Prepared exclusively for Zach

What others in the trenches say about

The Pragmatic Programmer. . .

“The cool thing about this book is that it’s great for keeping the
programming process fresh. [The book] helps you to continue to grow
and clearly comes from people who have been there.”
Kent Beck, author of Extreme Programming Explained:
Embrace Change

“I found this book to be a great mix of solid advice and wonderful

analogies!”
Martin Fowler, author of Refactoring and UML Distilled

“I would buy a copy, read it twice, then tell all my colleagues to run
out and grab a copy. This is a book I would never loan because I would
worry about it being lost.”
Kevin Ruland, Management Science, MSG-Logistics

“The wisdom and practical experience of the authors is obvious. The

topics presented are relevant and useful. . . . By far its greatest
strength for me has been the outstanding analogies—tracer bullets,
broken windows, and the fabulous helicopter-based explanation of the
need for orthogonality, especially in a crisis situation. I have little
doubt that this book will eventually become an excellent source of
useful information for journeymen programmers and expert mentors
alike.”
John Lakos, author of Large-Scale C++ Software Design

Prepared exclusively for Zach

 “This is the sort of book I will buy a dozen copies of when it comes out
so I can give it to my clients.”
Eric Vought, Software Engineer

“Most modern books on software development fail to cover the basics

of what makes a great software developer, instead spending their time
on syntax or technology where in reality the greatest leverage possible
for any software team is in having talented developers who really know
their craft well. An excellent book.”
Pete McBreen, Independent Consultant

“Since reading this book, I have implemented many of the practical

suggestions and tips it contains. Across the board, they have saved my
company time and money while helping me get my job done quicker!
This should be a desktop reference for everyone who works with code
for a living.”
Jared Richardson, Senior Software Developer,
iRenaissance, Inc.

“I would like to see this issued to every new employee at my

company. . . .”
Chris Cleeland, Senior Software Engineer,
Object Computing, Inc.

Prepared exclusively for Zach

 The Pragmatic Programmer

Prepared exclusively for Zach

 This page intentionally left blank

Prepared exclusively for Zach

 The Pragmatic Programmer
From Journeyman to Master
Andrew Hunt
David Thomas

ADDISON–WESLEY
An imprint of Addison Wesley Longman, Inc.
Reading, Massachusetts Harlow, England Menlo Park, California
Berkeley, California Don Mills, Ontario Sydney
Bonn Amsterdam Tokyo Mexico City

Prepared exclusively for Zach

 Many of the designations used by manufacturers and sellers to distinguish their products
are claimed as trademarks. Where those designations appear in this book, and Addison–
Wesley was aware of a trademark claim, the designations have been printed in initial
capital letters or in all capitals.

Lyrics from the song “The Boxer” on page 157 are Copyright c 1968 Paul Simon. Used by
permission of the Publisher: Paul Simon Music. Lyrics from the song “Alice’s Restaurant”
on page 220 are by Arlo Guthrie, c 1966, 1967 (renewed) by A PPLESEED M USIC I NC. All
Rights Reserved. Used by Permission.

The authors and publisher have taken care in the preparation of this book, but make
no express or implied warranty of any kind and assume no responsibility for errors or
omissions. No liability is assumed for incidental or consequential damages in connection
with or arising out of the use of the information or programs contained herein.

The publisher offers discounts on this book when ordered in quantity for special sales.
For more information, please contact:
AWL Direct Sales
Addison Wesley Longman, Inc.
One Jacob Way
Reading, Massachusetts 01867
(781) 944-3700

Visit AWL on the Web: www.awl.com/cseng

Library of Congress Cataloging-in-Publication Data

Hunt, Andrew, 1964 –

The Pragmatic Programmer / Andrew Hunt, David Thomas.
p. cm.
Includes bibliographical references.
ISBN 0-201-61622-X
1. Computer programming. I. Thomas, David, 1956– .
II. Title.
QA76.6.H857 1999
005.1--dc21 99–43581
CIP

Copyright c 2000 by Addison Wesley Longman, Inc.

All rights reserved. No part of this publication may be reproduced, stored in a retrieval
system, or transmitted, in any form or by any means, electronic, mechanical, photo-
copying, recording, or otherwise, without the prior written permission of the publisher.
Printed in the United States of America. Published simultaneously in Canada.

ISBN 0-201-61622-X
Text printed in the United States on recycled paper at Courier Stoughton in Stoughton, Massachusetts.
25th Printing February 2010

Prepared exclusively for Zach

 For Ellie and Juliet,
Elizabeth and Zachary,
Stuart and Henry

Prepared exclusively for Zach

 This page intentionally left blank

Prepared exclusively for Zach

 Contents
F OREWORD xiii

P REFACE xvii

1 A P RAGMATIC P HILOSOPHY 1
1. The Cat Ate My Source Code . . . . . . . . . . . . . . . . . 2
2. Software Entropy . . . . . . . . . . . . . . . . . . . . . . . . 4
3. Stone Soup and Boiled Frogs . . . . . . . . . . . . . . . . . 7
4. Good-Enough Software . . . . . . . . . . . . . . . . . . . . 9
5. Your Knowledge Portfolio . . . . . . . . . . . . . . . . . . . 12
6. Communicate! . . . . . . . . . . . . . . . . . . . . . . . . . 18

2 A P RAGMATIC A PPROACH 25
7. The Evils of Duplication . . . . . . . . . . . . . . . . . . . . 26
8. Orthogonality . . . . . . . . . . . . . . . . . . . . . . . . . . 34
9. Reversibility . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
10. Tracer Bullets . . . . . . . . . . . . . . . . . . . . . . . . . 48
11. Prototypes and Post-it Notes . . . . . . . . . . . . . . . . . 53
12. Domain Languages . . . . . . . . . . . . . . . . . . . . . . 57
13. Estimating . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

3 T HE B ASIC T OOLS 71
14. The Power of Plain Text . . . . . . . . . . . . . . . . . . . . 73
15. Shell Games . . . . . . . . . . . . . . . . . . . . . . . . . . 77
16. Power Editing . . . . . . . . . . . . . . . . . . . . . . . . . . 82
17. Source Code Control . . . . . . . . . . . . . . . . . . . . . . 86
18. Debugging . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
19. Text Manipulation . . . . . . . . . . . . . . . . . . . . . . . 99
20. Code Generators . . . . . . . . . . . . . . . . . . . . . . . . 102

ix

Prepared exclusively for Zach

 x C ONTENTS

4 P RAGMATIC PARANOIA 107

21. Design by Contract . . . . . . . . . . . . . . . . . . . . . . 109
22. Dead Programs Tell No Lies . . . . . . . . . . . . . . . . . 120
23. Assertive Programming . . . . . . . . . . . . . . . . . . . . 122
24. When to Use Exceptions . . . . . . . . . . . . . . . . . . . 125
25. How to Balance Resources . . . . . . . . . . . . . . . . . . 129

5 B END , OR B REAK 137

26. Decoupling and the Law of Demeter . . . . . . . . . . . . . 138
27. Metaprogramming . . . . . . . . . . . . . . . . . . . . . . . 144
28. Temporal Coupling . . . . . . . . . . . . . . . . . . . . . . . 150
29. It’s Just a View . . . . . . . . . . . . . . . . . . . . . . . . . 157
30. Blackboards . . . . . . . . . . . . . . . . . . . . . . . . . . 165

6 W HILE Y OU A RE C ODING 171

31. Programming by Coincidence . . . . . . . . . . . . . . . . 172
32. Algorithm Speed . . . . . . . . . . . . . . . . . . . . . . . . 177
33. Refactoring . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
34. Code That’s Easy to Test . . . . . . . . . . . . . . . . . . . 189
35. Evil Wizards . . . . . . . . . . . . . . . . . . . . . . . . . . 198

7 B EFORE THE P ROJECT 201

36. The Requirements Pit . . . . . . . . . . . . . . . . . . . . . 202
37. Solving Impossible Puzzles . . . . . . . . . . . . . . . . . . 212
38. Not Until You’re Ready . . . . . . . . . . . . . . . . . . . . 215
39. The Specification Trap . . . . . . . . . . . . . . . . . . . . 217
40. Circles and Arrows . . . . . . . . . . . . . . . . . . . . . . . 220

8 P RAGMATIC P ROJECTS 223

41. Pragmatic Teams . . . . . . . . . . . . . . . . . . . . . . . . 224
42. Ubiquitous Automation . . . . . . . . . . . . . . . . . . . . 230
43. Ruthless Testing . . . . . . . . . . . . . . . . . . . . . . . . 237
44. It’s All Writing . . . . . . . . . . . . . . . . . . . . . . . . . 248
45. Great Expectations . . . . . . . . . . . . . . . . . . . . . . 255
46. Pride and Prejudice . . . . . . . . . . . . . . . . . . . . . . 258

Prepared exclusively for Zach

 C ONTENTS xi

Appendices

A R ESOURCES 261
Professional Societies . . . . . . . . . . . . . . . . . . . . . . . . 262
Building a Library . . . . . . . . . . . . . . . . . . . . . . . . . . 262
Internet Resources . . . . . . . . . . . . . . . . . . . . . . . . . 266
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275

B A NSWERS TO E XERCISES 279

I NDEX 309

Prepared exclusively for Zach

 This page intentionally left blank

Prepared exclusively for Zach

 Foreword
As a reviewer I got an early opportunity to read the book you are hold-
ing. It was great, even in draft form. Dave Thomas and Andy Hunt have
something to say, and they know how to say it. I saw what they were
doing and I knew it would work. I asked to write this foreword so that I
could explain why.

Simply put, this book tells you how to program in a way that you can
follow. You wouldn’t think that that would be a hard thing to do, but it
is. Why? For one thing, not all programming books are written by pro-
grammers. Many are compiled by language designers, or the journalists
who work with them to promote their creations. Those books tell you
how to talk in a programming language—which is certainly important,
but that is only a small part of what a programmer does.

What does a programmer do besides talk in programming language?

Well, that is a deeper issue. Most programmers would have trouble
explaining what they do. Programming is a job filled with details, and
keeping track of those details requires focus. Hours drift by and the
code appears. You look up and there are all of those statements. If you
don’t think carefully, you might think that programming is just typing
statements in a programming language. You would be wrong, of course,
but you wouldn’t be able to tell by looking around the programming
section of the bookstore.

In The Pragmatic Programmer Dave and Andy tell us how to program in

a way that we can follow. How did they get so smart? Aren’t they just
as focused on details as other programmers? The answer is that they
paid attention to what they were doing while they were doing it—and
then they tried to do it better.

Imagine that you are sitting in a meeting. Maybe you are thinking
that the meeting could go on forever and that you would rather be
programming. Dave and Andy would be thinking about why they were

xiii

Prepared exclusively for Zach

 xiv F OREWORD

having the meeting, and wondering if there is something else they could
do that would take the place of the meeting, and deciding if that some-
thing could be automated so that the work of the meeting just happens
in the future. Then they would do it.

That is just the way Dave and Andy think. That meeting wasn’t some-
thing keeping them from programming. It was programming. And it
was programming that could be improved. I know they think this way
because it is tip number two: Think About Your Work.

So imagine that these guys are thinking this way for a few years.
Pretty soon they would have a collection of solutions. Now imagine
them using their solutions in their work for a few more years, and
discarding the ones that are too hard or don’t always produce results.
Well, that approach just about defines pragmatic. Now imagine them
taking a year or two more to write their solutions down. You might
think, That information would be a gold mine. And you would be right.

The authors tell us how they program. And they tell us in a way that we
can follow. But there is more to this second statement than you might
think. Let me explain.

The authors have been careful to avoid proposing a theory of software

development. This is fortunate, because if they had they would be
obliged to warp each chapter to defend their theory. Such warping is
the tradition in, say, the physical sciences, where theories eventually
become laws or are quietly discarded. Programming on the other hand
has few (if any) laws. So programming advice shaped around wanna-be
laws may sound good in writing, but it fails to satisfy in practice. This
is what goes wrong with so many methodology books.

I’ve studied this problem for a dozen years and found the most promise
in a device called a pattern language. In short, a pattern is a solution,
and a pattern language is a system of solutions that reinforce each
other. A whole community has formed around the search for these
systems.

This book is more than a collection of tips. It is a pattern language

in sheep’s clothing. I say that because each tip is drawn from experi-
ence, told as concrete advice, and related to others to form a system.
These are the characteristics that allow us to learn and follow a pattern
language. They work the same way here.

Prepared exclusively for Zach

 F OREWORD xv

You can follow the advice in this book because it is concrete. You won’t
find vague abstractions. Dave and Andy write directly for you, as if each
tip was a vital strategy for energizing your programming career. They
make it simple, they tell a story, they use a light touch, and then they
follow that up with answers to questions that will come up when you
try.

And there is more. After you read ten or fifteen tips you will begin to see
an extra dimension to the work. We sometimes call it QWAN, short for
the quality without a name. The book has a philosophy that will ooze
into your consciousness and mix with your own. It doesn’t preach. It
just tells what works. But in the telling more comes through. That’s the
beauty of the book: It embodies its philosophy, and it does so unpre-
tentiously.

So here it is: an easy to read—and use—book about the whole practice

of programming. I’ve gone on and on about why it works. You probably
only care that it does work. It does. You will see.

—Ward Cunningham

Prepared exclusively for Zach

 This page intentionally left blank

Prepared exclusively for Zach

 Preface
This book will help you become a better programmer.

It doesn’t matter whether you are a lone developer, a member of a large

project team, or a consultant working with many clients at once. This
book will help you, as an individual, to do better work. This book isn’t
theoretical—we concentrate on practical topics, on using your experi-
ence to make more informed decisions. The word pragmatic comes from
the Latin pragmaticus—“skilled in business”—which itself is derived
from the Greek , meaning “to do.” This is a book about doing.

Programming is a craft. At its simplest, it comes down to getting a

computer to do what you want it to do (or what your user wants it to do).
As a programmer, you are part listener, part advisor, part interpreter,
and part dictator. You try to capture elusive requirements and find a
way of expressing them so that a mere machine can do them justice.
You try to document your work so that others can understand it, and
you try to engineer your work so that others can build on it. What’s
more, you try to do all this against the relentless ticking of the project
clock. You work small miracles every day.

It’s a difficult job.

There are many people offering you help. Tool vendors tout the mir-
acles their products perform. Methodology gurus promise that their
techniques guarantee results. Everyone claims that their programming
language is the best, and every operating system is the answer to all
conceivable ills.

Of course, none of this is true. There are no easy answers. There is no

such thing as a best solution, be it a tool, a language, or an operat-
ing system. There can only be systems that are more appropriate in a
particular set of circumstances.

xvii

Prepared exclusively for Zach

 xviii P REFACE

This is where pragmatism comes in. You shouldn’t be wedded to any

particular technology, but have a broad enough background and expe-
rience base to allow you to choose good solutions in particular situ-
ations. Your background stems from an understanding of the basic
principles of computer science, and your experience comes from a wide
range of practical projects. Theory and practice combine to make you
strong.

You adjust your approach to suit the current circumstances and envi-
ronment. You judge the relative importance of all the factors affecting a
project and use your experience to produce appropriate solutions. And
you do this continuously as the work progresses. Pragmatic Program-
mers get the job done, and do it well.

Who Should Read This Book?

This book is aimed at people who want to become more effective and
more productive programmers. Perhaps you feel frustrated that you
don’t seem to be achieving your potential. Perhaps you look at col-
leagues who seem to be using tools to make themselves more produc-
tive than you. Maybe your current job uses older technologies, and you
want to know how newer ideas can be applied to what you do.

We don’t pretend to have all (or even most) of the answers, nor are
all of our ideas applicable in all situations. All we can say is that if
you follow our approach, you’ll gain experience rapidly, your produc-
tivity will increase, and you’ll have a better understanding of the entire
development process. And you’ll write better software.

What Makes a Pragmatic Programmer?

Each developer is unique, with individual strengths and weaknesses,
preferences and dislikes. Over time, each will craft his or her own
personal environment. That environment will reflect the programmer’s
individuality just as forcefully as his or her hobbies, clothing, or hair-
cut. However, if you’re a Pragmatic Programmer, you’ll share many of
the following characteristics:

Early adopter/fast adapter. You have an instinct for technologies

and techniques, and you love trying things out. When given some-

Prepared exclusively for Zach

 P REFACE xix

thing new, you can grasp it quickly and integrate it with the rest of
your knowledge. Your confidence is born of experience.

Inquisitive. You tend to ask questions. That’s neat—how did you

do that? Did you have problems with that library? What’s this BeOS
I’ve heard about? How are symbolic links implemented? You are a
pack rat for little facts, each of which may affect some decision
years from now.

Critical thinker. You rarely take things as given without first get-
ting the facts. When colleagues say “because that’s the way it’s
done,” or a vendor promises the solution to all your problems, you
smell a challenge.

Realistic. You try to understand the underlying nature of each

problem you face. This realism gives you a good feel for how diffi-
cult things are, and how long things will take. Understanding for
yourself that a process should be difficult or will take a while to
complete gives you the stamina to keep at it.

Jack of all trades. You try hard to be familiar with a broad range
of technologies and environments, and you work to keep abreast of
new developments. Although your current job may require you to
be a specialist, you will always be able to move on to new areas and
new challenges.

We’ve left the most basic characteristics until last. All Pragmatic Pro-
grammers share them. They’re basic enough to state as tips:

TIP 1

Care About Your Craft

We feel that there is no point in developing software unless you care

about doing it well.

TIP 2

Think! About Your Work

In order to be a Pragmatic Programmer, we’re challenging you to think

about what you’re doing while you’re doing it. This isn’t a one-time
audit of current practices—it’s an ongoing critical appraisal of every

Prepared exclusively for Zach

 xx P REFACE

decision you make, every day, and on every development. Never run on
auto-pilot. Constantly be thinking, critiquing your work in real time.
The old IBM corporate motto, THINK!, is the Pragmatic Programmer’s
mantra.

If this sounds like hard work to you, then you’re exhibiting the realistic
characteristic. This is going to take up some of your valuable time—time
that is probably already under tremendous pressure. The reward is a
more active involvement with a job you love, a feeling of mastery over
an increasing range of subjects, and pleasure in a feeling of continuous
improvement. Over the long term, your time investment will be repaid
as you and your team become more efficient, write code that’s easier to
maintain, and spend less time in meetings.

Individual Pragmatists, Large Teams

Some people feel that there is no room for individuality on large teams
or complex projects. “Software construction is an engineering disci-
pline,” they say, “that breaks down if individual team members make
decisions for themselves.”

We disagree.

The construction of software should be an engineering discipline. How-

ever, this doesn’t preclude individual craftsmanship. Think about the
large cathedrals built in Europe during the Middle Ages. Each took
thousands of person-years of effort, spread over many decades. Lessons
learned were passed down to the next set of builders, who advanced
the state of structural engineering with their accomplishments. But the
carpenters, stonecutters, carvers, and glass workers were all craftspeo-
ple, interpreting the engineering requirements to produce a whole that
transcended the purely mechanical side of the construction. It was their
belief in their individual contributions that sustained the projects:

We who cut mere stones must always be envisioning cathedrals.

— Quarry worker’s creed

Within the overall structure of a project there is always room for in-
dividuality and craftsmanship. This is particularly true given the cur-
rent state of software engineering. One hundred years from now, our
engineering may seem as archaic as the techniques used by medieval

Prepared exclusively for Zach

 P REFACE xxi

cathedral builders seem to today’s civil engineers, while our craftsman-

ship will still be honored.

It’s a Continuous Process

A tourist visiting England’s Eton College asked the gardener how he got
the lawns so perfect. “That’s easy,” he replied, “You just brush off the
dew every morning, mow them every other day, and roll them once a
week.”

“Is that all?” asked the tourist.

“Absolutely,” replied the gardener. “Do that for 500 years and you’ll
have a nice lawn, too.”

Great lawns need small amounts of daily care, and so do great pro-
grammers. Management consultants like to drop the word kaizen in
conversations. “Kaizen” is a Japanese term that captures the concept
of continuously making many small improvements. It was considered
to be one of the main reasons for the dramatic gains in productivity and
quality in Japanese manufacturing and was widely copied throughout
the world. Kaizen applies to individuals, too. Every day, work to refine
the skills you have and to add new tools to your repertoire. Unlike the
Eton lawns, you’ll start seeing results in a matter of days. Over the
years, you’ll be amazed at how your experience has blossomed and
your skills have grown.

How the Book Is Organized

This book is written as a collection of short sections. Each section is
self-contained, and addresses a particular topic. You’ll find numerous
cross references, which help put each topic in context. Feel free to read
the sections in any order—this isn’t a book you need to read front-to-
back.

Occasionally you’ll come across a box labeled Tip nn (such as Tip 1,

“Care About Your Craft” on page xix). As well as emphasizing points in
the text, we feel the tips have a life of their own—we live by them daily.
You’ll find a summary of all the tips on a pull-out card inside the back
cover.

Prepared exclusively for Zach

 xxii P REFACE

Appendix A contains a set of resources: the book’s bibliography, a list of

URLs to Web resources, and a list of recommended periodicals, books,
and professional organizations. Throughout the book you’ll find refer-
ences to the bibliography and to the list of URLs—such as [KP99] and
[URL 18], respectively.

We’ve included exercises and challenges where appropriate. Exercises

normally have relatively straightforward answers, while the challenges
are more open-ended. To give you an idea of our thinking, we’ve in-
cluded our answers to the exercises in Appendix B, but very few have
a single correct solution. The challenges might form the basis of group
discussions or essay work in advanced programming courses.

What’s in a Name?
“When I use a word,” Humpty Dumpty said, in rather a scornful tone, “it means
just what I choose it to mean—neither more nor less.”
Lewis Carroll, Through the Looking-Glass

Scattered throughout the book you’ll find various bits of jargon—either

perfectly good English words that have been corrupted to mean some-
thing technical, or horrendous made-up words that have been assigned
meanings by computer scientists with a grudge against the language.
The first time we use each of these jargon words, we try to define it,
or at least give a hint to its meaning. However, we’re sure that some
have fallen through the cracks, and others, such as object and rela-
tional database, are in common enough usage that adding a definition
would be boring. If you do come across a term you haven’t seen be-
fore, please don’t just skip over it. Take time to look it up, perhaps on
the Web, or maybe in a computer science textbook. And, if you get a
chance, drop us an e-mail and complain, so we can add a definition to
the next edition.

Having said all this, we decided to get revenge against the computer sci-
entists. Sometimes, there are perfectly good jargon words for concepts,
words that we’ve decided to ignore. Why? Because the existing jargon
is normally restricted to a particular problem domain, or to a partic-
ular phase of development. However, one of the basic philosophies of
this book is that most of the techniques we’re recommending are uni-
versal: modularity applies to code, designs, documentation, and team

Prepared exclusively for Zach

 P REFACE xxiii

organization, for instance. When we wanted to use the conventional

jargon word in a broader context, it got confusing—we couldn’t seem
to overcome the baggage the original term brought with it. When this
happened, we contributed to the decline of the language by inventing
our own terms.

Source Code and Other Resources

Most of the code shown in this book is extracted from compilable source
files, available for download from our Web site:

www.pragmaticprogrammer.com

There you’ll also find links to resources we find useful, along with
updates to the book and news of other Pragmatic Programmer devel-
opments.

Send Us Feedback
We’d appreciate hearing from you. Comments, suggestions, errors in
the text, and problems in the examples are all welcome. E-mail us at

[email protected]

Acknowledgments
When we started writing this book, we had no idea how much of a team
effort it would end up being.

Addison-Wesley has been brilliant, taking a couple of wet-behind-the-

ears hackers and walking us through the whole book-production pro-
cess, from idea to camera-ready copy. Many thanks to John Wait
and Meera Ravindiran for their initial support, Mike Hendrickson, our
enthusiastic editor (and a mean cover designer!), Lorraine Ferrier and
John Fuller for their help with production, and the indefatigable Julie
DeBaggis for keeping us all together.

Then there were the reviewers: Greg Andress, Mark Cheers, Chris Clee-
land, Alistair Cockburn, Ward Cunningham, Martin Fowler, Thanh
T. Giang, Robert L. Glass, Scott Henninger, Michael Hunter, Brian

Prepared exclusively for Zach

 xxiv P REFACE

Kirby, John Lakos, Pete McBreen, Carey P. Morris, Jared Richardson,

Kevin Ruland, Eric Starr, Eric Vought, Chris Van Wyk, and Deborra
Zukowski. Without their careful comments and valuable insights, this
book would be less readable, less accurate, and twice as long. Thank
you all for your time and wisdom.

The second printing of this book benefited greatly from the eagle eyes
of our readers. Many thanks to Brian Blank, Paul Boal, Tom Ekberg,
Brent Fulgham, Louis Paul Hebert, Henk-Jan Olde Loohuis, Alan Lund,
Gareth McCaughan, Yoshiki Shibata, and Volker Wurst, both for find-
ing the mistakes and for having the grace to point them out gently.

Over the years, we have worked with a large number of progressive

clients, where we gained and refined the experience we write about
here. Recently, we’ve been fortunate to work with Peter Gehrke on sev-
eral large projects. His support and enthusiasm for our techniques are
much appreciated.

This book was produced using LATEX, pic, Perl, dvips, ghostview, ispell,
GNU make, CVS, Emacs, XEmacs, EGCS, GCC, Java, iContract, and
SmallEiffel, using the Bash and zsh shells under Linux. The stagger-
ing thing is that all of this tremendous software is freely available. We
owe a huge “thank you” to the thousands of Pragmatic Programmers
worldwide who have contributed these and other works to us all. We’d
particularly like to thank Reto Kramer for his help with iContract.

Last, but in no way least, we owe a huge debt to our families. Not only
have they put up with late night typing, huge telephone bills, and our
permanent air of distraction, but they’ve had the grace to read what
we’ve written, time after time. Thank you for letting us dream.

Andy Hunt
Dave Thomas

Prepared exclusively for Zach

 Chapter 1

A Pragmatic Philosophy
What distinguishes Pragmatic Programmers? We feel it’s an attitude, a
style, a philosophy of approaching problems and their solutions. They
think beyond the immediate problem, always trying to place it in its
larger context, always trying to be aware of the bigger picture. After all,
without this larger context, how can you be pragmatic? How can you
make intelligent compromises and informed decisions?

Another key to their success is that they take responsibility for every-
thing they do, which we discuss in The Cat Ate My Source Code. Being
responsible, Pragmatic Programmers won’t sit idly by and watch their
projects fall apart through neglect. In Software Entropy, we tell you how
to keep your projects pristine.

Most people find change difficult to accept, sometimes for good reasons,
sometimes because of plain old inertia. In Stone Soup and Boiled Frogs,
we look at a strategy for instigating change and (in the interests of
balance) present the cautionary tale of an amphibian that ignored the
dangers of gradual change.

One of the benefits of understanding the context in which you work

is that it becomes easier to know just how good your software has to
be. Sometimes near-perfection is the only option, but often there are
trade-offs involved. We explore this in Good-Enough Software.

Of course, you need to have a broad base of knowledge and experience

to pull all of this off. Learning is a continuous and ongoing process.
In Your Knowledge Portfolio, we discuss some strategies for keeping the
momentum up.

Prepared exclusively for Zach

 2 C HAPTER 1 A P RAGMATIC P HILOSOPHY

Finally, none of us works in a vacuum. We all spend a large amount

of time interacting with others. Communicate! lists ways we can do this
better.

Pragmatic programming stems from a philosophy of pragmatic think-

ing. This chapter sets the basis for that philosophy.

1 The Cat Ate My Source Code

The greatest of all weaknesses is the fear of appearing weak.
J. B. Bossuet, Politics from Holy Writ, 1709

One of the cornerstones of the pragmatic philosophy is the idea of tak-

ing responsibility for yourself and your actions in terms of your career
advancement, your project, and your day-to-day work. A Pragmatic Pro-
grammer takes charge of his or her own career, and isn’t afraid to admit
ignorance or error. It’s not the most pleasant aspect of programming,
to be sure, but it will happen—even on the best of projects. Despite
thorough testing, good documentation, and solid automation, things
go wrong. Deliveries are late. Unforeseen technical problems come up.

These things happen, and we try to deal with them as professionally as

we can. This means being honest and direct. We can be proud of our
abilities, but we must be honest about our shortcomings—our igno-
rance as well as our mistakes.

Take Responsibility
Responsibility is something you actively agree to. You make a commit-
ment to ensure that something is done right, but you don’t necessarily
have direct control over every aspect of it. In addition to doing your own
personal best, you must analyze the situation for risks that are beyond
your control. You have the right not to take on a responsibility for an
impossible situation, or one in which the risks are too great. You’ll have
to make the call based on your own ethics and judgment.

When you do accept the responsibility for an outcome, you should ex-
pect to be held accountable for it. When you make a mistake (as we all
do) or an error in judgment, admit it honestly and try to offer options.

Prepared exclusively for Zach

 T HE C AT A TE M Y S OURCE C ODE 3

Don’t blame someone or something else, or make up an excuse. Don’t

blame all the problems on a vendor, a programming language, manage-
ment, or your coworkers. Any and all of these may play a role, but it is
up to you to provide solutions, not excuses.

If there was a risk that the vendor wouldn’t come through for you, then
you should have had a contingency plan. If the disk crashes—taking
all of your source code with it—and you don’t have a backup, it’s your
fault. Telling your boss “the cat ate my source code” just won’t cut it.

TIP 3

Provide Options, Don’t Make Lame Excuses

Before you approach anyone to tell them why something can’t be done,
is late, or is broken, stop and listen to yourself. Talk to the rubber
duck on your monitor, or the cat. Does your excuse sound reasonable,
or stupid? How’s it going to sound to your boss?

Run through the conversation in your mind. What is the other person
likely to say? Will they ask, “Have you tried this ” or “Didn’t you con-
sider that?” How will you respond? Before you go and tell them the bad
news, is there anything else you can try? Sometimes, you just know
what they are going to say, so save them the trouble.

Instead of excuses, provide options. Don’t say it can’t be done; explain

what can be done to salvage the situation. Does code have to be thrown
out? Educate them on the value of refactoring (see Refactoring, page
184). Do you need to spend time prototyping to determine the best way
to proceed (see Prototypes and Post-it Notes, page 53)? Do you need to
introduce better testing (see Code That’s Easy to Test, page 189, and
Ruthless Testing, page 237) or automation (see Ubiquitous Automation,
page 230) to prevent it from happening again? Perhaps you need ad-
ditional resources. Don’t be afraid to ask, or to admit that you need
help.

Try to flush out the lame excuses before voicing them aloud. If you
must, tell your cat first. After all, if little Tiddles is going to take the
blame. . . .

Prepared exclusively for Zach

 4 C HAPTER 1 A P RAGMATIC P HILOSOPHY

Related sections include:

Prototypes and Post-it Notes, page 53
Refactoring, page 184
Code That’s Easy to Test, page 189
Ubiquitous Automation, page 230
Ruthless Testing, page 237

Challenges
How do you react when someone—such as a bank teller, an auto mechanic,
or a clerk—comes to you with a lame excuse? What do you think of them
and their company as a result?

2 Software Entropy
While software development is immune from almost all physical laws,
entropy hits us hard. Entropy is a term from physics that refers to the
amount of “disorder” in a system. Unfortunately, the laws of thermo-
dynamics guarantee that the entropy in the universe tends toward a
maximum. When disorder increases in software, programmers call it
“software rot.”

There are many factors that can contribute to software rot. The most
important one seems to be the psychology, or culture, at work on a
project. Even if you are a team of one, your project’s psychology can
be a very delicate thing. Despite the best laid plans and the best peo-
ple, a project can still experience ruin and decay during its lifetime. Yet
there are other projects that, despite enormous difficulties and con-
stant setbacks, successfully fight nature’s tendency toward disorder
and manage to come out pretty well.

What makes the difference?

In inner cities, some buildings are beautiful and clean, while others
are rotting hulks. Why? Researchers in the field of crime and urban
decay discovered a fascinating trigger mechanism, one that very quickly
turns a clean, intact, inhabited building into a smashed and abandoned
derelict [WK82].

Prepared exclusively for Zach

 S OFTWARE E NTROPY 5

A broken window.

One broken window, left unrepaired for any substantial length of time,
instills in the inhabitants of the building a sense of abandonment—a
sense that the powers that be don’t care about the building. So another
window gets broken. People start littering. Graffiti appears. Serious
structural damage begins. In a relatively short space of time, the build-
ing becomes damaged beyond the owner’s desire to fix it, and the sense
of abandonment becomes reality.

The “Broken Window Theory” has inspired police departments in New

York and other major cities to crack down on the small stuff in order
to keep out the big stuff. It works: keeping on top of broken windows,
graffiti, and other small infractions has reduced the serious crime level.

TIP 4

Don’t Live with Broken Windows

Don’t leave “broken windows” (bad designs, wrong decisions, or poor

code) unrepaired. Fix each one as soon as it is discovered. If there is
insufficient time to fix it properly, then board it up. Perhaps you can
comment out the offending code, or display a "Not Implemented" mes-
sage, or substitute dummy data instead. Take some action to prevent
further damage and to show that you’re on top of the situation.

We’ve seen clean, functional systems deteriorate pretty quickly once

windows start breaking. There are other factors that can contribute to
software rot, and we’ll touch on some of them elsewhere, but neglect
accelerates the rot faster than any other factor.

You may be thinking that no one has the time to go around cleaning
up all the broken glass of a project. If you continue to think like that,
then you’d better plan on getting a dumpster, or moving to another
neighborhood. Don’t let entropy win.

Putting Out Fires

By contrast, there’s the story of an obscenely rich acquaintance of
Andy’s. His house was immaculate, beautiful, loaded with priceless
antiques, objets d’art, and so on. One day, a tapestry that was hang-
ing a little too close to his living room fireplace caught on fire. The fire

Prepared exclusively for Zach

 6 C HAPTER 1 A P RAGMATIC P HILOSOPHY

department rushed in to save the day—and his house. But before they
dragged their big, dirty hoses into the house, they stopped—with the
fire raging—to roll out a mat between the front door and the source of
the fire.

They didn’t want to mess up the carpet.

A pretty extreme case, to be sure, but that’s the way it must be with
software. One broken window—a badly designed piece of code, a poor
management decision that the team must live with for the duration
of the project—is all it takes to start the decline. If you find yourself
working on a project with quite a few broken windows, it’s all too easy
to slip into the mindset of “All the rest of this code is crap, I’ll just follow
suit.” It doesn’t matter if the project has been fine up to this point.
In the original experiment leading to the “Broken Window Theory,” an
abandoned car sat for a week untouched. But once a single window was
broken, the car was stripped and turned upside down within hours.

By the same token, if you find yourself on a team and a project where
the code is pristinely beautiful—cleanly written, well designed, and
elegant—you will likely take extra special care not to mess it up, just
like the firefighters. Even if there’s a fire raging (deadline, release date,
trade show demo, etc.), you don’t want to be the first one to make a
mess.

Related sections include:

Stone Soup and Boiled Frogs, page 7
Refactoring, page 184
Pragmatic Teams, page 224

Challenges
Help strengthen your team by surveying your computing “neighborhood.”
Choose two or three “broken windows” and discuss with your colleagues
what the problems are and what could be done to fix them.

Can you tell when a window first gets broken? What is your reaction? If
it was the result of someone else’s decision, or a management edict, what
can you do about it?

Prepared exclusively for Zach

 S TONE S OUP AND B OILED F ROGS 7

3 Stone Soup and Boiled Frogs

The three soldiers returning home from war were hungry. When they saw the
village ahead their spirits lifted—they were sure the villagers would give them
a meal. But when they got there, they found the doors locked and the windows
closed. After many years of war, the villagers were short of food, and hoarded
what they had.

Undeterred, the soldiers boiled a pot of water and carefully placed three stones
into it. The amazed villagers came out to watch.

“This is stone soup,” the soldiers explained. “Is that all you put in it?” asked
the villagers. “Absolutely—although some say it tastes even better with a few
carrots .” A villager ran off, returning in no time with a basket of carrots from
his hoard.

A couple of minutes later, the villagers again asked “Is that it?”

“Well,” said the soldiers, “a couple of potatoes give it body.” Off ran another
villager.

Over the next hour, the soldiers listed more ingredients that would enhance the
soup: beef, leeks, salt, and herbs. Each time a different villager would run off to
raid their personal stores.

Eventually they had produced a large pot of steaming soup. The soldiers removed
the stones, and they sat down with the entire village to enjoy the first square
meal any of them had eaten in months.

There are a couple of morals in the stone soup story. The villagers are
tricked by the soldiers, who use the villagers’ curiosity to get food from
them. But more importantly, the soldiers act as a catalyst, bringing
the village together so they can jointly produce something that they
couldn’t have done by themselves—a synergistic result. Eventually ev-
eryone wins.

Every now and then, you might want to emulate the soldiers.

You may be in a situation where you know exactly what needs doing
and how to do it. The entire system just appears before your eyes—you
know it’s right. But ask permission to tackle the whole thing and you’ll
be met with delays and blank stares. People will form committees, bud-
gets will need approval, and things will get complicated. Everyone will
guard their own resources. Sometimes this is called “start-up fatigue.”

Prepared exclusively for Zach

 8 C HAPTER 1 A P RAGMATIC P HILOSOPHY

It’s time to bring out the stones. Work out what you can reasonably
ask for. Develop it well. Once you’ve got it, show people, and let them
marvel. Then say “of course, it would be better if we added .” Pretend
it’s not important. Sit back and wait for them to start asking you to
add the functionality you originally wanted. People find it easier to join
an ongoing success. Show them a glimpse of the future and you’ll get
them to rally around.1

TIP 5

Be a Catalyst for Change

The Villagers’ Side

On the other hand, the stone soup story is also about gentle and grad-
ual deception. It’s about focusing too tightly. The villagers think about
the stones and forget about the rest of the world. We all fall for it, every
day. Things just creep up on us.

We’ve all seen the symptoms. Projects slowly and inexorably get totally
out of hand. Most software disasters start out too small to notice, and
most project overruns happen a day at a time. Systems drift from their
specifications feature by feature, while patch after patch gets added to
a piece of code until there’s nothing of the original left. It’s often the
accumulation of small things that breaks morale and teams.

TIP 6

Remember the Big Picture

We’ve never tried this—honest. But they say that if you take a frog and
drop it into boiling water, it will jump straight back out again. However,
if you place the frog in a pan of cold water, then gradually heat it, the
frog won’t notice the slow increase in temperature and will stay put
until cooked.

1. While doing this, you may be comforted by the line attributed to Rear Admiral Dr.
Grace Hopper: “It’s easier to ask forgiveness than it is to get permission.”

Prepared exclusively for Zach

 G OOD -E NOUGH S OFTWARE 9

Note that the frog’s problem is different from the broken windows issue
discussed in Section 2. In the Broken Window Theory, people lose the
will to fight entropy because they perceive that no one else cares. The
frog just doesn’t notice the change.

Don’t be like the frog. Keep an eye on the big picture. Constantly review
what’s happening around you, not just what you personally are doing.

Related sections include:

Software Entropy, page 4
Programming by Coincidence, page 172
Refactoring, page 184
The Requirements Pit, page 202
Pragmatic Teams, page 224

Challenges
While reviewing a draft of this book, John Lakos raised the following is-
sue: The soldiers progressively deceive the villagers, but the change they
catalyze does them all good. However, by progressively deceiving the frog,
you’re doing it harm. Can you determine whether you’re making stone
soup or frog soup when you try to catalyze change? Is the decision subjec-
tive or objective?

4 Good-Enough Software
Striving to better, oft we mar what’s well.
King Lear 1.4

There’s an old(ish) joke about a U.S. company that places an order for
100,000 integrated circuits with a Japanese manufacturer. Part of the
specification was the defect rate: one chip in 10,000. A few weeks later
the order arrived: one large box containing thousands of ICs, and a
small one containing just ten. Attached to the small box was a label
that read: “These are the faulty ones.”

Prepared exclusively for Zach

 10 C HAPTER 1 A P RAGMATIC P HILOSOPHY

If only we really had this kind of control over quality. But the real
world just won’t let us produce much that’s truly perfect, particularly
not bug-free software. Time, technology, and temperament all conspire
against us.

However, this doesn’t have to be frustrating. As Ed Yourdon described

in an article in IEEE Software [You95], you can discipline yourself to
write software that’s good enough—good enough for your users, for fu-
ture maintainers, for your own peace of mind. You’ll find that you are
more productive and your users are happier. And you may well find
that your programs are actually better for their shorter incubation.

Before we go any further, we need to qualify what we’re about to say.

The phrase “good enough” does not imply sloppy or poorly produced
code. All systems must meet their users’ requirements to be success-
ful. We are simply advocating that users be given an opportunity to
participate in the process of deciding when what you’ve produced is
good enough.

Involve Your Users in the Trade-Off

Normally you’re writing software for other people. Often you’ll remem-
ber to get requirements from them.2 But how often do you ask them
how good they want their software to be? Sometimes there’ll be no
choice. If you’re working on pacemakers, the space shuttle, or a low-
level library that will be widely disseminated, the requirements will
be more stringent and your options more limited. However, if you’re
working on a brand new product, you’ll have different constraints. The
marketing people will have promises to keep, the eventual end users
may have made plans based on a delivery schedule, and your company
will certainly have cash-flow constraints. It would be unprofessional to
ignore these users’ requirements simply to add new features to the pro-
gram, or to polish up the code just one more time. We’re not advocating
panic: it is equally unprofessional to promise impossible time scales
and to cut basic engineering corners to meet a deadline.

2. That was supposed to be a joke!

Prepared exclusively for Zach

 G OOD -E NOUGH S OFTWARE 11

The scope and quality of the system you produce should be specified
as part of that system’s requirements.

TIP 7

Make Quality a Requirements Issue

Often you’ll be in situations where trade-offs are involved. Surprisingly,

many users would rather use software with some rough edges today
than wait a year for the multimedia version. Many IT departments with
tight budgets would agree. Great software today is often preferable to
perfect software tomorrow. If you give your users something to play with
early, their feedback will often lead you to a better eventual solution (see
Tracer Bullets, page 48).

Know When to Stop

In some ways, programming is like painting. You start with a blank
canvas and certain basic raw materials. You use a combination of sci-
ence, art, and craft to determine what to do with them. You sketch
out an overall shape, paint the underlying environment, then fill in the
details. You constantly step back with a critical eye to view what you’ve
done. Every now and then you’ll throw a canvas away and start again.

But artists will tell you that all the hard work is ruined if you don’t
know when to stop. If you add layer upon layer, detail over detail, the
painting becomes lost in the paint.

Don’t spoil a perfectly good program by overembellishment and over-

refinement. Move on, and let your code stand in its own right for a
while. It may not be perfect. Don’t worry: it could never be perfect. (In
Chapter 6, page 171, we’ll discuss philosophies for developing code in
an imperfect world.)

Related sections include:

Tracer Bullets, page 48
The Requirements Pit, page 202
Pragmatic Teams, page 224
Great Expectations, page 255

Prepared exclusively for Zach

 12 C HAPTER 1 A P RAGMATIC P HILOSOPHY

Challenges
Look at the manufacturers of the software tools and operating systems that
you use. Can you find any evidence that these companies are comfortable
shipping software they know is not perfect? As a user, would you rather
(1) wait for them to get all the bugs out, (2) have complex software and
accept some bugs, or (3) opt for simpler software with fewer defects?

Consider the effect of modularization on the delivery of software. Will it

take more or less time to get a monolithic block of software to the required
quality compared with a system designed in modules? Can you find com-
mercial examples?

5 Your Knowledge Portfolio

An investment in knowledge always pays the best interest.
Benjamin Franklin

Ah, good old Ben Franklin—never at a loss for a pithy homily. Why, if we
could just be early to bed and early to rise, we’d be great programmers—
right? The early bird might get the worm, but what happens to the early
worm?

In this case, though, Ben really hit the nail on the head. Your knowledge
and experience are your most important professional assets.

Unfortunately, they’re expiring assets.3 Your knowledge becomes out of

date as new techniques, languages, and environments are developed.
Changing market forces may render your experience obsolete or irrele-
vant. Given the speed at which Web-years fly by, this can happen pretty
quickly.

As the value of your knowledge declines, so does your value to your

company or client. We want to prevent this from ever happening.

3. An expiring asset is something whose value diminishes over time. Examples include
a warehouse full of bananas and a ticket to a ball game.

Prepared exclusively for Zach

 Y OUR K NOWLEDGE P OR TFOLIO 13

Your Knowledge Portfolio

We like to think of all the facts programmers know about computing,
the application domains they work in, and all their experience as their
Knowledge Portfolios. Managing a knowledge portfolio is very similar to
managing a financial portfolio:

1. Serious investors invest regularly—as a habit.

2. Diversification is the key to long-term success.

3. Smart investors balance their portfolios between conservative and

high-risk, high-reward investments.

4. Investors try to buy low and sell high for maximum return.

5. Portfolios should be reviewed and rebalanced periodically.

To be successful in your career, you must manage your knowledge port-

folio using these same guidelines.

Building Your Portfolio

Invest regularly. Just as in financial investing, you must invest in
your knowledge portfolio regularly. Even if it’s just a small amount,
the habit itself is as important as the sums. A few sample goals are
listed in the next section.

Diversify. The more different things you know, the more valuable
you are. As a baseline, you need to know the ins and outs of the
particular technology you are working with currently. But don’t
stop there. The face of computing changes rapidly—hot technology
today may well be close to useless (or at least not in demand) to-
morrow. The more technologies you are comfortable with, the better
you will be able to adjust to change.

Manage risk. Technology exists along a spectrum from risky,

potentially high-reward to low-risk, low-reward standards. It’s not
a good idea to invest all of your money in high-risk stocks that
might collapse suddenly, nor should you invest all of it conserva-
tively and miss out on possible opportunities. Don’t put all your
technical eggs in one basket.

Prepared exclusively for Zach

 14 C HAPTER 1 A P RAGMATIC P HILOSOPHY

Buy low, sell high. Learning an emerging technology before it be-

comes popular can be just as hard as finding an undervalued stock,
but the payoff can be just as rewarding. Learning Java when it first
came out may have been risky, but it paid off handsomely for the
early adopters who are now at the top of that field.

Review and rebalance. This is a very dynamic industry. That hot

technology you started investigating last month might be stone cold
by now. Maybe you need to brush up on that database technology
that you haven’t used in a while. Or perhaps you could be bet-
ter positioned for that new job opening if you tried out that other
language. . . .

Of all these guidelines, the most important one is the simplest to do:

TIP 8

Invest Regularly in Your Knowledge Portfolio

Goals
Now that you have some guidelines on what and when to add to your
knowledge portfolio, what’s the best way to go about acquiring intellec-
tual capital with which to fund your portfolio? Here are a few sugges-
tions.

Learn at least one new language every year. Different languages

solve the same problems in different ways. By learning several dif-
ferent approaches, you can help broaden your thinking and avoid
getting stuck in a rut. Additionally, learning many languages is far
easier now, thanks to the wealth of freely available software on the
Internet (see page 267).

Read a technical book each quarter. Bookstores are full of techni-

cal books on interesting topics related to your current project. Once
you’re in the habit, read a book a month. After you’ve mastered the
technologies you’re currently using, branch out and study some
that don’t relate to your project.

Read nontechnical books, too. It is important to remember that

computers are used by people—people whose needs you are trying
to satisfy. Don’t forget the human side of the equation.

Prepared exclusively for Zach

 Y OUR K NOWLEDGE P OR TFOLIO 15

Take classes. Look for interesting courses at your local commu-

nity college or university, or perhaps at the next trade show that
comes to town.

Participate in local user groups. Don’t just go and listen, but

actively participate. Isolation can be deadly to your career; find out
what people are working on outside of your company.

Experiment with different environments. If you’ve worked only in

Windows, play with Unix at home (the freely available Linux is per-
fect for this). If you’ve used only makefiles and an editor, try an
IDE, and vice versa.

Stay current. Subscribe to trade magazines and other journals

(see page 262 for recommendations). Choose some that cover tech-
nology different from that of your current project.

Get wired. Want to know the ins and outs of a new language or
other technology? Newsgroups are a great way to find out what
experiences other people are having with it, the particular jargon
they use, and so on. Surf the Web for papers, commercial sites,
and any other sources of information you can find.

It’s important to continue investing. Once you feel comfortable with

some new language or bit of technology, move on. Learn another one.

It doesn’t matter whether you ever use any of these technologies on a

project, or even whether you put them on your resume. The process of
learning will expand your thinking, opening you to new possibilities and
new ways of doing things. The cross-pollination of ideas is important;
try to apply the lessons you’ve learned to your current project. Even if
your project doesn’t use that technology, perhaps you can borrow some
ideas. Get familiar with object orientation, for instance, and you’ll write
plain C programs differently.

Opportunities for Learning

So you’re reading voraciously, you’re on top of all the latest breaking
developments in your field (not an easy thing to do), and somebody
asks you a question. You don’t have the faintest idea what the answer
is, and freely admit as much.

Prepared exclusively for Zach

 16 C HAPTER 1 A P RAGMATIC P HILOSOPHY

Don’t let it stop there. Take it as a personal challenge to find the answer.
Ask a guru. (If you don’t have a guru in your office, you should be able
to find one on the Internet: see the box on on the facing page.) Search
the Web. Go to the library.4

If you can’t find the answer yourself, find out who can. Don’t let it rest.
Talking to other people will help build your personal network, and you
may surprise yourself by finding solutions to other, unrelated problems
along the way. And that old portfolio just keeps getting bigger. . . .

All of this reading and researching takes time, and time is already in
short supply. So you need to plan ahead. Always have something to
read in an otherwise dead moment. Time spent waiting for doctors and
dentists can be a great opportunity to catch up on your reading—but be
sure to bring your own magazine with you, or you might find yourself
thumbing through a dog-eared 1973 article about Papua New Guinea.

Critical Thinking
The last important point is to think critically about what you read
and hear. You need to ensure that the knowledge in your portfolio is
accurate and unswayed by either vendor or media hype. Beware of the
zealots who insist that their dogma provides the only answer—it may
or may not be applicable to you and your project.

Never underestimate the power of commercialism. Just because a Web

search engine lists a hit first doesn’t mean that it’s the best match; the
content provider can pay to get top billing. Just because a bookstore
features a book prominently doesn’t mean it’s a good book, or even
popular; they may have been paid to place it there.

TIP 9

Critically Analyze What You Read and Hear

Unfortunately, there are very few simple answers anymore. But with
your extensive portfolio, and by applying some critical analysis to the

4. In this era of the Web, many people seem to have forgotten about real live libraries
filled with research material and staff.

Prepared exclusively for Zach

 Y OUR K NOWLEDGE P OR TFOLIO 17

Care and Cultivation of Gurus

With the global adoption of the Internet, gurus suddenly are as close
as your Enter key. So, how do you find one, and how do you get one
to talk with you?
We find there are some simple tricks.
Know exactly what you want to ask, and be as specific as you
can be.
Frame your question carefully and politely. Remember that you’re
asking a favor; don’t seem to be demanding an answer.
Once you’ve framed your question, stop and look again for the
answer. Pick out some keywords and search the Web. Look for
appropriate FAQs (lists of frequently asked questions with an-
swers).
Decide if you want to ask publicly or privately. Usenet news-
groups are wonderful meeting places for experts on just about
any topic, but some people are wary of these groups’ public
nature. Alternatively, you can always e-mail your guru directly.
Either way, use a meaningful subject line. (“Need Help!!!”
doesn’t cut it.)
Sit back and be patient. People are busy, and it may take days to
get a specific answer.
Finally, please be sure to thank anyone who responds to you. And if
you see people asking questions you can answer, play your part and
participate.

torrent of technical publications you will read, you can understand the
complex answers.

Challenges
Start learning a new language this week. Always programmed in C++? Try
Smalltalk [URL 13] or Squeak [URL 14]. Doing Java? Try Eiffel [URL 10]
or TOM [URL 15]. See page 267 for sources of other free compilers and
environments.

Start reading a new book (but finish this one first!). If you are doing very
detailed implementation and coding, read a book on design and architec-
ture. If you are doing high-level design, read a book on coding techniques.

Prepared exclusively for Zach

 18 C HAPTER 1 A P RAGMATIC P HILOSOPHY

Get out and talk technology with people who aren’t involved in your cur-
rent project, or who don’t work for the same company. Network in your
company cafeteria, or maybe seek out fellow enthusiasts at a local user’s
group meeting.

6 Communicate!
I believe that it is better to be looked over than it is to be overlooked.
Mae West, Belle of the Nineties, 1934

Maybe we can learn a lesson from Ms. West. It’s not just what you’ve
got, but also how you package it. Having the best ideas, the finest code,
or the most pragmatic thinking is ultimately sterile unless you can com-
municate with other people. A good idea is an orphan without effective
communication.

As developers, we have to communicate on many levels. We spend

hours in meetings, listening and talking. We work with end users,
trying to understand their needs. We write code, which communicates
our intentions to a machine and documents our thinking for future
generations of developers. We write proposals and memos requesting
and justifying resources, reporting our status, and suggesting new
approaches. And we work daily within our teams to advocate our ideas,
modify existing practices, and suggest new ones. A large part of our day
is spent communicating, so we need to do it well.

We’ve put together a list of ideas that we find useful.

Know What You Want to Say

Probably the most difficult part of the more formal styles of commu-
nication used in business is working out exactly what it is you want
to say. Fiction writers plot out their books in detail before they start,
but people writing technical documents are often happy to sit down at
a keyboard, enter “1. Introduction,” and start typing whatever comes
into their heads next.

Plan what you want to say. Write an outline. Then ask yourself, “Does
this get across whatever I’m trying to say?” Refine it until it does.

Prepared exclusively for Zach

 C OMMUNICATE ! 19

This approach is not just applicable to writing documents. When you’re

faced with an important meeting or a phone call with a major client,
jot down the ideas you want to communicate, and plan a couple of
strategies for getting them across.

Know Your Audience

You’re communicating only if you’re conveying information. To do that,
you need to understand the needs, interests, and capabilities of your
audience. We’ve all sat in meetings where a development geek glazes
over the eyes of the vice president of marketing with a long monologue
on the merits of some arcane technology. This isn’t communicating: it’s
just talking, and it’s annoying.5

Form a strong mental picture of your audience. The acrostic WISDOM,

shown in Figure 1.1 on the following page, may help.

Say you want to suggest a Web-based system to allow your end users
to submit bug reports. You can present this system in many differ-
ent ways, depending on your audience. End users will appreciate that
they can submit bug reports 24 hours a day without waiting on the
phone. Your marketing department will be able to use this fact to boost
sales. Managers in the support department will have two reasons to
be happy: fewer staff will be needed, and problem reporting will be
automated. Finally, developers may enjoy getting experience with Web-
based client-server technologies and a new database engine. By making
the appropriate pitch to each group, you’ll get them all excited about
your project.

Choose Your Moment

It’s six o’clock on Friday afternoon, following a week when the auditors
have been in. Your boss’s youngest is in the hospital, it’s pouring rain
outside, and the commute home is guaranteed to be a nightmare. This
probably isn’t a good time to ask her for a memory upgrade for your PC.

As part of understanding what your audience needs to hear, you need

to work out what their priorities are. Catch a manager who’s just been
given a hard time by her boss because some source code got lost, and

5. The word annoy comes from the Old French enui, which also means “to bore.”

Prepared exclusively for Zach

 20 C HAPTER 1 A P RAGMATIC P HILOSOPHY

Figure 1.1. The WISDOM acrostic—understanding an audience

What do you want them to learn?

What is their interest in what you’ve got to say?
How sophisticated are they?
How much detail do they want?
Whom do you want to own the information?
How can you motivate them to listen to you?

you’ll have a more receptive listener to your ideas on source code repos-
itories. Make what you’re saying relevant in time, as well as in content.
Sometimes all it takes is the simple question “Is this a good time to talk
about...?”

Choose a Style
Adjust the style of your delivery to suit your audience. Some people
want a formal “just the facts” briefing. Others like a long, wide-ranging
chat before getting down to business. When it comes to written docu-
ments, some like to receive large bound reports, while others expect a
simple memo or e-mail. If in doubt, ask.

Remember, however, that you are half of the communication transac-

tion. If someone says they need a paragraph describing something and
you can’t see any way of doing it in less than several pages, tell them
so. Remember, that kind of feedback is a form of communication, too.

Make It Look Good

Your ideas are important. They deserve a good-looking vehicle to convey
them to your audience.

Too many developers (and their managers) concentrate solely on con-

tent when producing written documents. We think this is a mistake.
Any chef will tell you that you can slave in the kitchen for hours only
to ruin your efforts with poor presentation.

There is no excuse today for producing poor-looking printed docu-

ments. Modern word processors (along with layout systems such as
LATEX and troff) can produce stunning output. You need to learn just a
few basic commands. If your word processor supports style sheets, use

Prepared exclusively for Zach

 C OMMUNICATE ! 21

them. (Your company may already have defined style sheets that you
can use.) Learn how to set page headers and footers. Look at the sam-
ple documents included with your package to get ideas on style and
layout. Check the spelling, first automatically and then by hand. After
awl, their are spelling miss steaks that the chequer can knot ketch.

Involve Your Audience

We often find that the documents we produce end up being less im-
portant than the process we go through to produce them. If possible,
involve your readers with early drafts of your document. Get their feed-
back, and pick their brains. You’ll build a good working relationship,
and you’ll probably produce a better document in the process.

Be a Listener
There’s one technique that you must use if you want people to listen
to you: listen to them. Even if this is a situation where you have all the
information, even if this is a formal meeting with you standing in front
of 20 suits—if you don’t listen to them, they won’t listen to you.

Encourage people to talk by asking questions, or have them summarize

what you tell them. Turn the meeting into a dialog, and you’ll make
your point more effectively. Who knows, you might even learn some-
thing.

Get Back to People

If you ask someone a question, you feel they’re impolite if they don’t
respond. But how often do you fail to get back to people when they
send you an e-mail or a memo asking for information or requesting
some action? In the rush of everyday life, it’s easy to forget. Always
respond to e-mails and voice mails, even if the response is simply “I’ll
get back to you later.” Keeping people informed makes them far more
forgiving of the occasional slip, and makes them feel that you haven’t
forgotten them.

TIP 10

It’s Both What You Say and the Way You Say It

Unless you work in a vacuum, you need to be able to communicate. The

more effective that communication, the more influential you become.

Prepared exclusively for Zach

 22 C HAPTER 1 A P RAGMATIC P HILOSOPHY

E-Mail Communication
Everything we’ve said about communicating in writing applies equally
to electronic mail. E-mail has evolved to the point where it is a main-
stay of intra- and intercorporate communications. E-mail is used to
discuss contracts, to settle disputes, and as evidence in court. But
for some reason, people who would never send out a shabby paper
document are happy to fling nasty-looking e-mail around the world.
Our e-mail tips are simple:
Proofread before you hit SEND .

Check the spelling.

Keep the format simple. Some people read e-mail using propor-
tional fonts, so the ASCII art pictures you laboriously created will
look to them like hen-scratchings.
Use rich-text or HTML formatted mail only if you know that all your
recipients can read it. Plain text is universal.
Try to keep quoting to a minimum. No one likes to receive back
their own 100-line e-mail with “I agree” tacked on.
If you’re quoting other people’s e-mail, be sure to attribute it, and
quote it inline (rather than as an attachment).
Don’t flame unless you want it to come back and haunt you later.
Check your list of recipients before sending. A recent Wall Street
Journal article described an employee who took to distributing
criticisms of his boss over departmental e-mail, without realizing
that his boss was included on the distribution list.
Archive and organize your e-mail—both the important stuff you
receive and the mail you send.
As various Microsoft and Netscape employees discovered during the
1999 Department of Justice investigation, e-mail is forever. Try to give
the same attention and care to e-mail as you would to any written
memo or report.

Prepared exclusively for Zach

 C OMMUNICATE ! 23

Summary
Know what you want to say.
Know your audience.
Choose your moment.
Choose a style.
Make it look good.
Involve your audience.
Be a listener.
Get back to people.

Related sections include:

Prototypes and Post-it Notes, page 53
Pragmatic Teams, page 224

Challenges
There are several good books that contain sections on communications
within development teams [Bro95, McC95, DL99]. Make it a point to try
to read all three over the next 18 months. In addition, the book Dinosaur
Brains [Ber96] discusses the emotional baggage we all bring to the work
environment.

The next time you have to give a presentation, or write a memo advocating
some position, try working through the WISDOM acrostic on page 20 before
you start. See if it helps you understand how to position what you say. If
appropriate, talk to your audience afterward and see how accurate your
assessment of their needs was.

Prepared exclusively for Zach

 This page intentionally left blank

Prepared exclusively for Zach

 Chapter 2

A Pragmatic Approach
There are certain tips and tricks that apply at all levels of software
development, ideas that are almost axiomatic, and processes that are
virtually universal. However, these approaches are rarely documented
as such; you’ll mostly find them written down as odd sentences in dis-
cussions of design, project management, or coding.

In this chapter we’ll bring these ideas and processes together. The first
two sections, The Evils of Duplication and Orthogonality, are closely
related. The first warns you not to duplicate knowledge throughout
your systems, the second not to split any one piece of knowledge across
multiple system components.

As the pace of change increases, it becomes harder and harder to keep

our applications relevant. In Reversibility, we’ll look at some techniques
that help insulate your projects from their changing environment.

The next two sections are also related. In Tracer Bullets, we talk about
a style of development that allows you to gather requirements, test
designs, and implement code at the same time. If this sounds too good
to be true, it is: tracer bullet developments are not always applicable.
When they’re not, Prototypes and Post-it Notes shows you how to use
prototyping to test architectures, algorithms, interfaces, and ideas.

As computer science slowly matures, designers are producing increas-

ingly higher-level languages. While the compiler that accepts “make it
so” hasn’t yet been invented, in Domain Languages we present some
more modest suggestions that you can implement for yourself.

25

Prepared exclusively for Zach

 26 C HAPTER 2 A P RAGMATIC A PPROACH

Finally, we all work in a world of limited time and resources. You can
survive both of these scarcities better (and keep your bosses happier) if
you get good at working out how long things will take, which we cover
in Estimating.

By keeping these fundamental principles in mind during development,

you can write code that’s better, faster, and stronger. You can even
make it look easy.

7 The Evils of Duplication

Giving a computer two contradictory pieces of knowledge was Captain
James T. Kirk’s preferred way of disabling a marauding artificial intel-
ligence. Unfortunately, the same principle can be effective in bringing
down your code.

As programmers, we collect, organize, maintain, and harness knowl-

edge. We document knowledge in specifications, we make it come alive
in running code, and we use it to provide the checks needed during
testing.

Unfortunately, knowledge isn’t stable. It changes—often rapidly. Your

understanding of a requirement may change following a meeting with
the client. The government changes a regulation and some business
logic gets outdated. Tests may show that the chosen algorithm won’t
work. All this instability means that we spend a large part of our time
in maintenance mode, reorganizing and reexpressing the knowledge in
our systems.

Most people assume that maintenance begins when an application is

released, that maintenance means fixing bugs and enhancing features.
We think these people are wrong. Programmers are constantly in main-
tenance mode. Our understanding changes day by day. New require-
ments arrive as we’re designing or coding. Perhaps the environment
changes. Whatever the reason, maintenance is not a discrete activity,
but a routine part of the entire development process.

Prepared exclusively for Zach

 T HE E VILS OF D UPLICATION 27

When we perform maintenance, we have to find and change the rep-

resentations of things—those capsules of knowledge embedded in the
application. The problem is that it’s easy to duplicate knowledge in the
specifications, processes, and programs that we develop, and when we
do so, we invite a maintenance nightmare—one that starts well before
the application ships.

We feel that the only way to develop software reliably, and to make our
developments easier to understand and maintain, is to follow what we
call the DRY principle:

E VERY PIECE OF KNOWLEDGE MUST HAVE A SINGLE , UNAMBIGU -

OUS , AUTHORITATIVE REPRESENTATION WITHIN A SYSTEM .

Why do we call it DRY ?

TIP 11

DRY —D on’t R epeat Y ourself

The alternative is to have the same thing expressed in two or more

places. If you change one, you have to remember to change the others,
or, like the alien computers, your program will be brought to its knees
by a contradiction. It isn’t a question of whether you’ll remember: it’s a
question of when you’ll forget.

You’ll find the DRY principle popping up time and time again through-
out this book, often in contexts that have nothing to do with coding.
We feel that it is one of the most important tools in the Pragmatic Pro-
grammer’s tool box.

In this section we’ll outline the problems of duplication and suggest

general strategies for dealing with it.

How Does Duplication Arise?

Most of the duplication we see falls into one of the following categories:

Imposed duplication. Developers feel they have no choice—the

environment seems to require duplication.

Inadvertent duplication. Developers don’t realize that they are

duplicating information.

Prepared exclusively for Zach

 28 C HAPTER 2 A P RAGMATIC A PPROACH

Impatient duplication. Developers get lazy and duplicate because

it seems easier.

Interdeveloper duplication. Multiple people on a team (or on dif-

ferent teams) duplicate a piece of information.

Let’s look at these four i’s of duplication in more detail.

Imposed Duplication
Sometimes, duplication seems to be forced on us. Project standards
may require documents that contain duplicated information, or docu-
ments that duplicate information in the code. Multiple target platforms
each require their own programming languages, libraries, and devel-
opment environments, which makes us duplicate shared definitions
and procedures. Programming languages themselves require certain
structures that duplicate information. We have all worked in situations
where we felt powerless to avoid duplication. And yet often there are
ways of keeping each piece of knowledge in one place, honoring the
DRY principle, and making our lives easier at the same time. Here are
some techniques:

Multiple representations of information. At the coding level, we

often need to have the same information represented in different forms.
Maybe we’re writing a client-server application, using different lan-
guages on the client and server, and need to represent some shared
structure on both. Perhaps we need a class whose attributes mirror the
schema of a database table. Maybe you’re writing a book and want to
include excerpts of programs that you also will compile and test.

With a bit of ingenuity you can normally remove the need for dupli-
cation. Often the answer is to write a simple filter or code generator.
Structures in multiple languages can be built from a common metadata
representation using a simple code generator each time the software is
built (an example of this is shown in Figure 3.4, page 106). Class defini-
tions can be generated automatically from the online database schema,
or from the metadata used to build the schema in the first place. The
code extracts in this book are inserted by a preprocessor each time we
format the text. The trick is to make the process active: this cannot be
a one-time conversion, or we’re back in a position of duplicating data.

Prepared exclusively for Zach

 T HE E VILS OF D UPLICATION 29

Documentation in code. Programmers are taught to comment their

code: good code has lots of comments. Unfortunately, they are never
taught why code needs comments: bad code requires lots of comments.

The DRY principle tells us to keep the low-level knowledge in the code,
where it belongs, and reserve the comments for other, high-level expla-
nations. Otherwise, we’re duplicating knowledge, and every change
means changing both the code and the comments. The comments will
inevitably become out of date, and untrustworthy comments are worse
than no comments at all. (See It’s All Writing, page 248, for more infor-
mation on comments.)

Documentation and code. You write documentation, then you write

code. Something changes, and you amend the documentation and up-
date the code. The documentation and code both contain representa-
tions of the same knowledge. And we all know that in the heat of the
moment, with deadlines looming and important clients clamoring, we
tend to defer the updating of documentation.

Dave once worked on an international telex switch. Quite understand-

ably, the client demanded an exhaustive test specification and required
that the software pass all tests on each delivery. To ensure that the
tests accurately reflected the specification, the team generated them
programmatically from the document itself. When the client amended
their specification, the test suite changed automatically. Once the team
convinced the client that the procedure was sound, generating accep-
tance tests typically took only a few seconds.

Language issues. Many languages impose considerable duplication

in the source. Often this comes about when the language separates
a module’s interface from its implementation. C and C++ have header
files that duplicate the names and type information of exported vari-
ables, functions, and (for C++) classes. Object Pascal even duplicates
this information in the same file. If you are using remote procedure
calls or CORBA [URL 29], you’ll duplicate interface information between
the interface specification and the code that implements it.

There is no easy technique for overcoming the requirements of a lan-

guage. While some development environments hide the need for header
files by generating them automatically, and Object Pascal allows you to
abbreviate repeated function declarations, you are generally stuck with

Prepared exclusively for Zach

 30 C HAPTER 2 A P RAGMATIC A PPROACH

what you’re given. At least with most language-based issues, a header

file that disagrees with the implementation will generate some form of
compilation or linkage error. You can still get things wrong, but at least
you’ll be told about it fairly early on.

Think also about comments in header and implementation files. There

is absolutely no point in duplicating a function or class header com-
ment between the two files. Use the header files to document interface
issues, and the implementation files to document the nitty-gritty details
that users of your code don’t need to know.

Inadvertent Duplication
Sometimes, duplication comes about as the result of mistakes in the
design.

Let’s look at an example from the distribution industry. Say our anal-
ysis reveals that, among other attributes, a truck has a type, a license
number, and a driver. Similarly, a delivery route is a combination of a
route, a truck, and a driver. We code up some classes based on this
understanding.

But what happens when Sally calls in sick and we have to change
drivers? Both Truck and DeliveryRoute contain a driver. Which one
do we change? Clearly this duplication is bad. Normalize it according
to the underlying business model—does a truck really have a driver
as part of its underlying attribute set? Does a route? Or maybe there
needs to be a third object that knits together a driver, a truck, and a
route. Whatever the eventual solution, avoid this kind of unnormalized
data.

There is a slightly less obvious kind of unnormalized data that occurs

when we have multiple data elements that are mutually dependent.
Let’s look at a class representing a line:
class Line {
public:
Point start;
Point end;
double length;
};

At first sight, this class might appear reasonable. A line clearly has a
start and end, and will always have a length (even if it’s zero). But we

Prepared exclusively for Zach

 T HE E VILS OF D UPLICATION 31

have duplication. The length is defined by the start and end points:
change one of the points and the length changes. It’s better to make
the length a calculated field:

class Line {
public:
Point start;
Point end;
double length() { return start.distanceTo(end); }
};

Later on in the development process, you may choose to violate the DRY
principle for performance reasons. Frequently this occurs when you
need to cache data to avoid repeating expensive operations. The trick is
to localize the impact. The violation is not exposed to the outside world:
only the methods within the class have to worry about keeping things
straight.

class Line {
private:
bool changed;
double length;
Point start;
Point end;
public:
void setStart(Point p) { start = p; changed = true; }
void setEnd(Point p) { end = p; changed = true; }
Point getStart(void) { return start; }
Point getEnd(void) { return end; }
double getLength() {
if (changed) {
length = start.distanceTo(end);
changed = false;
}
return length;
}
};

This example also illustrates an important issue for object-oriented lan-

guages such as Java and C++. Where possible, always use accessor
functions to read and write the attributes of objects.1 It will make it
easier to add functionality, such as caching, in the future.

1. The use of accessor functions ties in with Meyer’s Uniform Access principle [Mey97b],
which states that “All services offered by a module should be available through a uni-
form notation, which does not betray whether they are implemented through storage or
through computation.”

Prepared exclusively for Zach

 32 C HAPTER 2 A P RAGMATIC A PPROACH

Impatient Duplication
Every project has time pressures—forces that can drive the best of us
to take shortcuts. Need a routine similar to one you’ve written? You’ll
be tempted to copy the original and make a few changes. Need a value
to represent the maximum number of points? If I change the header
file, the whole project will get rebuilt. Maybe I should just use a literal
number here; and here; and here. Need a class like one in the Java
runtime? The source is available, so why not just copy it and make the
changes you need (license provisions notwithstanding)?

If you feel this temptation, remember the hackneyed aphorism “short

cuts make for long delays.” You may well save some seconds now, but
at the potential loss of hours later. Think about the issues surround-
ing the Y2K fiasco. Many were caused by the laziness of developers
not parameterizing the size of date fields or implementing centralized
libraries of date services.

Impatient duplication is an easy form to detect and handle, but it takes

discipline and a willingness to spend time up front to save pain later.

Interdeveloper Duplication
On the other hand, perhaps the hardest type of duplication to detect
and handle occurs between different developers on a project. Entire
sets of functionality may be inadvertently duplicated, and that duplica-
tion could go undetected for years, leading to maintenance problems.
We heard firsthand of a U.S. state whose governmental computer sys-
tems were surveyed for Y2K compliance. The audit turned up more
than 10,000 programs, each containing its own version of Social Secu-
rity number validation.

At a high level, deal with the problem by having a clear design, a strong
technical project leader (see page 228 in Pragmatic Teams), and a well-
understood division of responsibilities within the design. However, at
the module level, the problem is more insidious. Commonly needed
functionality or data that doesn’t fall into an obvious area of responsi-
bility can get implemented many times over.

We feel that the best way to deal with this is to encourage active and
frequent communication between developers. Set up forums to discuss
common problems. (On past projects, we have set up private Usenet

Prepared exclusively for Zach

 T HE E VILS OF D UPLICATION 33

newsgroups to allow developers to exchange ideas and ask questions.

This provides a nonintrusive way of communicating—even across mul-
tiple sites—while retaining a permanent history of everything said.)
Appoint a team member as the project librarian, whose job is to facil-
itate the exchange of knowledge. Have a central place in the source
tree where utility routines and scripts can be deposited. And make a
point of reading other people’s source code and documentation, either
informally or during code reviews. You’re not snooping—you’re learning
from them. And remember, the access is reciprocal—don’t get twisted
about other people poring (pawing?) through your code, either.

TIP 12

Make It Easy to Reuse

What you’re trying to do is foster an environment where it’s easier to

find and reuse existing stuff than to write it yourself. If it isn’t easy,
people won’t do it. And if you fail to reuse, you risk duplicating knowl-
edge.

Related sections include:

Orthogonality, page 34
Text Manipulation, page 99
Code Generators, page 102
Refactoring, page 184
Pragmatic Teams, page 224
Ubiquitous Automation, page 230
It’s All Writing, page 248

Prepared exclusively for Zach

 34 C HAPTER 2 A P RAGMATIC A PPROACH

8 Orthogonality
Orthogonality is a critical concept if you want to produce systems that
are easy to design, build, test, and extend. However, the concept of
orthogonality is rarely taught directly. Often it is an implicit feature of
various other methods and techniques you learn. This is a mistake.
Once you learn to apply the principle of orthogonality directly, you’ll
notice an immediate improvement in the quality of systems you pro-
duce.

What Is Orthogonality?
“Orthogonality” is a term borrowed from geom- move parallel
etry. Two lines are orthogonal if they meet at to X -axis

right angles, such as the axes on a graph. In

vector terms, the two lines are independent. no change
on Y -axis
Move along one of the lines, and your position
projected onto the other doesn’t change.

In computing, the term has come to signify a kind of independence or

decoupling. Two or more things are orthogonal if changes in one do not
affect any of the others. In a well-designed system, the database code
will be orthogonal to the user interface: you can change the interface
without affecting the database, and swap databases without changing
the interface.

Before we look at the benefits of orthogonal systems, let’s first look at a

system that isn’t orthogonal.

A Nonorthogonal System
You’re on a helicopter tour of the Grand Canyon when the pilot, who
made the obvious mistake of eating fish for lunch, suddenly groans and
faints. Fortunately, he left you hovering 100 feet above the ground. You
rationalize that the collective pitch lever 2 controls overall lift, so lower-

2. Helicopters have four basic controls. The cyclic is the stick you hold in your right
hand. Move it, and the helicopter moves in the corresponding direction. Your left hand
holds the collective pitch lever. Pull up on this and you increase the pitch on all the
blades, generating lift. At the end of the pitch lever is the throttle. Finally you have two
foot pedals, which vary the amount of tail rotor thrust and so help turn the helicopter.

Prepared exclusively for Zach

 O R THOGONALITY 35

ing it slightly will start a gentle descent to the ground. However, when
you try it, you discover that life isn’t that simple. The helicopter’s nose
drops, and you start to spiral down to the left. Suddenly you discover
that you’re flying a system where every control input has secondary
effects. Lower the left-hand lever and you need to add compensating
backward movement to the right-hand stick and push the right pedal.
But then each of these changes affects all of the other controls again.
Suddenly you’re juggling an unbelievably complex system, where every
change impacts all the other inputs. Your workload is phenomenal:
your hands and feet are constantly moving, trying to balance all the
interacting forces.

Helicopter controls are decidedly not orthogonal.

Benefits of Orthogonality
As the helicopter example illustrates, nonorthogonal systems are in-
herently more complex to change and control. When components of
any system are highly interdependent, there is no such thing as a local
fix.

TIP 13

Eliminate Effects Between Unrelated Things

We want to design components that are self-contained: independent,

and with a single, well-defined purpose (what Yourdon and Constan-
tine call cohesion [YC86]). When components are isolated from one
another, you know that you can change one without having to worry
about the rest. As long as you don’t change that component’s external
interfaces, you can be comfortable that you won’t cause problems that
ripple through the entire system.

You get two major benefits if you write orthogonal systems: increased
productivity and reduced risk.

Gain Productivity
Changes are localized, so development time and testing time are
reduced. It is easier to write relatively small, self-contained compo-
nents than a single large block of code. Simple components can be

Prepared exclusively for Zach

 36 C HAPTER 2 A P RAGMATIC A PPROACH

designed, coded, unit tested, and then forgotten—there is no need

to keep changing existing code as you add new code.

An orthogonal approach also promotes reuse. If components have

specific, well-defined responsibilities, they can be combined with
new components in ways that were not envisioned by their original
implementors. The more loosely coupled your systems, the easier
they are to reconfigure and reengineer.

There is a fairly subtle gain in productivity when you combine

orthogonal components. Assume that one component does dis-
tinct things and another does things. If they are orthogonal and
you combine them, the result does things. However, if the
two components are not orthogonal, there will be overlap, and the
result will do less. You get more functionality per unit effort by
combining orthogonal components.

Reduce Risk
An orthogonal approach reduces the risks inherent in any development.

Diseased sections of code are isolated. If a module is sick, it is less

likely to spread the symptoms around the rest of the system. It
is also easier to slice it out and transplant in something new and
healthy.

The resulting system is less fragile. Make small changes and fixes to
a particular area, and any problems you generate will be restricted
to that area.

An orthogonal system will probably be better tested, because it will

be easier to design and run tests on its components.

You will not be as tightly tied to a particular vendor, product, or

platform, because the interfaces to these third-party components
will be isolated to smaller parts of the overall development.

Let’s look at some of the ways you can apply the principle of orthogo-
nality to your work.

Project Teams
Have you noticed how some project teams are efficient, with everyone
knowing what to do and contributing fully, while the members of other

Prepared exclusively for Zach

 O R THOGONALITY 37

teams are constantly bickering and don’t seem able to get out of each
other’s way?

Often this is an orthogonality issue. When teams are organized with lots
of overlap, members are confused about responsibilities. Every change
needs a meeting of the entire team, because any one of them might be
affected.

How do you organize teams into groups with well-defined responsibili-

ties and minimal overlap? There’s no simple answer. It depends partly
on the project and your analysis of the areas of potential change. It also
depends on the people you have available. Our preference is to start by
separating infrastructure from application. Each major infrastructure
component (database, communications interface, middleware layer,
and so on) gets its own subteam. Each obvious division of application
functionality is similarly divided. Then we look at the people we have
(or plan to have) and adjust the groupings accordingly.

You can get an informal measure of the orthogonality of a project team’s

structure. Simply see how many people need to be involved in dis-
cussing each change that is requested. The larger the number, the less
orthogonal the group. Clearly, an orthogonal team is more efficient.
(Having said this, we also encourage subteams to communicate con-
stantly with each other.)

Design
Most developers are familiar with the need to design orthogonal sys-
tems, although they may use words such as modular, component-based,
and layered to describe the process. Systems should be composed of
a set of cooperating modules, each of which implements functionality
independent of the others. Sometimes these components are organized
into layers, each providing a level of abstraction. This layered approach
is a powerful way to design orthogonal systems. Because each layer
uses only the abstractions provided by the layers below it, you have
great flexibility in changing underlying implementations without affect-
ing code. Layering also reduces the risk of runaway dependencies be-
tween modules. You’ll often see layering expressed in diagrams such as
Figure 2.1 on the next page.

There is an easy test for orthogonal design. Once you have your com-
ponents mapped out, ask yourself: If I dramatically change the require-

Prepared exclusively for Zach

 38 C HAPTER 2 A P RAGMATIC A PPROACH

Figure 2.1. Typical layer diagram

User Interface

Database Report Business

access engine logic

Application framework

Standard C library

Operating system

ments behind a particular function, how many modules are affected? In

an orthogonal system, the answer should be “one.”3 Moving a button
on a GUI panel should not require a change in the database schema.
Adding context-sensitive help should not change the billing subsystem.

Let’s consider a complex system for monitoring and controlling a heat-

ing plant. The original requirement called for a graphical user interface,
but the requirements were changed to add a voice response system with
touchtone telephone control of the plant. In an orthogonally designed
system, you would need to change only those modules associated with
the user interface to handle this: the underlying logic of controlling the
plant would remain unchanged. In fact, if you structure your system
carefully, you should be able to support both interfaces with the same
underlying code base. It’s Just a View, page 157, talks about writing de-
coupled code using the Model-View-Controller (MVC) paradigm, which
works well in this situation.

3. In reality, this is naive. Unless you are remarkably lucky, most real-world require-
ments changes will affect multiple functions in the system. However, if you analyze the
change in terms of functions, each functional change should still ideally affect just one
module.

Prepared exclusively for Zach

 O R THOGONALITY 39

Also ask yourself how decoupled your design is from changes in the
real world. Are you using a telephone number as a customer identifier?
What happens when the phone company reassigns area codes? Don’t
rely on the properties of things you can’t control.

Toolkits and Libraries

Be careful to preserve the orthogonality of your system as you introduce
third-party toolkits and libraries. Choose your technologies wisely.

We once worked on a project that required that a certain body of Java

code run both locally on a server machine and remotely on a client
machine. The alternatives for distributing classes this way were RMI
and CORBA. If a class were made remotely accessible using RMI, every
call to a remote method in that class could potentially throw an excep-
tion, which means that a naive implementation would require us to
handle the exception whenever our remote classes were used. Using
RMI here is clearly not orthogonal: code calling our remote classes
should not have to be aware of their locations. The alternative—using
CORBA—did not impose that restriction: we could write code that was
unaware of our classes’ locations.

When you bring in a toolkit (or even a library from other members of
your team), ask yourself whether it imposes changes on your code that
shouldn’t be there. If an object persistence scheme is transparent, then
it’s orthogonal. If it requires you to create or access objects in a special
way, then it’s not. Keeping such details isolated from your code has the
added benefit of making it easier to change vendors in the future.

The Enterprise Java Beans (EJB) system is an interesting example of

orthogonality. In most transaction-oriented systems, the application
code has to delineate the start and end of each transaction. With EJB,
this information is expressed declaratively as metadata, outside any
code. The same application code can run in different EJB transaction
environments with no change. This is likely to be a model for many
future environments.

Another interesting twist on orthogonality is Aspect-Oriented Program-

ming (AOP), a research project at Xerox Parc ([KLM 97] and [URL 49]).
AOP lets you express in one place behavior that would otherwise be
distributed throughout your source code. For example, log messages

Prepared exclusively for Zach

 40 C HAPTER 2 A P RAGMATIC A PPROACH

are normally generated by sprinkling explicit calls to some log function

throughout your source. With AOP, you implement logging orthogonally
to the things being logged. Using the Java version of AOP, you could
write a log message when entering any method of class Fred by coding
the aspect:

aspect Trace {
advise * Fred.*(..) {
static before {
Log.write("-> Entering " + thisJoinPoint.methodName);
}
}
}

If you weave this aspect into your code, trace messages will be gen-
erated. If you don’t, you’ll see no messages. Either way, your original
source is unchanged.

Coding
Every time you write code you run the risk of reducing the orthogonality
of your application. Unless you constantly monitor not just what you
are doing but also the larger context of the application, you might un-
intentionally duplicate functionality in some other module, or express
existing knowledge twice.

There are several techniques you can use to maintain orthogonality:

Keep your code decoupled. Write shy code—modules that don’t

reveal anything unnecessary to other modules and that don’t rely
on other modules’ implementations. Try the Law of Demeter [LH89],
which we discuss in Decoupling and the Law of Demeter, page 138.
If you need to change an object’s state, get the object to do it for you.
This way your code remains isolated from the other code’s imple-
mentation and increases the chances that you’ll remain orthogonal.

Avoid global data. Every time your code references global data,
it ties itself into the other components that share that data. Even
globals that you intend only to read can lead to trouble (for exam-
ple, if you suddenly need to change your code to be multithreaded).
In general, your code is easier to understand and maintain if you
explicitly pass any required context into your modules. In object-
oriented applications, context is often passed as parameters to

Prepared exclusively for Zach

 O R THOGONALITY 41

objects’ constructors. In other code, you can create structures con-

taining the context and pass around references to them.

The Singleton pattern in Design Patterns [GHJV95] is a way of

ensuring that there is only one instance of an object of a particular
class. Many people use these singleton objects as a kind of global
variable (particularly in languages, such as Java, that otherwise do
not support the concept of globals). Be careful with singletons—
they can also lead to unnecessary linkage.

Avoid similar functions. Often you’ll come across a set of functions

that all look similar—maybe they share common code at the start
and end, but each has a different central algorithm. Duplicate code
is a symptom of structural problems. Have a look at the Strategy
pattern in Design Patterns for a better implementation.

Get into the habit of being constantly critical of your code. Look for any
opportunities to reorganize it to improve its structure and orthogonal-
ity. This process is called refactoring, and it’s so important that we’ve
dedicated a section to it (see Refactoring, page 184).

Testing
An orthogonally designed and implemented system is easier to test.
Because the interactions between the system’s components are formal-
ized and limited, more of the system testing can be performed at the
individual module level. This is good news, because module level (or
unit) testing is considerably easier to specify and perform than integra-
tion testing. In fact, we suggest that every module have its own unit
test built into its code, and that these tests be performed automatically
as part of the regular build process (see Code That’s Easy to Test, page
189).

Building unit tests is itself an interesting test of orthogonality. What

does it take to build and link a unit test? Do you have to drag in a large
percentage of the rest of the system just to get a test to compile or link?
If so, you’ve found a module that is not well decoupled from the rest of
the system.

Bug fixing is also a good time to assess the orthogonality of the system
as a whole. When you come across a problem, assess how localized

Prepared exclusively for Zach

 42 C HAPTER 2 A P RAGMATIC A PPROACH

the fix is. Do you change just one module, or are the changes scat-
tered throughout the entire system? When you make a change, does it
fix everything, or do other problems mysteriously arise? This is a good
opportunity to bring automation to bear. If you use a source code con-
trol system (and you will after reading Source Code Control, page 86),
tag bug fixes when you check the code back in after testing. You can
then run monthly reports analyzing trends in the number of source
files affected by each bug fix.

Documentation
Perhaps surprisingly, orthogonality also applies to documentation. The
axes are content and presentation. With truly orthogonal documenta-
tion, you should be able to change the appearance dramatically without
changing the content. Modern word processors provide style sheets and
macros that help (see It’s All Writing, page 248).

Living with Orthogonality

Orthogonality is closely related to the DRY principle introduced on page
27. With DRY, you’re looking to minimize duplication within a system,
whereas with orthogonality you reduce the interdependency among the
system’s components. It may be a clumsy word, but if you use the prin-
ciple of orthogonality, combined closely with the DRY principle, you’ll
find that the systems you develop are more flexible, more understand-
able, and easier to debug, test, and maintain.

If you’re brought into a project where people are desperately struggling

to make changes, and where every change seems to cause four other
things to go wrong, remember the nightmare with the helicopter. The
project probably is not orthogonally designed and coded. It’s time to
refactor.

And, if you’re a helicopter pilot, don’t eat the fish .

Related sections include:

The Evils of Duplication, page 26
Source Code Control, page 86
Design by Contract, page 109
Decoupling and the Law of Demeter, page 138

Prepared exclusively for Zach

 O R THOGONALITY 43

Metaprogramming, page 144

It’s Just a View, page 157
Refactoring, page 184
Code That’s Easy to Test, page 189
Evil Wizards, page 198
Pragmatic Teams, page 224
It’s All Writing, page 248

Challenges
Consider the difference between large GUI-oriented tools typically found on
Windows systems and small but combinable command line utilities used
at shell prompts. Which set is more orthogonal, and why? Which is easier
to use for exactly the purpose for which it was intended? Which set is
easier to combine with other tools to meet new challenges?

C++ supports multiple inheritance, and Java allows a class to implement

multiple interfaces. What impact does using these facilities have on orthog-
onality? Is there a difference in impact between using multiple inheritance
and multiple interfaces? Is there a difference between using delegation and
using inheritance?

Exercises
1. You are writing a class called Split, which splits input lines into fields. Answer
on p. 279
Which of the following two Java class signatures is the more orthogonal
design?

class Split1 {
public Split1(InputStreamReader rdr) { ...
public void readNextLine() throws IOException { ...
public int numFields() { ...
public String getField(int fieldNo) { ...
}
class Split2 {
public Split2(String line) { ...
public int numFields() { ...
public String getField(int fieldNo) { ...
}

2. Which will lead to a more orthogonal design: modeless or modal dialog Answer
on p. 279
boxes?

3. How about procedural languages versus object technology? Which results Answer
on p. 280
in a more orthogonal system?

Prepared exclusively for Zach

 44 C HAPTER 2 A P RAGMATIC A PPROACH

9 Reversibility
Nothing is more dangerous than an idea if it’s the only one you have.
Emil-Auguste Chartier, Propos sur la religion, 1938

Engineers prefer simple, single solutions to problems. Math tests that

allow you to proclaim with great confidence that are much more
comfortable than fuzzy, warm essays about the myriad causes of the
French Revolution. Management tends to agree with the engineers: sin-
gle, easy answers fit nicely on spreadsheets and project plans.

If only the real world would cooperate! Unfortunately, while is today,

it may need to be tomorrow, and next week. Nothing is forever—and
if you rely heavily on some fact, you can almost guarantee that it will
change.

There is always more than one way to implement something, and there
is usually more than one vendor available to provide a third-party prod-
uct. If you go into a project hampered by the myopic notion that there is
only one way to do it, you may be in for an unpleasant surprise. Many
project teams have their eyes forcibly opened as the future unfolds:
“But you said we’d use database XYZ! We are 85% done coding the
project, we can’t change now!” the programmer protested. “Sorry, but
our company decided to standardize on database PDQ instead—for all
projects. It’s out of my hands. We’ll just have to recode. All of you will
be working weekends until further notice.”

Changes don’t have to be that Draconian, or even that immediate. But

as time goes by, and your project progresses, you may find yourself
stuck in an untenable position. With every critical decision, the project
team commits to a smaller target—a narrower version of reality that
has fewer options.

By the time many critical decisions have been made, the target becomes
so small that if it moves, or the wind changes direction, or a butterfly in
Tokyo flaps its wings, you miss.4 And you may miss by a huge amount.

4. Take a nonlinear, or chaotic, system and apply a small change to one of its inputs.
You may get a large and often unpredictable result. The clichéd butterfly flapping its
wings in Tokyo could be the start of a chain of events that ends up generating a tornado
in Texas. Does this sound like any projects you know?

Prepared exclusively for Zach

 R EVERSIBILITY 45

The problem is that critical decisions aren’t easily reversible.

Once you decide to use this vendor’s database, or that architectural

pattern, or a certain deployment model (client-server versus stand-
alone, for instance), you are committed to a course of action that cannot
be undone, except at great expense.

Reversibility
Many of the topics in this book are geared to producing flexible, adapt-
able software. By sticking to their recommendations—especially the
DRY principle (page 26), decoupling (page 138), and use of metadata
(page 144)—we don’t have to make as many critical, irreversible de-
cisions. This is a good thing, because we don’t always make the best
decisions the first time around. We commit to a certain technology only
to discover we can’t hire enough people with the necessary skills. We
lock in a certain third-party vendor just before they get bought out
by their competitor. Requirements, users, and hardware change faster
than we can get the software developed.

Suppose you decide, early in the project, to use a relational database

from vendor A. Much later, during performance testing, you discover
that the database is simply too slow, but that the object database
from vendor B is faster. With most conventional projects, you’d be out
of luck. Most of the time, calls to third-party products are entangled
throughout the code. But if you really abstracted the idea of a database
out—to the point where it simply provides persistence as a service—
then you have the flexibility to change horses in midstream.

Similarly, suppose the project begins as a client-server model, but then,

late in the game, marketing decides that servers are too expensive for
some clients, and they want a stand-alone version. How hard would
that be for you? Since it’s just a deployment issue, it shouldn’t take
more than a few days. If it would take longer, then you haven’t thought
about reversibility. The other direction is even more interesting. What
if the stand-alone product you are making needs to be deployed in a
client-server or n-tier fashion? That shouldn’t be hard either.

The mistake lies in assuming that any decision is cast in stone—and in

not preparing for the contingencies that might arise. Instead of carving

Prepared exclusively for Zach

 46 C HAPTER 2 A P RAGMATIC A PPROACH

decisions in stone, think of them more as being written in the sand at

the beach. A big wave can come along and wipe them out at any time.

TIP 14

There Are No Final Decisions

Flexible Architecture
While many people try to keep their code flexible, you also need to think
about maintaining flexibility in the areas of architecture, deployment,
and vendor integration.

Technologies such as CORBA can help insulate portions of a project

from changes in development language or platform. Is the performance
of Java on that platform not up to expectations? Recode the client in
C++, and nothing else needs to change. Is the rules engine in C++
not flexible enough? Switch over to a Smalltalk version. With a CORBA
architecture, you have to take a hit only for the component you are
replacing; the other components shouldn’t be affected.

Are you developing for Unix? Which one? Do you have all of the porta-
bility concerns addressed? Are you developing for a particular version
of Windows? Which one—3.1, 95, 98, NT, CE, or 2000? How hard will
it be to support other versions? If you keep decisions soft and pliable,
it won’t be hard at all. If you have poor encapsulation, high coupling,
and hard-coded logic or parameters in the code, it might be impossible.

Not sure how marketing wants to deploy the system? Think about it up
front and you can support a stand-alone, client-server, or n-tier model
just by changing a configuration file. We’ve written programs that do
just that.

Normally, you can simply hide a third-party product behind a well-

defined, abstract interface. In fact, we’ve always been able to do so
on any project we’ve worked on. But suppose you couldn’t isolate it
that cleanly. What if you had to sprinkle certain statements liberally
throughout the code? Put that requirement in metadata, and use some
automatic mechanism, such as Aspects (see page 39) or Perl, to insert
the necessary statements into the code itself. Whatever mechanism you

Prepared exclusively for Zach

 R EVERSIBILITY 47

use, make it reversible. If something is added automatically, it can be

taken out automatically as well.

No one knows what the future may hold, especially not us! So en-
able your code to rock-n-roll: to “rock on” when it can, to roll with
the punches when it must.

Related sections include:

Decoupling and the Law of Demeter, page 138
Metaprogramming, page 144
It’s Just a View, page 157

Challenges
Time for a little quantum mechanics with Schrödinger’s cat. Suppose you
have a cat in a closed box, along with a radioactive particle. The particle
has exactly a 50% chance of fissioning into two. If it does, the cat will
be killed. If it doesn’t, the cat will be okay. So, is the cat dead or alive?
According to Schrödinger, the correct answer is both. Every time a sub-
nuclear reaction takes place that has two possible outcomes, the universe
is cloned. In one, the event occurred, in the other it didn’t. The cat’s alive
in one universe, dead in another. Only when you open the box do you know
which universe you are in.

No wonder coding for the future is difficult.

But think of code evolution along the same lines as a box full of Schrö-
dinger’s cats: every decision results in a different version of the future.
How many possible futures can your code support? Which ones are more
likely? How hard will it be to support them when the time comes?

Dare you open the box?

Prepared exclusively for Zach

 48 C HAPTER 2 A P RAGMATIC A PPROACH

10 Tracer Bullets
Ready, fire, aim

There are two ways to fire a machine gun in the dark.5 You can find
out exactly where your target is (range, elevation, and azimuth). You
can determine the environmental conditions (temperature, humidity,
air pressure, wind, and so on). You can determine the precise speci-
fications of the cartridges and bullets you are using, and their inter-
actions with the actual gun you are firing. You can then use tables or
a firing computer to calculate the exact bearing and elevation of the
barrel. If everything works exactly as specified, your tables are correct,
and the environment doesn’t change, your bullets should land close to
their target.

Or you could use tracer bullets.

Tracer bullets are loaded at intervals on the ammo belt alongside reg-
ular ammunition. When they’re fired, their phosphorus ignites and
leaves a pyrotechnic trail from the gun to whatever they hit. If the trac-
ers are hitting the target, then so are the regular bullets.

Not surprisingly, tracer bullets are preferred to the labor of calcula-

tion. The feedback is immediate, and because they operate in the same
environment as the real ammunition, external effects are minimized.

The analogy might be violent, but it applies to new projects, particularly

when you’re building something that hasn’t been built before. Like the
gunners, you’re trying to hit a target in the dark. Because your users
have never seen a system like this before, their requirements may be
vague. Because you may be using algorithms, techniques, languages,
or libraries you aren’t familiar with, you face a large number of un-
knowns. And because projects take time to complete, you can pretty
much guarantee the environment you’re working in will change before
you’re done.

The classic response is to specify the system to death. Produce reams

of paper itemizing every requirement, tying down every unknown, and

5. To be pedantic, there are many ways of firing a machine gun in the dark, including
closing your eyes and spraying out bullets. But this is an analogy, and we’re allowed to
take liberties.

Prepared exclusively for Zach

 T RACER B ULLETS 49

constraining the environment. Fire the gun using dead reckoning. One
big calculation up front, then shoot and hope.

Pragmatic Programmers, however, tend to prefer using tracer bullets.

Code That Glows in the Dark

Tracer bullets work because they operate in the same environment and
under the same constraints as the real bullets. They get to the tar-
get fast, so the gunner gets immediate feedback. And from a practical
standpoint they’re a relatively cheap solution.

To get the same effect in code, we’re looking for something that gets us
from a requirement to some aspect of the final system quickly, visibly,
and repeatably.

TIP 15

Use Tracer Bullets to Find the Target

We once undertook a complex client-server database marketing project.

Part of its requirement was the ability to specify and execute temporal
queries. The servers were a range of relational and specialized data-
bases. The client GUI, written in Object Pascal, used a set of C libraries
to provide an interface to the servers. The user’s query was stored on
the server in a Lisp-like notation before being converted to optimized
SQL just prior to execution. There were many unknowns and many
different environments, and no one was too sure how the GUI should
behave.

This was a great opportunity to use tracer code. We developed the

framework for the front end, libraries for representing the queries, and
a structure for converting a stored query into a database-specific query.
Then we put it all together and checked that it worked. For that initial
build, all we could do was submit a query that listed all the rows in a
table, but it proved that the UI could talk to the libraries, the libraries
could serialize and unserialize a query, and the server could generate
SQL from the result. Over the following months we gradually fleshed
out this basic structure, adding new functionality by augmenting each
component of the tracer code in parallel. When the UI added a new
query type, the library grew and the SQL generation was made more
sophisticated.

Prepared exclusively for Zach

 50 C HAPTER 2 A P RAGMATIC A PPROACH

Tracer code is not disposable: you write it for keeps. It contains all the
error checking, structuring, documentation, and self-checking that any
piece of production code has. It simply is not fully functional. However,
once you have achieved an end-to-end connection among the compo-
nents of your system, you can check how close to the target you are,
adjusting if necessary. Once you’re on target, adding functionality is
easy.

Tracer development is consistent with the idea that a project is never

finished: there will always be changes required and functions to add. It
is an incremental approach.

The conventional alternative is a kind of heavy engineering approach:

code is divided into modules, which are coded in a vacuum. Modules
are combined into subassemblies, which are then further combined,
until one day you have a complete application. Only then can the appli-
cation as a whole be presented to the user and tested.

The tracer code approach has many advantages:

Users get to see something working early. If you have success-

fully communicated what you are doing (see Great Expectations,
page 255), your users will know they are seeing something imma-
ture. They won’t be disappointed by a lack of functionality; they’ll
be ecstatic to see some visible progress toward their system. They
also get to contribute as the project progresses, increasing their
buy-in. These same users will likely be the people who’ll tell you
how close to the target each iteration is.

Developers build a structure to work in. The most daunting piece

of paper is the one with nothing written on it. If you have worked
out all the end-to-end interactions of your application, and have
embodied them in code, then your team won’t need to pull as much
out of thin air. This makes everyone more productive, and encour-
ages consistency.

You have an integration platform. As the system is connected

end-to-end, you have an environment to which you can add new
pieces of code once they have been unit-tested. Rather than at-
tempting a big-bang integration, you’ll be integrating every day
(often many times a day). The impact of each new change is more
apparent, and the interactions are more limited, so debugging and
testing are faster and more accurate.

Prepared exclusively for Zach

 T RACER B ULLETS 51

You have something to demonstrate. Project sponsors and top

brass have a tendency to want to see demos at the most inconve-
nient times. With tracer code, you’ll always have something to show
them.

You have a better feel for progress. In a tracer code development,

developers tackle use cases one by one. When one is done, they
move to the next. It is far easier to measure performance and to
demonstrate progress to your user. Because each individual devel-
opment is smaller, you avoid creating those monolithic blocks of
code that are reported as 95% complete week after week.

Tracer Bullets Don’t Always Hit Their Target

Tracer bullets show what you’re hitting. This may not always be the tar-
get. You then adjust your aim until they’re on target. That’s the point.

It’s the same with tracer code. You use the technique in situations
where you’re not 100% certain of where you’re going. You shouldn’t
be surprised if your first couple of attempts miss: the user says “that’s
not what I meant,” or data you need isn’t available when you need it,
or performance problems seem likely. Work out how to change what
you’ve got to bring it nearer the target, and be thankful that you’ve
used a lean development methodology. A small body of code has low
inertia—it is easy and quick to change. You’ll be able to gather feed-
back on your application and generate a new, more accurate version
faster and at less cost than with any other method. And because every
major application component is represented in your tracer code, your
users can be confident that what they’re seeing is based on reality, not
just a paper specification.

Tracer Code versus Prototyping

You might think that this tracer code concept is nothing more than
prototyping under an aggressive name. There is a difference. With a
prototype, you’re aiming to explore specific aspects of the final system.
With a true prototype, you will throw away whatever you lashed to-
gether when trying out the concept, and recode it properly using the
lessons you’ve learned.

For example, say you’re producing an application that helps shippers

determine how to pack odd-sized boxes into containers. Among other

Prepared exclusively for Zach

 52 C HAPTER 2 A P RAGMATIC A PPROACH

problems, the user interface needs to be intuitive and the algorithms

you use to determine optimal packing are very complex.

You could prototype a user interface for your end users in a GUI tool.
You code only enough to make the interface responsive to user actions.
Once they’ve agreed to the layout, you might throw it away and recode
it, this time with the business logic behind it, using the target language.
Similarly, you might want to prototype a number of algorithms that
perform the actual packing. You might code functional tests in a high-
level, forgiving language such as Perl, and code low-level performance
tests in something closer to the machine. In any case, once you’d made
your decision, you’d start again and code the algorithms in their final
environment, interfacing to the real world. This is prototyping, and it is
very useful.

The tracer code approach addresses a different problem. You need to

know how the application as a whole hangs together. You want to show
your users how the interactions will work in practice, and you want to
give your developers an architectural skeleton on which to hang code.
In this case, you might construct a tracer consisting of a trivial imple-
mentation of the container packing algorithm (maybe something like
first-come, first-served) and a simple but working user interface. Once
you have all the components in the application plumbed together, you
have a framework to show your users and your developers. Over time,
you add to this framework with new functionality, completing stubbed
routines. But the framework stays intact, and you know the system
will continue to behave the way it did when your first tracer code was
completed.

The distinction is important enough to warrant repeating. Prototyp-

ing generates disposable code. Tracer code is lean but complete, and
forms part of the skeleton of the final system. Think of prototyping as
the reconnaissance and intelligence gathering that takes place before a
single tracer bullet is fired.

Related sections include:

Good-Enough Software, page 9
Prototypes and Post-it Notes, page 53
The Specification Trap, page 217
Great Expectations, page 255

Prepared exclusively for Zach

 P ROTOTYPES AND P OST - IT N OTES 53

11 Prototypes and Post-it Notes

Many different industries use prototypes to try out specific ideas; pro-
totyping is much cheaper than full-scale production. Car makers, for
example, may build many different prototypes of a new car design. Each
one is designed to test a specific aspect of the car—the aerodynamics,
styling, structural characteristics, and so on. Perhaps a clay model will
be built for wind tunnel testing, maybe a balsa wood and duct tape
model will do for the art department, and so on. Some car companies
take this a step further, and now do a great deal of modeling work on
the computer, reducing costs even further. In this way, risky or uncer-
tain elements can be tried out without committing to building the real
item.

We build software prototypes in the same fashion, and for the same
reasons—to analyze and expose risk, and to offer chances for correction
at a greatly reduced cost. Like the car makers, we can target a prototype
to test one or more specific aspects of a project.

We tend to think of prototypes as code-based, but they don’t always

have to be. Like the car makers, we can build prototypes out of different
materials. Post-it notes are great for prototyping dynamic things such
as workflow and application logic. A user interface can be prototyped
as a drawing on a whiteboard, as a nonfunctional mock-up drawn with
a paint program, or with an interface builder.

Prototypes are designed to answer just a few questions, so they are

much cheaper and faster to develop than applications that go into pro-
duction. The code can ignore unimportant details—unimportant to you
at the moment, but probably very important to the user later on. If you
are prototyping a GUI, for instance, you can get away with incorrect
results or data. On the other hand, if you’re just investigating compu-
tational or performance aspects, you can get away with a pretty poor
GUI, or perhaps even no GUI at all.

But if you find yourself in an environment where you cannot give up

the details, then you need to ask yourself if you are really building a
prototype at all. Perhaps a tracer bullet style of development would be
more appropriate in this case (see Tracer Bullets, page 48).

Prepared exclusively for Zach

 54 C HAPTER 2 A P RAGMATIC A PPROACH

Things to Prototype
What sorts of things might you choose to investigate with a prototype?
Anything that carries risk. Anything that hasn’t been tried before, or
that is absolutely critical to the final system. Anything unproven, exper-
imental, or doubtful. Anything you aren’t comfortable with. You can
prototype

Architecture
New functionality in an existing system
Structure or contents of external data
Third-party tools or components
Performance issues
User interface design

Prototyping is a learning experience. Its value lies not in the code pro-
duced, but in the lessons learned. That’s really the point of prototyping.

TIP 16

Prototype to Learn

How to Use Prototypes

When building a prototype, what details can you ignore?

Correctness. You may be able to use dummy data where appro-

priate.

Completeness. The prototype may function only in a very limited

sense, perhaps with only one preselected piece of input data and
one menu item.

Robustness. Error checking is likely to be incomplete or missing

entirely. If you stray from the predefined path, the prototype may
crash and burn in a glorious display of pyrotechnics. That’s okay.

Style. It is painful to admit this in print, but prototype code prob-

ably doesn’t have much in the way of comments or documenta-
tion. You may produce reams of documentation as a result of your
experience with the prototype, but comparatively very little on the
prototype system itself.

Prepared exclusively for Zach

 P ROTOTYPES AND P OST - IT N OTES 55

Since a prototype should gloss over details, and focus in on specific

aspects of the system being considered, you may want to implement
prototypes using a very high-level language—higher than the rest of
the project (maybe a language such as Perl, Python, or Tcl). A high-
level scripting language lets you defer many details (including specify-
ing data types) and still produce a functional (albeit incomplete or slow)
piece of code.6 If you need to prototype user interfaces, investigate tools
such as Tcl/Tk, Visual Basic, Powerbuilder, or Delphi.

Scripting languages work well as the “glue” to combine low-level pieces

into new combinations. Under Windows, Visual Basic can glue together
COM controls. More generally, you can use languages such as Perl
and Python to bind together low-level C libraries—either by hand, or
automatically with tools such as the freely available SWIG [URL 28].
Using this approach, you can rapidly assemble existing components
into new configurations to see how things work.

Prototyping Architecture
Many prototypes are constructed to model the entire system under con-
sideration. As opposed to tracer bullets, none of the individual modules
in the prototype system need to be particularly functional. In fact, you
may not even need to code in order to prototype architecture—you can
prototype on a whiteboard, with Post-it notes or index cards. What you
are looking for is how the system hangs together as a whole, again de-
ferring details. Here are some specific areas you may want to look for
in the architectural prototype:

Are the responsibilities of the major components well defined and

appropriate?

Are the collaborations between major components well defined?

Is coupling minimized?

Can you identify potential sources of duplication?

Are interface definitions and constraints acceptable?

6. If you are investigating absolute (instead of relative) performance, you will need to
stick to a language that is close in performance to the target language.

Prepared exclusively for Zach

 56 C HAPTER 2 A P RAGMATIC A PPROACH

Does every module have an access path to the data it needs during
execution? Does it have that access when it needs it?

This last item tends to generate the most surprises and the most valu-
able results from the prototyping experience.

How Not to Use Prototypes

Before you embark on any code-based prototyping, make sure that
everyone understands that you are writing disposable code. Prototypes
can be deceptively attractive to people who don’t know that they are just
prototypes. You must make it very clear that this code is disposable,
incomplete, and unable to be completed.

It’s easy to become misled by the apparent completeness of a demon-

strated prototype, and project sponsors or management may insist on
deploying the prototype (or its progeny) if you don’t set the right expec-
tations. Remind them that you can build a great prototype of a new
car out of balsa wood and duct tape, but you wouldn’t try to drive it in
rush-hour traffic!

If you feel there is a strong possibility in your environment or culture

that the purpose of prototype code may be misinterpreted, you may be
better off with the tracer bullet approach. You’ll end up with a solid
framework on which to base future development.

When used properly, a prototype can save you huge amounts of time,
money, pain, and suffering by identifying and correcting potential prob-
lem spots early in the development cycle—the time when fixing mis-
takes is both cheap and easy.

Related sections include:

The Cat Ate My Source Code, page 2
Communicate!, page 18
Tracer Bullets, page 48
Great Expectations, page 255

Exercises
Answer 4. Marketing would like to sit down and brainstorm a few Web-page designs
on p. 280
with you. They are thinking of clickable image maps to take you to other
pages, and so on. But they can’t decide on a model for the image—maybe

Prepared exclusively for Zach

 D OMAIN L ANGUAGES 57

it’s a car, or a phone, or a house. You have a list of target pages and
content; they’d like to see a few prototypes. Oh, by the way, you have 15
minutes. What tools might you use?

12 Domain Languages
The limits of language are the limits of one’s world.
Ludwig Wittgenstein

Computer languages influence how you think about a problem, and

how you think about communicating. Every language comes with a
list of features—buzzwords such as static versus dynamic typing, early
versus late binding, inheritance models (single, multiple, or none)—all
of which may suggest or obscure certain solutions. Designing a solution
with Lisp in mind will produce different results than a solution based
on C-style thinking, and vice versa. Conversely, and we think more
importantly, the language of the problem domain may also suggest a
programming solution.

We always try to write code using the vocabulary of the application

domain (see The Requirements Pit, page 210, where we suggest using a
project glossary). In some cases, we can go to the next level and actually
program using the vocabulary, syntax, and semantics—the language—
of the domain.

When you listen to users of a proposed system, they might be able to

tell you exactly how the system should work:
Listen for transactions defined by ABC Regulation 12.3 on a set of X.25
lines, translate them to XYZ Company’s format 43B, retransmit them on
the satellite uplink, and store for future analysis.

If your users have a number of such well-bounded statements, you

can invent a mini-language tailored to the application domain that ex-
presses exactly what they want:
From X25LINE1 (Format=ABC123) {
Put TELSTAR1 (Format=XYZ43B);
Store DB;
}

Prepared exclusively for Zach

 58 C HAPTER 2 A P RAGMATIC A PPROACH

This language need not be executable. Initially, it could be simply a

way of capturing the user’s requirements—a specification. However,
you may want to consider taking this a step further and actually imple-
menting the language. Your specification has become executable code.

After you’ve written the application, the users give you a new require-
ment: transactions with negative balances shouldn’t be stored, and
should be sent back on the X.25 lines in the original format:

From X25LINE1 (Format=ABC123) {

if (ABC123.balance < 0) {
Put X25LINE1 (Format=ABC123);
}
else {
Put TELSTAR1 (Format=XYZ43B);
Store DB;
}
}

That was easy, wasn’t it? With the proper support in place, you can pro-
gram much closer to the application domain. We’re not suggesting that
your end users actually program in these languages. Instead, you’re
giving yourself a tool that lets you work closer to their domain.

TIP 17

Program Close to the Problem Domain

Whether it’s a simple language to configure and control an application

program, or a more complex language to specify rules or procedures,
we think you should consider ways of moving your project closer to
the problem domain. By coding at a higher level of abstraction, you are
free to concentrate on solving domain problems, and can ignore petty
implementation details.

Remember that there are many users of an application. There’s the end
user, who understands the business rules and the required outputs.
There are also secondary users: operations staff, configuration and test
managers, support and maintenance programmers, and future genera-
tions of developers. Each of these users has their own problem domain,
and you can generate mini-environments and languages for all of them.

Prepared exclusively for Zach

 D OMAIN L ANGUAGES 59

Domain-Specific Errors
If you are writing in the problem domain, you can also perform
domain-specific validation, reporting problems in terms your users
can understand. Take our switching application on on the facing page.
Suppose the user misspelled the format name:
From X25LINE1 (Format=AB123)

If this happened in a standard, general-purpose programming lan-

guage, you might receive a standard, general-purpose error mes-
sage:
Syntax error: undeclared identifier

But with a mini-language, you would instead be able to issue an error

message using the vocabulary of the domain:
"AB123" is not a format. Known formats are ABC123,
XYZ43B, PDQB, and 42.

Implementing a Mini-Language
At its simplest, a mini-language may be in a line-oriented, easily parsed
format. In practice, we probably use this form more than any other.
It can be parsed simply using switch statements, or using regular
expressions in scripting languages such as Perl. The answer to Exercise
5 on page 281 shows a simple implementation in C.

You can also implement a more complex language, with a more formal
syntax. The trick here is to define the syntax first using a notation such
as BNF.7 Once you have your grammar specified, it is normally trivial to
convert it into the input syntax for a parser generator. C and C++ pro-
grammers have been using yacc (or its freely available implementation,
bison [URL 27]) for years. These programs are documented in detail in
the book Lex and Yacc [LMB92]. Java programmers can try javaCC,
which can be found at [URL 26]. The answer to Exercise 7 on page 282

7. BNF, or Backus-Naur Form, lets you specify context-free grammars recursively. Any
good book on compiler construction or parsing will cover BNF in (exhaustive) detail.

Prepared exclusively for Zach

 60 C HAPTER 2 A P RAGMATIC A PPROACH

shows a parser written using bison. As it shows, once you know the
syntax, it’s really not a lot of work to write simple mini-languages.

There’s another way of implementing a mini-language: extend an exist-

ing one. For example, you could integrate application-level functionality
with (say) Python [URL 9] and write something like8
record = X25LINE1.get(format=ABC123)
if (record.balance < 0):
X25LINE1.put(record, format=ABC123)
else:
TELSTAR1.put(record, format=XYZ43B)
DB.store(record)

Data Languages and Imperative Languages

The languages you implement can be used in two different ways.

Data languages produce some form of data structure used by an ap-

plication. These languages are often used to represent configuration
information.

For example, the sendmail program is used throughout the world for
routing e-mail over the Internet. It has many excellent features and
benefits, which are controlled by a thousand-line configuration file,
written using sendmail’s own configuration language:
Mlocal, P=/usr/bin/procmail,
F=lsDFMAw5:/|@qSPfhn9,
S=10/30, R=20/40,
T=DNS/RFC822/X-Unix,
A=procmail -Y -a $h -d $u

Obviously, readability is not one of sendmail’s strengths.

For years, Microsoft has been using a data language that can describe
menus, widgets, dialog boxes, and other Windows resources. Figure 2.2
on the next page shows an excerpt from a typical resource file. This is
far easer to read than the sendmail example, but it is used in exactly
the same way—it is compiled to generate a data structure.

Imperative languages take this a step further. Here the language is

actually executed, and so can contain statements, control constructs,
and the like (such as the script on page 58).

8. Thanks to Eric Vought for this example.

Prepared exclusively for Zach

 D OMAIN L ANGUAGES 61

Figure 2.2. Windows .rc file

MAIN_MENU MENU
{
POPUP "&File"
{
MENUITEM "&New", CM_FILENEW
MENUITEM "&Open...", CM_FILEOPEN
MENUITEM "&Save", CM_FILESAVE
}
}
MY_DIALOG_BOX DIALOG 6, 15, 292, 287
STYLE DS_MODALFRAME | WS_POPUP | WS_VISIBLE |
WS_CAPTION | WS_SYSMENU
CAPTION "My Dialog Box"
FONT 8, "MS Sans Serif"
{
DEFPUSHBUTTON "OK", ID_OK, 232, 16, 50, 14
PUSHBUTTON "Help", ID_HELP, 232, 52, 50, 14
CONTROL "Edit Text Control", ID_EDIT1,
"EDIT", WS_BORDER | WS_TABSTOP, 16, 16, 80, 56
CHECKBOX "Checkbox", ID_CHECKBOX1, 153, 65, 42, 38,
BS_AUTOCHECKBOX | WS_TABSTOP
}

You can also use your own imperative languages to ease program main-
tenance. For example, you may be asked to integrate information from
a legacy application into your new GUI development. A common way
of achieving this is by screen scraping; your application connects to
the mainframe application as if it were a regular human user, issuing
keystrokes and “reading” the responses it gets back. You could script
the interaction using a mini-language.9

locate prompt "SSN:"

type "%s" social_security_number
type enter
waitfor keyboardunlock
if text_at(10,14) is "INVALID SSN" return bad_ssn
if text_at(10,14) is "DUPLICATE SSN" return dup_ssn
# etc...

When the application determines it is time to enter a Social Security

number, it invokes the interpreter on this script, which then controls

9. In fact, you can buy tools that support just this kind of scripting. You can also inves-
tigate open-source packages such as Expect, which provide similar capabilities [URL 24].

Prepared exclusively for Zach

 62 C HAPTER 2 A P RAGMATIC A PPROACH

the transaction. If the interpreter is embedded within the application,

the two can even share data directly (for example, via a callback mech-
anism).

Here you’re programming in the maintenance programmer’s domain.

When the mainframe application changes, and the fields move around,
the programmer can simply update your high-level description, rather
than groveling around in the details of C code.

Stand-Alone and Embedded Languages

A mini-language doesn’t have to be used directly by the application to
be useful. Many times we may use a specification language to create
artifacts (including metadata) that are compiled, read-in, or otherwise
used by the program itself (see Metaprogramming, page 144).

For example, on page 100 we describe a system in which we used

Perl to generate a large number of derivations from an original schema
specification. We invented a common language to express the database
schema, and then generated all the forms of it we needed—SQL, C,
Web pages, XML, and others. The application didn’t use the specifica-
tion directly, but it relied on the output produced from it.

It is common to embed high-level imperative languages directly into

your application, so that they execute when your code runs. This is
clearly a powerful capability; you can change your application’s behav-
ior by changing the scripts it reads, all without compiling. This can
significantly simplify maintenance in a dynamic application domain.

Easy Development or Easy Maintenance?

We’ve looked at several different grammars, ranging from simple line-
oriented formats to more complex grammars that look like real lan-
guages. Since it takes extra effort to implement, why would you choose
a more complex grammar?

The trade-off is extendibility and maintenance. While the code for pars-
ing a “real” language may be harder to write, it will be much easier for
people to understand, and to extend in the future with new features
and functionality. Languages that are too simple may be easy to parse,
but can be cryptic—much like the sendmail example on page 60.

Prepared exclusively for Zach

 D OMAIN L ANGUAGES 63

Given that most applications exceed their expected lifetimes, you’re

probably better off biting the bullet and adopting the more complex
and readable language up front. The initial effort will be repaid many
times in reduced support and maintenance costs.

Related sections include:

Metaprogramming, page 144

Challenges
Could some of the requirements of your current project be expressed in
a domain-specific language? Would it be possible to write a compiler or
translator that could generate most of the code required?

If you decide to adopt mini-languages as a way of programming closer to

the problem domain, you’re accepting that some effort will be required to
implement them. Can you see ways in which the framework you develop
for one project can be reused in others?

Exercises
5. We want to implement a mini-language to control a simple drawing pack- Answer
on p. 281
age (perhaps a turtle-graphics system). The language consists of single-
letter commands. Some commands are followed by a single number. For
example, the following input would draw a rectangle.

P 2 # select pen 2
D # pen down
W 2 # draw west 2cm
N 1 # then north 1
E 2 # then east 2
S 1 # then back south
U # pen up

Implement the code that parses this language. It should be designed so

that it is simple to add new commands.

6. Design a BNF grammar to parse a time specification. All of the following Answer
on p. 282
examples should be accepted.

4pm, 7:38pm, 23:42, 3:16, 3:16am

7. Implement a parser for the BNF grammar in Exercise 6 using yacc, bison, Answer
on p. 282
or a similar parser-generator.

8. Implement the time parser using Perl. [Hint: Regular expressions make Answer
on p. 283
good parsers.]

Prepared exclusively for Zach

 64 C HAPTER 2 A P RAGMATIC A PPROACH

13 Estimating
Quick! How long will it take to send War and Peace over a 56k modem
line? How much disk space will you need for a million names and
addresses? How long does a 1,000-byte block take to pass through
a router? How many months will it take to deliver your project?

At one level, these are all meaningless questions—they are all missing
information. And yet they can all be answered, as long as you are com-
fortable estimating. And, in the process of producing an estimate, you’ll
come to understand more about the world your programs inhabit.

By learning to estimate, and by developing this skill to the point where

you have an intuitive feel for the magnitudes of things, you will be able
to show an apparent magical ability to determine their feasibility. When
someone says “we’ll send the backup over an ISDN line to the central
site,” you’ll be able to know intuitively whether this is practical. When
you’re coding, you’ll be able to know which subsystems need optimizing
and which ones can be left alone.

TIP 18

Estimate to Avoid Surprises

As a bonus, at the end of this section we’ll reveal the single correct
answer to give whenever anyone asks you for an estimate.

How Accurate Is Accurate Enough?

To some extent, all answers are estimates. It’s just that some are more
accurate than others. So the first question you have to ask yourself
when someone asks you for an estimate is the context in which your
answer will be taken. Do they need high accuracy, or are they looking
for a ballpark figure?

If your grandmother asks when you will arrive, she’s probably won-
dering whether to make you lunch or dinner. On the other hand, a
diver trapped underwater and running out of air is probably inter-
ested in an answer down to the second.

Prepared exclusively for Zach

 E STIMATING 65

What’s the value of ? If you’re wondering how much edging to

buy to put around a circular flower bed, then “3” is probably good
enough.10 If you’re in school, then maybe “ ” is a good approxima-
tion. If you’re in NASA, then maybe 12 decimal places will do.

One of the interesting things about estimating is that the units you
use make a difference in the interpretation of the result. If you say
that something will take about 130 working days, then people will be
expecting it to come in pretty close. However, if you say “Oh, about
six months,” then they know to look for it any time between five and
seven months from now. Both numbers represent the same duration,
but “130 days” probably implies a higher degree of accuracy than you
feel. We recommend that you scale time estimates as follows:

Duration Quote estimate in

1–15 days days
3–8 weeks weeks
8–30 weeks months
30+ weeks think hard before giving an estimate

So, if after doing all the necessary work, you decide that a project will
take 125 working days (25 weeks), you might want to deliver an esti-
mate of “about six months.”

The same concepts apply to estimates of any quantity: choose the units
of your answer to reflect the accuracy you intend to convey.

Where Do Estimates Come From?

All estimates are based on models of the problem. But before we get
too deeply into the techniques of building models, we have to mention
a basic estimating trick that always gives good answers: ask someone
who’s already done it. Before you get too committed to model building,
cast around for someone who’s been in a similar situation in the past.

10. “3” is also apparently good enough if you are a legislator. In 1897, Indiana State
Legislature House Bill No. 246 attempted to decree that henceforth should have the
value of “3”. The Bill was tabled indefinitely at its second reading when a mathematics
professor pointed out that their powers did not quite extend to passing laws of nature.

Prepared exclusively for Zach

 66 C HAPTER 2 A P RAGMATIC A PPROACH

See how their problem got solved. It’s unlikely you’ll ever find an exact
match, but you’d be surprised how many times you can successfully
draw on other’s experiences.

Understand What’s Being Asked

The first part of any estimation exercise is building an understanding
of what’s being asked. As well as the accuracy issues discussed above,
you need to have a grasp of the scope of the domain. Often this is
implicit in the question, but you need to make it a habit to think about
the scope before starting to guess. Often, the scope you choose will form
part of the answer you give: “Assuming there are no traffic accidents
and there’s gas in the car, I should be there in 20 minutes.”

Build a Model of the System

This is the fun part of estimating. From your understanding of the ques-
tion being asked, build a rough and ready bare-bones mental model.
If you’re estimating response times, your model may involve a server
and some kind of arriving traffic. For a project, the model may be the
steps that your organization uses during development, along with a
very rough picture of how the system might be implemented.

Model building can be both creative and useful in the long term. Often,
the process of building the model leads to discoveries of underlying
patterns and processes that weren’t apparent on the surface. You may
even want to reexamine the original question: “You asked for an esti-
mate to do X. However, it looks like Y, a variant of X, could be done in
about half the time, and you lose only one feature.”

Building the model introduces inaccuracies into the estimating pro-

cess. This is inevitable, and also beneficial. You are trading off model
simplicity for accuracy. Doubling the effort on the model may give you
only a slight increase in accuracy. Your experience will tell you when to
stop refining.

Break the Model into Components

Once you have a model, you can decompose it into components. You’ll
need to discover the mathematical rules that describe how these com-
ponents interact. Sometimes a component contributes a single value

Prepared exclusively for Zach

 E STIMATING 67

that is added into the result. Some components may supply multiply-
ing factors, while others may be more complicated (such as those that
simulate the arrival of traffic at a node).

You’ll find that each component will typically have parameters that af-
fect how it contributes to the overall model. At this stage, simply identify
each parameter.

Give Each Parameter a Value

Once you have the parameters broken out, you can go through and
assign each one a value. You expect to introduce some errors in this
step. The trick is to work out which parameters have the most impact
on the result, and concentrate on getting them about right. Typically,
parameters whose values are added into a result are less significant
than those that are multiplied or divided. Doubling a line speed may
double the amount of data received in an hour, while adding a 5 ms
transit delay will have no noticeable effect.

You should have a justifiable way of calculating these critical parame-

ters. For the queuing example, you might want to measure the actual
transaction arrival rate of the existing system, or find a similar sys-
tem to measure. Similarly, you could measure the current time taken
to serve a request, or come up with an estimate using the techniques
described in this section. In fact, you’ll often find yourself basing an
estimate on other subestimates. This is where your largest errors will
creep in.

Calculate the Answers

Only in the simplest of cases will an estimate have a single answer.
You might be happy to say “I can walk five cross-town blocks in 15
minutes.” However, as the systems get more complex, you’ll want to
hedge your answers. Run multiple calculations, varying the values of
the critical parameters, until you work out which ones really drive the
model. A spreadsheet can be a big help. Then couch your answer in
terms of these parameters. ”The response time is roughly three quarters
of a second if the system has a SCSI bus and 64MB memory, and one
second with 48MB memory.” (Notice how “three quarters of a second”
conveys a different feeling of accuracy than 750 ms.)

Prepared exclusively for Zach

 68 C HAPTER 2 A P RAGMATIC A PPROACH

During the calculation phase, you may start getting answers that seem
strange. Don’t be too quick to dismiss them. If your arithmetic is cor-
rect, your understanding of the problem or your model is probably
wrong. This is valuable information.

Keep Track of Your Estimating Prowess

We think it’s a great idea to record your estimates so you can see how
close you were. If an overall estimate involved calculating subestimates,
keep track of these as well. Often you’ll find your estimates are pretty
good—in fact, after a while, you’ll come to expect this.

When an estimate turns out wrong, don’t just shrug and walk away.
Find out why it differed from your guess. Maybe you chose some param-
eters that didn’t match the reality of the problem. Maybe your model
was wrong. Whatever the reason, take some time to uncover what hap-
pened. If you do, your next estimate will be better.

Estimating Project Schedules

The normal rules of estimating can break down in the face of the com-
plexities and vagaries of a sizable application development. We find that
often the only way to determine the timetable for a project is by gain-
ing experience on that same project. This needn’t be a paradox if you
practice incremental development, repeating the following steps.

Check requirements
Analyze risk
Design, implement, integrate
Validate with the users

Initially, you may have only a vague idea of how many iterations will be
required, or how long they may be. Some methods require you to nail
this down as part of the initial plan, but for all but the most trivial of
projects this is a mistake. Unless you are doing an application similar
to a previous one, with the same team and the same technology, you’d
just be guessing.

So you complete the coding and testing of the initial functionality and
mark this as the end of the first increment. Based on that experience,
you can refine your initial guess on the number of iterations and what

Prepared exclusively for Zach

 E STIMATING 69

can be included in each. The refinement gets better and better each
time, and confidence in the schedule grows along with it.

TIP 19

Iterate the Schedule with the Code

This may not be popular with management, who typically want a sin-
gle, hard-and-fast number before the project even starts. You’ll have to
help them understand that the team, their productivity, and the envi-
ronment will determine the schedule. By formalizing this, and refining
the schedule as part of each iteration, you’ll be giving them the most
accurate scheduling estimates you can.

What to Say When Asked for an Estimate

You say “I’ll get back to you.”

You almost always get better results if you slow the process down and
spend some time going through the steps we describe in this section.
Estimates given at the coffee machine will (like the coffee) come back to
haunt you.

Related sections include:

Algorithm Speed, page 177

Challenges
Start keeping a log of your estimates. For each, track how accurate you
turned out to be. If your error was greater than 50%, try to find out where
your estimate went wrong.

Exercises
9. You are asked “Which has a higher bandwidth: a 1Mbps communications Answer
on p. 283
line or a person walking between two computers with a full 4GB tape in
their pocket?” What constraints will you put on your answer to ensure that
the scope of your response is correct? (For example, you might say that the
time taken to access the tape is ignored.)

10. So, which has the higher bandwidth? Answer

on p. 284

Prepared exclusively for Zach

 This page intentionally left blank

Prepared exclusively for Zach

 Chapter 3

The Basic Tools

Every craftsman starts his or her journey with a basic set of good-
quality tools. A woodworker might need rules, gauges, a couple of saws,
some good planes, fine chisels, drills and braces, mallets, and clamps.
These tools will be lovingly chosen, will be built to last, will perform
specific jobs with little overlap with other tools, and, perhaps most im-
portantly, will feel right in the budding woodworker’s hands.

Then begins a process of learning and adaptation. Each tool will have
its own personality and quirks, and will need its own special handling.
Each must be sharpened in a unique way, or held just so. Over time,
each will wear according to use, until the grip looks like a mold of the
woodworker’s hands and the cutting surface aligns perfectly with the
angle at which the tool is held. At this point, the tools become conduits
from the craftsman’s brain to the finished product—they have become
extensions of his or her hands. Over time, the woodworker will add new
tools, such as biscuit cutters, laser-guided miter saws, dovetail jigs—
all wonderful pieces of technology. But you can bet that he or she will
be happiest with one of those original tools in hand, feeling the plane
sing as it slides through the wood.

Tools amplify your talent. The better your tools, and the better you
know how to use them, the more productive you can be. Start with a
basic set of generally applicable tools. As you gain experience, and as
you come across special requirements, you’ll add to this basic set. Like
the craftsman, expect to add to your toolbox regularly. Always be on the
lookout for better ways of doing things. If you come across a situation
where you feel your current tools can’t cut it, make a note to look for

71

Prepared exclusively for Zach

 72 C HAPTER 3 T HE B ASIC T OOLS

something different or more powerful that would have helped. Let need
drive your acquisitions.

Many new programmers make the mistake of adopting a single power

tool, such as a particular integrated development environment (IDE),
and never leave its cozy interface. This really is a mistake. We need to
be comfortable beyond the limits imposed by an IDE. The only way to
do this is to keep the basic tool set sharp and ready to use.

In this chapter we’ll talk about investing in your own basic toolbox.
As with any good discussion on tools, we’ll start (in The Power of Plain
Text) by looking at your raw materials, the stuff you’ll be shaping. From
there we’ll move to the workbench, or in our case the computer. How
can you use your computer to get the most out of the tools you use?
We’ll discuss this in Shell Games. Now that we have material and a
bench to work on, we’ll turn to the tool you’ll probably use more than
any other, your editor. In Power Editing, we’ll suggest ways of making
you more efficient.

To ensure that we never lose any of our precious work, we should al-
ways use a Source Code Control system—even for things such as our
personal address book! And, since Mr. Murphy was really an optimist
after all, you can’t be a great programmer until you become highly
skilled at Debugging.

You’ll need some glue to bind much of the magic together. We discuss
some possibilities, such as awk, Perl, and Python, in Text Manipulation.

Just as woodworkers sometimes build jigs to guide the construction of

complex pieces, programmers can write code that itself writes code. We
discuss this in Code Generators.

Spend time learning to use these tools, and at some point you’ll be sur-
prised to discover your fingers moving over the keyboard, manipulating
text without conscious thought. The tools will have become extensions
of your hands.

Prepared exclusively for Zach

 T HE P OWER OF P LAIN T EXT 73

14 The Power of Plain Text

As Pragmatic Programmers, our base material isn’t wood or iron, it’s
knowledge. We gather requirements as knowledge, and then express
that knowledge in our designs, implementations, tests, and documents.
And we believe that the best format for storing knowledge persistently
is plain text. With plain text, we give ourselves the ability to manipulate
knowledge, both manually and programmatically, using virtually every
tool at our disposal.

What Is Plain Text?

Plain text is made up of printable characters in a form that can be read
and understood directly by people. For example, although the following
snippet is made up of printable characters, it is meaningless.
Field19=467abe

The reader has no idea what the significance of 467abe may be. A better
choice would be to make it understandable to humans.
DrawingType=UMLActivityDrawing

Plain text doesn’t mean that the text is unstructured; XML, SGML, and
HTML are great examples of plain text that has a well-defined structure.
You can do everything with plain text that you could do with some
binary format, including versioning.

Plain text tends to be at a higher level than a straight binary encoding,

which is usually derived directly from the implementation. Suppose you
wanted to store a property called uses_menus that can be either TRUE
or FALSE. Using text, you might write this as
myprop.uses_menus=FALSE

Contrast this with 0010010101110101.

The problem with most binary formats is that the context necessary to
understand the data is separate from the data itself. You are artificially
divorcing the data from its meaning. The data may as well be encrypted;
it is absolutely meaningless without the application logic to parse it.
With plain text, however, you can achieve a self-describing data stream
that is independent of the application that created it.

Prepared exclusively for Zach

 74 C HAPTER 3 T HE B ASIC T OOLS

TIP 20

Keep Knowledge in Plain Text

Drawbacks
There are two major drawbacks to using plain text: (1) It may take
more space to store than a compressed binary format, and (2) it may
be computationally more expensive to interpret and process a plain text
file.

Depending on your application, either or both of these situations may

be unacceptable—for example, when storing satellite telemetry data, or
as the internal format of a relational database.

But even in these situations, it may be acceptable to store metadata

about the raw data in plain text (see Metaprogramming, page 144).

Some developers may worry that by putting metadata in plain text,

they’re exposing it to the system’s users. This fear is misplaced. Binary
data may be more obscure than plain text, but it is no more secure.
If you worry about users seeing passwords, encrypt them. If you don’t
want them changing configuration parameters, include a secure hash 1
of all the parameter values in the file as a checksum.

The Power of Text

Since larger and slower aren’t the most frequently requested features
from users, why bother with plain text? What are the benefits?

Insurance against obsolescence

Leverage
Easier testing

Insurance Against Obsolescence

Human-readable forms of data, and self-describing data, will outlive
all other forms of data and the applications that created them. Period.

1. MD5 is often used for this purpose. For an excellent introduction to the wonderful
world of cryptography, see [Sch95].

Prepared exclusively for Zach

 T HE P OWER OF P LAIN T EXT 75

As long as the data survives, you will have a chance to be able to use
it—potentially long after the original application that wrote it is defunct.

You can parse such a file with only partial knowledge of its format; with
most binary files, you must know all the details of the entire format in
order to parse it successfully.

Consider a data file from some legacy system2 that you are given. You
know little about the original application; all that’s important to you is
that it maintained a list of clients’ Social Security numbers, which you
need to find and extract. Among the data, you see
<FIELD10>123-45-6789</FIELD10>
...
<FIELD10>567-89-0123</FIELD10>
...
<FIELD10>901-23-4567</FIELD10>

Recognizing the format of a Social Security number, you can quickly

write a small program to extract that data—even if you have no infor-
mation on anything else in the file.

But imagine if the file had been formatted this way instead:
AC27123456789B11P
...
XY43567890123QTYL
...
6T2190123456788AM

You may not have recognized the significance of the numbers quite
as easily. This is the difference between human readable and human
understandable.

While we’re at it, FIELD10 doesn’t help much either. Something like
<SSNO>123-45-6789</SSNO>

makes the exercise a no-brainer—and ensures that the data will outlive
any project that created it.

Leverage
Virtually every tool in the computing universe, from source code man-
agement systems to compiler environments to editors and stand-alone
filters, can operate on plain text.

2. All software becomes legacy as soon as it’s written.

Prepared exclusively for Zach

 76 C HAPTER 3 T HE B ASIC T OOLS

The Unix Philosophy

Unix is famous for being designed around the philosophy of small,
sharp tools, each intended to do one thing well. This philosophy is
enabled by using a common underlying format—the line-oriented,
plain text file. Databases used for system administration (users and
passwords, networking configuration, and so on) are all kept as plain
text files. (Some systems, such as Solaris, also maintain a binary
form of certain databases as a performance optimization. The plain
text version is kept as an interface to the binary version.)
When a system crashes, you may be faced with only a minimal envi-
ronment to restore it (you may not be able to access graphics drivers,
for instance). Situations such as this can really make you appreciate
the simplicity of plain text.

For instance, suppose you have a production deployment of a large

application with a complex site-specific configuration file (sendmail
comes to mind). If this file is in plain text, you could place it under a
source code control system (see Source Code Control, page 86), so that
you automatically keep a history of all changes. File comparison tools
such as diff and fc allow you to see at a glance what changes have
been made, while sum allows you to generate a checksum to monitor
the file for accidental (or malicious) modification.

Easier Testing
If you use plain text to create synthetic data to drive system tests, then
it is a simple matter to add, update, or modify the test data without
having to create any special tools to do so. Similarly, plain text output
from regression tests can be trivially analyzed (with diff, for instance)
or subjected to more thorough scrutiny with Perl, Python, or some other
scripting tool.

Lowest Common Denominator

Even in the future of XML-based intelligent agents that travel the wild
and dangerous Internet autonomously, negotiating data interchange
among themselves, the ubiquitous text file will still be there. In fact, in

Prepared exclusively for Zach

 S HELL G AMES 77

heterogeneous environments the advantages of plain text can outweigh

all of the drawbacks. You need to ensure that all parties can communi-
cate using a common standard. Plain text is that standard.

Related sections include:

Source Code Control, page 86
Code Generators, page 102
Metaprogramming, page 144
Blackboards, page 165
Ubiquitous Automation, page 230
It’s All Writing, page 248

Challenges
Design a small address book database (name, phone number, and so on)
using a straightforward binary representation in your language of choice.
Do this before reading the rest of this challenge.

1. Translate that format into a plain text format using XML.

2. For each version, add a new, variable-length field called directions in

which you might enter directions to each person’s house.

What issues come up regarding versioning and extensibility? Which form

was easier to modify? What about converting existing data?

15 Shell Games
Every woodworker needs a good, solid, reliable workbench, somewhere
to hold work pieces at a convenient height while he or she works them.
The workbench becomes the center of the wood shop, the craftsman
returning to it time and time again as a piece takes shape.

For a programmer manipulating files of text, that workbench is the

command shell. From the shell prompt, you can invoke your full reper-
toire of tools, using pipes to combine them in ways never dreamt of by
their original developers. From the shell, you can launch applications,
debuggers, browsers, editors, and utilities. You can search for files,

Prepared exclusively for Zach

 78 C HAPTER 3 T HE B ASIC T OOLS

query the status of the system, and filter output. And by programming
the shell, you can build complex macro commands for activities you
perform often.

For programmers raised on GUI interfaces and integrated development

environments (IDEs), this might seem an extreme position. After all,
can’t you do everything equally well by pointing and clicking?

The simple answer is “no.” GUI interfaces are wonderful, and they can
be faster and more convenient for some simple operations. Moving files,
reading MIME-encoded e-mail, and typing letters are all things that
you might want to do in a graphical environment. But if you do all
your work using GUIs, you are missing out on the full capabilities of
your environment. You won’t be able to automate common tasks, or
use the full power of the tools available to you. And you won’t be able
to combine your tools to create customized macro tools. A benefit of
GUIs is WYSIWYG—what you see is what you get. The disadvantage is
WYSIAYG—what you see is all you get.

GUI environments are normally limited to the capabilities that their

designers intended. If you need to go beyond the model the designer
provided, you are usually out of luck—and more often than not, you
do need to go beyond the model. Pragmatic Programmers don’t just cut
code, or develop object models, or write documentation, or automate
the build process—we do all of these things. The scope of any one tool
is usually limited to the tasks that the tool is expected to perform.
For instance, suppose you need to integrate a code preprocessor (to
implement design-by-contract, or multi-processing pragmas, or some
such) into your IDE. Unless the designer of the IDE explicitly provided
hooks for this capability, you can’t do it.

You may already be comfortable working from the command prompt, in

which case you can safely skip this section. Otherwise, you may need
to be convinced that the shell is your friend.

As a Pragmatic Programmer, you will constantly want to perform ad hoc

operations—things that the GUI may not support. The command line is
better suited when you want to quickly combine a couple of commands
to perform a query or some other task. Here are a few examples.

Prepared exclusively for Zach

 S HELL G AMES 79

Find all .c files modified more recently than your Makefile.

Shell . . . find . -name ’*.c’ -newer Makefile -print

GUI . . . . . Open the Explorer, navigate to the correct directory,

click on the Makefile, and note the modification time.
Then bring up Tools/Find, and enter *.c for the file
specification. Select the date tab, and enter the date you
noted for the Makefile in the first date field. Then hit OK.

Construct a zip/tar archive of my source.

Shell . . . zip archive.zip *.h *.c – or –
tar cvf archive.tar *.h *.c

GUI . . . . . Bring up a ZIP utility (such as the shareware WinZip

[URL 41]), select “Create New Archive,” enter its name,
select the source directory in the add dialog, set the filter
to “*.c”, click “Add,” set the filter to “*.h”, click “Add,”
then close the archive.

Which Java files have not been changed in the last week?
Shell . . . find . -name ’*.java’ -mtime +7 -print

GUI . . . . . Click and navigate to “Find files,” click the “Named” field
and type in “*.java”, select the “Date Modified” tab. Then
select “Between.” Click on the starting date and type in
the starting date of the beginning of the project. Click on
the ending date and type in the date of a week ago today
(be sure to have a calendar handy). Click on “Find Now.”

Of those files, which use the awt libraries?

Shell . . . find . -name ’*.java’ -mtime +7 -print |
xargs grep ’java.awt’

GUI . . . . . Load each file in the list from the previous example
into an editor and search for the string “java.awt”. Write
down the name of each file containing a match.

Clearly the list could go on. The shell commands may be obscure or
terse, but they are powerful and concise. And, because shell commands
can be combined into script files (or command files under Windows

Prepared exclusively for Zach

 80 C HAPTER 3 T HE B ASIC T OOLS

systems), you can build sequences of commands to automate things

you do often.

TIP 21

Use the Power of Command Shells

Gain familiarity with the shell, and you’ll find your productivity soaring.
Need to create a list of all the unique package names explicitly imported
by your Java code? The following stores it in a file called “list.”

grep ’^import ’ *.java |

sed -e’s/.*import *//’ -e’s/;.*$//’ |
sort -u >list

If you haven’t spent much time exploring the capabilities of the com-
mand shell on the systems you use, this might appear daunting. How-
ever, invest some energy in becoming familiar with your shell and things
will soon start falling into place. Play around with your command shell,
and you’ll be surprised at how much more productive it makes you.

Shell Utilities and Windows Systems

Although the command shells provided with Windows systems are im-
proving gradually, Windows command-line utilities are still inferior to
their Unix counterparts. However, all is not lost.

Cygnus Solutions has a package called Cygwin [URL 31]. As well as

providing a Unix compatibility layer for Windows, Cygwin comes with a
collection of more than 120 Unix utilities, including such favorites as
ls, grep, and find. The utilities and libraries may be downloaded and
used for free, but be sure to read their license.3 The Cygwin distribution
comes with the Bash shell.

3. The GNU General Public License [URL 57] is a kind of legal virus that Open Source
developers use to protect their (and your) rights. You should spend some time reading
it. In essence, it says that you can use and modify GPL’d software, but if you distribute
any modifications they must be licensed according to the GPL (and marked as such), and
you must make source available. That’s the virus part—whenever you derive a work from
a GPL’d work, your derived work must also be GPL’d. However, it does not limit you in
any way when simply using the tools—the ownership and licensing of software developed
using the tools are up to you.

Prepared exclusively for Zach

 S HELL G AMES 81

Using Unix Tools Under Windows

We love the availability of high-quality Unix tools under Windows, and
use them daily. However, be aware that there are integration issues.
Unlike their MS-DOS counterparts, these utilities are sensitive to the
case of filenames, so ls a*.bat won’t find AUTOEXEC.BAT. You
may also come across problems with filenames containing spaces,
and with differences in path separators. Finally, there are interesting
problems when running MS-DOS programs that expect MS-DOS–style
arguments under the Unix shells. For example, the Java utilities from
JavaSoft use a colon as their CLASSPATH separator under Unix, but
use a semicolon under MS-DOS. As a result, a Bash or ksh script
that runs on a Unix box will run identically under Windows, but the
command line it passes to Java will be interpreted incorrectly.

Alternatively, David Korn (of Korn shell fame) has put together a pack-
age called UWIN. This has the same aims as the Cygwin distribution—it
is a Unix development environment under Windows. UWIN comes with
a version of the Korn shell. Commercial versions are available from
Global Technologies, Ltd. [URL 30]. In addition, AT&T allows free down-
loading of the package for evaluation and academic use. Again, read
their license before using.

Finally, Tom Christiansen is (at the time of writing) putting together

Perl Power Tools, an attempt to implement all the familiar Unix utilities
portably, in Perl [URL 32].

Related sections include:

Ubiquitous Automation, page 230

Challenges
Are there things that you’re currently doing manually in a GUI? Do you
ever pass instructions to colleagues that involve a number of individual
“click this button,” “select this item” steps? Could these be automated?

Whenever you move to a new environment, make a point of finding out

what shells are available. See if you can bring your current shell with you.

Investigate alternatives to your current shell. If you come across a problem

your shell can’t address, see if an alternative shell would cope better.

Prepared exclusively for Zach

 82 C HAPTER 3 T HE B ASIC T OOLS

16 Power Editing
We’ve talked before about tools being an extension of your hand. Well,
this applies to editors more than to any other software tool. You need
to be able to manipulate text as effortlessly as possible, because text
is the basic raw material of programming. Let’s look at some common
features and functions that help you get the most from your editing
environment.

One Editor
We think it is better to know one editor very well, and use it for all edit-
ing tasks: code, documentation, memos, system administration, and so
on. Without a single editor, you face a potential modern day Babel of
confusion. You may have to use the built-in editor in each language’s
IDE for coding, and an all-in-one office product for documentation, and
maybe a different built-in editor for sending e-mail. Even the keystrokes
you use to edit command lines in the shell may be different.4 It is diffi-
cult to be proficient in any of these environments if you have a different
set of editing conventions and commands in each.

You need to be proficient. Simply typing linearly and using a mouse

to cut and paste is not enough. You just can’t be as effective that way
as you can with a powerful editor under your fingers. Typing or
BACKSPACE ten times to move the cursor left to the beginning of a line
isn’t as efficient as typing a single key such as ^A , Home , or 0 .

TIP 22

Use a Single Editor Well

Choose an editor, know it thoroughly, and use it for all editing tasks.
If you use a single editor (or set of keybindings) across all text editing
activities, you don’t have to stop and think to accomplish text manip-
ulation: the necessary keystrokes will be a reflex. The editor will be

4. Ideally, the shell you use should have keybindings that match the ones used by
your editor. Bash, for instance, supports both vi and emacs keybindings.

Prepared exclusively for Zach

 P OWER E DITING 83

an extension of your hand; the keys will sing as they slice their way
through text and thought. That’s our goal.

Make sure that the editor you choose is available on all platforms you
use. Emacs, vi, CRiSP, Brief, and others are available across multiple
platforms, often in both GUI and non-GUI (text screen) versions.

Editor Features
Beyond whatever features you find particularly useful and comfortable,
here are some basic abilities that we think every decent editor should
have. If your editor falls short in any of these areas, then this may be
the time to consider moving on to a more advanced one.

Configurable. All aspects of the editor should be configurable to

your preferences, including fonts, colors, window sizes, and key-
stroke bindings (which keys perform what commands). Using only
keystrokes for common editing operations is more efficient than
mouse or menu-driven commands, because your hands never leave
the keyboard.

Extensible. An editor shouldn’t be obsolete just because a new

programming language comes out. It should be able to integrate
with whatever compiler environment you are using. You should be
able to “teach” it the nuances of any new language or text format
(XML, HTML version 9, and so on).

Programmable. You should be able to program the editor to per-

form complex, multistep tasks. This can be done with macros or
with a built-in scripting programming language (Emacs uses a vari-
ant of Lisp, for instance).

In addition, many editors support features that are specific to a partic-

ular programming language, such as:

Syntax highlighting
Auto-completion
Auto-indentation
Initial code or document boilerplate
Tie-in to help systems
IDE-like features (compile, debug, and so on)

Prepared exclusively for Zach

 84 C HAPTER 3 T HE B ASIC T OOLS

Figure 3.1. Sorting lines in an editor

import java.util.Vector; emacs: M-x sort-lines import java.awt.*;

import java.util.Stack; import java.net.URL;
import java.net.URL; import java.util.Stack;
import java.awt.*; import java.util.Vector;
vi: :.,+3!sort

A feature such as syntax highlighting may sound like a frivolous extra,

but in reality it can be very useful and enhance your productivity. Once
you get used to seeing keywords appear in a different color or font, a
mistyped keyword that doesn’t appear that way jumps out at you long
before you fire up the compiler.

Having the ability to compile and navigate directly to errors within the
editor environment is very handy on big projects. Emacs in particular
is adept at this style of interaction.

Productivity
A surprising number of people we’ve met use the Windows notepad
utility to edit their source code. This is like using a teaspoon as a
shovel—simply typing and using basic mouse-based cut and paste is
not enough.

What sort of things will you need to do that can’t be done in this way?

Well, there’s cursor movement, to start with. Single keystrokes that

move you in units of words, lines, blocks, or functions are far more
efficient than repeatedly typing a keystroke that moves you character
by character or line by line.

Or suppose you are writing Java code. You like to keep your import
statements in alphabetical order, and someone else has checked in a
few files that don’t adhere to this standard (this may sound extreme,
but on a large project it can save you a lot of time scanning through a
long list of import statements). You’d like to go quickly through a few
files and sort a small section of them. In editors such as vi and Emacs
you can do this easily (see Figure 3.1). Try that in notepad.

Some editors can help streamline common operations. For instance,

when you create a new file in a particular language, the editor can
supply a template for you. It might include:

Prepared exclusively for Zach

 P OWER E DITING 85

Name of the class or module filled in (derived from the filename)

Your name and/or copyright statements

Skeletons for constructs in that language (constructor and destruc-

tor declarations, for example)

Another useful feature is auto-indenting. Rather than having to indent

manually (by using space or tab), the editor automatically indents for
you at the appropriate time (after typing an open brace, for example).
The nice part about this feature is that you can use the editor to provide
a consistent indentation style for your project.5

Where to Go from Here

This sort of advice is particularly hard to write because virtually every
reader is at a different level of comfort and expertise with the editor(s)
they are currently using. So, to summarize, and to provide some guid-
ance on where to go next, find yourself in the left-hand column of the
chart, and look at the right-hand column to see what we think you
should do.

If this sounds like you Then think about

I use only basic features of Pick a powerful editor and learn it

many different editors. well.

I have a favorite editor, but I Learn them. Cut down the number of
don’t use all of its features. keystrokes you need to type.

I have a favorite editor and Try to expand and use it for more
use it where possible. tasks than you do already.

I think you are nuts. Notepad As long as you are happy and produc-
is the best editor ever made. tive, go for it! But if you find yourself
subject to “editor envy,” you may need
to reevaluate your position.

5. The Linux kernel is developed this way. Here you have geographically dispersed
developers, many working on the same pieces of code. There is a published list of settings
(in this case, for Emacs) that describes the required indentation style.

Prepared exclusively for Zach

 86 C HAPTER 3 T HE B ASIC T OOLS

What Editors Are Available?

Having recommended that you master a decent editor, which one do
we recommend? Well, we’re going to duck that question; your choice of
editor is a personal one (some would even say a religious one!). However,
in Appendix A, page 266, we list a number of popular editors and where
to get them.

Challenges
Some editors use full-blown languages for customization and scripting.
Emacs, for example, uses Lisp. As one of the new languages you are going
to learn this year, learn the language your editor uses. For anything you
find yourself doing repeatedly, develop a set of macros (or equivalent) to
handle it.

Do you know everything your editor is capable of doing? Try to stump your
colleagues who use the same editor. Try to accomplish any given editing
task in as few keystrokes as possible.

17 Source Code Control

Progress, far from consisting in change, depends on retentiveness. Those who
cannot remember the past are condemned to repeat it.
George Santayana, Life of Reason

One of the important things we look for in a user interface is the UNDO
key—a single button that forgives us our mistakes. It’s even better if the
environment supports multiple levels of undo and redo, so you can go
back and recover from something that happened a couple of minutes
ago. But what if the mistake happened last week, and you’ve turned
your computer on and off ten times since then? Well, that’s one of the
many benefits of using a source code control system: it’s a giant UNDO
key—a project-wide time machine that can return you to those halcyon
days of last week, when the code actually compiled and ran.

Source code control systems, or the more widely scoped configuration

management systems, keep track of every change you make in your
source code and documentation. The better ones can keep track of

Prepared exclusively for Zach

 S OURCE C ODE C ONTROL 87

compiler and OS versions as well. With a properly configured source

code control system, you can always go back to a previous version of
your software.

But a source code control system (SCCS6 ) does far more than undo
mistakes. A good SCCS will let you track changes, answering questions
such as: Who made changes in this line of code? What’s the difference
between the current version and last week’s? How many lines of code
did we change in this release? Which files get changed most often? This
kind of information is invaluable for bug-tracking, audit, performance,
and quality purposes.

An SCCS will also let you identify releases of your software. Once iden-
tified, you will always be able to go back and regenerate the release,
independent of changes that may have occurred later.

We often use an SCCS to manage branches in the development tree. For

example, once you have released some software, you’ll normally want
to continue developing for the next release. At the same time, you’ll
need to deal with bugs in the current release, shipping fixed versions
to clients. You’ll want these bug fixes rolled into the next release (if
appropriate), but you don’t want to ship code under development to
clients. With an SCCS you can generate branches in the development
tree each time you generate a release. You apply bug fixes to code in
the branch, and continue developing on the main trunk. Since the bug
fixes may be relevant to the main trunk as well, some systems allow you
to merge selected changes from the branch back into the main trunk
automatically.

Source code control systems may keep the files they maintain in a cen-
tral repository—a great candidate for archiving.

Finally, some products may allow two or more users to be working

concurrently on the same set of files, even making concurrent changes
in the same file. The system then manages the merging of these changes
when the files are sent back to the repository. Although seemingly risky,
such systems work well in practice on projects of all sizes.

6. We use the uppercase SCCS to refer to generic source code control systems. There
is also a specific system called “sccs,” originally released with AT&T System V Unix.

Prepared exclusively for Zach

 88 C HAPTER 3 T HE B ASIC T OOLS

TIP 23

Always Use Source Code Control

Always. Even if you are a single-person team on a one-week project.

Even if it’s a “throw-away” prototype. Even if the stuff you’re work-
ing on isn’t source code. Make sure that everything is under source
code control—documentation, phone number lists, memos to vendors,
makefiles, build and release procedures, that little shell script that
burns the CD master—everything. We routinely use source code con-
trol on just about everything we type (including the text of this book).
Even if we’re not working on a project, our day-to-day work is secured
in a repository.

Source Code Control and Builds

There is a tremendous hidden benefit in having an entire project under
the umbrella of a source code control system: you can have product
builds that are automatic and repeatable.

The project build mechanism can pull the latest source out of the repos-
itory automatically. It can run in the middle of the night after everyone’s
(hopefully) gone home. You can run automatic regression tests to en-
sure that the day’s coding didn’t break anything. The automation of
the build ensures consistency—there are no manual procedures, and
you won’t need developers remembering to copy code into some special
build area.

The build is repeatable because you can always rebuild the source as
it existed on a given date.

But My Team Isn’t Using Source Code Control

Shame on them! Sounds like an opportunity to do some evangelizing!
However, while you wait for them to see the light, perhaps you should
implement your own private source control. Use one of the freely avail-
able tools we list in Appendix A, and make a point of keeping your
personal work safely tucked into a repository (as well as doing what-
ever your project requires). Although this may seem to be duplication
of effort, we can pretty much guarantee it will save you grief (and save
your project money) the first time you need to answer questions such

Prepared exclusively for Zach

 S OURCE C ODE C ONTROL 89

as “What did you do to the xyz module?” and “What broke the build?”
This approach may also help convince your management that source
code control really works.

Don’t forget that an SCCS is equally applicable to the things you do

outside of work.

Source Code Control Products

Appendix A, page 271, gives URLs for representative source code control
systems, some commercial and others freely available. And many more
products are available—look for pointers to the configuration manage-
ment FAQ. For an introduction to the freely-available CVS version con-
trol system, see our book Pragmatic Version Control [TH03].

Related sections include:

Orthogonality, page 34
The Power of Plain Text, page 73
It’s All Writing, page 248

Challenges
Even if you are not able to use an SCCS at work, install RCS or CVS on a
personal system. Use it to manage your pet projects, documents you write,
and (possibly) configuration changes applied to the computer system itself.

Take a look at some of the Open Source projects for which publicly ac-
cessible archives are available on the Web (such as Mozilla [URL 51], KDE
[URL 54], and the Gimp [URL 55]). How do you get updates of the source?
How do you make changes—does the project regulate access or arbitrate
the inclusion of changes?

Prepared exclusively for Zach

 90 C HAPTER 3 T HE B ASIC T OOLS

18 Debugging
It is a painful thing
To look at your own trouble and know
That you yourself and no one else has made it
Sophocles, Ajax

The word bug has been used to describe an “object of terror” ever since
the fourteenth century. Rear Admiral Dr. Grace Hopper, the inventor
of COBOL, is credited with observing the first computer bug—literally,
a moth caught in a relay in an early computer system. When asked
to explain why the machine wasn’t behaving as intended, a technician
reported that there was “a bug in the system,” and dutifully taped it—
wings and all—into the log book.

Regrettably, we still have “bugs” in the system, albeit not the flying
kind. But the fourteenth century meaning—a bogeyman—is perhaps
even more applicable now than it was then. Software defects mani-
fest themselves in a variety of ways, from misunderstood requirements
to coding errors. Unfortunately, modern computer systems are still
limited to doing what you tell them to do, not necessarily what you
want them to do.

No one writes perfect software, so it’s a given that debugging will take
up a major portion of your day. Let’s look at some of the issues involved
in debugging and some general strategies for finding elusive bugs.

Psychology of Debugging
Debugging itself is a sensitive, emotional subject for many developers.
Instead of attacking it as a puzzle to be solved, you may encounter
denial, finger pointing, lame excuses, or just plain apathy.

Embrace the fact that debugging is just problem solving, and attack it
as such.

Having found someone else’s bug, you can spend time and energy lay-
ing blame on the filthy culprit who created it. In some workplaces this
is part of the culture, and may be cathartic. However, in the technical
arena, you want to concentrate on fixing the problem, not the blame.

Prepared exclusively for Zach

 D EBUGGING 91

TIP 24

Fix the Problem, Not the Blame

It doesn’t really matter whether the bug is your fault or someone else’s.
It is still your problem.

A Debugging Mindset
The easiest person to deceive is one’s self.
Edward Bulwer-Lytton, The Disowned

Before you start debugging, it’s important to adopt the right mindset.
You need to turn off many of the defenses you use each day to protect
your ego, tune out any project pressures you may be under, and get
yourself comfortable. Above all, remember the first rule of debugging:

TIP 25

Don’t Panic

It’s easy to get into a panic, especially if you are facing a deadline, or
have a nervous boss or client breathing down your neck while you are
trying to find the cause of the bug. But it is very important to step back
a pace, and actually think about what could be causing the symptoms
that you believe indicate a bug.

If your first reaction on witnessing a bug or seeing a bug report is

“that’s impossible,” you are plainly wrong. Don’t waste a single neuron
on the train of thought that begins “but that can’t happen” because
quite clearly it can, and has.

Beware of myopia when debugging. Resist the urge to fix just the symp-
toms you see: it is more likely that the actual fault may be several steps
removed from what you are observing, and may involve a number of
other related things. Always try to discover the root cause of a problem,
not just this particular appearance of it.

Where to Start
Before you start to look at the bug, make sure that you are work-
ing on code that compiled cleanly—without warnings. We routinely set

Prepared exclusively for Zach

 92 C HAPTER 3 T HE B ASIC T OOLS

compiler warning levels as high as possible. It doesn’t make sense to

waste time trying to find a problem that the compiler could find for you!
We need to concentrate on the harder problems at hand.

When trying to solve any problem, you need to gather all the relevant
data. Unfortunately, bug reporting isn’t an exact science. It’s easy to be
misled by coincidences, and you can’t afford to waste time debugging
coincidences. You first need to be accurate in your observations.

Accuracy in bug reports is further diminished when they come through

a third party—you may actually need to watch the user who reported
the bug in action to get a sufficient level of detail.

Andy once worked on a large graphics application. Nearing release, the

testers reported that the application crashed every time they painted
a stroke with a particular brush. The programmer responsible argued
that there was nothing wrong with it; he had tried painting with it, and
it worked just fine. This dialog went back and forth for several days,
with tempers rapidly rising.

Finally, we got them together in the same room. The tester selected
the brush tool and painted a stroke from the upper right corner to the
lower left corner. The application exploded. "Oh," said the programmer,
in a small voice, who then sheepishly admitted that he had made test
strokes only from the lower left to the upper right, which did not expose
the bug.

There are two points to this story:

You may need to interview the user who reported the bug in order
to gather more data than you were initially given.

Artificial tests (such as the programmer’s single brush stroke from

bottom to top) don’t exercise enough of an application. You must
brutally test both boundary conditions and realistic end-user usage
patterns. You need to do this systematically (see Ruthless Testing,
page 237).

Debugging Strategies
Once you think you know what is going on, it’s time to find out what
the program thinks is going on.

Prepared exclusively for Zach

 D EBUGGING 93

Bug Reproduction
No, our bugs aren’t really multiplying (although some of them are
probably old enough to do it legally). We’re talking about a different
kind of reproduction.
The best way to start fixing a bug is to make it reproducible. After all,
if you can’t reproduce it, how will you know if it is ever fixed?
But we want more than a bug that can be reproduced by following
some long series of steps; we want a bug that can be reproduced
with a single command. It’s a lot harder to fix a bug if you have to go
through 15 steps to get to the point where the bug shows up. Some-
times by forcing yourself to isolate the circumstances that display the
bug, you’ll even gain an insight on how to fix it.
See Ubiquitous Automation, page 230, for other ideas along these
lines.

Visualize Your Data

Often, the easiest way to discern what a program is doing—or what it
is going to do—is to get a good look at the data it is operating on. The
simplest example of this is a straightforward “variable name = data
value” approach, which may be implemented as printed text, or as
fields in a GUI dialog box or list.

But you can gain a much deeper insight into your data by using a
debugger that allows you to visualize your data and all of the inter-
relationships that exist. There are debuggers that can represent your
data as a 3D fly-over through a virtual reality landscape, or as a 3D
waveform plot, or just as simple structural diagrams, as shown in Fig-
ure 3.2 on the next page. As you single-step through your program,
pictures like these can be worth much more than a thousand words,
as the bug you’ve been hunting suddenly jumps out at you.

Even if your debugger has limited support for visualizing data, you
can still do it yourself—either by hand, with paper and pencil, or with
external plotting programs.

The DDD debugger has some visualization capabilities, and is freely

available (see [URL 19]). It is interesting to note that DDD works with

Prepared exclusively for Zach

 94 C HAPTER 3 T HE B ASIC T OOLS

Figure 3.2. Sample debugger diagram of a circular linked list. The arrows repre-
sent pointers to nodes.

1: list
(List *)
0x804db40

next
value = 85 value = 86
self = 0x804db40 self = 0x804db50
next = 0x804db50 next = 0x804db40
next

multiple languages, including Ada, C, C++, Fortran, Java, Modula, Pas-

cal, Perl, and Python (clearly an orthogonal design).

Tracing
Debuggers generally focus on the state of the program now. Sometimes
you need more—you need to watch the state of a program or a data
structure over time. Seeing a stack trace can only tell you how you got
here directly. It can’t tell you what you were doing prior to this call
chain, especially in event-based systems.

Tracing statements are those little diagnostic messages you print to

the screen or to a file that say things such as “got here” and “value of
x = 2.” It’s a primitive technique compared with IDE-style debuggers,
but it is peculiarly effective at diagnosing several classes of errors that
debuggers can’t. Tracing is invaluable in any system where time itself
is a factor: concurrent processes, real-time systems, and event-based
applications.

You can use tracing statements to “drill down” into the code. That is,
you can add tracing statements as you descend the call tree.

Trace messages should be in a regular, consistent format; you may

want to parse them automatically. For instance, if you needed to track
down a resource leak (such as unbalanced file opens/closes), you could
trace each open and each close in a log file. By processing the log

Prepared exclusively for Zach

 D EBUGGING 95

Corrupt Variables? Check Their Neighborhood

Sometimes you’ll examine a variable, expecting to see a small integer
value, and instead get something like 0x6e69614d. Before you roll
up your sleeves for some serious debugging, have a quick look at the
memory around this corrupted variable. Often it will give you a clue.
In our case, examining the surrounding memory as characters shows
us
20333231 6e69614d 2c745320 746f4e0a
1 2 3 M a i n S t , n N o t
2c6e776f 2058580a 31323433 00000a33
o w n , n X X 3 4 2 1 3 n 0 0

Looks like someone sprayed a street address over our counter. Now
we know where to look.

file with Perl, you could easily identify where the offending open was
occurring.

Rubber Ducking
A very simple but particularly useful technique for finding the cause
of a problem is simply to explain it to someone else. The other person
should look over your shoulder at the screen, and nod his or her head
constantly (like a rubber duck bobbing up and down in a bathtub).
They do not need to say a word; the simple act of explaining, step by
step, what the code is supposed to do often causes the problem to leap
off the screen and announce itself.7

It sounds simple, but in explaining the problem to another person you

must explicitly state things that you may take for granted when go-
ing through the code yourself. By having to verbalize some of these
assumptions, you may suddenly gain new insight into the problem.

7. Why “rubber ducking”? While an undergraduate at Imperial College in London, Dave

did a lot of work with a research assistant named Greg Pugh, one of the best developers
Dave has known. For several months Greg carried around a small yellow rubber duck,
which he’d place on his terminal while coding. It was a while before Dave had the courage
to ask. . . .

Prepared exclusively for Zach

 96 C HAPTER 3 T HE B ASIC T OOLS

Process of Elimination
In most projects, the code you are debugging may be a mixture of appli-
cation code written by you and others on your project team, third-party
products (database, connectivity, graphical libraries, specialized com-
munications or algorithms, and so on) and the platform environment
(operating system, system libraries, and compilers).

It is possible that a bug exists in the OS, the compiler, or a third-party

product—but this should not be your first thought. It is much more
likely that the bug exists in the application code under development.
It is generally more profitable to assume that the application code is
incorrectly calling into a library than to assume that the library itself is
broken. Even if the problem does lie with a third party, you’ll still have
to eliminate your code before submitting the bug report.

We worked on a project where a senior engineer was convinced that the

select system call was broken on Solaris. No amount of persuasion or
logic could change his mind (the fact that every other networking appli-
cation on the box worked fine was irrelevant). He spent weeks writing
work-arounds, which, for some odd reason, didn’t seem to fix the prob-
lem. When finally forced to sit down and read the documentation on
select, he discovered the problem and corrected it in a matter of min-
utes. We now use the phrase “select is broken” as a gentle reminder
whenever one of us starts blaming the system for a fault that is likely
to be our own.

TIP 26

“select” Isn’t Broken

Remember, if you see hoof prints, think horses—not zebras. The OS is

probably not broken. And the database is probably just fine.

If you “changed only one thing” and the system stopped working, that
one thing was likely to be responsible, directly or indirectly, no matter
how farfetched it seems. Sometimes the thing that changed is outside of
your control: new versions of the OS, compiler, database, or other third-
party software can wreak havoc with previously correct code. New bugs
might show up. Bugs for which you had a work-around get fixed, break-
ing the work-around. APIs change, functionality changes; in short, it’s
a whole new ball game, and you must retest the system under these

Prepared exclusively for Zach

 D EBUGGING 97

new conditions. So keep a close eye on the schedule when considering

an upgrade; you may want to wait until after the next release.

If, however, you have no obvious place to start looking, you can always
rely on a good old-fashioned binary search. See if the symptoms are
present at either of two far away spots in the code. Then look in the
middle. If the problem is present, then the bug lies between the start
and the middle point; otherwise, it is between the middle point and the
end. You can continue in this fashion until you narrow down the spot
sufficiently to identify the problem.

The Element of Surprise

When you find yourself surprised by a bug (perhaps even muttering
“that’s impossible” under your breath where we can’t hear you), you
must reevaluate truths you hold dear. In that linked list routine—the
one you knew was bulletproof and couldn’t possibly be the cause of this
bug—did you test all the boundary conditions? That other piece of code
you’ve been using for years—it couldn’t possibly still have a bug in it.
Could it?

Of course it can. The amount of surprise you feel when something goes
wrong is directly proportional to the amount of trust and faith you have
in the code being run. That’s why, when faced with a “surprising” fail-
ure, you must realize that one or more of your assumptions is wrong.
Don’t gloss over a routine or piece of code involved in the bug because
you “know” it works. Prove it. Prove it in this context, with this data,
with these boundary conditions.

TIP 27

Don’t Assume It—Prove It

When you come across a surprise bug, beyond merely fixing it, you need
to determine why this failure wasn’t caught earlier. Consider whether
you need to amend the unit or other tests so that they would have
caught it.

Also, if the bug is the result of bad data that was propagated through
a couple of levels before causing the explosion, see if better parame-
ter checking in those routines would have isolated it earlier (see the

Prepared exclusively for Zach

 98 C HAPTER 3 T HE B ASIC T OOLS

discussions on crashing early and assertions on pages 120 and 122,

respectively).

While you’re at it, are there any other places in the code that may be
susceptible to this same bug? Now is the time to find and fix them.
Make sure that whatever happened, you’ll know if it happens again.

If it took a long time to fix this bug, ask yourself why. Is there anything
you can do to make fixing this bug easier the next time around? Per-
haps you could build in better testing hooks, or write a log file analyzer.

Finally, if the bug is the result of someone’s wrong assumption, discuss

the problem with the whole team: if one person misunderstands, then
it’s possible many people do.

Do all this, and hopefully you won’t be surprised next time.

Debugging Checklist
Is the problem being reported a direct result of the underlying bug,
or merely a symptom?

Is the bug really in the compiler? Is it in the OS? Or is it in your

code?

If you explained this problem in detail to a coworker, what would

you say?

If the suspect code passes its unit tests, are the tests complete
enough? What happens if you run the unit test with this data?

Do the conditions that caused this bug exist anywhere else in the
system?

Related sections include:

Assertive Programming, page 122
Programming by Coincidence, page 172
Ubiquitous Automation, page 230
Ruthless Testing, page 237

Challenges
Debugging is challenge enough.

Prepared exclusively for Zach

 T EXT M ANIPULATION 99

19 Text Manipulation
Pragmatic Programmers manipulate text the same way woodworkers
shape wood. In previous sections we discussed some specific tools—
shells, editors, debuggers—that we use. These are similar to a wood-
worker’s chisels, saws, and planes—tools specialized to do one or two
jobs well. However, every now and then we need to perform some trans-
formation not readily handled by the basic tool set. We need a general-
purpose text manipulation tool.

Text manipulation languages are to programming what routers8 are

to woodworking. They are noisy, messy, and somewhat brute force.
Make mistakes with them, and entire pieces can be ruined. Some peo-
ple swear they have no place in the toolbox. But in the right hands,
both routers and text manipulation languages can be incredibly pow-
erful and versatile. You can quickly trim something into shape, make
joints, and carve. Used properly, these tools have surprising finesse
and subtlety. But they take time to master.

There is a growing number of good text manipulation languages. Unix

developers often like to use the power of their command shells, aug-
mented with tools such as awk and sed. People who prefer a more
structured tool like the object-oriented nature of Python [URL 9]. Some
people use Tcl [URL 23] as their tool of choice. We happen to prefer Ruby
[TFH04] and Perl [URL 8] for hacking out short scripts.

These languages are important enabling technologies. Using them, you

can quickly hack up utilities and prototype ideas—jobs that might take
five or ten times as long using conventional languages. And that mul-
tiplying factor is crucially important to the kind of experimenting that
we do. Spending 30 minutes trying out a crazy idea is a whole lot bet-
ter than spending five hours. Spending a day automating important
components of a project is acceptable; spending a week might not be.
In their book The Practice of Programming [KP99], Kernighan and Pike
built the same program in five different languages. The Perl version
was the shortest (17 lines, compared with C’s 150). With Perl you can

8. Here router means the tool that spins cutting blades very, very fast, not a device for
interconnecting networks.

Prepared exclusively for Zach

 100 C HAPTER 3 T HE B ASIC T OOLS

manipulate text, interact with programs, talk over networks, drive Web
pages, perform arbitrary precision arithmetic, and write programs that
look like Snoopy swearing.

TIP 28

Learn a Text Manipulation Language

To show the wide-ranging applicability of text manipulation languages,

here’s a sample of some applications we’ve developed over the last few
years.

Database schema maintenance. A set of Perl scripts took a plain

text file containing a database schema definition and from it gen-
erated:

– The SQL statements to create the database

– Flat data files to populate a data dictionary
– C code libraries to access the database
– Scripts to check database integrity
– Web pages containing schema descriptions and diagrams
– An XML version of the schema

Java property access. It is good OO programming style to restrict

access to an object’s properties, forcing external classes to get and
set them via methods. However, in the common case where a prop-
erty is represented inside the class by a simple member variable,
creating a get and set method for each variable is tedious and me-
chanical. We have a Perl script that modifies the source files and
inserts the correct method definitions for all appropriately flagged
variables.

Test data generation. We had tens of thousands of records of test

data, spread over several different files and formats, that needed to
be knitted together and converted into a form suitable for loading
into a relational database. Perl did it in a couple of hours (and in
the process found a couple of consistency errors in the original
data).

Book writing. We think it is important that any code presented

in a book should have been tested first. Most of the code in this

Prepared exclusively for Zach

 T EXT M ANIPULATION 101

book has been. However, using the DRY principle (see The Evils of
Duplication, page 26) we didn’t want to copy and paste lines of code
from the tested programs into the book. That would have meant
that the code was duplicated, virtually guaranteeing that we’d for-
get to update an example when the corresponding program was
changed. For some examples, we also didn’t want to bore you with
all the framework code needed to make our example compile and
run. We turned to Perl. A relatively simple script is invoked when
we format the book—it extracts a named segment of a source file,
does syntax highlighting, and converts the result into the typeset-
ting language we use.

C to Object Pascal interface. A client had a team of developers

writing Object Pascal on PCs. Their code needed to interface to a
body of code written in C. We developed a short Perl script that
parsed the C header files, extracting the definitions of all exported
functions and the data structures they used. We then generated
Object Pascal units with Pascal records for all the C structures,
and imported procedure definitions for all the C functions. This
generation process became part of the build, so that whenever the
C header changed, a new Object Pascal unit would be constructed
automatically.

Generating Web documentation. Many project teams are pub-

lishing their documentation to internal Web sites. We have writ-
ten many Perl programs that analyze database schemas, C or C++
source files, makefiles, and other project sources to produce the
required HTML documentation. We also use Perl to wrap the docu-
ments with standard headers and footers, and to transfer them to
the Web site.

We use text manipulation languages almost every day. Many of the

ideas in this book can be implemented more simply in them than in
any other language of which we’re aware. These languages make it easy
to write code generators, which we’ll look at next.

Related sections include:

The Evils of Duplication, page 26

Prepared exclusively for Zach

 102 C HAPTER 3 T HE B ASIC T OOLS

Exercises
Answer 11. Your C program uses an enumerated type to represent one of 100 states.
on p. 285
You’d like to be able to print out the state as a string (as opposed to a
number) for debugging purposes. Write a script that reads from standard
input a file containing
name
state_a
state_b
: :

Produce the file name.h, which contains

extern const char* NAME_names[];
typedef enum {
state_a,
state_b,
: :
} NAME;

and the file name.c, which contains

const char* NAME_names[] = {

"state_a",
"state_b",
: :
};

Answer 12. Halfway through writing this book, we realized that we hadn’t put the use
on p. 286
strict directive into many of our Perl examples. Write a script that goes
through the .pl files in a directory and adds a use strict at the end of
the initial comment block to all files that don’t already have one. Remember
to keep a backup of all files you change.

20 Code Generators
When woodworkers are faced with the task of producing the same thing
over and over, they cheat. They build themselves a jig or a template. If
they get the jig right once, they can reproduce a piece of work time after
time. The jig takes away complexity and reduces the chances of making
mistakes, leaving the craftsman free to concentrate on quality.

As programmers, we often find ourselves in a similar position. We need

to achieve the same functionality, but in different contexts. We need to
repeat information in different places. Sometimes we just need to pro-
tect ourselves from carpal tunnel syndrome by cutting down on repeti-
tive typing.

Prepared exclusively for Zach

 C ODE G ENERATORS 103

In the same way a woodworker invests the time in a jig, a programmer

can build a code generator. Once built, it can be used throughout the
life of the project at virtually no cost.

TIP 29

Write Code That Writes Code

There are two main types of code generators:

1. Passive code generators are run once to produce a result. From

that point forward, the result becomes freestanding—it is divorced
from the code generator. The wizards discussed in Evil Wizards,
page 198, along with some CASE tools, are examples of passive
code generators.

2. Active code generators are used each time their results are required.
The result is a throw-away—it can always be reproduced by the
code generator. Often, active code generators read some form of
script or control file to produce their results.

Passive Code Generators

Passive code generators save typing. They are basically parameterized
templates, generating a given output from a set of inputs. Once the
result is produced, it becomes a full-fledged source file in the project; it
will be edited, compiled, and placed under source control just like any
other file. Its origins will be forgotten.

Passive code generators have many uses:

Creating new source files. A passive code generator can produce

templates, source code control directives, copyright notices, and
standard comment blocks for each new file in a project. We have
our editors set up to do this whenever we create a new file: edit
a new Java program, and the new editor buffer will automatically
contain a comment block, package directive, and the outline class
declaration, already filled in.

Performing one-off conversions among programming languages. We

started writing this book using the troff system, but we switched to
LATEX after 15 sections had been completed. We wrote a code gen-
erator that read the troff source and converted it to LATEX. It was

Prepared exclusively for Zach

 104 C HAPTER 3 T HE B ASIC T OOLS

about 90% accurate; the rest we did by hand. This is an interest-

ing feature of passive code generators: they don’t have to be totally
accurate. You get to choose how much effort you put into the gen-
erator, compared with the energy you spend fixing up its output.

Producing lookup tables and other resources that are expensive to

compute at runtime. Instead of calculating trigonometric functions,
many early graphics systems used precomputed tables of sine and
cosine values. Typically, these tables were produced by a passive
code generator and then copied into the source.

Active Code Generators

While passive code generators are simply a convenience, their active
cousins are a necessity if you want to follow the DRY principle. With
an active code generator, you can take a single representation of some
piece of knowledge and convert it into all the forms your application
needs. This is not duplication, because the derived forms are dispos-
able, and are generated as needed by the code generator (hence the
word active).

Whenever you find yourself trying to get two disparate environments to

work together, you should consider using active code generators.

Perhaps you’re developing a database application. Here, you’re dealing

with two environments—the database and the programming language
you are using to access it. You have a schema, and you need to define
low-level structures mirroring the layout of certain database tables.
You could just code these directly, but this violates the DRY princi-
ple: knowledge of the schema would then be expressed in two places.
When the schema changes, you need to remember to change the corre-
sponding code. If a column is removed from a table, but the code base
is not changed, you might not even get a compilation error. The first
you’ll know about it is when your tests start failing (or when the user
calls).

An alternative is to use an active code generator—take the schema and

use it to generate the source code for the structures, as shown in Figure
3.3. Now, whenever the schema changes, the code used to access it also
changes, automatically. If a column is removed, then its corresponding
field in the structure will disappear, and any higher-level code that uses
that column will fail to compile. You’ve caught the error at compile time,

Prepared exclusively for Zach

 C ODE G ENERATORS 105

Figure 3.3. Active code generator creates code from a database schema

struct EmployeeRow
Schema
table employee active
struct EmployerRow
table employer code
table benefit generator
struct BenefitRow

not in production. Of course, this scheme works only if you make the
code generation part of the build process itself.9

Another example of melding environments using code generators hap-

pens when different programming languages are used in the same
application. In order to communicate, each code base will need some
information in common—data structures, message formats, and field
names, for example. Rather than duplicate this information, use a code
generator. Sometimes you can parse the information out of the source
files of one language and use it to generate code in a second language.
Often, though, it is simpler to express it in a simpler, language-neutral
representation and generate the code for both languages, as shown in
Figure 3.4 on the following page. Also see the answer to Exercise 13 on
page 286 for an example of how to separate the parsing of the flat file
representation from code generation.

Code Generators Needn’t Be Complex

All this talk of active this and passive that may leave you with the
impression that code generators are complex beasts. They needn’t be.
Normally the most complex part is the parser, which analyzes the input
file. Keep the input format simple, and the code generator becomes

9. Just how do you go about building code from a database schema? There are several
ways. If the schema is held in a flat file (for example, as create table statements), then
a relatively simple script can parse it and generate the source. Alternatively, if you use a
tool to create the schema directly in the database, then you should be able to extract the
information you need directly from the database’s data dictionary. Perl provides libraries
that give you access to most major databases.

Prepared exclusively for Zach

 106 C HAPTER 3 T HE B ASIC T OOLS

Figure 3.4. Generating code from a language-neutral representation. In the input

file, lines starting with ‘M’ flag the start of a message definition, ‘F’ lines define
fields, and ‘E’ is the end of the message.

# Add a product
# to the ’on-order’ list
M AddProduct
F id int
F name char[30]
F order_code int
E
gen
C era
te te
nera P as
ge cal

/* Add a product */ { Add a product }

/* to the ’on-order’ list */ { to the ’on-order’ list }
typedef struct { AddProductMsg = packed record
int id; id: LongInt;
char name[30]; name: array[0..29] of char;
int order_code; order_code: LongInt;
} AddProductMsg; end;

simple. Have a look at the answer to Exercise 13 (page 286): the actual
code generation is basically print statements.

Code Generators Needn’t Generate Code

Although many of the examples in this section show code generators
that produce program source, this needn’t always be the case. You can
use code generators to write just about any output: HTML, XML, plain
text—any text that might be an input somewhere else in your project.

Related sections include:

The Evils of Duplication, page 26
The Power of Plain Text, page 73
Evil Wizards, page 198
Ubiquitous Automation, page 230

Exercises
Answer 13. Write a code generator that takes the input file in Figure 3.4, and generates
on p. 286
output in two languages of your choice. Try to make it easy to add new
languages.

Prepared exclusively for Zach

 Chapter 4

Pragmatic Paranoia
TIP 30

You Can’t Write Perfect Software

Did that hurt? It shouldn’t. Accept it as an axiom of life. Embrace it.

Celebrate it. Because perfect software doesn’t exist. No one in the brief
history of computing has ever written a piece of perfect software. It’s
unlikely that you’ll be the first. And unless you accept this as a fact,
you’ll end up wasting time and energy chasing an impossible dream.

So, given this depressing reality, how does a Pragmatic Programmer

turn it into an advantage? That’s the topic of this chapter.

Everyone knows that they personally are the only good driver on Earth.
The rest of the world is out there to get them, blowing through stop
signs, weaving between lanes, not indicating turns, talking on the
telephone, reading the paper, and just generally not living up to our
standards. So we drive defensively. We look out for trouble before it
happens, anticipate the unexpected, and never put ourselves into a
position from which we can’t extricate ourselves.

The analogy with coding is pretty obvious. We are constantly interfac-

ing with other people’s code—code that might not live up to our high
standards—and dealing with inputs that may or may not be valid. So
we are taught to code defensively. If there’s any doubt, we validate all
information we’re given. We use assertions to detect bad data. We check
for consistency, put constraints on database columns, and generally
feel pretty good about ourselves.

107

Prepared exclusively for Zach

 108 C HAPTER 4 P RAGMATIC P ARANOIA

But Pragmatic Programmers take this a step further. They don’t trust
themselves, either. Knowing that no one writes perfect code, includ-
ing themselves, Pragmatic Programmers code in defenses against their
own mistakes. We describe the first defensive measure in Design by
Contract: clients and suppliers must agree on rights and responsibili-
ties.

In Dead Programs Tell No Lies, we want to ensure that we do no damage

while we’re working the bugs out. So we try to check things often and
terminate the program if things go awry.

Assertive Programming describes an easy method of checking along the

way—write code that actively verifies your assumptions.

Exceptions, like any other technique, can cause more harm than good
if not used properly. We’ll discuss the issues in When to Use Exceptions.

As your programs get more dynamic, you’ll find yourself juggling sys-
tem resources—memory, files, devices, and the like. In How to Balance
Resources, we’ll suggest ways of ensuring that you don’t drop any of
the balls.

In a world of imperfect systems, ridiculous time scales, laughable tools,

and impossible requirements, let’s play it safe.

When everybody actually is out to get you, paranoia is just good thinking.
Woody Allen

Prepared exclusively for Zach

 D ESIGN BY C ONTRACT 109

21 Design by Contract
Nothing astonishes men so much as common sense and plain dealing.
Ralph Waldo Emerson, Essays

Dealing with computer systems is hard. Dealing with people is even

harder. But as a species, we’ve had longer to figure out issues of human
interactions. Some of the solutions we’ve come up with during the last
few millennia can be applied to writing software as well. One of the best
solutions for ensuring plain dealing is the contract.

A contract defines your rights and responsibilities, as well as those of

the other party. In addition, there is an agreement concerning reper-
cussions if either party fails to abide by the contract.

Maybe you have an employment contract that specifies the hours you’ll
work and the rules of conduct you must follow. In return, the company
pays you a salary and other perks. Each party meets its obligations and
everyone benefits.

It’s an idea used the world over—both formally and informally—to help
humans interact. Can we use the same concept to help software mod-
ules interact? The answer is “yes.”

DBC
Bertrand Meyer [Mey97b] developed the concept of Design by Contract
for the language Eiffel.1 It is a simple yet powerful technique that
focuses on documenting (and agreeing to) the rights and responsibil-
ities of software modules to ensure program correctness. What is a
correct program? One that does no more and no less than it claims
to do. Documenting and verifying that claim is the heart of Design by
Contract (DBC, for short).

Every function and method in a software system does something. Be-

fore it starts that something, the routine may have some expectation of
the state of the world, and it may be able to make a statement about
the state of the world when it concludes. Meyer describes these expec-
tations and claims as follows:

1. Based in part on earlier work by Dijkstra, Floyd, Hoare, Wirth, and others. For more
information on Eiffel itself, see [URL 10] and [URL 11].

Prepared exclusively for Zach

 110 C HAPTER 4 P RAGMATIC P ARANOIA

Preconditions. What must be true in order for the routine to be

called; the routine’s requirements. A routine should never get called
when its preconditions would be violated. It is the caller’s respon-
sibility to pass good data (see the box on page 115).

Postconditions. What the routine is guaranteed to do; the state of

the world when the routine is done. The fact that the routine has
a postcondition implies that it will conclude: infinite loops aren’t
allowed.

Class invariants. A class ensures that this condition is always

true from the perspective of a caller. During internal processing of
a routine, the invariant may not hold, but by the time the routine
exits and control returns to the caller, the invariant must be true.
(Note that a class cannot give unrestricted write-access to any data
member that participates in the invariant.)

Let’s look at the contract for a routine that inserts a data value into
a unique, ordered list. In iContract, a preprocessor for Java available
from [URL 17], you’d specify it as

/**
* @invariant forall Node n in elements() |
* n.prev() != null
* implies
* n.value().compareTo(n.prev().value()) > 0
*/
public class dbc_list {
/**
* @pre contains(aNode) == false
* @post contains(aNode) == true
*/
public void insertNode(final Node aNode) {
// ...

Here we are saying that nodes in this list must always be in increas-
ing order. When you insert a new node, it can’t exist already, and we
guarantee that the node will be found after you have inserted it.

You write these preconditions, postconditions, and invariants in the

target programming language, perhaps with some extensions. For ex-
ample, iContract provides predicate logic operators—forall, exists,
and implies—in addition to normal Java constructs. Your assertions
can query the state of any object that the method can access, but be
sure that the query is free from any side effects (see page 124).

Prepared exclusively for Zach

 D ESIGN BY C ONTRACT 111

DBC and Constant Parameters

Often, a postcondition will use parameters passed into a method to
verify correct behavior. But if the routine is allowed to change the
parameter that’s passed in, you might be able to circumvent the con-
tract. Eiffel doesn’t allow this to happen, but Java does. Here, we use
the Java keyword final to indicate our intentions that the param-
eter shouldn’t be changed within the method. This isn’t foolproof—
subclasses are free to redeclare the parameter as non-final. Alterna-
tively, you can use the iContract syntax variable@pre to get the
original value of the variable as it existed on entry to the method.

The contract between a routine and any potential caller can thus be
read as
If all the routine’s preconditions are met by the caller, the routine shall
guarantee that all postconditions and invariants will be true when it
completes.

If either party fails to live up to the terms of the contract, then a rem-
edy (which was previously agreed to) is invoked—an exception is raised,
or the program terminates, for instance. Whatever happens, make no
mistake that failure to live up to the contract is a bug. It is not some-
thing that should ever happen, which is why preconditions should not
be used to perform things such as user-input validation.

TIP 31

Design with Contracts

In Orthogonality, page 34, we recommended writing “shy” code. Here,

the emphasis is on “lazy” code: be strict in what you will accept before
you begin, and promise as little as possible in return. Remember, if your
contract indicates that you’ll accept anything and promise the world in
return, then you’ve got a lot of code to write!

Inheritance and polymorphism are the cornerstones of object-oriented

languages and an area where contracts can really shine. Suppose you
are using inheritance to create an “is-a-kind-of” relationship, where one
class “is-a-kind-of” another class. You probably want to adhere to the
Liskov Substitution Principle [Lis88]:

Prepared exclusively for Zach

 112 C HAPTER 4 P RAGMATIC P ARANOIA

Subclasses must be usable through the base class interface without the
need for the user to know the difference.

In other words, you want to make sure that the new subtype you have
created really “is-a-kind-of” the base type—that it supports the same
methods, and that the methods have the same meaning. We can do
this with contracts. We need to specify a contract only once, in the base
class, to have it applied to every future subclass automatically. A sub-
class may, optionally, accept a wider range of input, or make stronger
guarantees. But it must accept at least as much, and guarantee as
much, as its parent.

For example, consider the Java base class java.awt.Component. You

can treat any visual component in AWT or Swing as a Component, with-
out knowing that the actual subclass is a button, a canvas, a menu, or
whatever. Each individual component can provide additional, specific
functionality, but it has to provide at least the basic capabilities de-
fined by Component. But there’s nothing to prevent you from creating
a subtype of Component that provides correctly named methods that
do the wrong thing. You can easily create a paint method that doesn’t
paint, or a setFont method that doesn’t set the font. AWT doesn’t have
contracts to catch the fact that you didn’t live up to the agreement.

Without a contract, all the compiler can do is ensure that a subclass

conforms to a particular method signature. But if we put a base class
contract in place, we can now ensure that any future subclass can’t
alter the meanings of our methods. For instance, you might want to
establish a contract for setFont such as the following, which ensures
that the font you set is the font you get:

/**
* @pre f != null
* @post getFont() == f
*/
public void setFont(final Font f) {
// ...

Implementing DBC
The greatest benefit of using DBC may be that it forces the issue of
requirements and guarantees to the forefront. Simply enumerating at
design time what the input domain range is, what the boundary con-
ditions are, and what the routine promises to deliver—or, more im-

Prepared exclusively for Zach

 D ESIGN BY C ONTRACT 113

portantly, what it doesn’t promise to deliver—is a huge leap forward in

writing better software. By not stating these things, you are back to pro-
gramming by coincidence (see page 172), which is where many projects
start, finish, and fail.

In languages that do not support DBC in the code, this might be as

far as you can go—and that’s not too bad. DBC is, after all, a design
technique. Even without automatic checking, you can put the contract
in the code as comments and still get a very real benefit. If nothing else,
the commented contracts give you a place to start looking when trouble
strikes.

Assertions
While documenting these assumptions is a great start, you can get
much greater benefit by having the compiler check your contract for
you. You can partially emulate this in some languages by using asser-
tions (see Assertive Programming, page 122). Why only partially? Can’t
you use assertions to do everything DBC can do?

Unfortunately, the answer is no. To begin with, there is no support

for propagating assertions down an inheritance hierarchy. This means
that if you override a base class method that has a contract, the asser-
tions that implement that contract will not be called correctly (unless
you duplicate them manually in the new code). You must remember to
call the class invariant (and all base class invariants) manually before
you exit every method. The basic problem is that the contract is not
automatically enforced.

Also, there is no built-in concept of “old” values; that is, values as

they existed at the entry to a method. If you’re using assertions to
enforce contracts, you must add code to the precondition to save any
information you’ll want to use in the postcondition. Compare this with
iContract, where the postcondition can just reference “variable@pre,”
or with Eiffel, which supports “old expression.”

Finally, the runtime system and libraries are not designed to support
contracts, so these calls are not checked. This is a big loss, because
it is often at the boundary between your code and the libraries it uses
that the most problems are detected (see Dead Programs Tell No Lies,
page 120 for a more detailed discussion).

Prepared exclusively for Zach

 114 C HAPTER 4 P RAGMATIC P ARANOIA

Language Support
Languages that feature built-in support of DBC (such as Eiffel and
Sather [URL 12]) check pre- and postconditions automatically in the
compiler and runtime system. You get the greatest benefit in this case
because all of the code base (libraries, too) must honor their contracts.

But what about more popular languages such as C, C++, and Java?
For these languages, there are preprocessors that process contracts
embedded in the original source code as special comments. The pre-
processor expands these comments to code that verifies the assertions.

For C and C++, you may want to investigate Nana [URL 18]. Nana doesn’t
handle inheritance, but it does use the debugger at runtime to monitor
assertions in a novel way.

For Java, there is iContract [URL 17]. It takes comments (in JavaDoc
form) and generates a new source file with the assertion logic included.

Preprocessors aren’t as good as a built-in facility. They can be messy to

integrate into your project, and other libraries you use won’t have con-
tracts. But they can still be very helpful; when a problem is discovered
this way—especially one that you would never have found—it’s almost
like magic.

DBC and Crashing Early

DBC fits in nicely with our concept of crashing early (see Dead Pro-
grams Tell No Lies, page 120). Suppose you have a method that cal-
culates square roots (such as in the Eiffel class DOUBLE). It needs a
precondition that restricts the domain to positive numbers. An Eiffel
precondition is declared with the keyword require, and a postcondi-
tion is declared with ensure, so you could write

sqrt: DOUBLE is
-- Square root routine
require
sqrt_arg_must_be_positive: Current >= 0;
--- ...
--- calculate square root here
--- ...
ensure
((Result*Result) - Current).abs <= epsilon*Current.abs;
-- Result should be within error tolerance
end;

Prepared exclusively for Zach

 D ESIGN BY C ONTRACT 115

Who’s Responsible?
Who is responsible for checking the precondition, the caller or the
routine being called? When implemented as part of the language, the
answer is neither: the precondition is tested behind the scenes after
the caller invokes the routine but before the routine itself is entered.
Thus if there is any explicit checking of parameters to be done, it
must be performed by the caller, because the routine itself will never
see parameters that violate its precondition. (For languages without
built-in support, you would need to bracket the called routine with a
preamble and/or postamble that checks these assertions.)
Consider a program that reads a number from the console, calcu-
lates its square root (by calling sqrt), and prints the result. The sqrt
function has a precondition—its argument must not be negative. If the
user enters a negative number at the console, it is up to the calling
code to ensure that it never gets passed to sqrt. This calling code
has many options: it could terminate, it could issue a warning and
read another number, or it could make the number positive and ap-
pend an “ ” to the result returned by sqrt. Whatever its choice, this
is definitely not sqrt’s problem.
By expressing the domain of the square root function in the precon-
dition of the sqrt routine, you shift the burden of correctness to the
caller—where it belongs. You can then design the sqrt routine se-
cure in the knowledge that its input will be in range.

If your algorithm for calculating the square root fails (or isn’t within the
specified error tolerance), you get an error message and a stack trace
to show you the call chain.

If you pass sqrt a negative parameter, the Eiffel runtime prints the
error “sqrt_arg_must_be_positive,” along with a stack trace. This
is better than the alternative in languages such as Java, C, and C++,
where passing a negative number to sqrt returns the special value
NaN (Not a Number). It may be some time later in the program that you
attempt to do some math on NaN, with surprising results.

It’s much easier to find and diagnose the problem by crashing early, at
the site of the problem.

Prepared exclusively for Zach

 116 C HAPTER 4 P RAGMATIC P ARANOIA

Other Uses of Invariants

So far we have discussed pre- and postconditions that apply to individ-
ual methods and invariants that apply to all methods within a class,
but there are other useful ways to use invariants.

Loop Invariants
Getting the boundary conditions right on a nontrivial loop can be prob-
lematic. Loops are subject to the banana problem (I know how to spell
“banana,” but I don’t know when to stop), fencepost errors (not know-
ing whether to count the fenceposts or the spaces between them), and
the ubiquitous “off by one” error [URL 52].

Invariants can help in these situations: a loop invariant is a statement

of the eventual goal of a loop, but is generalized so that it is also valid
before the loop executes and on each iteration through the loop. You
can think of it as a kind of miniature contract. The classic example is
a routine that finds the maximum value in an array.

int m = arr[0]; // example assumes arr.length > 0

int i = 1;
// Loop invariant: m = max(arr[0:i-1])
while (i < arr.length) {
m = Math.max(m, arr[i]);
i = i + 1;
}

(arr[m:n] is a notational convenience meaning a slice of the array from

index m to n.) The invariant must be true before the loop runs, and the
body of the loop must ensure that it remains true as the loop executes.
In this way we know that the invariant also holds when the loop termi-
nates, and therefore that our result is valid. Loop invariants can be
coded explicitly as assertions, but they are also useful as design and
documentation tools.

Semantic Invariants
You can use semantic invariants to express inviolate requirements, a
kind of “philosophical contract.”

We once wrote a debit card transaction switch. A major requirement

was that the user of a debit card should never have the same trans-
action applied to their account twice. In other words, no matter what

Prepared exclusively for Zach

 D ESIGN BY C ONTRACT 117

sort of failure mode might happen, the error should be on the side of
not processing a transaction rather than processing a duplicate trans-
action.

This simple law, driven directly from the requirements, proved to be

very helpful in sorting out complex error recovery scenarios, and guided
the detailed design and implementation in many areas.

Be sure not to confuse requirements that are fixed, inviolate laws with
those that are merely policies that might change with a new manage-
ment regime. That’s why we use the term semantic invariants—it must
be central to the very meaning of a thing, and not subject to the whims
of policy (which is what more dynamic business rules are for).

When you find a requirement that qualifies, make sure it becomes

a well-known part of whatever documentation you are producing—
whether it is a bulleted list in the requirements document that gets
signed in triplicate or just a big note on the common whiteboard that
everyone sees. Try to state it clearly and unambiguously. For example,
in the debit card example, we might write

E RR IN FAVOR OF THE CONSUMER .

This is a clear, concise, unambiguous statement that’s applicable in

many different areas of the system. It is our contract with all users of
the system, our guarantee of behavior.

Dynamic Contracts and Agents

Until now, we have talked about contracts as fixed, immutable specifi-
cations. But in the landscape of autonomous agents, this doesn’t need
to be the case. By the definition of “autonomous,” agents are free to
reject requests that they do not want to honor. They are free to renego-
tiate the contract—“I can’t provide that, but if you give me this, then I
might provide something else.”

Certainly any system that relies on agent technology has a critical de-
pendence on contractual arrangements—even if they are dynamically
generated.

Imagine: with enough components and agents that can negotiate their
own contracts among themselves to achieve a goal, we might just solve
the software productivity crisis by letting software solve it for us.

Prepared exclusively for Zach

 118 C HAPTER 4 P RAGMATIC P ARANOIA

But if we can’t use contracts by hand, we won’t be able to use them

automatically. So next time you design a piece of software, design its
contract as well.

Related sections include:

Orthogonality, page 34
Dead Programs Tell No Lies, page 120
Assertive Programming, page 122
How to Balance Resources, page 129
Decoupling and the Law of Demeter, page 138
Temporal Coupling, page 150
Programming by Coincidence, page 172
Code That’s Easy to Test, page 189
Pragmatic Teams, page 224

Challenges
Points to ponder: If DBC is so powerful, why isn’t it used more widely? Is
it hard to come up with the contract? Does it make you think about issues
you’d rather ignore for now? Does it force you to THINK!? Clearly, this is a
dangerous tool!

Exercises
Answer 14. What makes a good contract? Anyone can add preconditions and postcon-
on p. 288
ditions, but will they do you any good? Worse yet, will they actually do
more harm than good? For the example below and for those in Exercises
15 and 16, decide whether the specified contract is good, bad, or ugly, and
explain why.

First, let’s look at an Eiffel example. Here we have a routine for adding a
STRING to a doubly linked, circular list (remember that preconditions are
labeled with require, and postconditions with ensure).

-- Add a unique item to a doubly linked list,

-- and return the newly created NODE.
add_item (item : STRING) : NODE is
require
item /= Void -- ’/=’ is ’not equal’.
find_item(item) = Void -- Must be unique
deferred -- Abstract base class.
ensure
result.next.previous = result -- Check the newly
result.previous.next = result -- added node’s links.
find_item(item) = result -- Should find it.
end

Prepared exclusively for Zach

 D ESIGN BY C ONTRACT 119

15. Next, let’s try an example in Java—somewhat similar to the example in Answer
on p. 288
Exercise 14. insertNumber inserts an integer into an ordered list. Pre-
and postconditions are labeled as in iContract (see [URL 17]).

private int data[];

/**
* @post data[index-1] < data[index] &&
* data[index] == aValue
*/
public Node insertNumber (final int aValue)
{
int index = findPlaceToInsert(aValue);
...

16. Here’s a fragment from a stack class in Java. Is this a good contract? Answer
on p. 289
/**
* @pre anItem != null // Require real data
* @post pop() == anItem // Verify that it’s
* // on the stack
*/
public void push(final String anItem)

17. The classic examples of DBC (as in Exercises 14–16) show an implemen- Answer
on p. 289
tation of an ADT (Abstract Data Type)—typically a stack or queue. But not
many people really write these kinds of low-level classes.

So, for this exercise, design an interface to a kitchen blender. It will even-
tually be a Web-based, Internet-enabled, CORBA-fied blender, but for now
we just need the interface to control it. It has ten speed settings (0 means
off). You can’t operate it empty, and you can change the speed only one
unit at a time (that is, from 0 to 1, and from 1 to 2, not from 0 to 2).

Here are the methods. Add appropriate pre- and postconditions and an
invariant.
int getSpeed()
void setSpeed(int x)
boolean isFull()
void fill()
void empty()

18. How many numbers are in the series ? Answer

on p. 290

Prepared exclusively for Zach

 120 C HAPTER 4 P RAGMATIC P ARANOIA

22 Dead Programs Tell No Lies

Have you noticed that sometimes other people can detect that things
aren’t well with you before you’re aware of the problem yourself? It’s the
same with other people’s code. If something is starting to go awry with
one of our programs, sometimes it is a library routine that catches it
first. Maybe a stray pointer has caused us to overwrite a file handle with
something meaningless. The next call to read will catch it. Perhaps a
buffer overrun has trashed a counter we’re about to use to determine
how much memory to allocate. Maybe we’ll get a failure from malloc. A
logic error a couple of million instructions ago means that the selector
for a case statement is no longer the expected 1, 2, or 3. We’ll hit the
default case (which is one reason why each and every case/switch
statement needs to have a default clause—we want to know when the
“impossible” has happened).

It’s easy to fall into the “it can’t happen” mentality. Most of us have
written code that didn’t check that a file closed successfully, or that a
trace statement got written as we expected. And all things being equal,
it’s likely that we didn’t need to—the code in question wouldn’t fail
under any normal conditions. But we’re coding defensively. We’re look-
ing for rogue pointers in other parts of our program trashing the stack.
We’re checking that the correct versions of shared libraries were actu-
ally loaded.

All errors give you information. You could convince yourself that the
error can’t happen, and choose to ignore it. Instead, Pragmatic Pro-
grammers tell themselves that if there is an error, something very, very
bad has happened.

TIP 32

Crash Early

Crash, Don’t Trash

One of the benefits of detecting problems as soon as you can is that you
can crash earlier. And many times, crashing your program is the best
thing you can do. The alternative may be to continue, writing corrupted

Prepared exclusively for Zach

 D EAD P ROGRAMS T ELL N O L IES 121

data to some vital database or commanding the washing machine into

its twentieth consecutive spin cycle.

The Java language and libraries have embraced this philosophy. When
something unexpected happens within the runtime system, it throws
a RuntimeException. If not caught, this will percolate up to the top
level of the program and cause it to halt, displaying a stack trace.

You can do the same in other languages. If you don’t have an exception
mechanism, or if your libraries don’t throw exceptions, then make sure
you handle the errors yourself. In C, macros can be very useful for this:
#define CHECK(LINE, EXPECTED)
{ int rc = LINE;
if (rc != EXPECTED)
ut_abort(__FILE__, __LINE__, #LINE, rc, EXPECTED); }

void ut_abort(char *file, int ln, char *line, int rc, int exp) {
fprintf(stderr, "%s line %d n’%s’: expected %d, got %d n",
file, ln, line, exp, rc);
exit(1);
}

Then you can wrap calls that should never fail using
CHECK(stat("/tmp", &stat_buff), 0);

If it should fail, you’d get a message written to stderr:

source.c line 19
’stat("/tmp", &stat_buff)’: expected 0, got -1

Clearly it is sometimes inappropriate simply to exit a running program.

You may have claimed resources that might not get released, or you
may need to write log messages, tidy up open transactions, or inter-
act with other processes. The techniques we discuss in When to Use
Exceptions, page 125, will help here. However, the basic principle stays
the same—when your code discovers that something that was supposed
to be impossible just happened, your program is no longer viable. Any-
thing it does from this point forward becomes suspect, so terminate it
as soon as possible. A dead program normally does a lot less damage
than a crippled one.

Related sections include:

Design by Contract, page 109
When to Use Exceptions, page 125

Prepared exclusively for Zach

 122 C HAPTER 4 P RAGMATIC P ARANOIA

23 Assertive Programming
There is a luxury in self-reproach. When we blame ourselves we feel no one else
has a right to blame us.
Oscar Wilde, The Picture of Dorian Gray

It seems that there’s a mantra that every programmer must memorize

early in his or her career. It is a fundamental tenet of computing, a core
belief that we learn to apply to requirements, designs, code, comments,
just about everything we do. It goes

T HIS CAN NEVER HAPPEN

“This code won’t be used 30 years from now, so two-digit dates are fine.”
“This application will never be used abroad, so why internationalize it?”
“count can’t be negative.” “This printf can’t fail.”

Let’s not practice this kind of self-deception, particularly when coding.

TIP 33

If It Can’t Happen, Use Assertions to Ensure That It Won’t

Whenever you find yourself thinking “but of course that could never
happen,” add code to check it. The easiest way to do this is with asser-
tions. In most C and C++ implementations, you’ll find some form of
assert or _assert macro that checks a Boolean condition. These
macros can be invaluable. If a pointer passed in to your procedure
should never be NULL, then check for it:
void writeString(char *string) {
assert(string != NULL);
...

Assertions are also useful checks on an algorithm’s operation. Maybe

you’ve written a clever sort algorithm. Check that it works:

for (int i = 0; i < num_entries-1; i++) {

assert(sorted[i] <= sorted[i+1]);
}

Of course, the condition passed to an assertion should not have a side

effect (see the box on page 124). Also remember that assertions may be
turned off at compile time—never put code that must be executed into
an assert.

Prepared exclusively for Zach

 A SSER TIVE P ROGRAMMING 123

Don’t use assertions in place of real error handling. Assertions check

for things that should never happen: you don’t want to be writing code
such as
printf("Enter ’Y’ or ’N’: ");
ch = getchar();
assert((ch == ’Y’) || (ch == ’N’)); /* bad idea! */

And just because the supplied assert macros call exit when an as-
sertion fails, there’s no reason why versions you write should. If you
need to free resources, have an assertion failure generate an exception,
longjmp to an exit point, or call an error handler. Just make sure the
code you execute in those dying milliseconds doesn’t rely on the infor-
mation that triggered the assertion failure in the first place.

Leave Assertions Turned On

There is a common misunderstanding about assertions, promulgated
by the people who write compilers and language environments. It goes
something like this:
Assertions add some overhead to code. Because they check for things
that should never happen, they’ll get triggered only by a bug in the code.
Once the code has been tested and shipped, they are no longer needed,
and should be turned off to make the code run faster. Assertions are a
debugging facility.

There are two patently wrong assumptions here. First, they assume
that testing finds all the bugs. In reality, for any complex program you
are unlikely to test even a miniscule percentage of the permutations
your code will be put through (see Ruthless Testing, page 245). Second,
the optimists are forgetting that your program runs in a dangerous
world. During testing, rats probably won’t gnaw through a communi-
cations cable, someone playing a game won’t exhaust memory, and log
files won’t fill the hard drive. These things might happen when your
program runs in a production environment. Your first line of defense is
checking for any possible error, and your second is using assertions to
try to detect those you’ve missed.

Turning off assertions when you deliver a program to production is like

crossing a high wire without a net because you once made it across in
practice. There’s dramatic value, but it’s hard to get life insurance.

Even if you do have performance issues, turn off only those assertions
that really hit you. The sort example above may be a critical part of

Prepared exclusively for Zach

 124 C HAPTER 4 P RAGMATIC P ARANOIA

Assertions and Side Effects

It is embarrassing when the code we add to detect errors actually
ends up creating new errors. This can happen with assertions if eval-
uating the condition has side effects. For example, in Java it would be
a bad idea to code something such as
while (iter.hasMoreElements()) {
Test.ASSERT(iter.nextElement() != null);
Object obj = iter.nextElement();
// ....
}

The .nextElement() call in the ASSERT has the side effect of

moving the iterator past the element being fetched, and so the loop
will process only half the elements in the collection. It would be better
to write
while (iter.hasMoreElements()) {
Object obj = iter.nextElement();
Test.ASSERT(obj != null);
// ....
}

This problem is a kind of “Heisenbug”—debugging that changes the

behavior of the system being debugged (see [URL 52]).

your application, and may need to be fast. Adding the check means
another pass through the data, which might be unacceptable. Make
that particular check optional,2 but leave the rest in.

Related sections include:

Debugging, page 90
Design by Contract, page 109
How to Balance Resources, page 129
Programming by Coincidence, page 172

2. In C-based languages, you can either use the preprocessor or use if statements to
make assertions optional. Many implementations turn off code generation for the assert
macro if a compile-time flag is set (or not set). Otherwise, you can place the code within an
if statement with a constant condition, which many compilers (including most common
Java systems) will optimize away.

Prepared exclusively for Zach

 W HEN TO U SE E XCEPTIONS 125

Exercises
19. A quick reality check. Which of these “impossible” things can happen? Answer
on p. 290
1. A month with fewer than 28 days
2. stat(".",&sb) == -1 (that is, can’t access the current directory)
3. In C++: a = 2; b = 3; if (a + b != 5) exit(1);
4. A triangle with an interior angle sum
5. A minute that doesn’t have 60 seconds
6. In Java: (a + 1) <= a

20. Develop a simple assertion checking class for Java. Answer

on p. 291

24 When to Use Exceptions

In Dead Programs Tell No Lies, page 120, we suggested that it is good
practice to check for every possible error—particularly the unexpected
ones. However, in practice this can lead to some pretty ugly code; the
normal logic of your program can end up being totally obscured by error
handling, particularly if you subscribe to the “a routine must have a
single return statement” school of programming (we don’t). We’ve seen
code that looks something like the following:
retcode = OK;
if (socket.read(name) != OK) {
retcode = BAD_READ;
}
else {
processName(name);
if (socket.read(address) != OK) {
retcode = BAD_READ;
}
else {
processAddress(address);
if (socket.read(telNo) != OK) {
retcode = BAD_READ;
}
else {
// etc, etc...
}
}
}
return retcode;

Fortunately, if the programming language supports exceptions, you can

rewrite this code in a far neater way:

Prepared exclusively for Zach

 126 C HAPTER 4 P RAGMATIC P ARANOIA

retcode = OK;
try {
socket.read(name);
process(name);
socket.read(address);
processAddress(address);
socket.read(telNo);
// etc, etc...
}
catch (IOException e) {
retcode = BAD_READ;
Logger.log("Error reading individual: " + e.getMessage());
}
return retcode;

The normal flow of control is now clear, with all the error handling
moved off to a single place.

What Is Exceptional?
One of the problems with exceptions is knowing when to use them. We
believe that exceptions should rarely be used as part of a program’s
normal flow; exceptions should be reserved for unexpected events.
Assume that an uncaught exception will terminate your program and
ask yourself, “Will this code still run if I remove all the exception han-
dlers?” If the answer is “no,” then maybe exceptions are being used in
nonexceptional circumstances.

For example, if your code tries to open a file for reading and that file
does not exist, should an exception be raised?

Our answer is, “It depends.” If the file should have been there, then an
exception is warranted. Something unexpected happened—a file you
were expecting to exist seems to have disappeared. On the other hand,
if you have no idea whether the file should exist or not, then it doesn’t
seem exceptional if you can’t find it, and an error return is appropriate.

Let’s look at an example of the first case. The following code opens the
file /etc/passwd, which should exist on all Unix systems. If it fails, it
passes on the FileNotFoundException to its caller.

public void open_passwd() throws FileNotFoundException {

// This may throw FileNotFoundException...
ipstream = new FileInputStream("/etc/passwd");
// ...
}

Prepared exclusively for Zach

 W HEN TO U SE E XCEPTIONS 127

However, the second case may involve opening a file specified by the
user on the command line. Here an exception isn’t warranted, and the
code looks different:
public boolean open_user_file(String name)
throws FileNotFoundException {
File f = new File(name);
if (!f.exists()) {
return false;
}
ipstream = new FileInputStream(f);
return true;
}

Note that the FileInputStream call can still generate an exception,

which the routine passes on. However, the exception will be generated
under only truly exceptional circumstances; simply trying to open a file
that does not exist will generate a conventional error return.

TIP 34

Use Exceptions for Exceptional Problems

Why do we suggest this approach to exceptions? Well, an exception

represents an immediate, nonlocal transfer of control—it’s a kind of
cascading goto. Programs that use exceptions as part of their normal
processing suffer from all the readability and maintainability problems
of classic spaghetti code. These programs break encapsulation: rou-
tines and their callers are more tightly coupled via exception handling.

Error Handlers Are an Alternative

An error handler is a routine that is called when an error is detected.
You can register a routine to handle a specific category of errors. When
one of these errors occurs, the handler will be called.

There are times when you may want to use error handlers, either in-
stead of or alongside exceptions. Clearly, if you are using a language
such as C, which does not support exceptions, this is one of your few
other options (see the challenge on the next page). However, sometimes
error handlers can be used even in languages (such as Java) that have
a good exception handling scheme built in.

Prepared exclusively for Zach

 128 C HAPTER 4 P RAGMATIC P ARANOIA

Consider the implementation of a client-server application, using Java’s

Remote Method Invocation (RMI) facility. Because of the way RMI is
implemented, every call to a remote routine must be prepared to han-
dle a RemoteException. Adding code to handle these exceptions can
become tedious, and means that it is difficult to write code that works
with both local and remote routines. A possible work-around is to wrap
your remote objects in a class that is not remote. This class then imple-
ments an error handler interface, allowing the client code to register a
routine to be called when a remote exception is detected.

Related sections include:

Dead Programs Tell No Lies, page 120

Challenges
Languages that do not support exceptions often have some other nonlocal
transfer of control mechanism (C has longjmp/setjmp, for example). Con-
sider how you could implement some kind of ersatz exception mechanism
using these facilities. What are the benefits and dangers? What special
steps do you need to take to ensure that resources are not orphaned?
Does it make sense to use this kind of solution whenever you code in C?

Exercises
Answer 21. While designing a new container class, you identify the following possible
on p. 292
error conditions:

1. No memory available for a new element in the add routine

2. Requested entry not found in the fetch routine

3. null pointer passed to the add routine

How should each be handled? Should an error be generated, should an

exception be raised, or should the condition be ignored?

Prepared exclusively for Zach

 H OW TO B ALANCE R ESOURCES 129

25 How to Balance Resources

“I brought you into this world,” my father would say, “and I can take you out. It
don’t make no difference to me. I’ll just make another one like you.”
Bill Cosby, Fatherhood

We all manage resources whenever we code: memory, transactions,

threads, files, timers—all kinds of things with limited availability. Most
of the time, resource usage follows a predictable pattern: you allocate
the resource, use it, and then deallocate it.

However, many developers have no consistent plan for dealing with re-
source allocation and deallocation. So let us suggest a simple tip:

TIP 35

Finish What You Start

This tip is easy to apply in most circumstances. It simply means that

the routine or object that allocates a resource should be responsible for
deallocating it. Let’s see how it applies by looking at an example of some
bad code—an application that opens a file, reads customer information
from it, updates a field, and writes the result back. We’ve eliminated
error handling to make the example clearer.

void readCustomer(const char *fName, Customer *cRec) {

cFile = fopen(fName, "r+");
fread(cRec, sizeof(*cRec), 1, cFile);
}
void writeCustomer(Customer *cRec) {
rewind(cFile);
fwrite(cRec, sizeof(*cRec), 1, cFile);
fclose(cFile);
}
void updateCustomer(const char *fName, double newBalance) {
Customer cRec;
readCustomer(fName, &cRec);
cRec.balance = newBalance;
writeCustomer(&cRec);
}

At first sight, the routine updateCustomer looks pretty good. It seems

to implement the logic we require—reading a record, updating the bal-
ance, and writing the record back out. However, this tidiness hides a

Prepared exclusively for Zach

 130 C HAPTER 4 P RAGMATIC P ARANOIA

major problem. The routines readCustomer and writeCustomer are

tightly coupled3 —they share the global variable cFile. readCustomer
opens the file and stores the file pointer in cFile, and writeCustomer
uses that stored pointer to close the file when it finishes. This global
variable doesn’t even appear in the updateCustomer routine.

Why is this bad? Let’s consider the unfortunate maintenance program-

mer who is told that the specification has changed—the balance should
be updated only if the new value is not negative. She goes into the
source and changes updateCustomer:

void updateCustomer(const char *fName, double newBalance) {

Customer cRec;
readCustomer(fName, &cRec);
if (newBalance >= 0.0) {
cRec.balance = newBalance;
writeCustomer(&cRec);
}
}

All seems fine during testing. However, when the code goes into produc-
tion, it collapses after several hours, complaining of too many open files.
Because writeCustomer is not getting called in some circumstances,
the file is not getting closed.

A very bad solution to this problem would be to deal with the special
case in updateCustomer:
void updateCustomer(const char *fName, double newBalance) {
Customer cRec;
readCustomer(fName, &cRec);
if (newBalance >= 0.0) {
cRec.balance = newBalance;
writeCustomer(&cRec);
}
else
fclose(cFile);
}

This will fix the problem—the file will now get closed regardless of the
new balance—but the fix now means that three routines are coupled
through the global cFile. We’re falling into a trap, and things are going
to start going downhill rapidly if we continue on this course.

3. For a discussion of the dangers of coupled code, see Decoupling and the Law of
Demeter, page 138.

Prepared exclusively for Zach

 H OW TO B ALANCE R ESOURCES 131

The finish what you start tip tells us that, ideally, the routine that allo-
cates a resource should also free it. We can apply it here by refactoring
the code slightly:

void readCustomer(FILE *cFile, Customer *cRec) {

fread(cRec, sizeof(*cRec), 1, cFile);
}
void writeCustomer(FILE *cFile, Customer *cRec) {
rewind(cFile);
fwrite(cRec, sizeof(*cRec), 1, cFile);
}
void updateCustomer(const char *fName, double newBalance) {
FILE *cFile;
Customer cRec;
cFile = fopen(fName, "r+"); // >---
readCustomer(cFile, &cRec); // |
if (newBalance >= 0.0) { // |
cRec.balance = newBalance; // |
writeCustomer(cFile, &cRec); // |
} // |
fclose(cFile); // <---
}

Now all the responsibility for the file is in the updateCustomer routine.
It opens the file and (finishing what it starts) closes it before exiting. The
routine balances the use of the file: the open and close are in the same
place, and it is apparent that for every open there will be a correspond-
ing close. The refactoring also removes an ugly global variable.

Nest Allocations
The basic pattern for resource allocation can be extended for routines
that need more than one resource at a time. There are just two more
suggestions:

1. Deallocate resources in the opposite order to that in which you

allocate them. That way you won’t orphan resources if one resource
contains references to another.

2. When allocating the same set of resources in different places in

your code, always allocate them in the same order. This will reduce
the possibility of deadlock. (If process A claims resource1 and is
about to claim resource2, while process B has claimed resource2
and is trying to get resource1, the two processes will wait forever.)

It doesn’t matter what kind of resources we’re using—transactions,

memory, files, threads, windows—the basic pattern applies: whoever

Prepared exclusively for Zach

 132 C HAPTER 4 P RAGMATIC P ARANOIA

allocates a resource should be responsible for deallocating it. However,

in some languages we can develop the concept further.

Objects and Exceptions

The equilibrium between allocations and deallocations is reminiscent of
a class’s constructor and destructor. The class represents a resource,
the constructor gives you a particular object of that resource type, and
the destructor removes it from your scope.

If you are programming in an object-oriented language, you may find

it useful to encapsulate resources in classes. Each time you need a
particular resource type, you instantiate an object of that class. When
the object goes out of scope, or is reclaimed by the garbage collector,
the object’s destructor then deallocates the wrapped resource.

This approach has particular benefits when you’re working with lan-
guages such as C++, where exceptions can interfere with resource deal-
location.

Balancing and Exceptions

Languages that support exceptions can make resource deallocation
tricky. If an exception is thrown, how do you guarantee that every-
thing allocated prior to the exception is tidied up? The answer depends
to some extent on the language.

Balancing Resources with C++ Exceptions

C++ supports a try catch exception mechanism. Unfortunately, this
means that there are always at least two possible paths when exiting a
routine that catches and then rethrows an exception:

void doSomething(void) {
Node *n = new Node;
try {
// do something
}
catch (...) {
delete n;
throw;
}
delete n;
}

Prepared exclusively for Zach

 H OW TO B ALANCE R ESOURCES 133

Notice that the node we create is freed in two places—once in the rou-
tine’s normal exit path, and once in the exception handler. This is an
obvious violation of the DRY principle and a maintenance problem wait-
ing to happen.

However, we can use the semantics of C++ to our advantage. Local

objects are automatically destroyed on exiting from their enclosing
block. This gives us a couple of options. If the circumstances permit,
we can change “n” from a pointer to an actual Node object on the stack:
void doSomething1(void) {
Node n;
try {
// do something
}
catch (...) {
throw;
}
}

Here we rely on C++ to handle the destruction of the Node object auto-
matically, whether an exception is thrown or not.

If the switch from a pointer is not possible, the same effect can be
achieved by wrapping the resource (in this case, a Node pointer) within
another class.
// Wrapper class for Node resources
class NodeResource {
Node *n;
public:
NodeResource() { n = new Node; }
~NodeResource() { delete n; }
Node *operator->() { return n; }
};
void doSomething2(void) {
NodeResource n;
try {
// do something
}
catch (...) {
throw;
}
}

Now the wrapper class, NodeResource, ensures that when its objects
are destroyed the corresponding nodes are also destroyed. For conve-
nience, the wrapper provides a dereferencing operator ->, so that its
users can access the fields in the contained Node object directly.

Prepared exclusively for Zach

 134 C HAPTER 4 P RAGMATIC P ARANOIA

Because this technique is so useful, the standard C++ library provides

the template class auto_ptr, which gives you automatic wrappers for
dynamically allocated objects.
void doSomething3(void) {
auto_ptr<Node> p (new Node);
// Access the Node as p->...
// Node automatically deleted at end
}

Balancing Resources in Java

Unlike C++, Java implements a lazy form of automatic object destruc-
tion. Unreferenced objects are considered to be candidates for garbage
collection, and their finalize method will get called should garbage
collection ever claim them. While a convenience for developers, who
no longer get the blame for most memory leaks, it makes it difficult
to implement resource clean-up using the C++ scheme. Fortunately,
the designers of the Java language thoughtfully added a language fea-
ture to compensate, the finally clause. When a try block contains
a finally clause, code in that clause is guaranteed to be executed if
any statement in the try block is executed. It doesn’t matter whether
an exception is thrown (or even if the code in the try block executes a
return)—the code in the finally clause will get run. This means we
can balance our resource usage with code such as
public void doSomething() throws IOException {
File tmpFile = new File(tmpFileName);
FileWriter tmp = new FileWriter(tmpFile);
try {
// do some work
}
finally {
tmpFile.delete();
}
}

The routine uses a temporary file, which we want to delete, regardless

of how the routine exits. The finally block allows us to express this
concisely.

When You Can’t Balance Resources

There are times when the basic resource allocation pattern just isn’t
appropriate. Commonly this is found in programs that use dynamic

Prepared exclusively for Zach

 H OW TO B ALANCE R ESOURCES 135

data structures. One routine will allocate an area of memory and link
it into some larger structure, where it may stay for some time.

The trick here is to establish a semantic invariant for memory alloca-

tion. You need to decide who is responsible for data in an aggregate data
structure. What happens when you deallocate the top-level structure?
You have three main options:

1. The top-level structure is also responsible for freeing any substruc-

tures that it contains. These structures then recursively delete data
they contain, and so on.

2. The top-level structure is simply deallocated. Any structures that

it pointed to (that are not referenced elsewhere) are orphaned.

3. The top-level structure refuses to deallocate itself if it contains any

substructures.

The choice here depends on the circumstances of each individual data

structure. However, you need to make it explicit for each, and imple-
ment your decision consistently. Implementing any of these options in
a procedural language such as C can be a problem: data structures
themselves are not active. Our preference in these circumstances is to
write a module for each major structure that provides standard alloca-
tion and deallocation facilities for that structure. (This module can also
provide facilities such as debug printing, serialization, deserialization,
and traversal hooks.)

Finally, if keeping track of resources gets tricky, you can write your
own form of limited automatic garbage collection by implementing a
reference counting scheme on your dynamically allocated objects. The
book More Effective C++ [Mey96] dedicates a section to this topic.

Checking the Balance

Because Pragmatic Programmers trust no one, including ourselves, we
feel that it is always a good idea to build code that actually checks that
resources are indeed freed appropriately. For most applications, this
normally means producing wrappers for each type of resource, and us-
ing these wrappers to keep track of all allocations and deallocations.
At certain points in your code, the program logic will dictate that the
resources will be in a certain state: use the wrappers to check this.

Prepared exclusively for Zach

 136 C HAPTER 4 P RAGMATIC P ARANOIA

For example, a long-running program that services requests will prob-

ably have a single point at the top of its main processing loop where it
waits for the next request to arrive. This is a good place to ensure that
resource usage has not increased since the last execution of the loop.

At a lower, but no less useful level, you can invest in tools that (among
other things) check your running programs for memory leaks. Purify
(www.rational.com) and Insure++ (www.parasoft.com) are popular
choices.

Related sections include:

Design by Contract, page 109
Assertive Programming, page 122
Decoupling and the Law of Demeter, page 138

Challenges
Although there are no guaranteed ways of ensuring that you always free
resources, certain design techniques, when applied consistently, will help.
In the text we discussed how establishing a semantic invariant for major
data structures could direct memory deallocation decisions. Consider how
Design by Contract, page 109, could help refine this idea.

Exercises
Answer 22. Some C and C++ developers make a point of setting a pointer to NULL after
on p. 292
they deallocate the memory it references. Why is this a good idea?

Answer 23. Some Java developers make a point of setting an object variable to NULL
on p. 292
after they have finished using the object. Why is this a good idea?

Prepared exclusively for Zach

 Chapter 5

Bend, or Break
Life doesn’t stand still.

Neither can the code that we write. In order to keep up with today’s
near-frantic pace of change, we need to make every effort to write code
that’s as loose—as flexible—as possible. Otherwise we may find our
code quickly becoming outdated, or too brittle to fix, and may ultimately
be left behind in the mad dash toward the future.

In Reversibility, on page 44, we talked about the perils of irreversible

decisions. In this chapter, we’ll tell you how to make reversible deci-
sions, so your code can stay flexible and adaptable in the face of an
uncertain world.

First we need to look at coupling—the dependencies among modules

of code. In Decoupling and the Law of Demeter we’ll show how to keep
separate concepts separate, and decrease coupling.

A good way to stay flexible is to write less code. Changing code leaves
you open to the possibility of introducing new bugs. Metaprogramming
will explain how to move details out of the code completely, where they
can be changed more safely and easily.

In Temporal Coupling, we’ll look at two aspects of time as they relate to

coupling. Do you depend on the “tick” coming before the “tock”? Not if
you want to stay flexible.

A key concept in creating flexible code is the separation of a data model

from a view, or presentation, of that model. We’ll decouple models from
views in It’s Just a View.

137

Prepared exclusively for Zach

 138 C HAPTER 5 B END , OR B REAK

Finally, there’s a technique for decoupling modules even further by pro-

viding a meeting place where modules can exchange data anonymously
and asynchronously. This is the topic of Blackboards.

Armed with these techniques, you can write code that will “roll with the
punches.”

26 Decoupling and the Law of Demeter

Good fences make good neighbors.
Robert Frost, “Mending Wall”

In Orthogonality, page 34, and Design by Contract, page 109, we sug-

gested that writing “shy” code is beneficial. But “shy” works two ways:
don’t reveal yourself to others, and don’t interact with too many people.

Spies, dissidents, revolutionaries, and such are often organized into

small groups of people called cells. Although individuals in each cell
may know each other, they have no knowledge of those in other cells. If
one cell is discovered, no amount of truth serum will reveal the names
of others outside the cell. Eliminating interactions between cells pro-
tects everyone.

We feel that this is a good principle to apply to coding as well. Organize

your code into cells (modules) and limit the interaction between them.
If one module then gets compromised and has to be replaced, the other
modules should be able to carry on.

Minimize Coupling
What’s wrong with having modules that know about each other? Noth-
ing in principle—we don’t need to be as paranoid as spies or dissidents.
However, you do need to be careful about how many other modules you
interact with and, more importantly, how you came to interact with
them.

Suppose you are remodeling your house, or building a house from

scratch. A typical arrangement involves a “general contractor.” You hire
the contractor to get the work done, but the contractor may or may

Prepared exclusively for Zach

 D ECOUPLING AND THE L AW OF D EMETER 139

not do the construction personally; the work may be offered to various

subcontractors. But as the client, you are not involved in dealing with
the subcontractors directly—the general contractor assumes that set of
headaches on your behalf.

We’d like to follow this same model in software. When we ask an object
for a particular service, we’d like the service to be performed on our
behalf. We do not want the object to give us a third-party object that we
have to deal with to get the required service.

For example, suppose you are writing a class that generates a graph
of scientific recorder data. You have data recorders spread around the
world; each recorder object contains a location object giving its position
and time zone. You want to let your users select a recorder and plot its
data, labeled with the correct time zone. You might write

public void plotDate(Date aDate, Selection aSelection) {

TimeZone tz =
aSelection.getRecorder().getLocation().getTimeZone();
...
}

But now the plotting routine is unnecessarily coupled to three classes—

Selection, Recorder, and Location. This style of coding dramat-
ically increases the number of classes on which our class depends.
Why is this a bad thing? It increases the risk that an unrelated change
somewhere else in the system will affect your code. For instance, if Fred
makes a change to Location such that it no longer directly contains a
TimeZone, you have to change your code as well.

Rather than digging though a hierarchy yourself, just ask for what you
need directly:

public void plotDate(Date aDate, TimeZone aTz) {

...
}
plotDate(someDate, someSelection.getTimeZone());

We added a method to Selection to get the time zone on our behalf:

the plotting routine doesn’t care whether the time zone comes from
the Recorder directly, from some contained object within Recorder,
or whether Selection makes up a different time zone entirely. The
selection routine, in turn, should probably just ask the recorder for
its time zone, leaving it up to the recorder to get it from its contained
Location object.

Prepared exclusively for Zach

 140 C HAPTER 5 B END , OR B REAK

Traversing relationships between objects directly can quickly lead to

a combinatorial explosion1 of dependency relationships. You can see
symptoms of this phenomenon in a number of ways:

1. Large C or C++ projects where the command to link a unit test is

longer than the test program itself

2. “Simple” changes to one module that propagate through unrelated

modules in the system

3. Developers who are afraid to change code because they aren’t sure
what might be affected

Systems with many unnecessary dependencies are very hard (and ex-
pensive) to maintain, and tend to be highly unstable. In order to keep
the dependencies to a minimum, we’ll use the Law of Demeter to design
our methods and functions.

The Law of Demeter for Functions

The Law of Demeter for functions [LH89] attempts to minimize coupling
between modules in any given program. It tries to prevent you from
reaching into an object to gain access to a third object’s methods. The
law is summarized in Figure 5.1 on the next page.

By writing “shy” code that honors the Law of Demeter as much as pos-
sible, we can achieve our objective:

TIP 36

Minimize Coupling Between Modules

Does It Really Make a Difference?

While it sounds good in theory, does following the Law of Demeter really
help to create more maintainable code?

Studies have shown [BBM96] that classes in C++ with larger response
sets are more prone to error than classes with smaller response sets (a

1. If objects all know about each other, then a change to just one object can result
in the other objects needing changes.

Prepared exclusively for Zach

 D ECOUPLING AND THE L AW OF D EMETER 141

Figure 5.1. Law of Demeter for functions

class Demeter {
private:
A *a; The Law of Demeter for functions
int func(); states that any method of an
public:
object should call only methods
//...
void example(B& b); belonging to:
}

void Demeter::example(B& b) {
C c;
int f = func(); itself

any parameters that were

b.invert();
passed in to the method
a = new A();
a->setActive(); any objects it created

any directly held component

c.print();
objects
}

response set is defined to be the number of functions directly invoked

by methods of the class).

Because following the Law of Demeter reduces the size of the response
set in the calling class, it follows that classes designed in this way
will also tend to have fewer errors (see [URL 56] for more papers and
information on the Demeter project).

Using The Law of Demeter will make your code more adaptable and
robust, but at a cost: as a “general contractor,” your module must dele-
gate and manage any and all subcontractors directly, without involving
clients of your module. In practice, this means that you will be writing
a large number of wrapper methods that simply forward the request on
to a delegate. These wrapper methods will impose both a runtime cost
and a space overhead, which may be significant—even prohibitive—in
some applications.

As with any technique, you must balance the pros and cons for your
particular application. In database schema design it is common prac-
tice to “denormalize” the schema for a performance improvement: to

Prepared exclusively for Zach

 142 C HAPTER 5 B END , OR B REAK

Physical Decoupling
In this section we’re concerned largely with designing to keep things
logically decoupled within systems. However, there is another kind
of interdependence that becomes highly significant as systems grow
larger. In his book Large-Scale C++ Software Design [Lak96], John
Lakos addresses the issues surrounding the relationships among
the files, directories, and libraries that make up a system. Large
projects that ignore these physical design problems wind up with build
cycles that are measured in days and unit tests that may drag in the
entire system as support code, among other problems. Mr. Lakos
argues convincingly that logical and physical design must proceed in
tandem—that undoing the damage done to a large body of code by
cyclic dependencies is extremely difficult. We recommend this book
if you are involved in large-scale developments, even if C++ isn’t your
implementation language.

violate the rules of normalization in exchange for speed. A similar trade-

off can be made here as well. In fact, by reversing the Law of Demeter
and tightly coupling several modules, you may realize an important
performance gain. As long as it is well known and acceptable for those
modules to be coupled, your design is fine.

Otherwise, you may find yourself on the road to a brittle, inflexible

future. Or no future at all.

Related sections include:

Orthogonality, page 34
Reversibility, page 44
Design by Contract, page 109
How to Balance Resources, page 129
It’s Just a View, page 157
Pragmatic Teams, page 224
Ruthless Testing, page 237

Challenges
We’ve discussed how using delegation makes it easier to obey the Law of
Demeter and hence reduce coupling. However, writing all of the methods

Prepared exclusively for Zach

 D ECOUPLING AND THE L AW OF D EMETER 143

needed to forward calls to delegated classes is boring and error prone.

What are the advantages and disadvantages of writing a preprocessor that
generates these calls automatically? Should this preprocessor be run only
once, or should it be used as part of the build?

Exercises
24. We discussed the concept of physical decoupling in the box on on the facing Answer
on p. 293
page. Which of the following C++ header files is more tightly coupled to the
rest of the system?

person1.h: person2.h:
#include "date.h" class Date;
class Person1 { class Person2 {
private: private:
Date myBirthdate; Date *myBirthdate;
public: public:
Person1(Date &birthDate); Person2(Date &birthDate);
// ... // ...

25. For the example below and for those in Exercises 26 and 27, determine if Answer
on p. 293
the method calls shown are allowed according to the Law of Demeter. This
first one is in Java.
public void showBalance(BankAccount acct) {
Money amt = acct.getBalance();
printToScreen(amt.printFormat());
}

26. This example is also in Java. Answer

on p. 294
public class Colada {
private Blender myBlender;
private Vector myStuff;
public Colada() {
myBlender = new Blender();
myStuff = new Vector();
}
private void doSomething() {
myBlender.addIngredients(myStuff.elements());
}
}

27. This example is in C++. Answer

on p. 294
void processTransaction(BankAccount acct, int) {
Person *who;
Money amt;
amt.setValue(123.45);
acct.setBalance(amt);
who = acct.getOwner();
markWorkflow(who->name(), SET_BALANCE);
}

Prepared exclusively for Zach

 144 C HAPTER 5 B END , OR B REAK

27 Metaprogramming
No amount of genius can overcome a preoccupation with detail.
Levy’s Eighth Law

Details mess up our pristine code—especially if they change frequently.

Every time we have to go in and change the code to accommodate some
change in business logic, or in the law, or in management’s personal
tastes of the day, we run the risk of breaking the system—of introducing
a new bug.

So we say “out with the details!” Get them out of the code. While we’re
at it, we can make our code highly configurable and “soft”—that is,
easily adaptable to changes.

Dynamic Configuration
First, we want to make our systems highly configurable. Not just things
such as screen colors and prompt text, but deeply ingrained items such
as the choice of algorithms, database products, middleware technology,
and user-interface style. These items should be implemented as con-
figuration options, not through integration or engineering.

TIP 37

Configure, Don’t Integrate

Use metadata to describe configuration options for an application: tun-

ing parameters, user preferences, the installation directory, and so on.

What exactly is metadata? Strictly speaking, metadata is data about

data. The most common example is probably a database schema or
data dictionary. A schema contains data that describes fields (columns)
in terms of names, storage lengths, and other attributes. You should be
able to access and manipulate this information just as you would any
other data in the database.

We use the term in its broadest sense. Metadata is any data that
describes the application—how it should run, what resources it should
use, and so on. Typically, metadata is accessed and used at runtime,
not at compile time. You use metadata all the time—at least your pro-
grams do. Suppose you click on an option to hide the toolbar on your

Prepared exclusively for Zach

 M ETAPROGRAMMING 145

Web browser. The browser will store that preference, as metadata, in

some sort of internal database.

This database might be in a proprietary format, or it might use a stan-

dard mechanism. Under Windows, either an initialization file (using the
suffix .ini) or entries in the system Registry are typical. Under Unix,
the X Window System provides similar functionality using Application
Default files. Java uses Property files. In all of these environments, you
specify a key to retrieve a value. Alternatively, more powerful and flex-
ible implementations of metadata use an embedded scripting language
(see Domain Languages, page 57, for details).

The Netscape browser has actually implemented preferences using both

of these techniques. In Version 3, preferences were saved as simple
key/value pairs:
SHOW_TOOLBAR: False

Later, Version 4 preferences looked more like JavaScript:

user_pref("custtoolbar.Browser.Navigation_Toolbar.open", false);

Metadata-Driven Applications
But we want to go beyond using metadata for simple preferences. We
want to configure and drive the application via metadata as much as
possible. Our goal is to think declaratively (specifying what is to be
done, not how) and create highly dynamic and adaptable programs. We
do this by adopting a general rule: program for the general case, and
put the specifics somewhere else—outside the compiled code base.

TIP 38

Put Abstractions in Code, Details in Metadata

There are several benefits to this approach:

It forces you to decouple your design, which results in a more flex-

ible and adaptable program.

It forces you to create a more robust, abstract design by deferring

details—deferring them all the way out of the program.

Prepared exclusively for Zach

 146 C HAPTER 5 B END , OR B REAK

You can customize the application without recompiling it. You can
also use this level of customization to provide easy work-arounds
for critical bugs in live production systems.

Metadata can be expressed in a manner that’s much closer to the

problem domain than a general-purpose programming language
might be (see Domain Languages, page 57).

You may even be able to implement several different projects using

the same application engine, but with different metadata.

We want to defer definition of most details until the last moment, and
leave the details as soft—as easy to change—as we can. By crafting
a solution that allows us to make changes quickly, we stand a better
chance of coping with the flood of directional shifts that swamp many
projects (see Reversibility, page 44).

Business Logic
So you’ve made the choice of database engine a configuration option,
and provided metadata to determine the user-interface style. Can we
do more? Definitely.

Because business policy and rules are more likely to change than any
other aspect of the project, it makes sense to maintain them in a very
flexible format.

For example, your purchasing application may include various corpo-

rate policies. Maybe you pay small suppliers in 45 days and large ones
in 90 days. Make the definitions of the supplier types, as well as the
time periods themselves, configurable. Take the opportunity to gener-
alize.

Maybe you are writing a system with horrendous workflow require-

ments. Actions start and stop according to complex (and changing)
business rules. Consider encoding them in some kind of rule-based
(or expert) system, embedded within your application. That way, you’ll
configure it by writing rules, not cutting code.

Less complex logic can be expressed using a mini-language, removing

the need to recompile and redeploy when the environment changes.
Have a look at page 58 for an example.

Prepared exclusively for Zach

 M ETAPROGRAMMING 147

When to Configure
As mentioned in The Power of Plain Text, page 73, we recommend
representing configuration metadata in plain text—it makes life that
much easier.
But when should a program read this configuration? Many programs
will scan such things only at startup, which is unfortunate. If you need
to change the configuration, this forces you to restart the application.
A more flexible approach is to write programs that can reload their
configuration while they’re running. This flexibility comes at a cost: it
is more complex to implement.
So consider how your application will be used: if it is a long-running
server process, you will want to provide some way to reread and apply
metadata while the program is running. For a small client GUI appli-
cation that restarts quickly, you may not need to.
This phenomenon is not limited to application code. We’ve all been
annoyed at operating systems that force us to reboot when we install
some simple application or change an innocuous parameter.

An Example: Enterprise Java Beans

Enterprise Java Beans (EJB) is a framework for simplifying program-
ming in a distributed, transaction-based environment. We mention it
here because EJB illustrates how metadata can be used both to config-
ure applications and to reduce the complexity of writing code.

Suppose you want to create some Java software that will participate
in transactions across different machines, between different database
vendors, and with different thread and load-balancing models.

The good news is, you don’t have to worry about all that. You write
a bean—a self-contained object that follows certain conventions—and
place it in a bean container that manages much of the low-level detail
on your behalf. You can write the code for a bean without including any
transaction operations or thread management; EJB uses metadata to
specify how transactions should be handled.

Thread allocation and load balancing are specified as metadata to the

underlying transaction service that the container uses. This separation

Prepared exclusively for Zach

 148 C HAPTER 5 B END , OR B REAK

allows us great flexibility to configure the environment dynamically, at

runtime.

The bean’s container can manage transactions on the bean’s behalf in

one of several different styles (including an option where you control
your own commits and rollbacks). All of the parameters that affect the
bean’s behavior are specified in the bean’s deployment descriptor—a
serialized object that contains the metadata we need.

Distributed systems such as EJB are leading the way into a new world
of configurable, dynamic systems.

Cooperative Configuration
We’ve talked about users and developers configuring dynamic applica-
tions. But what happens if you let applications configure each other—
software that adapts itself to its environment? Unplanned, spur-of-the-
moment configuration of existing software is a powerful concept.

Operating systems already configure themselves to hardware as they

boot, and Web browsers update themselves with new components auto-
matically.

Your larger applications probably already have issues with handling dif-
ferent versions of data and different releases of libraries and operating
systems. Perhaps a more dynamic approach will help.

Don’t Write Dodo-Code

Without metadata, your code is not as adaptable or flexible as it could
be. Is this a bad thing? Well, out here in the real world, species that
don’t adapt die.

The dodo didn’t adapt to the presence of humans and their livestock
on the island of Mauritius, and quickly became extinct.2 It was the first
documented extinction of a species at the hand of man.

Don’t let your project (or your career) go the way of the dodo.

2. It didn’t help that the settlers beat the placid (read stupid) birds to death with clubs
for sport.

Prepared exclusively for Zach

 M ETAPROGRAMMING 149

Related sections include:

Orthogonality, page 34
Reversibility, page 44
Domain Languages, page 57
The Power of Plain Text, page 73

Challenges
For your current project, consider how much of the application might be
moved out of the program itself to metadata. What would the resultant
“engine” look like? Would you be able to reuse that engine in the context
of a different application?

Exercises
28. Which of the following things would be better represented as code within Answer
on p. 295
a program, and which externally as metadata?

1. Communication port assignments

2. An editor’s support for highlighting the syntax of various languages
3. An editor’s support for different graphic devices
4. A state machine for a parser or scanner
5. Sample values and results for use in unit testing

Prepared exclusively for Zach

 150 C HAPTER 5 B END , OR B REAK

28 Temporal Coupling
What is temporal coupling all about, you may ask. It’s about time.

Time is an often ignored aspect of software architectures. The only time

that preoccupies us is the time on the schedule, the time left until we
ship—but this is not what we’re talking about here. Instead, we are
talking about the role of time as a design element of the software itself.
There are two aspects of time that are important to us: concurrency
(things happening at the same time) and ordering (the relative positions
of things in time).

We don’t usually approach programming with either of these aspects in

mind. When people first sit down to design an architecture or write a
program, things tend to be linear. That’s the way most people think—
do this and then always do that. But thinking this way leads to tempo-
ral coupling: coupling in time. Method A must always be called before
method B; only one report can be run at a time; you must wait for the
screen to redraw before the button click is received. Tick must happen
before tock.

This approach is not very flexible, and not very realistic.

We need to allow for concurrency3 and to think about decoupling any

time or order dependencies. In doing so, we can gain flexibility and
reduce any time-based dependencies in many areas of development:
workflow analysis, architecture, design, and deployment.

Workflow
On many projects, we need to model and analyze the users’ workflows
as part of requirements analysis. We’d like to find out what can happen
at the same time, and what must happen in a strict order. One way to
do this is to capture their description of workflow using a notation such
as the UML activity diagram.4

3. We won’t go into the details of concurrent or parallel programming here; a good

computer science textbook should cover the basics, including scheduling, deadlock, star-
vation, mutual exclusion/semaphores, and so on.

4. For more information on all of the UML diagram types, see [FS97].

Prepared exclusively for Zach

 T EMPORAL C OUPLING 151

An activity diagram consists of a set of actions drawn as rounded boxes.

The arrow leaving an action leads to either another action (which can
start once the first action completes) or to a thick line called a synchro-
nization bar. Once all the actions leading into a synchronization bar are
complete, you can then proceed along any arrows leaving the bar. An
action with no arrows leading into it can be started at any time.

You can use activity diagrams to maximize parallelism by identifying

activities that could be performed in parallel, but aren’t.

TIP 39

Analyze Workflow to Improve Concurrency

For instance, in our blender project (Exercise 17, page 119), users may
initially describe their current workflow as follows.

1. Open blender
2. Open piña colada mix
3. Put mix in blender
4. Measure 1/2 cup white rum
5. Pour in rum
6. Add 2 cups of ice
7. Close blender
8. Liquefy for 2 minutes
9. Open blender
10. Get glasses
11. Get pink umbrellas
12. Serve

Even though they describe these actions serially, and may even per-
form them serially, we notice that many of them could be performed in
parallel, as we show in the activity diagram in Figure 5.2 on the next
page.

It can be eye-opening to see where the dependencies really exist. In

this instance, the top-level tasks (1, 2, 4, 10, and 11) can all happen
concurrently, up front. Tasks 3, 5, and 6 can happen in parallel later.

If you were in a piña colada-making contest, these optimizations may

make all the difference.

Prepared exclusively for Zach

 152 C HAPTER 5 B END , OR B REAK

Figure 5.2. UML activity diagram: making a piña colada

Open Measure
2. Open mix 1. 4.
blender rum

Put Add two Pour

3. 6. 5.
mix in cups ice in rum

Close
7.
blender
Get pink
11.
umbrellas
8. Liquefy
Get
10.
glasses
Open
9.
blender

12. Serve

Architecture
We wrote an On-Line Transaction Processing (OLTP) system a few years
ago. At its simplest, all the system had to do was read a request and
process the transaction against the database. But we wrote a three-
tier, multiprocessing distributed application: each component was an
independent entity that ran concurrently with all other components.
While this sounds like more work, it wasn’t: taking advantage of tem-
poral decoupling made it easier to write. Let’s take a closer look at this
project.

The system takes in requests from a large number of data communica-

tion lines and processes transactions against a back-end database.

The design addresses the following constraints:

Prepared exclusively for Zach

 T EMPORAL C OUPLING 153

Figure 5.3. OLTP architecture overview

Input
task #1
App.
App’n logic #1 Database
Input Database
task #2 handler
Queue App. Queue
logic #n
Input
task #n

Database operations take a relatively long time to complete.

For each transaction, we must not block communication services
while a database transaction is being processed.
Database performance suffers with too many concurrent sessions.
Multiple transactions are in progress concurrently on each data
line.

The solution that gave us the best performance and cleanest architec-
ture looked something like Figure 5.3.

Each box represents a separate process; processes communicate via

work queues. Each input process monitors one incoming communica-
tion line, and makes requests to the application server. All requests are
asynchronous: as soon as the input process makes its current request,
it goes back to monitoring the line for more traffic. Similarly, the appli-
cation server makes requests of the database process,5 and is notified
when the individual transaction is complete.

This example also shows a way to get quick and dirty load balancing
among multiple consumer processes: the hungry consumer model.

5. Even though we show the database as a single, monolithic entity, it is not. The
database software is partitioned into several processes and client threads, but this is
handled internally by the database software and isn’t part of our example.

Prepared exclusively for Zach

 154 C HAPTER 5 B END , OR B REAK

In a hungry consumer model, you replace the central scheduler with a

number of independent consumer tasks and a centralized work queue.
Each consumer task grabs a piece from the work queue and goes on
about the business of processing it. As each task finishes its work, it
goes back to the queue for some more. This way, if any particular task
gets bogged down, the others can pick up the slack, and each individual
component can proceed at its own pace. Each component is temporally
decoupled from the others.

TIP 40

Design Using Services

Instead of components, we have really created services—independent,

concurrent objects behind well-defined, consistent interfaces.

Design for Concurrency

The rising acceptance of Java as a platform has exposed more devel-
opers to multithreaded programming. But programming with threads
imposes some design constraints—and that’s a good thing. Those con-
straints are actually so helpful that we want to abide by them whenever
we program. It will help us decouple our code and fight programming
by coincidence (see page 172).

With linear code, it’s easy to make assumptions that lead to sloppy
programming. But concurrency forces you to think through things a bit
more carefully—you’re not alone at the party anymore. Because things
can now happen at the “same time,” you may suddenly see some time-
based dependencies.

To begin with, any global or static variables must be protected from

concurrent access. Now may be a good time to ask yourself why you
need a global variable in the first place. In addition, you need to make
sure that you present consistent state information, regardless of the
order of calls. For example, when is it valid to query the state of your
object? If your object is in an invalid state between certain calls, you
may be relying on a coincidence that no one can call your object at that
point in time.

Prepared exclusively for Zach

 T EMPORAL C OUPLING 155

Suppose you have a windowing subsystem where the widgets are first
created and then shown on the display in two separate steps. You aren’t
allowed to set state in the widget until it is shown. Depending on how
the code is set up, you may be relying on the fact that no other object
can use the created widget until you’ve shown it on the screen.

But this may not be true in a concurrent system. Objects must always
be in a valid state when called, and they can be called at the most awk-
ward times. You must ensure that an object is in a valid state any time
it could possibly be called. Often this problem shows up with classes
that define separate constructor and initialization routines (where the
constructor doesn’t leave the object in an initialized state). Using class
invariants, discussed in Design by Contract, page 109, will help you
avoid this trap.

Cleaner Interfaces
Thinking about concurrency and time-ordered dependencies can lead
you to design cleaner interfaces as well. Consider the C library routine
strtok, which breaks a string into tokens.

The design of strtok isn’t thread safe,6 but that isn’t the worst part:
look at the time dependency. You must make the first call to strtok
with the variable you want to parse, and all successive calls with a NULL
instead. If you pass in a non-NULL value, it restarts the parse on that
buffer instead. Without even considering threads, suppose you wanted
to use strtok to parse two separate strings at the same time:
char buf1[BUFSIZ];
char buf2[BUFSIZ];
char *p, *q;
strcpy(buf1, "this is a test");
strcpy(buf2, "this ain’t gonna work");
p = strtok(buf1, " ");
q = strtok(buf2, " ");
while (p && q) {
printf("%s %s n", p, q);
p = strtok(NULL, " ");
q = strtok(NULL, " ");
}

6. It uses static data to maintain the current position in the buffer. The static data
isn’t protected against concurrent access, so it isn’t thread safe. In addition, it clobbers
the first argument you pass in, which can lead to some nasty surprises.

Prepared exclusively for Zach

 156 C HAPTER 5 B END , OR B REAK

The code as shown will not work: there is implicit state retained in
strtok between calls. You have to use strtok on just one buffer at a
time.

Now in Java, the design of a string parser has to be different. It must

be thread safe and present a consistent state.

StringTokenizer st1 = new StringTokenizer("this is a test");

StringTokenizer st2 = new StringTokenizer("this test will work");
while (st1.hasMoreTokens() && st2.hasMoreTokens()) {
System.out.println(st1.nextToken());
System.out.println(st2.nextToken());
}

StringTokenizer is a much cleaner, more maintainable, interface. It

contains no surprises, and won’t cause mysterious bugs in the future,
as strtok might.

TIP 41

Always Design for Concurrency

Deployment
Once you’ve designed an architecture with an element of concurrency,
it becomes easier to think about handling many concurrent services:
the model becomes pervasive.

Now you can be flexible as to how the application is deployed: stand-

alone, client-server, or n-tier. By architecting your system as inde-
pendent services, you can make the configuration dynamic as well.
By planning for concurrency, and decoupling operations in time, you
have all these options—including the stand-alone option, where you
can choose not to be concurrent.

Going the other way (trying to add concurrency to a nonconcurrent

application) is much harder. If we design to allow for concurrency, we
can more easily meet scalability or performance requirements when the
time comes—and if the time never comes, we still have the benefit of a
cleaner design.

Isn’t it about time?

Prepared exclusively for Zach

 I T ’ S J UST A V IEW 157

Related sections include:

Design by Contract, page 109
Programming by Coincidence, page 172

Challenges
How many tasks do you perform in parallel when you get ready for work in
the morning? Could you express this in a UML activity diagram? Can you
find some way to get ready more quickly by increasing concurrency?

29 It’s Just a View

Still, a man hears
What he wants to hear
And disregards the rest
La la la. . .
Simon and Garfunkel, “The Boxer”

Early on we are taught not to write a program as a single big chunk, but
that we should “divide and conquer” and separate a program into mod-
ules. Each module has its own responsibilities; in fact, a good definition
of a module (or class) is that it has a single, well-defined responsibility.

But once you separate a program into different modules based on

responsibility, you have a new problem. At runtime, how do the objects
talk to each other? How do you manage the logical dependencies
between them? That is, how do you synchronize changes in state (or
updates to data values) in these different objects? It needs to be done
in a clean, flexible manner—we don’t want them to know too much
about each other. We want each module to be like the man in the song
and just hear what it wants to hear.

We’ll start off with the concept of an event. An event is simply a special
message that says “something interesting just happened” (interesting,
of course, lies in the eye of the beholder). We can use events to signal
changes in one object that some other object may be interested in.

Using events in this way minimizes coupling between those objects—

the sender of the event doesn’t need to have any explicit knowledge of

Prepared exclusively for Zach

 158 C HAPTER 5 B END , OR B REAK

the receiver. In fact, there could be multiple receivers, each one focused
on its own agenda (of which the sender is blissfully unaware).

We need to exercise some care in using events, however. In an early ver-

sion of Java, for example, one routine received all the events destined
for a particular application. Not exactly the road to easy maintenance
or evolution.

Publish/Subscribe
Why is it bad to push all the events through a single routine? It vio-
lates object encapsulation—that one routine now has to have intimate
knowledge of the interactions among many objects. It also increases the
coupling—and we’re trying to decrease coupling. Because the objects
themselves have to have knowledge of these events as well, you are
probably going to violate the DRY principle, orthogonality, and perhaps
even sections of the Geneva Convention. You may have seen this kind
of code—it is usually dominated by a huge case statement or multiway
if-then. We can do better.

Objects should be able to register to receive only the events they need,
and should never be sent events they don’t need. We don’t want to
spam our objects! Instead, we can use a publish/subscribe protocol,
illustrated using the UML sequence diagram in Figure 5.4 on the next
page.7

A sequence diagram shows the flow of messages among several objects,

with objects arranged in columns. Each message is shown as a labeled
arrow from the sender’s column to the receiver’s column. An asterisk in
the label means that more than one message of this type can be sent.

If we are interested in certain events generated by a Publisher, all we

have to do is register ourselves. The Publisher keeps track of all in-
terested Subscriber objects; when the Publisher generates an event
of interest, it will call each Subscriber in turn and notify them that
the event has occurred.

7. See also the Observer pattern in [GHJV95] for more information.

Prepared exclusively for Zach

 I T ’ S J UST A V IEW 159

Figure 5.4. Publish/subscribe protocol

Subscriber Subscriber
Publisher
one two

register

notify*

register

notify*

notify*

unsubscribe

notify*

There are several variations on this theme—mirroring other commu-

nication styles. Objects may use publish/subscribe on a peer-to-peer
basis (as we saw above); they may use a “software bus” where a
centralized object maintains the database of listeners and dispatches
messages appropriately. You might even have a scheme where critical
events get broadcast to all listeners—registered or not. One possible
implementation of events in a distributed environment is illustrated by
the CORBA Event Service, described in the box on the following page.

We can use this publish/subscribe mechanism to implement a very

important design concept: the separation of a model from views of the
model. Let’s start with a GUI-based example, using the Smalltalk design
in which this concept was born.

Model-View-Controller
Suppose you have a spreadsheet application. In addition to the num-
bers in the spreadsheet itself, you also have a graph that displays the

Prepared exclusively for Zach

 160 C HAPTER 5 B END , OR B REAK

The CORBA Event Service

The CORBA Event Service allows participating objects to send and
receive event notifications via a common bus, the event channel. The
event channel arbitrates event handling, and also decouples event
producers from event consumers. It works in two basic ways: push
and pull.
In push mode, event suppliers inform the event channel that an event
has occurred. The channel then automatically distributes that event
to all client objects that have registered interest.
In pull mode, clients periodically poll the event channel, which in turn
polls the supplier that offers event data corresponding to the request.
Although the CORBA Event Service can be used to implement all of
the event models discussed in this section, you can also view it as
a different animal. CORBA facilitates communication among objects
written in different programming languages running on geographi-
cally dispersed machines with different architectures. Sitting on top
of CORBA, the event service gives you a decoupled way of interact-
ing with applications around the world, written by people you’ve never
met, using programming languages you’d rather not know about.

numbers as a bar chart and a running total dialog box that shows the
sum of a column in the spreadsheet.

Obviously, we don’t want to have three separate copies of the data. So

we create a model—the data itself, with common operations to manip-
ulate it. Then we can create separate views that display the data in
different ways: as a spreadsheet, as a graph, or in a totals box. Each
of these views may have its own controller. The graph view may have a
controller that allows you to zoom in or out, or pan around the data,
for example. None of this affects the data itself, just that view.

This is the key concept behind the Model-View-Controller (MVC) idiom:

separating the model from both the GUI that represents it and the con-
trols that manage the view.8

8. The view and controller are tightly coupled, and in some implementations of MVC
the view and controller are a single component.

Prepared exclusively for Zach

 I T ’ S J UST A V IEW 161

By doing so, you can take advantage of some interesting possibilities.

You can support multiple views of the same data model. You can use
common viewers on many different data models. You can even support
multiple controllers to provide nontraditional input mechanisms.

TIP 42

Separate Views from Models

By loosening the coupling between the model and the view/controller,

you buy yourself a lot of flexibility at low cost. In fact, this technique
is one of the most important ways of maintaining reversibility (see
Reversibility, page 44).

Java Tree View

A good example of an MVC design can be found in the Java tree widget.
The tree widget (which displays a clickable, traversable tree) is actually
a set of several different classes organized in an MVC pattern.

To produce a fully functional tree widget, all you need to do is provide a

data source that conforms to the TreeModel interface. Your code now
becomes the model for the tree.

The view is created by the TreeCellRenderer and TreeCellEditor

classes, which can be inherited from and customized to provide differ-
ent colors, fonts, and icons in the widget. JTree acts as the controller
for the tree widget and provides some general viewing functionality.

Because we have decoupled the model from the view, we simplify the
programming a great deal. You don’t have to think about programming
a tree widget anymore. Instead, you just provide a data source.

Suppose the vice president comes up to you and wants a quick appli-
cation that lets her navigate the company’s organizational chart, which
is held in a legacy database on the mainframe. Just write a wrapper
that takes the mainframe data, presents it as a TreeModel, and voilà:
you have a fully navigable tree widget.

Now you can get fancy and start using the viewer classes; you can
change how nodes are rendered, and use special icons, fonts, or colors.
When the VP comes back and says the new corporate standards dictate

Prepared exclusively for Zach

 162 C HAPTER 5 B END , OR B REAK

the use of a Skull and Crossbones icon for certain employees, you can
make the changes to TreeCellRenderer without touching any other
code.

Beyond GUIs
While MVC is typically taught in the context of GUI development, it is
really a general-purpose programming technique. The view is an inter-
pretation of the model (perhaps a subset)—it doesn’t need to be graph-
ical. The controller is more of a coordination mechanism, and doesn’t
have to be related to any sort of input device.

Model. The abstract data model representing the target object.

The model has no direct knowledge of any views or controllers.
View. A way to interpret the model. It subscribes to changes in the
model and logical events from the controller.
Controller. A way to control the view and provide the model with
new data. It publishes events to both the model and the view.

Let’s look at a nongraphical example.

Baseball is a unique institution. Where else can you learn such gems of
trivia as “this has become the highest-scoring game played on a Tues-
day, in the rain, under artificial lights, between teams whose names
start with a vowel?” Suppose we were charged with developing software
to support those intrepid announcers who must dutifully report on the
scores, the statistics, and the trivia.

Clearly we need information on the game in progress—the teams play-

ing, the conditions, the player at bat, the score, and so on. These facts
form our models; they will be updated as new information arrives (a
pitcher is changed, a player strikes out, it starts raining ).

We’ll then have a number of view objects that use these models. One
view might look for runs so it can update the current score. Another
may receive notifications of new batters, and retrieve a brief summary
of their year-to-date statistics. A third viewer may look at the data
and check for new world records. We might even have a trivia viewer,
responsible for coming up with those weird and useless facts that thrill
the viewing public.

Prepared exclusively for Zach

 I T ’ S J UST A V IEW 163

Figure 5.5. Baseball reporting. Viewers subscribe to models.

Score
collector TV feed
generator

Scores
Batter
stats
Display Web page
filter formatter

Records

Conditions

Tele-
prompter
Trivia

subscribes to
model viewer

But we don’t want to flood the poor announcer with all of these views
directly. Instead, we’ll have each view generate notifications of “inter-
esting” events, and let some higher-level object schedule what gets
shown.9

These viewer objects have suddenly become models for the higher-level
object, which itself might then be a model for different formatting view-
ers. One formatting viewer might create the teleprompter script for the
announcer, another might generate video captions directly on the satel-
lite uplink, another might update the network’s or team’s Web pages
(see Figure 5.5).

This kind of model-viewer network is a common (and valuable) design

technique. Each link decouples raw data from the events that created
it—each new viewer is an abstraction. And because the relationships
are a network (not just a linear chain), we have a lot of flexibility. Each

9. The fact that a plane flies overhead probably isn’t interesting unless it’s the 100th
plane to fly overhead that night.

Prepared exclusively for Zach

 164 C HAPTER 5 B END , OR B REAK

model may have many viewers, and one viewer may work with multiple
models.

In advanced systems such as this one, it can be handy to have de-

bugging views—specialized views that show you in-depth details of the
model. Adding a facility to trace individual events can be a great time
saver as well.

Still Coupled (After All These Years)

Despite the decrease in coupling we have achieved, listeners and event
generators (subscribers and publishers) still have some knowledge of
each other. In Java, for instance, they must agree on common interface
definitions and calling conventions.

In the next section, we’ll look at ways of reducing coupling even further
by using a form of publish and subscribe where none of the participants
need know about each other, or call each other directly.

Related sections include:

Orthogonality, page 34
Reversibility, page 44
Decoupling and the Law of Demeter, page 138
Blackboards, page 165
It’s All Writing, page 248

Exercises
Answer 29. Suppose you have an airline reservation system that includes the concept
on p. 296
of a flight:

public interface Flight {

// Return false if flight full.
public boolean addPassenger(Passenger p);
public void addToWaitList(Passenger p);
public int getFlightCapacity();
public int getNumPassengers();
}

If you add a passenger to the wait list, they’ll be put on the flight automat-
ically when an opening becomes available.

Prepared exclusively for Zach

 B LACKBOARDS 165

There’s a massive reporting job that goes through looking for overbooked or
full flights to suggest when additional flights might be scheduled. It works
fine, but it takes hours to run.

We’d like to have a little more flexibility in processing wait-list passengers,

and we’ve got to do something about that big report—it takes too long to
run. Use the ideas from this section to redesign this interface.

30 Blackboards
The writing is on the wall. . .

You may not usually associate elegance with police detectives, pictur-
ing instead some sort of doughnut and coffee cliché. But consider how
detectives might use a blackboard to coordinate and solve a murder
investigation.

Suppose the chief inspector starts off by setting up a large blackboard

in the conference room. On it, he writes a single question:

H. D UMPTY (M ALE , E GG ): A CCIDENT OR MURDER ?

Did Humpty really fall, or was he pushed? Each detective may make
contributions to this potential murder mystery by adding facts, state-
ments from witnesses, any forensic evidence that might arise, and so
on. As the data accumulates, a detective might notice a connection and
post that observation or speculation as well. This process continues,
across all shifts, with many different people and agents, until the case
is closed. A sample blackboard is shown in Figure 5.6 on the next page.

Some key features of the blackboard approach are:

None of the detectives needs to know of the existence of any other

detective—they watch the board for new information, and add their
findings.

The detectives may be trained in different disciplines, may have

different levels of education and expertise, and may not even work
in the same precinct. They share a desire to solve the case, but
that’s all.

Prepared exclusively for Zach

 166 C HAPTER 5 B END , OR B REAK

Figure 5.6. Someone found a connection between Humpty’s gambling debts and
the phone logs. Perhaps he was getting threatening phone calls.

H. Dumpty (Male, Egg): Accident or Murder?

Photos Shell fragments Phone logs
King’s men Gambling debts
Eyewitnesses Graffiti Wife’s alibi

Detective 1 Detective 2 Detective 3

Different detectives may come and go during the course of the pro-
cess, and may work different shifts.

There are no restrictions on what may be placed on the blackboard.

It may be pictures, sentences, physical evidence, and so on.

We’ve worked on a number of projects that involved a workflow or

distributed data gathering process. With each, designing a solution
around a simple blackboard model gave us a solid metaphor to work
with: all of the features listed above using detectives are just as appli-
cable to objects and code modules.

A blackboard system lets us decouple our objects from each other com-
pletely, providing a forum where knowledge consumers and producers
can exchange data anonymously and asynchronously. As you might
guess, it also cuts down on the amount of code we have to write.

Blackboard Implementations
Computer-based blackboard systems were originally invented for use
in artificial intelligence applications where the problems to be solved
were large and complex—speech recognition, knowledge-based reason-
ing systems, and so on.

Modern distributed blackboard-like systems such as JavaSpaces and

T Spaces [URL 50, URL 25] are based on a model of key/value pairs first

Prepared exclusively for Zach

 B LACKBOARDS 167

popularized in Linda [CG90], where the concept was known as tuple

space.

With these systems, you can store active Java objects—not just data—
on the blackboard, and retrieve them by partial matching of fields (via
templates and wildcards) or by subtypes. For example, suppose you
had a type Author, which is a subtype of Person. You could search
a blackboard containing Person objects by using an Author template
with a lastName value of “Shakespeare.” You’d get Bill Shakespeare
the author, but not Fred Shakespeare the gardener.

The main operations in JavaSpaces are:

Name Function
read Search for and retrieve data from the space.
write Put an item into the space.
take Similar to read, but removes the item from
the space as well.
notify Set up a notification to occur whenever an
object is written that matches the template.

T Spaces supports a similar set of operations, but with different names

and slightly different semantics. Both systems are built like a database
product; they provide atomic operations and distributed transactions
to ensure data integrity.

Since we can store objects, we can use a blackboard to design algo-

rithms based on a flow of objects, not just data. It’s as if our detec-
tives could pin people to the blackboard—witnesses themselves, not
just their statements. Anyone can ask a witness questions in the pur-
suit of the case, post the transcript, and move that witness to another
area of the blackboard, where he might respond differently (if you allow
the witness to read the blackboard too).

A big advantage of systems such as these is that you have a single, con-
sistent interface to the blackboard. When building a conventional dis-
tributed application, you can spend a great deal of time crafting unique
API calls for every distributed transaction and interaction in the sys-
tem. With the combinatorial explosion of interfaces and interactions,
the project can quickly become a nightmare.

Prepared exclusively for Zach

 168 C HAPTER 5 B END , OR B REAK

Organizing Your Blackboard

When the detectives work on large cases, the blackboard may be-
come cluttered, and it may become difficult to locate data on the
board. The solution is to partition the blackboard and start to orga-
nize the data on the blackboard somehow.
Different software systems handle this partitioning in different ways;
some use fairly flat zones or interest groups, while others adopt a
more hierarchical treelike structure.

The blackboard style of programming removes the need for so many

interfaces, making for a more elegant and consistent system.

Application Example
Suppose we are writing a program to accept and process mortgage or
loan applications. The laws that govern this area are odiously com-
plex, with federal, state, and local governments all having their say.
The lender must prove they have disclosed certain things, and must
ask for certain information—but must not ask certain other questions,
and so on, and so on.

Beyond the miasma of applicable law, we also have the following prob-
lems to contend with.

There is no guarantee on the order in which data arrives. For

instance, queries for a credit check or title search may take a sub-
stantial amount of time, while items such as name and address
may be available immediately.

Data gathering may be done by different people, distributed across

different offices, in different time zones.

Some data gathering may be done automatically by other systems.

This data may arrive asynchronously as well.

Nonetheless, certain data may still be dependent on other data. For

instance, you may not be able to start the title search for a car until
you get proof of ownership or insurance.

Prepared exclusively for Zach

 B LACKBOARDS 169

Arrival of new data may raise new questions and policies. Suppose
the credit check comes back with a less than glowing report; now
you need these five extra forms and perhaps a blood sample.

You can try to handle every possible combination and circumstance

using a workflow system. Many such systems exist, but they can be
complex and programmer intensive. As regulations change, the work-
flow must be reorganized: people may have to change their procedures
and hard-wired code may have to be rewritten.

A blackboard, in combination with a rules engine that encapsulates the

legal requirements, is an elegant solution to the difficulties found here.
Order of data arrival is irrelevant: when a fact is posted it can trigger
the appropriate rules. Feedback is easily handled as well: the output of
any set of rules can post to the blackboard and cause the triggering of
yet more applicable rules.

TIP 43

Use Blackboards to Coordinate Workflow

We can use the blackboard to coordinate disparate facts and agents,

while still maintaining independence and even isolation among partici-
pants.

You can accomplish the same results with more brute-force methods,
of course, but you’ll have a more brittle system. When it breaks, all
the king’s horses and all the king’s men might not get your program
working again.

Related sections include:

The Power of Plain Text, page 73
It’s Just a View, page 157

Challenges
Do you use blackboard systems in the real world—the message board by
the refrigerator, or the big whiteboard at work? What makes them effective?
Are messages ever posted with a consistent format? Does it matter?

Prepared exclusively for Zach

 170 C HAPTER 5 B END , OR B REAK

Exercises
Answer 30. For each of the following applications, would a blackboard system be ap-
on p. 297
propriate or not? Why?

1. Image processing. You’d like to have a number of parallel processes

grab chunks of an image, process them, and put the completed chunk
back.

2. Group calendaring. You’ve got people scattered across the globe,

in different time zones, and speaking different languages, trying to
schedule a meeting.

3. Network monitoring tool. The system gathers performance statistics

and collects trouble reports. You’d like to implement some agents to
use this information to look for trouble in the system.

Prepared exclusively for Zach

 Chapter 6

While You Are Coding

Conventional wisdom says that once a project is in the coding phase,
the work is mostly mechanical, transcribing the design into executable
statements. We think that this attitude is the single biggest reason that
many programs are ugly, inefficient, poorly structured, unmaintain-
able, and just plain wrong.

Coding is not mechanical. If it were, all the CASE tools that people
pinned their hopes on in the early 1980s would have replaced program-
mers long ago. There are decisions to be made every minute—decisions
that require careful thought and judgment if the resulting program is
to enjoy a long, accurate, and productive life.

Developers who don’t actively think about their code are programming
by coincidence—the code might work, but there’s no particular rea-
son why. In Programming by Coincidence, we advocate a more positive
involvement with the coding process.

While most of the code we write executes quickly, we occasionally

develop algorithms that have the potential to bog down even the fastest
processors. In Algorithm Speed, we discuss ways to estimate the speed
of code, and we give some tips on how to spot potential problems before
they happen.

Pragmatic Programmers think critically about all code, including our

own. We constantly see room for improvement in our programs and
our designs. In Refactoring, we look at techniques that help us fix up
existing code even while we’re in the midst of a project.

Something that should be in the back of your mind whenever you’re

producing code is that you’ll someday have to test it. Make code easy

171

Prepared exclusively for Zach

 172 C HAPTER 6 W HILE Y OU A RE C ODING

to test, and you’ll increase the likelihood that it will actually get tested,
a thought we develop in Code That’s Easy to Test.

Finally, in Evil Wizards, we suggest that you should be careful of tools

that write reams of code on your behalf unless you understand what
they’re doing.

Most of us can drive a car largely on autopilot—we don’t explicitly com-

mand our foot to press a pedal, or our arm to turn the wheel—we just
think “slow down and turn right.” However, good, safe drivers are con-
stantly reviewing the situation, checking for potential problems, and
putting themselves into good positions in case the unexpected happens.
The same is true of coding—it may be largely routine, but keeping your
wits about you could well prevent a disaster.

31 Programming by Coincidence
Do you ever watch old black-and-white war movies? The weary sol-
dier advances cautiously out of the brush. There’s a clearing ahead:
are there any land mines, or is it safe to cross? There aren’t any indica-
tions that it’s a minefield—no signs, barbed wire, or craters. The soldier
pokes the ground ahead of him with his bayonet and winces, expecting
an explosion. There isn’t one. So he proceeds painstakingly through the
field for a while, prodding and poking as he goes. Eventually, convinced
that the field is safe, he straightens up and marches proudly forward,
only to be blown to pieces.

The soldier’s initial probes for mines revealed nothing, but this was
merely lucky. He was led to a false conclusion—with disastrous results.

As developers, we also work in minefields. There are hundreds of traps

just waiting to catch us each day. Remembering the soldier’s tale, we
should be wary of drawing false conclusions. We should avoid pro-
gramming by coincidence—relying on luck and accidental successes—
in favor of programming deliberately.

Prepared exclusively for Zach

 P ROGRAMMING BY C OINCIDENCE 173

How to Program by Coincidence

Suppose Fred is given a programming assignment. Fred types in some
code, tries it, and it seems to work. Fred types in some more code, tries
it, and it still seems to work. After several weeks of coding this way,
the program suddenly stops working, and after hours of trying to fix it,
he still doesn’t know why. Fred may well spend a significant amount of
time chasing this piece of code around without ever being able to fix it.
No matter what he does, it just doesn’t ever seem to work right.

Fred doesn’t know why the code is failing because he didn’t know why
it worked in the first place. It seemed to work, given the limited “testing”
that Fred did, but that was just a coincidence. Buoyed by false confi-
dence, Fred charged ahead into oblivion. Now, most intelligent people
may know someone like Fred, but we know better. We don’t rely on
coincidences—do we?

Sometimes we might. Sometimes it can be pretty easy to confuse a

happy coincidence with a purposeful plan. Let’s look at a few examples.

Accidents of Implementation
Accidents of implementation are things that happen simply because
that’s the way the code is currently written. You end up relying on
undocumented error or boundary conditions.

Suppose you call a routine with bad data. The routine responds in a
particular way, and you code based on that response. But the author
didn’t intend for the routine to work that way—it was never even con-
sidered. When the routine gets “fixed,” your code may break. In the
most extreme case, the routine you called may not even be designed
to do what you want, but it seems to work okay. Calling things in the
wrong order, or in the wrong context, is a related problem.

paint(g);
invalidate();
validate();
revalidate();
repaint();
paintImmediately(r);

Here it looks like Fred is desperately trying to get something out on the
screen. But these routines were never designed to be called this way;
although they seem to work, that’s really just a coincidence.

Prepared exclusively for Zach

 174 C HAPTER 6 W HILE Y OU A RE C ODING

To add insult to injury, when the component finally does get drawn,
Fred won’t try to go back and take out the spurious calls. “It works
now, better leave well enough alone. . . .”

It’s easy to be fooled by this line of thought. Why should you take the
risk of messing with something that’s working? Well, we can think of
several reasons:

It may not really be working—it might just look like it is.

The boundary condition you rely on may be just an accident. In

different circumstances (a different screen resolution, perhaps), it
might behave differently.

Undocumented behavior may change with the next release of the

library.

Additional and unnecessary calls make your code slower.

Additional calls also increase the risk of introducing new bugs of

their own.

For code you write that others will call, the basic principles of good
modularization and of hiding implementation behind small, well-docu-
mented interfaces can all help. A well-specified contract (see Design by
Contract, page 109) can help eliminate misunderstandings.

For routines you call, rely only on documented behavior. If you can’t,
for whatever reason, then document your assumption well.

Accidents of Context
You can have “accidents of context” as well. Suppose you are writing a
utility module. Just because you are currently coding for a GUI envi-
ronment, does the module have to rely on a GUI being present? Are you
relying on English-speaking users? Literate users? What else are you
relying on that isn’t guaranteed?

Implicit Assumptions
Coincidences can mislead at all levels—from generating requirements
through to testing. Testing is particularly fraught with false causalities
and coincidental outcomes. It’s easy to assume that X causes Y, but as
we said in Debugging, page 90: don’t assume it, prove it.

Prepared exclusively for Zach

 P ROGRAMMING BY C OINCIDENCE 175

At all levels, people operate with many assumptions in mind—but these

assumptions are rarely documented and are often in conflict between
different developers. Assumptions that aren’t based on well-established
facts are the bane of all projects.

TIP 44

Don’t Program by Coincidence

How to Program Deliberately

We want to spend less time churning out code, catch and fix errors as
early in the development cycle as possible, and create fewer errors to
begin with. It helps if we can program deliberately:

Always be aware of what you are doing. Fred let things get slowly
out of hand, until he ended up boiled, like the frog in Stone Soup
and Boiled Frogs, page 7.

Don’t code blindfolded. Attempting to build an application you don’t

fully understand, or to use a technology you aren’t familiar with, is
an invitation to be misled by coincidences.

Proceed from a plan, whether that plan is in your head, on the back
of a cocktail napkin, or on a wall-sized printout from a CASE tool.

Rely only on reliable things. Don’t depend on accidents or assump-

tions. If you can’t tell the difference in particular circumstances,
assume the worst.

Document your assumptions. Design by Contract, page 109, can

help clarify your assumptions in your own mind, as well as help
communicate them to others.

Don’t just test your code, but test your assumptions as well. Don’t
guess; actually try it. Write an assertion to test your assumptions
(see Assertive Programming, page 122). If your assertion is right,
you have improved the documentation in your code. If you discover
your assumption is wrong, then count yourself lucky.

Prioritize your effort. Spend time on the important aspects; more

than likely, these are the hard parts. If you don’t have fundamen-

Prepared exclusively for Zach

 176 C HAPTER 6 W HILE Y OU A RE C ODING

tals or infrastructure correct, brilliant bells and whistles will be

irrelevant.

Don’t be a slave to history. Don’t let existing code dictate future

code. All code can be replaced if it is no longer appropriate. Even
within one program, don’t let what you’ve already done constrain
what you do next—be ready to refactor (see Refactoring, page 184).
This decision may impact the project schedule. The assumption is
that the impact will be less than the cost of not making the change.1

So next time something seems to work, but you don’t know why, make
sure it isn’t just a coincidence.

Related sections include:

Stone Soup and Boiled Frogs, page 7
Debugging, page 90
Design by Contract, page 109
Assertive Programming, page 122
Temporal Coupling, page 150
Refactoring, page 184
It’s All Writing, page 248

Exercises
Answer 31. Can you identify some coincidences in the following C code fragment?
on p. 298
Assume that this code is buried deep in a library routine.

fprintf(stderr,"Error, continue?");
gets(buf);

Answer 32. This piece of C code might work some of the time, on some machines. Then
on p. 298
again, it might not. What’s wrong?

/* Truncate string to its last maxlen chars */

void string_tail(char *string, int maxlen) {
int len = strlen(string);
if (len > maxlen) {
strcpy(string, string + (len - maxlen));
}
}

1. You can also go too far here. We once knew a developer who rewrote all source he
was given because he had his own naming conventions.

Prepared exclusively for Zach

 A LGORITHM S PEED 177

33. This code comes from a general-purpose Java tracing suite. The function Answer
on p. 299
writes a string to a log file. It passes its unit test, but fails when one of the
Web developers uses it. What coincidence does it rely on?

public static void debug(String s) throws IOException {

FileWriter fw = new FileWriter("debug.log", true);
fw.write(s);
fw.flush();
fw.close();
}

32 Algorithm Speed
In Estimating, page 64, we talked about estimating things such as how
long it takes to walk across town, or how long a project will take to fin-
ish. However, there is another kind of estimating that Pragmatic Pro-
grammers use almost daily: estimating the resources that algorithms
use—time, processor, memory, and so on.

This kind of estimating is often crucial. Given a choice between two

ways of doing something, which do you pick? You know how long your
program runs with 1,000 records, but how will it scale to 1,000,000?
What parts of the code need optimizing?

It turns out that these questions can often be answered using common
sense, some analysis, and a way of writing approximations called the
“big O” notation.

What Do We Mean by Estimating Algorithms?

Most nontrivial algorithms handle some kind of variable input—sorting
strings, inverting an matrix, or decrypting a message with an
-bit key. Normally, the size of this input will affect the algorithm: the
larger the input, the longer the running time or the more memory used.

If the relationship were always linear (so that the time increased in
direct proportion to the value of ), this section wouldn’t be important.
However, most significant algorithms are not linear. The good news is
that many are sublinear. A binary search, for example, doesn’t need to
look at every candidate when finding a match. The bad news is that

Prepared exclusively for Zach

 178 C HAPTER 6 W HILE Y OU A RE C ODING

other algorithms are considerably worse than linear; runtimes or mem-

ory requirements increase far faster than . An algorithm that takes a
minute to process ten items may take a lifetime to process 100.

We find that whenever we write anything containing loops or recur-

sive calls, we subconsciously check the runtime and memory require-
ments. This is rarely a formal process, but rather a quick confirma-
tion that what we’re doing is sensible in the circumstances. However,
we sometimes do find ourselves performing a more detailed analysis.
That’s when the notation comes in useful.

The O ( ) Notation
The notation is a mathematical way of dealing with approximations.
When we write that a particular sort routine sorts records in
time, we are simply saying that the worst-case time taken will vary
as the square of . Double the number of records, and the time will
increase roughly fourfold. Think of the as meaning on the order of.
The notation puts an upper bound on the value of the thing we’re
measuring (time, memory, and so on). If we say a function takes
time, then we know that the upper bound of the time it takes will not
grow faster than . Sometimes we come up with fairly complex
functions, but because the highest-order term will dominate the value
as increases, the convention is to remove all low-order terms, and not
to bother showing any constant multiplying factors. is the
same as , which is equivalent to . This is actually a weakness
of the notation—one algorithm may be 1,000 times faster than
another algorithm, but you won’t know it from the notation.

Figure 6.1 shows several common notations you’ll come across,

along with a graph comparing running times of algorithms in each cat-
egory. Clearly, things quickly start getting out of hand once we get over
.

For example, suppose you’ve got a routine that takes 1 s to process 100
records. How long will it take to process 1,000? If your code is , then
it will still take 1 s. If it’s , then you’ll probably be waiting about
3 s. will show a linear increase to 10 s, while an will take
some 33 s. If you’re unlucky enough to have an routine, then sit
back for 100 s while it does its stuff. And if you’re using an exponential

Prepared exclusively for Zach

 A LGORITHM S PEED 179

Figure 6.1. Runtimes of various algorithms

: traveling salesman : heapsort

: selection sort

: sequential search

: binary search

: array access

Some common O ( ) notations

Constant (access element in array, simple

statements)
Logarithmic (binary search) [The notation
is shorthand for ]
Linear (sequential search)
Worse than linear, but not much worse (aver-
age runtime of quicksort, heapsort)
Square law (selection and insertion sorts)
Cubic (multiplication of 2 matrices)
Exponential (traveling salesman problem, set
partitioning)

Prepared exclusively for Zach

 180 C HAPTER 6 W HILE Y OU A RE C ODING

algorithm , you might want to make a cup of coffee—your routine

should finish in about years. Let us know how the universe ends.

The notation doesn’t apply just to time; you can use it to represent
any other resources used by an algorithm. For example, it is often use-
ful to be able to model memory consumption (see Exercise 35 on page
183).

Common Sense Estimation

You can estimate the order of many basic algorithms using common
sense.

Simple loops. If a simple loop runs from to , then the algo-

rithm is likely to be —time increases linearly with . Exam-
ples include exhaustive searches, finding the maximum value in
an array, and generating checksums.

Nested loops. If you nest a loop inside another, then your algo-
rithm becomes , where and are the two loops’ limits.
This commonly occurs in simple sorting algorithms, such as bub-
ble sort, where the outer loop scans each element in the array in
turn, and the inner loop works out where to place that element in
the sorted result. Such sorting algorithms tend to be .

Binary chop. If your algorithm halves the set of things it consid-

ers each time around the loop, then it is likely to be logarithmic,
(see Exercise 37, page 183). A binary search of a sorted list,
traversing a binary tree, and finding the first set bit in a machine
word can all be .

Divide and conquer. Algorithms that partition their input, work

on the two halves independently, and then combine the result can
be . The classic example is quicksort, which works by
partitioning the data into two halves and recursively sorting each.
Although technically , because its behavior degrades when it
is fed sorted input, the average runtime of quicksort is .

Combinatoric. Whenever algorithms start looking at the permuta-

tions of things, their running times may get out of hand. This is be-
cause permutations involve factorials (there are
permutations of the digits from 1 to 5). Time a combinatoric

Prepared exclusively for Zach

 A LGORITHM S PEED 181

algorithm for five elements: it will take six times longer to run it for
six, and 42 times longer for seven. Examples include algorithms for
many of the acknowledged hard problems—the traveling salesman
problem, optimally packing things into a container, partitioning a
set of numbers so that each set has the same total, and so on. Of-
ten, heuristics are used to reduce the running times of these types
of algorithms in particular problem domains.

Algorithm Speed in Practice

It’s unlikely that you’ll spend much time during your career writing sort
routines. The ones in the libraries available to you will probably outper-
form anything you may write without substantial effort. However, the
basic kinds of algorithms we’ve described earlier pop up time and time
again. Whenever you find yourself writing a simple loop, you know that
you have an algorithm. If that loop contains an inner loop, then
you’re looking at . You should be asking yourself how large
these values can get. If the numbers are bounded, then you’ll know
how long the code will take to run. If the numbers depend on external
factors (such as the number of records in an overnight batch run, or
the number of names in a list of people), then you might want to stop
and consider the effect that large values may have on your running
time or memory consumption.

TIP 45

Estimate the Order of Your Algorithms

There are some approaches you can take to address potential problems.
If you have an algorithm that is , try to find a divide and conquer
approach that will take you down to .

If you’re not sure how long your code will take, or how much memory
it will use, try running it, varying the input record count or whatever is
likely to impact the runtime. Then plot the results. You should soon get
a good idea of the shape of the curve. Is it curving upward, a straight
line, or flattening off as the input size increases? Three or four points
should give you an idea.

Also consider just what you’re doing in the code itself. A simple
loop may well perform better than a complex, one for smaller

Prepared exclusively for Zach

 182 C HAPTER 6 W HILE Y OU A RE C ODING

values of , particularly if the algorithm has an expensive

inner loop.

In the middle of all this theory, don’t forget that there are practical
considerations as well. Runtime may look like it increases linearly for
small input sets. But feed the code millions of records and suddenly
the time degrades as the system starts to thrash. If you test a sort
routine with random input keys, you may be surprised the first time
it encounters ordered input. Pragmatic Programmers try to cover both
the theoretical and practical bases. After all this estimating, the only
timing that counts is the speed of your code, running in the production
environment, with real data.2 This leads to our next tip.

TIP 46

Test Your Estimates

If it’s tricky getting accurate timings, use code profilers to count the
number of times the different steps in your algorithm get executed,
and plot these figures against the size of the input.

Best Isn’t Always Best

You also need to be pragmatic about choosing appropriate algorithms—
the fastest one is not always the best for the job. Given a small input
set, a straightforward insertion sort will perform just as well as a quick-
sort, and will take you less time to write and debug. You also need to
be careful if the algorithm you choose has a high setup cost. For small
input sets, this setup may dwarf the running time and make the algo-
rithm inappropriate.

Also be wary of premature optimization. It’s always a good idea to make

sure an algorithm really is a bottleneck before investing your precious
time trying to improve it.

2. In fact, while testing the sort algorithms used as an exercise for this section on
a 64MB Pentium, the authors ran out of real memory while running the radix sort with
more than seven million numbers. The sort started using swap space, and times degraded
dramatically.

Prepared exclusively for Zach

 A LGORITHM S PEED 183

Related sections include:

Estimating, page 64

Challenges
Every developer should have a feel for how algorithms are designed and
analyzed. Robert Sedgewick has written a series of accessible books on the
subject ([Sed83, SF96, Sed92] and others). We recommend adding one of
his books to your collection, and making a point of reading it.

For those who like more detail than Sedgewick provides, read Donald
Knuth’s definitive Art of Computer Programming books, which analyze a
wide range of algorithms [Knu97a, Knu97b, Knu98].

In Exercise 34, we look at sorting arrays of long integers. What is the

impact if the keys are more complex, and the overhead of key comparison
is high? Does the key structure affect the efficiency of the sort algorithms,
or is the fastest sort always fastest?

Exercises
34. We have coded a set of simple sort routines, which can be downloaded Answer
on p. 299
from our Web site (www.pragmaticprogrammer.com). Run them on vari-
ous machines available to you. Do your figures follow the expected curves?
What can you deduce about the relative speeds of your machines? What
are the effects of various compiler optimization settings? Is the radix sort
indeed linear?

35. The routine below prints out the contents of a binary tree. Assuming the Answer
on p. 300
tree is balanced, roughly how much stack space will the routine use while
printing a tree of 1,000,000 elements? (Assume that subroutine calls im-
pose no significant stack overhead.)

void printTree(const Node *node) {

char buffer[1000];
if (node) {
printTree(node->left);
getNodeAsString(node, buffer);
puts(buffer);
printTree(node->right);
}
}

36. Can you see any way to reduce the stack requirements of the routine in Answer
on p. 300
Exercise 35 (apart from reducing the size of the buffer)?

37. On page 180, we claimed that a binary chop is . Can you prove Answer
on p. 301
this?

Prepared exclusively for Zach

 184 C HAPTER 6 W HILE Y OU A RE C ODING

33 Refactoring
Change and decay in all around I see . . .
H. F. Lyte, “Abide With Me”

As a program evolves, it will become necessary to rethink earlier deci-

sions and rework portions of the code. This process is perfectly natural.
Code needs to evolve; it’s not a static thing.

Unfortunately, the most common metaphor for software development

is building construction (Bertrand Meyer [Mey97b] uses the term “Soft-
ware Construction”). But using construction as the guiding metaphor
implies these steps:

1. An architect draws up blueprints.

2. Contractors dig the foundation, build the superstructure, wire and

plumb, and apply finishing touches.

3. The tenants move in and live happily ever after, calling building
maintenance to fix any problems.

Well, software doesn’t quite work that way. Rather than construction,
software is more like gardening—it is more organic than concrete. You
plant many things in a garden according to an initial plan and condi-
tions. Some thrive, others are destined to end up as compost. You may
move plantings relative to each other to take advantage of the inter-
play of light and shadow, wind and rain. Overgrown plants get split
or pruned, and colors that clash may get moved to more aesthetically
pleasing locations. You pull weeds, and you fertilize plantings that are
in need of some extra help. You constantly monitor the health of the
garden, and make adjustments (to the soil, the plants, the layout) as
needed.

Business people are comfortable with the metaphor of building con-

struction: it is more scientific than gardening, it’s repeatable, there’s
a rigid reporting hierarchy for management, and so on. But we’re not
building skyscrapers—we aren’t as constrained by the boundaries of
physics and the real world.

The gardening metaphor is much closer to the realities of software

development. Perhaps a certain routine has grown too large, or is trying

Prepared exclusively for Zach

 R EFACTORING 185

to accomplish too much—it needs to be split into two. Things that don’t
work out as planned need to be weeded or pruned.

Rewriting, reworking, and re-architecting code is collectively known as

refactoring.

When Should You Refactor?

When you come across a stumbling block because the code doesn’t
quite fit anymore, or you notice two things that should really be merged,
or anything else at all strikes you as being “wrong,” don’t hesitate to
change it. There’s no time like the present. Any number of things may
cause code to qualify for refactoring:

Duplication. You’ve discovered a violation of the DRY principle

(The Evils of Duplication, page 26).

Nonorthogonal design. You’ve discovered some code or design

that could be made more orthogonal (Orthogonality, page 34).

Outdated knowledge. Things change, requirements drift, and your

knowledge of the problem increases. Code needs to keep up.

Performance. You need to move functionality from one area of the

system to another to improve performance.

Refactoring your code—moving functionality around and updating ear-

lier decisions—is really an exercise in pain management. Let’s face it,
changing source code around can be pretty painful: it was almost work-
ing, and now it’s really torn up. Many developers are reluctant to start
ripping up code just because it isn’t quite right.

Real-World Complications
So you go to your boss or client and say, “This code works, but I need
another week to refactor it.”

We can’t print their reply.

Time pressure is often used as an excuse for not refactoring. But this
excuse just doesn’t hold up: fail to refactor now, and there’ll be a far
greater time investment to fix the problem down the road—when there
are more dependencies to reckon with. Will there be more time available
then? Not in our experience.

Prepared exclusively for Zach

 186 C HAPTER 6 W HILE Y OU A RE C ODING

You might want to explain this principle to the boss by using a medical
analogy: think of the code that needs refactoring as a “growth.” Remov-
ing it requires invasive surgery. You can go in now, and take it out while
it is still small. Or, you could wait while it grows and spreads—but re-
moving it then will be both more expensive and more dangerous. Wait
even longer, and you may lose the patient entirely.

TIP 47

Refactor Early, Refactor Often

Keep track of the things that need to be refactored. If you can’t refactor
something immediately, make sure that it gets placed on the schedule.
Make sure that users of the affected code know that it is scheduled to
be refactored and how this might affect them.

How Do You Refactor?

Refactoring started out in the Smalltalk community, and, along with
other trends (such as design patterns), has started to gain a wider au-
dience. But as a topic it is still fairly new; there isn’t much published
on it. The first major book on refactoring ([FBB 99], and also [URL 47])
is being published around the same time as this book.

At its heart, refactoring is redesign. Anything that you or others on your

team designed can be redesigned in light of new facts, deeper under-
standings, changing requirements, and so on. But if you proceed to rip
up vast quantities of code with wild abandon, you may find yourself in
a worse position than when you started.

Clearly, refactoring is an activity that needs to be undertaken slowly,

deliberately, and carefully. Martin Fowler offers the following simple
tips on how to refactor without doing more harm than good (see the
box on page 30 in [FS97]):

1. Don’t try to refactor and add functionality at the same time.

2. Make sure you have good tests before you begin refactoring. Run
the tests as often as possible. That way you will know quickly if
your changes have broken anything.

Prepared exclusively for Zach

 R EFACTORING 187

Automatic Refactoring
Historically, Smalltalk users have always enjoyed a class browser
as part of the IDE. Not to be confused with Web browsers, class
browsers let users navigate through and examine class hierarchies
and methods.
Typically, class browsers allow you to edit code, create new methods
and classes, and so on. The next variation on this idea is the refac-
toring browser.
A refactoring browser can semiautomatically perform common refac-
toring operations for you: splitting up a long routine into smaller ones,
automatically propagating changes to method and variable names,
drag and drop to assist you in moving code, and so on.
As we write this book, this technology has yet to appear outside of
the Smalltalk world, but this is likely to change at the same speed
that Java changes—rapidly. In the meantime, the pioneering Small-
talk refactoring browser can be found online at [URL 20].

3. Take short, deliberate steps: move a field from one class to another,
fuse two similar methods into a superclass. Refactoring often in-
volves making many localized changes that result in a larger-scale
change. If you keep your steps small, and test after each step, you
will avoid prolonged debugging.

We’ll talk more about testing at this level in Code That’s Easy to Test,
page 189, and larger-scale testing in Ruthless Testing, page 237, but
Mr. Fowler’s point of maintaining good regression tests is the key to
refactoring with confidence.

It can also be helpful to make sure that drastic changes to a module—

such as altering its interface or its functionality in an incompatible
manner—break the build. That is, old clients of this code should fail
to compile. You can then quickly find the old clients and make the
necessary changes to bring them up to date.

So next time you see a piece of code that isn’t quite as it should be, fix
both it and everything that depends on it. Manage the pain: if it hurts
now, but is going to hurt even more later, you might as well get it over

Prepared exclusively for Zach

 Copyright 2002 Addison Wesley Longman, Inc

188 C HAPTER 6 W HILE Y OU A RE C ODING

with. Remember the lessons of Software Entropy, page 4: don’t live with
broken windows.

Related sections include:

The Cat Ate My Source Code, page 2
Software Entropy, page 4
Stone Soup and Boiled Frogs, page 7
The Evils of Duplication, page 26
Orthogonality, page 34
Programming by Coincidence, page 172
Code That’s Easy to Test, page 189
Ruthless Testing, page 237

Exercises
Answer 38. The following code has obviously been updated several times over the
on p. 302
years, but the changes haven’t improved its structure. Refactor it.
if (state == TEXAS) {
rate = TX_RATE;
amt = base * TX_RATE;
calc = 2*basis(amt) + extra(amt)*1.05;
}
else if ((state == OHIO) || (state == MAINE)) {
rate = (state == OHIO) ? OH_RATE : ME_RATE;
amt = base * rate;
calc = 2*basis(amt) + extra(amt)*1.05;
if (state == OHIO)
points = 2;
}
else {
rate = 1;
amt = base;
calc = 2*basis(amt) + extra(amt)*1.05;
}

Answer 39. The following Java class needs to support a few more shapes. Refactor the
on p. 303
class to prepare it for the additions.
public class Shape {
public static final int SQUARE = 1;
public static final int CIRCLE = 2;
public static final int RIGHT_TRIANGLE = 3;
private int shapeType;
private double size;
public Shape(int shapeType, double size) {
this.shapeType = shapeType;
this.size = size;
}
// ... other methods ...

Prepared exclusively for Zach

 C ODE T HAT ’ S E ASY TO T EST 189

public double area() {

switch (shapeType) {
case SQUARE: return size*size;
case CIRCLE: return Math.PI*size*size/4.0;
case RIGHT_TRIANGLE: return size*size/2.0;
}
return 0;
}
}

40. This Java code is part of a framework that will be used throughout your Answer
on p. 303
project. Refactor it to be more general and easier to extend in the future.

public class Window {

public Window(int width, int height) { ... }
public void setSize(int width, int height) { ... }
public boolean overlaps(Window w) { ... }
public int getArea() { ... }
}

34 Code That’s Easy to Test

The Software IC is a metaphor that people like to toss around when
discussing reusability and component-based development.3 The idea is
that software components should be combined just as integrated circuit
chips are combined. This works only if the components you are using
are known to be reliable.

Chips are designed to be tested—not just at the factory, not just when
they are installed, but also in the field when they are deployed. More
complex chips and systems may have a full Built-In Self Test (BIST) fea-
ture that runs some base-level diagnostics internally, or a Test Access
Mechanism (TAM) that provides a test harness that allows the external
environment to provide stimuli and collect responses from the chip.

We can do the same thing in software. Like our hardware colleagues,

we need to build testability into the software from the very beginning,
and test each piece thoroughly before trying to wire them together.

3. The term “Software IC” (Integrated Circuit) seems to have been invented in 1986 by
Cox and Novobilski in their Objective-C book Object-Oriented Programming [CN91].

Prepared exclusively for Zach

 190 C HAPTER 6 W HILE Y OU A RE C ODING

Unit Testing
Chip-level testing for hardware is roughly equivalent to unit testing
in software—testing done on each module, in isolation, to verify its
behavior. We can get a better feeling for how a module will react in
the big wide world once we have tested it throughly under controlled
(even contrived) conditions.

A software unit test is code that exercises a module. Typically, the unit
test will establish some kind of artificial environment, then invoke rou-
tines in the module being tested. It then checks the results that are
returned, either against known values or against the results from pre-
vious runs of the same test (regression testing).

Later, when we assemble our “software IC’s” into a complete system,

we’ll have confidence that the individual parts work as expected, and
then we can use the same unit test facilities to test the system as a
whole. We talk about this large-scale checking of the system in Ruthless
Testing, page 237.

Before we get that far, however, we need to decide what to test at the
unit level. Typically, programmers throw a few random bits of data at
the code and call it tested. We can do much better, using the ideas
behind design by contract.

Testing Against Contract

We like to think of unit testing as testing against contract (see Design by
Contract, page 109). We want to write test cases that ensure that a given
unit honors its contract. This will tell us two things: whether the code
meets the contract, and whether the contract means what we think it
means. We want to test that the module delivers the functionality it
promises, over a wide range of test cases and boundary conditions.

What does this mean in practice? Let’s look at the square root routine
we first encountered on page 114. Its contract is simple:
require
argument >= 0;
ensure
((Result * Result) - argument).abs <= epsilon*argument;

This tells us what to test:

Prepared exclusively for Zach

 C ODE T HAT ’ S E ASY TO T EST 191

Pass in a negative argument and ensure that it is rejected.

Pass in an argument of zero to ensure that it is accepted (this is

the boundary value).

Pass in values between zero and the maximum expressible argu-

ment and verify that the difference between the square of the result
and the original argument is less than some small fraction of the
argument.

Armed with this contract, and assuming that our routine does its own
pre- and postcondition checking, we can write a basic test script to
exercise the square root function.
public void testValue(double num, double expected) {
double result = 0.0;
try { // We may throw a
result = mySqrt(num); // precondition exception
}
catch (Throwable e) {
if (num < 0.0) // If input is < 0, then
return; // we’re expecting the
else // exception, otherwise
assert(false); // force a test failure
}
assert(Math.abs(expected-result) < epsilon*expected);
}

Then we can call this routine to test our square root function:
testValue(-4.0, 0.0);
testValue( 0.0, 0.0);
testValue( 2.0, 1.4142135624);
testValue(64.0, 8.0);
testValue(1.0e7, 3162.2776602);

This is a pretty simple test; in the real world, any nontrivial module is
likely to be dependent on a number of other modules, so how do we go
about testing the combination?

Suppose we have a module A that uses a LinkedList and a Sort. In

order, we would test:

1. LinkedList’s contract, in full

2. Sort’s contract, in full

3. A’s contract, which relies on the other contracts but does not di-
rectly expose them

Prepared exclusively for Zach

 192 C HAPTER 6 W HILE Y OU A RE C ODING

This style of testing requires you to test subcomponents of a module

first. Once the subcomponents have been verified, then the module
itself can be tested.

If LinkedList and Sort’s tests passed, but A’s test failed, we can be
pretty sure that the problem is in A, or in A’s use of one of those sub-
components. This technique is a great way to reduce debugging effort:
we can quickly concentrate on the likely source of the problem within
module A, and not waste time reexamining its subcomponents.

Why do we go to all this trouble? Above all, we want to avoid creating a

“time bomb”—something that sits around unnoticed and blows up at an
awkward moment later in the project. By emphasizing testing against
contract, we can try to avoid as many of those downstream disasters as
possible.

TIP 48

Design to Test

When you design a module, or even a single routine, you should design
both its contract and the code to test that contract. By designing code
to pass a test and fulfill its contract, you may well consider bound-
ary conditions and other issues that wouldn’t occur to you otherwise.
There’s no better way to fix errors than by avoiding them in the first
place. In fact, by building the tests before you implement the code, you
get to try out the interface before you commit to it.

Writing Unit Tests

The unit tests for a module shouldn’t be shoved in some far-away cor-
ner of the source tree. They need to be conveniently located. For small
projects, you can embed the unit test for a module in the module itself.
For larger projects, we suggest moving each test into a subdirectory.
Either way, remember that if it isn’t easy to find, it won’t be used.

By making the test code readily accessible, you are providing developers
who may use your code with two invaluable resources:

1. Examples of how to use all the functionality of your module

Prepared exclusively for Zach

 C ODE T HAT ’ S E ASY TO T EST 193

2. A means to build regression tests to validate any future changes to

the code

It’s convenient, but not always practical, for each class or module to
contain its own unit test. In Java, for example, every class can have its
own main. In all but the application’s main class file, the main routine
can be used to run unit tests; it will be ignored when the application
itself is run. This has the benefit that the code you ship still contains
the tests, which can be used to diagnose problems in the field.

In C++ you can achieve the same effect (at compile time) by using
#ifdef to compile unit test code selectively. For example, here’s a
very simple unit test in C++, embedded in our module, that checks our
square root function using a testValue routine similar to the Java one
defined previously:

#ifdef __TEST__
int main(int argc, char **argv)
{
argc--; argv++; // skip program name
if (argc < 2) { // do standard tests if no args
testValue(-4.0, 0.0);
testValue( 0.0, 0.0);
testValue( 2.0, 1.4142135624);
testValue(64.0, 8.0);
testValue(1.0e7, 3162.2776602);
}
else { // else use args
double num, expected;
while (argc >= 2) {
num = atof(argv[0]);
expected = atof(argv[1]);
testValue(num,expected);
argc -= 2;
argv += 2;
}
}
return 0;
}
#endif

This unit test will either run a minimal set of tests or, if given argu-
ments, allow you to pass data in from the outside world. A shell script
could use this ability to run a much more complete set of tests.

What do you do if the correct response for a unit test is to exit, or abort
the program? In that case, you need to be able to select the test to
run, perhaps by specifying an argument on the command line. You’ll

Prepared exclusively for Zach

 194 C HAPTER 6 W HILE Y OU A RE C ODING

also need to pass in parameters if you need to specify different starting

conditions for your tests.

But providing unit tests isn’t enough. You must run them, and run
them often. It also helps if the class passes its tests once in a while.

Using Test Harnesses

Because we usually write a lot of test code, and do a lot of testing, we’ll
make life easier on ourselves and develop a standard testing harness
for the project. The main shown in the previous section is a very simple
test harness, but usually we’ll need more functionality than that.

A test harness can handle common operations such as logging status,

analyzing output for expected results, and selecting and running the
tests. Harnesses may be GUI driven, may be written in the same target
language as the rest of the project, or may be implemented as a combi-
nation of makefiles and Perl scripts. A simple test harness is shown
in the answer to Exercise 41 on page 305.

In object-oriented languages and environments, you might create a

base class that provides these common operations. Individual tests can
subclass from that and add specific test code. You could use a standard
naming convention and reflection in Java to build a list of tests dynam-
ically. This technique is a nice way of honoring the DRY principle—you
don’t have to maintain a list of available tests. But before you go off and
start writing your own harness, you may want to investigate Kent Beck
and Erich Gamma’s xUnit at [URL 22]. You might also want to look at
our book Pragmatic Unit Testing [HT03] for an introduction to JUnit.

Regardless of the technology you decide to use, test harnesses should

include the following capabilities:

A standard way to specify setup and cleanup

A method for selecting individual tests or all available tests
A means of analyzing output for expected (or unexpected) results
A standardized form of failure reporting

Tests should be composable; that is, a test can be composed of subtests

of subcomponents to any depth. We can use this feature to test selected
parts of the system or the entire system just as easily, using the same
tools.

Prepared exclusively for Zach

 C ODE T HAT ’ S E ASY TO T EST 195

Ad Hoc Testing
During debugging, we may end up creating some particular tests on-
the-fly. These may be as simple as a print statement, or a piece of
code entered interactively in a debugger or IDE environment.
At the end of the debugging session, you need to formalize the ad
hoc test. If the code broke once, it is likely to break again. Don’t just
throw away the test you created; add it to the existing unit test.

For example, using JUnit (the Java member of the xUnit family), we
might write our square root test as follows:

public class JUnitExample extends TestCase {

public JUnitExample(final String name) {
super(name);
}
protected void setUp() {
// Load up test data...
testData.addElement(new DblPair(-4.0,0.0));
testData.addElement(new DblPair(0.0,0.0));
testData.addElement(new DblPair(64.0,8.0));
testData.addElement(new DblPair(Double.MAX_VALUE,
1.3407807929942597E154));
}
public void testMySqrt() {
double num, expected, result = 0.0;
Enumeration enum = testData.elements();
while (enum.hasMoreElements()) {
DblPair p = (DblPair)enum.nextElement();
num = p.getNum();
expected = p.getExpected();
testValue(num, expected);
}
}
public static Test suite() {
TestSuite suite= new TestSuite();
suite.addTest(new JUnitExample("testMySqrt"));
return suite;
}
}

JUnit is designed to be composable: we could add as many tests as we

wanted to this suite, and each of those tests could in turn be a suite.
In addition, you have your choice of a graphical or batch interface to
drive the tests.

Prepared exclusively for Zach

 196 C HAPTER 6 W HILE Y OU A RE C ODING

Build a Test Window

Even the best sets of tests are unlikely to find all the bugs; there’s some-
thing about the damp, warm conditions of a production environment
that seems to bring them out of the woodwork.

This means you’ll often need to test a piece of software once it has
been deployed—with real-world data flowing though its veins. Unlike a
circuit board or chip, we don’t have test pins in software, but we can
provide various views into the internal state of a module, without using
the debugger (which may be inconvenient or impossible in a production
application).

Log files containing trace messages are one such mechanism. Log mes-
sages should be in a regular, consistent format; you may want to parse
them automatically to deduce processing time or logic paths that the
program took. Poorly or inconsistently formatted diagnostics are just
so much “spew”—they are difficult to read and impractical to parse.

Another mechanism for getting inside running code is the “hot-key”

sequence. When this particular combination of keys is pressed, a diag-
nostic control window pops up with status messages and so on. This
isn’t something you normally would reveal to end users, but it can be
very handy for the help desk.

For larger, more complex server code, a nifty technique for providing a
view into its operation is to include a built-in Web server. Anyone can
point a Web browser to the application’s HTTP port (which is usually
on a nonstandard number, such as 8080) and see internal status, log
entries, and possibly even some sort of a debug control panel. This may
sound difficult to implement, but it’s not. Freely available and embed-
dable HTTP Web servers are available in a variety of modern languages.
A good place to start looking is [URL 58].

A Culture of Testing
All software you write will be tested—if not by you and your team,
then by the eventual users—so you might as well plan on testing it
thoroughly. A little forethought can go a long way toward minimizing
maintenance costs and help-desk calls.

Prepared exclusively for Zach

 C ODE T HAT ’ S E ASY TO T EST 197

Despite its hacker reputation, the Perl community has a very strong
commitment to unit and regression testing. The Perl standard module
installation procedure supports a regression test by invoking

% make test

There’s nothing magic about Perl itself in this regard. Perl makes it
easier to collate and analyze test results to ensure compliance, but the
big advantage is simply that it’s a standard—tests go in a particular
place, and have a certain expected output. Testing is more cultural than
technical; we can instill this testing culture in a project regardless of
the language being used.

TIP 49

Test Your Software, or Your Users Will

Related sections include:

The Cat Ate My Source Code, page 2
Orthogonality, page 34
Design by Contract, page 109
Refactoring, page 184
Ruthless Testing, page 237

Exercises
41. Design a test jig for the blender interface described in the answer to Exer- Answer
on p. 305
cise 17 on page 289. Write a shell script that will perform a regression test
for the blender. You need to test basic functionality, error and boundary
conditions, and any contractual obligations. What restrictions are placed
on changing the speed? Are they being honored?

Prepared exclusively for Zach

 198 C HAPTER 6 W HILE Y OU A RE C ODING

35 Evil Wizards
There’s no denying it—applications are getting harder and harder to
write. User interfaces in particular are becoming increasingly sophis-
ticated. Twenty years ago, the average application would have a glass
teletype interface (if it had an interface at all). Asynchronous terminals
would typically provide a character interactive display, while pollable
devices (such as the ubiquitous IBM 3270) would let you fill in an entire
screen before hitting SEND . Now, users expect graphical user interfaces,
with context-sensitive help, cut and paste, drag and drop, OLE integra-
tion, and MDI or SDI. Users are looking for Web-browser integration and
thin-client support.

All the time the applications themselves are getting more complex. Most
developments now use a multitier model, possibly with some middle-
ware layer or a transaction monitor. These programs are expected to be
dynamic and flexible, and to interoperate with applications written by
third parties.

Oh, and did we mention that we needed it all next week?

Developers are struggling to keep up. If we were using the same kind
of tools that produced the basic dumb-terminal applications 20 years
ago, we’d never get anything done.

So the tool makers and infrastructure vendors have come up with a

magic bullet, the wizard. Wizards are great. Do you need an MDI appli-
cation with OLE container support? Just click a single button, answer a
couple of simple questions, and the wizard will automatically generate
skeleton code for you. The Microsoft Visual C++ environment creates
over 1,200 lines of code for this scenario, automatically. Wizards are
hard at work in other contexts, too. You can use wizards to create server
components, implement Java beans, and handle network interfaces—
all complex areas where it’s nice to have expert help.

But using a wizard designed by a guru does not automatically make Joe
developer equally expert. Joe can feel pretty good—he’s just produced
a mass of code and a pretty spiffy-looking program. He just adds in the
specific application functionality and it’s ready to ship. But unless Joe
actually understands the code that has been produced on his behalf,
he’s fooling himself. He’s programming by coincidence. Wizards are a
one-way street—they cut the code for you, and then move on. If the

Prepared exclusively for Zach

 E VIL W IZARDS 199

code they produce isn’t quite right, or if circumstances change and you
need to adapt the code, you’re on your own.

We are not against wizards. On the contrary, we dedicate an entire

section (Code Generators, page 102) to writing your own. But if you do
use a wizard, and you don’t understand all the code that it produces,
you won’t be in control of your own application. You won’t be able to
maintain it, and you’ll be struggling when it comes time to debug.

TIP 50

Don’t Use Wizard Code You Don’t Understand

Some people feel that this is an extreme position. They say that develop-
ers routinely rely on things they don’t fully understand—the quantum
mechanics of integrated circuits, the interrupt structure of the proces-
sor, the algorithms used to schedule processes, the code in the supplied
libraries, and so on. We agree. And we’d feel the same about wizards
if they were simply a set of library calls or standard operating system
services that developers could rely on. But they’re not. Wizards gener-
ate code that becomes an integral part of Joe’s application. The wizard
code is not factored out behind a tidy interface—it is interwoven line by
line with functionality that Joe writes.4 Eventually, it stops being the
wizard’s code and starts being Joe’s. And no one should be producing
code they don’t fully understand.

Related sections include:

Orthogonality, page 34
Code Generators, page 102

Challenges
If you have a GUI-building wizard available, use it to generate a skeleton
application. Go through every line of code it produces. Do you understand
it all? Could you have produced it yourself? Would you have produced it
yourself, or is it doing things you don’t need?

4. However, there are other techniques that help manage complexity. We discuss two,
beans and AOP, in Orthogonality, page 34.

Prepared exclusively for Zach

 This page intentionally left blank

Prepared exclusively for Zach

 Chapter 7

Before the Project

Do you ever get the feeling that your project is doomed, even before it
starts? Sometimes it might be, unless you establish some basic ground
rules first. Otherwise, you might as well suggest that it be shut down
now, and save the sponsor some money.

At the very beginning of a project, you’ll need to determine the require-

ments. Simply listening to users is not enough: read The Requirements
Pit to find out more.

Conventional wisdom and constraint management are the topics of

Solving Impossible Puzzles. Whether you are performing requirements,
analysis, coding, or testing, difficult problems will crop up. Most of the
time, they won’t be as difficult as they first appear to be.

When you think you’ve got the problems solved, you may still not feel
comfortable with jumping in and starting. Is it simple procrastination,
or is it something more? Not Until You’re Ready offers advice on when it
may be prudent to listen to that cautionary voice inside your head.

Starting too soon is one problem, but waiting too long may be even
worse. In The Specification Trap, we’ll discuss the advantages of speci-
fication by example.

Finally, we’ll look at some of the pitfalls of formal development pro-

cesses and methodologies in Circles and Arrows. No matter how well
thought out it is, and regardless of which “best practices” it includes,
no method can replace thinking.

With these critical issues sorted out before the project gets under way,
you can be better positioned to avoid “analysis paralysis” and actually
begin your successful project.

201

Prepared exclusively for Zach

 202 C HAPTER 7 B EFORE THE P ROJECT

36 The Requirements Pit

Perfection is achieved, not when there is nothing left to add, but when there is
nothing left to take away. . . .
Antoine de St. Exupery, Wind, Sand, and Stars, 1939

Many books and tutorials refer to requirements gathering as an early

phase of the project. The word “gathering” seems to imply a tribe of
happy analysts, foraging for nuggets of wisdom that are lying on the
ground all around them while the Pastoral Symphony plays gently in
the background. “Gathering” implies that the requirements are already
there—you need merely find them, place them in your basket, and be
merrily on your way.

It doesn’t quite work that way. Requirements rarely lie on the surface.
Normally, they’re buried deep beneath layers of assumptions, miscon-
ceptions, and politics.

TIP 51

Don’t Gather Requirements—Dig for Them

Digging for Requirements

How can you recognize a true requirement while you’re digging through
all the surrounding dirt? The answer is both simple and complex.

The simple answer is that a requirement is a statement of something

that needs to be accomplished. Good requirements might include the
following:

An employee record may be viewed only by a nominated group of

people.

The cylinder-head temperature must not exceed the critical value,

which varies by engine.

The editor will highlight keywords, which will be selected depending

on the type of file being edited.

However, very few requirements are as clear-cut, and that’s what makes
requirements analysis complex.

Prepared exclusively for Zach

 T HE R EQUIREMENTS P IT 203

The first statement in the list above may have been stated by the users
as “Only an employee’s supervisors and the personnel department may
view that employee’s records.” Is this statement truly a requirement?
Perhaps today, but it embeds business policy in an absolute statement.
Policies change regularly, so we probably don’t want to hardwire them
into our requirements. Our recommendation is to document these poli-
cies separately from the requirement, and hyperlink the two. Make the
requirement the general statement, and give the developers the policy
information as an example of the type of thing they’ll need to support
in the implementation. Eventually, policy may end up as metadata in
the application.

This is a relatively subtle distinction, but it’s one that will have pro-
found implications for the developers. If the requirement is stated as
“Only personnel can view an employee record,” the developer may end
up coding an explicit test every time the application accesses these
files. However, if the statement is “Only authorized users may access
an employee record,” the developer will probably design and implement
some kind of access control system. When policy changes (and it will),
only the metadata for that system will need to be updated. In fact, gath-
ering requirements in this way naturally leads you to a system that is
well factored to support metadata.

The distinctions among requirements, policy, and implementation can

get very blurred when user interfaces are discussed. “The system must
let you choose a loan term” is a statement of requirement. “We need
a list box to select the loan term” may or may not be. If the users
absolutely must have a list box, then it is a requirement. If instead they
are describing the ability to choose, but are using listbox as an example,
then it may not be. The box on page 205 discusses a project that went
horribly wrong because the users’ interface needs were ignored.

It’s important to discover the underlying reason why users do a par-

ticular thing, rather than just the way they currently do it. At the end
of the day, your development has to solve their business problem, not
just meet their stated requirements. Documenting the reasons behind
requirements will give your team invaluable information when making
daily implementation decisions.

There’s a simple technique for getting inside your users’ requirements

that isn’t used often enough: become a user. Are you writing a system

Prepared exclusively for Zach

 204 C HAPTER 7 B EFORE THE P ROJECT

for the help desk? Spend a couple of days monitoring the phones with
an experienced support person. Are you automating a manual stock
control system? Work in the warehouse for a week.1 As well as giving
you insight into how the system will really be used, you’d be amazed at
how the request “May I sit in for a week while you do your job?” helps
build trust and establishes a basis for communication with your users.
Just remember not to get in the way!

TIP 52

Work with a User to Think Like a User

The requirements mining process is also the time to start to build a rap-
port with your user base, learning their expectations and hopes for the
system you are building. See Great Expectations, page 255, for more.

Documenting Requirements
So you are sitting down with the users and prying genuine require-
ments from them. You come across a few likely scenarios that describe
what the application needs to do. Ever the professional, you want to
write these down and publish a document that everyone can use as a
basis for discussions—the developers, the end users, and the project
sponsors.

That’s a pretty wide audience.

Ivar Jacobson [Jac94] proposed the concept of use cases to capture

requirements. They let you describe a particular use of the system—
not in terms of user interface, but in a more abstract fashion. Unfortu-
nately, Jacobson’s book was a little vague on details, so there are now
many different opinions on what a use case should be. Is it formal or
informal, simple prose or a structured document (like a form)? What
level of detail is appropriate (remember we have a wide audience)?

1. Does a week sound like a long time? It really isn’t, particularly when you’re looking
at processes in which management and workers occupy different worlds. Management
will give you one view of how things operate, but when you get down on the floor, you’ll
find a very different reality—one that will take time to assimilate.

Prepared exclusively for Zach

 T HE R EQUIREMENTS P IT 205

Sometimes the Interface Is the System

In an article in Wired magazine (January 1999, page 176), pro-
ducer and musician Brian Eno described an incredible piece of
technology—the ultimate mixing board. It does anything to sound that
can be done. And yet, instead of letting musicians make better music,
or produce a recording faster or less expensively, it gets in the way; it
disrupts the creative process.
To see why, you have to look at how recording engineers work. They
balance sounds intuitively. Over the years, they develop an innate
feedback loop between their ears and their fingertips—sliding faders,
rotating knobs, and so on. However, the interface to the new mixer
didn’t leverage off those abilities. Instead, it forced its users to type
on a keyboard or click a mouse. The functions it provided were com-
prehensive, but they were packaged in unfamiliar and exotic ways.
The functions the engineers needed were sometimes hidden behind
obscure names, or were achieved with nonintuitive combinations of
basic facilities.
That environment has a requirement to leverage existing skill sets.
While slavishly duplicating what already exists doesn’t allow for
progress, we must be able to provide a transition to the future.
For example, the recording engineers may have been better served
by some sort of touchscreen interface—still tactile, still mounted as a
traditional mixing board might be, yet allowing the software to go be-
yond the realm of fixed knobs and switches. Providing a comfortable
transition through familiar metaphors is one way to help get buy-in.
This example also illustrates our belief that successful tools adapt to
the hands that use them. In this case, it is the tools that you build for
others that must be adaptable.

One way of looking at use cases is to emphasize their goal-driven

nature. Alistair Cockburn has a paper that describes this approach,
as well as templates that can be used (strictly or not) as a starting
place ([Coc97a], also online at [URL 46]). Figure 7.1 on the following
page shows an abbreviated example of his template, while Figure 7.2
shows his sample use case.

By using a formal template as an aide-mémoire, you can be sure that

you include all the information you need in a use case: performance

Prepared exclusively for Zach

 206 C HAPTER 7 B EFORE THE P ROJECT

Figure 7.1. Cockburn’s use case template

A. CHARACTERISTIC INFORMATION
– Goal in context
– Scope
– Level
– Preconditions
– Success end condition
– Failed end condition
– Primary actor
– Trigger
B. MAIN SUCCESS SCENARIO
C. EXTENSIONS
D. VARIATIONS
E. RELATED INFORMATION
– Priority
– Performance target
– Frequency
– Superordinate use case
– Subordinate use cases
– Channel to primary actor
– Secondary actors
– Channel to secondary actors
F. SCHEDULE
G. OPEN ISSUES

characteristics, other involved parties, priority, frequency, and various

errors and exceptions that can crop up (“nonfunctional requirements”).
This is also a great place to record user comments such as “oh, except
if we get a xxx condition, then we have to do yyy instead.” The template
also serves as a ready-made agenda for meetings with your users.

This sort of organization supports the hierarchical structuring of use

cases—nesting more detailed use cases inside higher-level ones. For
example, post debit and post credit both elaborate on post transaction.

Use Case Diagrams

Workflow can be captured with UML activity diagrams, and conceptual-
level class diagrams can sometimes be useful for modeling the business

Prepared exclusively for Zach

 T HE R EQUIREMENTS P IT 207

Figure 7.2. A sample use case

USE CASE 5: BUY GOODS

A. CHARACTERISTIC INFORMATION
Goal in context: Buyer issues request directly to our company, expects
goods shipped and to be billed.
Scope: Company
Level: Summary
Preconditions: We know buyer, their address, etc.
Success end condition: Buyer has goods, we have money for the goods.
Failed end condition: We have not sent the goods, buyer has not sent
the money.
Primary actor: Buyer, any agent (or computer) acting for the customer
Trigger: Purchase request comes in.
B. MAIN SUCCESS SCENARIO
1. Buyer calls in with a purchase request.
2. Company captures buyer’s name, address, requested goods, etc.
3. Company gives buyer information on goods, prices, delivery dates, etc.
4. Buyer signs for order.
5. Company creates order, ships order to buyer.
6. Company ships invoice to buyer.
7. Buyer pays invoice.
C. EXTENSIONS
3a. Company is out of one of the ordered items: Renegotiate order.
4a. Buyer pays directly with credit card: Take payment by credit card (use
case 44).
7a. Buyer returns goods: Handle returned goods (use case 105).
D. VARIATIONS
1. Buyer may use phone in, fax in, Web order form, electronic interchange.
7. Buyer may pay by cash, money order, check, or credit card.
E. RELATED INFORMATION
Priority: Top
Performance target: 5 minutes for order, 45 days until paid
Frequency: 200/day
Superordinate use case: Manage customer relationship (use case 2).
Subordinate use cases: Create order (15). Take payment by credit card
(44). Handle returned goods (105).
Channel to primary actor: May be phone, file, or interactive
Secondary actors: Credit card company, bank, shipping service
F. SCHEDULE
Due date: Release 1.0
G. OPEN ISSUES
What happens if we have part of the order?
What happens if credit card is stolen?

Prepared exclusively for Zach

 208 C HAPTER 7 B EFORE THE P ROJECT

Figure 7.3. UML use cases—so simple a child could do it!

Go
home

at hand. But true use cases are textual descriptions, with a hierarchy
and cross-links. Use cases can contain hyperlinks to other use cases,
and they can be nested within each other.

It seems incredible to us that anyone would seriously consider docu-

menting information this dense using only simplistic stick people such
as Figure 7.3. Don’t be a slave to any notation; use whatever method
best communicates the requirements with your audience.

Overspecifying
A big danger in producing a requirements document is being too spe-
cific. Good requirements documents remain abstract. Where require-
ments are concerned, the simplest statement that accurately reflects
the business need is best. This doesn’t mean you can be vague—you
must capture the underlying semantic invariants as requirements, and
document the specific or current work practices as policy.

Requirements are not architecture. Requirements are not design, nor

are they the user interface. Requirements are need.

Seeing Further
The Year 2000 problem is often blamed on short-sighted programmers,
desperate to save a few bytes in the days when mainframes had less
memory than a modern TV remote control.

But it wasn’t the programmers’ doing, and it wasn’t really a memory

usage issue. If anything, it was the system analysts’ and designers’
fault. The Y2K problem came about from two main causes: a failure
to see beyond current business practice, and a violation of the DRY
principle.

Prepared exclusively for Zach

 T HE R EQUIREMENTS P IT 209

Businesses were using the two-digit shortcut long before computers

came on the scene. It was common practice. The earliest data process-
ing applications merely automated existing business processes, and
simply repeated the mistake. Even if the architecture required two-digit
years for data input, reporting, and storage, there should have been an
abstraction of a DATE that “knew” the two digits were an abbreviated
form of the real date.

TIP 53

Abstractions Live Longer than Details

Does “seeing further” require you to predict the future? No. It means
generating statements such as
The system makes active use of an abstraction of DATEs. The system
will implement DATE services, such as formatting, storage, and math
operations, consistently and universally.

The requirements will specify only that dates are used. It may hint
that some math may be done on dates. It may tell you that dates will
be stored on various forms of secondary storage. These are genuine
requirements for a DATE module or class.

Just One More Wafer-Thin Mint. . .

Many projects failures are blamed on an increase in scope—also known
as feature bloat, creeping featurism, or requirements creep. This is an
aspect of the boiled-frog syndrome from Stone Soup and Boiled Frogs,
page 7. What can we do to prevent requirements from creeping up on
us?

In the literature, you will find descriptions of many metrics, such as

bugs reported and fixed, defect density, cohesion, coupling, function
points, lines of code, and so on. These metrics may be tracked by hand
or with software.

Unfortunately, not many projects seem to track requirements actively.

This means that they have no way to report on changes of scope—
who requested a feature, who approved it, total number of requests
approved, and so on.

Prepared exclusively for Zach

 210 C HAPTER 7 B EFORE THE P ROJECT

The key to managing growth of requirements is to point out each new

feature’s impact on the schedule to the project sponsors. When the
project is a year late from initial estimates and accusations start flying,
it can be helpful to have an accurate, complete picture of how, and
when, requirements growth occurred.

It’s easy to get sucked into the “just one more feature” maelstrom, but
by tracking requirements you can get a clearer picture that “just one
more feature” is really the fifteenth new feature added this month.

Maintain a Glossary
As soon as you start discussing requirements, users and domain ex-
perts will use certain terms that have specific meaning to them. They
may differentiate between a “client” and a “customer,” for example. It
would then be inappropriate to use either word casually in the system.

Create and maintain a project glossary—one place that defines all the
specific terms and vocabulary used in a project. All participants in the
project, from end users to support staff, should use the glossary to
ensure consistency. This implies that the glossary needs to be widely
accessible—a good argument for Web-based documentation (more on
that in a moment).

TIP 54

Use a Project Glossary

It’s very hard to succeed on a project where the users and develop-
ers refer to the same thing by different names or, even worse, refer to
different things by the same name.

Get the Word Out

In It’s All Writing, page 248, we discuss publishing of project documents
to internal Web sites for easy access by all participants. This method of
distribution is particularly useful for requirements documents.

By presenting requirements as a hypertext document, we can better

address the needs of a diverse audience—we can give each reader what

Prepared exclusively for Zach

 T HE R EQUIREMENTS P IT 211

they want. Project sponsors can cruise along at a high level of abstrac-
tion to ensure that business objectives are met. Programmers can use
hyperlinks to “drill down” to increasing levels of detail (even referencing
appropriate definitions or engineering specifications).

Web-based distribution also avoids the typical two-inch-thick binder

entitled Requirements Analysis that no one ever reads and that becomes
outdated the instant ink hits paper.

If it’s on the Web, the programmers may even read it.

Related sections include:

Stone Soup and Boiled Frogs, page 7
Good-Enough Software, page 9
Circles and Arrows, page 220
It’s All Writing, page 248
Great Expectations, page 255

Challenges
Can you use the software you are writing? Is it possible to have a good feel
for requirements without being able to use the software yourself?

Pick a non-computer-related problem you currently need to solve. Gener-

ate requirements for a noncomputer solution.

Exercises
42. Which of the following are probably genuine requirements? Restate those Answer
on p. 307
that are not to make them more useful (if possible).

1. The response time must be less than 500 ms.

2. Dialog boxes will have a gray background.

3. The application will be organized as a number of front-end processes

and a back-end server.

4. If a user enters non-numeric characters in a numeric field, the system

will beep and not accept them.

5. The application code and data must fit within 256kB.

Prepared exclusively for Zach

 212 C HAPTER 7 B EFORE THE P ROJECT

37 Solving Impossible Puzzles

Gordius, the King of Phrygia, once tied a knot that no one could untie. It was said
that he who solved the riddle of the Gordian Knot would rule all of Asia. So along
comes Alexander the Great, who chops the knot to bits with his sword. Just a
little different interpretation of the requirements, that’s all and he did end up
ruling most of Asia.

Every now and again, you will find yourself embroiled in the middle of a
project when a really tough puzzle comes up: some piece of engineering
that you just can’t get a handle on, or perhaps some bit of code that
is turning out to be much harder to write than you thought. Maybe it
looks impossible. But is it really as hard as it seems?

Consider real-world puzzles—those devious little bits of wood, wrought

iron, or plastic that seem to turn up as Christmas presents or at garage
sales. All you have to do is remove the ring, or fit the T-shaped pieces
in the box, or whatever.

So you pull on the ring, or try to put the T’s in the box, and quickly
discover that the obvious solutions just don’t work. The puzzle can’t be
solved that way. But even though it’s obvious, that doesn’t stop people
from trying the same thing—over and over—thinking there must be a
way.

Of course, there isn’t. The solution lies elsewhere. The secret to solving
the puzzle is to identify the real (not imagined) constraints, and find
a solution therein. Some constraints are absolute; others are merely
preconceived notions. Absolute constraints must be honored, however
distasteful or stupid they may appear to be. On the other hand, some
apparent constraints may not be real constraints at all. For example,
there’s that old bar trick where you take a brand new, unopened cham-
pagne bottle and bet that you can drink beer out of it. The trick is to
turn the bottle upside down, and pour a small quantity of beer in the
hollow in the bottom of the bottle. Many software problems can be just
as sneaky.

Degrees of Freedom
The popular buzz-phrase “thinking outside the box” encourages us to
recognize constraints that might not be applicable and to ignore them.

Prepared exclusively for Zach

 S OLVING I MPOSSIBLE P UZZLES 213

But this phrase isn’t entirely accurate. If the “box” is the boundary of
constraints and conditions, then the trick is to find the box, which may
be considerably larger than you think.

The key to solving puzzles is both to recognize the constraints placed on

you and to recognize the degrees of freedom you do have, for in those
you’ll find your solution. This is why some puzzles are so effective; you
may dismiss potential solutions too readily.

For example, can you connect all of the dots in the following puzzle
and return to the starting point with just three straight lines—without
lifting your pen from the paper or retracing your steps [Hol78]?

You must challenge any preconceived notions and evaluate whether or

not they are real, hard-and-fast constraints.

It’s not whether you think inside the box or outside the box. The prob-
lem lies in finding the box—identifying the real constraints.

TIP 55

Don’t Think Outside the Box—Find the Box

When faced with an intractable problem, enumerate all the possible

avenues you have before you. Don’t dismiss anything, no matter how
unusable or stupid it sounds. Now go through the list and explain why
a certain path cannot be taken. Are you sure? Can you prove it?

Consider the Trojan horse—a novel solution to an intractable problem.

How do you get troops into a walled city without being discovered? You
can bet that “through the front door” was initially dismissed as suicide.

Categorize and prioritize your constraints. When woodworkers begin a

project, they cut the longest pieces first, then cut the smaller pieces
out of the remaining wood. In the same manner, we want to identify
the most restrictive constraints first, and fit the remaining constraints
within them.

By the way, a solution to the Four Posts puzzle is shown on page 307.

Prepared exclusively for Zach

 214 C HAPTER 7 B EFORE THE P ROJECT

There Must Be an Easier Way!

Sometimes you will find yourself working on a problem that seems
much harder than you thought it should be. Maybe it feels like you’re
going down the wrong path—that there must be an easier way than
this! Perhaps you are running late on the schedule now, or even despair
of ever getting the system to work because this particular problem is
“impossible.”

That’s when you step back a pace and ask yourself these questions:

Is there an easier way?

Are you trying to solve the right problem, or have you been dis-
tracted by a peripheral technicality?

Why is this thing a problem?

What is it that’s making it so hard to solve?

Does it have to be done this way?

Does it have to be done at all?

Many times a surprising revelation will come to you as you try to answer
one of these questions. Many times a reinterpretation of the require-
ments can make a whole set of problems go away—just like the Gordian
knot.

All you need are the real constraints, the misleading constraints, and
the wisdom to know the difference.

Challenges
Take a hard look at whatever difficult problem you are embroiled in today.
Can you cut the Gordian knot? Ask yourself the key questions we outlined
above, especially “Does it have to be done this way?”

Were you handed a set of constraints when you signed on to your current
project? Are they all still applicable, and is the interpretation of them still
valid?

Prepared exclusively for Zach

 N OT U NTIL Y OU ’ RE R EADY 215

38 Not Until You’re Ready

He who hesitates is sometimes saved.
James Thurber, The Glass in the Field

Great performers share a trait: they know when to start and when
to wait. The diver stands on the high-board, waiting for the perfect
moment to jump. The conductor stands before the orchestra, arms
raised, until she senses that the moment is right to start the piece.

You are a great performer. You too need to listen to the voice that whis-
pers “wait.” If you sit down to start typing and there’s some nagging
doubt in your mind, heed it.

TIP 56

Listen to Nagging Doubts—Start When You’re Ready

There used to be a style of tennis coaching called “inner tennis.” You’d

spend hours hitting balls over the net, not particularly trying for accu-
racy, but instead verbalizing just where the ball hit relative to some
target (often a chair). The idea was that the feedback would train your
subconscious and reflexes, so that you improved without consciously
knowing how or why.

As a developer, you’ve been doing the same kind of thing during your
entire career. You’ve been trying things and seeing which worked and
which didn’t. You’ve been accumulating experience and wisdom. When
you feel a nagging doubt, or experience some reluctance when faced
with a task, heed it. You may not be able to put your finger on exactly
what’s wrong, but give it time and your doubts will probably crystal-
lize into something more solid, something you can address. Software
development is still not a science. Let your instincts contribute to your
performance.

Good Judgment or Procrastination?

Everyone fears the blank sheet of paper. Starting a new project (or even
a new module in an existing project) can be an unnerving experience.
Many of us would prefer to put off making the initial commitment of

Prepared exclusively for Zach

 216 C HAPTER 7 B EFORE THE P ROJECT

starting. So how can you tell when you’re simply procrastinating, rather
than responsibly waiting for all the pieces to fall into place?

A technique that has worked for us in these circumstances is to start

prototyping. Choose an area that you feel will be difficult and begin
producing some kind of proof of concept. One of two things will typi-
cally happen. Shortly after starting, you may feel that you’re wasting
your time. This boredom is probably a good indication that your initial
reluctance was just a desire to put off the commitment to start. Give
up on the prototype, and hack into the real development.

On the other hand, as the prototype progresses you may have one of
those moments of revelation when you suddenly realize that some basic
premise was wrong. Not only that, but you’ll see clearly how you can put
it right. You’ll feel comfortable abandoning the prototype and launching
into the project proper. Your instincts were right, and you’ve just saved
yourself and your team a considerable amount of wasted effort.

When you make the decision to prototype as a way of investigating your

unease, be sure to remember why you’re doing it. The last thing you
want is to find yourself several weeks into serious development before
remembering that you started out writing a prototype.

Somewhat cynically, starting work on a prototype might also be more

politically acceptable than simply announcing that “I don’t feel right
about starting” and firing up solitaire.

Challenges
Discuss the fear-of-starting syndrome with your colleagues. Do others ex-
perience the same thing? Do they heed it? What tricks do they use to
overcome it? Can a group help overcome an individual’s reluctance, or is
that just peer pressure?

Prepared exclusively for Zach

 T HE S PECIFICATION T RAP 217

39 The Specification Trap

The Landing Pilot is the Non-Handling Pilot until the ‘decision altitude’ call, when
the Handling Non-Landing Pilot hands the handling to the Non-Handling Land-
ing Pilot, unless the latter calls ‘go-around,’ in which case the Handling Non-
Landing Pilot continues handling and the Non-Handling Landing Pilot continues
non-handling until the next call of ‘land’ or ‘go-around’ as appropriate. In view
of recent confusions over these rules, it was deemed necessary to restate them
clearly.
British Airways memorandum, quoted in Pilot Magazine,
December 1996

Program specification is the process of taking a requirement and reduc-

ing it down to the point where a programmer’s skill can take over. It is
an act of communication, explaining and clarifying the world in such a
way as to remove major ambiguities. As well as talking to the developer
who will be performing the initial implementation, the specification is a
record for future generations of programmers who will be maintaining
and enhancing the code. The specification is also an agreement with
the user—a codification of their needs and an implicit contract that the
final system will be in line with that requirement.

Writing a specification is quite a responsibility.

The problem is that many designers find it difficult to stop. They feel
that unless every little detail is pinned down in excruciating detail they
haven’t earned their daily dollar.

This is a mistake for several reasons. First, it’s naive to assume that a
specification will ever capture every detail and nuance of a system or its
requirement. In restricted problem domains, there are formal methods
that can describe a system, but they still require the designer to explain
the meaning of the notation to the end users—there is still a human
interpretation going on to mess things up. Even without the problems
inherent in this interpretation, it is very unlikely that the average user
knows going in to a project exactly what they need. They may say they
have an understanding of the requirement, and they may sign off on
the 200-page document you produce, but you can guarantee that once
they see the running system you’ll be inundated with change requests.

Second, there is a problem with the expressive power of language it-

self. All the diagramming techniques and formal methods still rely on

Prepared exclusively for Zach

 218 C HAPTER 7 B EFORE THE P ROJECT

natural language expressions of the operations to be performed.2 And

natural language is really not up to the job. Look at the wording of any
contract: in an attempt to be precise, lawyers have to bend the language
in the most unnatural ways.

Here’s a challenge for you. Write a short description that tells someone
how to tie bows in their shoelaces. Go on, try it!

If you are anything like us, you probably gave up somewhere around
“now roll your thumb and forefinger so that the free end passes under
and inside the left lace. . . .” It is a phenomenally difficult thing to do.
And yet most of us can tie our shoes without conscious thought.

TIP 57

Some Things Are Better Done than Described

Finally, there is the straightjacket effect. A design that leaves the coder
no room for interpretation robs the programming effort of any skill and
art. Some would say this is for the best, but they’re wrong. Often, it is
only during coding that certain options become apparent. While coding,
you may think “Look at that. Because of the particular way I coded this
routine, I could add this additional functionality with almost no effort” or
“The specification says to do this, but I could achieve an almost identical
result by doing it a different way, and I could do it in half the time.”
Clearly, you shouldn’t just hack in and make the changes, but you
wouldn’t even have spotted the opportunity if you were constrained by
an overly prescriptive design.

As a Pragmatic Programmer, you should tend to view requirements

gathering, design, and implementation as different facets of the same
process—the delivery of a quality system. Distrust environments where
requirements are gathered, specifications are written, and then coding
starts, all in isolation. Instead, try to adopt a seamless approach: spec-
ification and implementation are simply different aspects of the same
process—an attempt to capture and codify a requirement. Each should

2. There are some formal techniques that attempt to express operations algebraically,
but these techniques are rarely used in practice. They still require that the analysts
explain the meaning to the end users.

Prepared exclusively for Zach

 T HE S PECIFICATION T RAP 219

flow directly into the next, with no artificial boundaries. You’ll find that
a healthy development process encourages feedback from implementa-
tion and testing into the specification process.

Just to be clear, we are not against generating specifications. Indeed, we

recognize that there are times where incredibly detailed specifications
are demanded—for contractual reasons, because of the environment
where you work, or because of the nature of the product you are de-
veloping.3 Just be aware that you reach a point of diminishing, or even
negative, returns as the specifications get more and more detailed. Also
be careful about building specifications layered on top of specifications,
without any supporting implementation or prototyping; it’s all too easy
to specify something that can’t be built.

The longer you allow specifications to be security blankets, protecting

developers from the scary world of writing code, the harder it will be to
move on to hacking out code. Don’t fall into this specification spiral: at
some point, you need to start coding! If you find your team all wrapped
up in warm, comfy specifications, break them out. Look at prototyping,
or consider a tracer bullet development.

Related sections include:

Tracer Bullets, page 48

Challenges
The shoelace example mentioned in the text is an interesting illustration
of the problems of written descriptions. Did you consider describing the
process using diagrams rather than words? Photographs? Some formal
notation from topology? Models with wire laces? How would you teach a
toddler?

Sometimes a picture is worth more than any number of words. Sometimes

it is worthless. If you find yourself overspecifying, would pictures or special
notations help? How detailed do they have to be? When is a drawing tool
better than a whiteboard?

3. Detailed specifications are clearly appropriate for life-critical systems. We feel they
should also be produced for interfaces and libraries used by others. When your entire
output is seen as a set of routine calls, you’d better make sure those calls are well
specified.

Prepared exclusively for Zach

 220 C HAPTER 7 B EFORE THE P ROJECT

40 Circles and Arrows

[photographs] with circles and arrows and a paragraph on the back of each one
explaining what each one was, to be used as evidence against us. . .
Arlo Guthrie, “Alice’s Restaurant”

From structured programming, through chief programmer teams, CASE

tools, waterfall development, the spiral model, Jackson, ER diagrams,
Booch clouds, OMT, Objectory, and Coad/Yourdon, to today’s UML,
computing has never been short of methods intended to make pro-
gramming more like engineering. Each method gathers its disciples,
and each enjoys a period of popularity. Then each is replaced by the
next. Of all of them, perhaps only the first—structured programming—
has enjoyed a long life.

Yet some developers, adrift in a sea of sinking projects, keep clinging to

the latest fad just as shipwreck victims latch onto passing driftwood.
As each new piece floats by they painfully swim over, hoping it will be
better. At the end of the day, though, it doesn’t matter how good the
flotsam is, the developers are still aimlessly adrift.

Don’t get us wrong. We like (some) formal techniques and methods. But
we believe that blindly adopting any technique without putting it into
the context of your development practices and capabilities is a recipe
for disappointment.

TIP 58

Don’t Be a Slave to Formal Methods

Formal methods have some serious shortcomings.

Most formal methods capture requirements using a combination of

diagrams and some supporting words. These pictures represent the
designers’ understanding of the requirements. However in many
cases these diagrams are meaningless to the end users, so the
designers have to interpret them. Therefore, there is no real formal
checking of the requirements by the actual user of the system—
everything is based on the designers’ explanations, just as in old-
fashioned written requirements. We see some benefit in capturing
requirements this way, but we prefer, where possible, to show the
user a prototype and let them play with it.

Prepared exclusively for Zach

 C IRCLES AND A RROWS 221

Formal methods seem to encourage specialization. One group of

people works on a data model, another looks at the architecture,
while requirements gatherers collect use cases (or their equivalent).
We’ve seen this lead to poor communication and wasted effort.
There is also a tendency to fall back into the us versus them mental-
ity of designers against coders. We prefer to understand the whole
of the system we’re working on. It may not be possible to have an
in-depth grasp of every aspect of a system, but you should know
how the components interact, where the data lives, and what the
requirements are.

We like to write adaptable, dynamic systems, using metadata to

allow us to change the character of applications at runtime. Most
current formal methods combine a static object or data model with
some kind of event- or activity-charting mechanism. We haven’t yet
come across one that allows us to illustrate the kind of dynamism
we feel systems should exhibit. In fact, most formal methods will
lead you astray, encouraging you to set up static relationships
between objects that really should be knitted together dynamically.

Do Methods Pay Off?

In a 1999 CACM article [Gla99b], Robert Glass reviews the research
into the productivity and quality improvements gained using seven
different software development technologies (4GLs, structured tech-
niques, CASE tools, formal methods, clean room methodology, process
models, and object orientation). He reports that the initial hype sur-
rounding all of these methods was overblown. Although there is an
indication that some methods have benefits, these benefits start to
manifest themselves only after a significant productivity and quality
drop while the technique is adopted and its users train themselves.
Never underestimate the cost of adopting new tools and methods. Be
prepared to treat the first projects using these techniques as a learning
experience.

Should We Use Formal Methods?

Absolutely. But always remember that formal development methods
are just one more tool in the toolbox. If, after careful analysis, you
feel you need to use a formal method, then embrace it—but remember
who is in charge. Never become a slave to a methodology: circles and

Prepared exclusively for Zach

 222 C HAPTER 7 B EFORE THE P ROJECT

arrows make poor masters. Pragmatic Programmers look at methodolo-

gies critically, then extract the best from each and meld them into a set
of working practices that gets better each month. This is crucial. You
should work constantly to refine and improve your processes. Never
accept the rigid confines of a methodology as the limits of your world.

Don’t give in to the false authority of a method. People may walk into
meetings with an acre of class diagrams and 150 use cases, but all
that paper is still just their fallible interpretation of requirements and
design. Try not to think about how much a tool cost when you look at
its output.

TIP 59

Expensive Tools Do Not Produce Better Designs

Formal methods certainly have their place in development. However, if

you come across a project where the philosophy is “the class diagram is
the application, the rest is mechanical coding,” you know you’re looking
at a waterlogged project team and a long paddle home.

Related sections include:

The Requirements Pit, page 202

Challenges
Use case diagrams are part of the UML process for gathering requirements
(see The Requirements Pit, page 202). Are they an effective way of commu-
nicating with your users? If not, why are you using them?

How can you tell if a formal method is bringing your team benefits? What
can you measure? What constitutes an improvement? Can you distinguish
between benefits of the tool and increased experience on the part of team
members?

Where is the break-even point for introducing new methods to your team?
How do you evaluate the trade-off between future benefits and current
losses of productivity as the tool is introduced?

Are tools that work for large projects good for small ones? How about the
other way around?

Prepared exclusively for Zach

 Chapter 8

Pragmatic Projects
As your project gets under way, we need to move away from issues
of individual philosophy and coding to talk about larger, project-sized
issues. We aren’t going to go into specifics of project management, but
we will talk about a handful of critical areas that can make or break
any project.

As soon as you have more than one person working on a project, you
need to establish some ground rules and delegate parts of the project
accordingly. In Pragmatic Teams, we’ll show how to do this while hon-
oring the pragmatic philosophy.

The single most important factor in making project-level activities work

consistently and reliably is to automate your procedures. We’ll explain
why, and show some real-life examples in Ubiquitous Automation.

Earlier, we talked about testing as you code. In Ruthless Testing, we go

to the next step of project-wide testing philosophy and tools—especially
if you don’t have a large QA staff at your beck and call.

The only thing that developers dislike more than testing is documenta-
tion. Whether you have technical writers helping you or are doing it on
your own, we’ll show you how to make the chore less painful and more
productive in It’s All Writing.

Success is in the eye of the beholder—the sponsor of the project. The

perception of success is what counts, and in Great Expectations we’ll
show you some tricks to delight every project’s sponsor.

223

Prepared exclusively for Zach

 224 C HAPTER 8 P RAGMATIC P ROJECTS

The last tip in the book is a direct consequence of all the rest. In Pride
and Prejudice, we encourage you to sign your work, and to take pride
in what you do.

41 Pragmatic Teams
At Group L, Stoffel oversees six first-rate programmers, a managerial challenge
roughly comparable to herding cats.
The Washington Post Magazine, June 9, 1985

So far in this book we’ve looked at pragmatic techniques that help an

individual be a better programmer. Can these methods work for teams
as well?

The answer is a resounding “yes!” There are advantages to being a prag-

matic individual, but these advantages are multiplied manyfold if the
individual is working on a pragmatic team.

In this section we’ll look briefly at how pragmatic techniques can be

applied to teams as a whole. These notes are only a start. Once you’ve
got a group of pragmatic developers working in an enabling environ-
ment, they’ll quickly develop and refine their own team dynamics that
work for them.

Let’s recast some of the previous sections in terms of teams.

No Broken Windows
Quality is a team issue. The most diligent developer placed on a team
that just doesn’t care will find it difficult to maintain the enthusiasm
needed to fix niggling problems. The problem is further exacerbated
if the team actively discourages the developer from spending time on
these fixes.

Teams as a whole should not tolerate broken windows—those small

imperfections that no one fixes. The team must take responsibility for
the quality of the product, supporting developers who understand the
no broken windows philosophy we describe in Software Entropy, page
4, and encouraging those who haven’t yet discovered it.

Prepared exclusively for Zach

 P RAGMATIC T EAMS 225

Some team methodologies have a quality officer—someone to whom the

team delegates the responsibility for the quality of the deliverable. This
is clearly ridiculous: quality can come only from the individual contri-
butions of all team members.

Boiled Frogs
Remember the poor frog in the pan of water, back in Stone Soup and
Boiled Frogs, page 7? It doesn’t notice the gradual change in its en-
vironment, and ends up cooked. The same can happen to individuals
who aren’t vigilant. It can be difficult to keep an eye on your overall
environment in the heat of project development.

It’s even easier for teams as a whole to get boiled. People assume that
someone else is handling an issue, or that the team leader must have
OK’d a change that your user is requesting. Even the best-intentioned
teams can be oblivious to significant changes in their projects.

Fight this. Make sure everyone actively monitors the environment for
changes. Maybe appoint a chief water tester. Have this person check
constantly for increased scope, decreased time scales, additional fea-
tures, new environments—anything that wasn’t in the original agree-
ment. Keep metrics on new requirements (see page 209). The team
needn’t reject changes out of hand—you simply need to be aware that
they’re happening. Otherwise, it’ll be you in the hot water.

Communicate
It’s obvious that developers in a team must talk to each other. We gave
some suggestions to facilitate this in Communicate! on page 18. How-
ever, it’s easy to forget that the team itself has a presence within the
organization. The team as an entity needs to communicate clearly with
the rest of the world.

To outsiders, the worst project teams are those that appear sullen and
reticent. They hold meetings with no structure, where no one wants to
talk. Their documents are a mess: no two look the same, and each uses
different terminology.

Great project teams have a distinct personality. People look forward to

meetings with them, because they know that they’ll see a well-prepared

Prepared exclusively for Zach

 226 C HAPTER 8 P RAGMATIC P ROJECTS

performance that makes everyone feel good. The documentation they

produce is crisp, accurate, and consistent. The team speaks with one
voice.1 They may even have a sense of humor.

There is a simple marketing trick that helps teams communicate as

one: generate a brand. When you start a project, come up with a name
for it, ideally something off-the-wall. (In the past, we’ve named projects
after things such as killer parrots that prey on sheep, optical illusions,
and mythical cities.) Spend 30 minutes coming up with a zany logo,
and use it on your memos and reports. Use your team’s name liberally
when talking with people. It sounds silly, but it gives your team an
identity to build on, and the world something memorable to associate
with your work.

Don’t Repeat Yourself

In The Evils of Duplication, page 26, we talked about the difficulties of
eliminating duplicated work between members of a team. This duplica-
tion leads to wasted effort, and can result in a maintenance nightmare.
Clearly good communication can help here, but sometimes something
extra is needed.

Some teams appoint a member as the project librarian, responsible for

coordinating documentation and code repositories. Other team mem-
bers can use this person as the first port of call when they’re look-
ing for something. A good librarian will also be able to spot impending
duplication by reading the material that they’re handling.

When the project’s too big for one librarian (or when no one wants
to play the role), appoint people as focal points for various functional
aspects of the work. If people want to talk over date handling, they
should know to talk with Mary. If there’s a database schema issue, see
Fred.

And don’t forget the value of groupware systems and local Usenet news-
groups for communicating and archiving questions and answers.

1. The team speaks with one voice—externally. Internally, we strongly encourage lively,
robust debate. Good developers tend to be passionate about their work.

Prepared exclusively for Zach

 P RAGMATIC T EAMS 227

Orthogonality
Traditional team organization is based on the old-fashioned waterfall
method of software construction. Individuals are assigned roles based
on their job function. You’ll find business analysts, architects, design-
ers, programmers, testers, documenters, and the like.2 There is an
implicit hierarchy here—the closer to the user you’re allowed, the more
senior you are.

Taking things to the extreme, some development cultures dictate strict

divisions of responsibility; coders aren’t allowed to talk to testers, who
in turn aren’t allowed to talk to the chief architect, and so on. Some
organizations then compound the problem by having different sub-
teams report through separate management chains.

It is a mistake to think that the activities of a project—analysis, design,

coding, and testing—can happen in isolation. They can’t. These are
different views of the same problem, and artificially separating them
can cause a boatload of trouble. Programmers who are two or three
levels removed from the actual users of their code are unlikely to be
aware of the context in which their work is used. They will not be able
to make informed decisions.

TIP 60

Organize Around Functionality, Not Job Functions

We favor splitting teams functionally. Divide your people into small

teams, each responsible for a particular functional aspect of the final
system. Let the teams organize themselves internally, building on indi-
vidual strengths as they can. Each team has responsibilities to others
in the project, as defined by their agreed-upon commitments. The exact
set of commitments changes with each project, as does the allocation
of people into teams.

Functionality here does not necessarily mean end-user use cases. The
database access layer counts, as does the help subsystem. We’re look-
ing for cohesive, largely self-contained teams of people—exactly the

2. In The Rational Unified Process: An Introduction, the author identifies 27 separate

roles within a project team! [Kru98]

Prepared exclusively for Zach

 228 C HAPTER 8 P RAGMATIC P ROJECTS

same criteria we should be using when we modularize code. There are

warning signs that the team organization is wrong—a classic example
is having two subteams working on the same program module or class.

How does this functional style of organization help? Organize our re-
sources using the same techniques we use to organize code, using
techniques such as contracts (Design by Contract, page 109), decou-
pling (Decoupling and the Law of Demeter, page 138), and orthogonality
(Orthogonality, page 34), and we help isolate the team as a whole from
the effects of change. If the user suddenly decides to change database
vendors, only the database team should be affected. Should marketing
suddenly decide to use an off-the-shelf tool for the calendar function,
the calendar group takes a hit. Properly executed, this kind of group
approach can dramatically reduce the number of interactions between
individuals’ work, reducing time scales, increasing quality, and cutting
down on the number of defects. This approach can also lead to a more
committed set of developers. Each team knows that they alone are re-
sponsible for a particular function, so they feel more ownership of their
output.

However, this approach works only with responsible developers and

strong project management. Creating a pool of autonomous teams and
letting them loose without leadership is a recipe for disaster. The project
needs at least two “heads”—one technical, the other administrative.
The technical head sets the development philosophy and style, assigns
responsibilities to teams, and arbitrates the inevitable “discussions”
between people. The technical head also looks constantly at the big pic-
ture, trying to find any unnecessary commonality between teams that
could reduce the orthogonality of the overall effort. The administrative
head, or project manager, schedules the resources that the teams need,
monitors and reports on progress, and helps decide priorities in terms
of business needs. The administrative head might also act as the team’s
ambassador when communicating with the outside world.

Teams on larger projects need additional resources: a librarian who

indexes and stores code and documentation, a tool builder who pro-
vides common tools and environments, operational support, and so on.

This type of team organization is similar in spirit to the old chief pro-
grammer team concept, first documented in 1972 [Bak72].

Prepared exclusively for Zach

 P RAGMATIC T EAMS 229

Automation
A great way to ensure both consistency and accuracy is to automate
everything the team does. Why lay code out manually when your editor
can do it automatically as you type? Why complete test forms when the
overnight build can run tests automatically?

Automation is an essential component of every project team—important

enough for us to dedicate an entire section to it, starting on the follow-
ing page. To ensure that things get automated, appoint one or more
team members as tool builders to construct and deploy the tools that
automate the project drudgery. Have them produce makefiles, shell
scripts, editor templates, utility programs, and the like.

Know When to Stop Adding Paint

Remember that teams are made up of individuals. Give each member
the ability to shine in his or her own way. Give them just enough struc-
ture to support them and to ensure that the project delivers against
its requirements. Then, like the painter in Good-Enough Software, page
11, resist the temptation to add more paint.

Related sections include:

Software Entropy, page 4
Stone Soup and Boiled Frogs, page 7
Good-Enough Software, page 9
Communicate!, page 18
The Evils of Duplication, page 26
Orthogonality, page 34
Design by Contract, page 109
Decoupling and the Law of Demeter, page 138
Ubiquitous Automation, page 230

Challenges
Look around for successful teams outside the area of software develop-
ment. What makes them successful? Do they use any of the processes
discussed in this section?

Prepared exclusively for Zach

 230 C HAPTER 8 P RAGMATIC P ROJECTS

Next time you start a project, try convincing people to brand it. Give your
organization time to become used to the idea, and then do a quick audit to
see what difference it made, both within the team and externally.

Team Algebra: In school, we are given problems such as “If it takes 4 work-
ers 6 hours to dig a ditch, how long would it take 8 workers?” In real life,
however, what factors affect the answer to: “If it takes 4 programmers 6
months to develop an application, how long would it take 8 programmers?”
In how many scenarios is the time actually reduced?

42 Ubiquitous Automation
Civilization advances by extending the number of important operations we can
perform without thinking.
Alfred North Whitehead

At the dawn of the age of automobiles, the instructions for starting a

Model-T Ford were more than two pages long. With modern cars, you
just turn the key—the starting procedure is automatic and foolproof. A
person following a list of instructions might flood the engine, but the
automatic starter won’t.

Although computing is still an industry at the Model-T stage, we can’t

afford to go through two pages of instructions again and again for some
common operation. Whether it is the build and release procedure, code
review paperwork, or any other recurring task on the project, it has to
be automatic. We may have to build the starter and fuel injector from
scratch, but once it’s done, we can just turn the key from then on.

In addition, we want to ensure consistency and repeatability on the

project. Manual procedures leave consistency up to chance; repeatabil-
ity isn’t guaranteed, especially if aspects of the procedure are open to
interpretation by different people.

Prepared exclusively for Zach

 U BIQUITOUS A UTOMATION 231

All on Automatic
We were once at a client site where all the developers were using the
same IDE. Their system administrator gave each developer a set of
instructions on installing add-on packages to the IDE. These instruc-
tions filled many pages—pages full of click here, scroll there, drag this,
double-click that, and do it again.

Not surprisingly, every developer’s machine was loaded slightly differ-

ently. Subtle differences in the application’s behavior occurred when
different developers ran the same code. Bugs would appear on one
machine but not on others. Tracking down version differences of any
one component usually revealed a surprise.

TIP 61

Don’t Use Manual Procedures

People just aren’t as repeatable as computers are. Nor should we expect

them to be. A shell script or batch file will execute the same instruc-
tions, in the same order, time after time. It can be put under source
control, so you can examine changes to the procedure over time as well
(“but it used to work. . . ”).

Another favorite tool of automation is cron (or “at” on Windows NT). It

allows us to schedule unattended tasks to run periodically—usually in
the middle of the night. For example, the following crontab file speci-
fies that a project’s nightly command be run at five minutes past mid-
night every day, that the backup be run at 3:15 a.m. on weekdays, and
that expense_reports be run at midnight on the first of the month.

# MIN HOUR DAY MONTH DAYOFWEEK COMMAND

# ---------------------------------------------------------------
5 0 * * * /projects/Manhattan/bin/nightly
15 3 * * 1-5 /usr/local/bin/backup
0 0 1 * * /home/accounting/expense_reports

Using cron, we can schedule backups, the nightly build, Web site main-
tenance, and anything else that needs to be done—unattended, auto-
matically.

Prepared exclusively for Zach

 232 C HAPTER 8 P RAGMATIC P ROJECTS

Compiling the Project

Compiling the project is a chore that should be reliable and repeat-
able. We generally compile projects with makefiles, even when using an
IDE environment. There are several advantages in using makefiles. It
is a scripted, automatic procedure. We can add in hooks to generate
code for us, and run regression tests automatically. IDEs have their
advantages, but with IDEs alone it can be hard to achieve the level of
automation that we’re looking for. We want to check out, build, test,
and ship with a single command.

Generating Code
In The Evils of Duplication, page 26, we advocated generating code to
derive knowledge from common sources. We can exploit make’s depen-
dency analysis mechanism to make this process easy. It’s a pretty sim-
ple matter to add rules to a makefile to generate a file from some other
source automatically. For example, suppose we wanted to take an XML
file, generate a Java file from it, and compile the result.

.SUFFIXES: .java .class .xml

.xml.java:
perl convert.pl $< > $@
.java.class:
$(JAVAC) $(JAVAC_FLAGS) $<

Type make test.class, and make will automatically look for a file
named test.xml, build a .java file by running a Perl script, and then
compile that file to produce test.class.

We can use the same sort of rules to generate source code, header files,
or documentation automatically from some other form as well (see Code
Generators, page 102).

Regression Tests
You can also use the makefile to run regression tests for you, either for
an individual module or for an entire subsystem. You can easily test
the entire project with just one command at the top of the source tree,
or you can test an individual module by using the same command in a
single directory. See Ruthless Testing, page 237, for more on regression
testing.

Prepared exclusively for Zach

 U BIQUITOUS A UTOMATION 233

Recursive make
Many projects set up recursive, hierarchical makefiles for project
builds and testing. But be aware of some potential problems.
make calculates dependencies between the various targets it has to
build. But it can analyze only the dependencies that exist within one
single make invocation. In particular, a recursive make has no knowl-
edge of dependencies that other invocations of make may have. If
you are careful and precise, you can get the proper results, but it’s
easy to cause extra work unnecessarily—or miss a dependency and
not recompile when it’s needed.
In addition, build dependencies may not be the same as test depen-
dencies, and you may need separate hierarchies.

Build Automation
A build is a procedure that takes an empty directory (and a known com-
pilation environment) and builds the project from scratch, producing
whatever you hope to produce as a final deliverable—a CD-ROM mas-
ter image or a self-extracting archive, for instance. Typically a project
build will encompass the following steps.

1. Check out the source code from the repository.

2. Build the project from scratch, typically from a top-level makefile.

Each build is marked with some form of release or version number,
or perhaps a date stamp.

3. Create a distributable image. This procedure may entail fixing file

ownership and permissions, and producing all examples, docu-
mentation, README files, and anything else that will ship with the
product, in the exact format that will be required when you ship.3

4. Run specified tests ( make test ).

3. If you are producing a CD-ROM in ISO9660 format, for example, you would run the
program that produces a bit-for-bit image of the 9660 file system. Why wait until the
night before you ship to make sure it works?

Prepared exclusively for Zach

 234 C HAPTER 8 P RAGMATIC P ROJECTS

For most projects, this level of build is run automatically every night.
In this nightly build, you will typically run more complete tests than an
individual might run while building some specific portion of the project.
The important point is to have the full build run all available tests. You
want to know if a regression test failed because of one of today’s code
changes. By identifying the problem close to the source, you stand a
better chance of finding and fixing it.

When you don’t run tests regularly, you may discover that the appli-
cation broke due to a code change made three months ago. Good luck
finding that one.

Final Builds
Final builds, which you intend to ship as products, may have different
requirements from the regular nightly build. A final build may require
that the repository be locked, or tagged with the release number, that
optimization and debug flags be set differently, and so on. We like to
use a separate make target (such as make final) that sets all of these
parameters at once.

Remember that if the product is compiled differently from earlier ver-

sions, then you must test against this version all over again.

Automatic Administrivia
Wouldn’t it be nice if programmers could actually devote all of their time
to programming? Unfortunately, this is rarely the case. There is e-mail
to be answered, paperwork to be filled out, documents to be posted to
the Web, and so on. You may decide to create a shell script to do some
of the dirty work, but you still have to remember to run the script when
needed.

Because memory is the second thing you lose as you age,4 we don’t
want to rely on it too heavily. We can run scripts to perform proce-
dures for us automatically, based on the content of source code and
documents. Our goal is to maintain an automatic, unattended, content-
driven workflow.

4. What’s the first? I forget.

Prepared exclusively for Zach

 U BIQUITOUS A UTOMATION 235

Web Site Generation

Many development teams use an internal Web site for project commu-
nication, and we think this is a great idea. But we don’t want to spend
too much time maintaining the Web site, and we don’t want to let it get
stale or out of date. Misleading information is worse than no informa-
tion at all.

Documentation that is extracted from code, requirements analyses,

design documents, and any drawings, charts, or graphs all need to
be published to the Web on a regular basis. We like to publish these
documents automatically as part of the nightly build or as a hook into
the source code check-in procedure.

However it is done, Web content should be generated automatically

from information in the repository and published without human inter-
vention. This is really another application of the DRY principle: infor-
mation exists in one form as checked-in code and documents. The view
from the Web browser is simply that—just a view. You shouldn’t have
to maintain that view by hand.

Any information generated by the nightly build should be accessible on

the development Web site: results of the build itself (for example, the
build results might be presented as a one-page summary that includes
compiler warnings, errors, and current status), regression tests, per-
formance statistics, coding metrics and any other static analysis, and
so on.

Approval Procedures
Some projects have various administrative workflows that must be fol-
lowed. For instance, code or design reviews need to be scheduled and
followed through, approvals may need to be granted, and so on. We can
use automation—and especially the Web site—to help ease the paper-
work burden.

Suppose you wanted to automate code review scheduling and approval.

You might put a special marker in each source code file:
/* Status: needs_review */

A simple script could go through all of the source code and look for
all files that had a status of needs_review, indicating that they were
ready to be reviewed. You could then post a list of those files as a

Prepared exclusively for Zach

 236 C HAPTER 8 P RAGMATIC P ROJECTS

Web page, automatically send e-mail to the appropriate people, or even

schedule a meeting automatically using some calendar software.

You can set up a form on a Web page for the reviewers to register
approval or disapproval. After the review, the status can be automat-
ically changed to reviewed. Whether you have a code walk-through
with all the participants is up to you; you can still do the paperwork
automatically. (In an article in the April 1999 CACM, Robert Glass sum-
marizes research that seems to indicate that, while code inspection is
effective, conducting reviews in meetings is not [Gla99a].)

The Cobbler’s Children

The cobbler’s children have no shoes. Often, people who develop soft-
ware use the poorest tools to do the job.

But we have all the raw materials we need to craft better tools. We
have cron. We have make, Ant, and CruiseControl for automation (see
[Cla04]). And we have Ruby, Perl, and other high-level scripting lan-
guages for quickly developing custom tools, Web page generators, code
generators, test harnesses, and so on.

Let the computer do the repetitious, the mundane—it will do a better

job of it than we would. We’ve got more important and more difficult
things to do.

Related sections include:

The Cat Ate My Source Code, page 2
The Evils of Duplication, page 26
The Power of Plain Text, page 73
Shell Games, page 77
Debugging, page 90
Code Generators, page 102
Pragmatic Teams, page 224
Ruthless Testing, page 237
It’s All Writing, page 248

Challenges
Look at your habits throughout the workday. Do you see any repetitive
tasks? Do you type the same sequence of commands over and over again?

Prepared exclusively for Zach

 R UTHLESS T ESTING 237

Try writing a few shell scripts to automate the process. Do you always click
on the same sequence of icons repeatedly? Can you create a macro to do
all that for you?

How much of your project paperwork can be automated? Given the high
expense of programming staff,5 determine how much of the project’s bud-
get is being wasted on administrative procedures. Can you justify the
amount of time it would take to craft an automated solution based on
the overall cost savings it would achieve?

43 Ruthless Testing
Most developers hate testing. They tend to test gently, subconsciously
knowing where the code will break and avoiding the weak spots. Prag-
matic Programmers are different. We are driven to find our bugs now,
so we don’t have to endure the shame of others finding our bugs later.

Finding bugs is somewhat like fishing with a net. We use fine, small
nets (unit tests) to catch the minnows, and big, coarse nets (integration
tests) to catch the killer sharks. Sometimes the fish manage to escape,
so we patch any holes that we find, in hopes of catching more and more
slippery defects that are swimming about in our project pool.

TIP 62

Test Early. Test Often. Test Automatically.

We want to start testing as soon as we have code. Those tiny minnows

have a nasty habit of becoming giant, man-eating sharks pretty fast,
and catching a shark is quite a bit harder. But we don’t want to have
to do all that testing by hand.

5. For estimating purposes, you can figure an industry average of about US$100,000
per head—that’s salary plus benefits, training, office space and overhead, and so on.

Prepared exclusively for Zach

 238 C HAPTER 8 P RAGMATIC P ROJECTS

Many teams develop elaborate test plans for their projects. Sometimes
they will even use them. But we’ve found that teams that use auto-
mated tests have a much better chance of success. Tests that run with
every build are much more effective than test plans that sit on a shelf.

The earlier a bug is found, the cheaper it is to remedy. “Code a little,

test a little” is a popular saying in the Smalltalk world,6 and we can
adopt that mantra as our own by writing test code at the same time (or
even before) we write the production code.

In fact, a good project may well have more test code than production
code. The time it takes to produce this test code is worth the effort. It
ends up being much cheaper in the long run, and you actually stand a
chance of producing a product with close to zero defects.

Additionally, knowing that you’ve passed the test gives you a high de-
gree of confidence that a piece of code is “done.”

TIP 63

Coding Ain’t Done ’Til All the Tests Run

Just because you have finished hacking out a piece of code doesn’t
mean you can go tell your boss or your client that it’s done. It’s not.
First of all, code is never really done. More importantly, you can’t claim
that it is usable by anyone until it passes all of the available tests.

We need to look at three main aspects of project-wide testing: what to

test, how to test, and when to test.

What to Test
There are several major types of software testing that you need to per-
form:

Unit testing
Integration testing
Validation and verification

6. eXtreme Programming [URL 45] calls this concept “continuous integration, relent-
less testing.”

Prepared exclusively for Zach

 R UTHLESS T ESTING 239

Resource exhaustion, errors, and recovery

Performance testing
Usability testing

This list is by no means complete, and some specialized projects will

require various other types of testing as well. But it gives us a good
starting point.

Unit Testing
A unit test is code that exercises a module. We covered this topic by
itself in Code That’s Easy to Test, page 189. Unit testing is the founda-
tion of all the other forms of testing that we’ll discuss in this section.
If the parts don’t work by themselves, they probably won’t work well
together. All of the modules you are using must pass their own unit
tests before you can proceed.

Once all of the pertinent modules have passed their individual tests,
you’re ready for the next stage. You need to test how all the modules
use and interact with each other throughout the system.

Integration Testing
Integration testing shows that the major subsystems that make up the
project work and play well with each other. With good contracts in place
and well tested, any integration issues can be detected easily. Other-
wise, integration becomes a fertile breeding ground for bugs. In fact, it
is often the single largest source of bugs in the system.

Integration testing is really just an extension of the unit testing we’ve

described—only now you’re testing how entire subsystems honor their
contracts.

Validation and Verification

As soon as you have an executable user interface or prototype, you
need to answer an all-important question: the users told you what they
wanted, but is it what they need?

Does it meet the functional requirements of the system? This, too,

needs to be tested. A bug-free system that answers the wrong ques-
tion isn’t very useful. Be conscious of end-user access patterns and

Prepared exclusively for Zach

 240 C HAPTER 8 P RAGMATIC P ROJECTS

how they differ from developer test data (for an example, see the story
about brush strokes on page 92).

Resource Exhaustion, Errors, and Recovery

Now that you have a pretty good idea that the system will behave cor-
rectly under ideal conditions, you need to discover how it will behave
under real-world conditions. In the real world, your programs don’t
have limitless resources; they run out of things. A few limits your code
may encounter include:

Memory
Disk space
CPU bandwidth
Wall-clock time
Disk bandwidth
Network bandwidth
Color palette
Video resolution

You might actually check for disk space or memory allocation failures,
but how often do you test for the others? Will your application fit on
a screen with colors? Will it run on a screen
with -bit color without looking like a postage stamp? Will the batch
job finish before the archive starts?

You can detect environmental limitations, such as the video speci-

fications, and adapt as appropriate. Not all failures are recoverable,
however. If your code detects that memory has been exhausted, your
options are limited: you may not have enough resources left to do any-
thing except fail.

When the system does fail,7 will it fail gracefully? Will it try, as best
it can, to save its state and prevent loss of work? Or will it “GPF” or
“core-dump” in the user’s face?

7. Our copy editor wanted us to change this sentence to “If the system does fail .”
We resisted.

Prepared exclusively for Zach

 R UTHLESS T ESTING 241

Performance Testing
Performance testing, stress testing, or testing under load may be an
important aspect of the project as well.

Ask yourself if the software meets the performance requirements under

real-world conditions—with the expected number of users, or connec-
tions, or transactions per second. Is it scalable?

For some applications, you may need specialized testing hardware or

software to simulate the load realistically.

Usability Testing
Usability testing is different from the types of testing discussed so far.
It is performed with real users, under real environmental conditions.

Look at usability in terms of human factors. Were there any misun-

derstandings during requirements analysis that need to be addressed?
Does the software fit the user like an extension of the hand? (Not only
do we want our own tools to fit our hands, but we want the tools we
create for users to fit their hands as well.)

As with validation and verification, you need to perform usability test-

ing as early as you can, while there is still time to make corrections. For
larger projects, you may want to bring in human factors specialists. (If
nothing else, it’s fun to play with the one-way mirrors).

Failure to meet usability criteria is just as big a bug as dividing by zero.

How to Test
We’ve looked at what to test. Now we’ll turn our attention to how to
test, including:

Regression testing
Test data
Exercising GUI systems
Testing the tests
Testing thoroughly

Prepared exclusively for Zach

 242 C HAPTER 8 P RAGMATIC P ROJECTS

Design/Methodology Testing
Can you test the design of the code itself and the methodology you
used to build the software? After a fashion, yes you can. You do this
by analyzing metrics—measurements of various aspects of the code.
The simplest metric (and often the least interesting) is lines of code—
how big is the code itself?
There are a wide variety of other metrics you can use to examine
code, including:
McCabe Cyclomatic Complexity Metric (measures complexity of
decision structures)
Inheritance fan-in (number of base classes) and fan-out (number
of derived modules using this one as a parent)
Response set (see Decoupling and the Law of Demeter, page
138)
Class coupling ratios (see [URL 48])
Some metrics are designed to give you a “passing grade,” while oth-
ers are useful only by comparison. That is, you calculate these met-
rics for every module in the system and see how a particular module
relates to its brethren. Standard statistical techniques (such as mean
and standard deviation) are usually used here.
If you find a module whose metrics are markedly different from all
the rest, you need to ask yourself if that is appropriate. For some
modules, it may be okay to “blow the curve.” But for those that don’t
have a good excuse, it can indicate potential problems.

Regression Testing
A regression test compares the output of the current test with previous
(or known) values. We can ensure that bugs we fixed today didn’t break
things that were working yesterday. This is an important safety net,
and it cuts down on unpleasant surprises.

All of the tests we’ve mentioned so far can be run as regression tests,
ensuring that we haven’t lost any ground as we develop new code. We
can run regressions to verify performance, contracts, validity, and so
on.

Prepared exclusively for Zach

 R UTHLESS T ESTING 243

Test Data
Where do we get the data to run all these tests? There are only two
kinds of data: real-world data and synthetic data. We actually need
to use both, because the different natures of these kinds of data will
expose different bugs in our software.

Real-world data comes from some actual source. Possibly it has been
collected from an existing system, a competitor’s system, or a prototype
of some sort. It represents typical user data. The big surprises come as
you discover what typical means. This is most likely to reveal defects
and misunderstandings in requirements analysis.

Synthetic data is artificially generated, perhaps under certain statistical

constraints. You may need to use synthetic data for any of the following
reasons.

You need a lot of data, possibly more than any real-world sample
could provide. You might be able to use the real-world data as a
seed to generate a larger sample set, and tweak certain fields that
need to be unique.

You need data to stress the boundary conditions. This data may
be completely synthetic: date fields containing February 29, 1999,
huge record sizes, or addresses with foreign postal codes.

You need data that exhibits certain statistical properties. Want to

see what happens if every third transaction fails? Remember the
sort algorithm that slows to a crawl when handed presorted data?
You can present data in random or sorted order to expose this kind
of weakness.

Exercising GUI Systems

Testing GUI-intensive systems often requires specialized testing tools.
These tools may be based on a simple event capture/playback model,
or they may require specially written scripts to drive the GUI. Some
systems combine elements of both.

Less sophisticated tools enforce a high degree of coupling between the

version of software being tested and the test script itself: if you move
a dialog box or make a button smaller, the test may not be able to

Prepared exclusively for Zach

 244 C HAPTER 8 P RAGMATIC P ROJECTS

find it, and may fail. Most modern GUI testing tools use a number of
different techniques to get around this problem, and try to adjust to
minor layout differences.

However, you can’t automate everything. Andy worked on a graphics

system that allowed the user to create and display nondeterministic
visual effects which simulated various natural phenomena. Unfortu-
nately, during testing you couldn’t just grab a bitmap and compare the
output with a previous run, because it was designed to be different
every time. For situations such as this one, you may have no choice
but to rely on manual interpretation of test results.

One of the many advantages of writing decoupled code (see Decou-

pling and the Law of Demeter, page 138) is more modular testing. For
instance, for data processing applications that have a GUI front end,
your design should be decoupled enough so that you can test the ap-
plication logic without having a GUI present. This idea is similar to
testing your subcomponents first. Once the application logic has been
validated, it becomes easier to locate bugs that show up with the user
interface in place (it’s likely that the bugs were created by the user-
interface code).

Testing the Tests

Because we can’t write perfect software, it follows that we can’t write
perfect test software either. We need to test the tests.

Think of our set of test suites as an elaborate security system, designed

to sound the alarm when a bug shows up. How better to test a security
system than to try to break in?

After you have written a test to detect a particular bug, cause the bug
deliberately and make sure the test complains. This ensures that the
test will catch the bug if it happens for real.

TIP 64

Use Saboteurs to Test Your Testing

If you are really serious about testing, you might want to appoint a
project saboteur. The saboteur’s role is to take a separate copy of the

Prepared exclusively for Zach

 R UTHLESS T ESTING 245

source tree, introduce bugs on purpose, and verify that the tests will
catch them.

When writing tests, make sure that alarms sound when they should.

Testing Thoroughly
Once you are confident that your tests are correct, and are finding
bugs you create, how do you know if you have tested the code base
thoroughly enough?

The short answer is “you don’t,” and you never will. But there are prod-
ucts on the market that can help. These coverage analysis tools watch
your code during testing and keep track of which lines of code have
been executed and which haven’t. These tools help give you a general
feel for how comprehensive your testing is, but don’t expect to see 100%
coverage.

Even if you do happen to hit every line of code, that’s not the whole
picture. What is important is the number of states that your program
may have. States are not equivalent to lines of code. For instance, sup-
pose you have a function that takes two integers, each of which can be
a number from 0 to 999.
int test(int a, int b) {
return a / (a + b);
}

In theory, this three-line function has 1,000,000 logical states, 999,999

of which will work correctly and one that will not (when a + b equals
zero). Simply knowing that you executed this line of code doesn’t tell
you that—you would need to identify all possible states of the program.
Unfortunately, in general this is a really hard problem. Hard as in, “The
sun will be a cold hard lump before you can solve it.”

TIP 65

Test State Coverage, Not Code Coverage

Even with good code coverage, the data you use for testing still has a
huge impact, and, more importantly, the order in which you traverse
code may have the largest impact of all.

Prepared exclusively for Zach

 246 C HAPTER 8 P RAGMATIC P ROJECTS

When to Test
Many projects tend to leave testing to the last minute—right where it
will be cut against the sharp edge of a deadline.8 We need to start much
sooner than that. As soon as any production code exists, it needs to be
tested.

Most testing should be done automatically. It’s important to note that

by “automatically” we mean that the test results are interpreted auto-
matically as well. See Ubiquitous Automation, page 230, for more on this
subject.

We like to test as frequently as we can, and always before we check code

into the source repository. Some source code control systems, such as
Aegis, can do this automatically. Otherwise, we just type

% make test

Usually, it isn’t a problem to run regressions on all of the individual

unit tests and integration tests as often as needed.

But some tests may not be easily run on a such a frequent basis. Stress
tests, for instance, may require special setup or equipment, and some
hand holding. These tests may be run less often—weekly or monthly,
perhaps. But it is important that they be run on a regular, scheduled
basis. If it can’t be done automatically, then make sure it appears on
the schedule, with all the necessary resources allocated to the task.

Tightening the Net

Finally, we’d like to reveal the single most important concept in testing.
It is an obvious one, and virtually every textbook says to do it this way.
But for some reason, most projects still do not.

If a bug slips through the net of existing tests, you need to add a new
test to trap it next time.

8. dead line ded-lı̄n n (1864) a line drawn within or around a prison that a prisoner
passes at the risk of being shot—Webster’s Collegiate Dictionary.

Prepared exclusively for Zach

 R UTHLESS T ESTING 247

TIP 66

Find Bugs Once

Once a human tester finds a bug, it should be the last time a human
tester finds that bug. The automated tests should be modified to check
for that particular bug from then on, every time, with no exceptions, no
matter how trivial, and no matter how much the developer complains
and says, “Oh, that will never happen again.”

Because it will happen again. And we just don’t have the time to go
chasing after bugs that the automated tests could have found for us.
We have to spend our time writing new code—and new bugs.

Related sections include:

The Cat Ate My Source Code, page 2
Debugging, page 90
Decoupling and the Law of Demeter, page 138
Refactoring, page 184
Code That’s Easy to Test, page 189
Ubiquitous Automation, page 230

Challenges
Can you automatically test your project? Many teams are forced to answer
“no.” Why? Is it too hard to define the acceptable results? Won’t this make
it hard to prove to the sponsors that the project is “done”?

Is it too hard to test the application logic independent of the GUI? What
does this say about the GUI? About coupling?

Prepared exclusively for Zach

 248 C HAPTER 8 P RAGMATIC P ROJECTS

44 It’s All Writing

The palest ink is better than the best memory.
Chinese Proverb

Typically, developers don’t give much thought to documentation. At

best it is an unfortunate necessity; at worst it is treated as a low-priority
task in the hope that management will forget about it at the end of the
project.

Pragmatic Programmers embrace documentation as an integral part of

the overall development process. Writing documentation can be made
easier by not duplicating effort or wasting time, and by keeping docu-
mentation close at hand—in the code itself, if possible.

These aren’t exactly original or novel thoughts; the idea of wedding

code and documentation appears in Donald Knuth’s work on literate
programming and in Sun’s JavaDoc utility, among others. We want to
downplay the dichotomy between code and documentation, and instead
treat them as two views of the same model (see It’s Just a View, page
157). In fact, we want to go a little further and apply all of our pragmatic
principles to documentation as well as to code.

TIP 67

Treat English as Just Another Programming Language

There are basically two kinds of documentation produced for a project:

internal and external. Internal documentation includes source code
comments, design and test documents, and so on. External documen-
tation is anything shipped or published to the outside world, such as
user manuals. But regardless of the intended audience, or the role of
the writer (developer or technical writer), all documentation is a mir-
ror of the code. If there’s a discrepancy, the code is what matters—for
better or worse.

TIP 68

Build Documentation In, Don’t Bolt It On

We’ll start with internal documentation.

Prepared exclusively for Zach

 I T ’ S A LL W RITING 249

Comments in Code
Producing formatted documents from the comments and declarations
in source code is fairly straightforward, but first we have to ensure that
we actually have comments in the code. Code should have comments,
but too many comments can be just as bad as too few.

In general, comments should discuss why something is done, its pur-

pose and its goal. The code already shows how it is done, so comment-
ing on this is redundant—and is a violation of the DRY principle.

Commenting source code gives you the perfect opportunity to docu-

ment those elusive bits of a project that can’t be documented anywhere
else: engineering trade-offs, why decisions were made, what other alter-
natives were discarded, and so on.

We like to see a simple module-level header comment, comments for

significant data and type declarations, and a brief per-class and per-
method header, describing how the function is used and anything that
it does that is not obvious.

Variable names, of course, should be well chosen and meaningful. foo,

for instance, is meaningless, as is doit or manager or stuff. Hun-
garian notation (where you encode the variable’s type information in
the name itself) is utterly inappropriate in object-oriented systems.
Remember that you (and others after you) will be reading the code many
hundreds of times, but only writing it a few times. Take the time to spell
out connectionPool instead of cp.

Even worse than meaningless names are misleading names. Have you
ever had someone explain inconsistencies in legacy code such as, “The
routine called getData really writes data to disk”? The human brain
will repeatedly foul this up—it’s called the Stroop Effect [Str35]. You
can try the following experiment yourself to see the effects of such in-
terference. Get some colored pens, and use them to write down the
names of colors. However, never write a color name using that color
pen. You could write the word “blue” in green, the word “brown” in red,
and so on. (Alternatively, we have a sample set of colors already drawn
on our Web site at www.pragmaticprogrammer.com.) Once you have
the color names drawn, try to say aloud the color with which each word
is drawn, as fast as you can. At some point you’ll trip up and start read-
ing the names of the colors, and not the colors themselves. Names are

Prepared exclusively for Zach

 250 C HAPTER 8 P RAGMATIC P ROJECTS

deeply meaningful to your brain, and misleading names add chaos to

your code.

You can document parameters, but ask yourself if it is really necessary

in all cases. The level of comment suggested by the JavaDoc tool seems
appropriate:

/**
* Find the peak (highest) value within a specified date
* range of samples.
*
* @param aRange Range of dates to search for data.
* @param aThreshold Minimum value to consider.
* @return the value, or <code>null</code> if no value found
* greater than or equal to the threshold.
*/
public Sample findPeak(DateRange aRange, double aThreshold);

Here’s a list of things that should not appear in source comments.

A list of the functions exported by code in the file. There are pro-
grams that analyze source for you. Use them, and the list is guar-
anteed to be up to date.

Revision history. This is what source code control systems are

for (see Source Code Control, page 86). However, it can be useful to
include information on the date of last change and the person who
made it.9

A list of other files this file uses. This can be determined more
accurately using automatic tools.

The name of the file. If it must appear in the file, don’t maintain it
by hand. RCS and similar systems can keep this information up to
date automatically. If you move or rename the file, you don’t want
to have to remember to edit the header.

One of the most important pieces of information that should appear

in the source file is the author’s name—not necessarily who edited
the file last, but the owner. Attaching responsibility and accountabil-
ity to source code does wonders in keeping people honest (see Pride
and Prejudice, page 258).

9. This kind of information, as well as the filename, is provided by the RCS $Id$ tag.

Prepared exclusively for Zach

 I T ’ S A LL W RITING 251

The project may also require certain copyright notices or other legal
boilerplate to appear in each source file. Get your editor to insert these
for you automatically.

With meaningful comments in place, tools such as JavaDoc [URL 7] and

DOC++ [URL 21] can extract and format them to automatically produce
API-level documentation. This is one specific example of a more general
technique we use—executable documents.

Executable Documents
Suppose we have a specification that lists the columns in a database
table. We’ll then have a separate set of SQL commands to create the
actual table in the database, and probably some kind of programming
language record structure to hold the contents of a row in the table.
The same information is repeated three times. Change any one of these
three sources, and the other two are immediately out of date. This is a
clear violation of the DRY principle.

To correct this problem, we need to choose the authoritative source

of information. This may be the specification, it may be a database
schema tool, or it may be some third source altogether. Let’s choose
the specification document as the source. It’s now our model for this
process. We then need to find a way to export the information it con-
tains as different views—a database schema and a high-level language
record, for example.10

If your document is stored as plain text with markup commands (using

HTML, LATEX, or troff, for example), then you can use tools such as Perl
to extract the schema and reformat it automatically. If your document
is in a word processor’s binary format, then see the box on the following
page for some options.

Your document is now an integral part of the project development. The

only way to change the schema is to change the document. You are
guaranteeing that the specification, schema, and code all agree. You
minimize the amount of work you have to do for each change, and you
can update the views of the change automatically.

10. See It’s Just a View, page 157, for more on models and views.

Prepared exclusively for Zach

 252 C HAPTER 8 P RAGMATIC P ROJECTS

What if My Document Isn’t Plain Text?

Unfortunately, more and more project documents are now being writ-
ten using word processors that store the file on disk in some propri-
etary format. We say “unfortunately” because this severely restricts
your options to process the document automatically. However, you
still have a couple of options:
Write macros. Most sophisticated word processors now have
a macro language. With some effort you can program them to
export tagged sections of your documents into the alternative
forms you need. If programming at this level is too painful, you
could always export the appropriate section into a standard for-
mat plain text file, and then use a tool such as Perl to convert this
into the final forms.
Make the document subordinate. Rather than have the doc-
ument as the definitive source, use another representation. (In
the database example, you might want to use the schema as the
authoritative information.) Then write a tool that exports this in-
formation into a form that the document can import. Be careful,
however. You need to ensure that this information is imported ev-
ery time the document is printed, rather than just once when the
document is created.

We can generate API-level documentation from source code using tools

such as JavaDoc and DOC++ in a similar fashion. The model is the
source code: one view of the model can be compiled; other views are
meant to be printed out or viewed on the Web. Our goal is always to
work on the model—whether the model is the code itself or some other
document—and have all views updated automatically (see Ubiquitous
Automation, page 230, for more on automatic processes).

Suddenly, documentation isn’t so bad.

Technical Writers
Up until now, we’ve talked only about internal documentation—written
by the programmers themselves. But what happens when you have
professional technical writers involved in the project? All too often, pro-
grammers just throw material “over the wall” to technical writers and

Prepared exclusively for Zach

 I T ’ S A LL W RITING 253

let them fend for themselves to produce user manuals, promotional

pieces, and so on.

This is a mistake. Just because programmers aren’t writing these docu-

ments doesn’t mean that we can forsake pragmatic principles. We want
the writers to embrace the same basic principles that a Pragmatic Pro-
grammer does—especially honoring the DRY principle, orthogonality,
the model-view concept, and the use of automation and scripting.

Print It or Weave It
One problem inherent with published, paper documentation is that it
can become out of date as soon as it’s printed. Documentation of any
form is just a snapshot.

So we try to produce all documentation in a form that can be published

online, on the Web, complete with hyperlinks. It’s easier to keep this
view of the documentation up to date than to track down every existing
paper copy, burn it, and reprint and redistribute new copies. It’s also a
better way to address the needs of a wide audience. Remember, though,
to put a date stamp or version number on each Web page. This way the
reader can get a good idea of what’s up to date, what’s changed recently,
and what hasn’t.

Many times you need to present the same documentation in different

formats: a printed document, Web pages, online help, or perhaps a slide
show. The typical solution relies heavily on cut-and-paste, creating a
number of new independent documents from the original. This is a bad
idea: a document’s presentation should be independent of its content.

If you are using a markup system, you have the flexibility to implement
as many different output formats as you need. You can choose to have

<H1>Chapter Title</H1>

generate a new chapter in the report version of the document and title
a new slide in the slide show. Technologies such as XSL and CSS11 can
be used to generate multiple output formats from this one markup.

11. eXtensible Style Language and Cascading Style Sheets, two technologies designed
to help separate presentation from content.

Prepared exclusively for Zach

 254 C HAPTER 8 P RAGMATIC P ROJECTS

If you are using a word processor, you’ll probably have similar capa-
bilities. If you remembered to use styles to identify different document
elements, then by applying different style sheets you can drastically
alter the look of the final output. Most word processors now allow you
to convert your document to formats such as HTML for Web publishing.

Markup Languages
Finally, for large-scale documentation projects, we recommend looking
at some of the more modern schemes for marking up documentation.

Many technical authors now use DocBook to define their documents.

DocBook is an SGML-based markup standard that carefully identifies
every component in a document. The document can be passed through
a DSSSL processor to render it into any number of different formats.
The Linux documentation project uses DocBook to publish information
in RTF, TEX, info, PostScript, and HTML formats.

As long as your original markup is rich enough to express all the con-
cepts you need (including hyperlinks), translation to any other pub-
lishable form can be both easy and automatic. You can produce online
help, published manuals, product highlights for the Web site, and even
a tip-a-day calendar, all from the same source—which of course is
under source control and is built along with the nightly build (see Ubiq-
uitous Automation, page 230).

Documentation and code are different views of the same underlying

model, but the view is all that should be different. Don’t let documen-
tation become a second-class citizen, banished from the main project
workflow. Treat documentation with the same care you treat code, and
the users (and maintainers who follow) will sing your praises.

Related sections include:

The Evils of Duplication, page 26
Orthogonality, page 34
The Power of Plain Text, page 73
Source Code Control, page 86
It’s Just a View, page 157
Programming by Coincidence, page 172
The Requirements Pit, page 202
Ubiquitous Automation, page 230

Prepared exclusively for Zach

 G REAT E XPECTATIONS 255

Challenges
Did you write an explanatory comment for the source code you just wrote?
Why not? Pressed for time? Not sure if the code will really work—are you
just trying out an idea as a prototype? You’ll throw the code away after-
wards, right? It won’t make it into the project uncommented and experi-
mental, will it?

Sometimes it is uncomfortable to document the design of source code be-

cause the design isn’t clear in your mind; it’s still evolving. You don’t feel
that you should waste effort describing what something does until it actu-
ally does it. Does this sound like programming by coincidence (page 172)?

45 Great Expectations
Be astonished, O ye heavens, at this, and be horribly afraid...
Jeremiah 2:12

A company announces record profits, and its share price drops 20%.
The financial news that night explains that the company failed to meet
analysts’ expectations. A child opens an expensive Christmas present
and bursts into tears—it wasn’t the cheap doll the child was hoping for.
A project team works miracles to implement a phenomenally complex
application, only to have it shunned by its users because it doesn’t have
a help system.

In an abstract sense, an application is successful if it correctly imple-

ments its specifications. Unfortunately, this pays only abstract bills.

In reality, the success of a project is measured by how well it meets the

expectations of its users. A project that falls below their expectations
is deemed a failure, no matter how good the deliverable is in absolute
terms. However, like the parent of the child expecting the cheap doll,
go too far and you’ll be a failure, too.

TIP 69

Gently Exceed Your Users’ Expectations

However, the execution of this tip requires some work.

Prepared exclusively for Zach

 256 C HAPTER 8 P RAGMATIC P ROJECTS

Communicating Expectations
Users initially come to you with some vision of what they want. It may
be incomplete, inconsistent, or technically impossible, but it is theirs,
and, like the child at Christmas, they have some emotion invested in it.
You cannot just ignore it.

As your understanding of their needs develops, you’ll find areas where

their expectations cannot be met, or where their expectations are per-
haps too conservative. Part of your role is to communicate this. Work
with your users so that their understanding of what you’ll be deliv-
ering is accurate. And do this throughout the development process.
Never lose sight of the business problems your application is intended
to solve.

Some consultants call this process “managing expectations”—actively

controlling what users should hope to get from their systems. We think
this is a somewhat elitist position. Our role is not to control the hopes
of our users. Instead, we need to work with them to come to a common
understanding of the development process and the final deliverable,
along with those expectations they have not yet verbalized. If the team
is communicating fluently with the outside world, this process is almost
automatic; everyone should understand what’s expected and how it will
be built.

There are some important techniques that can be used to facilitate this
process. Of these, Tracer Bullets, page 48, and Prototypes and Post-
it Notes, page 53, are the most important. Both let the team construct
something that the user can see. Both are ideal ways of communicating
your understanding of their requirements. And both let you and your
users practice communicating with each other.

The Extra Mile

If you work closely with your users, sharing their expectations and com-
municating what you’re doing, then there will be few surprises when
the project gets delivered.

This is a BAD THING. Try to surprise your users. Not scare them, mind
you, but delight them.

Prepared exclusively for Zach

 G REAT E XPECTATIONS 257

Give them that little bit more than they were expecting. The extra bit of
effort it requires to add some user-oriented feature to the system will
pay for itself time and time again in goodwill.

Listen to your users as the project progresses for clues about what
features would really delight them. Some things you can add relatively
easily that look good to the average user include:

Balloon or ToolTip help

Keyboard shortcuts
A quick reference guide as a supplement to the user’s manual
Colorization
Log file analyzers
Automated installation
Tools for checking the integrity of the system
The ability to run multiple versions of the system for training
A splash screen customized for their organization

All of these things are relatively superficial, and don’t really overburden
the system with feature bloat. However, each tells your users that the
development team cared about producing a great system, one that was
intended for real use. Just remember not to break the system adding
these new features.

Related sections include:

Good-Enough Software, page 9
Tracer Bullets, page 48
Prototypes and Post-it Notes, page 53
The Requirements Pit, page 202

Challenges
Sometimes the toughest critics of a project are the people who worked on
it. Have you ever experienced disappointment that your own expectations
weren’t met by something you produced? How could that be? Maybe there’s
more than logic at work here.

What do your users comment on when you deliver software? Is their atten-
tion to the various areas of the application proportional to the effort you
invested in each? What delights them?

Prepared exclusively for Zach

 258 C HAPTER 8 P RAGMATIC P ROJECTS

46 Pride and Prejudice

You have delighted us long enough.
Jane Austen, Pride and Prejudice

Pragmatic Programmers don’t shirk from responsibility. Instead, we

rejoice in accepting challenges and in making our expertise well known.
If we are responsible for a design, or a piece of code, we do a job we can
be proud of.

TIP 70

Sign Your Work

Craftsmen of an earlier age were proud to sign their work. You should
be, too.

Project teams are still made up of people, however, and this rule can
cause trouble. On some projects, the idea of code ownership can cause
cooperation problems. People may become territorial, or unwilling to
work on common foundation elements. The project may end up like a
bunch of insular little fiefdoms. You become prejudiced in favor of your
code and against your coworkers.

That’s not what we want. You shouldn’t jealously defend your code
against interlopers; by the same token, you should treat other peo-
ple’s code with respect. The Golden Rule (“Do unto others as you would
have them do unto you”) and a foundation of mutual respect among
the developers is critical to make this tip work.

Anonymity, especially on large projects, can provide a breeding ground

for sloppiness, mistakes, sloth, and bad code. It becomes too easy to see
yourself as just a cog in the wheel, producing lame excuses in endless
status reports instead of good code.

While code must be owned, it doesn’t have to be owned by an individual.

In fact, Kent Beck’s successful eXtreme Programming method [URL 45]
recommends communal ownership of code (but this also requires
additional practices, such as pair programming, to guard against the
dangers of anonymity).

Prepared exclusively for Zach

 P RIDE AND P REJUDICE 259

We want to see pride of ownership. “I wrote this, and I stand behind my

work.” Your signature should come to be recognized as an indicator of
quality. People should see your name on a piece of code and expect it
to be solid, well written, tested, and documented. A really professional
job. Written by a real professional.

A Pragmatic Programmer.

Prepared exclusively for Zach

 This page intentionally left blank

Prepared exclusively for Zach

 Appendix A

Resources
The only reason we were able to cover so much ground in this book is
that we viewed many of our subjects from a high altitude. If we’d given
them the in-depth coverage they deserved, the book would have been
ten times longer.

We started the book with the suggestion that Pragmatic Programmers

should always be learning. In this appendix we’ve listed resources that
may help you with this process.

In the section Professional Societies, we give details of the IEEE and the
ACM. We recommend that Pragmatic Programmers join one (or both)
of these societies. Then, in Building a Library, we highlight periodicals,
books, and Web sites that we feel contain high-quality and pertinent
information (or that are just plain fun).

Throughout the book we referenced many software resources accessible

via the Internet. In the Internet Resources section, we list the URLs
of these resources, along with a short description of each. However,
the nature of the Web means that many of these links may well be
stale by the time you read this book. You could try one of the many
search engines for a more up-to-date link, or visit our Web site at www.
pragmaticprogrammer.com and check our links section.

Finally, this appendix contains the book’s bibliography.

261

Prepared exclusively for Zach

 262 A PPENDIX A R ESOURCES

Professional Societies
There are two world-class professional societies for programmers: the
Association for Computing Machinery (ACM)1 and the IEEE Computer
Society.2 We recommend that all programmers belong to one (or both) of
these societies. In addition, developers outside the United States may
want to join their national societies, such as the BCS in the United
Kingdom.

Membership in a professional society has many benefits. The confer-

ences and local meetings give you great opportunities to meet people
with similar interests, and the special interest groups and technical
committees give you the opportunity to participate in setting standards
and guidelines used around the world. You’ll also get a lot out of their
publications, from high-level discussions of industry practice to low-
level computing theory.

Building a Library
We’re big on reading. As we noted in Your Knowledge Portfolio, page 12,
a good programmer is always learning. Keeping current with books and
periodicals can help. Here are some that we like.

Periodicals
If you’re like us, you’ll save old magazines and periodicals until they’re
piled high enough to turn the bottom ones to flat sheets of diamond.
This means it’s worth being fairly selective. Here are a few periodicals
we read.

IEEE Computer. Sent to members of the IEEE Computer Society,

Computer has a practical focus but is not afraid of theory. Some
issues are oriented around a theme, while others are simply col-

1. ACM Member Services, PO Box 11414, New York, NY 10286, USA.

www.acm.org

2. 1730 Massachusetts Avenue NW, Washington, DC 20036-1992, USA.

www.computer.org

Prepared exclusively for Zach

 B UILDING A L IBRARY 263

lections of interesting articles. This magazine has a good signal-to-

noise ratio.

IEEE Software. This is another great bimonthly publication of the

IEEE Computer Society aimed at software practitioners.

Communications of the ACM. The basic magazine received by all

members of the ACM, CACM has been a standard in the indus-
try for decades, and has probably published more seminal articles
than any other source.

SIGPLAN. Produced by the ACM Special Interest Group on Pro-

gramming Languages, SIGPLAN is an optional addition to your
ACM membership. It is often used for publishing language specifi-
cations, along with articles of interest to everyone who likes looking
deeply into programming.

Dr. Dobbs Journal. A monthly magazine, available by subscription

and on newsstands, Dr. Dobbs is quirky, but has articles ranging
from bit-level practice to heavy theory.

The Perl Journal. If you like Perl, you should probably subscribe
to The Perl Journal (www.tpj.com).

Software Development Magazine. A monthly magazine focusing on

general issues of project management and software development.

Weekly Trade Papers

There are several weekly newspapers published for developers and their
managers. These papers are largely a collection of company press re-
leases, redressed as articles. However, the content is still valuable—it
lets you track what is going on, keep abreast of new product announce-
ments, and follow industry alliances as they are forged and broken.
Don’t expect a lot of in-depth technical coverage, though.

Prepared exclusively for Zach

 264 A PPENDIX A R ESOURCES

Books
Computing books can be expensive, but choose carefully and they’re
a worthwhile investment. You may want to check out our Pragmatic
Bookshelf titles at http://pragmaticprogrammer.com. Additionally,
here are a handful of the many other books we like.

Analysis and Design

Object-Oriented Software Construction, 2nd Edition. Bertrand
Meyer’s epic book on the fundamentals of object-oriented develop-
ment, all in about 1,300 pages [Mey97b].

Design Patterns. A design pattern describes a way to solve a par-

ticular class of problems at a higher level than a programming lan-
guage idiom. This now-classic book [GHJV95] by the Gang of Four
describes 23 basic design patterns, including Proxy, Visitor, and
Singleton.

Analysis Patterns. A treasure trove of high-level, architectural pat-

terns taken from a wide variety of real-world projects and distilled
in book form. A relatively quick way to gain the insight of many
years of modeling experience [Fow96].

Teams and Projects

The Mythical Man Month. Fred Brooks’ classic on the perils of
organizing project teams, recently updated [Bro95].

Dynamics of Software Development. A series of short essays on

building software in large teams, focusing on the dynamics between
team members, and between the team and the rest of the world
[McC95].

Surviving Object-Oriented Projects: A Manager’s Guide. Alistair

Cockburn’s “reports from the trenches” illustrate many of the perils
and pitfalls of managing an OO project—especially your first one.
Mr. Cockburn provides tips and techniques to get you through the
most common problems [Coc97b].

Prepared exclusively for Zach

 B UILDING A L IBRARY 265

Specific Environments
Unix. W. Richard Stevens has several excellent books including
Advanced Programming in the Unix Environment and the Unix Net-
work Programming books [Ste92, Ste98, Ste99].

Windows. Marshall Brain’s Win32 System Services [Bra95] is a

concise reference to the low-level APIs. Charles Petzold’s Program-
ming Windows [Pet98] is the bible of Windows GUI development.

C++. As soon as you find yourself on a C++ project, run, don’t walk,
to the bookstore and get Scott Meyer’s Effective C++, and possibly
More Effective C++ [Mey97a, Mey96]. For building systems of any
appreciable size, you need John Lakos’ Large-Scale C++ Software
Design [Lak96]. For advanced techniques, turn to Jim Coplien’s
Advanced C++ Programming Styles and Idioms [Cop92].

In addition, the O’Reilly Nutshell series (www.ora.com) gives quick,

comprehensive treatments of miscellaneous topics and languages such
as perl, yacc, sendmail, Windows internals, and regular expressions.

The Web
Finding good content on the Web is hard. Here are several links that we
check at least once a week.

Slashdot. Billed as “News for nerds. Stuff that matters,” Slashdot

is one of the net homes of the Linux community. As well as regular
updates on Linux news, the site offers information on technologies
that are cool and issues that affect developers.
www.slashdot.org

Cetus Links. Thousands of links on object-oriented topics.

www.cetus-links.org

WikiWikiWeb. The Portland Pattern Repository and patterns dis-

cussion. Not just a great resource, the WikiWikiWeb site is an in-
teresting experiment in collective editing of ideas.
www.c2.com

Prepared exclusively for Zach

 266 A PPENDIX A R ESOURCES

Internet Resources
The links below are to resources available on the Internet. They were
valid at the time of writing, but (the Net being what it is) they may well
be out of date by the time you read this. If so, you could try a general
search for the filenames, or come to the Pragmatic Programmer Web
site (www.pragmaticprogrammer.com) and follow our links.

Editors
Emacs and vi are not the only cross-platform editors, but they are freely
available and widely used. A quick scan through a magazine such as
Dr. Dobbs will turn up several commercial alternatives.

Emacs
Both Emacs and XEmacs are available on Unix and Windows platforms.

[ URL 1] The Emacs Editor

www.gnu.org
The ultimate in big editors, containing every feature that any editor has
ever had, Emacs has a near-vertical learning curve, but repays hand-
somely once you’ve mastered it. It also makes a great mail and news reader,
address book, calendar and diary, adventure game, .

[ URL 2] The XEmacs Editor

www.xemacs.org
Spawned from the original Emacs some years ago, XEmacs is reputed to
have cleaner internals and a better-looking interface.

vi
There are at least 15 different vi clones available. Of these, vim is prob-
ably ported to the most platforms, and so would be a good choice of
editor if you find yourself working in many different environments.

[ URL 3] The Vim Editor

ftp://ftp.fu-berlin.de/misc/editors/vim
From the documentation: “There are a lot of enhancements above vi: multi
level undo, multi windows and buffers, syntax highlighting, command line
editing, filename completion, on-line help, visual selection, etc .”

Prepared exclusively for Zach

 I NTERNET R ESOURCES 267

[ URL 4] The elvis Editor

elvis.the-little-red-haired-girl.org
An enhanced vi clone with support for X.

[ URL 5] Emacs Viper Mode

http://www.cs.sunysb.edu/~kifer/emacs.html
Viper is a set of macros that make Emacs look like vi. Some may doubt
the wisdom of taking the world’s largest editor and extending it to emulate
an editor whose strength is its compactness. Others claim it combines the
best of both worlds.

Compilers, Languages, and Development Tools

[ URL 6] The GNU C/C++ Compiler
www.fsf.org/software/gcc/gcc.html
One of the most popular C and C++ compilers on the planet. It also does
Objective-C. (At the time of writing, the egcs project, which previously
splintered from gcc, is in the process of merging back into the fold.)

[ URL 7] The Java Language from Sun

java.sun.com
Home of Java, including downloadable SDKs, documentation, tutorials,
news, and more.

[ URL 8] Perl Language Home Page

www.perl.com
O’Reilly hosts this set of Perl-related resources.

[ URL 9] The Python Language

www.python.org
The Python object-oriented programming language is interpreted and in-
teractive, with a slightly quirky syntax and a wide and loyal following.

[ URL 10] SmallEiffel

SmallEiffel.loria.fr
The GNU Eiffel compiler runs on any machine that has an ANSI C compiler
and a Posix runtime environment.

[ URL 11] ISE Eiffel

www.eiffel.com
Interactive Software Engineering is the originator of Design by Contract,
and sells a commercial Eiffel compiler and related tools.

Prepared exclusively for Zach

 268 A PPENDIX A R ESOURCES

[ URL 12] Sather

www.icsi.berkeley.edu/~sather
Sather is an experimental language that grew out of Eiffel. It aims to sup-
port higher-order functions and iteration abstraction as well as Common
Lisp, CLU, or Scheme, and to be as efficient as C, C++, or Fortran.

[ URL 13] VisualWorks

www.cincom.com
Home of the VisualWorks Smalltalk environment. Noncommercial versions
for Windows and Linux are available for free.

[ URL 14] The Squeak Language Environment

squeak.cs.uiuc.edu
Squeak is a freely available, portable implementation of Smalltalk-80 writ-
ten in itself; it can produce C code output for higher performance.

[ URL 15] The TOM Programming Language

www.gerbil.org/tom
A very dynamic language with roots in Objective-C.

[ URL 16] The Beowulf Project

www.beowulf.org
A project that builds high-performance computers out of networked clus-
ters of inexpensive Linux boxes.

[ URL 17] iContract—Design by Contract Tool for Java

www.reliable-systems.com
Design by Contract formalism of preconditions, postconditions, and invari-
ants, implemented as a preprocessor for Java. Honors inheritance, imple-
ments existential quantifiers, and more.

[ URL 18] Nana—Logging and Assertions for C and C++

www.gnu.org/software/nana/nana.html
Improved support for assertion checking and logging in C and C++. It also
provides some support for Design by Contract.

[ URL 19] DDD—Data Display Debugger

http://www.gnu.org/software/ddd/
A free graphical front end for Unix debuggers.

[ URL 20] John Brant’s Refactoring Browser

st-www.cs.uiuc.edu/users/brant/Refactory
A popular refactoring browser for Smalltalk.

Prepared exclusively for Zach

 I NTERNET R ESOURCES 269

[ URL 21] DOC++ Documentation Generator

www.zib.de/Visual/software/doc++/index.html
DOC++ is a documentation system for C/C++ and Java that generates
both LATEX and HTML output for sophisticated online browsing of your
documentation directly from the C++ header or Java class files.

[ URL 22] xUnit—Unit Testing Framework

www.XProgramming.com
A simple but powerful concept, the xUnit unit testing framework provides
a consistent platform for testing software written in a variety of languages.

[ URL 23] The Tcl Language

www.scriptics.com
Tcl (“Tool Command Language”) is a scripting language designed to be easy
to embed into an application.

[ URL 24] Expect—Automate Interaction with Programs

expect.nist.gov
An extension built on Tcl [URL 23], expect allows you to script interac-
tion with programs. As well as helping you write command files that (for
example) fetch files from remote servers or extend the power of your shell,
expect can be useful when performing regression testing. A graphical ver-
sion, expectk, lets you wrap non-GUI applications with a windowing front
end.

[ URL 25] T Spaces

www.almaden.ibm.com/cs/TSpaces
From their Web page: “T Spaces is a network communication buffer with
database capabilities. It enables communication between applications and
devices in a network of heterogeneous computers and operating systems.
T Spaces provides group communication services, database services, URL-
based file transfer services, and event notification services.”

[ URL 26] javaCC—Java Compiler-Compiler

www.webgain.com/products/java_cc
A parser generator that is tightly coupled to the Java language.

[ URL 27] The bison Parser Generator

www.gnu.org/software/bison/bison.html
bison takes an input grammar specification and generates from it the C
source code of a suitable parser.

Prepared exclusively for Zach

 270 A PPENDIX A R ESOURCES

[ URL 28] SWIG—Simplified Wrapper and Interface Generator

www.swig.org
SWIG is a software development tool that connects programs written in C,
C++, and Objective-C with a variety of high-level programming languages
such as Perl, Python, and Tcl/Tk, as well as Java, Eiffel, and Guile.

[ URL 29] The Object Management Group, Inc.

www.omg.org
The OMG is the steward of various specifications for producing distributed
object-based systems. Their work includes the Common Object Request
Broker Architecture (CORBA) and the Internet Inter-ORB Protocol (IIOP).
Combined, these specifications make it possible for objects to communi-
cate with each other, even if they are written in different languages and
run on different types of computers.

Unix Tools Under DOS

[ URL 30] The UWIN Development Tools
www.gtlinc.com/uwin.html
Global Technologies, Inc., Old Bridge, NJ
The UWIN package provides Windows Dynamic Link Libraries (DLLs) that
emulate a large portion of the Unix C level library interface. Using this
interface, GTL has ported a large number of Unix command-line tools to
Windows. See also [URL 31].

[ URL 31] The Cygnus Cygwin Tools

sourceware.cygnus.com/cygwin/
Cygnus Solutions, Sunnyvale, CA
The Cygnus package also emulates the the Unix C library interface, and
provides a large array of Unix command-line tools under the Windows op-
erating system.

[ URL 32] Perl Power Tools

www.perl.com/pub/language/ppt/
A project to reimplement the classic Unix command set in Perl, making the
commands available on all platforms that support Perl (and that’s a lot of
platforms).

Prepared exclusively for Zach

 I NTERNET R ESOURCES 271

Source Code Control Tools

[ URL 33] RCS—Revision Control System
www.cs.purdue.edu/homes/trinkle/RCS/
GNU source code control system for Unix and Windows NT.

[ URL 34] CVS—Concurrent Version System

www.cvshome.com
Freely available source code control system for Unix and Windows NT.
Extends RCS by supporting a client-server model and concurrent access
to files.

[ URL 35] Aegis Transaction-Based Configuration Management

http://www.canb.auug.org.au/~millerp/aegis.html
A process-oriented revision control tool that imposes project standards
(such as verifying that checked-in code passes tests).

[ URL 36] ClearCase

www.rational.com
Version control, workspace and build management, process control.

[ URL 37] MKS Source Integrity

www.mks.com
Version control and configuration management. Some versions incorporate
features allowing remote developers to work on the same files simultane-
ously (much like CVS).

[ URL 38] PVCS Configuration Management

www.merant.com
A source code control system, very popular for Windows systems.

[ URL 39] Visual SourceSafe

www.microsoft.com
A version control system that integrates with Microsoft’s visual develop-
ment tools.

[ URL 40] Perforce

www.perforce.com
A client-server software configuration management system.

Prepared exclusively for Zach

 272 A PPENDIX A R ESOURCES

Other Tools
[ URL 41] WinZip—Archive Utility for Windows
www.winzip.com
Nico Mak Computing, Inc., Mansfield, CT
A Windows-based file archive utility. Supports both zip and tar formats.

[ URL 42] The Z Shell

sunsite.auc.dk/zsh
A shell designed for interactive use, although it is also a powerful scripting
language. Many of the useful features of bash, ksh, and tcsh were incor-
porated into zsh; many original features were added.

[ URL 43] A Free SMB Client for Unix Systems

samba.anu.edu.au/pub/samba/
Samba lets you share files and other resources between Unix and Windows
systems. Samba includes:

An SMB server, to provide Windows NT and LAN Manager-style file

and print services to SMB clients such as Windows 95, Warp Server,
smbfs, and others.
A Netbios nameserver, which among other things gives browsing sup-
port. Samba can be the master browser on your LAN if you wish.
An ftp-like SMB client that allows you to access PC resources (disks
and printers) from Unix, Netware, and other operating systems.

Papers and Publications

[ URL 44] The comp.object FAQ
www.cyberdyne-object-sys.com/oofaq2
A substantial and well-organized FAQ for the comp.object newsgroup.

[ URL 45] eXtreme Programming

www.XProgramming.com
From the Web site: “In XP, we use a very lightweight combination of prac-
tices to create a team that can rapidly produce extremely reliable, efficient,
well-factored software. Many of the XP practices were created and tested as
part of the Chrysler C3 project, which is a very successful payroll system
implemented in Smalltalk.”

[ URL 46] Alistair Cockburn’s Home Page

members.aol.com/acockburn
Look for “Structuring Use Cases with Goals” and use case templates.

Prepared exclusively for Zach

 I NTERNET R ESOURCES 273

[ URL 47] Martin Fowler’s Home Page

ourworld.compuserve.com/homepages/martin_fowler
Author of Analysis Patterns and co-author of UML Distilled and Refactor-
ing: Improving the Design of Existing Code. Martin Fowler’s home page dis-
cusses his books and his work with the UML.

[ URL 48] Robert C. Martin’s Home Page

www.objectmentor.com
Good introductory papers on object-oriented techniques, including depen-
dency analysis and metrics.

[ URL 49] Aspect-Oriented Programming

www.parc.xerox.com/csl/projects/aop/
An approach to adding functionality to code, both orthogonally and declar-
atively.

[ URL 50] JavaSpaces Specification

java.sun.com/products/javaspaces
A Linda-like system for Java that supports distributed persistence and
distributed algorithms.

[ URL 51] Netscape Source Code

www.mozilla.org
The development source of the Netscape browser.

[ URL 52] The Jargon File

www.jargon.org
Eric S. Raymond
Definitions for many common (and not so common) computer industry
terms, along with a good dose of folklore.

[ URL 53] Eric S. Raymond’s Papers

www.tuxedo.org/~esr
Eric’s papers on The Cathedral and the Bazaar and Homesteading the Noo-
sphere describing the psychosocietal basis for and implications of the Open
Source movement.

[ URL 54] The K Desktop Environment

www.kde.org
From their Web page: “KDE is a powerful graphical desktop environment
for Unix workstations. KDE is an Internet project and truly open in every
sense.”

Prepared exclusively for Zach

 274 A PPENDIX A R ESOURCES

[ URL 55] The GNU Image Manipulation Program

www.gimp.org
Gimp is a freely distributed program used for image creation, composition,
and retouching.

[ URL 56] The Demeter Project

www.ccs.neu.edu/research/demeter
Research focused on making software easier to maintain and evolve using
Adaptive Programming.

Miscellaneous
[ URL 57] The GNU Project
www.gnu.org
Free Software Foundation, Boston, MA
The Free Software Foundation is a tax-exempt charity that raises funds
for the GNU project. The GNU project’s goal is to produce a complete, free,
Unix-like system. Many of the tools they’ve developed along the way have
become industry standards.

[ URL 58] Web Server Information

www.netcraft.com/survey/servers.html
Links to the home pages of over 50 different web servers. Some are com-
mercial products, while others are freely available.

Bibliography
[Bak72] F. T. Baker. Chief programmer team management of pro-
duction programming. IBM Systems Journal, 11(1):56–73,
1972.

[BBM96] V. Basili, L. Briand, and W. L. Melo. A validation of object-

oriented design metrics as quality indicators. IEEE Trans-
actions on Software Engineering, 22(10):751–761, October
1996.

[Ber96] Albert J. Bernstein. Dinosaur Brains: Dealing with All Those

Impossible People at Work. Ballantine Books, New York, NY,
1996.

Prepared exclusively for Zach

 B IBLIOGRAPHY 275

[Bra95] Marshall Brain. Win32 System Services. Prentice Hall, En-

glewood Cliffs, NJ, 1995.

[Bro95] Frederick P. Brooks, Jr. The Mythical Man Month: Essays

on Software Engineering. Addison-Wesley, Reading, MA, an-
niversary edition, 1995.

[CG90] N. Carriero and D. Gelenter. How to Write Parallel Programs:

A First Course. MIT Press, Cambridge, MA, 1990.

[Cla04] Mike Clark. Pragmatic Project Automation. The Pragmatic

Programmers, LLC, Raleigh, NC, and Dallas, TX, 2004.

[CN91] Brad J. Cox and Andrex J. Novobilski. Object-Oriented

Programming, An Evolutionary Approach. Addison-Wesley,
Reading, MA, 1991.

[Coc97a] Alistair Cockburn. Goals and use cases. Journal of Object

Oriented Programming, 9(7):35–40, September 1997.

[Coc97b] Alistair Cockburn. Surviving Object-Oriented Projects: A

Manager’s Guide. Addison Wesley Longman, Reading, MA,
1997.

[Cop92] James O. Coplien. Advanced C++ Programming Styles and

Idioms. Addison-Wesley, Reading, MA, 1992.

[DL99] Tom Demarco and Timothy Lister. Peopleware: Productive

Projects and Teams. Dorset House, New York, NY, second
edition, 1999.

[FBB 99] Martin Fowler, Kent Beck, John Brant, William Opdyke, and
Don Roberts. Refactoring: Improving the Design of Existing
Code. Addison Wesley Longman, Reading, MA, 1999.

[Fow96] Martin Fowler. Analysis Patterns: Reusable Object Models.

Addison Wesley Longman, Reading, MA, 1996.

[FS99] Martin Fowler and Kendall Scott. UML Distilled: Applying the
Standard Object Modeling Language. Addison Wesley Long-
man, Reading, MA, second edition, 1999.

[GHJV95] Erich Gamma, Richard Helm, Ralph Johnson, and John

Vlissides. Design Patterns: Elements of Reusable Object-
Oriented Software. Addison-Wesley, Reading, MA, 1995.

Prepared exclusively for Zach

 276 A PPENDIX A R ESOURCES

[Gla99a] Robert L. Glass. Inspections—Some surprising findings.

Communications of the ACM, 42(4):17–19, April 1999.

[Gla99b] Robert L. Glass. The realities of software technology payoffs.

Communications of the ACM, 42(2):74–79, February 1999.

[Hol78] Michael Holt. Math Puzzles and Games. Dorset Press, New
York, NY, 1978.

[HT03] Andy Hunt and Dave Thomas. Pragmatic Unit Testing In

Java with JUnit. The Pragmatic Programmers, LLC, Raleigh,
NC, and Dallas, TX, 2003.

[Jac94] Ivar Jacobson. Object-Oriented Software Engineering: A Use-

Case Driven Approach. Addison-Wesley, Reading, MA, 1994.

[KLM 97] Gregor Kiczales, John Lamping, Anurag Mendhekar, Chris

Maeda, Cristina Videira Lopes, Jean-Marc Loingtier, and
John Irwin. Aspect-oriented programming. In European
Conference on Object-Oriented Programming (ECOOP), vol-
ume LNCS 1241. Springer-Verlag, June 1997.

[Knu97a] Donald Ervin Knuth. The Art of Computer Programming: Fun-

damental Algorithms, volume 1. Addison Wesley Longman,
Reading, MA, third edition, 1997.

[Knu97b] Donald Ervin Knuth. The Art of Computer Programming:

Seminumerical Algorithms, volume 2. Addison Wesley Long-
man, Reading, MA, third edition, 1997.

[Knu98] Donald Ervin Knuth. The Art of Computer Programming:

Sorting and Searching, volume 3. Addison Wesley Longman,
Reading, MA, second edition, 1998.

[KP99] Brian W. Kernighan and Rob Pike. The Practice of Program-

ming. Addison Wesley Longman, Reading, MA, 1999.

[Kru98] Philippe Kruchten. The Rational Unified Process: An Intro-

duction. Addison Wesley Longman, Reading, MA, 1998.

[Lak96] John Lakos. Large-Scale C++ Software Design. Addison

Wesley Longman, Reading, MA, 1996.

Prepared exclusively for Zach

 B IBLIOGRAPHY 277

[LH89] Karl J. Lieberherr and Ian Holland. Assuring good style

for object-oriented programs. IEEE Software, pages 38–48,
September 1989.

[Lis88] Barbara Liskov. Data abstraction and hierarchy. SIGPLAN

Notices, 23(5), May 1988.

[LMB92] John R. Levine, Tony Mason, and Doug Brown. Lex and
Yacc. O’Reilly & Associates, Inc., Sebastopol, CA, second
edition, 1992.

[McC95] Jim McCarthy. Dynamics of Software Development. Mi-

crosoft Press, Redmond, WA, 1995.

[Mey96] Scott Meyers. More Effective C++: 35 New Ways to Improve

Your Programs and Designs. Addison-Wesley, Reading, MA,
1996.

[Mey97a] Scott Meyers. Effective C++: 50 Specific Ways to Improve

Your Programs and Designs. Addison Wesley Longman,
Reading, MA, second edition, 1997.

[Mey97b] Bertrand Meyer. Object-Oriented Software Construction.

Prentice Hall, Englewood Cliffs, NJ, second edition, 1997.

[Pet98] Charles Petzold. Programming Windows, The Definitive

Guide to the Win32 API. Microsoft Press, Redmond, WA, fifth
edition, 1998.

[Sch95] Bruce Schneier. Applied Cryptography: Protocols, Algo-

rithms, and Source Code in C. John Wiley & Sons, New York,
NY, second edition, 1995.

[Sed83] Robert Sedgewick. Algorithms. Addison-Wesley, Reading,

MA, 1983.

[Sed92] Robert Sedgewick. Algorithms in C++. Addison-Wesley,

Reading, MA, 1992.

[SF96] Robert Sedgewick and Phillipe Flajolet. An Introduction to

the Analysis of Algorithms. Addison-Wesley, Reading, MA,
1996.

[Ste92] W. Richard Stevens. Advanced Programming in the Unix En-

vironment. Addison-Wesley, Reading, MA, 1992.

Prepared exclusively for Zach

 278 A PPENDIX A R ESOURCES

[Ste98] W. Richard Stevens. Unix Network Programming, Volume 1:

Networking APIs: Sockets and Xti. Prentice Hall, Englewood
Cliffs, NJ, second edition, 1998.

[Ste99] W. Richard Stevens. Unix Network Programming, Volume

2: Interprocess Communications. Prentice Hall, Englewood
Cliffs, NJ, second edition, 1999.

[Str35] James Ridley Stroop. Studies of interference in serial verbal

reactions. Journal of Experimental Psychology, 18:643–662,
1935.

[TFH04] Dave Thomas, Chad Fowler, and Andy Hunt. Programming

Ruby, The Pragmatic Programmers’ Guide. The Pragmatic
Programmers, LLC, Raleigh, NC, and Dallas, TX, 2004.

[TH03] Dave Thomas and Andy Hunt. Pragmatic Version Control

Using CVS. The Pragmatic Programmers, LLC, Raleigh, NC,
and Dallas, TX, 2003.

[WK82] James Q. Wilson and George Kelling. The police and neigh-
borhood safety. The Atlantic Monthly, 249(3):29–38, March
1982.

[YC86] Edward Yourdon and Larry L. Constantine. Structured De-

sign: Fundamentals of a Discipline of Computer Program and
Systems Design. Prentice Hall, Englewood Cliffs, NJ, second
edition, 1986.

[You95] Edward Yourdon. When good-enough software is best. IEEE

Software, May 1995.

Prepared exclusively for Zach

 Appendix B

Answers to Exercises
Exercise 1: from Orthogonality on page 43
You are writing a class called Split, which splits input lines into fields. Which
of the following two Java class signatures is the more orthogonal design?

class Split1 {
public Split1(InputStreamReader rdr) { ...
public void readNextLine() throws IOException { ...
public int numFields() { ...
public String getField(int fieldNo) { ...
}
class Split2 {
public Split2(String line) { ...
public int numFields() { ...
public String getField(int fieldNo) { ...
}

Answer 1: To our way of thinking, class Split2 is more orthogonal. It con-

centrates on its own task, splitting lines, and ignores details such as where the
lines are coming from. Not only does this make the code easier to develop, but
it also makes it more flexible. Split2 can split lines read from a file, generated
by another routine, or passed in via the environment.

Exercise 2: from Orthogonality on page 43

Which will lead to a more orthogonal design: modeless or modal dialog boxes?

Answer 2: If done correctly, probably modeless. A system that uses mode-

less dialog boxes will be less concerned with what is going on at any particular
moment in time. It will likely have a better intermodule communications in-
frastructure than a modal system, which may have built-in assumptions about
the state of the system—assumptions that lead to increased coupling and de-
creased orthogonality.

279

Prepared exclusively for Zach

 280 A PPENDIX B A NSWERS TO E XERCISES

Exercise 3: from Orthogonality on page 43

How about procedural languages versus object technology? Which results in a
more orthogonal system?

Answer 3: This is a little tricky. Object technology can provide a more orthog-
onal system, but because it has more features to abuse, it is actually easier to
create a nonorthogonal system using objects than it is using a procedural lan-
guage. Features such as multiple inheritance, exceptions, operator overload-
ing, and parent-method overriding (via subclassing) provide ample opportunity
to increase coupling in nonobvious ways.

With object technology and a little extra effort, you can achieve a much more
orthogonal system. But while you can always write “spaghetti code” in a pro-
cedural language, object-oriented languages used poorly can add meatballs to
your spaghetti.

Exercise 4: from Prototypes and Post-it Notes on page 56

Marketing would like to sit down and brainstorm a few Web-page designs with
you. They are thinking of clickable image maps to take you to other pages, and
so on. But they can’t decide on a model for the image—maybe it’s a car, or a
phone, or a house. You have a list of target pages and content; they’d like to
see a few prototypes. Oh, by the way, you have 15 minutes. What tools might
you use?

Answer 4: Low-tech to the rescue! Draw a few cartoons with markers on a

whiteboard—a car, a phone, and a house. It doesn’t have to be great art; stick-
figure outlines are fine. Put Post-it notes that describe the contents of target
pages on the clickable areas. As the meeting progresses, you can refine the
drawings and placements of the Post-it notes.

Exercise 5: from Domain Languages on page 63

We want to implement a mini-language to control a simple drawing package
(perhaps a turtle-graphics system). The language consists of single-letter com-
mands. Some commands are followed by a single number. For example, the
following input would draw a rectangle.

P 2 # select pen 2
D # pen down
W 2 # draw west 2cm
N 1 # then north 1
E 2 # then east 2
S 1 # then back south
U # pen up

Implement the code that parses this language. It should be designed so that it
is simple to add new commands.

Prepared exclusively for Zach

 281

Answer 5: Because we want to make the language extendable, we’ll make the
parser table driven. Each entry in the table contains the command letter, a flag
to say whether an argument is required, and the name of the routine to call to
handle that particular command.
typedef struct {
char cmd; /* the command letter */
int hasArg; /* does it take an argument */
void (*func)(int, int); /* routine to call */
} Command;
static Command cmds[] = {
{ ’P’, ARG, doSelectPen },
{ ’U’, NO_ARG, doPenUp },
{ ’D’, NO_ARG, doPenDown },
{ ’N’, ARG, doPenDir },
{ ’E’, ARG, doPenDir },
{ ’S’, ARG, doPenDir },
{ ’W’, ARG, doPenDir }
};

The main program is pretty simple: read a line, look up the command, get the
argument if required, then call the handler function.
while (fgets(buff, sizeof(buff), stdin)) {
Command *cmd = findCommand(*buff);
if (cmd) {
int arg = 0;
if (cmd->hasArg && !getArg(buff+1, &arg)) {
fprintf(stderr, "’%c’ needs an argument n", *buff);
continue;
}
cmd->func(*buff, arg);
}
}

The function that looks up a command performs a linear search of the table,
returning either the matching entry or NULL.
Command *findCommand(int cmd) {
int i;
for (i = 0; i < ARRAY_SIZE(cmds); i++) {
if (cmds[i].cmd == cmd)
return cmds + i;
}
fprintf(stderr, "Unknown command ’%c’ n", cmd);
return 0;
}

Finally, reading the numeric argument is pretty simple using sscanf.

int getArg(const char *buff, int *result) {
return sscanf(buff, "%d", result) == 1;
}

Prepared exclusively for Zach

 282 A PPENDIX B A NSWERS TO E XERCISES

Exercise 6: from Domain Languages on page 63

Design a BNF grammar to parse a time specification. All of the following exam-
ples should be accepted.
4pm, 7:38pm, 23:42, 3:16, 3:16am

Answer 6: Using BNF, a time specification could be

time hour ampm

hour : minute ampm
hour : minute
ampm am pm
hour digit
digit digit
minute digit digit
digit 0 1 2 3 4 5 6 7 8 9

Exercise 7: from Domain Languages on page 63

Implement a parser for the BNF grammar in Exercise 6 using yacc, bison, or
a similar parser-generator.

Answer 7: We coded our example using bison, the GNU version of yacc. For
clarity, we’re just showing the body of the parser here. Look at the source on
our Web site for the full implementation.
time: spec END_TOKEN
{ if ($1 >= 24*60) yyerror("Time is too large");
printf("%d minutes past midnight n", $1);
exit(0);
}
;
spec: hour ’:’ minute
{ $$ = $1 + $3;
}
| hour ’:’ minute ampm
{ if ($1 > 11*60) yyerror("Hour out of range");
$$ = $1 + $3 + $4;
}
| hour ampm
{ if ($1 > 11*60) yyerror("Hour out of range");
$$ = $1 + $2;
}
;
hour: hour_num
{ if ($1 > 23) yyerror("Hour out of range");
$$ = $1 * 60;
};

Prepared exclusively for Zach

 283

minute: DIGIT DIGIT

{ $$ = $1*10 + $2;
if ($$ > 59) yyerror("minute out of range");
};
ampm: AM { $$ = AM_MINS; }
| PM { $$ = PM_MINS; }
;
hour_num: DIGIT { $$ = $1; }
| DIGIT DIGIT { $$ = $1*10 + $2; }
;

Exercise 8: from Domain Languages on page 63

Implement the time parser using Perl. [Hint: Regular expressions make good
parsers.]

Answer 8:
$_ = shift;
/^( d d?)(am|pm)$/ && doTime($1, 0, $2, 12);
/^( d d?):( d d)(am|pm)$/ && doTime($1, $2, $3, 12);
/^( d d?):( d d)$/ && doTime($1, $2, 0, 24);
die "Invalid time $_ n";
#
# doTime(hour, min, ampm, maxHour)
#
sub doTime($$$$) {
my ($hour, $min, $offset, $maxHour) = @_;
die "Invalid hour: $hour" if ($hour >= $maxHour);
$hour += 12 if ($offset eq "pm");
print $hour*60 + $min, " minutes past midnight n";
exit(0);
}

Exercise 9: from Estimating on page 69

You are asked “Which has a higher bandwidth: a 1Mbps communications line
or a person walking between two computers with a full 4GB tape in their
pocket?” What constraints will you put on your answer to ensure that the scope
of your response is correct? (For example, you might say that the time taken to
access the tape is ignored.)

Answer 9: Our answer must be couched in several assumptions:

The tape contains the information we need to be transferred.

We know the speed at which the person walks.
We know the distance between the machines.
We are not accounting for the time it takes to transfer information to and
from the tape.

Prepared exclusively for Zach

 284 A PPENDIX B A NSWERS TO E XERCISES

The overhead of storing data on a tape is roughly equal to the overhead of

sending it over a communications line.

Exercise 10: from Estimating on page 69

So, which has the higher bandwidth?

Answer 10: Subject to the caveats in Answer 9: A 4GB tape contains

bits, so a 1Mbps line would have to pump data for about seconds, or
roughly hours, to transfer the equivalent amount of information. If the person
is walking at a constant mph, then our two machines would need to be at
least miles apart for the communications line to outperform our courier.
Otherwise, the person wins.

Exercise 11: from Text Manipulation on page 102

Your C program uses an enumerated type to represent one of 100 states. You’d
like to be able to print out the state as a string (as opposed to a number)
for debugging purposes. Write a script that reads from standard input a file
containing
name
state_a
state_b
: :

Produce the file name.h, which contains

extern const char* NAME_names[];

typedef enum {
state_a,
state_b,
: :
} NAME;

and the file name.c, which contains

const char* NAME_names[] = {

"state_a",
"state_b",
: :
};

Prepared exclusively for Zach

 285

Answer 11: We implemented our answer using Perl.

my @consts;
my $name = <>;
die "Invalid format - missing name" unless defined($name);
chomp $name;
# Read in the rest of the file
while (<>) {
chomp;
s/^ s*//; s/ s*$//;
die "Invalid line: $_" unless /^( w+)$/;
push @consts, $_;
}
# Now generate the file
open(HDR, ">$name.h") or die "Can’t open $name.h: $!";
open(SRC, ">$name.c") or die "Can’t open $name.c: $!";
my $uc_name = uc($name);
my $array_name = $uc_name . "_names";
print HDR "/* File generated automatically - do not edit */ n";
print HDR "extern const char *$ {array_name}[];";
print HDR "typedef enum { n ";
print HDR join ", n ", @consts;
print HDR " n} $uc_name; n n";
print SRC "/* File generated automatically - do not edit */ n";
print SRC "const char *$ {array_name}[] = { n "";
print SRC join " ", n "", @consts;
print SRC " " n}; n";
close(SRC);
close(HDR);

Using the DRY principle, we won’t cut and paste this new file into our code.
Instead, we’ll #include it—the flat file is the master source of these constants.
This means that we’ll need a makefile to regenerate the header when the file
changes. The following extract is from the test bed in our source tree (available
on the Web site).

etest.c etest.h: etest.inc enumerated.pl

perl enumerated.pl etest.inc

Exercise 12: from Text Manipulation on page 102

Halfway through writing this book, we realized that we hadn’t put the use
strict directive into many of our Perl examples. Write a script that goes
through the .pl files in a directory and adds a use strict at the end of
the initial comment block to all files that don’t already have one. Remember to
keep a backup of all files you change.

Prepared exclusively for Zach

 286 A PPENDIX B A NSWERS TO E XERCISES

Answer 12: Here’s our answer, written in Perl.

my $dir = shift or die "Missing directory";

for my $file (glob("$dir/*.pl")) {
open(IP, "$file") or die "Opening $file: $!";
undef $/; # Turn off input record separator --
my $content = <IP>; # read whole file as one string.
close(IP);
if ($content !~ /^use strict/m) {
rename $file, "$file.bak" or die "Renaming $file: $!";
open(OP, ">$file") or die "Creating $file: $!";
# Put ’use strict’ on first line that
# doesn’t start ’#’
$content =~ s/^(?!#)/ nuse strict; n n/m;
print OP $content;
close(OP);
print "Updated $file n";
}
else {
print "$file already strict n";
}
}

Exercise 13: from Code Generators on page 106

Write a code generator that takes the input file in Figure 3.4, page 106, and
generates output in two languages of your choice. Try to make it easy to add
new languages.

Answer 13: We use Perl to implement our solution. It dynamically loads a

module to generate the requested language, so adding new languages is easy.
The main routine loads the back end (based on a command-line parameter),
then reads its input and calls code generation routines based on the content of
each line. We’re not particularly fussy about error handling—we’ll get to know
pretty quickly if things go wrong.

my $lang = shift or die "Missing language";

$lang .= "_cg.pm";
require "$lang" or die "Couldn’t load $lang";
# Read and parse the file
my $name;
while (<>) {
chomp;
if (/^ s*$/) { CG::blankLine(); }
elsif (/^ #(.*)/) { CG::comment($1); }
elsif (/^M s*(.+)/) { CG::startMsg($1); $name = $1; }
elsif (/^E/) { CG::endMsg($name); }
elsif (/^F s*( w+) s+( w+)$/)
{ CG::simpleType($1,$2); }

Prepared exclusively for Zach

 287

elsif (/^F s*( w+) s+( w+) [( d+) ]$/)

{ CG::arrayType($1,$2,$3); }
else {
die "Invalid line: $_";
}
}

Writing a language back end is simple: provide a module that implements the
required six entry points. Here’s the C generator:

#!/usr/bin/perl -w
package CG;
use strict;
# Code generator for ’C’ (see cg_base.pl)
sub blankLine() { print " n"; }
sub comment() { print "/*$_[0] */ n"; }
sub startMsg() { print "typedef struct { n"; }
sub endMsg() { print "} $_[0]; n n"; }
sub arrayType() {
my ($name, $type, $size) = @_;
print " $type $name [$size]; n";
}
sub simpleType() {
my ($name, $type) = @_;
print " $type $name; n";
}
1;

And here’s the one for Pascal:

#!/usr/bin/perl -w
package CG;
use strict;
# Code generator for ’Pascal’ (see cg_base.pl)
sub blankLine() { print " n"; }
sub comment() { print "{$_[0] } n"; }
sub startMsg() { print "$_[0] = packed record n"; }
sub endMsg() { print "end; n n"; }
sub arrayType() {
my ($name, $type, $size) = @_;
$size--;
print " $name: array[0..$size] of $type; n";
}
sub simpleType() {
my ($name, $type) = @_;
print " $name: $type; n";
}
1;

Prepared exclusively for Zach

 288 A PPENDIX B A NSWERS TO E XERCISES

Exercise 14: from Design by Contract on page 118

What makes a good contract? Anyone can add preconditions and postcondi-
tions, but will they do you any good? Worse yet, will they actually do more
harm than good? For the example below and for those in Exercises 15 and 16,
decide whether the specified contract is good, bad, or ugly, and explain why.

First, let’s look at an Eiffel example. Here we have a routine for adding a STRING
to a doubly linked, circular list (remember that preconditions are labeled with
require, and postconditions with ensure).

-- Add a unique item to a doubly linked list,

-- and return the newly created NODE.
add_item (item : STRING) : NODE is
require
item /= Void -- ’/=’ is ’not equal’.
find_item(item) = Void -- Must be unique
deferred -- Abstract base class.
ensure
result.next.previous = result -- Check the newly
result.previous.next = result -- added node’s links.
find_item(item) = result -- Should find it.
end

Answer 14: This Eiffel example is good. We require non-null data to be

passed in, and we guarantee that the semantics of a circular, doubly linked
list are honored. It also helps to be able to find the string we stored. Because
this is a deferred class, the actual class that implements it is free to use what-
ever underlying mechanism it wants to. It may choose to use pointers, or an
array, or whatever; as long as it honors the contract, we don’t care.

Exercise 15: from Design by Contract on page 119

Next, let’s try an example in Java—somewhat similar to the example in Exercise
14. insertNumber inserts an integer into an ordered list. Pre- and postcondi-
tions are labeled as in iContract (see [URL 17]).

private int data[];

/**
* @post data[index-1] < data[index] &&
* data[index] == aValue
*/
public Node insertNumber (final int aValue)
{
int index = findPlaceToInsert(aValue);
...

Answer 15: This is bad. The math in the index clause (index-1) won’t work
on boundary conditions such as the first entry. The postcondition assumes a
particular implementation: we want contracts to be more abstract than that.

Prepared exclusively for Zach

 289

Exercise 16: from Design by Contract on page 119

Here’s a fragment from a stack class in Java. Is this a good contract?

/**
* @pre anItem != null // Require real data
* @post pop() == anItem // Verify that it’s
* // on the stack
*/
public void push(final String anItem)

Answer 16: It’s a good contract, but a bad implementation. Here, the infa-
mous “Heisenbug” [URL 52] rears its ugly head. The programmer probably just
made a simple typo—pop instead of top. While this is a simple and contrived
example, side effects in assertions (or in any unexpected place in the code) can
be very difficult to diagnose.

Exercise 17: from Design by Contract on page 119

The classic examples of DBC (as in Exercises 14–16) show an implementation of
an ADT (Abstract Data Type)—typically a stack or queue. But not many people
really write these kinds of low-level classes.

So, for this exercise, design an interface to a kitchen blender. It will eventually
be a Web-based, Internet-enabled, CORBA-fied blender, but for now we just
need the interface to control it. It has ten speed settings (0 means off). You
can’t operate it empty, and you can change the speed only one unit at a time
(that is, from 0 to 1, and from 1 to 2, not from 0 to 2).

Here are the methods. Add appropriate pre- and postconditions and an invari-
ant.
int getSpeed()
void setSpeed(int x)
boolean isFull()
void fill()
void empty()

Answer 17: We’ll show the function signatures in Java, with the pre- and
postconditions labeled as in iContract.

First, the invariant for the class:

/**
* @invariant getSpeed() > 0
* implies isFull() // Don’t run empty
* @invariant getSpeed() >= 0 &&
* getSpeed() < 10 // Range check
*/

Next, the pre- and postconditions:

Prepared exclusively for Zach

 290 A PPENDIX B A NSWERS TO E XERCISES

/**
* @pre Math.abs(getSpeed() - x) <= 1 // Only change by one
* @pre x >= 0 && x < 10 // Range check
* @post getSpeed() == x // Honor requested speed
*/
public void setSpeed(final int x)
/**
* @pre !isFull() // Don’t fill it twice
* @post isFull() // Ensure it was done
*/
void fill()
/**
* @pre isFull() // Don’t empty it twice
* @post !isFull() // Ensure it was done
*/
void empty()

Exercise 18: from Design by Contract on page 119

How many numbers are in the series ?

Answer 18: There are 21 terms in the series. If you said 20, you just experi-
enced a fencepost error.

Exercise 19: from Assertive Programming on page 125

A quick reality check. Which of these “impossible” things can happen?

1. A month with fewer than 28 days

2. stat(".",&sb) == -1 (that is, can’t access the current directory)
3. In C++: a = 2; b = 3; if (a + b != 5) exit(1);
4. A triangle with an interior angle sum
5. A minute that doesn’t have 60 seconds
6. In Java: (a + 1) <= a

Answer 19:
1. September, 1752 had only 19 days. This was done to synchronize calen-
dars as part of the Gregorian Reformation.
2. The directory could have been removed by another process, you might not
have permission to read it, &sb might be invalid—you get the picture.
3. We sneakily didn’t specify the types of a and b. Operator overloading might
have defined +, =, or != to have unexpected behavior. Also, a and b may
be aliases for the same variable, so the second assignment will overwrite
the value stored in the first.
4. In non-Euclidean geometry, the sum of the angles of a triangle will not add
up to . Think of a triangle mapped on the surface of a sphere.

Prepared exclusively for Zach

 291

5. Leap minutes may have 61 or 62 seconds.

6. Overflow may leave the result of a + 1 negative (this can also happen in
C and C++).

Exercise 20: from Assertive Programming on page 125

Develop a simple assertion checking class for Java.

Answer 20: We chose to implement a very simple class with a single static
method, TEST, that prints a message and a stack trace if the passed condition
parameter is false.

package com.pragprog.util;
import java.lang.System; // for exit()
import java.lang.Thread; // for dumpStack()
public class Assert {
/** Write a message, print a stack trace and exit if
* our parameter is false.
*/
public static void TEST(boolean condition) {
if (!condition) {
System.out.println("==== Assertion Failed ====");
Thread.dumpStack();
System.exit(1);
}
}
// Testbed. If our argument is ’okay’, try an assertion that
// succeeds, if ’fail’ try one that fails
public static final void main(String args[]) {
if (args[0].compareTo("okay") == 0) {
TEST(1 == 1);
}
else if (args[0].compareTo("fail") == 0) {
TEST(1 == 2);
}
else {
throw new RuntimeException("Bad argument");
}
}
}

Exercise 21: from When to Use Exceptions on page 128

While designing a new container class, you identify the following possible error
conditions:

1. No memory available for a new element in the add routine

2. Requested entry not found in the fetch routine
3. null pointer passed to the add routine

Prepared exclusively for Zach

 292 A PPENDIX B A NSWERS TO E XERCISES

How should each be handled? Should an error be generated, should an excep-

tion be raised, or should the condition be ignored?

Answer 21: Running out of memory is an exceptional condition, so we feel

that case (1) should raise an exception.

Failure to find an entry is probably quite a normal occurrence. The application

that calls our collection class may well write code that checks to see if an entry
is present before adding a potential duplicate. We feel that case (2) should just
return an error.

Case (3) is more problematic—if the value null is significant to the application,
then it may be justifiably added to the container. If, however, it makes no sense
to store null values, an exception should probably be thrown.

Exercise 22: from How to Balance Resources on page 136

Some C and C++ developers make a point of setting a pointer to NULL after they
deallocate the memory it references. Why is this a good idea?

Answer 22: In most C and C++ implementations, there is no way of checking

that a pointer actually points to valid memory. A common mistake is to deal-
locate a block of memory and reference that memory later in the program. By
then, the memory pointed to may well have been reallocated to some other pur-
pose. By setting the pointer to NULL, the programmers hope to prevent these
rogue references—in most cases, dereferencing a NULL pointer will generate a
runtime error.

Exercise 23: from How to Balance Resources on page 136

Some Java developers make a point of setting an object variable to NULL after
they have finished using the object. Why is this a good idea?

Answer 23: By setting the reference to NULL, you reduce the number of point-
ers to the referenced object by one. Once this count reaches zero, the object is
eligible for garbage collection. Setting the references to NULL can be signifi-
cant for long-running programs, where the programmers need to ensure that
memory utilization doesn’t increase over time.

Prepared exclusively for Zach

 293

Exercise 24: from Decoupling and the Law of Demeter on page 143
We discussed the concept of physical decoupling in the box on page 142. Which
of the following C++ header files is more tightly coupled to the rest of the
system?

person1.h: person2.h:
#include "date.h" class Date;
class Person1 { class Person2 {
private: private:
Date myBirthdate; Date *myBirthdate;
public: public:
Person1(Date &birthDate); Person2(Date &birthDate);
// ... // ...

Answer 24: A header file is supposed to define the interface between the
corresponding implementation and the rest of the world. The header file itself
has no need to know about the internals of the Date class—it merely needs to
tell the compiler that the constructor takes a Date as a parameter. So, unless
the header file uses Dates in inline functions, the second snippet will work
fine.

What’s wrong with the first snippet? On a small project, nothing, except that
you are unnecessarily making everything that uses a Person1 class also in-
clude the header file for Date. Once this kind of usage gets common in a
project, you soon find that including one header file ends up including most
of the rest of the system—a serious drag on compilation times.

Exercise 25: from Decoupling and the Law of Demeter on page 143
For the example below and for those in Exercises 26 and 27, determine if the
method calls shown are allowed according to the Law of Demeter. This first one
is in Java.
public void showBalance(BankAccount acct) {
Money amt = acct.getBalance();
printToScreen(amt.printFormat());
}

Answer 25: The variable acct is passed in as a parameter, so the getBal-

ance call is allowed. Calling amt.printFormat(), however, is not. We don’t
“own” amt and it wasn’t passed to us. We could eliminate showBalance’s cou-
pling to Money with something like this:

void showBalance(BankAccount b) {
b.printBalance();
}

Prepared exclusively for Zach

 294 A PPENDIX B A NSWERS TO E XERCISES

Exercise 26: from Decoupling and the Law of Demeter on page 143
This example is also in Java.

public class Colada {

private Blender myBlender;
private Vector myStuff;
public Colada() {
myBlender = new Blender();
myStuff = new Vector();
}
private void doSomething() {
myBlender.addIngredients(myStuff.elements());
}
}

Answer 26: Since Colada creates and owns both myBlender and myStuff,
the calls to addIngredients and elements are allowed.

Exercise 27: from Decoupling and the Law of Demeter on page 143
This example is in C++.

void processTransaction(BankAccount acct, int) {

Person *who;
Money amt;
amt.setValue(123.45);
acct.setBalance(amt);
who = acct.getOwner();
markWorkflow(who->name(), SET_BALANCE);
}

Answer 27: In this case, processTransaction owns amt—it is created on

the stack. acct is passed in, so both setValue and setBalance are allowed.
But processTransaction does not own who, so the call who->name() is in
violation. The Law of Demeter suggests replacing this line with

markWorkflow(acct.name(), SET_BALANCE);

The code in processTransaction should not have to know which subobject

within a BankAccount holds the name—this structural knowledge should not
show through BankAccount’s contract. Instead, we ask the BankAccount for
the name on the account. It knows where it keeps the name (maybe in a
Person, in a Business, or in a polymorphic Customer object).

Prepared exclusively for Zach

 295

Exercise 28: from Metaprogramming on page 149

Which of the following things would be better represented as code within a
program, and which externally as metadata?

1. Communication port assignments

2. An editor’s support for highlighting the syntax of various languages
3. An editor’s support for different graphic devices
4. A state machine for a parser or scanner
5. Sample values and results for use in unit testing

Answer 28: There are no definitive answers here—the questions were

intended primarily to give you food for thought. However, this is what we think:

1. Communication port assignments. Clearly, this information should be

stored as metadata. But to what level of detail? Some Windows communi-
cations programs let you select only baud rate and port (say COM1 to COM4).
But perhaps you need to specify word size, parity, stop bits, and the duplex
setting as well. Try to allow the finest level of detail where practical.
2. An editor’s support for highlighting the syntax of various languages. This
should be implemented as metadata. You wouldn’t want to have to hack
code just because the latest version of Java introduced a new keyword.
3. An editor’s support for different graphic devices. This would probably be
difficult to implement strictly as metadata. You would not want to burden
your application with multiple device drivers only to select one at runtime.
You could, however, use metadata to specify the name of the driver and
dynamically load the code. This is another good argument for keeping the
metadata in a human-readable format; if you use the program to set up a
dysfunctional video driver, you may not be able to use the program to set
it back.
4. A state machine for a parser or scanner. This depends on what you are
parsing or scanning. If you are parsing some data that is rigidly defined
by a standards body and is unlikely to change without an act of Congress,
then hard coding it is fine. But if you are faced with a more volatile situa-
tion, it may be beneficial to define the state tables externally.
5. Sample values and results for use in unit testing. Most applications
define these values inline in the testing harness, but you can get better
flexibility by moving the test data—and the definition of the acceptable
results—out of the code itself.

Prepared exclusively for Zach

 296 A PPENDIX B A NSWERS TO E XERCISES

Exercise 29: from It’s Just a View on page 164

Suppose you have an airline reservation system that includes the concept of a
flight:

public interface Flight {

// Return false if flight full.
public boolean addPassenger(Passenger p);
public void addToWaitList(Passenger p);
public int getFlightCapacity();
public int getNumPassengers();
}

If you add a passenger to the wait list, they’ll be put on the flight automatically
when an opening becomes available.

There’s a massive reporting job that goes through looking for overbooked or full
flights to suggest when additional flights might be scheduled. It works fine, but
it takes hours to run.

We’d like to have a little more flexibility in processing wait-list passengers, and
we’ve got to do something about that big report—it takes too long to run. Use
the ideas from this section to redesign this interface.

Answer 29: We’ll take Flight and add some additional methods for main-
taining two lists of listeners: one for wait-list notification, and the other for
full-flight notification.

public interface Passenger {

public void waitListAvailable();
}
public interface Flight {
...
public void addWaitListListener(Passenger p);
public void removeWaitListListener(Passenger p);
public void addFullListener(FullListener b);
public void removeFullListener(FullListener b);
...
}
public interface BigReport extends FullListener {
public void FlightFullAlert(Flight f);
}

If we try to add a Passenger and fail because the flight is full, we can, option-
ally, put the Passenger on the wait list. When a spot opens up, waitList-
Available will be called. This method can then choose to add the Passenger
automatically, or have a service representative call the customer to ask if they
are still interested, or whatever. We now have the flexibility to perform different
behaviors on a per-customer basis.

Next, we want to avoid having the BigReport troll through tons of records look-
ing for full flights. By having BigReport registered as a listener on Flights,

Prepared exclusively for Zach

 297

each individual Flight can report when it is full—or nearly full, if we want.
Now users can get live, up-to-the-minute reports from BigReport instantly,
without waiting hours for it to run as it did previously.

Exercise 30: from Blackboards on page 170

For each of the following applications, would a blackboard system be appropri-
ate or not? Why?

1. Image processing. You’d like to have a number of parallel processes grab

chunks of an image, process them, and put the completed chunk back.
2. Group calendaring. You’ve got people scattered across the globe, in dif-
ferent time zones, and speaking different languages, trying to schedule a
meeting.
3. Network monitoring tool. The system gathers performance statistics and
collects trouble reports. You’d like to implement some agents to use this
information to look for trouble in the system.

Answer 30:
1. Image processing. For simple scheduling of a workload among the paral-
lel processes, a shared work queue may be more than adequate. You might
want to consider a blackboard system if there is feedback involved—that
is, if the results of one processed chunk affect other chunks, as in machine
vision applications, or complex 3D image-warp transforms.
2. Group calendaring. This might be a good fit. You can post scheduled
meetings and availability to the blackboard. You have entities functioning
autonomously, feedback from decisions is important, and participants may
come and go.
You might want to consider partitioning this kind of blackboard system
depending on who is searching: junior staff may care about only the im-
mediate office, human resources may want only English-speaking offices
worldwide, and the CEO may want the whole enchilada.
There is also some flexibility on data formats: we are free to ignore formats
or languages we don’t understand. We have to understand different for-
mats only for those offices that have meetings with each other, and we do
not need to expose all participants to a full transitive closure of all possi-
ble formats. This reduces coupling to where it is necessary, and does not
constrain us artificially.
3. Network monitoring tool. This is very similar to the mortgage/loan appli-
cation program described on page 168. You’ve got trouble reports sent in
by users and statistics reported automatically, all posting to the black-
board. A human or software agent can analyze the blackboard to diagnose

Prepared exclusively for Zach

 298 A PPENDIX B A NSWERS TO E XERCISES

network failures: two errors on a line might just be cosmic rays, but 20,000
errors and you’ve got a hardware problem. Just as the detectives solve the
murder mystery, you can have multiple entities analyzing and contributing
ideas to solve the network problems.

Exercise 31: from Programming by Coincidence on page 176

Can you identify some coincidences in the following C code fragment? Assume
that this code is buried deep in a library routine.

fprintf(stderr,"Error, continue?");
gets(buf);

Answer 31: There are several potential problems with this code. First, it
assumes a tty environment. That may be fine if the assumption is true, but
what if this code is called from a GUI environment where neither stderr nor
stdin is open?

Second, there is the problematic gets, which will write as many characters
as it receives into the buffer passed in. Malicious users have used this failing
to create buffer overrun security holes in many different systems. Never use
gets().

Third, the code assumes the user understands English.

Finally, no one in their right mind would ever bury user interaction such as
this in a library routine.

Exercise 32: from Programming by Coincidence on page 176

This piece of C code might work some of the time, on some machines. Then
again, it might not. What’s wrong?

/* Truncate string to its last maxlen chars */

void string_tail(char *string, int maxlen) {
int len = strlen(string);
if (len > maxlen) {
strcpy(string, string + (len - maxlen));
}
}

Answer 32: POSIX strcpy isn’t guaranteed to work for overlapping strings.
It might happen to work on some architectures, but only by coincidence.

Prepared exclusively for Zach

 299

Exercise 33: from Programming by Coincidence on page 177

This code comes from a general-purpose Java tracing suite. The function writes
a string to a log file. It passes its unit test, but fails when one of the Web
developers uses it. What coincidence does it rely on?

public static void debug(String s) throws IOException {

FileWriter fw = new FileWriter("debug.log", true);
fw.write(s);
fw.flush();
fw.close();
}

Answer 33: It won’t work in an applet context with security restrictions

against writing to the local disk. Again, when you have a choice of running
in GUI contexts or not, you may want to check dynamically to see what the
current environment is like. In this case, you may want to put a log file some-
where other than the local disk if it isn’t accessible.

Exercise 34: from Algorithm Speed on page 183

We have coded a set of simple sort routines, which can be downloaded from our
Web site (www.pragmaticprogrammer.com). Run them on various machines
available to you. Do your figures follow the expected curves? What can you
deduce about the relative speeds of your machines? What are the effects of
various compiler optimization settings? Is the radix sort indeed linear?

Answer 34: Clearly, we can’t give any absolute answers to this exercise. How-
ever, we can give you a couple of pointers.

If you find that your results don’t follow a smooth curve, you might want to
check to see if some other activity is using some of your processor’s power.
You probably won’t get good figures on a multiuser system, and even if you
are the only user you may find that background processes periodically take
cycles away from your programs. You might also want to check memory: if the
application starts using swap space, performance will nose dive.

It is interesting to experiment with different compilers and different optimiza-

tion settings. We found some that pretty startling speed-ups were possible by
enabling aggressive optimization. We also found that on the wider RISC archi-
tectures the manufacturer’s compilers often outperformed the more portable
GCC. Presumably, the manufacturer is privy to the secrets of efficient code
generation on these machines.

Prepared exclusively for Zach

 300 A PPENDIX B A NSWERS TO E XERCISES

Exercise 35: from Algorithm Speed on page 183

The routine below prints out the contents of a binary tree. Assuming the tree
is balanced, roughly how much stack space will the routine use while print-
ing a tree of 1,000,000 elements? (Assume that subroutine calls impose no
significant stack overhead.)

void printTree(const Node *node) {

char buffer[1000];
if (node) {
printTree(node->left);
getNodeAsString(node, buffer);
puts(buffer);
printTree(node->right);
}
}

Answer 35: The printTree routine uses about 1,000 bytes of stack space
for the buffer variable. It calls itself recursively to descend through the tree,
and each nested call adds another 1,000 bytes to the stack. It also calls itself
when it gets to the leaf nodes, but exits immediately when it discovers that the
pointer passed in is NULL. If the depth of the tree is , the maximum stack
requirement is therefore roughly .

A balanced binary tree holds twice as many elements at each level. A tree of
depth holds , or , elements. Our million-element
tree will therefore need , or 20 levels.

We’d therefore expect our routine to use roughly 21,000 bytes of stack.

Exercise 36: from Algorithm Speed on page 183

Can you see any way to reduce the stack requirements of the routine in Exer-
cise 35 (apart from reducing the size of the buffer)?

Answer 36: A couple of optimizations come to mind. First, the printTree

routine calls itself on leaf nodes, only to exit because there are no children.
That call increases the maximum stack depth by about 1,000 bytes. We can
also eliminate the tail recursion (the second recursive call), although this won’t
affect the worst-case stack usage.

while (node) {
if (node->left) printTree(node->left);
getNodeAsString(node, buffer);
puts(buffer);
node = node->right;
}

Prepared exclusively for Zach

 301

The biggest gain, however, comes from allocating just a single buffer, shared
by all invocations of printTree. Pass this buffer as a parameter to the recur-
sive calls, and only 1,000 bytes will be allocated, regardless of the depth of
recursion.
void printTreePrivate(const Node *node, char *buffer) {
if (node) {
printTreePrivate(node->left, buffer);
getNodeAsString(node, buffer);
puts(buffer);
printTreePrivate(node->right, buffer);
}
}
void newPrintTree(const Node *node) {
char buffer[1000];
printTreePrivate(node, buffer);
}

Exercise 37: from Algorithm Speed on page 183

On page 180, we claimed that a binary chop is . Can you prove this?

Answer 37: There are a couple of ways of getting there. One is to turn the
problem on its head. If the array has just one element, we don’t iterate around
the loop. Each additional iteration doubles the size of the array we can search.
The general formula for the array size is therefore , where is the
number of iterations. If you take logs to the base 2 of each side, you get
, which by the definition of logs becomes .

Exercise 38: from Refactoring on page 188

The following code has obviously been updated several times over the years,
but the changes haven’t improved its structure. Refactor it.
if (state == TEXAS) {
rate = TX_RATE;
amt = base * TX_RATE;
calc = 2*basis(amt) + extra(amt)*1.05;
}
else if ((state == OHIO) || (state == MAINE)) {
rate = (state == OHIO) ? OH_RATE : ME_RATE;
amt = base * rate;
calc = 2*basis(amt) + extra(amt)*1.05;
if (state == OHIO)
points = 2;
}
else {
rate = 1;
amt = base;
calc = 2*basis(amt) + extra(amt)*1.05;
}

Prepared exclusively for Zach

 302 A PPENDIX B A NSWERS TO E XERCISES

Answer 38: We might suggest a fairly mild restructuring here: make sure
that every test is performed just once, and make all the calculations common.
If the expression 2*basis(...)*1.05 appears in other places in the program,
we should probably make it a function. We haven’t bothered here.

We’ve added a rate_lookup array, initialized so that entries other than Texas,
Ohio, and Maine have a value of 1. This approach makes it easy to add values
for other states in the future. Depending on the expected usage pattern, we
might want to make the points field an array lookup as well.

rate = rate_lookup[state];
amt = base * rate;
calc = 2*basis(amt) + extra(amt)*1.05;
if (state == OHIO)
points = 2;

Exercise 39: from Refactoring on page 188

The following Java class needs to support a few more shapes. Refactor the class
to prepare it for the additions.

public class Shape {

public static final int SQUARE = 1;
public static final int CIRCLE = 2;
public static final int RIGHT_TRIANGLE = 3;
private int shapeType;
private double size;
public Shape(int shapeType, double size) {
this.shapeType = shapeType;
this.size = size;
}
// ... other methods ...

public double area() {

switch (shapeType) {
case SQUARE: return size*size;
case CIRCLE: return Math.PI*size*size/4.0;
case RIGHT_TRIANGLE: return size*size/2.0;
}
return 0;
}
}

Prepared exclusively for Zach

 303

Answer 39: When you see someone using enumerated types (or their equiv-
alent in Java) to distinguish between variants of a type, you can often improve
the code by subclassing:
public class Shape {
private double size;
public Shape(double size) {
this.size = size;
}
public double getSize() { return size; }
}
public class Square extends Shape {
public Square(double size) {
super(size);
}
public double area() {
double size = getSize();
return size*size;
}
}
public class Circle extends Shape {
public Circle(double size) {
super(size);
}
public double area() {
double size = getSize();
return Math.PI*size*size/4.0;
}
}
// etc...

Exercise 40: from Refactoring on page 189

This Java code is part of a framework that will be used throughout your project.
Refactor it to be more general and easier to extend in the future.
public class Window {
public Window(int width, int height) { ... }
public void setSize(int width, int height) { ... }
public boolean overlaps(Window w) { ... }
public int getArea() { ... }
}

Answer 40: This case is interesting. At first sight, it seems reasonable that a
window should have a width and a height. However, consider the future. Let’s
imagine that we want to support arbitrarily shaped windows (which will be
difficult if the Window class knows all about rectangles and their properties).

We’d suggest abstracting the shape of the window out of the Window class itself.

Prepared exclusively for Zach

 304 A PPENDIX B A NSWERS TO E XERCISES

public abstract class Shape {

// ...
public abstract boolean overlaps(Shape s);
public abstract int getArea();
}
public class Window {
private Shape shape;
public Window(Shape shape) {
this.shape = shape;
...
}
public void setShape(Shape shape) {
this.shape = shape;
...
}
public boolean overlaps(Window w) {
return shape.overlaps(w.shape);
}
public int getArea() {
return shape.getArea();
}
}

Note that in this approach we’ve used delegation rather than subclassing: a
window is not a “kind-of” shape—a window “has-a” shape. It uses a shape to
do its job. You’ll often find delegation useful when refactoring.

We could also have extended this example by introducing a Java interface that
specified the methods a class must support to support the shape functions.
This is a good idea. It means that when you extend the concept of a shape, the
compiler will warn you about classes that you have affected. We recommend
using interfaces this way when you delegate all the functions of some other
class.

Exercise 41: from Code That’s Easy to Test on page 197

Design a test jig for the blender interface described in the answer to Exercise
17 on page 289. Write a shell script that will perform a regression test for the
blender. You need to test basic functionality, error and boundary conditions,
and any contractual obligations. What restrictions are placed on changing the
speed? Are they being honored?

Prepared exclusively for Zach

 305

Answer 41: First, we’ll add a main to act as a unit test driver. It will accept
a very small, simple language as an argument: “E” to empty the blender, “F” to
fill it, digits 0-9 to set the speed, and so on.

public static void main(String args[]) {

// Create the blender to test
dbc_ex blender = new dbc_ex();
// And test it according to the string on standard input
try {
int a;
char c;
while ((a = System.in.read()) != -1) {
c = (char)a;
if (Character.isWhitespace(c)) {
continue;
}
if (Character.isDigit(c)) {
blender.setSpeed(Character.digit(c, 10));
}
else {
switch (c) {
case ’F’: blender.fill();
break;
case ’E’: blender.empty();
break;
case ’s’: System.out.println("SPEED: " +
blender.getSpeed());
break;
case ’f’: System.out.println("FULL " +
blender.isFull());
break;
default: throw new RuntimeException(
"Unknown Test directive");
}
}
}
}
catch (java.io.IOException e) {
System.err.println("Test jig failed: " + e.getMessage());
}
System.err.println("Completed blending n");
System.exit(0);
}

Next comes the shell script to drive the tests.

Prepared exclusively for Zach

 306 A PPENDIX B A NSWERS TO E XERCISES

#!/bin/sh
CMD="java dbc.dbc_ex"
failcount=0
expect_okay() {
if echo "$*" | $CMD #>/dev/null 2>&1
then
:
else
echo "FAILED! $*"
failcount=‘expr $failcount + 1‘
fi
}
expect_fail() {
if echo "$*" | $CMD >/dev/null 2>&1
then
echo "FAILED! (Should have failed): $*"
failcount=‘expr $failcount + 1‘
fi
}
report() {
if [ $failcount -gt 0 ]
then
echo -e " n n*** FAILED $failcount TESTS n"
exit 1 # In case we are part of something larger
else
exit 0 # In case we are part of something larger
fi
}
#
# Start the tests
#
expect_okay F123456789876543210E # Should run thru
expect_fail F5 # Fails, speed too high
expect_fail 1 # Fails, empty
expect_fail F10E1 # Fails, empty
expect_fail F1238 # Fails, skips
expect_okay FE # Never turn on
expect_fail F1E # Emptying while running
expect_okay F10E # Should be ok
report # Report results

The tests check to see if illegal speed changes are detected, if you try to empty
the blender while running, and so on. We put this in the makefile so we can
compile and run the regression test by simply typing

% make
% make test

Note that we have the test exit with 0 or 1 so we can use this as part of a larger
test as well.

There was nothing in the requirements that spoke of driving this component
via a script, or even using a language. End users will never see it. But we have
a powerful tool that we can use to test our code, quickly and exhaustively.

Prepared exclusively for Zach

 307

Exercise 42: from The Requirements Pit on page 211

Which of the following are probably genuine requirements? Restate those that
are not to make them more useful (if possible).

1. The response time must be less than 500 ms.

2. Dialog boxes will have a gray background.
3. The application will be organized as a number of front-end processes and
a back-end server.
4. If a user enters non-numeric characters in a numeric field, the system will
beep and not accept them.
5. The application code and data must fit within 256kB.

Answer 42:
1. This statement sounds like a real requirement: there may be constraints
placed on the application by its environment.
2. Even though this may be a corporate standard, it isn’t a requirement. It
would be better stated as “The dialog background must be configurable
by the end user. As shipped, the color will be gray.” Even better would be
the broader statement “All visual elements of the application (colors, fonts,
and languages) must be configurable by the end user.”
3. This statement is not a requirement, it’s architecture. When faced with
something like this, you have to dig deep to find out what the user is
thinking.
4. The underlying requirement is probably something closer to “The system
will prevent the user from making invalid entries in fields, and will warn
the user when these entries are made.”
5. This statement is probably a hard requirement.

A solution to the Four Posts

puzzle posed on page 213.

Prepared exclusively for Zach

 This page intentionally left blank

Prepared exclusively for Zach

 Index
A turning off, 123
Association for Computing Machinery
Accessor function, 31
(ACM), 262
ACM, see Association for Computing
Communications of the ACM, 263
Machinery
SIGPLAN, 263
Active code generator, 104
Assumptions, testing, 175
Activity diagram, 150
“at” command, 231
Advanced C++ Programming Styles and
Audience, 21
Idioms, 265
needs, 19
Advanced Programming in the Unix
auto_ptr, 134
Environment, 264
Automation, 230
Aegis transaction-based configuration
approval procedures, 235
management, 246, 271
build, 88, 233
Agent, 76, 117, 297
compiling, 232
Algorithm
cron, 231
binary chop, 180
documentation, 251
choosing, 182
scripts, 234
combinatoric, 180
team, 229
divide-and-conquer, 180
testing, 29, 238
estimating, 177, 178
Web site generation, 235
linear, 177
awk, 99
notation, 178, 181
quicksort, 180
runtime, 181 B
sublinear, 177 Backus-Naur Form (BNF), 59n
Allocations, nesting, 131 Base class, 112
Analysis Patterns, 264 bash shell, 80, 82n
Anonymity, 258 Bean, see Enterprise Java Beans (EJB)
AOP, see Aspect-Oriented Programming Beck, Kent, 194, 258
Architecture Beowulf project, 268
deployment, 156 “Big ” notation, 177
flexibility, 46 “Big picture”, 8
prototyping, 55 Binary chop, 97, 180
temporal decoupling, 152 Binary format, 73
Art of Computer Programming, 183 problems parsing, 75
Artificial intelligence, marauding, 26 bison, 59, 269
Aspect-Oriented Programming (AOP), BIST, see Built-In Self Test
39, 273 Blackboard system, 165
Assertion, 113, 122, 175 partitioning, 168
side effects, 124 workflow, 169

309

Prepared exclusively for Zach

 310 I NDEX

Blender example Call, routine, 115, 173

contract for, 119, 289 Cascading Style Sheets (CSS), 253
regression test jig, 305 Cat
workflow, 151 blaming, 3
BNF, see Backus-Naur Form (BNF) herding, 224
Boiled frog, 8, 175, 225 Schrödinger’s, 47
Boundary condition, 173, 243 Catalyzing change, 8
Brain, Marshall, 265 Cathedrals, xx
Branding, 226 Cetus links, 265
Brant, John, 268 Change, catalyzing, 8
“Broken Window Theory”, 5 Christiansen, Tom, 81
vs. stone soup, 9 Class
Brooks, Fred, 264 assertions, 113
Browser, class, 187 base, 112
Browser, refactoring, 187, 268 coupling, 139, 142
Bug, 90 coupling ratios, 242
failed contract as, 111 encapsulating resource, 132
see also Debugging; Error invariant, 110, 113
Build number of states, 245
automation, 88, 233 resource allocation, 132
dependencies, 233 subclass, 112
final, 234 wrapper, 132, 133, 135, 141
nightly, 231 Class browser, 187
refactoring, 187 ClearCase, 271
Built-In Self Test (BIST), 189 Cockburn, Alistair, xxiii, 205, 264, 272
Business logic, 146 Code generator, 28, 102
Business policy, 203 active, 104
makefiles, 232
C parsers, 105
passive, 103
C language
assertions, 122 Code profiler, 182
DBC, 114 Code reviews, 33, 236
duplication, 29 Coding
error handling, 121 algorithm speed, 177
error messages, 115 comments, 29, 249
macros, 121 coupled, 130
Object Pascal interface, 101 coverage analysis, 245
C++ language, 46 database schema, 104
assertions, 122 defensive, 107
auto_ptr, 134 and documentation, 29, 248
books, 265 estimating, 68
DBC, 114 exceptions, 125
decoupling, 142 implementation, 173
DOC++, 251, 269 iterative, 69
duplication, 29 “lazy”, 111
error messages, 115 metrics, 242
exceptions, 132 modules, 138
unit tests, 193 multiple representations, 28
Caching, 31 orthogonality, 34, 36, 40

Prepared exclusively for Zach

 I NDEX 311

ownership, 258 Concurrency, 150

prototypes, 55 design, 154
server code, 196 interfaces, 155
“shy”, 40, 138 and Programming by Coincidence,
specifications, 219 154
tracer bullets, 49–51 requirements analysis of, 150
unit testing, 190, 192 workflow, 150
see also Coupled code; Decoupled Concurrent Version System (CVS), 271
code; Metadata; Source code Configuration
control system (SCCS) cooperative, 148
Cohesion, 35 dynamic, 144
COM, see Component Object Model metadata, 147
Combinatorial explosion, 140, 167 Configuration management, 86, 271
Combinatoric algorithm, 180 Constantine, Larry L., 35
Command shell, 77 Constraint management, 213
bash, 80 Constructor, 132
Cygwin, 80 initialization, 155
vs. GUI, 78 Contact, authors’ e-mail, xxiii
UWIN, 81 Context, use instead of globals, 40
Windows, 80 Contract, 109, 174
Comment, 29, 249 see also Design by contract (DBC)
avoiding duplication, 29 Controller (MVC), 162
DBC, 113 Coplien, Jim, 265
parameters, 250 CORBA, see Common Object Request
types of, 249 Broker
unnecessary, 250 Coupled code, 130
see also Documentation coupling ratios, 242
Common Object Request Broker minimizing, 138, 158
(CORBA), 29, 39, 46 performance, 142
Event Service, 160 temporal coupling, 150
Communicating, 18 see also Decoupled code
audience, 19, 21 Coverage analysis, 245
duplication, 32 Cox, Brad J., 189n
Crash, 120
e-mail, 22
Critical thinking, 16
and formal methods, 221
cron, 231
presentation, 20
CSS, see Cascading Style Sheets
style, 20
CVS, see Concurrent Version System
teams, 225
Cygwin, 80, 270
users, 256
writing, 18
Communications of the ACM, 263
Comp.object FAQ, 272
D
Data
Compiling, 232 blackboard system, 169
compilers, 267 caching, 31
DBC, 113 dictionary, 144
warnings and debugging, 92 dynamic data structures, 135
Component Object Model (COM), 55 global, 40
Component-based systems, see language, 60
Modular system normalizing, 30

Prepared exclusively for Zach

 312 I NDEX

readable vs. understandable, 75 Dependency, reducing, see Modular

test, 100, 243 system; Orthogonality
views, 160 Deployment, 156
visualizing, 93 Deployment descriptor, 148
see also Metadata Design
Data Display Debugger (DDD), 93, 268 accessor functions, 31
Database concurrency, 154
active code generator, 104 context, 174
schema, 105f, 141, 144 deployment, 156
schema maintenance, 100 design/methodology testing, 242
DBC, see Design by contract metadata, 145
DDD, see Data Display Debugger orthogonality, 34, 37
Deadline, 6, 246 physical, 142
Deadlock, 131 refactoring, 186
Debugging, 90 using services, 154
assertions, 123 Design by contract (DBC), 109, 155
binary search, 97 and agents, 117
bug location, 96 assertions, 113
bug reproduction, 93 class invariant, 110
checklist, 98 as comments, 113
compiler warnings and, 92 dynamic contracts, 117
corrupt variables, 95 iContract, 268
“Heisenbug”, 124 language support, 114
rubber ducking, 95 list insertion example, 110
and source code branching, 87 pre- and postcondition, 110, 113,
surprise bug, 97 114
and testing, 92, 195 predicates, 110
time bomb, 192 unit testing, 190
tracing, 94 Design Patterns, 264
view, 164 observer, 158
visualizing data, 93 singleton, 41
Decision making, 46 strategy, 41
Decoupled code, 38, 40 Destructor, 132
architecture, 152 Detectives, 165
blackboard system, 166 Development tree, 87
Law of Demeter, 140 Development, iterative, 69
metadata, 145 Divide-and-conquer algorithm, 180
minimizing coupling, 138 DOC++ documentation generator, 251,
modular testing, 244 269
physical decoupling, 142 DocBook, 254
temporal coupling, 150 Documentation
workflow, 150 automatic updating, 251
see also Coupled code and code, 29, 248
Defensive coding, 107 comments, 29, 113, 249, 251
Delegation, 304 executable, 251
Delphi, 55 formats, 253
see also Object Pascal HTML, 101
Demeter project, 274 hypertext, 210
Demeter, Law of, 140 internal/external, 248

Prepared exclusively for Zach

 I NDEX 313

invariant, 117 Effective C++, 265

mark-up languages, 254 Eiffel, 109, 114, 267
orthogonality, 42 EJB, see Enterprise Java Beans
outline, 18 elvis editor, 267
requirements, 204 Emacs editor, 84, 266
technical writers, 252 Viper vi emulator, 267
word processors, 252, 254 Embedded mini-language, 62, 145
writing specifications, 218 Embellishment, 11
see also Comment; Web Encapsulation, object, 127, 158
documentation Eno, Brian, 205
Dodo, 148 Enterprise Java Beans (EJB), 39, 147
Domain, problem, 58, 66 Entropy, 4
Don’t repeat yourself, see DRY
Error
principle
DBC messages, 115
Downloading source code, see Example
design, 30
code
domain-specific, 59
Dr. Dobbs Journal, 263
early crash, 120
DRY principle, 27, 29, 42
log messages, 196
see also Duplication
orthogonality, 41
Duck, rubber, see Rubber duck
testing, 240, 247
Dumpty, Humpty, xxii, 165
Duplication, 26 see also Exception
code generators avoid, 28 Error handler, 127
and code reviews, 33 Estimating, 64
design errors, 30 accuracy, 64
documentation and code, 29 algorithms, 177, 178
DRY principle, 27, 29 iterative, 69
interdeveloper, 32 models, 66
in languages, 29 problem domain, 66
multiple representations, 28 project schedules, 68
teams, 226 records, 68
under time pressure, 32 testing, 182
types of, 27 Eton College, xxi
Dynamic configuration, 144 Event, 157
Dynamic data structure, 135 Event channel, 160
Dynamics of Software Development, 264 Example code
add logging, 40
E airline reservations, 164, 296
assert macro, 122
E-mail, 22
address for feedback, xxiii auto_ptr example, 134

Editor, 82 bad resource balancing, 129, 130

auto-indenting, 85 downloading, xxiii
cursor movement, 84 exception error handling, 125
features, 83 good resource balancing, 131
generating code, 103 JavaDoc example, 250
how many to learn, 82 method chaining, 139
template, 84 normalized class, 31
types of, 266 open password file, 126
Windows notepad, 84 open user file, 127

Prepared exclusively for Zach

 314 I NDEX

resources and exceptions, 132, log, 196

133 makefile, 232
side effect, 124 source, 103
spaghetti error handling, 125 Final build, 234
square root, 190 Fish, dangers of, 34
string parsing with Flexibility, 46
StringTokenizer, 156 Formal methods, 220, 221
string parsing with strtok, 155 Four Posts Puzzle, 213
unnormalized class, 30 Fowler, Martin, xxiii, 186, 273
Example code by name Free Software Foundation, see GNU
AOP, 40 Project
Misc.java, 156 Frog, boiled, see Boiled frog
assert, 122 Function
bad_balance.c, 129, 130 accessor, 31
balance.cc, 134 Law of Demeter for s, 140
balance.c, 131–133 similar, 41
class Line, 30, 31
exception, 125
findPeak, 250 G
interface Flight, 164, 296 Gamma, Erich, 194
misc.c, 155 Garbage collection, 134
openpasswd.java, 126 Gardening metaphor, 184
openuserfile.java, 127 Gehrke, Peter, xxiv
plotDate, 139 Glass, Robert, 221, 236
side_effect, 124 Global variables, 40, 130, 154
spaghetti, 125 Glossary, project, 210
sqrt, 190 GNU Project, 274
Exception, 121 C/C++ compiler, 267
effects of, 127 General Public License (GPL), 80
and error handlers, 127 GNU Image Manipulation Program
missing files, 126 (GIMP), 274
resource balancing, 132 SmallEiffel, 267
when to use, 125 “Good-enough software”, see Software,
Excuses, 3 quality
Executable document, 251 Gordian knot, 212
expect, 269 Goto, 127
Expert, see Guru GUI system
Expiring asset, 12 vs. command shell, 78
eXtensible Style Language (XSL), 253 interface, 78
Extinction, 148 testing, 244
eXtreme Programming, 238n, 258, 272 Guru, 17, 198

F H
Feature creep, 10 Hash, secure, 74
Feedback, e-mail address, xxiii Header file, 29
File “Heisenbug”, 124, 289
exception, 126 Helicopter, 34n
header, 29 Hopper, Grace, 8n, 90
implementation, 29 “Hot-key” sequence, 196

Prepared exclusively for Zach

 I NDEX 315

HTTP Web server, 196

Human factors, 241
J
Jacobson, Ivar, 204
Humpty Dumpty, xxii, 165
Jargon, xxii, 210
Hungarian notation, 249
Jargon file, 273
Hungry consumer model, 153 Java, 46, 267
Hypertext document, 210 code generation, 232
DBC, 114

I Enterprise Java Beans, 39, 147

error messages, 115
iContract, 110, 114, 268 exceptions, 121
IDE, see Integrated Development iContract, 110, 114, 268
Environment javaCC, 59, 269
IEEE Computer Society, 262 JavaDoc, 248, 251
IEEE Computer, 262 JavaSpaces, 166, 273
IEEE Software, 263 JUnit, 195
Imperative language, 60 multithreaded programming, 154
Implementation property access, 100
accidents, 173 property files, 145
coding, 173 resource balancing, 134
specifications, 219 RMI, 128
Imposed duplication, 28 string parser, 156
Inadvertent duplication, 30 tree view, 161
Indentation, automatic, 85 unit tests, 193
Independence, see Orthogonality and Windows shells, 81
Infrastructure, 37 JavaDoc, see Java
Inheritance, 111
assertions, 113
fan-in/fan-out, 242
K
K Desktop Environment, 273
Inner tennis, 215
Kaizen, xxi, 14
Inspection, code, see Code reviews
see also Knowledge portfolio
Insure++, 136
Kernighan, Brian, 99
Integrated circuit, 189n
Keybinding, 82
Integrated Development Environment
Kirk, James T., 26
(IDE), 72, 232
Knowledge
Integration platform, 50
producers and consumers, 166
Integration testing, 239
Knowledge portfolio, 12
Interface
building, 13
blackboard system, 168
critical thinking, 16
C/Object Pascal, 101
learning and reading, 14
concurrency, 155
researching, 15
error handler, 128
Knuth, Donald, 183, 248
GUI, 78
Korn, David, 81
prototyping, 55
Kramer, Reto, xxiv
user, 203
Kruchten, Phillipe, 227n
Invariant, 110, 113, 155
loop, 116
semantic, 116, 135 L
ISO9660 format, 233n Lakos, John, xxiv, 9, 142, 265
Iterative development, 69 Lame excuses, 3

Prepared exclusively for Zach

 316 I NDEX

Language, programming McCabe Cyclomatic Complexity Metric,

conversions, 103, 105 242
DBC, 114 Member variables, see Accessor
domain, 57 functions
duplication in, 29 Memory allocation, 135
learning, 14 Metadata, 144, 203
prototypes, 55 business logic, 146
scripting, 55, 145 configuration, 147
specification, 58, 62 controlling transactions, 39
text manipulation, 99 decoupled code, 145
see also Mini-language and formal methods, 221
Large-Scale C++ Software Design, 142, in plain text, 74
265 Metric, 242
LATEX system, 103 Meyer, Bertrand, 31n, 109, 184, 264
Law of Demeter, 140 Meyer, Scott, 265
Lawns, care of, xxi Microsoft Visual C++, 198
Layered design, 37 Microsoft Windows, 46
Layered system, see Modular system Mini-language, 59
“lazy” code, 111 data language, 60
Lex and Yacc, 59 embedded, 62
Librarian, see Project librarian imperative, 60
Library code, 39 parsing, 62
Linda model, 167 stand-alone, 62
Linear algorithms, 177 Mixing board, 205
Linux, 15, 254, 265 MKS Source Integrity, 271
Liskov Substitution Principle, 111 Model, 160
Listening, 21 calculations, 67
Literate programming, 248 components and parameters, 66
Logging, 39, 196 and estimating, 66
see also Tracing executable documents, 251
Lookup table, 104 view, 162
Loop Model-view-controller (MVC), 38, 160
nested, 180 Modular system, 37
simple, 180 coding, 138
Loop invariant, 116 prototyping, 55
resource allocation, 135
reversibility, 45
M testing, 41, 190, 244
More Effective C++, 265
Macro, 78, 86
Mozilla, 273
assertions, 122
Multithreaded programming, 154
documentation, 252
MVC, see Model-view-controller
error handling, 121
The Mythical Man Month, 264
Maintenance, 26
imperative languages, 61
Makefile, 232 N
recursive, 233 Name, variable, 249
Managing expectations, 256 Nana, 114, 268
Mark-up language, 254 Nest allocations, 131
Martin, Robert C., 273 Nested loop, 180

Prepared exclusively for Zach

 I NDEX 317

Netscape, 145, 273 Parallel programming, 150

Newsgroup, 15, 17, 33 Parrots, killer, see Branding
Nonorthogonal system, 34 Parsing, 59
Normalize, 30 code generators, 105
Novobilski, Andrew J., 189n log messages, 196
mini-language, 62

O strings, 155
Partitioning, 168
notation, 178, 181
Pascal, 29
Object
Passive code generator, 103
coupling, 140n
Performance testing, 241
destruction, 133, 134
Perl, 55, 62, 99
persistence, 39
C/Object Pascal interface, 101
publish/subscribe protocol, 158
database schema generation, 100
singleton, 41
home page, 267
valid/invalid state, 154
Java property access, 100
viewer, 163
power tools, 270
Object Management Group (OMG), 270
test data generation, 100
Object Pascal, 29
testing, 197
C interface, 101
and typesetting, 100
Object-Oriented Programming, 189n
Unix utilities in, 81
Object-Oriented Software Construction,
web documentation, 101
264
Perl Journal, 263
Obsolescence, 74
Persistence, 39, 45
OLTP, see On-Line Transaction
Petzold, Charles, 265
Processing system
Pike, Rob, 99
OMG, see Object Management Group
Pilot
On-Line Transaction Processing system
landing, handling, etc., 217
(OLTP), 152
who ate fish, 34
Options, providing, 3
Plain text, 73
Ordering, see Workflow
vs. binary format, 73
Orthogonality, 34
drawbacks, 74
coding, 34, 36, 40
executable documents, 251
design, 37
leverage, 75
documentation, 42
obsolescence, 74
DRY principle, 42
and easier testing, 76
nonorthogonal system, 34
Unix, 76
productivity, 35
Polymorphism, 111
project teams, 36, 227
Post-it note, 53, 55
testing, 41
Powerbuilder, 55
toolkits & libraries, 39
The Practice of Programming, 99
see also Modular system
Pragmatic programmer
Over embellishment, 11
characteristics, xviii
e-mail address, xxiii
P Web site, xxiii
Pain management, 185 Pre- and postcondition, 110, 113, 114
paint() method, 173 Predicate logic, 110
Painting, 11 Preprocessor, 114
Papua New Guinea, 16 Presentation, 20

Prepared exclusively for Zach

 318 I NDEX

Problem domain, 58, 66 Refactoring browser, 187, 268

metadata, 146 Refinement, excessive, 11
Problem solving, 213 Regression, 76, 197, 232, 242
checklist for, 214 Relationship
Productivity, 10, 35 has-a, 304
Programming by coincidence, 173 kind-of, 111, 304
Programming staff Releases, and SCCS, 87
expense of, 237 Remote Method Invocation (RMI), 128
Programming Windows, 265 exception handling, 39
Project Remote procedure call (RPC), 29, 39
glossary, 210 Repository, 87
“heads”, 228 Requirement, 11, 202
saboteur, 244 business problem, 203
schedules, 68 changing, 26
see also Automation; creep, 209
Team, project DBC, 110
Project librarian, 33, 226 distribution, 211
Prototyping, 53, 216 documenting, 204
architecture, 55 in domain language, 58
disposable code, 56 expressing as invariant, 116
kinds of, 54 formal methods, 220
and programming languages, 55 glossary, 210
and tracer code, 51 over specifying, 208
using, 54 and policy, 203
Publish/subscribe protocol, 158 usability testing, 241
Pugh, Greg, 95n user interface, 203
Purify, 136 Researching, 15
PVCS Configuration Management, 271 Resource balancing, 129
Python, 55, 99, 267 C++ exceptions, 132
checking, 135
Q coupled code, 130
dynamic data structures, 135
Quality
encapsulation in class, 132
control, 9
Java, 134
requirements, 11
nest allocations, 131
teams, 225
Response set, 141, 242
Quarry worker’s creed, xx
Responsibility, 2, 250, 258
Quicksort algorithm, 180
Reuse, 33, 36
Reversibility, 44
R flexible architecture, 46
Revision Control System (RCS), 250,
Rational Unified Process, 227n
Raymond, Eric S., 273 271
RCS, see Revision Control System Risk management, 13
Real-world data, 243 orthogonality, 36
Refactoring, 5, 185 RMI, see Remote Method Invocation
automatic, 187 Rock-n-roll, 47
and design, 186 RPC, see Remote procedure call
testing, 187 Rubber ducking, 3, 95
time constraints, 185 Rules engine, 169

Prepared exclusively for Zach

 I NDEX 319

S Source code control system (SCCS), 86

Aegis, 246
Saboteur, 244
builds using, 88
Samba, 272
CVS, 271
Sample programs, see Example code
development tree, 87
Sather, 114, 268
plain text and, 76
SCCS, see Source code control system
RCS, 250, 271
Schedule, project, 68
repository, 87
Schrödinger, Erwin (and his cat), 47
tools, 271
Scope, requirement, 209
Specialization, 221
Screen scraping, 61
Specification, 58
Scripting language, 55, 145
implementation, 219
Secure hash, 74
language, 62
sed, 99
as security blanket, 219
Sedgewick, Robert, 183
writing, 218
Self-contained components, see
Spy cells, 138
Orthogonality; Cohesion
Squeak, 268
Semantic invariant, 116, 135
Stand-alone mini-language, 62
sendmail program, 60
“Start-up fatigue”, 7
Sequence diagram, 158
Starting a project
Server code, 196
problem solving, 212
Services, design using, 154
prototyping, 216
Shell, command, 77
specifications, 217
vs. GUI, 78
see also Requirement
see also Command shell
Stevens, W. Richard, 264
“Shy code”, 40
Stone soup, 7
Side effect, 124
vs. broken windows, 9
SIGPLAN, 263
Stone-cutter’s creed, xx
Simple loop, 180
String parser, 155
Singleton object, 41
Stroop effect, 249
Slashdot, 265
strtok routine, 155
SmallEiffel, 267
Structured walkthroughs, see Code
Smalltalk, 46, 186, 187, 268, 272
reviews
Software
Style sheet, 20, 254
development technologies, 221
Style, communication, 20
quality, 9
Subclass, 112
requirements, 11
Sublinear algorithm, 177
Software bus, 159
Supplier, see Vendor
“Software Construction”, 184
Surviving Object-Oriented Projects: A
Software Development Magazine, 263
Manager’s Guide, 264
Software IC, 189n
SWIG, 55, 270
“Software rot”, 4
Synchronization bar, 151
Solaris, 76
Syntax highlighting, 84
Source code
Synthetic data, 243
cat eating, 3
documentation, see Comments
downloading, see Example code T
duplication in, 29 T Spaces, 166, 269
generating, 103 TAM, see Test Access Mechanism
reviews, see Code reviews Tcl, 55, 99, 269

Prepared exclusively for Zach

 320 I NDEX

Team, project, 36, 224 Trade paper, 263

automation, 229 Trade-offs, 249
avoiding duplication, 32 Transactions, EJB, 39
code review, 236 Tree widget, 161
communication, 225 troff system, 103
duplication, 226 Tuple space, 167
functionality, 227
organization, 227
pragmatism in, xx
U
UML, see Unified modeling language
quality, 225 (UML)
tool builders, 229 UNDO key, 86
Technical writer, 252 Unified modeling language (UML)
Template, use case, 205 activity diagram, 150
Temporal coupling, 150 sequence diagram, 158
Test Access Mechanism (TAM), 189 use case diagram, 208
Test harness, 194 Uniform Access Principle, 31n
Testing Unit testing, 190
automated, 238 DBC, 190
from specification, 29 modules, 239
bug fixing, 247 test harness, 194
coverage analysis, 245 test window, 196
and culture, 197 writing tests, 193
debugging, 92, 196 Unix, 46, 76
design/methodology, 242 Application Default files, 145
effectiveness, 244 books, 264
estimates, 182 Cygwin, 270
frequency, 246 DOS tools, 270
GUI systems, 244 Samba, 272
integration, 239 UWIN, 81, 270
orthogonality, 36, 41 Unix Network Programming, 264
performance, 241 Usability testing, 241
role of plain text, 76 Use case, 204
refactoring, 187 diagrams, 206
regression, 76, 197, 232, 242 Usenet newsgroup, 15, 17, 33
resource exhaustion, 240 User
saboteur, 244 expectations, 256
test data, 100, 243 groups, 18
usability, 241 interface, 203
validation and verification, 239 requirements, 10
see also Unit testing UWIN, 81, 270
Text manipulation language, 99
TOM programming language, 268
Toolkits, 39
V
Variable
Tools, adaptable, 205 corrupt, 95
Tracer code, 49 global, 130, 154
advantages of, 50 name, 249
and prototyping, 51 Vendor
Tracing, 94 libraries, 39
see also Logging reducing reliance on, 36, 39, 46

Prepared exclusively for Zach

 I NDEX 321

vi editor, 266 Unix utilities, 80, 81

View UWIN, 81
debugging, 164 WinZip, 272
executable documents, 251 WISDOM acrostic, 20
Java tree view, 161 Wizard, 198
model-view-controller, 160, 162 Word processor, 252, 254
model-viewer network, 162 Workflow, 150
vim editor, 266 blackboard system, 169
Visual Basic, 55 content-driven, 234
Visual C++, 198 Wrapper, 132, 133, 135, 141
Visual SourceSafe, 271 Writing, 18
VisualWorks, 268 see also Documentation
www.pragmaticprogrammer.com, xxiii

W WYSIWYG, see What You See Is What

You Get
Walkthoughs, see Code reviews
Warnings, compilation, 92
Web documentation, 101, 210, 253
automatic generation, 235
X
XEmacs editor, 266
news and information, 265
Xerox Parc, 39
Web server, 196
XSL, see eXtensible Style Language
Web site, pragmatic programmer, xxiii
xUnit, 194, 269
What You See Is What You Get
(WYSIWYG), 78
WikiWikiWeb, 265
Win32 System Services, 265
Y
yacc, 59
Windows, 46 Yourdon, Edward, 10, 35
“at” command, 231 Y2K problem, 32, 208
books, 265
Cygwin, 80
metadata, 145 Z
notepad, 84 Z shell, 272

Prepared exclusively for Zach


  
THE TRUSTED TECHNOLOGY LEARNING SOURCE

InformIT is a brand of Pearson and the online presence

for the world’s leading technology publishers. It’s your source
for reliable and qualified content and knowledge, providing
access to the top brands, authors, and contributors from
the tech community.


at
 
Looking for a book, eBook, or training video on a new technology? Seeking
timely and relevant information and tutorials? Looking for expert opinions,
advice, and tips? InformIT has the solution.

• Learn about new releases and special promotions by

subscribing to a wide variety of newsletters.
Visit informit.com /newsletters.

• Access FREE podcasts from experts at informit.com /podcasts.

• Read the latest author articles and sample chapters at
informit.com /articles.
• Access thousands of books and videos in the Safari Books
Online digital library at safari.informit.com.
• Get tips from expert blogs at informit.com /blogs.

Visit informit.com /learn to discover all the ways you can access the
hottest technology content.

Are You Part of the
 Crowd?

Connect with Pearson authors and editors via RSS feeds, Facebook,
Twitter, YouTube, and more! Visit informit.com /socialconnect.

  
THE TRUSTED TECHNOLOGY LEARNING SOURCE

Prepared exclusively for Zach

 THIS PRODUCT
informit.com/register

Register the Addison-Wesley, Exam Registering your products can unlock

Cram, Prentice Hall, Que, and the following benefits:
Sams products you own to unlock • Access to supplemental content,
great benefits. including bonus chapters,
source code, or project files.
To begin the registration process,
• A coupon to be used on your
simply go to informit.com/register
next purchase.
to sign in or create an account.
You will then be prompted to enter Registration benefits vary by product.
the 10- or 13-digit ISBN that appears Benefits will be listed on your Account
on the back cover of your product. page under Registered Products.

About InformIT — THE TRUSTED TECHNOLOGY LEARNING SOURCE

INFORMIT IS HOME TO THE LEADING TECHNOLOGY PUBLISHING IMPRINTS

Addison-Wesley Professional, Cisco Press, Exam Cram, IBM Press, Prentice Hall
Professional, Que, and Sams. Here you will gain access to quality and trusted content and
resources from the authors, creators, innovators, and leaders of technology. Whether you’re
looking for a book on a new technology, a helpful article, timely newsletters, or access to
the Safari Books Online digital library, InformIT has a solution for you.

 



THE TRUSTED TECHNOLOGY LEARNING SOURCE
Addison-Wesley | Cisco Press | Exam Cram
IBM Press | Que | Prentice Hall | Sams

SAFARI BOOKS ONLINE

Prepared exclusively for Zach

 The Pragmatic Programmer
This card summarizes the tips and checklists found in The Pragmatic Programmer.
For more information about T HE P RAGMATIC P ROGRAMMERS LLC, source code for the
examples, up-to-date pointers to Web resources, and an online bibliography, visit us
at www.pragmaticprogrammer.com.
Quick Reference Guide

TIPS 1 TO 22

1. Care About Your Craft . . . . . . . . . . . . . . . . . . . . . . . . . . . xix 12. Make It Easy to Reuse . . . . . . . . . . . . . . . . . . . . . . . . . . 33

Why spend your life developing software unless If it’s easy to reuse, people will. Create an environ-
you care about doing it well? ment that supports reuse.

2. Think! About Your Work . . . . . . . . . . . . . . . . . . . . . . . . . xix 13. Eliminate Effects Between Unrelated Things . . . . 35
Turn off the autopilot and take control. Constantly Design components that are self-contained, inde-
critique and appraise your work. pendent, and have a single, well-defined purpose.

3. Provide Options, Don’t Make Lame Excuses . . . . . . . . 3 14. There Are No Final Decisions . . . . . . . . . . . . . . . . . . . 46
Instead of excuses, provide options. Don’t say it No decision is cast in stone. Instead, consider each
can’t be done; explain what can be done. as being written in the sand at the beach, and plan
for change.
4. Don’t Live with Broken Windows . . . . . . . . . . . . . . . . . . .5
Fix bad designs, wrong decisions, and poor code 15. Use Tracer Bullets to Find the Target . . . . . . . . . . . 49
when you see them. Tracer bullets let you home in on your target by
trying things and seeing how close they land.
5. Be a Catalyst for Change . . . . . . . . . . . . . . . . . . . . . . . . . . .8
You can’t force change on people. Instead, show 16. Prototype to Learn . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .54
them how the future might be and help them par- Prototyping is a learning experience. Its value lies
ticipate in creating it. not in the code you produce, but in the lessons you
learn.
6. Remember the Big Picture . . . . . . . . . . . . . . . . . . . . . . . . . 8
Don’t get so engrossed in the details that you forget 17. Program Close to the Problem Domain . . . . . . . . . . 58
to check what’s happening around you. Design and code in your user’s language.

7. Make Quality a Requirements Issue . . . . . . . . . . . . . . 11 18. Estimate to Avoid Surprises . . . . . . . . . . . . . . . . . . . . 64

Involve your users in determining the project’s real Estimate before you start. You’ll spot potential
quality requirements. problems up front.

8. Invest Regularly in Your Knowledge Portfolio . . . . . 14 19. Iterate the Schedule with the Code . . . . . . . . . . . . . .69
Make learning a habit. Use experience you gain as you implement to refine
the project time scales.
9. Critically Analyze What You Read and Hear . . . . . . 16
Don’t be swayed by vendors, media hype, or 20. Keep Knowledge in Plain Text . . . . . . . . . . . . . . . . . . .74
dogma. Analyze information in terms of you and Plain text won’t become obsolete. It helps leverage
your project. your work and simplifies debugging and testing.

10. It’s Both What You Say and the Way You Say It . . 21 21. Use the Power of Command Shells . . . . . . . . . . . . . . 80
There’s no point in having great ideas if you don’t Use the shell when graphical user interfaces don’t
communicate them effectively. cut it.

11. DRY —D on’t R epeat Yourself . . . . . . . . . . . . . . . . . . . . 27 22. Use a Single Editor Well . . . . . . . . . . . . . . . . . . . . . . . . 82
Every piece of knowledge must have a single, un- The editor should be an extension of your hand;
ambiguous, authoritative representation within a make sure your editor is configurable, extensible,
system. and programmable.

1
Prepared exclusively for Zach
 TIPS 23 TO 46

23. Always Use Source Code Control . . . . . . . . . . . . . . . 88 35. Finish What You Start . . . . . . . . . . . . . . . . . . . . . . . . . 129
Source code control is a time machine for your Where possible, the routine or object that allocates
work—you can go back. a resource should be responsible for deallocating
it.
24. Fix the Problem, Not the Blame . . . . . . . . . . . . . . . . . 91
It doesn’t really matter whether the bug is your 36. Minimize Coupling Between Modules . . . . . . . . . . 140
fault or someone else’s—it is still your problem, Avoid coupling by writing “shy” code and applying
and it still needs to be fixed. the Law of Demeter.

25. Don’t Panic When Debugging . . . . . . . . . . . . . . . . . . . 91 37. Configure, Don’t Integrate . . . . . . . . . . . . . . . . . . . . . 144
Take a deep breath and THINK ! about what could Implement technology choices for an application as
be causing the bug. configuration options, not through integration or
engineering.
26. “select” Isn’t Broken . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
It is rare to find a bug in the OS or the compiler, 38. Put Abstractions in Code, Details in Metadata . 145
or even a third-party product or library. The bug is Program for the general case, and put the specifics
most likely in the application. outside the compiled code base.

27. Don’t Assume It—Prove It . . . . . . . . . . . . . . . . . . . . . . . 97 39. Analyze Workflow to Improve Concurrency . . . . 151
Prove your assumptions in the actual environ- Exploit concurrency in your user’s workflow.
ment—with real data and boundary conditions.
40. Design Using Services . . . . . . . . . . . . . . . . . . . . . . . . . 154
Design in terms of services—independent, concur-
28. Learn a Text Manipulation Language . . . . . . . . . . 100
rent objects behind well-defined, consistent inter-
You spend a large part of each day working with
faces.
text. Why not have the computer do some of it for
you?
41. Always Design for Concurrency . . . . . . . . . . . . . . . 156
Allow for concurrency, and you’ll design cleaner in-
29. Write Code That Writes Code . . . . . . . . . . . . . . . . . . 103
terfaces with fewer assumptions.
Code generators increase your productivity and
help avoid duplication.
42. Separate Views from Models . . . . . . . . . . . . . . . . . . .161
Gain flexibility at low cost by designing your appli-
30. You Can’t Write Perfect Software . . . . . . . . . . . . . . 107 cation in terms of models and views.
Software can’t be perfect. Protect your code and
users from the inevitable errors. 43. Use Blackboards to Coordinate Workflow . . . . . .169
Use blackboards to coordinate disparate facts and
31. Design with Contracts . . . . . . . . . . . . . . . . . . . . . . . . .111 agents, while maintaining independence and isola-
Use contracts to document and verify that code tion among participants.
does no more and no less than it claims to do.
44. Don’t Program by Coincidence . . . . . . . . . . . . . . . . 175
32. Crash Early . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120 Rely only on reliable things. Beware of accidental
A dead program normally does a lot less damage complexity, and don’t confuse a happy coincidence
than a crippled one. with a purposeful plan.

33. Use Assertions to Prevent the Impossible . . . . . 122 45. Estimate the Order of Your Algorithms . . . . . . . . 181
Assertions validate your assumptions. Use them Get a feel for how long things are likely to take be-
to protect your code from an uncertain world. fore you write code.

34. Use Exceptions for Exceptional Problems . . . . . 127 46. Test Your Estimates . . . . . . . . . . . . . . . . . . . . . . . . . . .182
Exceptions can suffer from all the readability and Mathematical analysis of algorithms doesn’t tell
maintainability problems of classic spaghetti code. you everything. Try timing your code in its target
Reserve exceptions for exceptional things. environment.

Prepared exclusively for Zach

 TIPS 47 TO 70

47. Refactor Early, Refactor Often . . . . . . . . . . . . . . . . . 186 59. Costly Tools Don’t Produce Better Designs . . . . 222
Just as you might weed and rearrange a gar- Beware of vendor hype, industry dogma, and the
den, rewrite, rework, and re-architect code when aura of the price tag. Judge tools on their merits.
it needs it. Fix the root of the problem.
60. Organize Teams Around Functionality . . . . . . . . 227
48. Design to Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .192 Don’t separate designers from coders, testers from
Start thinking about testing before you write a line
data modelers. Build teams the way you build
of code.
code.
49. Test Your Software, or Your Users Will . . . . . . . . .197
Test ruthlessly. Don’t make your users find bugs 61. Don’t Use Manual Procedures . . . . . . . . . . . . . . . . . 231
for you. A shell script or batch file will execute the same
instructions, in the same order, time after time.
50. Don’t Use Wizard Code You Don’t Understand . 199
Wizards can generate reams of code. Make sure 62. Test Early. Test Often. Test Automatically. . . . . 237
you understand all of it before you incorporate it Tests that run with every build are much more ef-
into your project. fective than test plans that sit on a shelf.
51. Don’t Gather Requirements—Dig for Them . . . . . 202
Requirements rarely lie on the surface. They’re 63. Coding Ain’t Done ’Til All the Tests Run . . . . . . . 238
buried deep beneath layers of assumptions, mis- ’Nuff said.
conceptions, and politics.
64. Use Saboteurs to Test Your Testing . . . . . . . . . . . .244
52. Work with a User to Think Like a User . . . . . . . . . 204 Introduce bugs on purpose in a separate copy of
It’s the best way to gain insight into how the sys- the source to verify that testing will catch them.
tem will really be used.

53. Abstractions Live Longer than Details . . . . . . . . .209 65. Test State Coverage, Not Code Coverage . . . . . . . 245
Invest in the abstraction, not the implementation. Identify and test significant program states. Just
Abstractions can survive the barrage of changes testing lines of code isn’t enough.
from different implementations and new technolo-
gies. 66. Find Bugs Once . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
Once a human tester finds a bug, it should be the
54. Use a Project Glossary . . . . . . . . . . . . . . . . . . . . . . . . 210 last time a human tester finds that bug. Automatic
Create and maintain a single source of all the spe- tests should check for it from then on.
cific terms and vocabulary for a project.
67. English is Just a Programming Language . . . . . . 248
55. Don’t Think Outside the Box—Find the Box . . . . 213
Write documents as you would write code: honor
When faced with an impossible problem, identify
the DRY principle, use metadata, MVC, automatic
the real constraints. Ask yourself: “Does it have to
generation, and so on.
be done this way? Does it have to be done at all?”

56. Start When You’re Ready . . . . . . . . . . . . . . . . . . . . . . 215 68. Build Documentation In, Don’t Bolt It On . . . . . . 248
You’ve been building experience all your life. Don’t Documentation created separately from code is
ignore niggling doubts. less likely to be correct and up to date.

57. Some Things Are Better Done than Described . 218 69. Gently Exceed Your Users’ Expectations . . . . . . 255
Don’t fall into the specification spiral—at some Come to understand your users’ expectations, then
point you need to start coding. deliver just that little bit more.
58. Don’t Be a Slave to Formal Methods . . . . . . . . . . . 220
Don’t blindly adopt any technique without putting 70. Sign Your Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 258
it into the context of your development practices Craftsmen of an earlier age were proud to sign their
and capabilities. work. You should be, too.

Prepared exclusively for Zach

 Checklists
✔ Languages to Learn . . . . . . . . . . . . . . . . . . . page 17 ✔ Law of Demeter for Functions . . . . . page 141
Tired of C, C++, and Java? Try CLOS, Dylan, Eif- An object’s method should call only methods be-
fel, Objective C, Prolog, Smalltalk, or TOM. Each longing to:
of these languages has different capabilities and a Itself
different “flavor.” Try a small project at home using Any parameters passed in
one or more of them.
Objects it creates
Component objects
✔ The WISDOM Acrostic . . . . . . . . . . . . . . . . . page 20
What do you want them to learn?
What is their interest in what you’ve got to say?
How sophisticated are they?
✔ How to Program Deliberately .. . . . . page 172
How much detail do they want? Stay aware of what you’re doing.
Whom do you want to own the information? Don’t code blindfolded.
How can you motivate them to listen to you?
Proceed from a plan.
Rely only on reliable things.
✔ How to Maintain Orthogonality . . . . .page 34 Document your assumptions.
Design independent, well-defined components. Test assumptions as well as code.
Keep your code decoupled. Prioritize your effort.
Avoid global data. Don’t be a slave to history.
Refactor similar functions.

✔ Things to prototype . . . . . . . . . . . . . . . . . . . page 53 ✔ When to Refactor . . . . . . . . . . . . . . . . . . . . .page 185

You discover a violation of the DRY principle.
Architecture
You find things that could be more orthogonal.
New functionality in an existing system
Your knowledge improves.
Structure or contents of external data
The requirements evolve.
Third-party tools or components
You need to improve performance.
Performance issues
User interface design
✔ Cutting the Gordian Knot . . . . . . . . . . page 212
✔ Architectural Questions . . . . . . . . . . . . . page 55 When solving impossible problems, ask yourself:
Is there an easier way?
Are responsibilities well defined?
Am I solving the right problem?
Are the collaborations well defined?
Why is this a problem?
Is coupling minimized?
What makes it hard?
Can you identify potential duplication?
Do I have to do it this way?
Are interface definitions and constraints accept-
able? Does it have to be done at all?
Can modules access needed data—when
needed? ✔ Aspects of Testing . . . . . . . . . . . . . . . . . . . page 237
Unit testing
✔ Debugging Checklist . . . . . . . . . . . . . . . . . .page 98 Integration testing
Is the problem being reported a direct result of Validation and verification
the underlying bug, or merely a symptom? Resource exhaustion, errors, and recovery
Is the bug really in the compiler? Is it in the Performance testing
OS? Or is it in your code?
Usability testing
If you explained this problem in detail to a
Testing the tests themselves
coworker, what would you say?
If the suspect code passes its unit tests, are the
tests complete enough? What happens if you
run the unit test with this data?
Checklists from The Pragmatic Programmer, by Andrew Hunt
Do the conditions that caused this bug exist and David Thomas. Visit www.pragmaticprogrammer.com .
anywhere else in the system? Copyright c 2000 by Addison Wesley Longman, Inc.

Prepared exclusively for Zach