not furthered discussed in this chapter. The constants that are only needed
by a single handler are encapsulated. The constants at the global level include those
that are needed by more than one handler or that are used in any handler’s test.
After the constants, the structure definitions and sample instances are written. In our
game we only have one structure definition: world. This is followed by the code
for each handler. The code for each handler includes the function definition, sample
expressions, and its tests. The results from the steps of the design recipe must still
be displayed for all global functions. The sample expressions and tests for auxiliary
functions are not included given that they are encapsulated. Finally, the program
ends with the refactored run function.

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 517
M. T. Morazán, Animated Problem Solving, Texts in Computer Science,
https://doi.org/10.1007/978-3-030-85091-3_23
518 23 Aliens Attack Version 5

Fig. 105 Structure of Aliens Attack version 5

(require 2htdp/image)
(require 2htdp/universe)
;; Data Definitions
..
.
;; Constants
..
.
(define-struct world (rocket aliens dir shots))
;; Sample world instances
..
.
<draw-world code>
..
.
<process-key code>
..
.
<process-tick code>
..
.
<game-over? code>
..
.
<draw-last-world code>
..
.
<run code>

Each code section is discussed in the remaining sections of this chapter. The
presented refactoring is based on the code presented in previous chapters. Quite a
bit of editing is required, but mostly the editing is straightforward: encapsulating
functions and simplifying definitions. There are a small number of testing updates
or removal (due to encapsulation). The test updates involve making sure that every
line of code is covered by the tests.

128 Constants

Two varieties of constants must be identified: those needed globally and those needed
locally. Global constants are defined to be those that are needed by more than one
global function or by the tests of a global function. All other constants ought to be
encapsulated. We organize the global constants in three broad categories: general
constants needed to define elements of the video game, constants to define images,
and constants to define an initial world.
128 Constants 519

The general constants needed to define elements of the video game include
those needed to define a scene, to define coordinates inside a scene, and NO-SHOT.
Examining the game’s code reveals the following general constants:
;; General Constants
(define MAX-CHARS-HORIZONTAL 20)
(define MAX-CHARS-VERTICAL 15)
(define IMAGE-WIDTH 30)
(define IMAGE-HEIGHT 30)
(define MIN-IMG-X 0)
(define MAX-IMG-X (sub1 MAX-CHARS-HORIZONTAL))
(define MIN-IMG-Y 0)
(define MAX-IMG-Y (sub1 MAX-CHARS-VERTICAL))
(define E-SCENE-W (* MAX-CHARS-HORIZONTAL IMAGE-WIDTH))
(define E-SCENE-H (* MAX-CHARS-VERTICAL IMAGE-HEIGHT))
(define NO-SHOT ’no-shot)
Observe that no changes are made to any of these constants. Each of these constants
is either needed by more than one function or test or is needed by another constant
definition.
The constants needed to define images that must remain global are those used only
to test the draw-world function. Image constants used while the game is running
are encapsulated in draw-world. The global image constants used in testing are
;; Drawing Testing Constants
(define SHOT-COLOR2 'skyblue)
(define ALIEN-COLOR2 'orange)
(define WINDOW2-COLOR 'white)
(define FUSELAGE2-COLOR 'orange)
(define NACELLE2-COLOR 'brown)
(define ALIEN-IMG2 . . .)
(define SHOT-IMG2 . . .)
(define FUSELAGE2 . . .)
(define WINDOW2 (ellipse 3 10 'solid WINDOW2-COLOR))
(define SINGLE-BOOSTER2 . . .)
(define BOOSTER2 (beside SINGLE-BOOSTER2 SINGLE-BOOSTER2))
(define ROCKET-MAIN2 . . .)
(define NACELLE2 . . .)
(define ROCKET-IMG2 . . .)
The values of all these constants remain unchanged from how they are defined in
Chap. 4. In the interest of clarity and brevity some of the code above is omitted as
indicated by the ellipsis.
520 23 Aliens Attack Version 5

The constants to define initial worlds are those needed to construct the initial
world used in the run function or in the testing of the handlers. These constants are
;; Constants for INIT-WORLD
(define AN-IMG-X (/ MAX-CHARS-HORIZONTAL 2))
(define INIT-ROCKET AN-IMG-X)
(define INIT-DIR 'right)
(define INIT-SHOT NO-SHOT)
(define INIT-LOA . . .)
(define INIT-LOS '())

;; Constants for INIT-WORLD2

(define INIT-ROCKET2 15)
(define INIT-ALIEN2 (make-posn 3 MAX-IMG-Y))
(define DIR2 'left)
(define SHOT2 (make-posn AN-IMG-X
(/ (sub1 MAX-CHARS-VERTICAL) 2)))
These constants remain mostly unchanged. To build SHOT2 AN-IMG-X is used
to eliminate having to evaluate (/ MAX-CHARS-HORIZONTAL 2) twice. We shall
now focus on the definition of INIT-LOA. Section 87 discusses the design of
create-alien-army using interval processing to define INIT-LOA. We take the
opportunity to discuss a new approach using ISL+’s build-list abstract function.
The signature for this function is
<X> natnum (natnum → X) → (listof X)
Let us call the given natural number k. This abstract function builds a list of size
k with the results of applying the given function to every integer, n, in the interval
[0..k-1] from low to high.
Four constants may be defined to build an initial army of aliens using build-list:
(define NUM-ALIENS 18)
(define NUM-ALIEN-LINES 3)
(define ALIENS-PER-LINE (/ NUM-ALIENS NUM-ALIEN-LINES))
(define STARTING-X-LINE (add1 (/ (- MAX-CHARS-HORIZONTAL
ALIENS-PER-LINE)
2)))
The first constant is the number of aliens in the initial army. The second is the
number of rows in the alien army. To make the design easy to understand we make
the first constant divisible by the second constant and their quotient is the number
of aliens per line (the third constant). Finally, the alien army ought to start in the
middle of the horizontal axis and at the top of the scene. This means that the number
of empty image-x coordinates before and after a row of aliens is given by dividing
by 2 the difference of the maximum number of characters on the horizontal axis and
the aliens per row. Adding 1 to this integer gives us the initial image-x coordinate
for each row (the fourth constant).
128 Constants 521

Modular arithmetic (first discussed in Sect. 30) may be used to compute the
image-x and image-y coordinates of each alien as follows:
x = (+ STARTING-X-LINE (remainder n ALIENS-PER-LINE))
y = (quotient n ALIENS-PER-LINE)
Here n is the number of the aliens to be created (e.g., the 12th or the 3rd ). Using
these formulas we define INIT-LOA as follows:
(define INIT-LOA
(build-list NUM-ALIENS
(λ (n)
(make-posn (+ STARTING-X-LINE
(remainder n
ALIENS-PER-LINE))
(quotient n
ALIENS-PER-LINE)))))
The function given to build-list creates a new alien instance using the alien
number it gets as input. Take time to compare the above approach to create an initial
army of aliens with the approach first presented in Sect. 87. You ought to be able to
appreciate how much more concise and elegant the code is now.

* Ex. 297 — Modify the constants to increase the number of aliens and the
number of rows in the initial alien army. Make sure a row of aliens fits and able
to move within confines of the scene.
*** Ex. 298 — Modify the constants to have the alien army in the initial world
displayed in the following formation:
522 23 Aliens Attack Version 5

129 Structure Definitions

Given that new characteristics are not being added to Aliens Attack there are no
changes made to the world structure nor to the sample worlds used in testing. They
remain as
;; A world is a structure: (make-world rocket loa dir los)
(define-struct world (rocket aliens dir shots))

(define INIT-WORLD (make-world INIT-ROCKET INIT-LOA

INIT-DIR INIT-LOS))

(define INIT-WORLD2 (make-world INIT-ROCKET2 (list INIT-ALIEN2)

DIR2 (list SHOT2)))

(define WORLD3 (make-world 7 (list (make-posn 3 3))

'right (list (make-posn 3 3))))
If INIT-WORLD were not used in testing, it could be encapsulated inside of run.
Since it is used in testing, it is left at the global level for it to be in scope for all
the tests. In contrast, the structure definition cannot be encapsulated unless all the
handlers are encapsulated with it (given that the need the world selector functions).
In general, if a structure definition is to be encapsulated, then all functions that use
the structure’s selectors, constructor, or predicate must be encapsulated with it.

130 Encapsulating and Refactoring Handlers

Each handler is a function. Therefore, a handler may encapsulate any constants and
auxiliary functions only it uses. In addition, the handler and any of its auxiliary
functions may be refactored to use abstract functions. You will be well-served,
however, if you refactor and test before encapsulating. Always keep in mind that
the process of refactoring may (accidentally) introduce a bug. It is better to catch
bugs before a refactored function is encapsulated. Once encapsulated the refactored
function cannot be directly tested.

130.1 The draw-world Handler

One of the goals is to encapsulate within the drawing handler anything that only
it uses. Inspecting the video game code immediately reveals that color constants,
the empty scene, the image-y coordinate of the rocket, and all the image-creating
functions (designed in Chap. 4) along with constants defined using them may be
encapsulated. We note that the image-creating functions do not require refactoring
130 Encapsulating and Refactoring Handlers 523

because they are simple and do not exhibit repetition among them. The definitions
to encapsulate are
(define SHOT-COLOR 'orange)
(define ALIEN-COLOR 'black)
(define WINDOW-COLOR 'darkgray)
(define FUSELAGE-COLOR 'green)
(define NACELLE-COLOR 'red)
(define E-SCENE-COLOR 'pink)
(define E-SCENE
(empty-scene E-SCENE-W E-SCENE-H E-SCENE-COLOR))
(define ROCKET-Y (sub1 MAX-CHARS-VERTICAL))

;; color → image Purpose: Create an alien image of the

;; given color
(define (mk-alien-img a-color) . . .)
(define ALIEN-IMG (mk-alien-img ALIEN-COLOR))

;; color → image Purpose: Create shot image of the

;; given color
(define (mk-shot-img a-color) . . .)
(define SHOT-IMG (mk-shot-img SHOT-COLOR))

;; color → image Purpose: Create the fuselage image

;; of the given color
(define (mk-fuselage-img a-color) . . .)
(define FUSELAGE (mk-fuselage-img FUSELAGE-COLOR))
(define FUSELAGE-W (image-width FUSELAGE))
(define FUSELAGE-H (image-width FUSELAGE))

;; color → image Purpose: Create rocket window image

(define (mk-window-img a-color) . . .)
(define WINDOW (mk-window-img WINDOW-COLOR))

;; color → image Purpose: Create single booster image

(define (mk-single-booster-img a-color) . . .)
(define SINGLE-BOOSTER
(mk-single-booster-img NACELLE-COLOR))

;; image → image Purpose: Create booster image

(define (mk-booster-img a-sb-img) . . .)
(define BOOSTER (mk-booster-img SINGLE-BOOSTER))

;; image image image → image Purpose: Create the main

;; rocket image
(define (mk-rocket-main-img a-window a-fuselage a-booster)
. . .)
524 23 Aliens Attack Version 5

(define ROCKET-MAIN
(mk-rocket-main-img WINDOW FUSELAGE BOOSTER))
;; image color → image Purpose: Create a rocket nacelle
;; image
(define (mk-nacelle-img a-rocket-main-img a-color) . . .)
(define NACELLE (mk-nacelle-img ROCKET-MAIN NACELLE-COLOR))

;; image image → ci Purpose: Create a rocket ci

(define (mk-rocket-ci a-rocket-main-img a-nacelle-img) . . .)
(define ROCKET-IMG (mk-rocket-ci ROCKET-MAIN NACELLE))
In the interest of brevity and clarity the bodies of the functions are not included
above.
In Sect. 125 the functions to draw a list of aliens and a list of shots were refactored
to
;; loa scene → scene
;; Purpose: To draw the given loa in the given scene
(define draw-loa (λ (a-lox scn)
(draw-lox draw-alien a-lox scn)))

;; los scene → scene

;; Purpose: To draw the given los in the given scene
(define draw-los (λ (a-lox scn)
(draw-lox draw-shot a-lox scn)))
Recall that the auxiliary function each uses (i.e., draw-alien and draw-shot) is
not recursive and is only referenced once (i.e., in the above functions). This suggests
using a λ-expression instead of defining the two auxiliary functions. The above
functions are refactored to
;; loa scene → scene
;; Purpose: To draw the given loa in the given scene
(define draw-loa (draw-lox-maker (λ (an-alien scn)
(draw-ci ALIEN-IMG
(posn-x an-alien)
(posn-y an-alien)
scn))))

;; los scene → scene

;; Purpose: To draw the given los in the given scene
(define draw-los (draw-lox-maker (λ (a-shot scn)
(if (eq? a-shot NO-SHOT)
scn
(draw-ci SHOT-IMG
(posn-x a-shot)
(posn-y a-shot)
scn)))))
130 Encapsulating and Refactoring Handlers 525

Implemented in this manner there is no longer a need for draw-alien and for
draw-shot to be defined. They may be deleted from the game’s program. En-
capsulating the above functions means that draw-ci (refactored using encapsula-
tion in Fig. 99), draw-lox-maker, and draw-lox may also be encapsulated inside
draw-world given that they are not referenced elsewhere.
To complete the refactoring of draw-world add draw-rocket and draw-world
(from Chap. 16) as local functions. The body of the local ought to call the locally
defined draw-world. The tests for (the new global) draw-world remain unchanged.

130.2 The process-key Handler

This handler is the only one that uses process-shooting, move-rckt-left, and
move-rckt-right as auxiliary functions. Therefore, these functions ought to be
encapsulated inside process-key. Observe, however, that move-rckt-left and
move-rckt-right are very similar:
;; rocket → rocket
;; Purpose: Move the given rocket right
(define (move-rckt-right a-rocket)
(if (< a-rocket (sub1 MAX-CHARS-HORIZONTAL))
(add1 a-rocket)
a-rocket))

;; rocket → rocket
;; Purpose: Move the given rocket left
(define (move-rckt-left a-rocket)
(if (> a-rocket 0)
(sub1 a-rocket)
a-rocket))
This suggests using a curried function to eliminate the repetitions and to create
the rocket-moving functions. There are three significant differences: the comparing
predicate, the second value given to this predicate, and the function used to create a
new rocket. Abstracting over the functions yields
(define (make-rocket-mover cmp val f)
(λ (rckt) (if (cmp rckt val) (f rckt) rckt)))
The rocket-moving functions are refactored to
;; rocket → rocket
;; Purpose: Move the given rocket left
(define move-rckt-left (make-rocket-mover > 0 sub1))

;; rocket → rocket
;; Purpose: Move the given rocket right
(define move-rckt-right
(mk-rckt-mvr < (sub1 MAX-CHARS-HORIZONTAL) add1))
526 23 Aliens Attack Version 5

The above three functions, process-shooting, and process-key from Aliens

Attack version 4 are encapsulated using a local-expression in the body of
process-key for Aliens Attack version 5. The body of the local-expression calls
the local process-key. The tests for this handler remain unchanged.

130.3 The process-tick Handler

This is the handler with the most auxiliary functions. We shall comment on all the
functions that ought to be encapsulated. We start with the refactored function to
move a list of aliens from Sect. 125:
;; dir → (alien → alien)
;; Purpose: Return a function to move an alien in the
;; given direction
(define (alien-mover-maker a-dir)
(λ (an-alien) (move-alien an-alien a-dir)))

;; (listof alien) dir → (listof alien)

;; Purpose: To move the given list of aliens in the
;; given direction
(define (move-loa a-loa dir)
(map (alien-mover-maker dir) a-loa))
It is unlikely that move-loa can be further simplified. The alien-mover-maker
function, on the other hand, refers to a function, move-alien, that is only ref-
erenced once. This suggests encapsulating it inside alien-mover-maker. Before
doing so, however, observe that move-alien is the only function that refers to
move-down-image-y, move-left-image-x, and move-right-image-x. There-
fore, these functions ought to be encapsulated inside move-alien. The resulting
function is
;; dir → (alien → alien)
;; Purpose: Return a function to move an alien in the
;; given direction
(define (move-alien-maker a-dir)
(local
[;; alien → alien
;; Purpose: Move the given alien in a-dir
(define (move-alien an-alien)
(local [;; image-y<max → image-y
;; Purpose: To move the given image-y<max down
(define (move-down-image-y an-img-y<max)
(add1 an-img-y<max))
130 Encapsulating and Refactoring Handlers 527

;; image-x>min → image-x
;; Purpose: Move the given image-x>min left
(define (move-left-image-x an-img-x>min)
(sub1 an-img-x>min))

;; image-x<max → image-x
;; Purpose: Move the given image-x<max right
(define (move-right-image-x an-img-x<max)
(add1 an-img-x<max))]
(cond [(eq? a-dir 'right)
(make-posn (move-right-image-x (posn-x an-alien))
(posn-y an-alien))]
[(eq? a-dir 'left)
(make-posn (move-left-image-x (posn-x an-alien))
(posn-y an-alien))]
[else (make-posn
(posn-x an-alien)
(move-down-image-y (posn-y an-alien)))])))]
move-alien))
Observe that move-alien is locally defined, but it is only referenced once and it is
not recursive. This suggests using a λ-expression to represent it instead of locally
defining it. The auxiliary functions to move image coordinates become local to the
new λ-expression as follows:
;; dir → (alien → alien)
;; Purpose: Return a function to move an alien in the
;; given direction
(define (move-alien-maker a-dir)
(lambda (an-alien)
(local [;; image-y<max → image-y
;; Purpose: To move the given image-y<max down
(define (move-down-image-y an-img-y<max)
(add1 an-img-y<max))

;; image-x>min → image-x
;; Purpose: Move the given image-x>min left
(define (move-left-image-x an-img-x>min)
(sub1 an-img-x>min))

;; image-x<max → image-x
;; Purpose: Move the given image-x<max right
(define (move-right-image-x an-img-x<max)
(add1 an-img-x<max))]
(cond [(eq? a-dir 'right)
(make-posn (move-right-image-x (posn-x an-alien))
(posn-y an-alien))]
528 23 Aliens Attack Version 5

[(eq? a-dir 'left)

(make-posn (move-left-image-x (posn-x an-alien))
(posn-y an-alien))]
[else (make-posn
(posn-x an-alien)
(move-down-image-y (posn-y an-alien)))]))))
In this form it is easier to see that alien-mover-maker is a curried function to
create specialized versions of an alien-moving function. This is the function to
encapsulate inside process-tick.

** Ex. 299 — The conditional in the above function contains some repetition.
Use abstraction to eliminate repetitions. Is the resulting function easier to un-
derstand?

In Sect. 119 the function to move a list of shots is refactored to

;; los → los
;; Purpose: To move the given list of shots
(define (move-los a-los)
(local [;; shot → shot
;; Purpose: To move the given shot
(define (local-move-shot a-shot)
(cond [(eq? a-shot NO-SHOT) a-shot]
[else
(cond [(= (posn-y a-shot) MIN-IMG-Y)
NO-SHOT]
[else
(make-posn
(posn-x a-shot)
(move-up-image-y (posn-y a-shot)))]))
(map local-move-shot a-los))
Once again, we see a local function defined that is not recursive and is only refer-
enced once. This suggests using a λ-expression. This yields the following refactored
move-los:
;; los → los
;; Purpose: To move the given list of shots
(define (move-los a-los)
(map (lambda (a-shot)
(cond [(eq? a-shot NO-SHOT) a-shot]
[(= (posn-y a-shot) MIN-IMG-Y) NO-SHOT]
[else (make-posn
(posn-x a-shot)
(sub1 (posn-y a-shot)))]))
a-los))
130 Encapsulating and Refactoring Handlers 529

In this example the nonrecursive function only referenced once, move-up-image,

is inlined instead of making it local to the λ-expression. In this case, inlining seems
not to affect the function’s readability. Be warned, however, that inlining may be
taken too far and render a function incomprehensible. In fact, the same is true for
abstraction. You have done too much abstraction when the code is too far removed
from its concrete goals. There must always be a balance between code readability
and refactoring techniques like abstraction and inlining.
Another auxiliary function used by process-tick is
;; loa dir → dir
;; Purpose: Return new aliens direction
(define (new-dir-after-tick a-loa old-dir)
(cond [(eq? old-dir 'down)
(new-dir-after-down a-loa)]
[(eq? old-dir 'left)
(new-dir-after-left a-loa)]
[else (new-dir-after-right a-loa)]))
This function is designed in Sect. 93.1. It references three auxiliary functions that
are simple and may be inlined without sacrificing readability as follows:
;; loa dir → dir
;; Purpose: Return new aliens direction
(define (new-dir-after-tick a-loa old-dir)
(cond [(eq? old-dir 'down)
(if (any-alien-at-left-edge? a-loa) 'right 'left)]
[(eq? old-dir 'left)
(if (any-alien-at-left-edge? a-loa) 'down 'left)]
[else (if (any-alien-at-right-edge? a-loa)
'down
'right)]))
Examining the code reveals that now this function is the only function that ref-
erences any-alien-at-left-edge? and any-alien-at-right-edge?. These
functions, therefore, ought to be encapsulated. Before doing so, however, we refactor
them to use ormap as suggested in Sect. 116.3. The new refactored function is
;; loa dir → dir
;; Purpose: Return new aliens direction
(define (new-dir-after-tick a-loa old-dir)
(local [;; loa → Boolean
;; Purpose: Determine if any alien is at scene’s
;; right edge
(define (any-alien-at-right-edge? a-loa)
(ormap (lambda (an-alien)
(= (posn-x an-alien) MAX-IMG-X))
a-loa))
530 23 Aliens Attack Version 5

;; loa → Boolean
;; Purpose: Determine if any alien is at scene’s
;; left edge
(define (any-alien-at-left-edge? a-loa)
(ormap (lambda (an-alien)
(= (posn-x an-alien) MIN-IMG-X))
a-loa))]
(cond [(eq? old-dir 'down)
(if (any-alien-at-left-edge? a-loa) 'right 'left)]
[(eq? old-dir 'left)
(if (any-alien-at-left-edge? a-loa) 'down 'left)]
[else (if (any-alien-at-right-edge? a-loa)
'down
'right)])))
You may be thinking that we can streamline the function by inlining the two local
functions in the body of the cond-expression. Although there is no technical difficulty
in doing so, this is likely to reduce the readability of the code. Therefore, in the interest
of making sure our code clearly communicates how the problem is solved we refrain
from further inlining. You may, of course, disagree and perform the inlining of the
local functions.
** Ex. 300 — Inline any-alien-at-left-edge? and
any-alien-at-right-edge? in the function above. Why is the result more
or less readable?

Section 77.2 discusses the design of remove-hit-aliens and Sect. 77.3 dis-
cusses the design of remove-shots. Both of these auxiliary functions are only used
by process-tick:
;; loa los → loa
;; Purpose: Remove aliens hit by any shot
(define (remove-hit-aliens a-loa a-los)
(cond [(empty? a-loa) '()]
[(hit-by-any-shot? (first a-loa) a-los)
(remove-hit-aliens (rest a-loa) a-los)]
[else (cons (first a-loa)
(remove-hit-aliens (rest a-loa) a-los))]))

;; los loa → los

;; Purpose: Remove NO-SHOTs and hit shots
(define (remove-shots a-los a-loa)
(cond [(empty? a-los) a-los]
[(or (eq? (first a-los) NO-SHOT)
(hit-any-alien? (first a-los) a-loa))
(remove-shots (rest a-los) a-loa)]
[else (cons (first a-los)
(remove-shots (rest a-los) a-loa))]))
130 Encapsulating and Refactoring Handlers 531

Immediately, encapsulation comes to mind. Take a minute, however, to think about

the functions themselves. Can they be simplified? Both of these functions are re-
moving elements from the list they get as input. In other words, these functions are
good candidates to be simplified using filter. For remove-hit-aliens filter
needs a predicate that negates the result of testing if an alien is hit by any shot. For
remove-shots filter needs a predicate that negates the result of testing if a shot
is NO-SHOT or has been hit by any alien. Both of these needed predicates may be
implemented using a λ-expression and ormap as follows:
;; loa los → loa
;; Purpose: Remove aliens hit by any shot
(define (remove-hit-aliens a-loa a-los)
(filter (lambda (an-alien)
(not (ormap (lambda (a-shot) (hit? a-shot an-alien))
a-los)))
a-loa))

;; los loa → los

;; Purpose: Remove NO-SHOTs and hit shots
(define (remove-shots a-los a-loa)
(filter (lambda (a-shot)
(not (or (eq? a-shot NO-SHOT)
(ormap (lambda (an-alien)
(hit? a-shot an-alien))
a-loa))))
a-los))
These functions may now be encapsulated in process-tick. Observe that hit? is
now referenced only by these two functions in the entire program. This means that
hit? may not be encapsulated in either. It also means that hit? may be encapsulated
in process-tick because it will be in scope for the functions that need them.
The last function that needs to be encapsulated is process-tick from Aliens
Attack version 4. The body of the local-expression in process-tick for Aliens
Attack version 5 calls the local process-tick.
Observe that running the existing tests after encapsulation fails to test all the code.
Specifically, not all the code in hit? and in move-los is covered by the existing tests.
None of the tests for process-tick has an alien hit nor test moving a NO-SHOT. To
remedy this we add the following test:
(check-expect (process-tick
(make-world INIT-ROCKET
(list (make-posn 2 5))
'left
(list (make-posn 1 6) NO-SHOT)))
(make-world INIT-ROCKET
'()
'left
'()))
532 23 Aliens Attack Version 5

Observe that in this test the alien is hit by the first shot and the second shot is
NO-SHOT. This means that all the code is now fully tested. The important lesson to
remember is that encapsulation may remove tests that cover code not covered by the
tests for the function encapsulated into and, thus, requires new tests to be developed.

130.4 The game-over? Handler

The only references to any-alien-reached-earth? and any-aliens-alive?

are in this handler and draw-last-world. This suggests that these two functions
should not be encapsulated. For the sake of argument, however, let us proceed as if
only game-over? references these two auxiliary functions. That is, let us not decide
a priori that these functions should not be encapsulated and engage in speculative
encapsulation. After encapsulation the refactored function is28
;; world → Boolean
;; Purpose: Detect if the game is over
(define (game-over? a-world)
(local [;; (listof alien) → Boolean
;; Purpose: Determine if any alien has reached earth
(define (any-alien-reached-earth? a-loa)
(local [(define (alien-reached-earth? an-alien)
(= (posn-y an-alien) MAX-IMG-Y))]
(ormap alien-reached-earth? a-loa)))
;; loa Boolean → Boolean
;; Purpose: Determine if there is an alien in the
;; given loa
(define (any-aliens-alive? a-loa) (cons? a-loa))]
(or (any-alien-reached-earth? (world-aliens a-world))
(not (any-aliens-alive? (world-aliens a-world))))))
Observe that negating the value returned by any-aliens-alive? is the same
as testing if the given list of aliens is empty. Further observe that the body of
any-alien-reached-earth? is easily inlined into the local-expression’s body.
Finally, observe that alien-reached-earth? may be represented using a λ-
expression given that it is not recursive and is only needed once. These observa-
tions mean that the local-expression may be removed resulting in the following
refactored handler:
;; world → Boolean
;; Purpose: Detect if the game is over
(define (game-over? a-world)
(or (ormap (λ (an-alien) (= (posn-y an-alien) MAX-IMG-Y))
(world-aliens a-world))
(empty? (world-aliens a-world))))
28
The use of ormap-pred is substituted with ISL+’s ormap.
130 Encapsulating and Refactoring Handlers 533

The tests for this handler are unchanged. This refactoring exercise illustrates two
important lessons. The first is that logic is a useful tool to simplify functions. In this
example, it led to the elimination of any-aliens-alive? and the use of not. The
second is that speculative encapsulation may make us aware of possible simplifica-
tions.

** Ex. 301 — A fellow student proposes that the following game-over? im-
plementation communicates better how the problem is solved:
;; world → Boolean
;; Purpose: Detect if the game is over
(define (game-over? a-world)
(local [(define ANY-REACHED-EARTH
(ormap (lambda (an-alien)
(= (posn-y an-alien) MAX-IMG-Y))
(world-aliens a-world)))
(define ALL-ALIENS-DEFEATED
(empty? (world-aliens a-world)))]
(or ANY-REACHED-EARTH ALL-ALIENS-DEFEATED)))

Is she right? What is better and worse about this implementation? Remember
to consider how an or-expression is evaluated (discussed in Sect. 13.1).

130.5 The draw-last-world Handler

Recall that this handler must determine why game-over? returns #true. Given that
in Aliens Attack 4 it references the same auxiliary functions as game-over? we
can employ the same observations that led to game-over?’s refactoring: logic and
inlining. The refactored function is
;; world → scene throws error
;; Purpose: To draw the game’s final scene
(define (draw-last-world a-world)
(cond [(ormap (lambda (an-alien)
(= (posn-y an-alien) MAX-IMG-Y))
(world-aliens a-world))
(place-image (text "EARTH WAS CONQUERED!" 36 'red)
(/ E-SCENE-W 2)
(/ E-SCENE-H 4)
(draw-world a-world))]
534 23 Aliens Attack Version 5

[(empty? (world-aliens a-world))

(place-image (text "EARTH WAS SAVED!" 36 'green)
(/ E-SCENE-W 2)
(/ E-SCENE-H 4)
(draw-world a-world))]
[else
(error (format "draw-last-world: Given world has ~s
aliens and none have reached earth."
(length (world-aliens a-world))))]))
Remember that the same refactoring strategies may work for different functions that
refer to the same auxiliary functions.

131 Refactoring run

The run function is the only function that references TICK-RATE. Therefore,
TICK-RATE’s definition ought to be encapsulated. This straightforward refactoring
yields
;; string → world
;; Purpose: To run the game
(define (run a-name)
(local [(define TICK-RATE 1/4)]
(big-bang INIT-WORLD
[on-draw draw-world]
[name a-name]
[on-key process-key]
[on-tick process-tick TICK-RATE]
[stop-when game-over? draw-last-world])))
Recall that INIT-WORLD cannot be encapsulated because it is used to test the
handlers.
The only task remaining is to decide where to place accum and id. These functions
are only referenced by draw-lox, which is encapsulated in draw-world. Therefore,
these functions may also be encapsulated inside draw-world.
This ends the development of Aliens Attack version 5. Take time to appreciate
how much better organized and clearer the game’s code is now. Do you feel that
the code for Aliens Attack version 5 better communicates how the game works than
the code for Aliens Attack version 4? Have you developed a bigger appreciation for
encapsulation, anonymous functions, and abstract functions?
132 What Have We Learned in This Chapter? 535

132 What Have We Learned in This Chapter?

The important lessons of Chap. 23 are summarized as follows:

• Use build-list to create a list of a given size k with the results obtained from
applying a given function to the integers in [0..k-1] from low to high.
• If a structure definition is to be encapsulated, then all functions that use the
structure’s selectors, constructor, or predicate must be encapsulated with it.
• Refactor and test before encapsulating.
• Constants and auxiliary functions that are only used by a function, f, ought to be
encapsulated in f.
• There must always be a balance between code readability and refactoring tech-
niques like abstraction and inlining.
• Encapsulation may remove tests that cover code not covered by the tests for the
function encapsulated into and, thus, requires new tests to be developed.
• Logic is a useful tool used to simplify functions.
• Speculative encapsulation may make us aware of possible simplifications.
• The same refactoring strategies may work for different functions that refer to the
same auxiliary functions.
Chapter 24
For-Loops and Pattern Matching

Abstract functions traverse data of arbitrary size to compute a value. For example,
fact (from Sect. 81) traverses the integers in [0..n] to compute n! and map
traverses a list to compute a value. In essence, they iterate through the elements
contained in an instance of data of arbitrary size. Every time a function that traverses
data of arbitrary size is written a conditional expression is needed. This repetition
suggests an abstraction is needed. In ISL+, this abstraction is provided by for-loops.
A for-loop generates the sequences of values to be iterated over and combines the
values obtained from evaluating an expression (known as its body) to return a value.
They allow programmers to dispense with the coding of conditional expressions. In
addition, they allow programmers to iterate over multiple pieces of data.
The conditionals written to process data of arbitrary size distinguish between
the subtypes of a type and use selector functions to access the components of any
compound subtype. For example, to process a list a conditional expression is used
to distinguish between an empty- and a cons-list. If the given list is a cons-list,
the selector functions first and rest are used to access its components. This is
done every time a list-processing function is written and an abstraction is needed to
avoid the repetition. A similar situation arises when several functions to process a
given defined structure are written. For example, a function to process a posn uses
the selectors posn-x and posn-y. In ISL+, a match-expression provides program-
mers with an abstraction to eliminate this repetitive practice. A match-expression
dispatches on the type of the data and may introduce local variables to capture the
values in a compound piece of data. The use of such expressions can also signif-
icantly improve code readability as done by local-expressions in Sect. 121.2. A
match-expression allows programmers to dispense with writing a local-expression
for this purpose. Determining a data type to introduce local variables or control
program evaluation is called pattern matching.
This chapter explores problem solving and program design using for-loops and
pattern matching. Neither of these increases our computational power. We can com-
pute anything that is possible without them. That is, these are abstractions to make
program development easier and increase readability. In order to use them you must

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 537
M. T. Morazán, Animated Problem Solving, Texts in Computer Science,
https://doi.org/10.1007/978-3-030-85091-3_24
538 24 For-Loops and Pattern Matching

require the 2htdp/abstraction teachpack. That is, the following must appear in
your program before any code:
(require 2htdp/abstraction)

133 For-Loops

For-loops come in two general varieties in ISL+: those that extend the variables
in scope for the body of the for-loop and those that extend the variables in scope
for the body and each subsequent variable introduced by the for-loop. The first we
shall call for-loops and the second we shall call for*-loops. Both varieties require
comprehension clauses that declare local variables and create the values iterated
over.

133.1 for-loops

ISL+ offers 6 different for-loops with the following syntax:

expr ::= (for/list (clause+) expr)
::= (for/and (clause+) expr)
::= (for/or (clause+) expr)
::= (for/sum (clause+) expr)
::= (for/product (clause+) expr)
::= (for/string (clause+) expr)
clause ::= [variable expr]
A for-loop has a keyword identifying the type of loop, one or more comprehension
clauses, and an expression for the loop’s body. A comprehension clause declares a
variable the sequence of values to iterate over. For each iteration step the variable
takes on the next value in the sequence. The scope of a comprehension variable is
the body of the for-loop. For each value a comprehension variable takes on the
body of the for-loop is evaluated. The type of for-loop used defines how the values
obtained from evaluating the loop’s body are combined:
for/list conses all the values of type X
for/and ands all the Booleans
for/or ors all the Booleans
for/sum adds all the numbers
for/product multiplies all the numbers
for/string appends all the strings
The body of the loop must always evaluate to a value of the expected type. Otherwise,
an error is thrown. For example, the body of a for/string must always evaluate to
a string and the body of a for/and must always evaluate to a Boolean.
133 For-Loops 539

A comprehension clause declares a variable that shadows previous declarations

of the same variable. In addition, it generates the values the variable iterates over.
The type and value of the embedded expression determine the values generated as
follows:
List The elements of the given list from the first to the last element
Natural number The integers in [0..n-1], where n is the given natural number
String The substrings of length 1 from left to right of the given string
For example, the comprehension [i (build-list 4 (λ (x) x))] means that i
is used to iterate over the list elements 0, 1, 2, and 3 and the comprehension [s
"dog"] means that s is used to iterate over the substrings of length one: “d,” “o,”
and “g.”
Consider the list mapping function from Sect. 116.5:
;; <X Y> (X → Y) (listof X) → (listof Y)
;; Purpose: Return the values from applying the given function
;; to the given list’s elements
(define (map-f f a-lox)
(if (empty? a-lox)
'()
(cons (f (first a-lox))
(map-f f (rest a-lox)))))
This function distinguishes between the subtypes of a list and uses list selector
functions to access the components of a non-empty list as it is traversed. Further
observe that it returns a list. It returns the empty list when the given list is empty.
Otherwise, it conses all the results obtained from applying f to the first list element.
This makes it an ideal candidate for refactoring using for/list. The body of the
for-loop must only specify what do with each of the values in the loop. In this case,
apply the given f to it. This results in the following refactored function:
;; <X Y> (X → Y) (listof X) → (listof Y)
;; Purpose: Return the values from applying the given function
;; to the given list’s elements
(define (map-f f a-lox)
(for/list ([an-x a-lox])
(f an-x)))
The scope of an-x is the body of the loop. The loop conses the results of applying f to
each element of a-lox. Take time to appreciate what has been achieved. The need to
type the if-expression, the returned empty list, the cons application expression, and
the recursive call is eliminated. These elements all exist but remain hidden inside the
implementation of for/list. More importantly, perhaps, the above function tells
540 24 For-Loops and Pattern Matching

us that anything implemented using map can also be implemented using a for/list
loop. Consider, for example, the function to move a list of shots from Sect. 130.3:
;; los → los
;; Purpose: To move the given list of shots
(define (move-los a-los)
(map (lambda (a-shot)
(cond [(eq? a-shot NO-SHOT) a-shot]
[(= (posn-y a-shot) MIN-IMG-Y) NO-SHOT]
[else (make-posn
(posn-x a-shot)
(sub1 (posn-y a-shot)))]))
a-los))
This function may be refactored to use a for-loop as follows:
(define (move-los a-los)
(for/list ([a-shot a-los])
(cond [(eq? a-shot NO-SHOT) a-shot]
[(= (posn-y a-shot) MIN-IMG-Y) NO-SHOT]
[else (make-posn
(posn-x a-shot)
(sub1 (posn-y a-shot)))])))
Observe that the body of the for-loop is the same as the body of the λ-expression
in the previous implementation.
Loops in ISL+ may traverse multiple values of arbitrary size at the same time.
At each iteration step the next value in each of the sequences created is used to
evaluate the body of the loop. The loop stops when the end of any sequence is
reached. Whatever remains of the other sequences is ignored. Consider, for example,
the following function from Sect. 111:
;; (listof num) (listof num) → (listof num)
;; Purpose: Return a list with the products of corresponding
;; elements in the given lists
;; Assumption: The given lists are of the same length
(define (mlist L1 L2)
(if (empty? L1)
'()
(cons (* (first L1) (first L2))
(mlist (rest L1) (rest L2)))))
Observe that both lists are simultaneously traversed. To refactor this function using
a for-loop two comprehensions are needed: one for each list. The body of the loop
simply has to multiply the corresponding elements as the values are iterated over.
The resulting function is
(define (mlist L1 L2)
(for/list ([v1 L1] [v2 L2])
(* v1 v2)))
133 For-Loops 541

How do you feel about the result? Do you find it more elegant? Feeling comfortable
with loops is a matter of practice, but you can see that they do offer the opportunity to
dispense with the need to use a conditional expression to distinguish among subtypes
and to use selector functions.
Consider writing a predicate to determine if all the elements of a given list of
numbers are less than a given threshold. There are several options you may choose
from to design this function: use structural recursion on a list of numbers, use
andmap, or use a for/and loop. Perhaps, the first thing that comes to mind is to
iterate through the values of the list. For each value in the list the result obtained
from testing if it is less than the given threshold is anded with the result obtained
from processing the rest of the list. This problem analysis suggests using a for/and
loop.
The following sample instances of a (listof number) are used to develop the
sample expressions using a for/and loop below:
;; Sample instances of lon
(define L0 '())
(define L1 '(89 33 77 56 12 8 7))
(define L2 '(8 31 37 44 12 2 4))

;; Sample expressions for all-lt

(define L0-VAL (for/and ([v L0]) (< v 887)))
(define L1-VAL (for/and ([v L1]) (< v 100)))
(define L2-VAL (for/and ([v L2]) (< v 20)))
Observe that all sample instances of a lon are processed the same way. The com-
prehension builds the sequence of values to be iterated over and, therefore, the loop
is indifferent to the subtype of list it is processing.
Abstracting over the sample expressions yields the following function:
;; all-lt: (listof number) number → boolean
;; Purpose: Determine if numbers given list are < given number
(define (all-lt L threshold)
(for/and ([v L])
(< v threshold)))
To complete the steps of the design recipe the following tests are used:
;; Tests using sample computations for all-lt
(check-expect (all-lt L0 887) L0-VAL)
(check-expect (all-lt L1 100) L1-VAL)
(check-expect (all-lt L2 20) L2-VAL)

;; Tests using sample values for all-lt

(check-expect (all-lt '() 5) #true)
(check-expect (all-lt (list -1 2 -3) 0) #false)
(check-expect (all-lt (list 3 -8 -4) 10) #true)
542 24 For-Loops and Pattern Matching

This completes the design of the function. You ought to run the tests and confirm
that they all pass.

** Ex. 302 — Using a for-loop, design and implement a function that takes
as input a list of strings and that returns a list containing their lengths.

*** Ex. 303 — Using a for-loop, design and implement a function to compute
n!.
*** Ex. 304 — Using a for-loop, design and implement a function that adds
the corresponding elements of 4 lists of numbers.

** Ex. 305 — Using a for-loop, design and implement a predicate to deter-

mine if all the numbers in a list of numbers are divisible by 3.

** Ex. 306 — Using a for-loop, design and implement a predicate to deter-

mine if any image in a list of images has more than 10,000 pixels.

** Ex. 307 — Using a for-loop, design and implement a function that takes
as input a string containing only letters and that returns the string in all lower
case.
*** Ex. 308 — Using a for-loop, design and implement a function that takes
a list of symbols as input and that returns a list of symbols in which every 'a in
the given list is replaced with a 'z.

133.2 for*-loops

The second kind of ISL+ loop, for*-loop, has similar syntax and the same 6 varieties
as for-loops. The syntax only varies by the use of a *:
expr ::= (for*/list (clause+) expr)
::= (for*/and (clause+) expr)
::= (for*/or (clause+) expr)
::= (for*/sum (clause+) expr)
::= (for*/product (clause+) expr)
::= (for*/string (clause+) expr)
The difference with for-loops is the scope of the comprehension variables. While in
a for-loop the scope of a comprehension variable is the loop’s body, in a for*-loop
the scope of a comprehension variable is the remaining comprehension clauses and
the loop’s body. In addition, the sequences of values are iterated over differently. For
example, for every value taken on by the first comprehension variable the second
sequence of values is iterated over and for every value taken on by the second
comprehension variable the third sequence of values is iterated. This pattern is
133 For-Loops 543

repeated for each subsequent comprehension. Observe that there is no difference

between a for-loop and a for*-loop if there is only one comprehension clause.
Consider, for example, the problem of computing the cross-product of two lists.
The cross-product of two lists, L1 and L2, is all pairs, (a b), where a is a member
of L1 and b is a member of L2. For instance, the cross-product of '(a b) and '(7 8 9)
is
'((a 7) (a 8) (a 9)
(b 7) (b 8) (b 9))
In this list every value in '(a b) is paired with every value in '(7 8 9). How can
such a list be created? It is straightforward to see that all the values in the first list
must be iterated over. What needs to be done for each of these values? Given that
each of these values must be matched with every value in the second list, we can see
that the values in the second list must be iterated over to form the needed pairs. All
of the pairs must be consed together to create the needed list. This suggests using a
for*/list. The body of the loop creates a pair by listing the current value of the
first given list with the current value of the second list.
To satisfy the next step of the design recipe we define the following sample lists:
;; Sample instances of (listof X)
(define L0 '())
(define L1 '(i j c d))
(define L2 '(7 6 3))
These sample lists are used to write sample expressions for the function that we
shall call cross-product. The sample expressions ought to cover all possible
combinations of the inputs. Given that there are 2 varieties of the first input and two
varieties of the second input we need 4 sample expressions (i.e., 2 × 2 = 4):
;; Sample expressions for cross-product
(define L0-L0-VAL (for*/list ([v1 L0] [v2 L0])
(list v1 v2)))
(define L0-L1-VAL (for*/list ([v1 L0] [v2 L1])
(list v1 v2)))
(define L2-L0-VAL (for*/list ([v1 L2] [v2 L0])
(list v1 v2)))
(define L1-L2-VAL (for*/list ([v1 L1] [v2 L2])
(list v1 v2)))
Observe that the inputs are processed using a for*/list loop and that the body of
the loop creates a pair using the current values of each list being iterated over. For
every value v1 in the first list the values of the second list are iterated over using v2.
Abstracting over the sample expressions yields the following function:
;; <X Y> (listof X) (listof Y) → (listof (list X Y))
;; Purpose: To compute the cross product of the two lists
(define (cross-product a-lox a-loy)
(for*/list ([v1 a-lox] [v2 a-loy])
(list v1 v2)))
544 24 For-Loops and Pattern Matching

The tests using sample computations are straightforward to write:

;; Tests using sample computations for cross-product
(check-expect (cross-product L0 L0) L0-L0-VAL)
(check-expect (cross-product L0 L1) L0-L1-VAL)
(check-expect (cross-product L2 L0) L2-L0-VAL)
(check-expect (cross-product L1 L2) L1-L2-VAL)
Writing the tests using sample values requires more care. We must have a clear
understanding of the ordering of the pairs in the result. Observe that for each value
in a-lox the values of a-loy are traversed. This means in the resulting list the
first set of pairs corresponds to the first element of a-lox, the second set of pairs
corresponds to the second element of a-lox, and so on. In the interest of readability
to clearly communicate how the function works we also have these tests cover all
possible combinations of the inputs as follows:
;; Tests using sample values for cross-product
(check-expect (cross-product '() '()) '())
(check-expect (cross-product '() '(x y z)) '())
(check-expect (cross-product '(x y z) '()) '())
(check-expect (cross-product '(a b) '(7 8 9))
(list (list 'a 7) (list 'a 8) (list 'a 9)
(list 'b 7) (list 'b 8) (list 'b 9)))
An important realization is that the values of a comprehension may be themselves
be iterated over by using subsequent comprehensions. For example, consider flat-
tening a (listof (listof symbol)). Every element of the given list is a list of
symbols and a list containing all the elements of all the sublists needs to be computed.
For instance, given '((o p) (q r s t) (u)) the result is '(o p q r s t u).
How can this be accomplished? Each sublist must be processed. This may be
done by iterating over the elements of the given list. What needs to be done with
each sublist? Each element of a sublist must be consed into the result. This may
be accomplished by iterating through the sublist and returning each element. This
analysis suggests using a for*/list loop. We can now develop sample instances
of the data to process and sample expressions illustrating how to process the sample
instances as follows:
;; Sample instances of (listof (listof symbol))
(define L0 '())
(define L1 ’((a b c) (d e f g h) (i j)))

;; Sample expressions for flatten

(define L0-VAL (for*/list ([a-los L0] (a-symbol a-los))
a-symbol))
(define L1-VAL (for*/list ([a-los L1] (a-symbol a-los))
a-symbol))
Observe that values taken on by a-los in both sample expressions are of type
(listof symbol). Each of the values in an instance of a-los is taken on by
133 For-Loops 545

a-symbol. The body of the loop simply returns a-symbol to have it consed into
the result.
Abstracting over the sample expressions yields the following function:
;; (listof (listof symbol)) → (listof symbol)
;; Purpose: Flatten the given (listof (listof symbol))
(define (flatten-lolos a-lolos)
(for*/list ([a-los a-lolos] [a-symbol a-los])
a-symbol))
This function is tested as follows:
;; Tests using sample computations for flatten-lolos
(check-expect (flatten-lolos L0) L0-VAL)
(check-expect (flatten-lolos L1) L1-VAL)

;; Tests using sample values for flatten-lolos

(check-expect (flatten-lolos '((x) (y) (z))) '(x y z))
(check-expect (flatten-lolos '((k l m n))) '(k l m n))
(check-expect (flatten-lolos '((o p) () (q r s t) (u)))
'(o p q r s t u))
Observe that the tests using sample values test non-empty lists with varying lengths.
Each of these lists contains lists of symbols of varying length including the empty
list in the third test.
In closing the presentation on for-loops it is opportune to highlight that virtually
all modern programming provide them. In some other programming languages, how-
ever, for-loops do not compute a value. Instead, for-loops are strictly a mechanism
for controlling the flow of the program. Using such looping constructs may some-
times result in faster executing code. This, however, is not always true. Performance
greatly depends on the optimizations done by the compiler (i.e., the program that
translates a program into machine code). Experimentation and empirical data are
required to determine if a speed-up is achieved.

*** Ex. 309 — For Aliens Attack a (listof alien) and a (listof shot)
are defined. Using a for*-loop, design and implement a predicate to determine
if any alien has been hit by any shot.

** Ex. 310 — Using a for*-loop, design and implement a function that given
a posn in the first quadrant of the Cartesian plane generates all the posns with
integer coordinates in the sub-plane defined by the origin and the given posn.
546 24 For-Loops and Pattern Matching

134 Pattern Matching

Whenever there is a variety in the data we have used a conditional expression to

process said data. If the data is compound selector functions are used over and over
again in every function. This repetition calls for an abstraction. Pattern matching is
used to eliminate this type of repetition based on the type of data processed. In ISL+
pattern matching is done using a match-expression that has the following syntax:
expr ::= (match expr
[pattern expr]+ )
::= (match expr
[pattern expr]+ )
[else expr]+ )
pattern ::= literal
::= variable
::= (cons pattern pattern)
::= (structure-name pattern∗)
::= (? predicate-name)
In parenthesis there is the keyword match followed by an expression. This expression
is the matching expression. After the matching expression there are 1 or more
pattern stanzas. A pattern stanza has in square brackets a pattern to match and a
corresponding expression to evaluate if the pattern is matched. There is an optional
else stanza that contains the default expression used when none of the patterns are
matched. A match-expression looks like a cond-expression, but it checks the type
of the value of the matching expression. First the matching expression is evaluated
and then its value type is checked against each pattern from top to bottom. If a
pattern is matched, the corresponding expression is evaluated to obtain the value of
the match-expression. If none of the stanzas yield a match, then an error is thrown.
A pattern may be a constant (e.g., 3 or “hat”). There is a match if the value of the
expression equals the constant. A pattern may (always) match a variable. Matching a
variable means that a declaration that takes on the value of the matching expression
and that shadows any previous declarations of the variable is made. The scope of
the new variable is the corresponding expression in the match stanza. A pattern
may also match a list that matches a pattern for the first element and a pattern for
the second element. If either of these patterns is a variable, the variable takes on
the corresponding value in the list. The scope of this variable is the corresponding
expression and previous declarations are shadowed. A pattern may also match a
structure that matches a pattern for each of its components. Any of the components
may match a variable that, as before, shadows any previous declaration of the same
variable and whose scope is the corresponding expression in the match stanza.
Finally, a pattern may match by testing a predicate.
134 Pattern Matching 547

134.1 Illustrative Example

To make the semantics of match-expressions clear consider the program in Fig. 106.
The tests are omitted from the figure and are discussed in detail below. To start
consider the following tests:
(check-expect (f X) X-VAL)
(check-expect (f B) B-VAL)
(check-expect (f "Matthew Flatt")
"Matched a string: Matthew Flatt")
The first test passes because the value of X, 42, is given as an argument to f and
is matched in the first stanza of f’s match-expression. The second two tests pass
because they both provide f with a string. The given string does not match the
numeric constant, 42, in the first stanza and is matched in the second stanza because
when given to string? this predicate returns #true.
Now consider what happens for these tests:
(check-expect (f DIMITROVA) DIMITROVA-VAL)
(check-expect (f (make-student "Robby Findler" 'freshman 4))
"Matched a student whose gpa is: 4")
For the arguments provided to f the first two stanzas fail to match because a student
instance is neither the number value 42 nor a string. The third stanza signals a
match because the input is a student instance. This match instantiates three (local)
variables, n, y, and g, for, respectively, the given student’s name, year, and gpa. The
scope for these local variables is the expression in the third stanza. The student’s
gpa, g, is used to build the returned string.
The following tests give f a list that starts with a Boolean:
(check-expect (f BLST) BLST-VAL)
(check-expect
(f '(#false))
"Matched a list that starts with a Boolean: #false")
The first three stanzas in the match-expression do not detect a match. The fourth
stanza, on the other hand, does detect a match because the value of a-value is a
list and the first element of the list is a Boolean. A local variable is instantiated for
the rest of the list, but not for the first list element. The string returned by the fourth
stanza is constructed using the first list element and, therefore, the selector first
must be used to access the needed value.
The fifth stanza detects a match for an arbitrary list and instantiates, B and C, two
local variables, respectively, for the first list element and the rest of the list. The local
variable B shadows the global declaration of B. Consider the following tests:
(check-expect (f NLST) NLST-VAL)
(check-expect
(f '("P. Achten" "J. Hughes" "P. Koopman"))
"Matched a list whose first element is: \"P. Achten\"")
548 24 For-Loops and Pattern Matching

Fig. 106 A function to illustrate the semantics of match-expressions

(require 2htdp/abstraction)
;; A student is a structure: (make-student string symbol number>=0)
(define-struct student (name year gpa))
(define X 42)
(define B "Basia Mucha")
(define DIMITROVA (make-student "Rositsa Abrasheva" senior 3.94))
(define BLST (#true #true #true))
(define NLST (1 2 3))
(define ELST ())
;; Any string
;; Purpose: Illustrate the semantics of match-expressions
(define (f a-value)
(match a-value
[42 "Matched 42"]
[(? string?) (string-append "Matched a string: " a-value)]
[(student n y g)
(string-append "Matched a student whose gpa is: "
(number->string g))]
[(cons (? boolean?) r)
(string-append "Matched a list that starts with a Boolean: "
(boolean->string (first a-value)))]
[(cons B C)
(format "Matched a list whose first element is: ~s" B)]
[else "Nothing is matched."]))

;; Sample expressions for f

(define X-VAL "Matched 42")
(define B-VAL (string-append "Matched a string: " B))
(define DIMITROVA-VAL
(string-append "Matched a student whose gpa is: "
(number->string (student-gpa DIMITROVA))))
(define BLST-VAL
(string-append "Matched a list that starts with a Boolean: "
(boolean->string (first BLST))))
(define NLST-VAL (format "Matched a list whose first element is: ~s"
(first NLST)))
(define ELST-VAL "Nothing is matched.")

The first three stanzas do not detect a match because a-value is not 42, a string,
or a student in either test. The fourth stanza does not detect a match because the
given lists are of a different type: they do not have a Boolean as the first list element.
For the fifth stanza, the first test instantiates B to 1 (which is not a Boolean) and C to
'(2 3). The local B bound to 1 shadows the B bound to "Basia Mucha". The second
test instantiates B to "P. Achten" (which is not a Boolean) and C to '("J. Hughes"
"P. Koopman"). The local B, once again, shadows the B bound to "Basia Mucha".
The expression in the fifth stanza uses the value of (the local) B to build the returned
list.29

29
The backslashes in the expected value of the second test is DrRacket’s convention to indicate
that a double quote, ", is part of the string.
134 Pattern Matching 549

Finally consider these tests:

(check-expect (f '()) ELST-VAL)
(check-expect (f (square 10 "solid" "green"))
"Nothing is matched.")
The value provided to f in each test is not 42, a string, a student, or a cons-list.
Therefore, the first five stanzas do not detect a match and the returned string is given
by the default stanza.

134.2 Refactoring Using Pattern Matching

All the abstraction techniques studied can effectively make programs shorter and
easier to understand. We shall refactor the program developed in Sect. 106 to evaluate
arithmetic expressions. The complete set of functions needed is displayed in Fig. 107.
The first observation is that sum-lon, subt-lon, and mult-lon are all very similar
because they are list-summarizing operations. This suggests refactoring using accum.
The refactoring of sum-lon and mult-lon is
;; lon → number
;; Purpose: To sum the given lon
(define (sum-lon a-lon) (accum 0 id + a-lon))

;; lon → number Purpose: To multiply the given lon

(define (mult-lon a-lon) (accum 1 id * a-lon))
The refactoring of subt-lon requires a little algebraic manipulation. Observe that
(- a b c) is equivalent to (* -1 (+ (* -1 a) b c)). Furthermore, a match-
expression may be used to eliminate the use of first and rest. These observations
yield
;; lon → number throws error
;; Purpose: To subtract the given lon
(define (subt-lon a-lon)
(match a-lon
['() (error "No numbers provided to -.")]
[(cons fnum rnums)
(* -1 (accum 0 id + (cons (* -1 fnum) rnums)))]))
550 24 For-Loops and Pattern Matching

The bodies of sum-lon and subt-lon have an expression to sum a list of numbers.
To eliminate mostly repeated code we abstract to create a new sum-lon function and
obtain
;; X → X Purpose: Return the given input
(define (id x) x)

;; (listof number) → number Purpose: Sum given lon

(define (sum-lon a-lon) (accum 0 id + a-lon))

;; lon → number Purpose: To sum the given lon

(define (sum-lon a-lon) (sum-lon a-lon))

;; lon → number throws error

;; Purpose: To subtract the given lon
(define (subt-lon a-lon)
(match a-lon
['() (error "No numbers provided to -.")]
[(cons fnum rnums)
(* -1 (sum-lon (cons (* -1 fnum) rnums)))
The new sum-lon function only takes a list of numbers as input. Along with
mult-lon, these functions are only used in apply-f and may be encapsulated.
Observe, however, that the bodies of subt-lon and mult-lon are easily inlined
into the body of apply-f. The resulting function is displayed as part of the refactored
program in Fig. 108.
We observe that eval-args is also a list-summarizing function and may be
refactored to use accum:
;; (listof sexpr) → (listof number) throws error
;; Purpose: To evaluate the sexprs in the given list
(define (eval-args a-losexpr)
(accum '() (λ (a-sexpr) (eval-sexpr a-sexpr)) cons args))
In this form it is easily inlined into eval-slist. In turn, eval-slist is now easily
inlined into eval-sexpr. Inlining these functions eliminates the need to define
them. Finally, a match-expression may be used in eval-sexpr to eliminate the use
of first and rest. The refactored eval-sexpr is displayed in Fig. 108.
To complete the refactoring the refactored eval-sexpr, accum, and apply-f
are made local inside a global eval-sexpr. The body of the local-expression calls
the locally defined eval-sexpr. This completes the transformation of the program
in Fig. 107 to the one in Fig. 108. Take time to appreciate how much shorter and
simpler the resulting program is. Do you want to program with or without using
abstraction? It is perfectly fine not to feel comfortable with all the abstractions yet.
As your experience increases you will become more comfortable with abstraction.
134 Pattern Matching 551

Fig. 107 Functions for evaluating arithmetic expressions from Sect. 106
;; <X Y Z> Z (X Y) (Y Z Z) (listof X) Z
;; Purpose: Summarize given list
(define (accum base-val ffirst comb L)
(if (empty? L) base-val
(comb (ffirst (first L)) (accum base-val ffirst comb (rest L)))))
;; X X Purpose: Return the given X
(define (id an-x) an-x)
;; (listof sexpr) (listof number) throws error
;; Purpose: To evaluate the sexprs in the given list
(define (eval-args a-losexpr)
(if (empty? a-losexpr) ()
(cons (eval-sexpr (first a-losexpr)) (eval-args (rest a-losexpr)))))
;; slist number throws error Purpose: To evaluate the given slist
(define (eval-slist sl) (apply-f (first sl) (eval-args (rest sl))))
;; sexpr number throws error Purpose: To evaluate the given sexpr
(define (eval-sexpr a-sexpr)
(cond [(number? a-sexpr) a-sexpr]
[else (eval-slist a-sexpr)]))
;; function (listof number) number throws error
;; Purpose: To apply the given function to the given numbers
(define (apply-f a-function a-lon)
(cond [(eq? a-function +) (sum-lon a-lon)]
[(eq? a-function -) (subt-lon a-lon)]
[else (mult-lon a-lon)]))
;; lon number Purpose: To sum the given lon
(define (sum-lon a-lon)
(if (empty? a-lon) 0 (+ (first a-lon) (sum-lon (rest a-lon)))))
;; lon number throws error Purpose: To subtract the given lon
(define (subt-lon a-lon)
(if (empty? a-lon) (error "No numbers provided to -.")
(- (first a-lon) (sum-lon (rest a-lon)))))
;; lon number Purpose: To multiply the given lon
(define (mult-lon a-lon)
(if (empty? a-lon) 1 (* (first a-lon) (mult-lon (rest a-lon)))))

134.3 Designing Using Pattern Matching

Consider a function that consumes a list of numbers and then returns a list of the
even numbers in the given list doubled. How can this problem be solved? Reason
about the varieties of (listof number). If the given list is empty, then the result is
the empty list because there are no numbers to process. If the given list is not empty,
then we must distinguish between two types of (listof number): those that start
with an even number and those that start with an odd number. In other words, we are
redefining a list of numbers to be:
;; Data Definition
;; A (listof number) (lon) is either:
;; 1. '()
;; 2. (cons even-number lon)
;; 3. (cons odd-number lon)
552 24 For-Loops and Pattern Matching

Fig. 108 The refactored program from Fig. 107

;; sexpr number throws error
;; Purpose: To evaluate the given sexpr
(define (eval-sexpr a-sexpr)
(local [;; <X Y Z> Z (X Y) (Y Z Z) (listof X) Z
;; Purpose: Summarize given list
(define (accum base-val ffirst comb a-lox)
(match a-lox
[ () base-val]
[(cons x xs)
(comb (ffirst x) (accum base-val ffirst comb xs))]))
;; function (listof number) number throws error
;; Purpose: To apply the given function to the given numbers
(define (apply-f a-function a-lon)
(local [;; X X Purpose: Return the given input
(define (id x) x)
;; (listof number) number Purpose: Sum given lon
(define (sum-lon a-lon) (accum 0 id + a-lon))]
(cond [(eq? a-function +) (sum-lon a-lon)]
[(eq? a-function -)
(match a-lon
[ () (error "No numbers provided to -.")]
[(cons fnum rnums)
(* -1 (sum-lon (cons (* -1 fnum) rnums)))])]
[else (accum 1 id * a-lon)])))
;; sexpr number throws error
;; Purpose: Evaluate the given sexpr
(define (eval-sexpr a-sexpr)
(match a-sexpr
[(cons op args)
(apply-f
op
(accum () (λ (a-sexpr) (eval-sexpr a-sexpr)) cons args))]
[else a-sexpr]))]
(eval-sexpr a-sexpr)))

If the given list starts with an even number, the double of the first number is consed
to the result of recursively processing the rest of the given list. If the given list starts
with an odd number, the answer is obtained by recursively processing the rest of the
given list.
This problem analysis allows us to develop the following sample lists of numbers
and sample expressions:
;; Sample instances of lon
(define ELON '())
(define LON1 '(0 1 2 3 4 5))
(define LON2 '(7 8 9))
134 Pattern Matching 553

;; Sample expressions for extract-double-evens

(define ELON-VAL '())
(define LON1-VAL (cons (* 2 (first LON1))
(extract-double-evens (rest LON1))))
(define LON2-VAL (extract-double-evens (rest LON2)))
Observe there is one sample list for each subtype of (listof number). The sample
expressions are written in accordance to the problem analysis above.
The (listof number) varieties suggest that a match-expression may be used
to distinguish among them. The match-expression needs one stanza for each variety.
Each stanza may declare local variables instead of using the selectors first and
rest. The following is an implementation of these observations:
;; (listof number) → (listof number)
;; Purpose: Return a list of doubled evens in the given list
(define (extract-double-evens a-lon)
(match a-lon
['() '()]
[(cons (? even?) rlon)
(cons (* 2 (first a-lon)) (extract-double-evens rlon))]
[(cons flon rlon) (extract-double-evens rlon)]))
Each variety is processed as suggested by the sample expressions. The only difference
is that the local variables introduced by the match-expression are used to reference
the list components. Observe that in the second stanza a variable is not declared for
the first number in the list and, therefore, first is used to reference it.
Finally, the function is tested as follows:
;; Tests using sample computations for extract-double-evens
(check-expect (extract-double-evens ELON) ELON-VAL)
(check-expect (extract-double-evens LON1) LON1-VAL)
(check-expect (extract-double-evens LON2) LON2-VAL)

;; Tests using sample values for extract-double-evens

(check-expect (extract-double-evens '(-3 -7 -11)) '())
(check-expect (extract-double-evens '(88 10 120))
'(176 20 240))
(check-expect (extract-double-evens '(1 2 4 8))
'(4 8 16))
The tests using sample values clearly illustrate the function works when the given
list contains only odd, only even, or a combination of add and even numbers.

*** Ex. 311 — Using a match-expression refactor filter-pred from

Sect. 116.6.
* Ex. 312 — Using a match-expression write a function to compute the length
of a list.
554 24 For-Loops and Pattern Matching

**** Ex. 313 — Using a match-expression refactor move-los developed in

Sect. 130.3.
** Ex. 314 — Using a match-expression write a predicate to determine if a
given list is a (listof string).

** Ex. 315 — Using a match-expression refactor andmap-pred in Sect. 116.4.

Do you prefer the refactored function? Why or why not?

135 What Have We Learned in This Chapter?

The important lessons of Chap. 24 are summarized as follows:

• A for-loop generates the sequences of values to be iterated over and combines
the values obtained from evaluating its body to return a value.
• A match-expression dispatches on the type of the data and may introduce local
variables to capture the values in a compound piece of data.
• Code readability can be greatly increased and code repetition greatly reduced
using for-loops and match-expressions.
• In order to use for-loops and match-expressions an ISL+ program must require
the 2htdp/abstraction teachpack.
• A comprehension clause declares a (local) variable to iterate over the values
generated from a list, a natural number, or a string.
• ISL+ offers 6 different for-loops with the following syntax:
expr ::= (for/list (clause+) expr)
::= (for/and (clause+) expr)
::= (for/or (clause+) expr)
::= (for/sum (clause+) expr)
::= (for/product (clause+) expr)
::= (for/string (clause+) expr)
clause ::= [variable <expr>]
• In a for-loop the scope of a comprehension variable is the loop’s body.
• ISL+ offers 6 different for*-loops with the following syntax:
expr ::= (for*/list (clause+) expr)
::= (for*/and (clause+) expr)
::= (for*/or (clause+) expr)
::= (for*/sum (clause+) expr)
::= (for*/product (clause+) expr)
::= (for*/string (clause+) expr)
• In a for*-loop the scope of a comprehension variable is the remaining compre-
hension clauses and the loop’s body.
• Pattern matching is used to process data based on its type.
135 What Have We Learned in This Chapter? 555

• Pattern matching eliminates the repetitive use of selector functions by declaring

(local) variables to capture the values contained in compound data.
• A match-expression that has the following syntax:
expr ::= (match expr
[pattern expr]+ )
::= (match expr
[pattern expr]+ )
[else expr]+ )
pattern ::= literal
::= variable
::= (cons pattern pattern)
::= (structure-name pattern∗)
::= (? predicate-name)
• Functions may be designed or refactored using for-loops or pattern matching.
Chapter 25
Interfaces and Objects

Previous chapters have explored the idea that functions are data. In Chap.20 functions
were passed as input to other functions. In Chap. 22 functions were returned as the
value of function calls. This chapter explores the other side of the coin. Is data
a function? Intuitively, perhaps, the immediate answer that comes to mind is an
unequivocal no. A posn, for example, is not a function given that it does not compute
anything. It stores two values called x and y. You cannot apply a posn to arguments.
The same feels true about lists and trees. If L is a (listof X), what meaning can (L
'first) possibly have? Indeed, it may not be intuitive to think of data as a function.
This lack of intuition, however, is more than anything else a product of training.
After all, has it ever been suggested in a high school Mathematics textbook that a
number is a function? It turns out, however, that data can be a function. That is, data
may be represented using a function. Traditionally, data is thought of separately from
functions. This occurs despite the fact that whenever we create a data definition we
have in mind functions that are valid on the defined data. In other words, we expect
certain behavior from the data. For example, the program for Aliens Attack defines
an alien. It is expected that the alien can be moved right, left, or down, that the
alien may be determined to be at the right or left edge of the scene, and that the alien
may be determined to be hit by a shot. It is not expected that an alien can provide
its name or its distance to the origin on a Cartesian plane. It becomes clear that an
instance of a type is information (i.e., zero or more values) and the functions that are
valid on it.
Indeed, much of this textbook is about defining types: valid values and valid
functions. A posn, for instance, is not simply x and y values. It is also the ability
to extract the x value (i.e., posn-x), to extract the y value (i.e., posn-y), and to
determine if something is a posn (i.e., posn?). Given a posn, the signature and
purpose statement of these selectors, and this predicate, we understand the expected
behavior. Furthermore, we can add new operations and their expected behavior. For
example, we may add a function to compute the distance to the origin. Why, therefore,
should anyone strictly think of a data definition and the operations on the defined type
as two separate entities? Chap. 21 discussed encapsulating related functions. These

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 557
M. T. Morazán, Animated Problem Solving, Texts in Computer Science,
https://doi.org/10.1007/978-3-030-85091-3_25
558 25 Interfaces and Objects

functions can be the operations valid on some data type. If valid type operations are
encapsulated, then there is no technical impediment to also encapsulating the values
that define an instance of a type. To do so we must first learn how to define the
operations valid on a type.

136 Interfaces

A data definition defines a type and in the case of compound data it also defines
the type of the components. It states nothing about the valid type operations. An
interface defines the behavior of a defined type. In other words, an interface specifies
the operations that are valid on a type. Consider the problem of computing the
distance to the origin for a given point on a three-dimensional plane. Following the
steps of the design recipe yields the program displayed in Fig. 109. There are a few
unstated assumptions in the program. First, it is assumed that there is a function
to construct a 3Dposn. Second, it is assumed that there are selector functions for
the components of a 3Dposn. Third, the values of x, y, z, and the selectors are
stored separately from the function dist-origin. Would all this had been clear to
you before reading the preceding chapters? Would you have known that there is a
function 3Dposn-x to extract the x value of an instance of a 3Dposn?
In addition to defining a type, an interface is developed to define the expected
behavior of a type. An interface outlines the valid operations and the returned type.
For a 3Dposn the interface is:
Request x: number
Request y: number
Request z: number
Request distance: number
The interface makes it clear to any reader or user which are the valid operations on a
3Dposn. Observe that an interface says nothing about how a data type and its valid
operations are implemented.
Take a moment to ponder what has just been done. The data definition and the inter-
face explicitly relate x, y, z, 3Dposn-x, 3Dposn-y, 3Dposn-z, and dist-origin.
If they are related, then we ought to be able to encapsulate them into a single pack-
age. Whenever a 3Dposn is constructed, the package returned ought to be able to
perform all the operations in the interface. How can a 3Dposn instance perform
many operations? How does it know what operations to perform?
To achieve this a technique called message-passing is used. An interface is imple-
mented by a constructor function that returns a message-processing function. This is
a curried function that receives as input a message requesting a service. For example,
this function may get the message 'getx requesting the x value of the 3Dposn. An
136 Interfaces 559

Fig. 109 Distance to the origin program for a 3D point

;; A 3Dposn is a structure: (make-3Dposn number number number)
(define-struct 3Dposn (x y z))

;; Sample instances of 3Dposn

(define ORIGIN (make-3Dposn 0 0 0))
(define A3DPOSN (make-3Dposn 2 3 5))

;; 3Dposn number
;; Purpose: Return the distance to the origin of the given 3Dposn
(define (dist-origin a-3dposn)
(sqrt (+ (sqr (3Dposn-x a-3dposn))
(sqr (3Dposn-y a-3dposn))
(sqr (3Dposn-z a-3dposn)))))

;; Sample expressions for dist-origin

(define ORIGIND (sqrt (+ (sqr (3Dposn-x ORIGIN))
(sqr (3Dposn-y ORIGIN))
(sqr (3Dposn-z ORIGIN)))))
(define A3DPOSND (sqrt (+ (sqr (3Dposn-x A3DPOSN))
(sqr (3Dposn-y A3DPOSN))
(sqr (3Dposn-z A3DPOSN)))))

;; Tests using sample computations for dist-origin

(check-within (dist-origin ORIGIN) ORIGIND 0.01)
(check-within (dist-origin A3DPOSN) A3DPOSND 0.01)

;; Tests using sample values for dist-origin

(check-within (dist-origin (make-3Dposn 10 20 30)) 37.42 0.01)

interface, therefore, must specify the messages used to request a service. We can
now refine the 3Dposn interface to be:
'getx: number
'gety: number
'getz: number
'd2o: number
Observe that there is a unique message (in this case a symbol) associated with each
service. The idea is that the message-processing function determines what value to
compute by examining the message it gets as input. Observe that embedded in the
interface definition is a data definition for a message. A message is an enumeration
type: either 'getx, 'gety, 'getz, or 'd2o.
A constructor function that implements an interface is called a class. A class
encapsulates the values of and the operations on a type. It defines a constructor for
instances of a type. The value returned by a class is called an object. An object is an
instance of an interface and knows how to perform all the services in the interface
using message-passing. The 3Dposn class is displayed in Fig. 110. The class takes as
input 3 numbers and returns a 3Dposn. It is named make-3Dposn to easily identify
its role as a constructor for 3Dposns. Its body is a local-expression that defines an
560 25 Interfaces and Objects

Fig. 110 Interface implementation for 3Dposn

(require 2htdp/abstraction)

;; number number number 3Dposn Purpose: Return a 3Dposn object

(define (make-3Dposn x y z)
(local [;; 3Dposn number Purpose: Return distance to origin
(define (dist-origin x y z) (sqrt (+ (sqr x) (sqr y) (sqr z))))
;; message 3Dposn service throws error
;; Purpose: To manage messages for a 3Dposn
(define (manager m)
(match m
[ getx x]
[ gety y]
[ getz z]
[ d2o (dist-origin x y z)]
[else (error (string-append "Unknown message to 3Dposn: "
(symbol->string m)))]))]
manager))
;; Sample 3Dposn objects
(define ORIGIN (make-3Dposn 0 0 0))
(define A3DPOSN (make-3Dposn 2 3 5))

;; Tests using sample computations for 3Dposn

(check-within (ORIGIN getx) 0 0.01)
(check-within (A3DPOSN gety) 3 0.01)
(check-within (ORIGIN getz) 0 0.01)
(check-within (A3DPOSN d2o) 6.16 0.01)

;; Tests using sample values for 3Dposn

(check-within ((make-3Dposn 10 20 30) d2o)
37.42 0.01)
(check-error (A3DPOSN move-r)
"Unknown message to 3Dposn: move-r")

auxiliary function for any value that needs to be computed (in this case only distance
to origin) and the message-processing function called manager. The manager takes
as input a message and returns (the value of) a service defined in the interface.
The body of manager is a cond-expression to distinguish the message varieties. If
a service requires no computation, a value is directly returned. If computation is
required (like computing the distance to the origin), a local function is called. In
this case, manager is a guarded function that throws an error. It is also an object
given that it knows how to compute all the services in the interface and, therefore,
the local-expression returns it.
Testing interfaces requires defining one or more objects and writing tests to check
services are correctly provided. In Fig. 110 two 3Dposn objects are defined. The tests
check the result obtained from passing each message to an object. For example, the
x coordinate of ORIGIN is obtained using (ORIGIN 'getx)—passing the message
'getx to ORIGIN. The expected value is 0. Finally, the tests using sample values use
inlined uses of the constructor to test services and errors.
136 Interfaces 561

A 3Dposn, which is thought of as data, is a function. Specifically, it is an instance

of the curried function manager (i.e., an object) that is specialized for the values of
x, y, z given to make-3Dposn. For ORIGIN it is an 3Dposn object in which x = y =
z = 0. For A3DPOSN it is an 3Dposn object in which x = 2, y = 3, and z = 5. Just like
functions can be data, we see that data can be functions.

136.1 Improving the Human Interface

Message-passing may reduce the readability of the code. For example, does
(A3DPOSN 'd20) communicate to others that this expression represents A3DPOSN's
distance to the origin? Unless you are intimately familiar with the message-passing
protocol it is likely that this expression is meaningless. Furthermore, it is unlikely
that any programmer (including yourself) will permanently remember the message-
passing protocol in the near and far future. This will make it unnecessarily more
difficult to refine the program.
To mitigate this problem wrapper functions for the services provided by an
interface may be written. A wrapper function hides the details of the implementation.
In this case, it hides the details of message-passing. The idea is to allow programmers
to use 3Dposns without forcing them to know how they are implemented (much like
you do not know how posns are implemented in ISL+). A wrapper function is needed
for each service in the interface. It takes as input an object (and any additional inputs if
any) and its body applies the interface to the appropriate message. Wrapper functions
are designed following the steps of the design recipe.
Figure 111 displays the wrapper functions for 3Dposn. Adding this code to the
one displayed in Fig. 110 allows programmers to use a nicer version of the defined
interface. Instead of explicitly using message-passing, they can use the wrapper
functions. Observe that now programmers have the same interface as the one used
in Fig. 109. Writing wrapper functions does not provide a programmer with new
computational powers, but it is an abstraction that liberates a programmer from the
details of a message-passing protocol.

136.2 Services that Require More Input

After an interface is implemented, it may be necessary to add more services. This

means expanding the message-processing function and, if necessary, design (local)
auxiliary functions. If a value may be computed using only the information stored in
an object, then adding a service is no different than what was done for dist-origin
for 3Dposn. For example, adding a service to determine if a given 3Dposn object is
on the x-axis may be done by comparing the x value to 0.
562 25 Interfaces and Objects

Fig. 111 Wrapper functions for 3Dposn

;; 3Dposn number
;; Purpose: Return the x of the given 3Dposn
(define (3Dposn-x a-3dposn) (a-3dposn getx))

;; Sample expressions for 3Dposn-x

(define ORIGINX (ORIGIN getx)) (define A3DPOSNX (A3DPOSN getx))

;; Tests using sample computations and values for 3Dposn-x

(check-within (3Dposn-x ORIGIN) ORIGINX 0.01)
(check-within (3Dposn-x A3DPOSN) A3DPOSNX 0.01)
(check-within (3Dposn-x (make-3Dposn 10 20 30)) 10 0.01)

;; 3Dposn number
;; Purpose: Return the y of the given 3Dposn
(define (3Dposn-y a-3dposn) (a-3dposn gety))

;; Sample expressions for 3Dposn-y

(define ORIGINY (ORIGIN gety)) (define A3DPOSNY (A3DPOSN gety))

;; Tests using sample computations and values for 3Dposn-y

(check-within (3Dposn-y ORIGIN) ORIGINY 0.01)
(check-within (3Dposn-y A3DPOSN) A3DPOSNY 0.01)
(check-within (3Dposn-y (make-3Dposn 10 20 30)) 20 0.01)

;; 3Dposn number
;; Purpose: Return the z of the given 3Dposn
(define (3Dposn-z a-3dposn) (a-3dposn getz))

;; Sample expressions for 3Dposn-z

(define ORIGINZ (ORIGIN getz)) (define A3DPOSNZ (A3DPOSN getz))

;; Tests using sample computations and values for 3Dposn-z

(check-within (3Dposn-z ORIGIN) ORIGINZ 0.01)
(check-within (3Dposn-z A3DPOSN) A3DPOSNZ 0.01)
(check-within (3Dposn-z (make-3Dposn 10 20 30)) 30 0.01)

;; 3Dposn number
;; Purpose: Return distance to origin of given 3Dposn
(define (dist-origin a-3dposn) (a-3dposn d20))

;; Sample expressions for dist-origin

(define ORIGIND (ORIGIN d20)) (define A3DPOSND (A3DPOSN d20))

;; Tests using sample computations and values for dist-origin

(check-within (dist-origin ORIGIN) ORIGIND 0.01)
(check-within (dist-origin A3DPOSN) A3DPOSND 0.01)
(check-within (dist-origin (make-3Dposn 10 20 30)) 37.42 0.01)

If a service requires further input, the answer cannot be computed using only the
values stored in an object. That is, the object providing the service (usually referred
to as this) needs information beyond that which it stores. For instance, consider
136 Interfaces 563

adding a service that computes the distance to a given 3Dposn object. In addition to
this object another 3Dposn object is needed. This other object is unknown when
the interface is implemented and, therefore, cannot be provided as input. It is similar
to receiving a message. When the interface is written, there is no way to know which
messages will actually be received as input. The solution for messages is to make
a curried function that consumes the extra input (i.e., a message). The same design
tactic may be employed to add services that require extra input. The interface must
return a function that consumes the extra input.
To illustrate the technique let us add a service to compute the distance of this
3Dposn to a given 3Dposn. The first step is to update the interface as follows:
'getx number
'gety number
'getz number
'd2o number
'd2p 3Dposn → number
The data definition of a message is expanded to include 'd2p for the new distance-
computing service. Given that extra input is needed, the interface returns a function
that consumes the extra input, a 3Dposn, and that returns a number for the distance
between this and the given 3Dposn.
The next step is to refine the manager function to include the new service. This
means adding a stanza to the condition for the new service as follows:
;; message → 3Dposn service throws error
;; Purpose: To manage messages for a 3Dposn
(define (manager m)
(match m
['getx x]
['gety y]
['getz z]
['d2o (dist-origin x y z)]
['d2p distance]
[else (error (string-append "Unknown message to 3Dposn: "
(symbol->string m)))]))
The new stanza matches the new message and returns the distance function (yet
to be written). The distance function must satisfy the return type specified in the
interface definition.
Now, the distance function is designed and implemented. Keep in mind that this
is a local function inside the 3Dposn class. Therefore, this function has in scope all
the variables declared in the class. This is where the power of currying is exploited.
The previous inputs (x, y, and z) are used to compute the distance to the 3Dposn
564 25 Interfaces and Objects

received as input. After following the steps of the design recipe the following local
function is added to the 3Dposn class:
;; 3Dpson arrow number
;; Purpose: Compute the distance from this to the given 3Dposn
(define (distance a-3dposn)
(sqrt (+ (sqr (- x (3Dposn-x a-3dposn)))
(sqr (- y (3Dposn-y a-3dposn)))
(sqr (- z (3Dposn-z a-3dposn))))))
Observe that this function uses the coordinates of this and of the given 3Dposn to
compute the distance.
The final step is to develop a wrapper function for the new distance service. As
before, this is done following the steps of the design recipe. The resulting function is
(define B3DPOSN (make-3Dposn 1 1 1))

;; 3Dposn 3Dposn → number

;; Purpose: Return the distance between the given 3Dposns
(define (3Dposn-distance p1 p2) ((p1 'd2p) p2))

;; Sample expressions for 3Dposn-distance

(define ORIGINDP ((ORIGIN 'd2p) ORIGIN))
(define A3DPOSNDP ((A3DPOSN 'd2p) B3DPOSN))

;; Tests using sample computations 3Dposn-distance

(check-within (3Dposn-distance ORIGIN ORIGIN)
ORIGINDP
0.01)
(check-within (3Dposn-distance A3DPOSN B3DPOSN)
A3DPOSNDP
0.01)

;; Tests using sample values for 3Dposn-distance

(check-within (3Dposn-distance (make-3Dposn 10 20 30)
(make-3Dposn 2 3 4))
32.07
0.01)
A new 3Dposn instance is defined to facilitate writing tests. The sample expressions
apply a first 3Dposn object to the message 'd2p. This returns the object’s distance
function. This function is applied to a second 3Dposn to compute the distance.
The function is obtained by abstracting over the sample expressions. The tests are
developed as always using sample computations and sample values.
137 A Design Recipe for Interfaces 565

137 A Design Recipe for Interfaces

The systematic steps to design the implementation of an interface may now be

enumerated. The design recipe for an interface is:
1. Identify the values that must be stored and the services that must be provided.
2. Develop an interface data definition and a data definition for messages.
3. Develop a function template for the class that consumes the values that must be
stored and whose body is a local-expression returning the message-processing
function.
4. Specialize the signature, purpose, class header, and message-processing function.
5. Write and make local the auxiliary functions needed by the message-passing
function.
6. Write and test a wrapper function for each service.
The first step is problem analysis. It asks you to identify the information that spe-
cializes each object and the services that an object must provide. For every piece
of information identified there must be a parameter. The services must include a
selector for each piece of information that specializes an object. The second step
asks you to develop an interface data definition and a message data definition. There
must be a message for each service identified in Step 1.
The third step asks for the development of a function template for the class. Its
parameters correspond to the values identified in Step 1 to specialize an object.
The name of the class is also the name of the constructor. The function’s body is
a local-expression to encapsulate all needed functions that returns the message-
processing function. The fourth step has you develop the definition template for local
message-passing function. This function must contain a conditional that processes
a message as defined in Step 2. The fifth step asks you to develop all the auxiliary
functions needed by the message-processing function and make them local. Each
auxiliary function is developed using the design recipe.
The sixth step has you write a wrapper function for each service in the interface
developed in Step 2. These functions take as input the object that is providing the
service and the extra input, if any, required. The body of each wrapper function
applies the object providing the service to the appropriate message. If extra input is
needed, the function returned by the object is applied to the extra input.

* Ex. 316 — Following the steps of the design recipe for interfaces, design and
implement an interface for posn. The services offered by a posn are accessing
the x and y coordinates.

*** Ex. 317 — Following the steps of the design recipe for interfaces, design
and implement an interface for alien as defined for Aliens Attack 5.

**** Ex. 318 — Following the steps of the design recipe for interfaces, design
and implement an interface for car. A car has three characteristics: gas tank
size in liters, maximum speed in kilometers per hour, and kilometers per liter.
566 25 Interfaces and Objects

In addition to accessing its components, the services offered include maximum

distance that may be traveled in kilometers, maximum distance that may be
traveled in miles, and the number of liters used to travel a given distance in
kilometers.
*** Ex. 319 — The 3Dposn implementation has code repetition in
dist-origin and distance. Refactor the 3Dposn code to eliminate this rep-
etition.
** Ex. 320 — Add two 3Dposn services one to move the x, y, and z coordinates
by some given amounts and one to determine if this is close to a given 3Dposn.
A 3Dposn is close if the distance to it is less than or equal to 2.

138 Interfaces and Union Types

Designing interfaces for data with variety requires individually reasoning about each
variety. This should not come as a surprise given that individually reasoning about
each variety is how functions to process a union type are designed. Consider the
functions developed in Sect. 72 to compute the length and the sum of a list of quizzes:
;; A quiz grade (qg) is a number in [0..100]

;; A list of quiz grades (loq) is a (listof qg)

;; loq → number
;; Purpose: To compute the length of the given loq
(define (length-loq a-loq)
(if (empty? a-loq)
0
(+ 1 (length-loq (rest a-loq)))))

;; loq → number
;; Purpose: To compute the sum of the given loq
(define (sum-loq a-loq)
(if (empty? a-loq)
0
(+ (first a-loq) (sum-loq (rest a-loq)))))
Observe that a loq is a union type that has two varieties: empty and non-empty.
For this reason every function that processes a loq has (repeatedly) a conditional
expression just like length-loq and sum-loq. To write the conditional expression
each data variety is individually reasoned about. That is, an answer is formulated for
the empty list and an answer is formulated for the non-empty list.
139 An Abbreviated (listof X) Interface 567

Consider the implications of both varieties of loq offering services. The empty
loq knows how to compute its length and sum. The non-empty loq knows how to
compute its length and sum. This means that a conditional expression is not needed
because a given list, regardless of its variety, knows how to compute its length and
sum. This is called polymorphic dispatch. Polymorphic dispatch, in essence, is
the automatic process of selecting which implementation of an operation to use. It
means that programmers may dispense writing a conditional expression to process a
union type. To achieve this, the code that provides the answer for one variety must be
separated from the code that provides the answer for another variety. For instance,
the code to compute the length of an empty loq must be separated from the code to
compute the length of a non-empty loq. In fact, this must be done for every service
that a loq offers.
How is this separation of code and conditional expression elimination achieved
for a union type? For each subtype there must be a class that encapsulates the code
for that subtype. Observe that each subtype must offer the same services. That is, all
varieties have a common interface. Each class only implements the services for the
subtype it is written for. For instance, the empty loq needs to return 0 for its length
and 0 for its sum. The non-empty loq needs to return 1 plus the length of the rest of
the list for its length and the first quiz plus the sum of the rest of the quizzes for its
sum.

139 An Abbreviated (listof X) Interface

To illustrate interface design for data with variety, an abbreviated version of (listof
X) is implemented. Abbreviated means that this implementation shall not contain all
the functions associated with lists in ISL+. This implementation shall only include
the familiar empty?, first, rest, cons, and map. In addition, it includes the ability
to transform any list implemented using our interface into an ISL+ list.
The steps of the design recipe are outlined. The first three steps guide the rest of
the design. Carefully outlining these steps greatly facilitates the remaining steps.

139.1 Step 1: Values and Services

Two classes are required: one for the empty (listof X) and one for the non-empty
(listof X). The empty (listof X) needs to store no values. The non-empty
(listof X) needs to store two values: an X for the first element of the list and a
(listof X) for the rest of the elements of the list.
The following services are offered by a (listof X):
• Determine if the list is the empty list.
• Access the first element of the list.
• Access the rest of the list.
• Add a new element to the front of this list.
568 25 Interfaces and Objects

• Apply a function to every element of the list and return a list of the results.
• Transform this list into an ISL+ list.

139.2 Step 2: Interface and Message Definitions

Based on the services outlined in Step 1, 6 return types and 6 message varieties need
to be defined. These may be defined simultaneously in the definition of an interface.
Observe that empty? and the transformation into an ISL+ list return a value with no
need for further input. The services first or rest either return a value or throw an
error with no need for further input. An error is thrown if the first element or rest
of the elements are requested from the empty list. Finally, adding an element to the
front of the list or mapping a function requires further input. This means that these
services must return a function to consume this input.
In order not to confuse ISL+ lists with lists created using the interface devel-
oped here, they are denoted differently. ISL+ lists are denoted as (listof X) and
(listof Y). Lists created using the interface developed are denoted, respectively,
as listofx and listofy.
The interface for a listofx is
;; A listofx is an interface offering
;; 'empty: Boolean
;; 'first: X throws error
;; 'rest: listofx throws error
;; 'cons: X → listofx
;; 'map: (X → Y) → listofy
;; '2Rlst: (listof X)
A message is one of the 6 symbols listed in the interface. Determining if this list
is empty returns a Boolean. Accessing the first element of this list returns a value
of type X or throws an error. Accessing the rest of this list returns a listofx
or throws an error. Adding a value to the front of this list returns a function that
consumes a value of type X and that returns a listofx. Mapping a function onto
this list returns a function that consumes a function of type (X →Y) to map a
single list element and that returns a listofy. Finally, converting this returns an
ISL+ (listof X).

139.3 Step 3: Class Function Template

There are a few things known about what needs to be done to implement a class for
listofx. These include the need for a local message-processing function, a local
function to add an element to the front of this, and a function to map a given
function onto this list. In addition, it is known that wrapper functions and tests
139 An Abbreviated (listof X) Interface 569

covering both subtypes are needed for each service. The class template captures all
of these and as a result is quite long. Do not let its length intimidate you. Developing
such a detailed class template makes the next steps of the design recipe much easier.
It is a good time investment to develop a detailed class template. The class template
for listofx is
;; . . . → listofx
;; Purpose: Return a . . . listofx object
(define (class-for-listofx . . .)
(local
..
[ .
;; X → listofx
;; Purpose: Add given X to the front of this list
(define (add2front an-x) . . .)
;; (X → Y) → (listof Y)
;; Purpose: Map the given function to this list
(define (map f) . . .)
;; message → service throws error
;; Purpose: Provide service for the given message
(define (manager m)
(match m
['empty? . . .]
['first . . .]
['rest . . .]
['cons . . .]
['map . . .]
['2Rlst . . .]
[else
(error
(format "Unknown list service requested: ~s" m))]))]
manager))

;; Sample listofx objects

(define L0 . . .) (define L1 . . .). . .

;; WRAPPER FUNCTIONS
;; listofx → Boolean
;; Purpose: Determine if given listofx is empty
(define (listofx-empty? lox-o) . . .)
;; Sample expressions for listofx-empty?
(define L0E . . .) (define L1E . . .). . .
;; Tests using sample computations for listofx-empty?
(check-expect (listofx-empty? L0) L0E)
(check-expect (listofx-empty? L1) L1E). . .
;; Tests using sample values for listofx-empty?
(check-expect (listofx-empty? . . .) . . .). . .
570 25 Interfaces and Objects

;; listofx → X throws error

;; Purpose: Return first element of the given listofx
(define (listofx-first lox-o) . . .)
;; Sample expressions for listofx-first
(define L1F . . .). . .
;; Tests using sample computations for listofx-first
(check-expect (listofx-first L1) L1F). . .
;; Tests using sample values for listofx-first
(check-error (listofx-first . . .) . . .)
(check-expect (listofx-first . . .) . . .). . .

;; listofx → listofx throws error

;; Purpose: Return rest of the given listofx
(define (listofx-rest lox-o) . . .)
;; Sample expressions for listofx-rest
(define L1R . . .)
;; Tests using sample computations for listofx-rest
(check-expect (listofx-rest . . .)) . . .). . .
;; Tests using sample values for listofx-rest
(check-error (listofx-rest L0) . . .)
(check-expect (listofx-rest . . .) . . .). . .

;; <X Y> listofx (X → Y) → listofx

;; Purpose: Map given function onto the given listofx
(define (listofx-map lox-o f) . . .)
;; Sample expressions for listofx-map
(define L0M . . .). . .(define L1M . . .)
;; Tests using sample computations for listofx-map
(check-expect (listofx-map L0 . . .) . . .)
(check-expect (listofx-map L1 . . .) . . .). . .
;; Tests using sample values for listofx-map
(check-expect (listofx-map . . . . . .) . . .). . .

;; listofx → Boolean
;; Purpose: Convert the given listofx to a (listof X)
(define (listofx-2Rlst lox-o) . . .)
;; Sample expressions for listofx-2Rlst
(define L0RL . . .). . .(define L1RL . . .)
;; Tests using sample computations for listofx-2Rlst
(check-expect (listofx-2Rlst L0) L0RL)
(check-expect (listofx-2Rlst L1) L1RL). . .
;; Tests using sample values for listofx-2Rlst
(check-expect (listofx-2Rlst . . .) . . .). . .
Observe that the template definition for the message-processing function, manager,
is specialized for the message data definition developed as part of the interface in
140 The Empty (listof X) Class 571

Step 2. Furthermore, the template definition suggests making it a guarded function

to signal an error when the signature is violated by a programmer. Also observe
that the local-expression returns the message-processing function. The template
suggests defining at least two sample listofx objects: one for each variety. In the
above template it is assumed that L0 is an empty listofx object. Finally, for each
wrapper function the steps of the design recipe are followed. The sample expressions
and the tests using sample computations for listofx-first and listofx-rest
have at least one test (not two). This is because L0 throws an error when used with
these functions. The errors, as practiced in this textbook, are tested as part of the
tests using sample values.

140 The Empty (listof X) Class

This section designs the class that implements the interface for the empty listofx.
Steps 4–5 are outlined.

140.1 Step 4: Signature, Purpose, Class Header, and

Message-Passing Function

According to the problem analysis done in Step 1 the empty listofx has no char-
acteristics that must be stored. This means that the signature ought to be
;; → listofx
That is, a function with no input that returns a listofx interface. Unfortunately,
ISL+ does not allow functions with 0 parameters. To overcome this the class function
shall have a dummy parameter that is never referenced. Its sole purpose is to allow
for the development of the class function. This makes the signature, purpose, and
class header for the empty listofx:
;; Z → listofx
;; Purpose: Return an empty listofx object
(define (mtList dont-care-param)
Z is the type of the parameter. It may be any type given that it is never used. This
class returns an empty listofx object. The name of the class, mtList, is chosen to
provide any reader of the code an idea of what the class implements.
The message-processing function takes as input a message as defined in Step 2.
The conditional in the function definition template from Step 3 is specialized to
return a value if no further input is required and a function if more input is required
as outlined in the interface developed in Step 2. This function is designed to only
implement the services for the empty listofx. This means that for 'empty? it must
return #true, for 'first and 'rest it must throw an error, for 'cons and 'map it
572 25 Interfaces and Objects

must return a function, and for '2Rlist it must return '(). The function may be
implemented as follows:
;; message → service throws error
;; Purpose: Provide service for the given message
(define (manager m)
(match m
['empty? #true]
['first (error "first requested from the empty list")]
['rest (error "rest requested from the empty list")]
['cons add2front]
['map map]
['2Rlst '()]
[else
(error
(format "Unknown list service requested: ~s" m))]))
Here add2front and map are two auxiliary functions (to be written) that implement
the two services that require further input. This completes Step 4 of the design recipe.

140.2 Step 5: Auxiliary Functions

This class, as suggested by the class template, needs at least two auxiliary functions.
The first is the function to implement the cons operation. This function must consume
a value of type X and create a new non-empty listofx. To do so it must use the
constructor for the other variety of listofx. This function must provide as input
the given X value and this list. How do we reference this? Recall that this list
is an object represented by the message-processing function. Therefore, this is
the manager function. Assuming the other class is named consList, this auxiliary
function is
;; X → listofx
;; Purpose: Add given X to the front of this list
(define (add2front an-x) (consList an-x manager))
The mapping function must apply the given function to every element of this
list and return a listofy. There are no elements in this list. Therefore, this list
must only return an empty listofy. The map function for the empty listofx class
is
;; (X → Y) → (listof Y)
;; Purpose: Map the given function to this list
(define (map f) (mtList 'D))
Here 'D is a dummy value to construct an empty listofy. Given that neither of these
functions requires auxiliary functions, this step of the design recipe is completed.
141 The Non-Empty (listof X) Class 573

141 The Non-Empty (listof X) Class

We now focus on the design of the class for the non-empty listofx. As with the
design of the empty class, steps 4–5 are outlined.

141.1 Step 4: Signature, Purpose, Class Header, and

Message-Passing Function

According to the problem analysis done in Step 1 the non-empty listofx has two
characteristics that must be stored: the first element and the rest of the elements. This
means that the signature, purpose, and class header ought to be
;; X listofx → listofx
;; Purpose: Return a nonempty listofx object
(define (consList first rest)
This class returns a non-empty listofx object. The name of the class, consList,
and the name of the parameters, first and rest, are chosen to provide any reader
of the code an idea of what they represent.
The message-processing function takes the same input and serves the same pur-
pose as in the empty listofx class. That is, this function is designed to only
implement the services for the non-empty listofx. This means that for 'empty?
it must return #false, for 'first it must return first, for 'rest it must return
rest, for 'cons it must return add2front, for 'map it must return map, and for
'2Rlist it must return the consing of first and the conversion of rest. The
function may be implemented as follows:
;; message → service throws error
;; Purpose: Provide service for the given message
(define (manager m)
(match m
['empty? #false]
['first first]
['rest rest]
['cons add2front]
['map map]
['2Rlst (cons first (rest '2Rlst))]
[else
(error
(format "Unknown list service requested: ~s" m))]))
As in the empty listofx class, add2front and map are two auxiliary functions
(to be written) that implement the two services that require further input. Pay close
attention to the value returned for '2Rlst. Observe that this, call it L, is converted
with the expression (L '2Rlst). The rest of L is recursively processed using (rest
574 25 Interfaces and Objects

'2Rlst). The only difference is that there is no need to write a conditional expression
thanks to polymorphic dispatch. Remember that rest is a listofx object and it
knows how to convert its value regardless of the list subtype that it is.
This completes Step 4 of the design recipe. Take your time to understand and ap-
preciate that we have structural recursion without a conditional expression. Together
the stanzas for '2Rlst in the empty and non-empty classes are, in essence, the same
code you write as when writing a recursive function. It is simply split among the
classes for each subtype. The code is written differently using message-passing, but
the design is the same.

141.2 Step 5: Auxiliary Functions

As the empty listofx class, this class needs two auxiliary functions: add2front
and map. The add2front function must consume a value of type X and create a new
non-empty listofx. To do so it must use the constructor for a non-empty listofx
using the given value and this. Using the constructor for a non-empty listofx
sounds like a recursive call and it is, but remember that the constructor returns a
curried function. There is no recursive traversal here. The recursive call has a single
step that returns an object. Therefore, it is safe to call consList. The function is
implemented the same way as in the empty listofx class:
;; X → listofx
;; Purpose: Add given X to the front of this list
(define (add2front an-x) (consList an-x manager))
If you are thinking that such repetition calls for an abstraction, you are correct. To
abstract away from a class you need to implement an abstract class. An abstract
class contains all the common code among the classes for different subtypes. We
will not delve into abstract classes in this textbook, but you shall learn about them
in an object-oriented programming class.
The mapping function, as expected, must apply the given function to every element
of this list and return a listofy. To do so it must recursively process the rest of
this. The recursive call is implemented using polymorphic dispatch (just as the
converting service). The map function for the non-empty listofx class is
;; (X → Y) → (listof Y)
;; Purpose: Map the given function to this list
(define (map f) (consList (f first) ((rest 'map) f)))
Observe that ((rest 'map) f) is requesting a service from rest. Specifically, it
is requesting its mapping function to provide f to it. Once again, no conditional is
required for this recursive process because rest knows how to map a function to
its elements regardless of the listofx subtype that it is. Given that neither of these
functions requires auxiliary functions, this step of the design recipe is completed.
142 Step 6: Wrapper Functions and Tests 575

142 Step 6: Wrapper Functions and Tests

This class needs 6 wrapper functions as suggested by the class function template
from Step 3. Start by specializing the sample listofx objects. This may be done as
follows:
;; Sample listofx objects
(define L0 (mtList 'dummyval))
(define L1 (consList
1
(consList
2
(consList
3
(mtList 'dummyval)))))
These are empty and non-empty listofx. These constants are used to write tests
for the wrapper functions.
The wrapper functions are designed using the steps of the design recipe as sug-
gested in the class function template. The wrapper function to determine if a given
listofx is empty requires no further input. This part of the class function template
may be specialized as follows:
;; listofx → Boolean
;; Purpose: Determine if the given listofx is empty
(define (listofx-empty? lox-o) (lox-o 'empty?))

;; Sample expressions for listofx-empty?

(define L0E (L0 'empty?))
(define L1E (L1 'empty?))

;; Tests using sample computations for listofx-empty?

(check-expect (listofx-empty? L0) L0E)
(check-expect (listofx-empty? L1) L1E)

;; Tests using sample values for listofx-empty?

(check-expect (listofx-empty? (mtList 'dummyval)) #true)
(check-expect
(listofx-empty? (consList 'hi (mtList 'dummyval)))
#false)
576 25 Interfaces and Objects

The wrapper function to access the first element of a given listofx requires no
further input. This function, however, may throw an error given that the interface
for this service states that an error may be thrown. This part of the class function
template may be specialized as follows:
;; listofx → X throws error
;; Purpose: Return first element of given list
(define (listofx-first lox-o) (lox-o 'first))

;; Sample expressions for listofx-first

(define L1F (L1 'first))

;; Tests using sample computations for listofx-first

(check-expect (listofx-first L1) L1F)

;; Tests using sample values for listofx-first

(check-error (listofx-first L0)
"first requested from the empty list")
(check-error (listofx-first (mtList 'dummyval))
"first requested from the empty list")
(check-expect
(listofx-first (consList 'hi (mtList 'dummyval)))
'hi)
Observe that the signature specifies that an error may be thrown. In addition, there
is only one sample expression and it uses L1. There is no sample expression using
L0 because for this operation it throws an error. Instead, L0 is used to test the error
as part of the tests using sample values.
The wrapper function to access the rest of a given listofx requires no further
input. This function, however, may throw an error as specified in the interface. This
part of the class function template may be specialized as follows:
;; listofx → listofx throws error
;; Purpose: Return rest of given list
(define (listofx-rest lox-o) (lox-o 'rest))

;; Sample expressions for listofx-rest

(define L1R (L1 'rest))

;; Tests using sample computations for listofx-rest

(check-expect (listofx-2Rlst (listofx-rest L1)) '(2 3))

;; Tests using sample values for listofx-rest

(check-error (listofx-rest L0)
"rest requested from the empty list")
(check-error (listofx-rest (mtList 'dummyval))
"rest requested from the empty list")
142 Step 6: Wrapper Functions and Tests 577

(check-expect
(listofx-2Rlst
(listofx-rest
(consList 'hi
(mtList 'dummyval))))
'())
Once again, observe that the signature specifies that an error may be thrown and
there is only one sample expression. L0 is used to test the error as part of the tests
using sample values.
The wrapper function to add a given value to the front of a given listofx
requires further input. This means that this wrapper function needs to apply the
function returned by the list object to the given value. In addition, recall that function
equivalence is an unsolvable problem (discussed in Sect. 115). This means that the
value of a listofx object cannot be tested for equality with another listofx
object (they are both functions). To overcome this problem, a listofx may be
converted into an ISL+ list for testing. This part of the class function template may
be specialized as follows:
;; listofx X → listofx
;; Purpose: Add the given value to the front of the
;; given listofx
(define (listofx-cons lox-o an-x) ((lox-o 'cons) an-x))

;; Sample expressions for listofx-cons

(define L0C ((L0 'cons) 31))
(define L1C ((L1 'cons) 0))

;; Tests using sample computations for listofx-cons

(check-expect (listofx-2Rlst (listofx-cons L0 31))
(listofx-2Rlst L0C))
(check-expect (listofx-2Rlst (listofx-cons L1 0))
(listofx-2Rlst L1C))

;; Tests using sample values for listofx-cons

(check-expect
(listofx-2Rlst
(listofx-cons (consList "hi" (mtList 'dummyval))
"hey"))
'("hey" "hi"))
Observe that the result to any call to listofx-cons is converted into an ISL+ list
to test its value.
The wrapper function to map a function onto a given listofx requires further
input. This wrapper function needs to apply the function returned by the list object
to the given function. As with listofx-cons, a returned listofx is converted into
578 25 Interfaces and Objects

an ISL+ list for testing. This part of the class function template may be specialized
as follows:
;; <X Y> listofx (X → Y) → listofy
;; Purpose: Map given function onto given list
(define (listofx-map lox-o f) ((lox-o 'map) f))

;; Sample expressions for listofx-map

(define L0M ((L0 'map) sub1))
(define L1M ((L1 'map) (λ (n) (* 2 n))))

;; Tests using sample computations for listofx-map

(check-expect (listofx-2Rlst (listofx-map L0 sub1))
(listofx-2Rlst L0M))
(check-expect
(listofx-2Rlst (listofx-map L1 (λ (n) (* 2 n))))
(listofx-2Rlst L1M))

;; Tests using sample values for listofx-map

(check-expect
(listofx-2Rlst
(listofx-map
(consList "hi" (mtList 'dummyval))
string-length))
'(2))
Observe that the result to any call to listofx-map is converted into an ISL+ list to
test its value.
Finally, the wrapper function to convert a given listofx into an ISL+ list requires
no further input. It applies the given listofx object to '2Rlst. This part of the
class function template may be specialized as follows:
;; listofx → Boolean
;; Purpose: Convert given listofx to a ISL+ list
(define (listofx-2Rlst lox-o) (lox-o '2Rlst))

;; Sample expressions for listofx-2Rlst

(define L0RL (L0 '2Rlst))
(define L1RL (L1 '2Rlst))

;; Tests using sample computations for listofx-2Rlst

(check-expect (listofx-2Rlst L0) L0RL)
(check-expect (listofx-2Rlst L1) L1RL)
143 What Have We Learned in This Chapter? 579

;; Tests using sample values for listofx-2Rlst

(check-expect (listofx-2Rlst (mtList 'dummyval)) '())
(check-expect (listofx-2Rlst
(consList 'hi (mtList 'dummyval)))
'(hi))

*** Ex. 321 — Add the service to filter a list to listofx.

*** Ex. 322 — Add the services to or-map and and-map a list to listofx.

**** Ex. 323 — Add a list-summarizing service to listofx.

***** Ex. 324 — Design and implement an interface for (btof X) (intro-
duced in Chap. 17).

***** Ex. 325 — Design and implement an interface for natural numbers (in-
troduced in Chap. 14).

143 What Have We Learned in This Chapter?

The important lessons of Chap. 25 are summarized as follows:

• Data may be a function just like a function may be data.
• An instance of a type may contain values and the operations that are valid on it.
• An interface specifies the operations that are valid on a type.
• An interface is implemented by a constructor function called a class that returns
a message-processing function.
• A class encapsulates the values of and the operations on a type.
• A class defines a constructor for instances of a type.
• The value returned by a class is called an object which is an instance of an
interface.
• Testing interfaces requires defining one or more objects and writing tests to check
services are correctly provided.
• To improve the human interface wrapper functions hide the details of an interface
implementation.
• To implement an interface for a union type there must be a class that encapsulates
the code for each subtype.
• Polymorphic dispatch allows programmers to implement structural recursion
without conditional expressions.
• Code is written differently when programming with objects, but the design is the
same as programming without objects.
• Objects may be converted to write tests.
 Part V
Distributed Programming
Chapter 26
Introduction to Distributed Programming

You are probably very familiar with distributed programming as a user. In fact, you
probably use distributed programs every day. Such programs include text messaging
systems, multiplayer video games, and social media apps. As a user of a distributed
program you solicit services from another program (usually running on another
computer). For example, you may click on a link in your social media app. The act
of clicking on the link sends a message to a computer operated by the social media
company requesting the contents associated with the link. In turn the social media
company’s computer sends your app a message with the contents of the link. This is
an example of two programs cooperating to solve a problem. Program cooperation
is at the heart of many modern software systems today.
Dividing a problem into several tasks and writing a program for each task that
communicates with the programs for other tasks are called distributed programming.
The tasks cooperate to solve a problem. Each task defines a component that is a
program that solves the task. Each component itself may be divided into subtasks
and may be solved using one or more computers. The components cooperate by
communicating with each other using message-passing. Messages are exchanged
via a network (e.g., the internet). For messages to be exchanged a communication
protocol must be designed. A communication protocol defines the messages that
may be exchanged and when messages are exchanged. For example, a text message
is not sent to your friend until you hit the Enter key or the Send button. Sending a
text message in this manner means you are adhering to a communication protocol:
the message must be a string and the message is only sent after you hit the right
button or key. There may be, of course, a variety in the messages you are allowed to
send (e.g., strings, numbers, etc.). Breaking the communication protocol means that
a message is not sent or is not delivered and communication fails to take place. That
is, there is a breakdown in cooperation.
Messages cannot be arbitrary. That is, there are a finite number of data types that
are suitable for transmission. For example, a number is suitable for transmission
but a posn is unsuitable for transmission. If a component needs to send data that is
unsuitable for transmission, the data must be marshaled. Marshalling is the process
of transforming data that is unsuitable for transmission into data that is suitable
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 583
M. T. Morazán, Animated Problem Solving, Texts in Computer Science,
https://doi.org/10.1007/978-3-030-85091-3_26
584 26 Introduction to Distributed Programming

Fig. 112 Protocol diagram for sending m from clienti to clientj

m
Time

Client Client Server

for transmission. To transmit a posn, for example, it must be marshaled into a list
of two numbers for transmission. The component receiving marshaled data must
unmarshal it. Unmarshalling is the process of reconstructing the original data from
marshaled data. For example, a component receiving a marshaled posn as a list of
two numbers must reconstruct the original posn. Marshalling and unmarshalling
functions are inverses of each other. That is, (unmarshal (marshal x)) = x and
(marshal (unmarshal message)) = message. The use of inverse functions is
actually quite common in programming as mentioned in Chap. 2. You may frequently
use encrypt and decrypt to protect sensitive data or compress and uncompress to
reduce file size. Now, you also know about marshalling and unmarshalling data for
transmission.
A pervasively used distributed system architecture is the client–server architec-
ture. A server is a program that provides services or coordinates the cooperation
among clients. A client is a program that performs a task (usually) in cooperation
with other clients to solve a problem. A client on one computer requests services
from the server that typically runs on another computer. All communication between
clients occurs through the server. That is, if Clienti needs to send a message, m,
to Clientj, then Clienti sends m to the server and the server sends m to Clientj.
This communication chain is part of the communication protocol. In a commu-
nication protocol there can be many communication chains for different events. For
example, in a multiplayer Aliens Attack there may be a communication chain that
is started when a player shoots and another that starts when a player moves the
rocket. A communication protocol may be specified using protocol diagrams for
different event-based communication chains. In a protocol diagram the horizontal
axis represents the components and the vertical access represents time (which grows
from top to bottom). Messages are represented by solid arrows from source to des-
tination at a slight angle. The angle is used to emphasize that communication is not
instantaneous. That is, time elapses as messages travel from source to destination.
Dashed arrows are used to represent communication that is implemented by the API
(e.g., a client registering with a server). Figure 112 displays the protocol diagram
144 A Design Recipe for Distributed Programming 585

to send a message, m, from Clienti to Clientj. Time grows from top to bottom.
The first step in the communication chain is Clienti sending m to the server. This
is indicated in the diagram by a solid arrow from Clienti to the Server. Observe
that the arrow’s negative slope clearly indicates that time elapses as m travels from
Clienti to the Server. At the Server, time elapses as m is processed before it
is sent out to Clientj. This is indicated by the gap between the incoming arrow
and the outgoing arrow at the Server. Finally, the outgoing arrow’s positive slope
clearly indicates that time elapses as m travels from the Server to Clientj. When
m arrives at Clientj the communication chain successfully ends.
Servers are generally described by a spectrum that goes from thin to think. A thin
server is one that provides a minimal number of services like message broadcasting.
The server itself does no or very little actual computing. A thick server, on the other
hand, provides services that involve actual computing that directly contributes to
solving a problem. For example, a server may be responsible for computing a value.

144 A Design Recipe for Distributed Programming

As you may have already deduced, distributed programming entails many charac-
teristics that are not present when a program has a single task (e.g., like computing
n!). It is important to carefully design the different components (e.g., clients and
server) and the communication protocol. The magnitude of these tasks may seem
overwhelming at this moment, but practicing the design and implementation of dis-
tributed programs will make you feel more comfortable. As with other topics in
this textbook, there is a design recipe to help guide your development. This design
recipe, however, is less prescriptive than the previously discussed design recipes. It
does not tell you when to use a certain type of expression nor does it dictate what the
parameters to a function must be. Instead, it guides you through the development of
a distributed program assuming that you have mastered the design recipes previously
studied to write functions. Each step still has a specific outcome that gets you closer
to writing a program to solve a problem.
The design recipe for distributed programming is:
1. Divide the problem into components.
2. Draft data definitions for the different components.
3. Design a communication protocol.
4. Design marshalling and unmarshalling functions.
5. Design and implement the components.
6. Test your program.
Step 1 is problem analysis. It asks you to outline how the components cooperate to
solve a problem. This step clearly defines the task (or tasks) carried out by a compo-
nent. For example, the server may be responsible for relaying rocket movements by a
player to all other players in Aliens Attack, while the clients are responsible for draw-
586 26 Introduction to Distributed Programming

ing the world. There are at least two components for every distributed application: a
server and a client.
Step 2 asks you to define the types (or refine the types of an existing program)
required by each component. The types required for each component are not neces-
sarily the same. A server, for example, requires a type definition for messages that
may be sent to it. A client, on the other hand, requires a possibly different type for
the messages it may receive.
Step 3 asks you to develop a communication protocol. This protocol must capture
all the communication chains that may occur. A communication chain is sparked by
an event. For example, in multiplayer Aliens Attack a communication chain is started
when a player shoots. The communication chain may be that the shot created is sent
to the server and the server sends the new shot to all the other players. As part of this
step you must develop data definitions for to-server messages and to-client
messages. These data definitions are used to design the message-processing function
for the server and for the clients.
Step 4 asks you to design marshalling and unmarshalling functions. If the data
that needs to be transmitted in a message is suitable for transmission, there may be
no need to develop these functions for this type of data. If the data to be transmitted
is unsuitable for transmission, then a marshalling and unmarshalling function is
needed to, respectively, create a message suitable for transmission and reconstruct
the original data on the receiving component. Commonly, messages are tagged to
easily distinguish the different varieties. This means that the need for marshalling
and unmarshalling is expected for most applications.
Step 5 asks you to develop the programs for each component. This means that you
need to develop at least two programs: one for the server and one for each client. The
client program may or may not be the same for all clients. Nonetheless, there must
be a separate program (different or copy) for each client. Observe that this means
that a distributed program is written in at least two different files.
Step 6 asks you to run and test your program. As always, if any tests fail you must
redesign.

145 More on the Universe API

The development of distributed programs in this textbook uses the API offered by the
universe teachpack. The universe teachpack provides the functionality to develop
distributed multiplayer games. Each player and the server are components. As you
know, a player manages a world and is executed using a big-bang-expression. A
server manages a universe (e.g., a collection of players) and is executed using a
universe-expression. The players in a universe exchange messages with the server.
All communication occurs through the server.
The universe teachpack provides two functions to create messages:
make-package and make-bundle. The first is used by a client to create a structure
that contains a (possibly new) world and a to-server message. The second is used by
145 More on the Universe API 587

the universe server to create a structure that contains a (possibly new) universe, a
list of mails to any of the players, and a list of worlds to be disconnected from the
universe. Observe that a bundle contains an arbitrary number of mails and not an
arbitrary number of to-world messages. A mail is a structure, built using make-mail,
that contains the recipient player and a to-client message.
A message must be an S-expression as defined by the universe API. A
universe S-expression is defined as follows:
A universe S-expression (sexpr) is either a:
1. string
2. symbol
3. number
4. Boolean
5. character
6 (listof sexpr)
Nothing else is suitable for transmission in a universe program. In particular observe
that structures may not be transmitted. If a structure must be transmitted, you must
implement marshalling and unmarshalling functions to do so.
The big-bang syntax required to run a player, as you know, specifies the handlers
that update the game or render the game to the screen. If the player is part of a universe,
it must also register with the server and specify a handler to process messages. To
register with the host a string containing the internet address of the computer running
the server is needed. The string for the internet address of your computer may be
obtained by examining the value of ISL+’s LOCALHOST variable (simply type it at
the prompt in the interactions window). During development and testing you can run
all components on your computer by using LOCALHOST to register players with the
server. For example, the run function for a player may look like this:
;; string → world
;; Purpose: To run the game
(define (run a-name)
(big-bang
INIT-WORLD
(on-draw draw-world)
(on-key process-key)
(on-tick update-world)
(stop-when game-over?)
(register LOCALHOST)
(on-receive process-message)
(name a-name)))
There are two new big-bang clauses here. The register clause tells the universe
teachpack the internet address where the server is running. If this clause contains
LOCALHOST, then the server is running on your computer. If this clause contains a
literal string, like "127.0.0.1", then the server is running at the computer found at
that internet address. It is noteworthy that handlers that may create a new world, like
588 26 Introduction to Distributed Programming

process-tick, may return a world or a package. A package is returned when a

message is sent to the server. Otherwise, it suffices to return a world. The world in
a package becomes the next world.
Similarly, the syntax to run the universe server specifies the initial universe and the
event handlers to manage events and update the universe. These events may include
the arrival of a new message, a registration request from a new world, a clock tick,
and a world disconnecting from the universe. For example, the run-server function
for the universe may look like this:
;; Z → universe
;; Purpose: Run the universe server
(define (run-server a-z)
(universe
initU
(on-new add-new-world)
(on-msg process-message)
(on-disconnect rm-world)
(on-tick process-tick)))
The function has a dummy parameter that is never used because ISL+ does not
allow a function to have 0 parameters. Its sole purpose is to allow us to write
a server-running function. The first expression in the universe-expression is the
initial universe (which you define just like you define INIT-WORLD for a player). The
on-new clause specifies the handler to process the request from a new world to join
the universe. The on-msg clause specifies the handler to process arriving messages.
The on-disconnect clause specifies the handler to process the departing of a world
from the universe. The on-tick clause specifies the handler to process clock ticks.
Of these, the initial universe and the on-new and on-msg clauses are required.
Any other handlers are optional and may be used if needed by the application. The
universe teachpack represents the clients connected to the server using a structure
called iworld. The only characteristic of an iworld that we may access is its
name using iworld-name. In order to allow programmers to write tests for server
functions the teachpack has three predefined iworlds that may be used: iworld1,
iworld2, and iworld3. The following are the signatures required by the universe
API for the handlers specified in run-server above:
add-new-world: universe iworld → bundle or universe
process-message: universe iworld message → bundle or universe
rm-world: universe iworld → bundle or universe
process-tick: universe → bundle or universe
A bundle is returned when there is at least one mail to be sent to a player. The
universe in the bundle becomes the next universe. If no mail needs to be sent,
then it suffices for a handler to return a universe.
This syntactical set-up provides a framework to design your distributed program
in a top-down manner. That is, starting from the handlers and then moving on to
auxiliary functions. Be aware that the above description only highlights the features
146 A Chat Application 589

of the universe API that are used in this textbook. If you are interested in further
details about the universe teachpack, you are strongly encouraged to read the
universe documentation found in DrRacket’s Help Desk.

146 A Chat Application

To illustrate the use of the design recipe and of the universe teachpack for dis-
tributed programming the development of a chat tool is presented. The chat tool
developed is not as sophisticated as the chat tools you are probably familiar with, but
it is an interesting first application to develop as you explore distributed program-
ming. The steps of the design recipe are outlined.

146.1 The Components

The chat tool is designed to allow users to share messages with everyone that is
connected to the server. A user types a string of at most length 20 and sends it to the
group of connected users. The server receives a message from a user and broadcasts
it to the rest of the users. There are, therefore, only two components to design: the
chat client and the chat server. All clients in this application execute the same code.
The client is responsible for drawing the status of the chat displaying the latest four
messages and the message partially typed by the user. The partially typed message
may, of course, be empty. The client is also responsible for processing keystrokes by
the user. If the user hits the Enter key when the new message is not empty, the new
message is added as the last message received by the user and is sent to server for
it to be broadcasted to the other users. If the user hits the Backspace key when the
partially written message is not empty, then its last character is deleted. The Shift
and Tab keys are ignored. Any other key is added to the message as long as the
length of the new message is at most 20.
The server is responsible for adding new users to the universe. A new user is
added only if her name is different from the name of every other user. When a new
user is added, a message is sent to all users informing them of the name of the new
user. The server is also responsible for processing the messages that arrive from the
clients. For each message a new mail is created for every user except the sender of
the message.

146.2 Data Definitions

The data definitions needed for the server and for the client may differ for some
applications. In other applications they may be the same. Yet in other applications
some data definitions may be shared, while others are not. Our chat application is of
the third variety.
590 26 Introduction to Distributed Programming

To design the client program a data definition for a text message and a world are
needed. We may define a text message, tm, entered by a client as
(define MAX-TM-LEN 20)

;; An text message (tm) is a string of length <= MAX-TM-LEN of

;; keystrokes that does not contain a return, a backspace, a shift,
;; or a tab.

A constant is defined to represent the maximum length of a text message. A text

message is an enumeration type. It is all strings of keystrokes that do not contain the
outlined keystrokes. The world needs to track the four most recent text messages and
the partially written text message. This is compound data of finite size that suggests
using a structure. The world may be defined as follows:
;; A world is a structure: (make-world tm tm tm tm tm)
;; that contains 5 text messages from left to right:
;; partially written to fourth most recent.
(define-struct world (tm1 tm2 tm3 tm4 tm5))

;; World function template

;; world . . . → . . .
;; Purpose:
;; (define (f-on-world a-world)
;; (. . .(world-tm1 a-world). . .
;; . . .(world-tm2 a-world). . .
;; . . .(world-tm3 a-world). . .
;; . . .(world-tm4 a-world). . .
;; . . .(world-tm5 a-world). . .))
;; Sample worlds
;; (define WORLD-0 (make-world . . . . . . . . . . . . . . .)) . . .
;;
;; Sample expressions for f-on-world
;; (define WORLD-0-VAL . . . WORLD-0 . . .)
;;
;; Tests using sample computations for f-on-world
;; (check-expect (f-on-world WORLD-0 . . .) WORLD-0-VAL) . . .
;;
;; Tests using sample values for f-on-world
;; (check-expect (f-on-world (make-world . . .) . . .) . . .) . . .

;; Sample worlds
(define INIT-WORLD (make-world "" "" "" "" ""))
(define A-WORLD
(make-world "Wanna hang?" "Good thnx" "Good and you?"
"Hi, how are you" "Hi"))
(define B-WORLD (make-world "12345678901234567890" "Guess a number"
"" "" ""))

Here tm1 is the partially written text message and tm5 is the fourth most recent
text message. The template, as expected, outlines all the steps of the design recipe
that must be completed. The sample worlds include the initial world containing all
empty text messages, a world that has no empty text messages, and a world that has
both empty and non-empty text messages and a full partially written message.
146 A Chat Application 591

To design the server program a data definition for a universe is needed. The server
needs to track the worlds in the universe to decide whether a new world may be
connected or to create a list of mails to broadcast a message. Thus, we may define a
universe as follows:
;; A universe is a (listof iworld)

;; Universe function template

;; universe . . . → . . .
;; Purpose:
;; (define (f-on-universe a-universe)
;; (if (empty? a-universe)
;; ...
;; (. . . (first a-universe) . . .
;; . . . (f-on-universe (rest a-universe)) . . .)))
;; Sample universes
;; (define UNIV-0 '())
;; (define UNIV-1 . . .) . . .
;;
;; Sample expressions for f-on-universe
;; (define UNIV-0-VAL . . . UNIV-0 . . .)
;; (define UNIV-1-VAL . . . UNIV-1 . . .)
;;
;; Tests using sample computations for f-on-universe
;; (check-expect (f-on-universe UNIV-0 . . .) UNIV-0-VAL)
;; (check-expect (f-on-universe UNIV-1 . . .) UNIV-1-VAL). . .
;;
;; Tests using sample values for f-on-universe
;; (check-expect (f-on-universe . . . . . .) . . .) . . .

;; ;; Sample universes
(define INIT-UNIV '())
(define A-UNIV (list iworld1 iworld2))
A universe is the list of iworlds connected to the server. The function template to
process a universe is obtained by specializing the template for a (listof X). This
means that in this application the universe is a subtype of (listof X). Therefore,
abstract functions on a (listof X), like map and filter, may be used to process a
universe. There are two sample universes given that there are two varieties. The
sample non-empty universe is built using sample iworlds offered by the API.
592 26 Introduction to Distributed Programming

Fig. 113 Protocol diagram for sending a tm from clienti

Time tm

tm
..
.
tm

Client1 ... Client Client Server

146.3 Communication Protocol

A communication chain is sparked by the keystroke Enter. Think carefully about

what ought to happen when Clienti sends a message. After hitting Enter, the
partially written message is complete and must be sent to the server. Upon receiving
this message, the server must broadcast it to the other users. The protocol diagram
displayed in Fig. 113 illustrates this communication chain. In the diagram there are n
clients and a server. The arrows indicate when messages are sent. The broadcast mes-
sages all start at the same point in the server. The vertical dots indicate that all clients
get the broadcast message except Clienti indicated by no incoming arrow to it.
Observe that there is an outgoing solid arrow from Clienti to the Server. This
means that Clienti may need to marshal its message and that the Server might have
to unmarshal the message. Given that the message, a tm, is suitable for transmission,
marshalling and unmarshalling is not necessary. What do the solid arrows from the
Server to Client1 through Clientn tell us? Essentially the same. The server and
the clients, once again, do not need to marshal and unmarshal because the message
is suitable for transmission.
A communication chain is also sparked when a new client joins the server. The
server must broadcast a message to all the clients that a new user has joined the chat.
The protocol diagram in Fig. 114 illustrates this communication chain. The dashed
line from Clienti to the server is a registration message sent to the universe
system. That is, dashed lines indicate that it is not a message constructed by our
code. Upon allowing the client to join the universe server, the server broadcasts
the new user to all the clients in the universe. To reduce the number of message
varieties we specify that this broadcast message must also be a tm. No marshalling
or unmarshalling is necessary because the message, a tm, is suitable for transmission.
We can now define to-server and to-client messages as follows:
;; A to-server message is a tm

;; A to-client message is a tm
146 A Chat Application 593

Fig. 114 Protocol diagram for a new client joining

Register
Time

tm
..
.
tm

Client1 ... Client Client Server

This concludes this step of the design recipe. An important lesson to remember is that
a component at the beginning of a solid arrow needs the function, if any, to marshal
the message and the component at the end of a solid arrow needs the function, if any,
to unmarshal the message.

146.4 Marshalling and Unmarshalling

The design of the communication protocol revealed that there is only one type of
to-server message and only one type of to-client message. Given that there
is no variety in these types and the message is suitable for transmission, there is no
need to develop marshalling and unmarshalling functions.

146.5 Component Implementation

The different components are independently designed using the protocol diagrams
to guide you. The protocol diagrams need to illustrate of the types of messages
that must be sent for different event-driven communication chains leading to data
definitions for to-server and to-client messages.
594 26 Introduction to Distributed Programming

Fig. 115 The draw-world handler

;; world image Purpose: To draw the given world
(define (draw-world a-world)
(local [(define WIDTH 270) (define HEIGHT 170) (define VSPACE 10)
(define E-SCENE (empty-scene WIDTH HEIGHT))
;; tm image Purpose: Convert the given text to an image
(define (make-tm-img a-tm) (text a-tm 24 "brown"))
;;image natnum>0 image image Purpose:Place given tm image
(define (place-tm img factor scn)
(place-image img
(add1 (/ (image-width img) 2))
(* factor (+ VSPACE (/ (image-height img) 2)))
scn))
;; world image Purpose: To draw the given world
(define (draw-world w)
(local [(define IMG5 (make-tm-img (world-tm5 w)))
(define IMG4 (make-tm-img (world-tm4 w)))
(define IMG3 (make-tm-img (world-tm3 w)))
(define IMG2 (make-tm-img (world-tm2 w)))
(define IMG1 (make-tm-img (world-tm1 w)))]
(add-line
(place-tm IMG5 1
(place-tm IMG4 2
(place-tm IMG3 3
(place-tm IMG2 4 (place-tm IMG1 6 E-SCENE)))))
0 (* 5 (+ VSPACE (/ (image-height IMG2) 2)))
(sub1 WIDTH) (* 5 (+ VSPACE (/ (image-height IMG2) 2)))
"red")))]
(draw-world a-world)))

146.5.1 Client

The client, according to our design, must draw the world, process keystrokes, have
a (unique) name, register with the server, and process messages. The run function is
;; string → world
;; Purpose: Run the chat program
(define (run a-name)
(big-bang
INIT-WORLD
(on-draw draw-world)
(on-key process-key)
(name a-name)
(register LOCALHOST)
(on-receive process-message)))
The draw-world function must draw the four most recent tms and the partially
written tm. This can be rendered by drawing the four most recent tms above each
other with the least recent on the top and most recent on the bottom. After this a
red line may be drawn to visually separate them from the partially written tm at the
146 A Chat Application 595

Fig. 116 The draw-world handler’s tests

;; Tests using sample values for draw-world

(check-expect (draw-world INIT-WORLD) )

(check-expect (draw-world A-WORLD) )

(check-expect (draw-world (make-world "12345678901234567890" "" "" "" ""))

bottom of the scene. No messages need to be sent to the server when the world is
rendered. The function to draw the world is displayed in Fig. 115 and its tests are
displayed in Fig. 116. Functions to create and place tm images are encapsulated.
Observe that there can only be tests using sample values because encapsulated
functions may not be used to write tests.
The manner in which process-key processes keystrokes is exemplified by the
sample expressions in Fig. 117. If the given key is Enter (i.e., "\r") and the partially
written tm is empty, the keystroke is ignored. Otherwise, following the protocol
diagram in Fig. 113, a package is created that sends the partially written tm to the
server and updates the most recently tms. If the given key is Backspace (i.e., “\b”),
the keystroke is ignored if the partially written tm is the empty string. Otherwise,
the last character of the partially written tm is removed. If the given key is either
"Shift" or Tab (i.e., "\t"), the keystroke is ignored. Any other keystroke is added
to the partially written tm if its length is less than or equal to MAX-TM-LEN and
ignored otherwise. The function definition derived from the sample expressions is
displayed in Fig. 118. The fact the function may need to send a message to the
server is reflected in the signature that states the return value may be a world or a
package. The function locally defines variables for the five tms in the given world.
This avoids peppering the conditional expression with expressions using world-
selectors. The tests are displayed in Fig. 117. The tests using sample values illustrate
how the function is expected to behave, while the tests using sample computations
illustrate that the function computes the values in the manner outlined by the sample
expressions.
The remaining handler to design is process-message. The protocol diagrams
in Figs. 113 and 114 inform us that the only message that a client receives is a tm.
The received tm becomes the most recently received message and the other messages
are moved to the next oldest slots. The fourth most recent message is discarded to
596 26 Introduction to Distributed Programming

Fig. 117 The process-key handler sample expressions and tests

;; Sample expressions for process-key
(define INITW-RET INIT-WORLD)
(define AW-RET (make-package
(make-world "" (world-tm1 A-WORLD) (world-tm2 A-WORLD)
(world-tm3 A-WORLD) (world-tm4 A-WORLD))
(world-tm1 A-WORLD)))
(define INITW-BACK INIT-WORLD)
(define AW-BACK (make-world
(substring (world-tm1 A-WORLD)
0
(sub1 (string-length (world-tm1 A-WORLD))))
(world-tm2 A-WORLD) (world-tm3 A-WORLD)
(world-tm4 A-WORLD) (world-tm5 A-WORLD)))
(define AW-SHIFT A-WORLD)
(define AW-TAB A-WORLD)
(define BW-M B-WORLD)
(define INITW-A (make-world (string-append (world-tm1 INIT-WORLD) "A")
"" "" "" ""))

;; Tests using sample computations for process-key

(check-expect (process-key INIT-WORLD "\r") INITW-RET)
(check-expect (process-key A-WORLD "\r") AW-RET)
(check-expect (process-key INIT-WORLD "\b") INITW-BACK)
(check-expect (process-key A-WORLD "\b") AW-BACK)
(check-expect (process-key A-WORLD "shift") AW-SHIFT)
(check-expect (process-key A-WORLD "\t") AW-TAB)
(check-expect (process-key B-WORLD "M") BW-M)
(check-expect (process-key INIT-WORLD "A") INITW-A)

;; Tests using sample values for process-key

(check-expect (process-key (make-world "" "B" "A" "" "") "C")
(make-world "C" "B" "A" "" ""))
(check-expect (process-key (make-world "" "B" "A" "" "") "\r")
(make-world "" "B" "A" "" ""))
(check-expect (process-key (make-world "C" "B" "A" "" "") "\b")
(make-world "" "B" "A" "" ""))

make room for the new tm. The message-processing handler may be implemented
as follows:
;; world message → world
;; Purpose: Process the given message
(define (process-message a-world a-message)
(make-world (world-tm1 a-world)
a-message
(world-tm2 a-world)
(world-tm3 a-world)
(world-tm4 a-world)))
146 A Chat Application 597

Fig. 118 The process-key function definition

;; world key world or package
;; Purpose: Return next world after the given keystroke
(define (process-key a-world a-key)
(local [(define tm1 (world-tm1 a-world))
(define tm2 (world-tm2 a-world))
(define tm3 (world-tm3 a-world))
(define tm4 (world-tm4 a-world))
(define tm5 (world-tm5 a-world))]
(cond [(string=? a-key "\r")
(if (string=? tm1 "") a-world
(make-package (make-world "" tm1 tm2 tm3 tm4) tm1))]
[(string=? a-key "\b")
(if (string=? tm1 "") a-world
(make-world (substring tm1 0 (sub1 (string-length tm1)))
tm2 tm3 tm4 tm5))]
[(or (string=? a-key "shift") (string=? a-key "\t")) a-world]
[else (if (= (string-length tm1) MAX-TM-LEN) a-world
(make-world (string-append tm1 a-key)
tm2 tm3 tm4 tm5))])))

;; Sample expressions for process-message

(define IWM (make-world (world-tm1 INIT-WORLD)
"It’s me"
(world-tm2 INIT-WORLD)
(world-tm3 INIT-WORLD)
(world-tm4 INIT-WORLD)))
(define BWM (make-world (world-tm1 B-WORLD)
"I think so!"
(world-tm2 B-WORLD)
(world-tm3 B-WORLD)
(world-tm4 B-WORLD)))

;; Tests using sample computations for process-message

(check-expect (process-message INIT-WORLD "It’s me") IWM)
(check-expect (process-message B-WORLD "I think so!") BWM)

;; Tests using sample computations for process-message

(check-expect
(process-message
(make-world "What d" "You and I?" "" "" "")
"No way")
(make-world "What d" "No way" "You and I?" "" ""))
Observe that the sample expressions and the test using a sample value illustrate
that the newly arriving tm becomes the new world’s most recent message. They
also illustrate that the partially written tm remains unchanged when a new message
arrives.
598 26 Introduction to Distributed Programming

Fig. 119 Sample expressions and tests for adding a world handler
;; Sample expressions for add-new-world
(define ADD-INITU
(make-bundle (list iworld1)
(map
( (iw)
(make-mail iw (string-append (iworld-name iworld3)
" has joined")))
INIT-UNIV)
()))
(define ADD-AUNIV
(make-bundle (list iworld3 iworld1 iworld2)
(map
( (iw)
(make-mail iw (string-append (iworld-name iworld3)
" has joined")))
A-UNIV)
()))

(define ADD-REPEAT A-UNIV)

;; Tests using sample computations for add-new-world

(check-expect (add-new-world INIT-UNIV iworld1) ADD-INITU)
(check-expect (add-new-world A-UNIV iworld3) ADD-AUNIV)
(check-expect (add-new-world A-UNIV iworld1) ADD-REPEAT)

;; Tests using sample values for add-new-world

(check-expect (add-new-world (list iworld1) iworld2)
(make-bundle (list iworld2 iworld1)
(list (make-mail iworld1 "iworld2 has joined"))
()))

Given that there are no auxiliary functions needed and that all the required
handlers are implemented, the design of the client code is done. Run the tests and
verify that they all pass. At this point do not call the run function because there is no
server for the client to connect to. Once you are satisfied that the tests pass, save two
or more copies of the program in the same directory. Each copy will be a different
component when executed.

146.5.2 Server

According to our design the server must process new clients attempting to register
and must process messages. To add a new iworld to the server the name of the
incoming iworld must be different from the name of any iworld already in the
server. If it is not, the given world is not added and the universe remains unchanged.
This is illustrated by the third sample expression in Fig. 119 that simply is the value of
the given universe. If the name of the arriving iworld is different, then it is added to
the universe. This is illustrated by the first and second sample expressions in Fig. 119.
146 A Chat Application 599

Fig. 120 The handler to add a new world to the universe

;; universe iworld bundle Purpose: Add the given world to the universe
;; ASSUMPTION: The name of the new iworld is a string
(define (add-new-world a-universe an-iworld)
(if (member? (iworld-name an-iworld) (map iworld-name a-universe))
a-universe
(local
[(define new-univ (cons an-iworld a-universe))
(define new-mails
(map
( (iw)
(make-mail
iw
(string-append (iworld-name an-iworld) " has joined")))
a-universe))]
(make-bundle new-univ new-mails ()))))

Observe that the outgoing arrows from the server in Fig. 114 are implemented using
map to create a mail for each iworld in the universe. The created mails have a tm
informing the recipient of the new client that has joined the chat. In the first sample
expression the incoming iworld is added to an empty universe. In the second sample
expression the incoming iworld is added to the front of a non-empty universe.
In both cases no worlds are disconnected from the universe and, therefore, the third
argument to make-bundle is '(). The tests illustrate that the function correctly
computes the values of the sample expressions and illustrate the creation of a bundle
using concrete values. Abstracting over the sample expressions yields the function
displayed in Fig. 120.
The incoming arrows to the server in the protocol diagrams in Figs. 113 and 114
inform us that the server only receives messages containing tms. The handler for
processing messages, therefore, only needs to forward an incoming message to all
the worlds except the sender as illustrated in Fig. 113. To achieve this the universe
must be filtered to exclude the iworld that sent the message and mails must be
created for the remaining iworlds. This is illustrated by the sample expressions in
Fig. 121. Locally a list of mails is created using filter and map. The body of the
local-expression creates a bundle with an unchanged universe, the new mails, and
an empty list of iworlds to disconnect. It may seem silly to write a sample expression
using an empty universe (i.e., INIT-UNIV) because there are no iworlds that could
have sent the message. Although this observation is correct, do not forget that the
goal is to validate the function for all universe subtypes. Therefore, a test using the
empty universe must be included. Abstracting over the sample expressions yields
the server’s process-message function.
This completes the design of the server. Run the tests to make sure they all pass.
600 26 Introduction to Distributed Programming

Fig. 121 The universe’s handler to process a message

;; universe iworld message bundle
;; Purpose: To process the given message from the given world
(define (process-message a-univ an-iw a-mess)
(local [(define new-mails (map ( (iw) (make-mail iw a-mess))
(filter
( (iw)
(not (equal? (iworld-name an-iw)
(iworld-name iw))))
a-univ)))]
(make-bundle a-univ new-mails ())))

;; Sample Expressions for process-message

(define AUNIV-MESS
(local [(define new-mails
(map ( (iw) (make-mail iw "Hi!"))
(filter ( (iw) (not (equal? (iworld-name iworld2)
(iworld-name iw))))
A-UNIV)))]
(make-bundle A-UNIV new-mails ())))

(define IUNIV-MESS
(local [(define new-mails
(map ( (iw) (make-mail iw "Hi!"))
(filter ( (iw) (not (equal? (iworld-name iworld2)
(iworld-name iw))))
INIT-UNIV)))]
(make-bundle INIT-UNIV new-mails ’())))

;; Tests using sample computations for process-message

(check-expect (process-message A-UNIV iworld2 "Hi!") AUNIV-MESS)
(check-expect (process-message INIT-UNIV iworld2 "Hi!") IUNIV-MESS)

;; Tests using sample values for process-message

(check-expect (process-message (list iworld2 iworld3) iworld3 "OK")
(make-bundle (list iworld2 iworld3)
(list (make-mail iworld2 "OK"))
()))

146.6 Running the Chat Tool

To run the chat tool on your machine first, run the server and then run one or more
clients. Type messages in each of the clients and see how the messages sent appear
in the other clients. Figure 122 displays a snapshot of a chat session with two clients.
The first client to join is Marco that gets a message when Francisco joins. Both
clients in the snapshot have partially written messages that only they can see because
they have not been sent yet. The window in the middle is the server window. In it you
can see all the server events like clients joining, messages received, and messages
sent.
147 What Have We Learned in This Chapter? 601

Fig. 122 A snapshot of the chat tool running on LOCALHOST

Once you are fairly sure that the chat tool is working, you may now use it to
chat with your fellow classmates. Pick a classmate to run the server and get their
internet address. All clients need to substitute LOCALHOST with a string containing
the internet address of the classmate running the server. Once all clients are ready
have the chosen classmate run the server and a client. Once the server is running call
the run function and chat away!

** Ex. 326 — Chat tools are susceptible to security breaches or misuse. In

our chat tool, for example, a joining client may have a name that is too long.
Redesign the tool reject clients with names that are too long and send such
clients a tm informing them.

*** Ex. 327 — The chat tool does not display the name of the sender of the
message. Redesign the chat tool to display the name of the sender and the
message all in one line. You may want to increase the maximum length of a tm
for this.
***** Ex. 328 — Redesign the chat tool to allow users to send emojis like a
smiley or a heart. Note that an emoji is an image and images are not suitable for
transmission.

147 What Have We Learned in This Chapter?

The important lessons of Chap. 26 are summarized as follows:

• Distributing programming is ubiquitous in modern society.
• Dividing a problem into several tasks and writing a program for each task that
communicates with the programs for other tasks is called distributed program-
ming.
• In distributed programming components cooperate with each other to solve a
problem using message-passing.
602 26 Introduction to Distributed Programming

• A communication protocol defines the messages that may be exchanged in a

communication chain sparked by an event. The design of this protocol is part of
distributed programming.
• If necessary, data is marshaled by a sending component to make it suitable for
transmission.
• If necessary, a component receiving a message unmarshals it to reconstruct the
original data.
• Marshalling and unmarshalling are an example of the common practice of using
inverse functions in programming.
• A pervasively used distributed system architecture is the client–server architec-
ture. A server provides services and a client requests services using message-
passing.
• A communication protocol is partly specified using protocol diagrams for different
event-based communication chains.
• In a protocol diagram, a component at the beginning of a solid arrow needs the
function, if any, to marshal the message and the component at the end of a solid
arrow needs the function, if any, to unmarshal the message.
• A thin server is one that provides a minimal number of services or very little
computing.
• A thick server provides services that involve computing that directly contribute
to solving a problem.
• The design recipe for distributed programming is:
1. Divide the problem into components.
2. Draft data definitions for the different components.
3. Design a communication protocol.
4. Design marshalling and unmarshalling functions.
5. Design and implement the components.
6. Test your program.
• The universe teachpack provides the functionality to develop distributed multi-
player games.
Chapter 27
Aliens Attack Version 6

Our next goal is to refine Aliens Attack to allow for multiple players. The game
is used to explore how to design a distributed program using both a thin and a
thick server. First, however, this chapter explores a single-player game refinement
that makes incorporating multiple players easier. The goal is to have all the world-
related data definitions needed for multiple players in place and implemented for a
single-player game. This new implementation of Aliens Attack will serve as the base
code to design a multiple player game. To a player the game will look exactly the
same as before. That is, no new features are added to the game. All that is refined is
the implementation of the existing game.
The refinement is inspired in the expected changes that are needed for a multiplayer
game. Consider the snapshot of multiplayer Aliens Attack displayed in Fig. 123. Take
your time to think about what changes are needed and what elements remain the same
in the world data definition. One difference is that the game has multiple rockets.
The rest of the elements remain the same. That is, there are still multiple aliens, a
direction, and multiple shots. Our initial task, therefore, is clear: define a world
that can have multiple rockets. In this version of the game the number of players
remains at 1 and, therefore, there is no need for a server, a communication protocol,
and all the other necessary features for a distributed program.

148 Refining the world Data Definition

The world data definition needs to change from having a rocket to having multiple
rockets that are all allies in repelling the invading alien army. It may be tempting to
simply change the world data definition to have allies instead of a single rocket. We
need, however, to be careful and perform a more detailed problem analysis. Think
about how a player ought to start the game. Should a player start, as in the single-
player game, with a world that has a full army of aliens moving in some hardwired
direction with no shots? This is likely to be fine for the first player that joins the game.
How about the players that join afterwards? These players cannot start with a world
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 603
M. T. Morazán, Animated Problem Solving, Texts in Computer Science,
https://doi.org/10.1007/978-3-030-85091-3_27
604 27 Aliens Attack Version 6

Fig. 123 Multiple-player Aliens attack

that has a full army of aliens and no shots because the first player may have already
started shooting and neutralizing invaders. This means that the starting world for a
player must be provided by the server. For the first player, it will be what is called
INIT-WORLD in the single-player game. For the other players, it must be the state of
the game when they join.
The above analysis allows us to take a top-down approach to refining the world
data definition. It informs us that there must be variety in the world data definition.
When a player joins the game, her world must be uninitialized and wait for the
server to provide the value of the world. After the server provides this value, the
world becomes a structure. We may now define the world type as follows:
;; A world is either
;; 1. 'uninitialized
;; 2. a structure: (make-world lor loa dir los)
(define-struct world (allies aliens dir shots))
Observe that world is now a union type with two varieties. Furthermore, the second
variety is a structure that contains a lor, instead of rocket, to represent the allies.
The template for a function on a world is:
;; TEMPLATE FOR FUNCTIONS ON A WORLD
;; world . . . → . . .
;; Purpose:
148 Refining the world Data Definition 605

;; (define (f-on-world w . . .)
;; (if (eq? a-world 'uninitialized)
;; ...
;; (. . . (world-allies w). . . (world-aliens w)
;; . . . (world-dir w). . . (world-shots w))))
;;
;; ;; Sample instances of world
;; (define WORLD1 'uninitialized)
;; (define WORLD2 (make-world . . . . . . . . . . . .))
;;
;; ;; Sample expressions for f-on-world
;; (define WORLD1-VAL . . . WORLD1 . . .)
;; (define WORLD2-VAL . . . WORLD2 . . .) . . .
;;
;; ;; Tests using sample computations for f-on-world
;; (check-expect (f-on-world WORLD1 . . .) WORLD1-VAL)
;; (check-expect (f-on-world WORLD2 . . .) WORLD2-VAL) . . .
;;
;; ;; Tests using sample values for f-on-world
;; (check-expect (f-on-world . . . . . .) . . . ) . . .
This new world function template informs us that the refined functions that process
a world must now contain a conditional in the body of the function. In addition, there
must be at least two sample worlds, two sample expressions, and two tests using
sample computations. This is the road map that guides the refinements of the game’s
handlers. It is also noteworthy that the tests using sample values must now construct
worlds based on the new data definition.
The next step is to define lor. Given that the number of allies is not known in
advance, it is data of arbitrary size. The natural choice that comes to mind is a list.
We define lor as follows:
;; An lor is a (listof ally)
The template for a lor is given by specializing the template for a (listof X) with
X = ally.
Can ally be replaced with rocket to complete the data definition refinements?
This requires further problem analysis. In the game each player is expected to control
their own rocket. Assume that an ally is simply a rocket. Consider the player, Pi ,
moving her rocket. Which rocket in Pi ’s lor ought to be moved? If there is only
one rocket in the list, the answer is trivial. What if there are multiple rockets? How
does the program decide which to move? Clearly, Pi wants to move her rocket and
not another player’s rocket. This means that an ally cannot simply be a rocket
because we need to distinguish which player owns each rocket. To do so each ally
606 27 Aliens Attack Version 6

rocket is associated with the name of the world that “owns” it. An ally is, therefore,
defined as follows:
;; An ally is a structure, (make-ally rocket string), with
;; a player’s rocket and name
(define-struct ally (rocket name))
With this data definition, it becomes clear which ally needs to be moved when Pi
moves her rocket: the ally that has Pi ’s name. The template for a function on an ally
is:
;; TEMPLATE FOR FUNCTIONS ON AN ALLY
;; ally . . . → . . .
;; Purpose:
;; (define (f-on-ally an-ally . . .)
;; (. . . (world-allies an-ally) . . . (world-aliens an-ally)
;; . . . (world-dir an-ally) . . . (world-shots an-ally))))
;;
;; ;; Sample instances of ally
;; (define ALLY1 (make-ally . . . . . .))
;;
;; ;; Sample expressions for f-on-ally
;; (define ALLY1-VAL . . . ALLY1 . . .) . . .
;;
;; ;; Tests using sample computations for f-on-ally
;; (check-expect (f-on-ally ALLY1 . . .) ALLY-VAL) . . .
;;
;; ;; Tests using sample values for f-on-ally
;; (check-expect (f-on-ally . . . . . .) . . . ) . . .
Given that there are no further data definitions to develop, we define ally, lor,
and world sample instances as follows:
(define MY-NAME "Yoli Ortega")

;; Sample instances of ally

(define INIT-ALLY (make-ally INIT-ROCKET MY-NAME))
(define INIT-ALLY2 (make-ally INIT-ROCKET2 MY-NAME))

;; Sample instances of lor

(define INIT-ALLIES (list INIT-ALLY))
(define INIT-ALLIES2 (list INIT-ALLY2))

(define INIT-WORLD (make-world INIT-ALLIES INIT-LOA

INIT-DIR INIT-LOS))
(define INIT-WORLD2 (make-world INIT-ALLIES2 (list INIT-ALIEN2)
DIR2 (list SHOT2)))
149 The draw-world Refinement 607

(define WORLD3 (make-world (list (make-ally 7 MY-NAME))

(list (make-posn 3 3))
'right
(list (make-posn 3 3))))
(define UNINIT-WORLD 'uninitialized)
A constant is defined for the world name and it is used to create ally instances.
The initial ally contains the player’s rocket and the initial lor only contains this
initial ally. This initial lor is used to build the initial world. The other constants
used to build this world are unchanged from Aliens Attack version 5 (in Chap. 23).
The initial world in this version of the game is not uninitialized because the goal is
to refine the single-player game to use the data definitions that allow multiple player
support.
This is reflected in the new definition of the run function:
; string → world
; Purpose: To run the game
(define (run a-name)
(local [(define TICK-RATE 1/4)]
(big-bang
INIT-WORLD
[on-draw draw-world]
[name MY-NAME]
[on-key process-key]
[on-tick process-tick TICK-RATE]
[stop-when game-over? draw-last-world])))
The world argument to big-bang is not 'uninitialized. Further observe that
this function now has a dummy parameter, a-name, that is never used. The name
associated with the world is now defined by the constant MY-NAME. In this manner,
the name associated with the world is made the same as the name associated with
this world’s rocket.
The following sections outline the refinements of the handlers. These sections
describe the end result and do not outline the entire development process. That is,
they present the final encapsulated functions and not the full development using all
the steps of the design recipe. As you read, remember that all auxiliary functions
were developed and tested before being made local to the appropriate handler.

149 The draw-world Refinement

The draw-world handler from Aliens Attack version 5 has the following structure:
;; world → scene
;; Purpose: To draw the world in E-SCENE
(define (draw-world a-world)
(local [. . .]
(draw-world a-world)))
608 27 Aliens Attack Version 6

According to the template for functions on a world, a conditional expression is

needed to distinguish among the world varieties. This conditional may be placed in
the body of the local-expression or in the body of the local draw-world function.
We shall take the first option to keep changes to local functions to a minimum.
We may, therefore, refactor the sample expressions to distinguish among the world
varieties as follows:
;; Sample expressions for draw-world
(define WORLD-SCN1 (if (eq? INIT-WORLD UNINIT-WORLD)
E-SCENE
(draw-world INIT-WORLD)))

(define WORLD-SCN2 (if (eq? UNINIT-WORLD UNINIT-WORLD)

E-SCENE
(draw-world UNINIT-WORLD)))
The sample expressions are valid because there is a local function, draw-world,
by the same name that draws the world when it is a structure. Therefore, after
abstracting over the sample expressions, the first refinement of this handler is:
;; world → scene
;; Purpose: To draw the world in E-SCENE
(define (draw-world a-world)
(local [. . .]
(if (eq? a-world UNINIT-WORLD)
E-SCENE
(draw-world a-world))))
Observe that the local draw-world function is only called when the given world
is a structure. That is, there is now an assumption made about the type of world
given to the local draw-world function. Also observe that E-SCENE is used in the
sample expressions. This means that it can no longer be local to draw-world. The
following definitions are moved to the global level in order for them to be in scope
for the sample expressions:
(define E-SCENE-COLOR S'pink)
(define E-SCENE
(empty-scene E-SCENE-W E-SCENE-H E-SCENE-COLOR))
The new local draw-world function requires small refinements:
;; world → scene
;; Purpose: To draw the world in E-SCENE
;; ASSUMPTION: The given world is a structure
(define (draw-world a-world)
(draw-los (world-shots a-world)
(draw-loa (world-aliens a-world)
(draw-allies (world-allies a-world)
E-SCENE))))
149 The draw-world Refinement 609

Observe that the assumption is made explicit for the benefit of any reader of the
code. In addition, an auxiliary function to draw the allies is needed. The allies are
drawn in the empty scene just like the rocket is drawn in Aliens Attack version 5.
The design of a local function draw-allies uses draw-lox-maker developed
in Sect. 125. This function may be used to create a function to draw the allies as
follows:
;; lor scene → scene
;; Purpose: Draw the given allies in the given scene
(define draw-allies
(draw-lox-maker (λ (an-ally scn)
(if (string=? (ally-name an-ally) MY-NAME)
(draw-rocket (ally-rocket an-ally) scn)
(draw-ally (ally-rocket an-ally)
scn)))))
The function provided as input to draw-lox-maker draws an ally by distinguishing
between the ally representing the world’s rocket and the rest of the allies. The
world’s rocket is drawn using the existing draw-rocket function. The other allies
must be drawn differently so that the player can easily see her rocket. This may
be accomplished by implementing a new local function, draw-ally, that uses a
different rocket image than that used by draw-rocket as follows:
;; rocket scene → scene
;; Purpose: To draw the rocket in the given scene
(define (draw-ally a-rocket a-scene)
(draw-ci ROCKET-IMG2 a-rocket ROCKET-Y a-scene))
This means that we now have two functions, draw-rocket and draw-ally, that are
almost identical. To eliminate the code repetition we may abstract over the functions
to create a curried function that returns a rocket-drawing function that is specialized
using the rocket image used. This results in the following local functions:
;; image → (rocket scene → scene)
;; Purpose: Create a rocket drawing function
(define (draw-ally-maker rocket-img)
(local [;; rocket scene → scene
;; Purpose: To draw the rocket in the given scene
(define (draw-r a-rocket a-scene)
(draw-ci rocket-img a-rocket ROCKET-Y a-scene))]
draw-r))

;; rocket scene → scene

;; Purpose: To draw the rocket in the given scene
(define draw-rocket (draw-ally-maker ROCKET-IMG))

;; rocket scene → scene

;; Purpose: To draw the rocket in the given scene
(define draw-ally (draw-ally-maker ROCKET-IMG2))
610 27 Aliens Attack Version 6

Both draw-rocket and draw-ally are refactored to use draw-ally-maker. Fi-

nally, ROCKET-IMG2 is a locally defined variable in draw-world. The following are
all the new local constant definitions used to create it:
(define FUSELAGE-COLOR2 'gold)
(define FUSELAGE2 (mk-fuselage-img FUSELAGE-COLOR2))
(define ROCKET-MAIN2 (mk-rocket-main-img WINDOW
FUSELAGE2
BOOSTER))
(define ROCKET-IMG2 (mk-rocket-ci ROCKET-MAIN2 NACELLE))
You may, of course, feel free to personalize your game by designing your own rocket
images using colors of your choice.
The final step is to refine the tests for draw-world as follows:
;; Tests using sample computations for draw-world
(check-expect (draw-world INIT-WORLD) WORLD-SCN1)
(check-expect (draw-world UNINIT-WORLD) WORLD-SCN2)

;; Tests using sample computations for draw-world

;; Added NO-SHOT to fully test draw-shot
(check-expect (draw-world (make-world (list INIT-ALLY)
(list INIT-ALIEN2)
DIR2
(list SHOT2 NO-SHOT)))

)
A test to draw the uninitialized world is added. The test using INIT-WORLD remains
unchanged. The test using a sample value is refactored to draw a world that has an
lor.

150 The process-key Refinement

The process-key handler from Aliens Attack version 5 has the following structure:
150 The process-key Refinement 611

;; world key → world

;; Purpose: Process a key event to return next world
(define (process-key a-world a-key)
(local [. . .]
(process-key a-world a-key)))
As suggested by the new template for functions on a world, a conditional expression is
needed to process an instance of the new world data definition. We can immediately
update the sample expressions to illustrate the use of such a conditional expression:
;; Sample expressions for process-key
(define KEY-RVAL (if (eq? INIT-WORLD UNINIT-WORLD)
INIT-WORLD
(process-key INIT-WORLD "right")))
(define KEY-LVAL (if (eq? INIT-WORLD UNINIT-WORLD)
INIT-WORLD
(process-key INIT-WORLD "left")))
(define KEY-SVAL (if (eq? INIT-WORLD UNINIT-WORLD)
INIT-WORLD
(process-key INIT-WORLD " ")))
(define KEY-SVAL2 (if (eq? INIT-WORLD2 UNINIT-WORLD)
INIT-WORLD2
(process-key INIT-WORLD2 " ")))
(define KEY-OVAL (if (eq? INIT-WORLD2 UNINIT-WORLD)
INIT-WORLD2
(process-key INIT-WORLD2 "m")))
(define KEY-NAOA (if (eq? UNINIT-WORLD UNINIT-WORLD)
UNINIT-WORLD
(process-key UNINIT-WORLD " ")))
Once again, such sample expressions can only be written because process-key has
a local function with the same name. Observe that process-key is only called when
the world is a structure. Abstraction over the sample expressions yields a handler
with the following structure:
;; world key → world
;; Purpose: Process a key event to return next world
(define (process-key a-world a-key)
(local [. . .]
(if (eq? a-world UNINIT-WORLD)
a-world
(process-key a-world a-key))))
The tests using sample computations do not need to be updated. The tests using
sample values, on the other hand, do need to be updated to properly construct world
instances. These instances only have the player’s rocket in the list of allies and are
refined as follows:
612 27 Aliens Attack Version 6

;; Tests using sample values for process-key

(check-expect (process-key
(make-world
(list (make-ally
(sub1 MAX-CHARS-HORIZONTAL)
MY-NAME))
INIT-LOA
'right
INIT-LOS)
"right")
(make-world
(list (make-ally
(sub1 MAX-CHARS-HORIZONTAL)
MY-NAME))
INIT-LOA
'right
INIT-LOS))
(check-expect (process-key
(make-world (list (make-ally 0 MY-NAME))
INIT-LOA
'left
INIT-LOS)
"left")
(make-world (list (make-ally 0 MY-NAME))
INIT-LOA
'left
INIT-LOS))
(check-expect (process-key
(make-world (list (make-ally 0 MY-NAME))
INIT-LOA
'left
INIT-LOS)
"o")
(make-world (list (make-ally 0 MY-NAME))
INIT-LOA
'left
INIT-LOS))
(check-expect (process-key INIT-WORLD2 ";") INIT-WORLD2)
(check-expect (process-key (make-world
(list (make-ally 0 MY-NAME))
INIT-LOA
'left
INIT-LOS)
" ")
150 The process-key Refinement 613

(make-world (list (make-ally 0 MY-NAME))

INIT-LOA
'left
(cons (make-posn 0 MAX-IMG-Y) '())))
(check-expect (process-key (make-world
(list (make-ally 0 MY-NAME))
INIT-LOA
'left
(cons SHOT2 '()))
"left")
(make-world (list (make-ally 0 MY-NAME))
INIT-LOA
'left
(cons SHOT2 '())))
The local function process-key processes a world and, therefore, must be
refined to use the new world data definition. If the player moves her rocket, then
the ally that must be moved is the one that has the player’s MY-NAME. If the player
shoots, then the image-x coordinate for the new shot is the player’s rocket image-x
coordinate. The function is updated as follows:
;; world key → world
;; Purpose: Process a key event to return next world
;; ASSUMPTION: The given world is a structure
(define (process-key a-world a-key)
(cond [(key=? a-key "right")
(make-world (move-ally-right MY-NAME
(world-allies a-world))
(world-aliens a-world)
(world-dir a-world)
(world-shots a-world))]
[(key=? a-key "left")
(make-world (move-ally-left MY-NAME
(world-allies a-world))
(world-aliens a-world)
(world-dir a-world)
(world-shots a-world))]
[(key=? a-key " ")
(make-world (world-allies a-world)
(world-aliens a-world)
(world-dir a-world)
(cons (process-shooting
(ally-rocket
(get-ally MY-NAME
(world-allies a-world))))
(world-shots a-world)))]
[else a-world]))
614 27 Aliens Attack Version 6

The assumption that the world is a structure is made explicit. Observe that three
new auxiliary functions are needed: move-ally-right, move-ally-left, and
get-ally.
Let us start with get-ally. This function must extract the ally that has MY-NAME
in it. If this function works, then the signature for process-shooting is satisfied
by the above function. The extraction of the correct ally may be done using filter
as follows:
;; string lor → ally
;; Purpose: Extract ally with given name
;; ASSUMPTIONS: There is a single ally with given name
(define (get-ally a-name a-lor)
(first (filter (λ (an-ally)
(string=? a-name (ally-name an-ally)))
a-lor)))
Observe that it is assumed that there is a single ally with the given name in the given
lor. This means that the list returned by filter contains a single ally and all that
is needed to extract the ally using first. The function given to filter compares
the name given to get-ally with the name of a given ally. Remember that this
function was first designed and tested before being made local to process-key.
The functions to move the player’s rocket are virtually the same. They only
vary by the function used to move the player’s rocket: move-rckt-right or
move-rckt-left. This suggests creating a curried function to create specialized
ally-moving functions. The input to this function is a rocket-moving function and
is implemented as follows:
;; (rocket → rocket) → (string lor → lor)
;; Purpose: Make an ally-moving function
(define (make-ally-mover move-rckt)
(λ (a-name a-lor)
(map (λ (an-ally)
(if (string=? a-name (ally-name an-ally))
(make-ally (move-rckt (ally-rocket an-ally))
(ally-name an-ally))
an-ally))
a-lor)))
The returned function takes as input a name and an lor. The given lor is traversed
using map. The function given to map compares the name given to the returned
function with the name of the ally in the given lor. If they match a new ally is
constructed using the rocket-moving function used to specialize the ally-moving
function. Otherwise, the ally from the given lor is unchanged. The ally-moving
functions are implemented as follows:
;; string lor → lor
;; Purpose: Move ally with given name right
(define move-ally-right (make-ally-mover move-rckt-right))
151 The process-tick Refinement 615

;; string lor → lor

;; Purpose: Move ally with given name left
(define move-ally-left (make-ally-mover move-rckt-left))

151 The process-tick Refinement

As with the previous handlers, process-tick must be refined using the new world
data definition. The sample expressions may be refined to be:
;; Sample expressions for process-tick
(define AFTER-TICK-WORLD1 (if (eq? INIT-WORLD UNINIT-WORLD)
INIT-WORLD
(process-tick INIT-WORLD)))

(define AFTER-TICK-WORLD2 (if (eq? INIT-WORLD2 UNINIT-WORLD)

INIT-WORLD2
(process-tick INIT-WORLD2)))

(define AFTER-TICK-UNINITW (if (eq? UNINIT-WORLD UNINIT-WORLD)

UNINIT-WORLD
(process-tick UNINIT-WORLD)))
As with process-key, writing such expressions only works because process-tick
has a local function with the same name. Abstracting over the sample expression
yields a new process-tick with the following structure:
;; world → world
;; Purpose: Create a new world after a clock tick
(define (process-tick a-world)
(local [. . .]
(if (eq? a-world UNINIT-WORLD)
a-world
(process-tick a-world))))
The tests using sample values are also updated to conform to the new world data
definition. They now create world sample instances with an lor:
(check-expect
(process-tick (make-world
(list (make-ally INIT-ROCKET MY-NAME))
(cons (make-posn 1 5) '())
'left
INIT-LOS))
616 27 Aliens Attack Version 6

(make-world (list (make-ally INIT-ROCKET MY-NAME))

(cons (make-posn MIN-IMG-X 5) '())
'down
INIT-LOS))

(check-expect
(process-tick (make-world
(list (make-ally INIT-ROCKET MY-NAME))
(list (make-posn 2 5))
'left
(list (make-posn 1 6) NO-SHOT)))
(make-world (list (make-ally INIT-ROCKET MY-NAME))
'()
'left
'()))

(check-expect
(process-tick (make-world
(list (make-ally INIT-ROCKET2 MY-NAME))
(list (make-posn
(- MAX-CHARS-HORIZONTAL 2)
10))
'right
(list SHOT2)))
(make-world (list (make-ally INIT-ROCKET2 MY-NAME))
(cons (make-posn MAX-IMG-X 10) '())
'down
(list (make-posn (posn-x SHOT2)
(sub1 (posn-y SHOT2))))))

(check-expect
(process-tick (make-world
(list (make-ally INIT-ROCKET2 MY-NAME))
(list (make-posn MAX-IMG-X 2))
'down
(list (make-posn 15 6))))
(make-world (list (make-ally INIT-ROCKET2 MY-NAME))
(list (make-posn MAX-IMG-X 3))
'left
(list (make-posn 15 5))))

(check-expect
(process-tick (make-world
(list (make-ally INIT-ROCKET2 MY-NAME))
(list (make-posn MIN-IMG-X 2))
'down
(cons (make-posn 2 MIN-IMG-Y) '())))
152 The game-over? Refinement 617

(make-world (list (make-ally INIT-ROCKET2 MY-NAME))

(list (make-posn MIN-IMG-X 3))
'right
'()))
The local process-tick function must also be updated given that it produces
a world. This function does not process the given world. Therefore, its refinement
only needs to use the proper selector for the allies. The refined function is:
;; world → world
;; Purpose: Create a new world after a clock tick
;; ASSUMPTION: The given world is a structure
(define (process-tick a-world)
(make-world
(world-allies a-world)
(remove-hit-aliens (move-loa (world-aliens a-world)
(world-dir a-world))
(move-los (world-shots a-world)))
(new-dir-after-tick (move-loa (world-aliens a-world)
(world-dir a-world))
(world-dir a-world))
(remove-shots (move-los (world-shots a-world))
(move-loa (world-aliens a-world)
(world-dir a-world)))))
Once again the assumption that the given world is a structure is made explicit for
the benefit of any reader of the code.

152 The game-over? Refinement

The game-over? predicate must return #false, as before, when the given world’s
list of aliens is empty or when any of the aliens has reached earth. In addition, it
must return #false if the given world is uninitialized. This problem analysis allows
the refinement of the sample expressions to be:
(define GAME-OVER1 (and (not (eq? INIT-WORLD2 UNINIT-WORLD))
(or (ormap
(λ (an-alien)
(= (posn-y an-alien)
MAX-IMG-Y))
(world-aliens INIT-WORLD2))
(empty? (world-aliens INIT-WORLD2)))))
618 27 Aliens Attack Version 6

(define GAME-OVER2 (and (not (eq? WORLD3 UNINIT-WORLD))

(or (ormap
(λ (an-alien)
(= (posn-y an-alien)
MAX-IMG-Y))
(world-aliens WORLD3))
(empty? (world-aliens WORLD3)))))

(define GAME-NOT-OVER (and (not (eq? INIT-WORLD

UNINIT-WORLD))
(or (ormap
(λ (an-alien)
(= (posn-y an-alien)
MAX-IMG-Y))
(world-aliens INIT-WORLD))
(empty?
(world-aliens INIT-WORLD)))))

(define GAME-NOT-DONE (and (not (eq? UNINIT-WORLD

UNINIT-WORLD))
(or (ormap
(λ (an-alien)
(= (posn-y an-alien)
MAX-IMG-Y))
(world-aliens UNINIT-WORLD))
(empty?
(world-aliens UNINIT-WORLD)))))
Abstracting over the sample expressions yields the refined handler:
;; world → Boolean
;; Purpose: Detect if the game is over
(define (game-over? a-world)
(and (not (eq? a-world UNINIT-WORLD))
(or (ormap (λ (an-alien)
(= (posn-y an-alien) MAX-IMG-Y))
(world-aliens a-world))
(empty? (world-aliens a-world)))))
Once again, the tests using sample values must be updated to construct worlds
that have a lor. They now become:
;; Tests using sample values for game-over?
(check-expect (game-over?
(make-world (list (make-ally 8 MY-NAME))
(list (make-posn 0 3))
'right
NO-SHOT))
#false)
153 What Have We Learned in This Chapter? 619

(check-expect (game-over?
(make-world (list (make-ally 8 MY-NAME))
(list (make-posn 0 MAX-IMG-Y))
'right
(list (make-posn 12 11))))
#true)
(check-expect (game-over?
(make-world (list (make-ally 8 MY-NAME))
(list (make-posn 0 5))
'right
(list (make-posn 0 5))))
#false)
This completes the needed refinements. Make sure that all the tests pass and play
the game. This refinement is the one that is used to create different implementations
of a multiplayer game.

* Ex. 329 — Refactor draw-ally-maker to use a λ-expression instead of a

local-expression.

**** Ex. 330 — Change the data definition of the world to include a timer. The
timer is a natural number and is used to control how often the player can shoot.
The player may shoot when the timer is 0. When a player shoots the new world
constructed has a nonzero timer. At every clock tick the timer is decremented,
and the player must wait until the timer reaches 0 to shoot again.

153 What Have We Learned in This Chapter?

The important lessons of Chap. 27 are summarized as follows:

• Sometimes it is possible to refine data definitions and corresponding code before
introducing design elements that are specific to distributed programming.
• A change in data representation in a program does not change the interface with
the user.
• When refining a sequential program into a distributed program, it may help
simplify the task to first get the sequential program working with the expected
data representation for the distributed version of the program.
• During a refinement problem analysis must be performed for the functions that
process or return instances of a refined type (or, if you prefer, a refined data
definition).
620 27 Aliens Attack Version 6

• When a data definition is refined, sample expressions and tests using sample values
for functions that process or produce instances of the refined data definition must
also be refined.
• Data definition refinements may create the need for new data definitions and new
auxiliary functions.
Chapter 28
Aliens Attack Version 7

This chapter refines Aliens Attack 6 into a multiplayer game. Although the design
and implementation of a distributed multiplayer game may seem overwhelming
remember that the design recipe for distributed programming helps you manage the
complexity of its development. The design recipe divides the design into several
smaller and manageable steps. Collectively these steps lead to a distributed program
that is readable to others.
Each section outlines a step of the design recipe. Take time to understand how
the different steps are interrelated. Remember that the steps build on each other.
Therefore, it is important to understand a step before proceeding to the next step.

154 Components

In Aliens Attack 6 a player’s program draws the game, detects when the game is
over, and updates the game by processing key events, clock ticks, and messages from
the server. An intuitive design idea for a multiplayer game is to have each player be a
component that performs the same tasks. In addition, each player component sends
a message to a server when a player induced change occurs. A message is sent to
the server when the player shoots or moves the rocket. A message is received from
the server when it joins the game and when another player shoots, moves it rocket,
arrives, or departs.
The server receives messages from a player, manages the joining of new play-
ers, and manages the departure of players. When a message arrives from a player,
Playeri, the server broadcasts the message to the other players. When a player joins
the game, the server provides it with its starting world and broadcast a message to
all players that they have a new ally. The starting world depends on when Playeri
arrives. If Playeri is the first player to join, the server provides it with the initial
world. If Playeri is not the first player to join, then the server requests the world
from an existing player and sends it to Playeri. In order to properly select recipients
of messages, each player component must have a distinct name. The server rejects
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 621
M. T. Morazán, Animated Problem Solving, Texts in Computer Science,
https://doi.org/10.1007/978-3-030-85091-3_28
622 28 Aliens Attack Version 7

Fig. 124 The components of Aliens attack version 7

Server
Manage arrival of new players.
Manage the departure of new players.
Broadcast new shots and rocket moves.
Process messages from the players.

...

Player1 Playern
Draw the world. Draw the world.
Update the world. ... Update the world.
Detect end of the game. Detect end of the game.
Process messages from the server. Process messages from the server.

any new player that has a name already associated with an existing player. When a
player departs the game, the server sends a message to all the other players to remove
the corresponding ally.
Figure 124 visually summarizes the design idea with n players. There are n+1
components: each player and the server. The players and the server exchange mes-
sages to advance the game. Observe that the server is restricted to managing arriving
and departing players and to broadcasting changes made to the game by any player.
It does not perform any task to update the state of the game. That is, this design is
for a thin server. Finally, keep in mind that each player and the server is a separate
file. Therefore, a game with n players has n+1 programs running. Each component
is written in a separate file.

155 Data Definitions

In order to send messages to specific players, the server needs to track the players in
the game. Given that each player is represented as an iworld in our API the universe
is defined as follows:
;; A universe is a (listof iworld), where each iworld
;; has a unique name

;; Sample instances of universe

(define INIT-UNIV '())
(define A-UNIV (list iworld1 iworld2))
The sample universes illustrate each variety. The non-empty sample universe
uses a subset of the sample iworlds provided by the API. Observe that each iworld
in A-UNIV has a unique name. The comments and code above go in the server
program.
156 Communication Protocol 623

Fig. 125 Protocol diagram for a Playeri shooting

marshaled ne
w shot
Time

shot
ed new
marshal
..
.

shot
new
ha led
mars

Player1 ... Player ... Player Server

The data definitions for the world are defined in Sect. 148. The only detail that
we highlight is that each player must have a unique value for its MY-NAME constant.
This is the value used to give the world a name in the big-bang expression and is
used by the server to distinguish the iworlds it tracks.

156 Communication Protocol

The communication chains are sparked by game-changing events that must be com-
municated to one or more players. Think carefully about what events cause a change
in the game that must be communicated to the players. On a player’s side a com-
munication chain is sparked when the player shoots or the player moves her rocket
given that either a new shot or a new ally is created. A communication chain is not
sparked when the clock ticks. Although with every clock tick the state of the game
changes, such a change occurs automatically in all the players. Thus, there is nothing
to communicate. On the server side a communication chain is sparked when a player
joins or leaves the game given that the allies gain or lose an ally.

156.1 Player-Sparked Communication Chains

When a player shoots a new shot is created. This shot does not exist for the other
players. Therefore, the new shot must be communicated to the other players. Figure
125 displays the communication protocol diagram for the communication chain
generated by Playeri shooting. When Playeri shoots, the marshaled new shot is
sent to the server. Upon receiving the message, the server broadcasts it to the rest
of the players. Take note that the new shot is not communicated back to Playeri
given that the new shot already exists in Playeri’s world. This diagram informs
624 28 Aliens Attack Version 7

Fig. 126 Communication protocol diagram for a Playeri moving her rocket
marshaled ro
Time cket moved

et moved
ed rock
marshal
..
.

ved
t mo
d rocke
hale
mars

Player1 ... Player ... Player Server

us that a marshaled new shot message is both a to-server and a to-player message
given that such a message may be an incoming message for both the server and the
players.
When a player moves her rocket, a new ally is created containing the new rocket.
This new ally does not exist for the other players. Therefore, the new ally must be
communicated to the other players. Figure 126 displays the communication protocol
diagram for the communication chain generated by Playeri moving her rocket.
When Playeri moves the rocket, the marshaled new ally is sent to the server.
Upon receiving the message, the server broadcasts it to the rest of the players. This
diagram informs us that a marshaled rocket moved message is both a to-server and
a to-player message.

156.2 Server-Sparked Communication Chains

When a player leaves the game (e.g., they lose their internet connection or close
the game’s window), the other players lose an ally. The server must inform the
remaining players to remove from their world the ally that left the game. Fig. 127
displays the communication protocol diagram for the communication chain generated
by Playeri leaving the game. When Playeri leaves the server gets a message (that is
not generated by our distributed program) indicating that Playeri has disconnected
from the server. The server broadcast a message to remove an ally to all the
remaining players. This diagram informs us that a marshaled remove ally message
is a to-player message.
There are two cases that need to be distinguished when a player tries to join the
game. The first case is when the universe is empty. That is, the player is the first to join
the game. Figure 128 displays the protocol diagram for the sparked communication
chain. The server gets an API-generated message that a new component wants to
join. If the server accepts the new player, it sends a marshaled start world message
156 Communication Protocol 625

Fig. 127 Communication protocol diagram for a Playeri leaving the game
component di
Time sconnect

ally
remove
..
.

lly
ve a
remo

Player1 ... Player ... Player Server

Fig. 128 Communication protocol diagram for Playeri joining an empty universe
new componen
t
Time

art world
marshaled st

Player Server

to the player. This protocol diagram informs us the marshaled start world message
is a to-player message.
The second case is when a player joins a non-empty universe. Figure 129 displays
the communication protocol diagram for the sparked communication chain. The
server gets an API-generated new component message. If the player is admitted, the
server must send out several messages. It sends a marshaled new ally message to
all the players in the universe and a marshaled message to any player (in Fig. 129
it is Player1) requesting the value of the world. Observe that in Fig. 129 all these
messages originate at the same point. This is because all these messages are sent
out when the new player is admitted to the game. A player, upon receiving a world
request message, sends a marshaled world back message to the server. When the
server receives a world back message, it sends a start world message to the player
that just joined the game. This protocol diagram informs us that a marshaled world
back message is a to-server message. It further informs us that a marshaled add ally
message, a marshaled send world message, and a marshaled start world message are
to-player messages.
626 28 Aliens Attack Version 7

Fig. 129 Communication protocol diagram for Playeri joining a non-empty uni-
verse
new componen
t

marshaled add ally

..
.
Time

d ally
led ad
marsha
d
nd worl
ale d se
marsh
marshaled world back

marshaled start world

Player1 ... Player ... Player Server

156.3 Message Data Definitions

Armed with the knowledge of the message types that need to be exchanged, to-player
message and to-server message data definitions must be developed. According to
our communication protocol, there are six varieties of to-player messages. Think
about what each to-player message must contain. Given that an ally is unsuitable
for transmission, a rocket move message must contain a marshaled ally. Similarly, a
new ally message must contain a marshaled ally. A new shot message must contain
a marshaled shot given that a shot is unsuitable for transmission. A remove ally
message must contain the name of the ally that must be removed. A send world
message must contain the name of the (new) ally that the value of the world is
intended for. Finally, a start message must contain a marshaled world.

156.3.1 Marshaled Data Definition

The marshaled data is defined as follows:

;; A marshaled ally (mr) is a (list image-x string)
;; A marshaled alien (ma) is a (list image-x image-y)
;; A marshaled shot (ms) is either
;; 1. (list image-x image-y)
;; 2. 'no-shot
;; A marshaled world (mw) is a (list (listof mr)
;; (listof ma)
;; direction
;; (listof ms))
Observe that the marshaled varieties of data contain structure data suitable for
transmission. There is variety in ms because there is variety in shot. A mw itself
156 Communication Protocol 627

contains marshaled allies, marshaled aliens, and marshaled shots. Based on the data
definitions, we may develop function templates and sample instances of marshaled
data as follows:
#| ;; Sample instances of mr
(define MR1 . . .) . . .
;; Sample expressions for f-on-mr
(define MR1-VAL . . .) . . .
;; mr . . . → . . .
;; Purpose:
(define (f-on-mr an-mr . . .)
(local [(define rocket (first an-mr))
(define name (second an-mr))]
. . .))
;; Tests using sample computations for f-on-mr
(check-expect (f-on-mr MR1 . . .) MR1-VAL) . . .
;; Tests using sample values for f-on-mr
(check-expect (f-on-mr . . . . . .) . . .) . . .

;; Sample instances of ma
(define MA1 . . .) . . .
;; Sample expressions for f-on-ma
(define MA1-VAL . . .) . . .
;; ma . . . → . . .
;; Purpose:
(define (f-on-ma an-ma . . .)
(local [(define img-x (first an-ma))
(define img-y (second an-ma))]
. . .))
;; Tests using sample computations for f-on-ma
(check-expect (f-on-ma MA1 . . .) MA1-VAL) . . .
;; Tests using sample values for f-on-ma
(check-expect (f-on-ma . . . . . .) . . .) . . .

;; Sample instances of ms
(define MS1 . . .) . . .
;; Sample expressions for f-on-ms
(define MS1-VAL . . .) . . .
;; ms . . . → . . .
;; Purpose:
(define (f-on-ms an-ms . . .)
(if (list? ms)
(local [(define img-x (first an-ms))
(define img-y (second an-ms))]
. . .)
an-ms . . .))
628 28 Aliens Attack Version 7

;; Tests using sample computations for f-on-ms

(check-expect (f-on-ms MS1 . . .) MS1-VAL) . . .
;; Tests using sample values for f-on-ms
(check-expect (f-on-ms . . . . . .) . . .) . . .

;; Sample instances of mw
(define MW1 . . .) . . .
;; Sample expressions for f-on-mw
(define MW1-VAL . . .) . . .
;; mw . . . → . . .
;; Purpose:
(define (f-on-mw an-mw . . .)
(local [(define lomr (first an-ms))
(define loma (second an-ms))
(define dir (third an-ms))
(define loms (fourth an-ms))]
. . .))
;; Tests using sample computations for f-on-mw
(check-expect (f-on-mw MW1 . . .) MW1-VAL) . . .
;; Tests using sample values for f-on-mw
(check-expect (f-on-mw . . . . . .) . . .) . . .
|#
(define MR1 '(10 "iworld1"))
(define MR2 '(8 "iworld3"))
(define MR3 `(11 ,MY-NAME))
(define MALLY1 (list 10 "Rolando"))
(define MALLY2 (list 10 "Margarita"))
(define MA '(14 3))
(define MALIEN1 (list 0 7))
(define MALIEN2 (list 9 4))
(define MS1 NO-SHOT)
(define MS2 '(2 2))
(define MSHOT1 NO-SHOT)
(define MSHOT2 (list 11 9))
(define MLOS '((2 2)))
(define MW '(((7 "iworld1") (3 "iworld2"))
((15 2))
left
((8 5) (7 2))))
(define MWORLD1 (list (list (list 16 "Cristian")
(list 7 "Laura")
(list 6 "Walter"))
(list (list 13 4)
(list 11 11))
'right
156 Communication Protocol 629

(list (list 2 3)
(list 12 8)
(list 5 14))))
(define MWORLD2 (list (list (list 0 "Marce")
(list 8 "Chaty")
(list 4 "Maggie")
(list 2 "Christy"))
'()
'right
'()))
Observe that the templates above are more sophisticated than those developed in the
beginning parts of this textbook. They use a local-expression to define the different
components of a message instead of littering the function body with multiple uses of
list selectors that may carry no meaning for readers trying to understand the design.
The larger than usual sample instances are to facilitate test writing.

156.3.2 To-Player Message Data Definition

Given the data definitions for marshaled data, a to-player message is defined as
follows:
;; A to-player message (tpm) is either
;; 1. (list 'rckt-move ma)
;; 2. (list 'new-shot ms)
;; 3. (list 'new-ally ma)
;; 4. (list 'rm-ally string)
;; 5. (list 'send-world string)
;; 6. (cons 'start mw)
Each tpm has a unique tag to identify it and the appropriate data that is marshaled
if necessary. The template for a function on a tpm and sample tpm instances are
defined as follows:
#| TEMPLATE FOR A FUNCTION on a tpm
;; Sample instances of tpm
(define ST-MSG . . .)
(define MA-MSG . . .)
(define NS-MSG . . .)
(define NA-MSG . . .)
(define RM-MSG . . .)
(define SW-MSG . . .) . . .
;; Sample expressions for f-on-tpm
(define ST-MSG-VAL . . .)
(define MA-MSG-VAL . . .)
(define NS-MSG-VAL . . .)
630 28 Aliens Attack Version 7

(define NA-MSG-VAL . . .)
(define RM-MSG-VAL . . .)
(define SW-MSG-VAL . . .) . . .
;; tpm . . . → . . .
;; Purpose:
(define (f-on-tpm a-tpm . . .)
(local [(define tag (first a-tpm))]
(cond [(eq? tag 'rckt-move)
(local [(define ally (unmarshal-ally
(second a-tpm)))]
. . .)]
[(eq? tag 'new-shot)
(if (list? (second a-tpm))
(local [(define shot (unmarshal-shot
(second a-tpm)))]
(if (eq? shot NO-SHOT)
...
. . .)))]
[(eq? tag 'new-ally)
(local [(define ally (unmarshal-ally
(second a-tpm)))]
. . .)]
[(eq? tag 'rm-ally)
(local [(define name (second a-tpm))]
. . .)]
[(eq? tag 'send-world)
(local [(define name (second a-tpm))]
. . .)]
[(eq? tag 'start)
(local [(define world (unmarshal-world
(rest a-tpm)))
(define allies (world-allies world))
(define aliens (world-aliens world))
(define dir (world-dir world))
(define shots (world-shots world))]
. . .)]
[else
(error (format "Unknown message tag received: ~s"
(first msg)))])))
;; Tests using sample computations for f-on-tpm
(check-expect (f-on-tpm ST-MSG . . .) ST-MSG-VAL)
(check-expect (f-on-tpm MA-MSG . . .) MA-MSG-VAL)
(check-expect (f-on-tpm NS-MSG . . .) NS-MSG-VAL)
(check-expect (f-on-tpm NA-MSG . . .) NA-MSG-VAL)
(check-expect (f-on-tpm RM-MSG . . .) RM-MSG-VAL)
156 Communication Protocol 631

(check-expect (f-on-tpm SW-MSG . . .) SW-MSG-VAL) . . .

;; Tests using sample values for f-on-tpm
(check-expect (f-on-tpm . . . . . .) . . .) . . .
|#
(define RM-MSG (list 'rckt-move MR2))
(define RM-MSG2 (list 'rckt-move MR3))
(define NS-MSG (list 'new-shot MS1))
(define NS-MSG2 (list 'new-shot MS2))
(define NA-MSG (list 'new-ally MR2))
(define RA-MSG (list 'rm-ally "world1"))
(define RA-MSG2 (list 'rm-ally "Margarita"))
(define SW-MSG (list 'send-world "world2"))
(define SW-MSG2 (list 'send-world "Fernando"))
(define ST-MSG (cons 'start MW))
Observe that each stanza uses a local-expression to capture the components of
the given tpm. Furthermore, the template suggests unmarshalling the marshaled
components and defining local variables for their values. For instance, the ally
marshaled in a new ally message is unmarshaled and locally defined.

156.3.3 To-Server Message Data Definition

According to our communication protocol, there are three varieties of to-server

messages. Two of these varieties, a rocket move message and a new shot message,
are also tpms. Think about what the third variety, a world back message, needs to
contain. It must contain a marshaled world. It must also contain the destination of
the marshaled world. That is, the name of the player that is awaiting to start playing.
A to-server message is defined as follows:
;; A to-server message (tsm) is either
;; 1. (list 'rckt-move ma)
;; 2. (list 'new-shot ms)
;; 3. (cons 'world-back (cons string mw))
The template for a function on a tsm is:
#| TEMPLATE FOR FUNCTIONS ON A tsm
;; Sample instances of tsm
(define TSM1 . . .)
(define TSM2 . . .)
(define TSM3 . . .) . . .
;; Sample expressions for f-on-tsm
(define TSM1-VAL . . .)
(define TSM2-VAL . . .)
(define TSM3-VAL . . .) . . .
;; tsm . . . → . . .
632 28 Aliens Attack Version 7

;; Purpose:
(define (f-on-tsm a-tsm . . .)
(local [(define tag (first a-tsm))]
(cond [(eq? tag 'rckt-move) . . .]
[(eq? tag 'new-shot) . . .]
[(eq? tag 'world-back) . . .]
[else
(error
(format "Unknown to-server message type: ~s"
tag))])))
;; Sample expressions for f-on-tsm
(define TSM1-VAL . . .)
(define TSM2-VAL . . .)
(define TSM3-VAL . . .) . . .
;; Tests using sample computations for f-on-tsm
(check-expect (f-on-tsm TSM1 . . .) TSM1-VAL)
(check-expect (f-on-tsm TSM2 . . .) TSM2-VAL)
(check-expect (f-on-tsm TSM3 . . .) TSM3-VAL) . . .
;; Tests using sample values for f-on-tsm
(check-expect (f-on-tsm . . . . . .) . . .) . . .
|#
This template, unlike the template for functions on a tpm, does not suggest unmar-
shalling the components in a given tsm. This is because our design is for a thin
server that is expected to do little more than broadcast the messages it receives. It
has no need to unmarshal the components because it does not process them.
The sample instances for tsms are mostly straight-forward to write because they
contain the same types of marshaled values in them as tpms. We may write sample
values as follows:
;; Sample instances of tsm
(define SRM-MSG RM-MSG)
(define SNS-MSG NS-MSG)
(define SWB-MSG (cons 'world-back (cons "iworld2" MW)))

157 Marshalling and Unmarshalling

Marshalling and unmarshalling, respectively, make data unfit for transmission into
data that is fit for transmission and back. In Sect. 156.3.1 four varieties of marshaled
data are defined explicitly: marshaled ally (mr), marshaled alien (ma), marshaled shot
(ms), and marshaled world (mw). Using the generic data definition for (listof X)
a list of mr, a list of ma, and a list of ms are also defined as part of an mw. We shall
develop marshalling and unmarshalling functions for each of these types.
Consider marshalling and unmarshalling an ally. To marshal an ally, you need an
ally as input and you need to output an mr. The mr is constructed by accessing the
157 Marshalling and Unmarshalling 633

components of the given ally. The marshalling function is obtained by specializing

the template for functions on an ally as follows:
;; Sample expressions for marshal-ally
(define M-ALLY1 (list (ally-rocket INIT-ALLY)
(ally-name INIT-ALLY)))
(define M-ALLY2 (list (ally-rocket INIT-ALLY2)
(ally-name INIT-ALLY2)))

;; ally → mr
;; Purpose: Marshal the given ally
(define (marshal-ally an-ally)
(list (ally-rocket an-ally) (ally-name an-ally)))

;; Tests using sample computations for marshal-ally

(check-expect (marshal-ally INIT-ALLY) M-ALLY1)
(check-expect (marshal-ally INIT-ALLY2) M-ALLY2)

;; Tests using sample values for marshal-ally

(check-expect (marshal-ally (make-ally 12 "Cordula"))
(list 12 "Cordula"))
The unmarshalling function must take as input an mr and return an ally. The
function is obtained by specializing the template for functions on an mr as follows:
;; Sample expressions for unmarshal-ally
(define UALLY1 (make-ally (first MALLY1)
(second MALLY1)))
(define UALLY2 (make-ally (first MALLY2)
(second MALLY2)))

;; mr → ally
;; Purpose: Unmarshal the given marshaled ally
(define (unmarshal-ally ma)
(local [(define rocket (first ma))
(define name (second ma))]
(make-ally rocket name)))

;; Tests using sample computations for unmarshal-ally

(check-expect (unmarshal-ally MALLY1) UALLY1)
(check-expect (unmarshal-ally MALLY2) UALLY2)

;; Tests using sample values for unmarshal-ally

(check-expect (unmarshal-ally (list 12 "Cordula"))
(make-ally 12 "Cordula"))
Observe that the tests using sample values for both functions illustrate that they are
inverses of each other.
634 28 Aliens Attack Version 7

Consider marshalling and unmarshalling an alien. To marshal an alien you

need an alien as input and you need to output an ma. The ma is constructed by
accessing the components of the given alien. The marshalling function is obtained
by specializing the template for functions on an alien as follows:
;; Sample expressions for marshal-alien
(define MINIT-ALIEN2 (list (posn-x INIT-ALIEN2)
(posn-y INIT-ALIEN2)))

;; alien → ma
;; Purpose: Marshal the given alien
(define (marshal-alien an-alien)
(list (posn-x an-alien) (posn-y an-alien)))

;; Tests using sample computations for marshal-alien

(check-expect (marshal-alien INIT-ALIEN2) MINIT-ALIEN2)

;; Tests using sample values for unmarshal-alien

(check-expect (marshal-alien (make-posn 7 2)) (list 7 2))
The unmarshalling function must take as input an ma and return an alien. The
function is obtained by specializing the template for functions on an ma as follows:
;; Sample expressions for unmarshal-alien
(define UALIEN2 (make-posn (first MINIT-ALIEN2)
(second MINIT-ALIEN2)))

;; ma → alien
;; Purpose: Unmarshal the given marshaled alien
(define (unmarshal-alien ma)
(local [(define img-x (first ma))
(define img-y (second ma))]
(make-posn img-x img-y)))

;; Tests using sample computations for unmarshal-alien

(check-expect (unmarshal-alien MINIT-ALIEN2) UALIEN2)

;; Tests using sample values for unmarshal-alien

(check-expect (unmarshal-alien (list 7 2)) (make-posn 7 2))
Observe that the tests using sample values for both functions illustrate that they are
inverses of each other.
Consider marshalling and unmarshalling a shot. To marshal a shot you need
a shot as input and you need to output an ms. The ms is constructed by accessing
the components of the given shot if it is a structure. Otherwise, there is no need
to transform the shot because it is fit for transmission. The marshalling function is
obtained by specializing the template for functions on an shot as follows:
157 Marshalling and Unmarshalling 635

;; Sample expressions for marshal-shot

(define M-NO-SHOT NO-SHOT)
(define M-SHOT2 (list (posn-x SHOT2) (posn-y SHOT2)))

;; shot → ms
;; Purpose: Marshal the given shot
(define (marshal-shot a-shot)
(if (eq? a-shot NO-SHOT)
NO-SHOT
(list (posn-x a-shot) (posn-y a-shot))))

;; Tests using sample computations for marshal-shot

(check-expect (marshal-shot NO-SHOT) M-NO-SHOT)
(check-expect (marshal-shot SHOT2) M-SHOT2)

;; Tests using sample values for marshal-shot

(check-expect (marshal-shot (make-posn 2 2)) (list 2 2))
The unmarshalling function must take as input an ms and return an shot. The
function is obtained by specializing the template for functions on an ms as follows:
;; Sample expressions for unmarshal-shot
(define USHOT1 NO-SHOT)
(define USHOT2 (make-posn (first MSHOT2) (second MSHOT2)))

;; ms → shot
;; Purpose: Unmarshal the given marshaled shot
(define (unmarshal-shot ms)
(if (list? ms)
(local [(define img-x (first ms))
(define img-y (second ms))]
(make-posn img-x img-y))
ms))

;; Tests using sample computations for unmarshal-shot

(check-expect (unmarshal-shot NO-SHOT) USHOT1)
(check-expect (unmarshal-shot MSHOT2) USHOT2)

;; Tests using sample values for unmarshal-shot

(check-expect (unmarshal-shot (list 2 2)) (make-posn 2 2))
Observe that the tests using sample values for both functions illustrate that they are
inverses of each other.
Consider marshalling and unmarshalling a world. To marshal a world you need
a world as input and you need to output an mw. The mw is constructed by accessing
the components of the given world. What if the given world is uninitialized? How
can an uninitialized world be marshaled? Indeed, attempting to do so would be an
error. We can, however, assume that the given world is always structure because only
636 28 Aliens Attack Version 7

a player in the game may receive a message to send the world. If a player receives
such a message, then its world is initialized. The marshalling function is obtained by
specializing the template for functions on an world as follows:
;; Sample expressions for marshal-world
(define MWORLD3 (list (map marshal-ally
(world-allies INIT-WORLD))
(map marshal-alien
(world-aliens INIT-WORLD))
(world-dir INIT-WORLD)
(map marshal-shot
(world-shots INIT-WORLD))))

(define MWORLD4 (list (map marshal-ally

(world-allies INIT-WORLD2))
(map marshal-alien
(world-aliens INIT-WORLD2))
(world-dir INIT-WORLD2)
(map marshal-shot
(world-shots INIT-WORLD2))))

;; world → mw
;; Purpose: Marshal the given world
;; ASSUMPTION: The given world is a structure
(define (marshal-world a-world)
(list (map marshal-ally (world-allies a-world))
(map marshal-alien (world-aliens a-world))
(world-dir a-world)
(map marshal-shot (world-shots a-world))))

;; Tests using sample computations for marshal-world

(check-expect (marshal-world INIT-WORLD) MWORLD3)
(check-expect (marshal-world INIT-WORLD2) MWORLD4)

;; Tests using sample values for marshal-world

(check-expect
(marshal-world (make-world (list (make-ally 5 "Luis")
(make-ally 8 "Cova"))
(list (make-posn 2 17))
"left"
'()))
(list (list (list 5 "Luis")
(list 8 "Cova"))
(list (list 2 17))
"left"
'()))
157 Marshalling and Unmarshalling 637

Observe that the assumption is explicitly stated for any reader of the code, thus,
explaining why the definition lacks a conditional expression to distinguish among
world varieties. In addition, observe that the code, using map, to create the mar-
shaled lists of mr, ma, and ms is inlined instead of creating separate functions. The
unmarshalling function must take as input an mw and return a world. The function
is obtained by specializing the template for functions on an mw as follows:
;; Sample expressions for unmarshal-world
(define UWORLD1 (make-world
(map unmarshal-ally (first MWORLD1))
(map unmarshal-alien (second MWORLD1))
(third MWORLD1)
(map unmarshal-shot (fourth MWORLD1))))

;; mw → world
;; Purpose: Unmarshal the given world
(define (unmarshal-world mw)
(local [(define lomr (first mw))
(define loma (second mw))
(define dir (third mw))
(define loms (fourth mw))]
(make-world (map unmarshal-ally lomr)
(map unmarshal-alien loma)
dir
(map unmarshal-shot loms))))

;; Sample expressions for unmarshal-world

(define UWORLD1 (make-world
(map unmarshal-ally (first MWORLD1))
(map unmarshal-alien (second MWORLD1))
(third MWORLD1)
(map unmarshal-shot (fourth MWORLD1))))

(define UWORLD2 (make-world

(map unmarshal-ally (first MWORLD2))
(map unmarshal-alien (second MWORLD2))
(third MWORLD2)
(map unmarshal-shot (fourth MWORLD2))))

;; Tests using sample computations for unmarshal-world

(check-expect (unmarshal-world MWORLD1) UWORLD1)
(check-expect (unmarshal-world MWORLD2) UWORLD2)
638 28 Aliens Attack Version 7

;; Tests using sample values for unmarshal-world

(check-expect (unmarshal-world
(list (list (list 5 "Neil")
(list 6 "Constance")
(list 7 "Madrid")
(list 8 "Skyler"))
(list (list 0 4))
"down"
'()))
(make-world (list (make-ally 5 "Neil")
(make-ally 6 "Constance")
(make-ally 7 "Madrid")
(make-ally 8 "Skyler"))
(list (make-posn 0 4))
"down"
'()))
Once again, observe that the tests using sample values for both functions illustrate
that they are inverses of each other.

158 Component Implementation

Component implementation refers to the refinement of existing code to make it a

component of a distributed program or to the design of new components. For Aliens
Attack 7 there are two goals: the refinement of the player code and the design of the
server.

158.1 Player Component

The refinement of existing code to design a component involves the adoption of

new data definitions and the implementation of the communication protocol. For
Aliens Attack 7, all the new data definitions are associated with the communication
protocol. Therefore, all code refinement focuses on functions that send messages to
the server and on the function to process tpms.

158.1.1 The process-key Refinement

A function that must send messages to the server is process-key. This means that
some of the player’s outgoing arrows in the protocol diagrams in Figs. 125, 126, 127,
128, and 129 must be implemented by this handler. According to Fig. 125 a new
shot message must be sent when the player shoots. This means that when the player
158 Component Implementation 639

Fig. 130 Refined local process-key function

;; world key world or package
(define (process-key a-world a-key)
(cond [(key=? a-key "right")
(local
[(define nw (make-world (move-ally-right
MY-NAME (world-allies a-world))
(world-aliens a-world)
(world-dir a-world)
(world-shots a-world)))
(define na (get-ally MY-NAME (world-allies nw)))]
(make-package nw (list rckt-move (marshal-ally na))))]
[(key=? a-key "left")
(local
[(define nw (make-world (move-ally-left MY-NAME
(world-allies a-world))
(world-aliens a-world)
(world-dir a-world)
(world-shots a-world)))
(define na (get-ally MY-NAME (world-allies nw)))]
(make-package nw (list rckt-move (marshal-ally na))))]
[(key=? a-key " ")
(local [(define ns (process-shooting
(ally-rocket
(get-ally
MY-NAME (world-allies a-world)))))
(define nw (make-world (world-allies a-world)
(world-aliens a-world)
(world-dir a-world)
(cons ns (world-shots a-world))))]
(make-package nw (list new-shot (marshal-shot ns))))]
[else a-world]))

shoots, a new shot is constructed using the player’s rocket coordinate and a new
world is constructed by adding the new shot to the list of shots. These new values
are used to create a package. The message to the server is constructed by making a
list containing the 'new-shot tag and the marshaled new shot. These refinements
are implemented in the third stanza of the conditional for the local process-key
function displayed in Fig. 130.
According to Fig. 126, a rocket move message must be sent when the player moves
her rocket. This means that a package must be constructed with a new world that
has the updated ally for the player and a message that contains the 'rckt-move tag
and the marshaled updated ally. These refinements are implemented in the first two
stanzas of the conditional for the local process-key function displayed in Fig. 130.
This completes the refinements for process-key code. Observe, however, that
the function may return a package or a world. This means that process-key’s
signature must also be refined. In other words, we have done more than just refactor
code. The updated signature is also displayed in Fig. 130. It is also noteworthy that
the only outgoing arrow from a player that still needs to be implemented is the one
for a world back message in Fig. 129. All of the incoming arrows to a player, of
course, still need to be implemented by the function that processes tpms.
640 28 Aliens Attack Version 7

158.1.2 The process-message Design

The function to process tpms is designed by specializing the template for functions
on a tpm and the protocol diagrams in Figs. 125, 126, 127, 128, and 129. Develop
a sample expression for each subtype of tpm one at a time using the expressions in
the f-on-tpm definition template. Consider an incoming arrow to a player with an
ally move message in Fig. 126. What needs to be done? For a ally move message,
a local variable is defined to capture the ally embedded in the message and a new
world is constructed by substituting in the list of allies. This is done, for example, as
follows:
(define PM-RMOVE (local [(define ally (unmarshal-ally
(second RM-MSG2)))]
(make-world (replace-ally
ally
(world-allies INIT-WORLD))
(world-aliens INIT-WORLD)
(world-dir INIT-WORLD)
(world-shots INIT-WORLD))))
Note that an auxiliary function is needed to replace an ally in a list of allies.
Consider an incoming arrow to a player with a new shot message in Fig. 125. What
needs to be done? For a new shot message, a local variable is defined to capture the
shot embedded in the message and a new world is constructed by adding the new
shot to the front of the list of shots. This is done, for example, as follows:
(define PM-NSHOT (local [(define shot (unmarshal-shot
(second NS-MSG)))]
(if (eq? shot NO-SHOT)
INIT-WORLD
(make-world
(world-allies INIT-WORLD)
(world-aliens INIT-WORLD)
(world-dir INIT-WORLD)
(cons shot
(world-shots INIT-WORLD))))))

(define PM-NSHOT2 (local [(define shot (unmarshal-shot

(second NS-MSG2)))]
(if (eq? shot NO-SHOT)
INIT-WORLD
(make-world
(world-allies INIT-WORLD)
(world-aliens INIT-WORLD)
(world-dir INIT-WORLD)
(cons shot
(world-shots INIT-WORLD))))))
158 Component Implementation 641

There are two sample expressions because there are two subtypes of shot. Consider
an incoming arrow to a player with a new ally message in Figs. 128 and 129. What
needs to be done? For both cases, a local variable needs to be defined to capture the
ally embedded in the message and a new world is constructed by adding the new
ally to the front of the list of allies. This is done, for example, as follows:
(define PM-NALLY (local [(define ally (unmarshal-ally
(second NA-MSG)))]
(make-world
(cons ally (world-allies INIT-WORLD))
(world-aliens INIT-WORLD)
(world-dir INIT-WORLD)
(world-shots INIT-WORLD))))
Consider an incoming arrow to a player with a remove ally message in Fig. 127.
What needs to be done? For a remove ally message, a local variable is defined to
capture the ally-name embedded in the message and a new world is constructed by
removing the ally from the list of allies. This is done, for example, as follows:
(define PM-RMALLY (local [(define name (second RA-MSG2))]
(make-world (remove-ally
name
(world-allies INIT-WORLD2))
(world-aliens INIT-WORLD2)
(world-dir INIT-WORLD2)
(world-shots INIT-WORLD2))))
Observe that an auxiliary function to remove an ally for a (listof ally) is
needed. Consider an incoming arrow to a player with a send world message in
Fig. 129. What needs to be done? For a send world message, a local variable is
defined to capture the world-name embedded in the message and a package needs to
be constructed because the outgoing arrow with world back message in Fig. 129 must
be implemented. The package is constructed with the world and its marshalling. The
outgoing message must also contain the proper tag and the name of the destination
world. This is done, for example, as follows:
(define PM-SWORLD (local [(define name (second SW-MSG2))]
(make-package
WORLD3
(cons 'world-back
(cons name
(marshal-world WORLD3))))))
Finally, consider an incoming arrow to a player with a start message in Fig. 129.
What needs to be done? For a start message, a local variable is defined to capture
the world embedded in the message. Local variables are also defined to capture
the components of the newly arrived world instance. These components are used to
build a new world by adding the player’s rocket to the list of allies. This is done, for
example, as follows:
642 28 Aliens Attack Version 7

(define PM-START (local [(define world (unmarshal-world

(rest ST-MSG)))
(define allies (world-allies world))
(define aliens (world-aliens world))
(define dir (world-dir world))
(define shots (world-shots world))]
(make-world (cons INIT-ALLY allies)
aliens
dir
shots)))
Based on the sample expressions, the process-message function definition
displayed in Fig. 131 is written. Observe that with this function implementation all
the player’s incident incoming and outgoing arrows are implemented. The only arrow
incidences remained to implement are those for the server.
The final step to complete the design of process-message is to write unit
tests. Writing the tests is the same as any other function. The tests using sample
computations use sample instances and the values of the sample expressions. The
tests using sample computations need to test that the proper error is generated when
an unknown message type is provided and illustrate with explicit instances one or
more worlds expected after receiving a message. The tests may be implemented as
follows:
;; Tests using sample computations for process-message
(check-expect (process-message INIT-WORLD RM-MSG2) PM-RMOVE)
(check-expect (process-message INIT-WORLD NS-MSG) PM-NSHOT)
(check-expect (process-message INIT-WORLD NA-MSG) PM-NALLY)
(check-expect (process-message INIT-WORLD2 RA-MSG2) PM-RMALLY)
(check-expect (process-message WORLD3 SW-MSG2) PM-SWORLD)
(check-expect (process-message INIT-WORLD ST-MSG) PM-START)

;; Tests using sample computations for process-message

(check-expect (process-message (make-world
(list (make-ally 6 "Doris"))
(list (make-posn 4 9))
"right"
'())
(list 'send-world "Don Marco"))
(make-package
(make-world
(list (make-ally 6 "Doris"))
(list (make-posn 4 9))
"right"
'())
(cons
'world-back
158 Component Implementation 643

Fig. 131 The player’s process-message function definition

;; world tpm world or package Purpose: Process given to-player message
(define (process-message a-world a-tpm)
(local [(define tag (first a-tpm))]
(cond [(eq? tag rckt-move)
(local [(define ally (unmarshal-ally (second a-tpm)))]
(make-world (replace-ally ally (world-allies a-world))
(world-aliens a-world) (world-dir a-world)
(world-shots a-world)))]
[(eq? tag new-shot)
(local [(define shot (unmarshal-shot (second a-tpm)))]
(if (eq? shot NO-SHOT) a-world
(make-world (world-allies a-world) (world-aliens a-world)
(world-dir a-world)
(cons shot (world-shots a-world)))))]
[(eq? tag new-ally)
(local [(define ally (unmarshal-ally (second a-tpm)))]
(make-world (cons ally (world-allies a-world))
(world-aliens a-world) (world-dir a-world)
(world-shots a-world)))]
[(eq? tag rm-ally)
(local [(define name (second a-tpm))]
(make-world (remove-ally name (world-allies a-world))
(world-aliens a-world) (world-dir a-world)
(world-shots a-world)))]
[(eq? tag send-world)
(local [(define name (second a-tpm))]
(make-package a-world
(cons world-back
(cons name (marshal-world a-world)))))]
[(eq? tag start)
(local [(define world (unmarshal-world (rest a-tpm)))
(define allies (world-allies world))
(define aliens (world-aliens world))
(define dir (world-dir world))
(define shots (world-shots world))]
(make-world (cons INIT-ALLY allies) aliens dir shots))]
[else (error (format "Unknown message type received: ~s"
(first a-tpm)))])))

(cons
"Don Marco"
(list (list (list 6 "Doris"))
(list (list 4 9))
"right"
'())))))

(check-error (process-message INIT-WORLD

(list 'send-w "Magna"))
"Unknown message type received: send-w")
644 28 Aliens Attack Version 7

The auxiliary functions needed by process-message are straight-forward list-

processing exercises. They are presented here in the interest of thoroughness. The
function to remove an ally from an lor may be implemented using filter as
follows:
;; string lor → lor
;; Purpose: Remove given ally for given list of allies
(define (remove-ally a-name a-lor)
(filter (λ (a) (not (string=? (ally-name a) a-name)))
a-lor))

;; Sample expressions for remove-ally

(define RM-ALLIES1 (filter (λ (a)
(not (string=? (ally-name a)
"Quintana")))
INIT-ALLIES))
(define RM-ALLIES2 (filter (λ (a)
(not (string=? (ally-name a)
"Marco")))
ALLIES2))

;; Tests using sample computations for remove-ally

(check-expect (remove-ally "Quintana" INIT-ALLIES) RM-ALLIES1)
(check-expect (remove-ally "Marco" ALLIES2) RM-ALLIES2)

;; Tests using sample values for remove-ally

(check-expect (remove-ally "Driscoll"
(list (make-ally 4 "Manfred")
(make-ally 7 "Cordula")
(make-ally 1 "Catherina")))
(list (make-ally 4 "Manfred")
(make-ally 7 "Cordula")
(make-ally 1 "Catherina")))
(check-expect (remove-ally "Driscoll"
(list (make-ally 12 "Sakas")
(make-ally 14 "Davila")
(make-ally 11 "Driscoll")
(make-ally 17 "Morazan")))
(list (make-ally 12 "Sakas")
(make-ally 14 "Davila")
(make-ally 17 "Morazan")))
The function to replace an ally from an lor may be implemented using map as
follows:
158 Component Implementation 645

;; ally lor → lor

;; Purpose: Replace given ally in given lor
(define (replace-ally an-ally a-lor)
(map (λ (a) (if (string=? (ally-name a) (ally-name an-ally))
an-ally
a))
a-lor))

;; Sample expressions for replace-ally

(define RP-ALLIES1 (map (λ (a)
(if (string=? (ally-name a)
(ally-name INIT-ALLY2))
INIT-ALLY2
a))
INIT-ALLIES))
(define RP-ALLIES2 (map (λ (a)
(if (string=? (ally-name a)
(ally-name INIT-ALLY))
INIT-ALLY
a))
ALLIES2))

;; Tests using sample computations for replace-ally

(check-expect (replace-ally INIT-ALLY2 INIT-ALLIES) RP-ALLIES1)
(check-expect (replace-ally INIT-ALLY ALLIES2) RP-ALLIES2)

;; Tests using sample values for replace-ally

(check-expect (replace-ally (make-ally 5 "Manfred")
(list (make-ally 4 "Manfred")
(make-ally 7 "Cordula")
(make-ally 1 "Catherina")))
(list (make-ally 5 "Manfred")
(make-ally 7 "Cordula")
(make-ally 1 "Catherina")))
(check-expect (replace-ally (make-ally 10 "Mirek")
(list (make-ally 12 "Sakas")
(make-ally 14 "Davila")
(make-ally 11 "Driscoll")
(make-ally 17 "Morazan")))
(list (make-ally 12 "Sakas")
(make-ally 14 "Davila")
(make-ally 11 "Driscoll")
(make-ally 17 "Morazan")))
Finally, the updated run function for the player is:
646 28 Aliens Attack Version 7

;; string → world
;; Purpose: To run the game
(define (run a-name)
(local [(define TICK-RATE 1/4)]
(big-bang
INIT-WORLD
[on-draw draw-world]
[name MY-NAME]
[on-key process-key]
[on-tick process-tick TICK-RATE]
[stop-when game-over? draw-last-world]
[register LOCALHOST]
[on-receive process-message])))

** Ex. 331 — A conditional expression and local-expressions are used to

implement process-message. Simplify the implementation by refactoring it
to use pattern matching.

** Ex. 332 — The unmarshalling functions are implemented using local-

expressions. Simplify their implementation by refactoring them to use pattern
matching.

158.2 Server Component

Our initial problem analysis states that the server receives messages from the player,
manages the joining of new players, and manages the departure of players. This means
that three handlers are needed. In addition, the server must broadcast messages to
selected players. This means it must keep track of the players that are in the game.
Based on this, we may define the universe and the run-server function as follows:
;; A universe is a (listof iworld), where each iworld has
;; a unique name

;; Sample instances of universe

(define INIT-UNIV '())
(define A-UNIV (list iworld1 iworld2))

;; Z → universe
;; Purpose: Run the chat server
(define (run-server a-z)
(universe INIT-UNIV
(on-new add-player)
(on-msg process-message)
(on-disconnect rm-player)))
158 Component Implementation 647

The defined universe instances are used to test the handlers and observe that the
second sample universe is built using the sample iworld instances provided by
the API.
Each of the handlers is independently designed using the data definitions and
the communication protocol defined in the previous sections of this chapter. As
you proceed through the design, take note of the importance of understanding the
communication protocol to properly have the components cooperate.

158.2.1 The add-player Handler

This handler manages the joining of new players. If the joining player’s name is not
already used in the universe, it is allowed to join the game. Otherwise it is rejected.
When a player is allowed to join, as outlined by Figs. 128 and 129, there are two
cases. If it is the first player to join, the protocol diagram in Fig. 128 informs us that
initial world must be sent to it. If it is not the first player to join, the protocol diagram
in Fig. 129 informs us that the existing players must be sent a new ally message and
that the first world in the universe must be sent a send world message.
Develop at least one sample expression for each of the three possibilities. To reject
a player, a bundle may be created that does not change the given universe, that
creates no mails, and that disconnects the player that is attempting to join the game.
A sample expression for a player with a repeated name is:
(define RPT-ADD (make-bundle A-UNIV '() (list iworld1)))
Observe that A-UNIV is unchanged and there are no mails constructed. By not
constructing any mails, the elements of the communication protocol that must still
be implemented remain unchanged. By disconnecting the player from the server,
the player’s game window only displays a message stating that the universe has
disappeared.
To add a player to an empty universe, a bundle is created with a universe that
contains the new player, a start message with the 'start tag and the initial world
marshaled, and an empty list of players to disconnect. A sample expression for this
is:
(define EMP-ADD (make-bundle
(cons iworld1 INIT-UNIV)
(list (make-mail
iworld1
(cons 'start
(marshal-world INIT-WORLD))))
'()))
Observe that the constructed mail implements the arrow from the server to Playeri
in Fig. 128.
To add a player to a non-empty universe, a bundle is created with a universe
that contains the new player, a list of mails containing a send world message to the
first player in the universe and add ally messages to the existing players, and an
648 28 Aliens Attack Version 7

Fig. 132 The server’s add-world function definition

;; universe iworld bundle
;; Purpose: Add new world to the universe
(define (add-player a-univ an-iw)
(cond
[(member? (iworld-name an-iw) (map iworld-name a-univ))
(make-bundle a-univ () (list an-iw))]
[(empty? a-univ)
(make-bundle
(cons an-iw a-univ)
(list (make-mail an-iw (cons start (marshal-world INIT-WORLD))))
())]
[else
(make-bundle
(cons an-iw a-univ)
(cons (make-mail (first a-univ)
(list send-world (iworld-name an-iw)))
(map
( (iw)
(make-mail iw
(list new-ally
(list INIT-ROCKET (iworld-name an-iw)))))
a-univ))
())]))

empty list of players to disconnect. The list new ally messages may be constructed
using map to traverse the existing universe. A sample expression for this is:
(define NEW-ADD (make-bundle
(cons iworld3 A-UNIV)
(cons (make-mail
(first A-UNIV)
(list 'send-world
(iworld-name iworld3)))
(map
(λ (iw)
(make-mail
iw
(list 'new-ally
(list
INIT-ROCKET
(iworld-name iworld3)))))
A-UNIV))
'()))
Observe that the list of emails implements the add ally and send world messages
from the server in Fig. 129.
To implement the handler, the template for functions on a universe is specialized
by abstracting over the sample expressions. The result is displayed in Fig. 132. The
158 Component Implementation 649

signature clearly states that a bundle is returned. Each of the three possible cases
has a stanza in the conditional expression. Observe that the initial world, the initial
rocket, and marshal-world must be defined in the server’s program. You may
copy the necessary definitions into the server’s program.
The handler is tested as follows:
;; Tests using sample computations for add-player
(check-expect (add-player A-UNIV iworld1) RPT-ADD)
(check-expect (add-player INIT-UNIV iworld1) EMP-ADD)
(check-expect (add-player A-UNIV iworld3) NEW-ADD)

;; Tests using sample values for add-player

(check-expect
(add-player (list iworld2 iworld3) iworld1)
(make-bundle (list iworld1 iworld2 iworld3)
(cons (make-mail
(first (list iworld2 iworld3))
(list 'send-world "iworld1"))
(map
(λ (iw)
(make-mail
iw
(list 'new-ally
(list INIT-ROCKET "iworld1"))))
(list iworld2 iworld3)))
'()))
Observe that the sample iworlds provided by the API are extensively used in the
test using sample values.

158.2.2 The rm-player Handler

This handler removes a player that disconnects from the universe. To do so a new
universe is constructed by removing the player and, as indicated by the protocol
diagram in Fig. 127, by sending a remove ally message to the remaining players. The
following sample expression illustrates how this may be done:
(define IW1-RM (local
[(define new-univ
(filter (λ (iw)
(not (string=?
(iworld-name iw)
(iworld-name iworld1))))
A-UNIV))]
(make-bundle
new-univ
650 28 Aliens Attack Version 7

Fig. 133 The server’s rm-player function definition

;; universe iworld bundle
;; Purpose: Remove a player from the game
(define (rm-player a-univ an-iw)
(local
[(define new-univ (filter ( (iw)
(not (string=? (iworld-name iw)
(iworld-name an-iw))))
a-univ))]
(make-bundle
new-univ
(map ( (iw)
(make-mail iw
(list rm-ally (iworld-name an-iw))))
new-univ)
())))

(map (λ (iw)
(make-mail iw
(list 'rm-ally
(iworld-name iworld1))))
new-univ)
'())))
Observe that the list of mails implements the remove ally arrows in Fig. 127 and that
no players are explicitly removed from the server.
Based on the sample expression and Fig. 127, the handler’s implementation is
displayed in Fig. 133. The function returns a bundle because messages must be sent
out. The handler is tested as follows:
;; Tests using sample computations for rm-player
(check-expect (rm-player A-UNIV iworld1) IW1-RM)

;; Tests using sample values for rm-player

(check-expect (rm-player (list iworld1 iworld2 iworld3)
iworld2)
(make-bundle (list iworld1 iworld3)
(map
(λ (iw)
(make-mail
iw
(list 'rm-ally "iworld2")))
(list iworld1 iworld3))
'()))
158 Component Implementation 651

158.2.3 The process-message Handler

The process-message handler must process all incoming tsms. According to the
protocol diagrams in Figs. 125, 126, 127, 128, and 129, the server always sends out
one or more messages when it receives a tsm. This means that this handler must
return a bundle. To process a tsm, we may define local variables for the tag of the
message and for the list of players that a message must be sent to. When the message
is a world back message, the mailing list contains only the recipient player’s iworld.
Otherwise, it contains all the iworlds in the universe other than the player that
sent the message.
Sample expressions are displayed in Fig. 134. Observe that for each the tag and
the mailing list are captured in local variables. For rocket move and a new shot
messages, the server broadcasts them to all the players except the sender as defined
by the protocol diagrams in Figs. 126 and 125. For a world back message, the
marshaled world embedded in the message is sent to the recipient embedded in the
message as a start message to implement the outgoing start message arrow from the
server to Playeri in Fig. 129.
Observe that the local variables are defined the same way for all three tsm sub-
types. Therefore, instead of repeating the same code three times, the local variables
may be defined outside the conditional that distinguishes between the tsm subtypes
as displayed in Fig. 135. Each stanza in the conditional expression creates a bundle
and is obtained from abstracting away the concrete values for the universe, the
sending iworld, and the tsm in the sample expressions. Finally, the handler is tested
as follows:
;; Tests using sample computations for process-message
(check-expect (process-message A-UNIV iworld1 SRM-MSG) PM-RM)
(check-expect (process-message A-UNIV iworld1 SNS-MSG) PM-NS)
(check-expect (process-message A-UNIV iworld2 SWB-MSG) PM-WB)

;; Tests using sample values for process-message

(check-error (process-message A-UNIV
iworld2
'(rocket-move ((5 "iworld2"))))
(format "Unknown message received by server: ~s."
('rocket-move ((5 "iworld2")))))

** Ex. 333 — Improve the tests using sample values for process-message.

** Ex. 334 — Improve the process-message implementation using pattern

matching.

* Ex. 335 — Explain what happens if a player with a repeated name is not
disconnected from the server in add-player. Try it out by temporarily changing
the code.
652 28 Aliens Attack Version 7

Fig. 134 Sample expressions for the server’s process-message handler

(define PM-RM
(local [(define tag (first SRM-MSG))
(define send-list
(if (eq? tag world-back)
(filter
( (iw)
(string=? (iworld-name iw) (second SRM-MSG)))
A-UNIV)
(filter ( (iw)
(not (string=? (iworld-name iw)
(iworld-name iworld1))))
A-UNIV)))]
(make-bundle A-UNIV
(map ( (iw) (make-mail iw SRM-MSG))
send-list)
())))
(define PM-NS
(local [(define tag (first SNS-MSG))
(define send-list
(if (eq? tag world-back)
(filter ( (iw)
(string=? (iworld-name iw)
(second SNS-MSG)))
A-UNIV)
(filter ( (iw)
(not (string=? (iworld-name iw)
(iworld-name iworld1))))
A-UNIV)))]
(make-bundle A-UNIV
(map ( (iw) (make-mail iw SNS-MSG)) send-list)
())))
(define PM-WB
(local [(define tag (first SWB-MSG))
(define send-list
(if (eq? tag world-back)
(filter ( (iw)
(string=? (iworld-name iw) (second SWB-MSG)))
A-UNIV)
(filter ( (iw)
(not (string=? (iworld-name iw)
(iworld-name iworld1))))
A-UNIV)))]
(make-bundle A-UNIV
(map ( (iw)
(make-mail (first send-list)
(cons start (rest (rest SWB-MSG)))))
send-list)
())))
159 A Subtle Bug 653

Fig. 135 The servers process-message function definition

;; universe iworld tsm → bundle
;; Purpose: To process the given message from the given world
(define (process-message a-univ an-iw a-tsm)
(local
[(define tag (first a-tsm))
(define send-list (if (eq? tag world-back)
(filter ( (iw) (string=? (iworld-name iw)
(second a-tsm)))
a-univ)
(filter
( (iw) (not (string=? (iworld-name iw)
(iworld-name an-iw))))
a-univ)))]
(cond [(eq? tag rckt-move)
(make-bundle a-univ
(map ( (iw) (make-mail iw a-tsm)) send-list)
())]
[(eq? tag new-shot)
(make-bundle a-univ
(map ( (iw) (make-mail iw a-tsm)) send-list)
())]
[(eq? tag world-back)
(make-bundle a-univ
(list (make-mail (first send-list)
(cons start
(rest (rest a-tsm)))))
())]
[else
(error (format "Unknown message received by server: ~s."
a-tsm))])))

159 A Subtle Bug

Before proceeding make sure you have at least two copies of the player’s program
saved (each with a unique MY-NAME value), the players’ tests and the server tests
pass, and you play the game a few times. To play the game as multiple players on
your machine remember to first call run-server in the server file and then run for
each of the player files.
What do you notice, if anything, after playing the game a few times on your
computer? If you run it enough times, you are bound to see that there is a synchro-
nization bug. That is, the state of the game is different for different players. Figure
136 displays two snapshots of the same game at the same time for two different
players. Figure 136a displays the game that Playeri sees and Fig. 136b displays the
game that Playerj sees. Observe that the army of aliens is different. How is this
possible? All the changes made by one player are communicated to all other players.
Therefore, the game state ought to be the same for all players.
654 28 Aliens Attack Version 7

Fig. 136 Snapshot of two different players in the same game at the same time

(a) Snapshot of Playeri ’s Game. (b) Snapshot of Playerj ’s Game.

Fig. 137 New shot only exists in Playeri’s game

(a) Snapshot of Playeri ’s Game. (b) Snapshot of Playerj ’s Game.

The problem is due to each player updating its world independently of the other
players and to messages taking time to travel from a player to the server and then to
a receiving player. Let us take a closer look at the consequences of this. Consider
what happens when Playeri shoots. The new shot is added to Playeri’s world
and a new shot message is sent to the server. Playerj has not yet been informed of
the new shot by Playeri. The snapshots in Fig. 137, with new shot only existing
in Playeri’s world, capture the state of the game for each player.
Let us assume that messages are very fast, and they take only one clock tick to
travel from Playeri to the server and then to Playerj. During that clock tick, the
new shot has moved in Playeri’s world, but no movement has taken place by the
copy of the shot received by Playerj. The two worlds are no longer synchronized.
The snapshots in Fig. 138 capture the state of the game for each player.
After the following clock tick, the shot moves in both worlds. In Playeri’s world
the shot hits the alien, but in Playerj’s world the shot has not reached the alien
and, thus, does not hit it. The snapshots in Fig. 139 capture the state of the game for
each player. This means that the alien is removed from Playeri’s world and is not
removed from Playerj’s world. Thus, the army of aliens is different in these two
worlds.
159 A Subtle Bug 655

Fig. 138 New shot arrives at Playerj’s game

(a) Snapshot of Playeri ’s Game. (b) Snapshot of Playerj ’s Game.

Fig. 139 New shot hits Alien only in Playeri’s game

(a) Snapshot of Playeri ’s Game. (b) Snapshot of Playerj ’s Game.

Synchronization bugs in distributed programming are hard to pinpoint because

they do not always manifest themselves. Sometimes an application may run without
seeing the bug and other times the bug is seen. There is no test we can write to protect
ourselves from this potential bug. Unfortunately, it is not the only subtle distributed
programming bug that may occur. Another common bug is deadlock. Deadlock
occurs when two (or more) components are waiting for each other to perform an
action. For example, two components may be waiting for each other to signal that
they are ready to perform a task. Componenti is waiting for Componentj to signal
it is ready and Componentj is waiting for Componenti to signal that it is ready. The
two components are deadlocked waiting for a message that cannot possibly ever be
sent. You will learn more about detecting and avoiding such bugs in an operating
systems course or in a distributed programming course.
What can we do now to fix multiplayer Aliens Attack now? Given that the problem
stems from each player making changes to its own copy of the world, the solution
is to have only one component allowed to make changes to the world and all the
other components get the same copy of the world whenever it is changed. The
natural choice is to only allow the server to make changes to the world and in this
manner maintain all the players synchronized. That is, the solution is to implement
multiplayer Aliens Attack using a thick server.
656 28 Aliens Attack Version 7

Does this mean that thin servers are useless? No, it does not mean that. Thin
servers are perfectly fine when clients do not need to be synchronized or when
clients are automatically synchronized by the nature of the application. For example,
thin servers work well for turn-based games, like Tic Tac Toe or chess, where players
take turns to make changes to the game. The players are synchronized after each
change. Now you have a criteria for choosing the kind of server that you ought to
implement for a distributed program.

*** Ex. 336 — Design and implement a two-player Tic Tac Toe game.

**** Ex. 337 — Design and implement a two-player Concentration game.

***** Ex. 338 — Design and implement a multiplayer Snakes and Ladders
game.

160 What Have We Learned in This Chapter?

The important lessons of Chap. 28 are summarized as follows:

• Following the steps of the design recipe for distributed programming is an effective
way of managing the complexity of writing distributed programs.
• Component implementation refers to the refinement of existing code to make it
part of a distributed program or to the design of completely new components.
• The refinement of existing code to design a component involves the adoption of
new data definitions and the implementation of the communication protocol.
• All the arrows in a communication protocol diagram must be implemented by
code that sends a message from the source to the destination.
• Synchronization and deadlock are distributed programming bugs that do not
always manifest themselves with every program execution.
• Thin servers are a good choice when clients do not need to be synchronized or
when clients are automatically synchronized by the nature of the application.
Chapter 29
Aliens Attack Version 8

This chapter addresses the design of a multiplayer Aliens Attack using a think server.
In this version of the game only one component, namely the server, is allowed to
make changes to the state of the game. The players, therefore, must be carefully
designed to not make changes to their local copy of the world. Players still need to
have the ability to move their rocket and to shoot. When such actions are taken by a
player, however, the player’s program does not change the world. Instead, a message
is sent to the server and the server creates a new world and sends the new world to
the players. Given that the server is the only component allowed to update the world,
the players cannot become unsynchronized.
The sections of this chapter outline the steps of the design recipe for distributed
computing. As with the material in Chap. 28, take your time to understand how the
steps are interrelated and build on each other.

161 The Components

Figure 140 outlines the components for Aliens Attack 8. The server is responsible
for managing new players that connect to the server. The server does not start the
game until the first player connects. If the arriving player’s world has a name
different from all other worlds connected to the server, then it is allowed to join the
game. Otherwise, it is not allowed to join the game. The server is also responsible for
managing the departure of players. As with Aliens Attack 7, the departing world must
be removed from the universe and the corresponding ally must also be removed. In
addition, the server is responsible for updating the world. This may occur in one of
two ways: a clock tick or a message arrives from a player requesting their rocket
be moved or a shot be created. Whenever the server updates the game, it sends all
the players the new world.
A player component is responsible for drawing the world, processing keystrokes,
detecting the end of the game, and processing messages from the server. The world
is rendered and the end of the game is detected as is done in Aliens Attack 7.
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 657
M. T. Morazán, Animated Problem Solving, Texts in Computer Science,
https://doi.org/10.1007/978-3-030-85091-3_29
658 29 Aliens Attack Version 8

Fig. 140 The components of Aliens attack version 8

Server
Manage arrival of new players.
Manage the departure of new players.
Process messages from players.
Update the world after a clock tick.

...

Player1 Playern
Draw the world. Draw the world.
Process keystrokes. ... Process keystrokes.
Detect end of the game. Detect end of the game.
Process messages from server. Process messages from server.

Processing keystrokes is different. When the player moves her rocket or shoots, a
message is sent to the server requesting the desired change be made. Receiving a
message from the server means that the server has changed the value of the world
and the world embedded in the message becomes the new world.

162 Data Definitions

The data definitions for the world are the same as those defined in Sect. 148. That
is, a world is either 'uninitialized or is a structure that contains a list of allies
(lor), a list of aliens (loa), a direction, and a list of shots (los). Remember
that each player must have a constant representing a unique name. To facilitate test
development, we define the following world sample instance in addition to those
found in Sect. 148:
(define WORLD4 (make-world (list (make-ally 8 "iworld3")
(make-ally 5 "iworld2"))
(list (make-posn 8 2))
'right
(list (make-posn 8 4))))
In order to send messages to the all players, the server needs to track the players in
the game. As in Aliens Attack 7, this may be done using a list of iworlds. In addition
the server must maintain the state of the game. Therefore, the (current) world must
be part of the universe. Given that the universe is always a finite number of elements
(i.e., two), it may be defined using a structure. The universe, the template for
functions on a universe, and sample instances of universeare as follows:
163 Communication Protocol 659

;; A universe (univ) is a structure:

;; (make-univ (listof iworld) world)
(define-struct univ (iws game))

;; Template for a function on a univ

#| ;; Sample instances of univ
(define UNIV1 (make-univ . . . . . .))

univ . . . → . . .
Purpose:
(define (f-on-univ a-univ . . .)
(. . .(f-on-loiw (univ-iws a-univ)). . .
. . .(f-on-world (univ-world a-univ). . .)))

;; Sample expressions for f-on-univ

(define UNIV1-VAL . . .) . . .

;; Tests using sample computations for f-on-univ

(check-expect (f-on-univ UNIV1 . . .) UNIV1-VAL) . . .

;; Tests using sample values for f-on-univ

(check-expect (f-on-univ . . . . . .) . . .) . . . |#

;; Sample instances of universe

(define INIT-UNIV (make-univ '() UNINIT-WORLD))
(define OTHR-UNIV (make-univ (list iworld1 iworld2) WORLD3))
(define OTHR-UNIV2 (make-univ (list iworld3 iworld2) WORLD4))
Observe that the initial universe, INIT-UNIV, has an empty list of iworlds and an
uninitialized world. The list of iworlds is empty because no players have joined
when the server starts running. The world is uninitialized because the game ought
to start when the first player joins. Finally, the other two universes are defined using
our API’s sample iworlds to test the server’s handlers.

163 Communication Protocol

As in Aliens Attack 7, there are player- and server-sparked communication chains.

According to our problem analysis, player-sparked communication chains start with
keystroke events when a player moves her rocket or shoots. Server-sparked commu-
nication chains start when a new player joins or departs the game and when the state
of the game is updated on a clock tick.
660 29 Aliens Attack Version 8

Fig. 141 Protocol diagram for a Playeri shooting

new shot re
quest

marshaled world
..
Time

.
rld ..
led wo .
marsha

orl d
hal ed w
mars

Player1 ... Player ... Player Server

Fig. 142 Protocol diagram for a Playeri moving her rocket

move reques
t

marshaled world
..
Time

.
rld ..
led wo .
marsha

orl d
haled w
mars

Player1 ... Player ... Player Server

163.1 Player-Sparked Communication Chains

Consider what needs to happen when a player attempts to shoot. Recall that the
player may not change the state of the game. Unlike Aliens Attack version 7, a new
shot may not be added to the player’s los and a new shot message must be sent to
the server. Therefore, when a player shoots, a request to add a new shot to the world
is sent to the server. The server processes the request to create a new shot, updates
the world, and sends the new world to all the players. This communication chain
is captured in Fig. 141. Observe that the new world is sent to the player, Playeri,
that requested the new shot be created. In this manner, the shooting player receives
a new world containing the new shot.
Consider what needs to happen when a player moves her rocket. Once again,
a player cannot update the ally that contains her rocket and modify the world.
Instead, a move request is sent to the server. The server updates the world and sends
the new world to all the players. Figure 142 captures this communication chain. As
163 Communication Protocol 661

Fig. 143 Protocol diagram for a Playeri leaving the game

component di
sconnect

marshaled world
Time

..
.

orl d
haled w
mars

Player1 ... Player ... Player Server

with shooting, the server sends the new world to the player that requested her rocket
be moved.

163.2 Server-Sparked Communication Chains

Consider the communication chain that must be implemented when a player leaves
the game. The server updates the world. The new world must be sent to all the
players. Figure 143 captures this communication chain. Observe that when Playeri
leaves the game, the new world is sent to all the remaining players.
Consider the communication chain that is sparked when a new player is admitted
to the game. The server must update the world to contain the new ally. This
new world must be transmitted to all the players. This communication chain is
captured in Fig. 144. Observe that, unlike a player joining in Aliens Attack version
7 described using multiple protocol diagrams (Figs. 128 and 129) for two different
communication chains, there is a single protocol diagram for Aliens Attack 8. In
terms of communication, there is no need to distinguish between a player being the
first or not to join the game. Although the server may need to update the world
differently in these two cases, the communication chain is the same. Regardless of
how the server updates the world, only a marshaled world is sent to all the players.
There is no need to request the value of the world from an existing player as depicted
in Fig. 129.
The final server-sparked communication chain starts on a clock tick. The server
updates the world and the new world is sent to all the players. Figure 145 displays
the protocol diagram for a clock tick at the server. Observe that this communication
chain is sparked without any player involvement. That is, it is sparked by a server
event.
662 29 Aliens Attack Version 8

Fig. 144 Protocol diagram for a Playeri joining the game

new componen
t

marshaled world
..
Time

.
rld ..
led wo .
marsha

orl d
hal ed w
mars

Player1 ... Player ... Player Server

Fig. 145 Protocol diagram for a server clock tick

marshaled world
..
Time

.
rld ..
led wo .
marsha

d
worl
ha led
mars

Player1 ... Player ... Player Server

163.3 Message Data Definitions

According to the protocol diagrams in Figs. 141, 142, 143, 144 and 145, there is only
one type of to-player message. Such a message contains a marshaled world. Given
that the world data definition has not been changed, we use the same data definition
for a marshaled world (mw) found in Sect. 156.3.1. We define a to-player message as
follows:
#| ;; A to-player message (tpm) is: (cons 'world mw)

tpm . . . → . . .
Purpose:
(define (f-on-tpm a-tpm . . .)
(. . .(f-on-mw (rest a-tpm) . . .). . .))

Sample instances of tpm

163 Communication Protocol 663

(define A-TPM (cons 'world . . .))

Sample expressions for f-on-tpm

(define A-TPM-VAL (f-on-mw (rest A-TPM) . . .))

Tests using sample for computations for f-on-mw

(check-expect (f-on-tpm A-TPM . . .) A-TPM-VAL) . . .

Tests using sample for values for f-on-mw

(check-expect (f-on-tpm . . . . . .) . . .) . . . |#

;; Sample instances of tpm

(define TPM1 (cons 'world
(list (list (list 10 "San Martin"))
(list (list 5 8))
'right
(list (list 12 4)))))

(define TPM2 (cons 'world

(list (list (list 0 "Juarez"))
(list (list 5 3))
'right
(list (list 7 7)))))
Observe that the tag is 'world. Technically, for this version of the game it is
unnecessary to have a tag, given that there is only one variety of tpm. However, it
is good practice to be consistent when designing. By convention, every message is
tagged. Such consistency makes it easier for others to understand the design of the
code and is likely to make future refinements easier to implement.
The protocol diagrams in Figs. 141 and 142 inform us that there are two types of
to-server messages. The first subtype is a new shot request. Think about what needs
to be contained in such a message. The server knows what ally shot, given that
it has access to the iworld that sent the message. Thus, this subtype of to-server
message needs to contain nothing more than a tag. The second subtype is a move
request. What content, if any, must the message contain? When a player moves her
rocket, there are two options: right or left. Therefore, in addition to a tag this subtype
must contain information, say the key pressed, to distinguish the direction. Based on
this analysis, we may define a to-server message as follows:
#| A to-server message (tsm) is either:
1. (list 'move key)
2. (list 'shoot)

;; tsm . . . → . . .
;; Purpose:
(define (f-on-tsm a-tsm . . .)
664 29 Aliens Attack Version 8

(local [(define tag (first a-tsm))]

(cond [(eq? tag 'move) . . .]
[(eq? tag 'shoot) . . .]
[else
(error
(format "Unknown to-server message type ~s"
a-tsm))])))

;; Sample instances of tsm

(define MV-TSM . . .)
(define SH-TSM . . .) . . .

;; Sample expressions for f-on-tsm

(define MV-TSM-VAL . . .)
(define SH-TSM-VAL . . .)

;; Tests using sample computations for f-on-tsm

(check-expect (f-on-tsm MV-TSM . . .) MV-TSM-VAL)
(check-expect (f-on-tsm SH-TSM . . .) SH-TSM-VAL) . . .

;; Tests using sample values for f-on-tsm

(check-expect (f-on-tsm . . . . . .) . . .) . . . |#

;; Sample instances of tsm

(define MV-LEFT (list 'move "left"))
(define MV-RGHT (list 'move "right"))
(define SHOOT (list 'shoot))

164 Marshalling and Unmarshalling

Figures 141 142, 143, 144, and 145 inform us that the server needs to marshal a world
and the players need to unmarshal a world. Given that the world data definition is
unchanged from Aliens Attack version 7, the marshal-world, unmarshal-world,
and auxiliary marshalling functions from Sect. 157 may be used in this version of
the game. There is no need to design these functions again.
Observe that tsms only contain data that is suitable for transmission. Therefore,
there is no need to implement marshalling and unmarshalling functions for these
messages.
165 Component Implementation 665

165 Component Implementation

As with Aliens Attack version 7, one of the goals is to reuse as much of the code
as possible from Aliens Attack 6. Based on our problem analysis from Sect. 161, we
can outline how to distribute the handlers from Aliens Attach version 6 among the
player and server components of this version of the game. The player components
need the following, possibly refined, handlers from Aliens Attack version 6:
1. draw-world
2. game-over? (and draw-last-world)
3. process-key
In addition, the player component needs a handler to process tpms. We can, therefore,
define the run function as follows:
;; Z → world
;; Purpose: To run the game
(define (run a-z)
(big-bang
UNINIT-WORLD
[on-draw draw-world]
[on-key process-key]
[stop-when game-over? draw-last-world]
[on-receive process-message]
[name MY-NAME]
[register LOCALHOST]))
Observe that there is not stanza for tick processing. This follows from observing that
it is the server that must update the world when the clock ticks.
The server component needs the following, possibly refined, handlers from Aliens
Attack version 6:
1. process-tick
2. process-key
In addition, the server component needs handlers to process tsms, to add players, and
to remove players. We can, therefore, define the run-server function as follows:
;; Z → univ
;; Purpose: To run the server
(define (run-server a-z)
(local [(define TICK-RATE 1/4)]
(universe
INIT-UNIV
(on-tick process-tick TICK-RATE)
(on-msg process-message)
(on-new add-player)
(on-disconnect rm-player))))
666 29 Aliens Attack Version 8

Observe that process-key is not a handler. This is because key events do not
directly affect the actions of the server. When a tsm is received, the server uses (a
refined version of) process-key to update the game.

165.1 Player Component

This subsection outlines the players’ handlers for Aliens Attack 8. As stated above,
a goal is to reuse existing Aliens Attack version 6 code as possible. We expect
a significant amount of code to be reusable because the world data definition is
unchanged.

165.1.1 The draw-world and game-over? Handlers

Neither of these handlers change the game in Aliens attack version 6. This means that
there is no need for them to be refined to send messages to the server. Furthermore,
given that the world data definition is unchanged, there are no refinements required
at all. The same holds true for the auxiliary function draw-last-world.

165.1.2 The process-key Handler

In this version of the game, the process-key handler may not update the game.
Therefore, a refinement is needed. Think carefully about what this handler must
do. As in Aliens Attack version 6, if the world in uninitialized then nothing needs
to be done and the existing world is returned. Otherwise, the key pressed must be
processed. If the key pressed is "left" or "right", then according to the protocol
diagram displayed in Fig. 142 a move request must be sent to the server. Similarly, if
the key pressed is " ", then according to the protocol diagram displayed in Fig. 141
a new shot request must be sent to the server. In both cases, a package is constructed
using the current world and the appropriate tsm.
Given that process-key is an encapsulated function, we shall explore its design
by writing tests using sample values. When the player presses the right or left arrow,
a package is created with the given world and a 'move message containing the key
pressed. Tests for the "right" and "left" key events may be written as follows:
;; Tests using sample values for process-key
(check-expect (process-key (make-world
(list (make-ally 10 MY-NAME))
(list (make-posn 7 2))
'right
'())
"right")
165 Component Implementation 667

(make-package (make-world
(list (make-ally 10 MY-NAME))
(list (make-posn 7 2))
'right
'())
(list 'move "right")))

(check-expect (process-key (make-world

(list (make-ally 10 MY-NAME))
(list (make-posn 7 2))
'right
'())
"left")
(make-package (make-world
(list (make-ally 10 MY-NAME))
(list (make-posn 7 2))
'right
'())
(list 'move "left")))
Observe that the world given to process-key is the same as the world used to
construct the package. Thus, process-key handler does not change the world. Also
observe that the tsm used to construct the package contains the key pressed by the
player.
When the player presses the space key, a package is created with the given world
and a 'shoot message. The following test illustrates how the package is created:
(check-expect (process-key (make-world
(list (make-ally 10 MY-NAME))
(list (make-posn 7 2))
'left
'())
" ")
(make-package (make-world
(list (make-ally 10 MY-NAME))
(list (make-posn 7 2))
'left
'())
(list 'shoot)))
Once again, observe that the world given to process-key remains unchanged. As
per the tsm data definition, the shoot request message contains nothing else other
than the 'shoot tag.
If any other key is pressed by the player, the world remains unchanged and no
message is sent to the server. The following test illustrates that the given world is
returned unchanged:
668 29 Aliens Attack Version 8

(check-expect (process-key (make-world

(list (make-ally 10 MY-NAME))
(list (make-posn 7 2))
'right
'())
"d")
(make-world (list (make-ally 10 MY-NAME))
(list (make-posn 7 2))
'right
'()))
Based on the design idea and the tests above, the player’s process-key handler
is defined as follows:
;; world key → world or package
;; Purpose: Process a key event to return next world
(define (process-key a-world a-key)
(local
[;; world key → world
;; Purpose: Process a key event to return next world
;; ASSUMPTION: The given world is a structure
(define (process-key a-world a-key)
(cond [(or (key=? a-key "right")
(key=? a-key "left"))
(make-package a-world (list 'move a-key))]
[(key=? a-key " ")
(make-package a-world (list 'shoot))]
[else a-world]))]
(if (eq? a-world UNINIT-WORLD)
a-world
(process-key a-world a-key))))
Observe that as in Aliens Attack version 6, the local process-key is only called if
the world is initialized. The local process-key may return a world or a package
as exemplified by the tests above. In addition, its body distinguishes between the
keys that must be processed. It does not change the given world and, if necessary,
creates a package.

165.1.3 The process-message Handler

This handler needs to process the single variety of tpm that has a marshaled world
embedded in it. To do so, the embedded marshaled world needs to be unmarshaled.
The following sample expressions illustrate how this is done:
;; Sample expressions for process-message
(define PM-TPM1 (unmarshal-world (rest TPM1)))
(define PM-TPM2 (unmarshal-world (rest TPM2)))
165 Component Implementation 669

Observe that the world required as input by the API and the message’s tag are not
used. Abstracting over the sample expressions yields the following function:
;; world tpm → world
;; Purpose: Update the world with the given tpm
(define (process-message a-world a-tpm)
(unmarshal-world (rest a-tpm)))
The tests using sample computations are written using sample worlds, sam-
ple tpms, and the value of the sample expressions. Tests using sample values are
written using a sample world, and an explicit world message to illustrate that the
unmarshaled world obtained from the embedded marshaled world becomes the
new world value. The tests may be written as follows:
;; Tests using sample computations for process-message
(check-expect (process-message INIT-WORLD TPM1) PM-TPM1)
(check-expect (process-message INIT-WORLD2 TPM2) PM-TPM2)

;; Tests using sample values for process-message

(check-expect (process-message WORLD3
(list
'world
(list (list 9 "Bolivar"))
(list (list 7 2))
'left
'()))
(make-world (list (make-ally 9 "Bolivar"))
(list (make-posn 7 2))
'left
'()))
This completes the design of the player component. Make sure that the player’s
file has all the constants needed by the handlers and the tests. Do not forget to run
the tests and verify that they pass.

165.2 Server Component

This subsection outlines the server’s handlers for Aliens Attack 8. As with the player
component, we shall reuse as much code as possible from Aliens Attack version 6.
This reuse mostly refers to the tick handler and the auxiliary function for the message
handler to process player requests.
670 29 Aliens Attack Version 8

165.2.1 The process-tick Handler

According to the API the tick handler must take as input a universe and return
a universe. This signature is different from the process-tick function from
Aliens Attack version 6 and, therefore, a new universe-processing function is needed.
Think carefully how a universe ought to be processed. If the given universe is
INIT-UNIV, it means that a player has not joined and the universe ought to remain
unchanged. Otherwise, the universe ought to be updated. The list of iworlds remains
unchanged. The game, on the other hand, needs to be updated. This may be achieved
by calling process-tick from Aliens Attack version 6 with the universe’s game
value. This means that the process-tick and all its local auxiliary functions from
Aliens Attack version 6 need to be local to the server’s process-tick. In addition,
as indicated by the protocol diagram in Fig. 145, the new world value must be sent
to all the players.
The handler is outlined as follows:
;; univ → bundle
;; Purpose: Create a new universe after a clock tick
(define (process-tick a-univ)
.
(local [..
;; world → world
;; Purpose: Create new world after a clock tick
;; ASSUMPTION: The world is a structure
(define (process-tick a-world) . . . )]
(if (equal? a-univ INIT-UNIV)
(make-bundle a-univ '() '())
(local [(define new-game
(process-tick
(univ-game a-univ)))]
(make-bundle
(make-univ (univ-iws a-univ) new-game)
(map (λ (iw)
(make-mail
iw
(cons 'world
(marshal-world new-game))))
(univ-iws a-univ))
'())))))
In this design the function always returns a bundle even when no messages are
sent to the players. The local-expression encapsulates process-tick and all its
auxiliary functions from Aliens Attack version 6. The body of the local-expression
distinguishes between a universe that has and that does not have at least one player.
Observe that in both cases the list of iworlds is unchanged and no iworlds are
disconnected from the universe. No mails are constructed when the game has not
165 Component Implementation 671

started. When the game has already started, all players get the new world value that
is locally defined (i.e., new-game).
The handler is tested in a similar fashion to testing done in previous versions of
the game. That is, the tests illustrate that changes to the aliens, the direction,
and the shots for the given universe’s world instance are correctly made and that
the list of iworlds remains unchanged. Tests may be written as follows:
;; Tests using sample values for process-tick
(check-expect
(process-tick
(make-univ (list iworld1 iworld3)
(make-world (list (make-ally 9 "iworld1")
(make-ally 2 "iworld3"))
(list (make-posn 2 5))
'left
(list (make-posn 3 6) NO-SHOT))))
(make-bundle
(make-univ (list iworld1 iworld3)
(make-world (list (make-ally 9 "iworld1")
(make-ally 2 "iworld3"))
(list (make-posn 1 5))
'left
(list (make-posn 3 5))))
(list (make-mail iworld1
(list 'world
(list (list 9 "iworld1")
(list 2 "iworld3"))
(list (list 1 5))
'left
(list (list 3 5))))
(make-mail iworld3
(list 'world
(list (list 9 "iworld1")
(list 2 "iworld3"))
(list (list 1 5))
'left
(list (list 3 5)))))
'()))

(check-expect
(process-tick
(make-univ (list iworld3)
(make-world (list (make-ally 6 "iworld3"))
(list (make-posn
(- MAX-CHARS-HORIZONTAL 2)
672 29 Aliens Attack Version 8

10))
'right
(list SHOT2))))
(make-bundle
(make-univ
(list iworld3)
(make-world (list (make-ally 6 "iworld3"))
(list (make-posn
(sub1 MAX-CHARS-HORIZONTAL)
10))
'down
(list (make-posn (posn-x SHOT2)
(sub1 (posn-y SHOT2))))))
(list (make-mail
iworld3
(list 'world
(list (list 6 "iworld3"))
(list (list (sub1 MAX-CHARS-HORIZONTAL) 10))
'down
(list (list (posn-x SHOT2)
(sub1 (posn-y SHOT2)))))))
'()))
(check-expect (process-tick
(make-univ
(list iworld2)
(make-world (list (make-ally 14 "iworld2"))
(list (make-posn MAX-IMG-X 2))
'down
(list (make-posn 15 6)))))
(make-bundle
(make-univ
(list iworld2)
(make-world
(list (make-ally 14 "iworld2"))
(list (make-posn MAX-IMG-X 3))
'left
(list (make-posn 15 5))))
(list (make-mail
iworld2
(list 'world
(list (list 14 "iworld2"))
(list (list MAX-IMG-X 3))
'left
(list (list 15 5)))))
'()))
165 Component Implementation 673

(check-expect (process-tick
(make-univ
(list iworld1)
(make-world (list (make-ally 3 "iworld1"))
(list (make-posn MIN-IMG-X 2))
'down
(list (make-posn 2 MIN-IMG-Y)))))
(make-bundle
(make-univ
(list iworld1)
(make-world
(list (make-ally 3 "iworld1"))
(list (make-posn MIN-IMG-X 3))
'right
'()))
(list (make-mail
iworld1
(list
'world
(list (list 3 "iworld1"))
(list (list MIN-IMG-X 3))
'right
'())))
'()))

165.2.2 The process-message Handler

This handler is written by specializing the template for functions on a tsm. If the
given tsm’s tag is shoot or move, then the universe’s game needs to be updated.
In both cases this may be accomplished by refining process-key from Aliens
Attack version 6. Think carefully about why a refinement is needed. The code for
process-key always moves the ally or creates a shot for a single player. In this
version of the game the server needs to do so for an arbitrary player. If process-key
is made local to process-message,then the name of the iworld making the request
and the universe’s world value are in scope and may be used to correctly move or
shoot. This means that process-key only needs the key to process as input. The
key is used as expected to move an ally or to create a shot for an ally. The value
returned by process-key, therefore, must be a bundle that has a univ with the new
world value and the mails containing this new value for the players.
Figure 146 outlines the server’s process-message handler. The signature clearly
states that the handler may throw an error. This occurs when a message with an unrec-
ognized tag is received. If the message’s tag is 'shoot or 'move, then process-key
674 29 Aliens Attack Version 8

Fig. 146 The server’s process-message handler

;; univ iworld tsm bundle throws error
;; Purpose: Process the message to create new universe
;; ASSUMPTION: The given univ is not INIT-UNIV
(define (process-message a-univ an-iw a-tsm)
(local [(define tag (first a-tsm)) (define name (iworld-name an-iw))
(define game (univ-game a-univ))
;; key bundle Purpose: Create a bundle for player request
(define (process-key a-key)
..
(local [ .
(define nw
(make-world
(cond [(key=? a-key "right")
(move-ally-right name (world-allies game))]
[(key=? a-key "left")
(move-ally-left name (world-allies game))]
[else (world-allies game)])
(world-aliens game) (world-dir game)
(if (key=? a-key " ")
(cons (process-shooting
(ally-rocket (get-ally name
(world-allies game))))
(world-shots game))
(world-shots game))))]
(make-bundle
(make-univ (univ-iws a-univ) nw)
(map
( (iw)
(make-mail iw (cons world (marshal-world nw))))
(univ-iws a-univ))
())))]
(if (or (eq? tag shoot) (eq? tag move))
(process-key (if (eq? tag shoot) " "
(second a-tsm)))
(error (format "Unknown to-server message type ~s" a-tsm)))))

is called with the key that needs to be processed. When the tag is 'move, the needed
key is extracted from the given tsm. The function process-key locally defines all
.
the functions that were locally defined for Aliens Attack 6 (indicated by the ..). In ad-
dition, it locally defines nw for the new world created. This new world is created by
processing the given key. The list of allies in this new world is computed by determin-
ing the given key’s variety. Observe that move-ally-right and move-ally-left
are called with name instead of MY-NAME as done in Aliens Attack version 6. This
is how moving an arbitrary ally is achieved. The aliens and direction in the
new world remain unchanged. The list of shots for the new world is obtained by
determining if the given key is " ". If not, the shots remain unchanged. If so, a
new shot is added to the game. Observe that get-ally is called with name instead
of MY-NAME as done in Aliens Attack version 6. This is how creating a shot for
an arbitrary ally is achieved. Finally, the body of the local-expression returns a
165 Component Implementation 675

bundle. The new universe is constructed using the existing list of iworlds and
the new world. The list of mails implements all the arrows from the server to all
the players in Figs. 141 and 142. No iworlds are disconnected from the server when
a player’s request is processed.
Testing the handler requires at least one test for the error thrown and one test for
each tsm subtype. Given that functions are encapsulated, the only viable testing is
using sample values. Testing the error generated is done by providing a tsm with an
invalid tag:
(check-error
(process-message OTHR-UNIV iworld2 (list 'move-left 'left))
(format "Unknown to-server message type: ~s"
(list 'move-left 'left)))
Testing the processing of a move message is done with two different tsms: one for
each direction. The tests are written using a sample universe and sample tsms.
The expected bundles illustrate that the list of iworlds remains unchanged and that
the proper ally is moved in the proper direction. Sample tests are as follows:
(check-expect (process-message OTHR-UNIV iworld2 MV-LEFT)
(make-bundle
(make-univ (list iworld1 iworld2)
(make-world
(list (make-ally 7 "iworld1")
(make-ally 8 "iworld2"))
(list (make-posn 3 3))
'right
(list (make-posn 1 2))))
(list
(make-mail iworld1
(list 'world
(list (list 7 "iworld1")
(list 8 "iworld2"))
(list (list 3 3))
'right
(list (list 1 2))))
(make-mail iworld2
(list 'world
(list (list 7 "iworld1")
(list 8 "iworld2"))
(list (list 3 3))
'right
(list (list 1 2)))))
'()))
676 29 Aliens Attack Version 8

(check-expect (process-message OTHR-UNIV iworld1 MV-RGHT)

(make-bundle
(make-univ (list iworld1 iworld2)
(make-world
(list (make-ally 8 "iworld1")
(make-ally 9 "iworld2"))
(list (make-posn 3 3))
'right
(list (make-posn 1 2))))
(list
(make-mail iworld1
(list 'world
(list (list 8 "iworld1")
(list 9 "iworld2"))
(list (list 3 3))
'right
(list (list 1 2))))
(make-mail iworld2
(list 'world
(list (list 8 "iworld1")
(list 9 "iworld2"))
(list (list 3 3))
'right
(list (list 1 2)))))
'()))
Testing the processing of a shoot message illustrates that a new shot is created for
the proper ally for a given sample universe that otherwise remains unchanged. It
also illustrates that the universe’s iworlds also remain unchanged. A sample test
is the following:
(check-expect
(process-message OTHR-UNIV iworld1 SHOOT)
(make-bundle
(make-univ
(list iworld1 iworld2)
(make-world
(list (make-ally 7 "iworld1") (make-ally 9 "iworld2"))
(list (make-posn 3 3))
'right
(list (make-posn 7 14) (make-posn 1 2))))
(list
(make-mail
iworld1
(list
'world
(list (list 7 "iworld1") (list 9 "iworld2"))
(list (list 3 3))
165 Component Implementation 677

'right
(list (list 7 14) (list 1 2))))
(make-mail
iworld2
(list
'world
(list (list 7 "iworld1") (list 9 "iworld2"))
(list (list 3 3))
'right
(list (list 7 14) (list 1 2)))))
'()))

165.2.3 The add-player Handler

The handler to add new players must reject iworlds that have a name that is already
in use in the universe’s list of iworlds. In such a case, the handler returns a
bundle with an unchanged universe, an empty list of mails, and a one-iworld
list to disconnect from the server that contains the iworld attempting to connect.
The following sample expression illustrates the design:
;; Sample expressions for add-player
(define RPT-ADD (make-bundle OTHR-UNIV '() (list iworld1)))
OTHR-UNIV already has an iworld with the name "iworld1". The world trying to
connect is rejected. If the iworld connecting does not have a name already in use,
two cases must be distinguished: this is the first player that joins or it is not. If it is the
first player to join, the initial world containing the ally that just joined is used to
create a new universe. Otherwise, the new ally is added to the existing game. In
both cases the new iworld is added to the universe’s list of iworlds. According
to the protocol diagram in Fig. 144, a world message must be sent to all players.
The new world value containing the new ally must be marshaled to send to all the
players. The following sample expressions illustrate how this may be achieved:
(define
EMP-ADD
(local
[(define new-iws (cons iworld2 (univ-iws INIT-UNIV)))
(define game (univ-game INIT-UNIV))
(define new-game
(if (equal? game UNINIT-WORLD)
(make-world (list (make-ally
INIT-ROCKET
(iworld-name iworld2)))
INIT-LOA
INIT-DIR
INIT-LOS)
678 29 Aliens Attack Version 8

(make-world (cons (make-ally

INIT-ROCKET
(iworld-name iworld2))
(world-allies game))
(world-aliens game)
(world-dir game)
(world-shots game))))]
(make-bundle (make-univ new-iws new-game)
(map (λ (iw)
(make-mail
iw
(cons 'world
(marshal-world new-game))))
new-iws)
'())))

(define
NEW-ADD
(local
[(define new-iws (cons iworld3 (univ-iws OTHR-UNIV)))
(define game (univ-game OTHR-UNIV))
(define new-game (if (equal? game UNINIT-WORLD)
(make-world
(list (make-ally
INIT-ROCKET
(iworld-name iworld3)))
INIT-LOA
INIT-DIR
INIT-LOS)
(make-world
(cons (make-ally
INIT-ROCKET
(iworld-name iworld3))
(world-allies game))
(world-aliens game)
(world-dir game)
(world-shots game))))]
(make-bundle (make-univ new-iws new-game)
(map (λ (iw)
(make-mail
iw
(cons 'world
(marshal-world new-game))))
new-iws)
'())))
165 Component Implementation 679

Local variables are defined for the new list of iworlds and for the new world value.
The new list of iworlds is obtained by adding the joining iworld to the list of
iworlds in the given universe. An if-expression is used to define the new world
value by distinguishing whether or not the given world is the first to join. The body of
the local-expression returns a bundle with the new universe and a list of mails
with the marshaled new world.
Based on the sample expressions, the add-player handler is defined as follows:
;; universe iworld → bundle
;; Purpose: Add new world to the universe
(define (add-player a-univ an-iw)
(if (member?
(iworld-name an-iw)
(map iworld-name (univ-iws a-univ)))
(make-bundle a-univ '() (list an-iw))
(local [(define new-iws (cons an-iw (univ-iws a-univ)))
(define game (univ-game a-univ))
(define new-game (if (equal? game UNINIT-WORLD)
(make-world
(list (make-ally
INIT-ROCKET
(iworld-name an-iw)))
INIT-LOA
INIT-DIR
INIT-LOS)
(make-world
(cons (make-ally
INIT-ROCKET
(iworld-name an-iw))
(world-allies game))
(world-aliens game)
(world-dir game)
(world-shots game))))]
(make-bundle
(make-univ new-iws new-game)
(map
(λ (iw)
(make-mail
iw
(cons 'world (marshal-world new-game))))
new-iws)
'()))))
The handler is tested as follows:
;; Tests using sample computations for add-player
(check-expect (add-player OTHR-UNIV iworld1) RPT-ADD)
680 29 Aliens Attack Version 8

(check-expect (add-player INIT-UNIV iworld2) EMP-ADD)

(check-expect (add-player OTHR-UNIV iworld3) NEW-ADD)

;; Tests using sample values for add-player

(check-expect
(add-player (make-univ
(list iworld2 iworld3)
(make-world
(list (make-ally 7 "iworld2")
(make-ally 9 "iworld3"))
(list (make-posn 3 3))
'right
(list (make-posn 3 3))))
iworld1)
(make-bundle (make-univ
(list iworld1 iworld2 iworld3)
(make-world
(list (make-ally 10 "iworld1")
(make-ally 7 "iworld2")
(make-ally 9 "iworld3"))
(list (make-posn 3 3))
'right
(list (make-posn 3 3))))
(map
(λ (iw)
(make-mail
iw
(cons
'world
(marshal-world
(make-world (list (make-ally 10 "iworld1")
(make-ally 7 "iworld2")
(make-ally 9 "iworld3"))
(list (make-posn 3 3))
'right
(list (make-posn 3 3)))))))
(list iworld1 iworld2 iworld3))
'()))
Observe that the test using sample values illustrates how a new player, iworld1, is
added to the universe and how all the world mails are created.
165 Component Implementation 681

165.2.4 The rm-player Handler

The handler to remove a player needs to create a new world value by eliminating
the ally from the disconnected iworld and by removing the disconnected iworld
from the list of iworlds. This may be accomplished by filtering the list of allies and
the list of iworlds. In addition, as specified by the protocol diagram in Fig. 143, the
new world value must be sent to all the (remaining) players. The following sample
expressions illustrate how this is accomplished:
;; Sample expressions for rm-player
(define RM-IW1 (local
[(define iws (univ-iws OTHR-UNIV))
(define game (univ-game OTHR-UNIV))
(define new-iws (filter
(λ (iw)
(not (string=?
(iworld-name iworld1)
(iworld-name iw))))
iws))
(define new-game (make-world
(filter
(λ (a)
(not (string=?
(iworld-name iworld1)
(ally-name a))))
(world-allies game))
(world-aliens game)
(world-dir game)
(world-shots game)))]
(make-bundle (make-univ new-iws new-game)
(map (λ (iw)
(make-mail
iw
(cons 'world
(marshal-world
new-game))))
new-iws)
'())))

(define RM-IW2 (local [(define iws (univ-iws OTHR-UNIV2))

(define game (univ-game OTHR-UNIV2))
(define new-iws
(filter
(λ (iw)
(not (string=?
(iworld-name iworld2)
682 29 Aliens Attack Version 8

(iworld-name iw))))
iws))
(define new-game
(make-world
(filter
(λ (a)
(not (string=?
(iworld-name iworld2)
(ally-name a))))
(world-allies game))
(world-aliens game)
(world-dir game)
(world-shots game)))]
(make-bundle (make-univ new-iws new-game)
(map (λ (iw)
(make-mail
iw
(cons 'world
(marshal-world
new-game))))
new-iws)
'())))
Observe that filter is used to remove the disconnected iworld and its corre-
sponding ally. The creation of the tpms is done using map and the new list of
iworlds.
Abstracting over the sample expressions yields the function for this handler:
;; univ iworld → bundle
;; Purpose: Remove given iw from universe and game
;; ASSUMPTION: Given univ is not INIT-UNIV
(define (rm-player a-univ an-iw)
(local [(define iws (univ-iws a-univ))
(define game (univ-game a-univ))
(define new-iws (filter
(λ (iw)
(not (string=? (iworld-name an-iw)
(iworld-name iw))))
iws))
(define new-game (make-world
(filter
(λ (a)
(not (string=? (iworld-name an-iw)
(ally-name a))))
(world-allies game))
(world-aliens game)
(world-dir game)
(world-shots game)))]
166 A Subtle Problem 683

(make-bundle (make-univ new-iws new-game)

(map (λ (iw)
(make-mail iw (cons 'world
(marshal-world
new-game))))
new-iws)
'())))
The function is tested as follows:
;; Tests using sample computations for rm-player
(check-expect (rm-player OTHR-UNIV iworld1) RM-IW1)
(check-expect (rm-player OTHR-UNIV2 iworld2) RM-IW2)

;; Tests using sample computations for rm-player

(check-expect (rm-player (make-univ
(list iworld3)
(make-world
(list (make-ally 8 "iworld3"))
'()
'down
'()))
iworld3)
(make-bundle
(make-univ
'()
(make-world '() '() 'down '()))
'()
'()))
Observe that the test using sample values explicitly shows everything that is removed
from the universe.
This completes the design of Aliens Attack version 8. Make sure that you have
the needed constants in the server and player files. Run the tests. Once all the tests
pass, play the game on your computer. Observe that the players, unlike Aliens Attack
version 7, do not become unsynchronized. This is because all players get the same
game from the server.

166 A Subtle Problem

Play the game with one or two friends. Assuming your internet connection is fast
enough the game ought to run smoothly. However, as the number of players increases,
the game may become choppy. That is, the game becomes sluggish or momentarily
stops. This may be due to communication overhead. Communication overhead is
the proportion of time that is spent exchanging messages instead of advancing the
684 29 Aliens Attack Version 8

state of the game. If communication overhead is large enough, the game becomes
slow.
In distributed programming communication is necessary and, therefore, some
degree of communication overhead is necessary. It is desirable, however, to keep
this overhead small to make programs faster. There is no universal solution to this
problem. That is, implementing a solution may or may not produce the desired
result. Common approaches to reduce communication overhead are to limit the
number messages exchanged and reduce the size of the messages exchanged. In
Aliens Attack 8, for example, the world value is always sent to the players. This is
done despite the fact that only one component of the world has changed. A different
communication protocol may transmit to the players only the components that have
changed. This may or may not be effective but is worth trying if communication
overhead makes the game slow. The exercises have you explore the mitigation of
communication overhead.
***** Ex. 339 — Design and implement a communication protocol that only
transmits to the players the parts of the world that have changed in a message.
For example, when a move request is satisfied, a single ally is changed and
nothing else changes in the game. Instead of sending the entire world to all
players, only the changed ally may be sent. Is the resulting game noticeably
faster or slower?
***** Ex. 340 — Design and implement a communication protocol that does
not send every single change to the player as they happen. Instead, the server
may wait for a number of changes, n, to occur before sending a message to the
players. Is the resulting game noticeably faster or slower?

167 What Have We Learned in This Chapter?

The important lessons of Chap. 29 are summarized as follows:

• Allowing only one component to change shared data (e.g., the world value) in a
distributed program solves synchronization problems.
• The communication protocol for a distributed program using a thick server can be
significantly simpler than the communication protocol when using a thin server.
• Components not allowed to change the value of shared data make requests to the
server for changes to be made.
• Consistency in design makes it easier for others to understand the design of the
code and is likely to make future refinements easier to implement.
• Code reuse is an effective way to manage the complexity of refinements.
• Communication overhead, the proportion of time that is spent exchanging mes-
sages instead of computing, may make applications slow.
• The impact of communication overhead may be mitigated by reducing the fre-
quency and the size of messages.
 Part VI
Epilogue
Chapter 30
Advice for Future Steps

Congratulations! You have reached the end of your first step into the world of problem
solving and program design. There is still much more you can learn about problem
solving and programming, but there is no doubt that now you have a solid foundation
to continue on this journey. This journey is inevitable even for those who do not
aspire to become Computer Scientists. Remember that problem solving is at the
heart of many human activities.
Although the book builds on much of your background knowledge (e.g., high
school algebra), you likely feel you have a different understanding now of this
knowledge. The book has emphasized the use of types to organize your thoughts
during the problem solving process. At the heart of this process is how elements in the
real or an imaginary world are represented in a program. The chosen representation
may (and should) be exploited to find solutions to problems. Remember that if you
know the type of data to be processed, then you know something about what the
solution to a problem may look like.
All good things must continue . . .

168 Advice for Computer Science Students

Be patient and apply the skills you have learned in the future. As your studies
progress, you will discover that few textbooks on programming emphasize design.
This makes it difficult sometimes to understand the programs presented. When you
see such code, tease out the details. Try to formulate the steps of the design recipe to
enlighten you about what the program does. This a good skill to develop given that
large pieces of software evolve over years of development in which programmers
come and go. Documenting the design of a program is a service that you and others
will appreciate when you have to maintain code.
Where can you go from here? The most fundamental piece of advice is to read
and learn more about problem solving using a computer. You will be well-served
if you practice your design skills by learning about a new programming language
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 687
M. T. Morazán, Animated Problem Solving, Texts in Computer Science,
https://doi.org/10.1007/978-3-030-85091-3_30
688 30 Advice for Future Steps

every semester and summer. The skills you have developed are directly applicable
to solving problems using programming languages beyond BSL, BSL+, and ISL+.
As you advance, you will discover abstractions that are not covered in this textbook.
You are, however, well-prepared to learn about them. You will also be well-served to
explore topics covered in this book in more depth, such as big-O notation (i.e., com-
plexity), generic programming, and distributed programming. I cannot recommend
strongly enough to take courses in the implementation of programming languages:
truly understand the technology that is central to Computer Science.

169 Advice for Non-Computer Science Students

You may feel excited, overwhelmed, or both after completing this textbook. There is
no doubt that you are a better problem solver now. The truth is that you are likely to
program throughout your life. Perhaps not using a programming language as done
in this textbook, but if you are problem solving then you are programming. You may
write essays, diagnose a patient, or create a piece of music. What do these activities
have in common with programming? They process data and are refined until you are
satisfied with the result. Is this truly different than finally designing Aliens Attack 8?
Do you not refine several drafts of an essay? Use the lessons you have absorbed and
apply them to domains other than programming. The famous popular saying don’t
reinvent the wheel is a call for abstraction. If you think about it carefully, you will see
that the steps of the design recipe are applicable to problem solving in any context.
You are now in a much better position to understand and, therefore, use software.
If you use a spreadsheet, you are programming. If you adjust the settings of a
thermostat, you are programming. If you are driving, you are programming. No?
When you put your blinker on before turning, are you not sending a message to other
drivers? If you begin to realize that problem solving and programming are fully
intertwined with life, then you are now in a position to be a better problem solver.