Visual Prolog for Tyros

Eduardo Costa
 PREFACE
This book started as a personal project. My intention was simply to
write a tutorial on Logic Programming for my son. The success of the book,
however, was immense, and I received many suggestions on how to improve
the text, or the code, and encouragement to keep on with the work. Mail
came from Saudi Arabia, China, Russia, Spain, France, Brazil, Canada, and
many other places. Although the space is short to list all people who wrote
me about the book, I want to acknowledge the collaboration and enthusiasm
of the following persons:

• Elena Efimova, who made many corrections on the text, and in the
historical background of Mathematics and Logics.

• Mark Safronov, who translated the book to Russian, and made many
corrections on the contents of the original. By the way, the Russian
translation has a much better layout than the English original.

• Thomas W. de Boer, who prepared a parallel edition of the book, with

longer explanations, and a text appropriate to rank beginners.

• Stuart Cumming. I often give examples that show what one should
avoid from the point of software engineering. Stuart pointed out that
people could be misled by these examples; he suggests that, whenever
an implementation is not robust, I should emphasize its weak points.
Although I agree with Stuart, I thought it better to remove the negative
examples altogether.

• Rose Shapiro, who corrected my Latin, and my description of many

historical events.

• Yuri Ilyin, who helped me with his expertise; without him, this book
would not be written.

• Reny Cury, who went through the manuscript correcting typos.

• Philippos Apolinarius who helped me with his knowledge of Botanical

Latin, and Chinese.

• Thomas Linder Puls and Elizabeth Safro, from PDC, for support and
encouragement.
Contents

I Savoir-faire 11
1 Introduction 13
1.1 Creating a project in VIP . . . . . . . . . . . . . . . . . . . . 13
1.1.1 Create a new GUI project: name . . . . . . . . . . . . 14
1.1.2 Compile and execute the program . . . . . . . . . . . . 15
1.2 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.3 Notions of Logics: Ancient Greeks . . . . . . . . . . . . . . . . 16

2 Forms 17
2.1 Create a form: folder/name . . . . . . . . . . . . . . . . . . . . 17
2.2 Enable the task menu: e.g. File/New option . . . . . . . . . . . 17
2.3 In CodeExpert, add code to project tree item . . . . . . . . . . 18
2.4 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.5 Notions of Logics: Aristotle’s Syllogistic . . . . . . . . . . . . 22
2.5.1 Valid syllogisms . . . . . . . . . . . . . . . . . . . . . . 24

3 Mouse events 27
3.1 Add code to MouseDownListener . . . . . . . . . . . . . . . . 27
3.2 onPaint . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.3 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.4 Notions of Logics: Boolean algebra . . . . . . . . . . . . . . . 30
3.5 Argument forms . . . . . . . . . . . . . . . . . . . . . . . . . . 31

4 Less Figures 33
4.1 Task menu . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.2 Project Tree . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
4.2.1 Code Expert . . . . . . . . . . . . . . . . . . . . . . . . 35
4.3 Create Project Item . . . . . . . . . . . . . . . . . . . . . . . . 36
4.4 Create a new class: folder/name . . . . . . . . . . . . . . . . . 39
4.5 Edit field contents . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.6 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

1
2 CONTENTS

4.7 Notions of Logics: Predicate calculus . . . . . . . . . . . . . . 40

5 Horn Clauses 41
5.1 Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
5.2 Predicates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
5.3 Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
5.4 Multi-solutions . . . . . . . . . . . . . . . . . . . . . . . . . . 46
5.4.1 A program that uses multi solution predicates . . . . . 47
5.5 Logical junctors . . . . . . . . . . . . . . . . . . . . . . . . . . 48
5.6 Implication . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
5.7 Horn Clauses . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
5.8 Declarations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
5.9 Drawing predicates . . . . . . . . . . . . . . . . . . . . . . . . 53
5.10 GDI Object . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
5.11 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
5.12 Notions of Logics: Meaning of Horn clauses . . . . . . . . . . . 56

6 Console Applications 57
6.1 Cut . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
6.2 Lists . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
6.3 List schemes . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
6.4 String operations . . . . . . . . . . . . . . . . . . . . . . . . . 68
—Parsing into tokens: fronttoken . . . . . . . . . . . . . . . . . . 68
—String concatenation: concatList . . . . . . . . . . . . . . . . . 68
—String concatenation: concat/2 and concat/3 . . . . . . . . . . 68
—Term to String conversion: toTerm . . . . . . . . . . . . . . . . 69
—Term to String conversion: hasdomain . . . . . . . . . . . . . . 69
—Term to String conversion: toString . . . . . . . . . . . . . . . . 69
—Term to String conversion: format . . . . . . . . . . . . . . . . . 70
6.4.1 Useful String Predicates . . . . . . . . . . . . . . . . . 71
—String processing: Useful predicates . . . . . . . . . . . . . . . . 71
6.5 Notions of Logics: Grammar for Predicates . . . . . . . . . . . 76

7 Grammars 79
7.1 Parsing grammar . . . . . . . . . . . . . . . . . . . . . . . . . 80
7.2 Generating grammar . . . . . . . . . . . . . . . . . . . . . . . 81
7.3 Why Prolog? . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
7.4 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
7.5 Notions of Logics: Natural deduction . . . . . . . . . . . . . . 84
CONTENTS 3

8 Painting 87
8.1 onPainting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
8.2 Custom Control . . . . . . . . . . . . . . . . . . . . . . . . . . 91
8.3 Notions of Logics: Resolution principle . . . . . . . . . . . . . 93

9 Data types 95
9.1 Primitive data types . . . . . . . . . . . . . . . . . . . . . . . 95
9.2 Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
9.3 Sets of numbers . . . . . . . . . . . . . . . . . . . . . . . . . . 97
9.4 Irrational Numbers . . . . . . . . . . . . . . . . . . . . . . . . 100
9.5 Real Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
9.6 Mathematics . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
9.7 Format . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
9.8 domains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
9.8.1 Lists . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
9.8.2 Functors . . . . . . . . . . . . . . . . . . . . . . . . . . 104
9.9 Notions of Logics: Horn clauses . . . . . . . . . . . . . . . . . 108

10 How to solve it in Prolog 109

10.1 Utilities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110

11 Facts 129
11.1 Class file . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
11.1.1 Reading and writing a string . . . . . . . . . . . . . . . 132
11.2 Constants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133

12 Classes and Objects 135

12.1 Object facts . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137

13 Giuseppe Peano 139

13.1 Turtle graphics . . . . . . . . . . . . . . . . . . . . . . . . . . 139
13.2 Turtle states . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
13.3 Recursion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
13.4 Peano’s curve . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
13.5 Latino Sine Flexione . . . . . . . . . . . . . . . . . . . . . . . 143
13.6 Quotations from Key to Interlingua . . . . . . . . . . . . . . . 144
13.7 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149

14 L-Systems 151
14.1 Class draw . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
14.2 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
4 CONTENTS

15 Games 155
15.1 Object Facts . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162

16 Animation 163
16.1 dopaint . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
16.2 Handling the timer . . . . . . . . . . . . . . . . . . . . . . . . 166
16.3 How the program works . . . . . . . . . . . . . . . . . . . . . 166

17 Text Editor 167

17.1 Saving and Loading files . . . . . . . . . . . . . . . . . . . . . 168

18 Printing 171

19 Tab forms and other goodies 173

19.1 Famous programs . . . . . . . . . . . . . . . . . . . . . . . . . 174
19.2 Botany . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176
19.3 Handling Chinese . . . . . . . . . . . . . . . . . . . . . . . . . 182
19.4 Regular expressions . . . . . . . . . . . . . . . . . . . . . . . . 186
19.5 A MIDI player . . . . . . . . . . . . . . . . . . . . . . . . . . 189
19.6 Midi format . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191

20 Bugs 195
20.1 Type error . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
20.2 Non-procedure inside a procedure . . . . . . . . . . . . . . . . 197
20.3 Non-determ condition . . . . . . . . . . . . . . . . . . . . . . . 198
20.4 Impossible to determine the type . . . . . . . . . . . . . . . . 198
20.5 Flow pattern . . . . . . . . . . . . . . . . . . . . . . . . . . . 200

21 A Data Base Manager 201

21.1 Database manager . . . . . . . . . . . . . . . . . . . . . . . . 201
21.2 btr class . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
21.3 The database manager . . . . . . . . . . . . . . . . . . . . . . 207

22 Books and articles 209

22.1 Grammars . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
22.2 Databases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
22.3 Programming Techniques . . . . . . . . . . . . . . . . . . . . . 210

II Artificial Intelligence 213

23 Search 215
CONTENTS 5

23.1 States . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215

23.2 Search tree . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
23.3 Breadth First Strategy . . . . . . . . . . . . . . . . . . . . . . 218
23.4 Depth First Strategy . . . . . . . . . . . . . . . . . . . . . . . 222
23.5 Heuristic search . . . . . . . . . . . . . . . . . . . . . . . . . . 223

24 Neural Networks 229

24.1 Neuron description . . . . . . . . . . . . . . . . . . . . . . . . 234
24.2 Implementation of a multi-layer network . . . . . . . . . . . . 235
24.3 Running the two layer neural network . . . . . . . . . . . . . . 240
24.4 A Historical Perspective . . . . . . . . . . . . . . . . . . . . . 241

25 Alpha Beta Pruning 243

25.1 Move generation . . . . . . . . . . . . . . . . . . . . . . . . . . 244
6 CONTENTS
List of Figures

1.1 Task menu: New Project . . . . . . . . . . . . . . . . . . . . . 13

1.2 Project Settings . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.3 Project tree . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.4 Building the project . . . . . . . . . . . . . . . . . . . . . . . 15
1.5 An empty application . . . . . . . . . . . . . . . . . . . . . . . 15

2.1 Adding a new item to the Project Tree . . . . . . . . . . . . . 18

2.2 Create a new form . . . . . . . . . . . . . . . . . . . . . . . . 19
2.3 Resizing the form window . . . . . . . . . . . . . . . . . . . . 19
2.4 Project Tree/Task Menu . . . . . . . . . . . . . . . . . . . . . 20
2.5 Enabling the Task Menu . . . . . . . . . . . . . . . . . . . . . 20
2.6 Going to the Code Expert . . . . . . . . . . . . . . . . . . . . 21
2.7 Dialog and Window Expert . . . . . . . . . . . . . . . . . . . 21
2.8 Pop-up Form . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.9 Barbara . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.10 Darii . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

3.1 PaintResponder . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.2 Paint Responder . . . . . . . . . . . . . . . . . . . . . . . . . 30

4.1 Create a new package . . . . . . . . . . . . . . . . . . . . . . . 37

4.2 A form with an edit field . . . . . . . . . . . . . . . . . . . . . 37
4.3 The fn class . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

5.1 Interesting predicates . . . . . . . . . . . . . . . . . . . . . . . 43

5.2 mapDataBase/draw.cl . . . . . . . . . . . . . . . . . . . . . . 45
5.3 mapDataBase/draw.pro . . . . . . . . . . . . . . . . . . . . . 46
5.4 draw.cl and draw.pro . . . . . . . . . . . . . . . . . . . . . . . 50
5.5 Utah cities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
5.6 Class draw . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

6.1 Reduce in Prolog . . . . . . . . . . . . . . . . . . . . . . . . . 65

7
8 LIST OF FIGURES

6.2 Zip in Prolog . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

7.1 Die Welt ist alles, was der Fall ist. . . . . . . . . . . . . . . . . 84

8.1 A form with a custom control . . . . . . . . . . . . . . . . . . 92

9.1 A thinking Greek . . . . . . . . . . . . . . . . . . . . . . . . . 101

9.2 List declaration . . . . . . . . . . . . . . . . . . . . . . . . . . 105

10.1 Minmax tree . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128

11.1 Reading a string from file . . . . . . . . . . . . . . . . . . . . 133

13.1 Files curve.cl and curve.pro . . . . . . . . . . . . . . . . . . . 141

13.2 File turtle.cl . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
13.3 turtle.pro . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
13.4 Peano’s curve . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
13.5 Inserting a file into an editor . . . . . . . . . . . . . . . . . . . 149
13.6 A dictionary of “difficult” LsF words . . . . . . . . . . . . . . 150

15.1 Lawn form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156

15.2 draw.cl and draw.pro . . . . . . . . . . . . . . . . . . . . . . . 160
15.3 objstate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161

16.1 click class handles the timer . . . . . . . . . . . . . . . . . . . 164

16.2 dopaint.pro . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165

17.1 The insert custom control button . . . . . . . . . . . . . . . . 168

17.2 Choose Custom Control dialog . . . . . . . . . . . . . . . . . . 168
17.3 A form with Edit Control . . . . . . . . . . . . . . . . . . . . 169
17.4 User control properties . . . . . . . . . . . . . . . . . . . . . . 169
17.5 Code for onSave . . . . . . . . . . . . . . . . . . . . . . . . . . 170
17.6 Code for onLoad . . . . . . . . . . . . . . . . . . . . . . . . . 170

19.1 listEdict methods . . . . . . . . . . . . . . . . . . . . . . . . . 181

19.2 File: ideograms.pro . . . . . . . . . . . . . . . . . . . . . . . . 184
19.3 Midi generator . . . . . . . . . . . . . . . . . . . . . . . . . . 194

20.1 A buggy program . . . . . . . . . . . . . . . . . . . . . . . . . 197

20.2 A bug caused by a non deterministic predicate . . . . . . . . . 199
20.3 Deterministic predicate required. . . . . . . . . . . . . . . . . 199

23.1 A search tree . . . . . . . . . . . . . . . . . . . . . . . . . . . 216

23.2 Puzzle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
LIST OF FIGURES 9

24.1 A three input perceptron . . . . . . . . . . . . . . . . . . . . . 230

24.2 A three input perceptron . . . . . . . . . . . . . . . . . . . . . 234
24.3 A neural network that can learn how to calculate a xor-gate . 235
10 LIST OF FIGURES
 Part I

Savoir-faire

11
Chapter 1

Introduction

Figure 1.1: Task menu: New Project

1.1 Creating a project in VIP

Let us create an empty project, to which you will add functionality later
on. The environment that you will use to develop your programs is called
IDE, that is the acronym for Integrated Development Environment. When
you enter the Visual Prolog IDE, you will get an environment like the one
shown in figure 1.1; we will refer to the menu of the IDE as the task menu.

13
14 Introduction

The system of windows and dialogs that you will create to communicate with
potential users of your programs is called a Graphical User Interface, or GUI
for short.

1.1.1 Create a new GUI project: name

This step is quite easy. Choose the option Project/New from the task menu,
as shown in figure 1.1. Then, fill in the Project Settings dialog as in figure 1.2.
Press the button OK and you will get the project tree (figure 1.3).

Figure 1.2: Project Settings

Figure 1.3: Project tree

1.1 Creating a project in VIP 15

1.1.2 Compile and execute the program

To compile the program, choose the option Build/Build from the task menu,
as shown in figure 1.4. To execute it, choose Build/Execute (see figure 1.4)
from the task menu, and a window like the one in figure 1.5 will pop up on
the screen. In order to exit the program, all you have to do is to click on the
X-shaped button that appears on the upper right hand corner of the Window
frame; if you prefer, choose the option File/Exit option of the application
task menu.

Figure 1.4: Building the project

Figure 1.5: An empty application

16 Introduction

1.2 Examples
The example in this chapter shows how to run an application, and is in di-
rectory plot1. In the next chapter, you will add functionality to the plot1
project, which means that will make the corresponding application do some-
thing useful.

1.3 Notions of Logics: Ancient Greeks

The Ancient Greeks devised a kind of society that was very different from
anything that existed before them. In other civilizations, decisions of political
or legal nature were taken by a king, a small group of aristocrats, or a king’s
deputy; these people could commit to writing the basis for their decisions;
when they did so, they would refer the resulting codex to a god, or to a
traditional hero. As for the Greeks, the laws and political decisions were
proposed by citizens for whom feats, lots, or life experience won the right
to a seat in an assembly or in a court of law. The proposition was voted,
and enforced only if the majority of the voters supported it. Let us see how
Pericles describes the situation.

Our constitution does not copy the laws of neighboring states;

we are rather a model to others than imitators ourselves. Its
administration favors the many instead of the few; this is why it
is called a democracy. If we look to the laws, they afford equal
justice to all in their private differences; if no social standing,
advancement in public life falls to reputation for capacity, class
considerations not being allowed to interfere with merit; nor again
does poverty bar the way: If a man is able to serve the state, he
is not hindered by the obscurity of his condition from proposing
laws, and political actions.

In this kind of society, if someone wanted to advance his interests, or make

inroads for his views, he would need to convince others, prove that he was
right, win debates and disputes. Therefore, necessity forced the Greeks to
invent Logics. Of course they developed other methods of winning an argu-
ment besides Logics: The art of telling lies, crowd psychology, the science
of disguising a fallacy, ornate language, and flattery were among the Greek
methods of winning a debate. However, this book will focus on Logics.
Chapter 2

Forms

In this chapter, you will add a form to the empty project that you have
created in chapter 1. A form is a window pane where you can add components
like buttons, that trigger actions, edit fields, that you can use to input text,
and canvas, where you can draw things.

2.1 Create a form: folder/name

To create a form, choose the option File/New in New Package from the task
menu, as in figure 2.1. Select Form from the left pane and fill in the Create
Project Item dialog as in figure 2.2. The new form name is query. Since you
have chosen the option New in New Package, Visual Prolog will create the
query form in a package named after the form. After you press the Create
button of the Create Project Item dialog, the IDE shows you a prototype
of the new form, as in figure 2.3. You may want to change the size of the
window rectangle, making it slightly larger than the prototype. In order to
do this, click and hold the mouse at the lower right corner and drag it, as
you do when you resize a normal window.

2.2 Enable the task menu: e.g. File/New option

When you ran the empty application, you certainly took notice of the fact
that the File/New option is disabled. To enable it, click on the TaskMenu.mnu
branch of the Project tree (figure 2.4). Then, open the tree that appears in
the lower part of the TaskMenu dialog, and remove the tick from the radio
button Disabled that corresponds to the &New\tF7 option, as in figure 2.5.

17
18 Forms

2.3 In CodeExpert, add code to project tree item

To add code to the File/New option, click on the TaskWindow.win branch of

the project tree with the right button of the mouse, which opens a floating
menu. Choose the option Code Expert (figure 2.6). As in figure 2.7, click on

Finally, press the button Add (refer to figure 2.7), and double click on the
branch id_file_new->onFileNew. This will open a text editor, with the
following snippet:

clauses
onFileNew(_Source, _MenuTag).

Build the application; then modify this snippet to:

clauses
onFileNew(W, _MenuTag) :- X= query::new(W), X:show().

Build the program again, choosing the option Build/Build from the task
menu, as in figure 1.4. Execute the program, and you will see that, whenever
you choose the File/New option, a new form is created.

Figure 2.1: Adding a new item to the Project Tree

2.3 In CodeExpert, add code to project tree item 19

Figure 2.2: Create a new form

Figure 2.3: Resizing the form window

20 Forms

Figure 2.4: Project Tree/Task Menu

Figure 2.5: Enabling the Task Menu

2.3 In CodeExpert, add code to project tree item 21

Figure 2.6: Going to the Code Expert

Figure 2.7: Dialog and Window Expert

22 Forms

2.4 Example
The example in this chapter shows how one creates a form by choosing the
option File/New, then filling in the Create Project Item dialog.

2.5 Notions of Logics: Aristotle’s Syllogistic

For Aristotle, a proposition consists of two terms, a subject, and a predicate,
and a quantifier: ‘every’, ‘no’, ‘some’, ‘not every’. Subject comes from the
Latin rendering of the Greek term ÍpokeÐmenon, which means what (or whom)
the proposition is about. Predicate is what the preposition talks about in
the subject.
The Portuguese Logician, medical doctor, and mathematician Pedro Julião
(also known as Peter of Spain) made important contributions to many fields
of knowledge, before being elected pope under the name of John XXI; for
instance, he is considered the father of the ophthalmology, and of the pre-

Figure 2.8: Pop-up Form

2.5 Notions of Logics: Aristotle’s Syllogistic 23

ventive medicine. He wrote a very influential Logical treatise (Summulae

Logicales), and is credited with introducing the following abreviations for
the Logical quantifiers:

Quantifier Type Abreviation Example

a Universal afirmative P aQ Every P is Q
e Universal negative P eQ No P is Q
i Particular afirmative P iQ Some P is Q
o Particular negative P oQ Some P is not Q

Propositions are classified in contraries, contradictories, and subcontraries.

Here are the definitions of each one of these classes:

a. Two propositions P and Q are contradictories if and only if

i. P and Q cannot both be true, and

ii. necessarily, either P , or Q is true.

Types a and o are contradictories; the same happens to e and i.

b. P and Q are contraries if and only if

i. P and Q cannot both be true, and

ii. P and Q can both be false.

a and e are contraries.

c. P and Q are subcontraries if and only if

i. P and Q cannot both be false, and

ii. P and Q can both be true.

i and o are subcontraries.

A syllogism is an argument consisting of exactly three propositions (two

premises and a conclusion) in which there appear a total of exactly three
terms, each of which is used exactly twice. Let P represent the predicate
of the conclusion (major term), and S the subject of the conclusion (minor
term). Aristotle called these terms the extremes. M represents the term
that is common to the premises, but it is absent in the conclusion; it is called
the middle term. The premise containing P is called the major premise; the
premise containing S is the minor premise.
24 Forms

Let an asterisk ∗ represent any of the four quantifiers (‘a’, ‘e’, ‘i’, ‘o’). In
this case, the syllogisms are classified in three figures, to wit,

F irst F igure Second F igure T hird F igure

P ∗M M ∗P P ∗M
M ∗S M ∗S S∗M
P ∗S P ∗S P ∗S

When we replace each asterisk with a quantifier, we get a tuple of proposi-

tions: {major premise, minor premise, conclusion}. According to Pe-
ter of Spain, each distinct tuple ({a, a, a}, {a, i, o} . . .) constitutes a modus,
or mood. In order to calculate how many moods there exist, we need to
review our lessons of combinatorics.
Permutations are the ways of building lists of things; in a list, the order
matters. On the other hand, combinations are groups of things, and the
order of the elements of a group does not matter. Since the order of the
quantifiers in a syllogism does matter, we are dealing with permutations.
We need to permute three elements from a set of four elements; repetition
is not disallowed. If you have n things to choose from, and you choose p of
them, then the number of possible permutations is given by

n
| × n ×{zn × . . . n} = np
p times
In the present case, one has to choose three quantifiers from the set {a, e, i, o}.
Therefore, there are 43 = 64 possibilities for each figure; since we have three
figures, the number of possibilities grows to 192.

2.5.1 Valid syllogisms

Not all of canditate syllogisms are logically valid. Aristotle recognized four
valid moods each for the first and second figures, and six moods in the case
of the third figure. On the other hand, Modern Logics recognizes four modes
each for all three figures. Peter of Spain invented Latin mneumonics to help
us remember the valid moods; in his scheme, the vowels indicate the mood;
for instance, BARBARA indicates {a, a, a}, and is a valid mood for the first
figure. The cultors of Classical Latin will notice that the names are based
2.5 Notions of Logics: Aristotle’s Syllogistic 25

on the Medieval Latin pronunciation, where a word like Cæsare was spelled
Cesare, and pronounced accordingly. Anyway, here is the complete list of
the valid moods:

First figure Second figure Third figure

{a, a, a} – Barbara {e, a, e} – Cesare {o, a, o} – Bocardo
{e, a, e} – celarent {a, e, e} – Camestres {e, i, o} – Ferison
{a, i, i} – Darii {e, i, o} – Festino {i, a, i} – Disamis
{e, i, o} – ferio {a, o, o} – baroco {a, i, i} – Datisi

Figure 2.9: Barbara

According to [Lewis Carroll], Venn proposed a diagram to check whether

a syllogism is valid. Figure 2.9 illustrates Venn’s method for Barbara. The
first premise of Barbara says that all M is P; therefore, one shades the part
of M that is not in P (left hand side of the figure). The second premisse
says that All S is M; therefore, one must shade the part of S that is outside
M. The remaining unshaded part of S is entirely inside P. Then, one can
conclude that All S is P.
M aP
SaM
SaP
Let us now consider the Darii syllogism, that says

∀P is M
∃M is S
∃S is P
Where the symbol ∀ means All, and ∃ means Some. Venn claims that the
universal premise should be diagrammed before diagramming a particular
premise. Therefore, at the left hand side of figure 2.10, we have shaded
the area of P that is not on M , as required by the premise ∀P is M . The
26 Forms

next step is to place the ∃ symbol in the area corresponding to the premise
∃M is S. Figure 2.10 shows that the ∃ symbol could be in the SM P̃ area or
in the SP M area. Since we do not know exactly which area it is in, we put
the ∃ symbol on the line. The conclusion is that Some S is P.

Figure 2.10: Darii

Alice Liddell’ s father wanted to give her a very good education. When
he noticed that there did not exist any good Greek-English dictionaries,
he and his friend Scott sat and prepared a very good one, that almost all
students of Greek use with pleasure and profit. He also hired the great
writer [Lewis Carroll] as a teacher of Logics for his two daughters. In order
to make his classes interesting, [Lewis Carroll] wrote a few books for children:
Alice in the Wonderland, Through the Looking Glass, and the Game of Logics.
The example below is from the Game of Logics.

X
+
Y
M
+

No philosophers are conceited;

Some conceited persons are not-gamblers.
Therefore some not-gamblers are not philosophers.

The following solution to the above syllogism has been kindly supplied to
me by Mr. Venn himself: The Minor Premiss that some of the constituents
in my’ must be saved — mark it with a cross. The Major declares that all
xm must be destroyed; erase it. Then as some my’ is to be saved, it must
clearly be my’x’. That is, there must exist my’x’; or, eliminating m, y’x’. In
common phraseology, Some not-gamblers are not philosophers.
Chapter 3

Mouse events

In the first two chapters of this document, you learned how to build an appli-
cation with a form, that pops up whenever you choose the option File/New,
from the application menu.

3.1 Add code to MouseDownListener

Click on the query.frm branch of the project tree in order to reopen the query
form, as shown on figure 2.3; press the button Events of the Properties-dialog.
This should open an Event list, as shown on the figure below.

27
28 Mouse events

Click on the MouseDownListener event, and replace the onMouseDown/4 pro-

cedure with the following code:

clauses
onMouseDown(S, Point, _ShiftControlAlt, _Button) :-
W= S:getVPIWindow(), Point= pnt(X, Y),
vpi::drawText(W, X, Y, "Hello, World!").

Build the program, and execute it. Choose

the option File/New to create a new form.
Whenever you click at any point within the
form, the program will plot a famous greeting
expression.

3.2 onPaint
You have used the event handler onMouseDown/4 to paint something on the
form. There are programmers that do not think that this is a good idea. In
fact, there are languages that make this approach very difficult. Java is one
of these languages, a very popular one, by the way. In Java, one should do
all paintings and drawings inside the following method:

public abstract void doPaint(java.awt.Graphics g)

Of course, one can get around this restriction, even in Java. Anyway, let us
learn how to put all drawing activities inside the event handler onPaint/3.

1. Create a project, whose name is plot0.

2. Add a new query form in a new query package.

3. Enable the task menu option File/New, and add the snippet

clauses
onFileNew(S, _MenuTag) :-
Q= query::new(S), Q:show().

to ProjTree/Taskwindow.win/Menu/TaskMenu/id_file/id_file_new.

4. Build the application.

3.2 onPaint 29

The next step is to add the snippet

class facts
mousePoint:pnt := pnt(-1, -1).

predicates
onMouseDown : drawWindow::mouseDownListener.

clauses
onMouseDown(_S, Point, _ShiftControlAlt, _Button) :-
mousePoint := Point,
Point= pnt(X, Y),
R= rct(X-8, Y-8, X+60, Y+8),
invalidate(R).

to the MouseDownListener. You are ready to implement the event han-

dler onPaint for query. The event handler onPaint will be called when-
ever a rectangular region of the form becomes obsolete or invalid. When
does a painting become invalid? When the window is covered completely or
partially by another window, or when you invalidate it using the predicate
invalidade(R), where R is a rectangle. The rectangle is described by the
coordinates of its upper left angle, and its lower right angle.

R= rct(X-8, Y-8, X+60, Y+8)

You have probably noticed that the event

handler onMouseDown/4 invalidates a rectan-
gular region around the mouse click. Then
the event handler onPaint will act on this in-
valid region by repainting it. To bring this
into effect, add the snippet of figure 3.2 to
the PaintResponder. From the Project Tree,
click on the query.frm branch. Press the Event
button of the Properties-dialog (see figures 3.1,
and 2.3, page 19); you will find the PaintRe-
sponder in the list of events that will appear
on the Properties-dialog. This new approach
requires you to understand a lot of new things, Figure 3.1: PaintResponder
like the if-then-else construction, and the idea
of global variable. Do not worry about these things for now. You will learn
about them at the appropriate time.
30 Mouse events

predicates
onPaint : drawWindow::paintResponder.

clauses
onPaint(_Source, _Rectangle, GDIObject) :-
if pnt(X, Y)= mousePoint, X>0
then
mousePoint := pnt(-1, -1),
GDIObject:drawText(pnt(X, Y), "Hello, World!", -1)
else succeed() end if.

Figure 3.2: Paint Responder

3.3 Examples
This chapter shows how to handle a mouse event, such as a left click.

3.4 Notions of Logics: Boolean algebra

Propositional Calculus is a formal system in which formulæ representing
propositions can be formed by combining atomic propositions using logical
connectives. A formula of the PC obeys the following grammar rules:

1. The α-Set contains atomic formulæ, in general, lowercase letters of the

Latin alphabet. Any member of the α-Set is a well formed formula.

2. If p is a formula, then ¬p is a formula. The symbol ¬ represents the

logical negation. Therefore, ¬p is true if p is false, and false otherwise.

3. If p and q are formulæ, then p ∨ q (logical disjunction), p ∧ q (logical

conjunction), p → q, p ≡ q are also formulae. Semantically, the formula
p ∨ q means p or q, and is true if either p or q is true; p ∧ q is true only
if both p and q are true; p → q means that p implies q, i.e., if p is true
then q is true. Finally, p ≡ q means that p is equivalent to q: If p is
true, then q is true as well, and if q is true, then p is also true.

Truth tables. In order to establish the truth value of a formula, one often
resorts to truth tables. Let us learn about truth tables through examples.
Let us consider the well formed formula p ∧ q. If 1 represents the constant
true, and 0 represents false, the truth table of the negation(¬p), of logical
3.5 Argument forms 31

conjunction(p ∧ q), and of the logical disjunction(p ∨ q) is given by

p q ¬p p∧q p∨q
0 0 1 0 0
0 1 1 0 1
1 0 0 0 1
1 1 0 1 1
Informally, the formula p → q means that

if p is true, then q is also true

A reasonable truth table that conveys such a meaning is given below.

p q p→q
0 0 1
0 1 1
1 0 0
1 1 1
It is possible to prove that P → Q is equivalent to the formula ¬P ∨ Q; let
us verify this assertion.

p q ¬p ¬p ∨ q p→q
0 0 1 1 1
0 1 1 1 1
1 0 0 0 0
1 1 0 1 1

3.5 Argument forms

Logical programmers often use a kind of argument called modus tollendo
ponens. In logical operator notation, this simple and very useful argument
form is written thus:
p∨q

¬p

`q
where ` represents the logical assertion. Roughly speaking, we are told that
either p or q is true; then we are told that p is not true; so we infer that it
has to be q that is true.
32 Mouse events

A very important argument form is the modus ponens. It is important

because it drives Prolog’s inference engine.

q←p

p
`q
You have learned that p → q means if p then q, i.e., p implies q. Therefore,
q ← p means that q is implied by p, or q if p. In fewer words, the first line
of the argument says that q is true, if p is true; the second line says that
p is true; then man or machine can infer that q is true. Let us consider a
concrete example.

happy(M an) ← hearsSocrates(M an)

hearsSocrates(plato)
` happy(plato)
Another valid argument is called modus tollens:

p→q

¬q
` ¬p
Chapter 4

Less Figures

Before going ahead, let us find out a way to describe our projects without
having to resort to figures.

4.1 Task menu

The task menu, shown in in figures 1.1 and 1.4, is the main menu of VIP IDE.
The notation A/B refers to one of its options. For instance, use the option
Project/New of the task menu to create a new project (figure 1.1). It has
six forms, to wit: General, Directories, Build Options, Version Information,
File Templates, and Run Options. In most cases, you need to fill in only the
General form.

General
Project Name: factorial
UI Strategy: Object-oriented GUI(pfc/gui)
Target Type: Exe
Base Directory: C:\vispro

When you find the instructional step:

Create a new GUI project: factorial (section 1.1.1).

you should enter the Project Settings dialog (by choosing the Project/New
option of the VDE task menu) and fill in the General form as shown above.
Do it mutatis mutandis, since your computer may not have space available
on drive C:, or you may want to put your programs in a directory different
from C:\vispro.

33
34 Less Figures

4.2 Project Tree

The easiest way to navigate through files and resources is to click on the
corresponding item of the Project Tree:

If you left click on a folder, it will open and present its contents. If you click
on an item with the right button of the mouse, a floating menu opens, as
shown in the figure. If I want you to go to the code expert, and add a snippet
to the TaskWindow.win, I will say:

In CodeExpert, add code to TaskWindow/TaskWindow.win(section 2.3)

To carry out this instruction, go to the project tree and do the following:

1. Click on the TaskWindow folder to open it, if it is closed.

2. Right click on the TaskWindow.win branch of the project tree to deploy

the floating menu, which has the following options:

Edit
Attribute
Delete
Code Expert
File Properties...

3. Finally, choose the option Code Expert.

4.2 Project Tree 35

4.2.1 Code Expert

The code expert also has a tree shape.

As its name reveals, the code expert

is used to insert code, or snippets, into
the many files of your project. To reach
the code expert, you must right click on
the item of the project tree to which you
want to add code. Then, choose the op-
tion Code Expert from the floating menu.
In the case of forms, there is a short-
cut to the Code Expert. If you left click
the branch of the form on the Project
Tree, a prototype of the form will pop
up, together with the Properties-dialog,
and two component boxes, one for Con-
trols, and another for Layout schemes
(see figure 2.3); if you press the button
Events on the Properties-dialog, you will
get a menu of events to choose from. In
past chapters, you had the opportunity
to work with two event handlers, to wit,
the MouseDownListener (page 27), and
the PaintResponder (page 29).
Forms have components, like buttons,
and edit fields. On the top of the
Properties-dialog, there is a list box for
selecting these components; if you select
one of these components, and press the
Events-button you will get a list of the
events that correspond to the selected
component. More on that later on.
To navigate through the Code Expert
tree, and reach the point where you want
to insert code, left-click on the appropri-
ate branches of the tree. If you want the
code expert to add a prototype on a leaf,
click on the leaf, and press the button Add that will appear at the bottom of
the dialog. Then, click on the leaf again, to reach the code.
36 Less Figures

If I ask you: In CodeExpert, add (section 2.3)

clauses
onFileNew(W, _MenuTag) :- S= query::new(W), S:show().
to TaskWindow.win/CodeExpert/Menu/TaskMenu/id_file/id_file_new,
here are the steps that you must follow:
• Project tree — Open the folder TaskWindow of the project tree, right
click on TaskWindow.win to open its menu, and choose the option
Code Expert (figure 2.6).
• Code Expert— Click on Menu → click on TaskMenu → click on id_file.
Select id_file_new and press the button Add to generate prototype
code. Finally, click on id_file_new->onFileNew. Refer to figures 2.6,
2.7, and to sec. 2.3. Add the requested code to the file TaskWindow.pro.

clauses
onFileNew(W, _MenuTag) :- S= query::new(W), S:show().

4.3 Create Project Item

To add a new item to the project tree, choose the option
File/New in New Package
from the task menu if you want to place the item inside a new package named
after it. If you want to insert the item into an existing package, choose the
option File/New in Existing Package. Be careful to place the new item,
or the new package, into the desired folder. In the example below, one places
the package containing the query form on the root of factorial. Always
choose names that mean something. For instance, if your package holds
computer graphics, you could choose a name like canvasFolder; if it contains
queries, a good name is formcontainer; and so on. Example:
• Create a new package on the root of the factorial-project (fig-
ure 4.1). Let the name of the package be formcontainer.
• Create a new form (query) inside the formcontainer-package, by
selecting the branch corresponding to the formcontainer-package in
the Project Tree, and choosing the option File/New in Existing Package
from the task menu (section 2.1). In order to put the window inside
a package, be sure that the package folder is selected before going to
File/New in Existing Package.
4.3 Create Project Item 37

Figure 4.1: Create a new package

When you create a form, the IDE shows shows the Form Window dialog
(see figure 4.2). You can take the opportunity to resize the window and add
an edit field (handle: edit_ctl) and a button (handle: factorial_ctl), as
shown in the figure. You can also reach the Form Window dialog by left
clicking on the form name in the project tree.

Figure 4.2: A form with an edit field

If you “double-click” on ProjectTree/TaskMenu.mnu, you get a dialog like

the one shown in figure 2.5. You can unfold the tree of this menu specification
by clicking on its option branches, as already mentioned. You may also need
to enable a menu option, as in section 2.2. It is also possible to create a new
38 Less Figures

menu entry, as shown in the figure below, where I have pressed the button
(icon) labelled New Item and created the entry Drawing Tools by filling in
the dialog. As you can see from the figure, the new option is enabled from
the start. The symbol & in the &File entry makes F a shortcut key.

Let us start a new project from scratch, to see what we can do with fewer
figures and check whether we have a terse description of a VIP project.

Create a new GUI project: factorial (section 1.1.1).

Add a new package to the project tree: factorial/formcontainer.

Create a new form: formcontainer/query (section 4.3). Add an Edit Field

(edit_ctl ) and a Button (factorial_ctl) to it, as in figure 4.2.

Enable the TaskMenu File/New option. Enable File/New (section 2.2),

and add (section 2.3):

clauses
onFileNew(W, _MenuTag) :- S= query::new(W), S:show().

to TaskWindow.win/Menu/TaskMenu/id_file->id_file_new->onFileNew.

Build the application to insert the new classes into the project.
4.4 Create a new class: folder/name 39

4.4 Create a new class: folder/name

To create a new class and put it inside the package formcontainer, select the
folder formcontainer in the Project Tree, and choose the option File/New in
Existing Package from the IDE task menu. Select the option Class from the
Project Item dialog, and fill in the class Name field with fn. Make sure to
untick the Create Objects box.
When you press the Create button, VIP will display the files fn.pro
and fn.cl, that contain a prototype of the fn class. It is our job to add
functionality to these files, by replacing the contents of listing of page 40
for them. Then, build the application once more, to make sure that Visual
Prolog inserts the fn-class into the project.

4.5 Edit field contents

It is quite tricky to get the contents of an edit field. This is true for Visual C,
Delphi, Clean, etc. To do justice to Visual Prolog, I must say that it offers
the easiest access to controls, as compared to other languages. However, it
is still tricky. To make things somewhat easier, the IDE stores a handle to
the edit field window in a fact variable. To see how to make good use of this
convenience, open the query.frm by clicking on its branch in the project
tree.

Choose the button:factorial_ctl from the Properties-dialog list box, press

the Event-button, and click on the ClickResponder entry of the Event-list;
then add the following snipet to the factorial_ctl button code:

clauses
onFactorialClick(_S) = button::defaultAction() :-
fn::calculate(edit_ctl:getText()).

Build and run the program.

40 Less Figures

4.6 Examples
In this chapter, you have learned how to cre-
ate a form with a button(factorial_ctl, and
an edit field (edit_ctl); you have also learned
how to access the contents of the edit field.
Below is the fn-class, that implements the fac-
torial function.
%File: fn.cl
class fn
predicates
classInfo : core::classInfo.
calculate:(string) procedure.
end class fn

% File fn.pro
implement fn
open core
class predicates
fact:(integer, integer)
procedure (i,o). Figure 4.3: The fn class
clauses
classInfo("forms/fn", "1.0").

fact(0, 1) :- !.
fact(N, N*F) :- fact(N-1, F).
calculate(X) :- N= toterm(X),
fact(N, F), stdio::write(F, "\n").
end implement fn

4.7 Notions of Logics: Predicate calculus

Propositional calculus does not have variables or quantifiers. This was fixed
when Friedrich Ludwig Gottlob Frege revealed Predicate Calculus to mankind.
However, Frege’s notation was hard do deal with; modern notation was in-
troduced by Giuseppe Peano. Let us see how Aristotle’s propositions look
like in the modern Predicate Calculus notation.

All a is b ∀X(a(X) → b(X))

Some a is b ∃X|(a(X) ∧ b(X))
Chapter 5

Horn Clauses

5.1 Functions
I am sure you know what a function is. It is possible that you do not know
the mathematical definition of function, but you have a feeling for functions,
acquired by the use of calculators and computer programs or by attending
one of those Pre-Algebra courses. A function has a functor, that is the name
of the function, and arguments. E.g.: sin(X) is a function. Another function
is mod(X, Y ), that gives the remainder of a division. When you want to use
a function, you substitute constant values for the variables, or arguments.
For instance, if you want to find the remainder of 13 divided by 2, you can
type mod(13, 2) into your calculator (if it has this function, of course). If
you want the sinus of π/3, you can type sin(3.1416/3).
One could say that functions are maps between possible values for the
argument and a value in the set of calculation results. Domain is the set of
possible values for the argument. Image is the set of calculation results. In
the case of the sinus function, the elements of the domain is the set of real
numbers. There is an important thing to remember. Mathematicians insist
that a function must produce only one value for a given argument. Then, if a
calculation
√ produces more than one value, it is not a function. For instance,
4 can be 2 or −2. Then, the square root of a number is not a function;
however, you can force it into the straightjack of the function definition, by
saying that only values greater or equal to zero are part of the image.
What about functions of many arguments? For example, the function
max(X, Y ) takes two arguments, and returns the greatest one. In this case,
you can consider that it has only one argument, that is a tuple of values.
Thus, the argument of max(5, 2) is the pair (5, 2). Mathematicians say that
the domain of such a function is the Cartesian product R × R.

41
42 Horn Clauses

There are functions whose functor is placed between its arguments. This
is the case of the arithmetic operations, where one often writes 5 + 7 instead
of +(5, 7).

5.2 Predicates
Predicates are functions whose domain is the set {verum, falsum}, or, if
you do not like the Latin names used by logicians, you can always opt for
the English equivalent: {true, false}. There are a few predicates that are
well known to any person who tried his hand on programming, or even by a
student who is taking Pre-Algebra. Here they are:

X > Y is true if X greater than Y , false otherwise

X < Y is true if X less than Y , false otherwise
X = Y is true if X equal to Y , false otherwise

A predicate with one argument tells of an attribute or feature of its argument.

One could say that such a predicate acts like an adjective. In C, ~X is true
if X is false, and false otherwise. In other computer languages, there are
predicates equivalent to this one. More examples of one slot predicates:

positive(X): true if X is positive, false otherwise

exists("bb.txt"): true if file "bb.txt" exists, false otherwise

A predicate with more than one argument shows that there exists a relation-
ship between its arguments. In the case of X = Y , the relationship is the
equality between X and Y. It would be interesting to have a programming
language, with predicates stating attributes and relationships other than the
few offered by calculators and main stream computer languages. For in-
stance, many people, during their lifetime, have felt a compelling need for
the prediction of one of the predicates listed in figure 5.1, specially the third
one. Prolog is a computer language that was invented to fulfill such a need.

5.3 Solutions
Suppose that you have a predicate city(Name, Point) that gives the coor-
dinates of a city on a map. The city/2 predicate has the domain 1

city:(string Name, pnt Position).

1
N.B. Predicates are functions whose domain can be any Cartesian product, but whose
image is always the set {true, false}.
5.3 Solutions 43

and can be implemented as a database of facts:

city("Salt Lake", pnt(30, 40)).

city("Logan", pnt(100, 120)).
city("Provo", pnt(100, 200)).
city("Yellowstone", pnt(200, 100)).

This predicate checks whether the given position of a given city is correct,
when one is not sure about it. Here are some queries that one could pose to
the city/2 predicate.

city("Salt Lake", pnt(30, 40)) → true

city("Logan", pnt(100, 200)) → false
city("Provo", pnt(100, 200)) → true

There is no question that you could find a use for such a predicate. However,
a predicate that returns the coordinates of a city, given the name, would be
even more useful.

city("Salt Lake", P) → P= pnt(30, 40)

In this new kind of predicate, symbols that start with uppercase are called
variables. Examples of variables: X, Y, Wh, Who, B, A, Xs, Temperature,
Humidity, Rate. Therefore, if you want to know whether a symbol is a
variable, check its first letter. If it is uppercase or even the underscore sign(_),
you are dealing with a variable.

positive(X) is true if X is positive, false otherwise.

rain(Temperature, Pressure, Humidity) is true if it is likely to rain at

a given temperature, pressure, and humidity. For instance

rain(100, 1.2, 90)

returns true if it is likely to rain when your instruments show 100o F ,

1.2 atmospheres, and 90% of relative humidity.

invest(Rate, StandardDeviation, Risk) Given a return Rate, Standard-

Deviation, and acceptable Risk, this predicate returns true if you
should choose an investment.

Figure 5.1: Interesting predicates

44 Horn Clauses

When you use a variable, as P in the query city("Salt Lake", P), you
want to know what one must substitute for P in order to make the predicate
city("Salt Lake", P) true. The answer is P= pnt(30, 40). Sophocles
said that your hand should not be faster than your mind, nor fail to keep
pace with it. Therefore, let us implement the city/2 predicate.

Project Settings. Enter the Project Settings dialog by choosing the option
Project/New from the VDE task menu, and fill it in.

General
Project Name: mapDataBase
UI Strategy: Object-oriented (pfc/GUI)
Target Type: Exe
Base Directory: C:\vispro
Sub-Directory: mapDataBase\

Create Project Item: Form. Select the root node of the Project Tree.
Choose the option File/New in New Package from the task menu. In
the Create Project Item dialog, select the option Form. Fill in the
dialog:

Name: map

Add the following buttons to the new form: Logan, Salt Lake, Provo.

Window Edit. Resize the new form rectangle.The form window must have
a sufficient size to show our “map”. Leave a generous empty space in
the center of the form.

Build/Build This is important: From the task menu, Build/Build the

project, otherwise the system will accuse an error in the next step.

Project Tree/TaskMenu.mnu. Enable File/New.

Project Tree/TaskWindow.win/Code Expert. Add

clauses
onFileNew(S, _MenuTag) :-
X= map::new(S), X:show().

to Menu/TaskMenu/id_file/id_file_new/onFileNew. Build/Build the

project again (it is better to be safe than sorry).
5.3 Solutions 45

% File: draw.cl
class draw
open core, vpiDomains
predicates
classInfo : core::classInfo.
drawThem:(windowHandle, string) procedure.
end class draw

Figure 5.2: mapDataBase/draw.cl

Create class. Create a draw class, as explained in section 4.4. To create

a new class, select the root of the Project Tree, and choose the option
File/New in New Package from the IDE task menu. The class name
is draw and the Create Objects box is off. Build the project, in order
to insert a prototype of the new class into the project tree. Then, edit
draw.cl and draw.pro as shown in figures 5.2 and 5.3.

To call the drawThem predicate using the city-buttons, go to the project

tree, and open the map.frm form, if it is not already open; in the Properties-
dialog, choose the logan_ctl button from the component list box; press the
Event button, and add the following snippet

clauses
onLoganClick(S) = button::defaultAction() :-
Parent= S:getParent(),
P= Parent:getVPIWindow(),
draw::drawThem(P, "Logan").

to the ClickResponder.

Repeat the operation for "Salt Lake" and "Provo". Don’t forget to change
the names of the buttons to logan_ctl and provo_ctl, and saltlake_ctl
respectively; change also the city name "Logan" in "drawThem" to "Provo"
or "Salt Lake". Build the project and execute the program. If you do not
remember how to build and execute the program, take a look at section 1.1.2.
In the new application, choose the option File/New. A new form will pop
up. Whenever you click on a button, the program draws the corresponding
city.
46 Horn Clauses

% File:draw.pro
implement draw
open core, vpiDomains, vpi

constants
className = "draw".
classVersion = "".

class facts
city:(string, pnt).

clauses
classInfo(className, classVersion).

city("Salt Lake", pnt(30, 40)).

city("Logan", pnt(100, 120)).
city("Provo", pnt(100, 80)).
city("Yellowstone", pnt(200, 100)).

drawThem(Win, Name) :- B= brush(pat_solid, color_red),

winSetBrush(Win, B),
city(Name, P), !, P= pnt(X1, Y1),
X2= X1+20, Y2= Y1+20,
drawEllipse(Win, rct(X1, Y1, X2, Y2)).
drawThem(_Win, _Name).
end implement draw

Figure 5.3: mapDataBase/draw.pro

5.4 Multi-solutions
In the last section, you saw predicates that are used for attributing solutions
to their variable, not for checking whether a relationship is true or false.
In the example, the predicate city/2 is used to retrieve only one solution.
However, there are situations that call for more than a single solution. Let
conn/2 be a predicate that establishes a connection between two cities.
conn(pnt(30, 40), pnt(100, 120)).
conn(pnt(100, 120), pnt(100, 200)).
conn(pnt(30, 40), pnt(200, 100)).
..
.
5.4 Multi-solutions 47

You can use it to retrieve all connections between cities, as the example
shows:

conn(pnt(30, 40), W). → W= pnt(100, 120)

→ W=pnt(200, 100)
conn(X, Y). → X=pnt(30, 40)/Y=pnt(100, 120)
→ X= pnt(100, 120)/Y=pnt(100, 200)
→ X=pnt(30, 40)/Y=pnt(200, 100))
Consider the query:

conn(pnt(30, 40), W)?

The answer could be W= pnt(100,120), but it could also be W= pnt(200, 100).

5.4.1 A program that uses multi solution predicates

Let us create a program to demonstrate how nice the multi solution feature
of Prolog is, a feature that it does not share with other languages.

Project Settings. Enter the Project Settings dialog by choosing the option
Project/New from the IDE task menu, and fill it in.

Project Name: drawMap

UI Strategy: Object-oriented GUI (pfc/gui)

Create Project Item: Package. Select the drawMap node of the project
tree. Choose the option File/New in New Package from the task menu.
In the Create Project Item dialog, select Package and fill in the form:

Name: plotter
Parent Directory:

Create Project Item: Form. Select the plotter node of the project tree.
Choose the option File/New in Existing Package from the task menu.
In the Create Project Item dialog, select the option Form, and fill it.

Name: map
Package: plotter.pack (plotter\)

Include the form. From the task menu, choose Build/Build.

Project Tree/TaskMenu.mnu. Enable File/New.

48 Horn Clauses

Project Tree/TaskWindow.win/Code Expert. Add

clauses
onFileNew(S, _MenuTag) :- F= map::new(S), F:show().

to Menu/TaskMenu/id_file/id_file_new/onFileNew

Create class. Create a draw class inside the package plotter, as explained
in section 4.4. Untick the box Create Objects. Put class code in files
draw.cl and draw.pro, as shown in figure 5.4. Build the application.

ProjectTree/map.frm Open the map.frm and add the following snippet

to the PaintResponder:

clauses
onPaint(S, _Rectangle, _GDIObject) :-
W= S:getVPIWindow(),
draw::drawThem(W).

If you build and run this program, you should obtain a window with a map,
such as the one in figure 5.5, whenever you press File/New.

5.5 Logical junctors

I guess that you have had contact with the logical and, that appears in
languages like C or Pascal:

if ((X > 2) && (X < 4)) {. . . }

If P1 and P2 are predicates, the expression “P1 and P2 ” is true if both P1

and P2 are true. A sequence of predicates linked by logical and-s is called
conjunction. In C,

(X>2) && (X<4)

is a conjunction. In Prolog, the predicates of a conjunction are separated by

commas. Then, the expression (X>2) && (X<4) becomes

X>2, X<4

The logical and is called a junctor.

5.6 Implication 49

5.6 Implication
An implication is a junctor represented by the symbol :-, that means if.
Therefore,

drawThem(Win) :- connections(Win), drawCities(Win).

means that you drawThem on Win if you draw connections on Win and
drawCities on Win.

5.7 Horn Clauses

A Horn clause can be a sole predicate. For instance, in the listing below,
there are four one predicate Horn clauses.

city("Salt Lake", pnt(30, 40)).

city("Logan", pnt(100, 120)).
city("Provo", pnt(100, 200)).
city("Yellowstone", pnt(200, 100)).

One predicate Horn clauses are called facts. In the example, the facts state
a relationship between a city and its coordinates. The domain of these pred-
icates is a set of pairs, each pair containing the name of the city and its
coordinates. A Horn clause can also take the form

H :- T1, T2, T3. . .

where Ti and H are predicates. Thus,

drawThem(Win) :- connections(Win),
drawCities(Win).

is an instance of Horn clause. In a Horn clause, the part that comes before
the :- sign is called the head. The part that comes after the sign :- is the
tail. In the example, the head is drawThem(W), and the tail is

connections(Win), drawCities(Win)

A set of Horn clauses with the same head defines a predicate. For instance,
the four city Horn clauses define the city/2 predicate, and

drawThem(Win) :- connections(Win), drawCities(Win).

defines drawThem/1.
50 Horn Clauses

% File: draw.cl
class draw
open core, vpiDomains
predicates
drawThem:(windowHandle) procedure.
end class draw

% File: draw.pro
implement draw
open core, vpiDomains, vpi
class facts
city:(string Name, pnt Position).
conn:(pnt, pnt).
class predicates
connections:( windowHandle).
drawCities:(windowHandle).
clauses
city("Salt Lake", pnt(30, 40)).
city("Logan", pnt(100, 120)).
city("Provo", pnt(100, 160)).
city("Yellowstone", pnt(200, 100)).

conn(pnt(30, 40) , pnt(100, 120)).

conn(pnt(100, 120), pnt(100, 160)).
conn(pnt(30, 40), pnt(200, 100)).

drawCities(W) :- city(N, P), P= pnt(X1, Y1), X2= X1+10, Y2= Y1+10,

drawEllipse(W, rct(X1, Y1, X2, Y2)),
drawText(W, X1, Y1, N), fail.
drawCities(_Win).

connections(Win) :- conn(P1, P2), drawLine(Win, P1, P2), fail.

connections(_Win).

drawThem(Win) :- connections(Win), drawCities(Win).

end implement draw

Figure 5.4: draw.cl and draw.pro

5.8 Declarations 51

Figure 5.5: Utah cities

5.8 Declarations
The definition of a predicate is not complete without type/flow pattern dec-
laration. Here is the type and flow (or mode) declaration of drawThem/1.

predicates
drawThem:(windowHandle) procedure (i).

The type declaration says that the argument Win of drawThem(Win) is of

type windowHandle. The mode declaration states that the argument of
drawThem/1 must be an input, i.e., it is a constant, not a free variable.
In this case, the predicate inputs data from its argument. In general, you do
not need to provide mode declaration. The compiler will infer the modes. If
it fails to perform the inference, it will issue an error message, and you can
add a mode declaration. If a predicate is supposed to be used only inside
its class, it is declared as class predicate. Examples:

class predicates
connections:( windowHandle).
drawCities:(windowHandle).

One predicate Horn clauses may be declared as facts. For example:

class facts
city:(string Name, pnt Position).
conn:(pnt, pnt).
52 Horn Clauses

Later on, you will see that facts can be asserted or retracted from the data
base. The arguments of conn/2 are a pair of pnt values. The type pnt is
defined in class vpiDomains. To use it in class draw, I have two options. I
can spell out the class to which pnt belongs
class facts
conn:(vpiDomains::pnt, vpi::Domains::pnt).
or else, I can open that class inside draw. I have chosen the second option:
open core, vpiDomains, vpi

Determinism declarations
To declare that a predicate has a single solution, or many solutions, one uses
the following keywords:

determ. A determ predicate can fail, or succeed with one solution.

procedure. This kind of predicate always succeeds, and has a single solu-
tion. The predicates connections/1 and drawCities/1 defined in figure 5.4
are procedures, and could have been declared thus:
class predicates
connections:( windowHandle) procedure (i).
drawCities:(windowHandle) procedure (i).

multi. A multi predicate never fails, and has many solutions.

nondeterm. A nondeterm predicate can fail, or succeed in many ways.

The facts city/2 and connection/2 are nondeterm, and accept the following
declaration:
class facts
city:(string Name, pnt Position) nondeterm.
conn:(pnt, pnt) nondeterm.
If a predicate has many solutions, and one of its solution fails to satisfy a
predicate in a conjunction, Prolog backtracks, and offer another solution in
an effort to satisfy the conjunction. Consider the Horn clause:
connections(Win) :- conn(P1, P2),
drawLine(Win, P1, P2), fail.
connections(_Win).
5.9 Drawing predicates 53

The nondeterm fact conn(P1, P2) provides two points, that are used by the
procedure drawLine(Win, P1, P2) to draw a straight line. Then, Prolog
tries to satisfy the predicate fail, that always fails, as its name shows.
Therefore, Prolog backtracks, and tries another solution, until it runs out
of options; then it tries the second clause of connection/1, which always
succeeds.

5.9 Drawing predicates

My sixteen year old son thinks that the only use for computer languages is
games, drawings, graphics, etc. If you are of the same opinion, you must
learn drawing predicates in depth. They require a handle2 to a window that
will support figures and plots. This is how you may obtain the handle:

clauses
onPaint(S, _Rectangle, _GDIObject) :-
W= S:getVPIWindow(), draw::drawThem(W).

In the draw class, the handle W is passed around. For instance, in the clause

drawThem(Win) :- connections(Win), drawCities(Win).

it is passed to connections/1, where it is the first argument of drawLine/3:

connections(Win) :- conn(P1, P2), drawLine(Win, P1, P2), fail.

The predicate drawLine(Win, P1, P2) draws a line from P1 to P2 on the

window Win. As you know, P1 and P2 are points, such as pnt(10, 20).
Another predicate that you are acquainted with is

drawEllipse(W, rct(X1, Y1, X2, Y2))

that draws an ellipse on the window W. The ellipse is inscribed inside the
rectangle rct(X1, Y1, X2, Y2), where X1, Y1 are the coordinates of the
upper left corner, and X2, Y2 are the coordinates of the lower right corner.

5.10 GDI Object

In the last example, one performed the drawing from the onPaint event
handler. When this happens, it may be a good idea to use methods from the
so called GDI Object. The example that follows shows how this works.
2
You do not need to know what a handle is for the time being.
54 Horn Clauses

Project Settings. Create the following project:

Project Name: drawMapObj

UI Strategy: Object-oriented GUI (pfc/gui)
Target type: Exe
Base Directory: C:\vispro

Create a package: plotter.

Create a form inside plotter: map.

Project Tree/TaskMenu.mnu. Enable File/New. From the task menu,

Build/Build the application, in order to include the map into the project.

Project Tree/TaskWindow.win/Code Expert. Add

clauses
onFileNew(S, _MenuTag) :-
F= map::new(S), F:show().

to Menu/TaskMenu/id_file/id_file_new/onFileNew

Create class. Create a draw class inside package plotter. Remember to

untick “Create objects”. You will find the new version of class draw in
figure 5.6. Build the application.

Add clauses
onPaint(_S, _Rectangle, GDIObject) :-
draw::drawThem(GDIObject).
to the PaintResponder of the map.frm.
5.10 GDI Object 55

%File: draw.cl
class draw
open core
predicates
drawThem:(windowGDI).
end class draw

% File: draw.pro
implement draw
open core, vpiDomains, vpi
class facts
city:(string Name, pnt Position).
conn:(pnt, pnt).
class predicates
connections:( windowGDI).
drawCities:(windowGDI).
clauses
city("Salt Lake", pnt(30, 40)).
city("Logan", pnt(100, 120)).
city("Provo", pnt(100, 160)).

conn(pnt(30, 40) , pnt(100, 120)).

conn(pnt(100, 120), pnt(100, 160)).

drawCities(W) :- city(N, P), P= pnt(X1, Y1),

X2= X1+10, Y2= Y1+10,
W:drawEllipse(rct(X1, Y1, X2, Y2)),
W:drawText(pnt(X1, Y1), N), fail.
drawCities(_Win).

connections(W) :- conn(P1, P2), W:drawLine(P1, P2), fail.

connections(_W).

drawThem(Win) :- connections(Win), drawCities(Win).

end implement draw

Figure 5.6: Class draw

56 Horn Clauses

5.11 Examples
There are three examples in this chapter: mapDataBase draws the position
of the three main cities of Utah, when one presses the appropriate button;
drawMap draws a coarse road map of Utah; drawMapObj is the GDI version
of drawMap.

5.12 Notions of Logics: Meaning of Horn clauses

A Horn clause has the form

H : −T1 , T2 , T3 . . .

The meaning of this clause is

If T1 &T2 &T3 . . . are true,

then H is also true.

You have seen before that this statement is equivalent to

H is true or the conjunction T1 &T2 &T3 . . . is not true.

This last interpretation sheds light on the meaning of the neck symbol, where
the semicolon stands for the connective or, and the minus sign stands for the
logical negation. Then,

H : −T

means: H is true, or not T. To finish our study of the Prolog connectives, let
us remind you that the comma stands for and, the semicolon stands for or,
and -> stands for if. . . then.
Chapter 6

Console Applications

Since people like graphic applications, we started with them. However, they
add details that sway your attention from what really matters. Thus, let us
see a few basic programming schemes without resorting to a graphic interface.

6.1 Cut
What should you do if you don’t want the system either to backtrack or try
another clause after finding a solution? In this case, you insert an exclamation
mark at a given point in the tail of the Horn clause. After finding the
exclamation mark, the system interrupts its search for new solutions. The
exclamation mark is called cut. Let us illustrate the use of cuts with an
algorithm that is very popular among Prolog programmers. Let us assume
that you want to write a predicate that finds the factorial of a number.
Mathematicians define factorial as:
factorial(0) → 1
factorial(n) → n × factorial(n − 1)
Using Horn clauses, this definition becomes:
fact(N, 1) :- N<1, !.
fact(N, N*F1) :- fact(N-1, F1).
The cut prevents Prolog from trying the second clause for N = 0. In other
words, if you pose the query
fact(0, F)?
the program succeeds in using the first clause for N = 0, and returns F = 1.
Without the cut, it will assume that the second clause of the definition also

57
58 Console Applications

provides a solution. Then, it will try to use it, and will go astray. The
cut ensures that factorial is a function and, as such, maps each value in
the domain to one and only one value in the image set. To implement the
factorial function, follow the following directive:
Create a New Project. Choose the option Project/New and fill the Project
Settings dialog thus:
General
Project Name: facfun
UI Strategy: console
Pay attention to the fact that we will use the console strategy, not GUI.
Build. Choose the option Build/Build from the task menu to add a pro-
totype of class facfun into the project tree. Edit facfun.pro as shown
below. Build the project again, and execute it using Run in Window.
Write a number to the prompt, and you will get its factorial. NB: To
test a console program, use Run in Window, not Execute.
%File: main.pro
implement main

class predicates
fact:(integer N, integer Res) procedure (i,o).
clauses
classinfo("facfun", "1.0").

fact(N, 1) :- N<1, !.
fact(N, N*F) :- fact(N-1, F).

run():- console::init(),
fact(stdio::read(), F), stdio::write(F), stdio::nl.
end implement main

goal
mainExe::run(main::run).

6.2 Lists
There is a book about Lisp that says: A list is a an ordered sequence of
elements, where ordered means that the order matters. In Prolog, a list is
6.2 Lists 59

put between brackets, and is supposed to have a head and a tail:

List Type Head Tail

[3, 4, 5, 6, 7] integer 3 [4, 5, 6, 7]
["wo3", "ni3", "ta1"] string* "wo3" ["ni3", "ta1"]
[4] integer* 4 []
[3.4, 5.6, 2.3] real* 3.4 [5.6, 2.3]
You can match a pattern of variables with a list. For instance, if you match

[X|Xs]

with the list [3.4, 5.6, 2.3] you will get X = 3.4 and Xs = [5.6, 2.3], i.e., X
matches the head of the list and Xs matches the tail. Of course, you can use
another pair of variables instead of X and Xs in the pattern [X|Xs]. Thus,
[A|B], [X|L], [Head|T ail], [F irst|Rest] and [P |Q] are equivalent patterns.
More examples of matches for the pattern [X|Xs]:

Pattern List X Xs
[X|Xs] [3.4, 5.6, 2.3] X = 3.4 Xs = [5.6, 2.3]
[X|Xs] [5.6, 2.3] X = 5.6 Xs=[2.3]
[X|Xs] [2.3] X = 2.3 Xs = []
[X|Xs] [] Does not match.

As you can see, the pattern [X|Xs] only matches lists with at least one
element. Here is a pattern that only matches lists with at least two elements:

Pattern List X1, X2 Xs

[X1, X2|Xs] [3, 5, 2, 7] X1 = 3, X2 = 5 Xs = [2, 7]
[X1, X2|Xs] [2, 7] X1 = 2, X2 = 7 Xs=[]
[X1, X2|Xs] [7] Does not match.
[X1, X2|Xs] [] Does not match.
60 Console Applications

Let avg be a small project that calculates the length of a list of real numbers.
As in the case of factorial, make sure that avg is a console application.
General
Project Name: avg
UI Strategy: console
Build the project to insert the avg class into the project tree. Then, edit
avg.pro as shown below. Execute using Run in Window, as before.
/* File: main.pro */
implement main
open core, console
domains
rList= real*.
class predicates
len:(rList, real) procedure (i, o).
clauses
classInfo("avg", "1.0").

len([], 0) :- !.
len([_X|Xs], 1.0+S) :- len(Xs, S).

run():- console::init(), List= read(),

len(List, A),
write(A), nl.
end implement main
goal
mainExe::run(main::run).

Let us examine the concepts behind this short program. The first thing that
you have done is to create a list domain:
domains
rList= real*.
Then, you define a predicate len/2, that inputs a list from the first argument,
and outputs its length to the second argument. Finally, you define a run()
predicate to test your program:
run():- console::init(), % Initialize the console.
hasdomain(rList, List), % Create a rList variable.
List= read(), % Read List from console.
len(List, A), % Find the length of List.
write(A), nl. % Output the length.
6.2 Lists 61

The domain declaration is handy if you are dealing with complex algebraic
data types. However, it is not required for such a simple structure as a list of
reals. Below you will find the same program without the domain declaration.
/* File: main.pro */
implement main
open core, console
class predicates
len:(real*, real) procedure (i, o).
clauses
classInfo("avg", "1.0").

len([], 0) :- !.
len([_X|Xs], 1.0+S) :- len(Xs, S).

run():- console::init(), List= read(),

len(List, A),
write(A), nl.
end implement main
goal
mainExe::run(main::run).

This program has a drawback: One can use len/2 only for calculating the
length of a list of reals. It would be nice if len/2 could take any kind of list.
This should be possible if one substitutes a type variable for real*. However,
before checking whether this scheme really works, it is necessary to devise
a way of informing L= read() which will be the eventual type of the input
list, and feed the information to len/2. Fortunately there is a predicate that
creates a variable for any given type. You will discover below how to use it.
implement main /* in file main.pro */
open core, console
class predicates
len:(Element*, real) procedure (i, o).
clauses
classInfo("avg", "1.0").
len([], 0) :- !.
len([_X|Xs], 1.0+S) :- len(Xs, S).

run():- console::init(),
hasdomain(real_list, L), L= read(),
len(L, A), write(A), nl.
end implement main
goal mainExe::run(main::run).
62 Console Applications

The next step in our scheme of things is to insert sum/2 clauses to main.pro.
sum([], 0) :- !.
sum([X|Xs], S+X) :- sum(Xs, S).
Let us examine what happens if one calls sum([3.4, 5.6, 2.3], S).
1. sum([3.4, 5.6, 2.3], S), matches
sum([X|Xs], S+X) :- sum(Xs, S), with X=3.4, Xs=[5.6, 2.3],
yielding sum([3.4, 5.6, 2.3], S[5.6,2.3] + 3.4) :- sum([5.6, 2.3], S[5.6,2.3] )

2. sum([5.6, 2.3], S[5.6,2.3] ), matches

sum([X|Xs], S+X) :- sum(Xs, S), with X=5.6, Xs=[2.3],
yielding sum([5.6, 2.3], S[2.3] + 5.6) :- sum([2.3], S[2.3] )

3. sum([2.3], S[2.3] ) matches

sum([X|Xs], S+X) :- sum(Xs, S), with X=2.3, Xs=[],
yielding sum([2.3], S[] + 2.3) :- sum([], S[] )

4. sum([], S[] ) matches sum([], 0.0) yielding S[] = 0.

After reaching the bottom of the list, the computer must roll back to the start
of the computation. What is worse, it must store the values of every X that
it finds on its way down, in order to perform the addition S + X when it rolls
back. The traditional way to prevent the roll back is to use accumulators.
Here is a definition that uses an accumulator to sum the elements of a list:
add([], A, A).
add([X|Xs], A, S) :- add(Xs, X+A, S).
Let us see how the computer calculates this second version of list addition.
1. add([3,4, 5.6, 2.3], 0.0, S)
matches add([X|Xs], A, S) :- add(Xs, X+A, S)
yielding add([5.6, 2.3], 0+3.4, S)

2. add([5.6, 2.3], 0.0+3.4, S)

matches add([X|Xs], A, S) :- add(Xs, X+A, S)
yielding add([2.3], 0+3.4+5.6, S)

3. add([2.3], 0.0+3.4+5.6, S)
matches add([X|Xs], A, S) :- add(Xs, X+A, S)
yielding add([ ], 0+3.4+5.6+2.3, S)
that matches add([], A, A) yielding S = 11.3.
You could use add/3 to calculate the average value of a list of numbers.
6.2 Lists 63

len([], 0) :- !.
len([_X|Xs], 1.0+S) :- len(Xs, S).

add([], A, A) :- !.
add([X|Xs], A, S) :- add(Xs, X+A, S).

sum(Xs, S) :- add(Xs, 0.0, S).

avg(Xs, A/L) :- sum(Xs, A), len(Xs, L).

The above program goes through the list twice, first to calculate the sum, and
second to calculate the length. Using two accumulators, you can calculate
the length and the sum of the list together.

%File: main.pro
implement main
open core, console

class predicates
avg:(real*, real, real, real) procedure (i, i, i, o).
clauses
classInfo("avg", "1.0").

avg([], S, L, S/L).
avg([X|Xs], S, L, A) :-
avg( Xs,
X+S, % add X to the sum acc
L+1.0, % increments the length acc
A).

run():- console::init(),
List= read(),
avg(List, 0, 0, A),
write(A), nl.
end implement main

goal
mainExe::run(main::run).

The above listing shows a program to calculate the average value of a list of
real numbers. One makes use of two accumulators, one for the sum, and the
other for the number of elements.
64 Console Applications

6.3 List schemes

Let us see a few important schemes of list programming. I thought that was
obvious that I didn’t invent these schemes. However, since a few persons
wrote me asking questions about their origin, I added two entries to the
reference list: [John Hughes], and [Wadler & Bird].

Reduction
The scheme used to calculate the sum of the elements of a list is called
reduction, as it reduces the dimension of the input. In fact, lists can be
thought of as one-dimensional data, and the sum is a zero-dimensional datum.
There are two ways of performing a reduction: recursive and iterative.

%Recursive reduction
class predicates
sum:(real*, real) procedure (i, o).
clauses
sum([], 0) :- !.
sum([X|Xs], S+X) :- sum(Xs, S).

The term iterative comes from the Latin word ITERUM, that means again.
You can also imagine that it comes from ITER/ITINERIS= way, road. An
iterative program goes through the code again, and again.

%Iterative reduction
red([], A, A) :- !.
red([X|Xs], A, S) :- red(Xs, X+A, S).
sumation(Xs, S) :- red(Xs, 0.0, S).

Let us assume that, instead of performing an addition, you want to perform

a multiplication. In this case, you can write the following program:

%Iterative reduction
red([], A, A) :- !.
red([X|Xs], A, S) :- red(Xs, X*A, S).
product(Xs, S) :- red(Xs, 0.0, S).

The two pieces of code are identical, except for a single detail: one of them
has X+A in the second argument of the iterative call of the predicate red/3,
while the other has X*A. [Wadler & Bird] and [John Hughes] suggests that
one capture this kind of similarity in higher order predicates or functions.
Before them, when he received the Turing award, [John Backus] made the
6.3 List schemes 65

same proposition. Let us see that one can use this prestigious method of
programming in Visual Prolog.
In the program of figure 6.1, I have declared a domain that is a two-place
function. In the section containing the class predicates, I have defined a few
functions belonging to this domain. Finally, I defined a red/4 predicate that
capture the structure of the reduction operation.

implement main
open core
domains
pp2 = (real Argument1, real Argument2) -> real ReturnType.
clauses
classInfo("main", "hi_order").

class predicates
sum:pp2.
prod:pp2.

red:(pp2, real*, real, real) procedure (i, i, i, o).

clauses
sum(X, Y)= Z :- Z=X+Y.
prod(X, Y)= Z :- Z=X*Y.

red(_P, [], A, A) :- !.
red(P, [X|Xs], A, Ans) :- red(P, Xs, P(X, A), Ans).

run():- console::init(),
red(prod, [1,2,3,4,5], 1, X),
stdio::write(X), stdio::nl,
succeed().
end implement main
goal
mainExe::run(main::run).

Figure 6.1: Reduce in Prolog

66 Console Applications

ZIP
There is a device used for continuous clothing closure invented by the Cana-
dian electrical engineer Gideon Sundback. The device drives one crazy when
it does not work, and is the cause of so many accidents with young boys,
that doctors and nurses need special training to deal with it.
class predicates
dotproduct:(real*, real*, real) procedure (i, i, o).
clauses
dotproduct([], _, 0) :- !.
dotproduct(_, [], 0) :- !.
dotproduct([X|Xs], [Y|Ys], X*Y+R) :- dotproduct(Xs, Ys, R).
The example shows a zipping scheme followed by a reduction step, in order
to perform the dot product of two lists. In the expression X*Y+R, one uses
the multiplication to close one clasp of the zipper.
We can capture the zipping scheme in a higher-order predicate, the same
way we have captured the reducing scheme. In fact, we can close each clasp
of the zipper with the help of the same type of functions that we have used
to perform a reduction step. In figure 6.2, you can find a (not very efficient)
definition of zip in Prolog. I added this listing because more than one person
wrote me requesting for an in-depth treatment of this important topic; I be-
lieve that I cannot provide an in-depth treatment, since this would go beyond
the scope of this book; after all, my goal is to avoid subjects that present
steep learning curves to beginners; however, the examples in figures 6.1,
and 6.2 should be enough for showing how to write higher-order predicates
in Visual Prolog. People who want to explore the possibilities of higher-order
Logics, and functions may refer themselves to books like [Wadler & Bird], or
in papers like [John Hughes].
The predicate zip has four arguments. The first argument is a two-
place function; in listing 6.2, I provide two of such functions, but you can
define others. The second and third arguments contain the lists to be zipped
together. The fourth argument contains the output of the operation. The
definition of zip/4 is simple:
zip(_P, [], _, []) :- !.
zip(_P, _, [], []) :- !.
zip(P, [X|Xs], [Y|Ys], [Z|As]) :- Z= P(X, Y),
zip(P, Xs, Ys, As).
If either input list becomes empty, the first or the second clause stop the
recursion, producing the empty list as output; otherwise, the third clause
6.3 List schemes 67

clasps the heads of the two input lists, and proceeds to the rest of the input
lists. I believe that this should be enough for a starting point.

implement main
open core
domains
pp2 = (real Argument1, real Argument2) -> real ReturnType.
clauses
classInfo("main", "zip").
class predicates
sum:pp2.
prod:pp2.
red:(pp2, real*, real, real) procedure (i, i, i, o).
zip:(pp2, real*, real*, real*) procedure (i, i, i, o).
dotprod:(real*, real*, real) procedure (i, i, o).
clauses
sum(X, Y)= Z :- Z=X+Y.
prod(X, Y)= Z :- Z=X*Y.

zip(_P, [], _, []) :- !.

zip(_P, _, [], []) :- !.
zip(P, [X|Xs], [Y|Ys], [Z|As]) :- Z= P(X, Y), zip(P, Xs, Ys, As).

red(_P, [], A, A) :- !.
red(P, [X|Xs], A, Ans) :- red(P, Xs, P(X, A), Ans).

dotprod(Xs, Ys, D) :- zip(prod, Xs, Ys, Z), red(sum, Z, 0, D).

run():- console::init(),
V1= [1,2,3,4,5], V2= [2,2,2,2,2],
dotprod( V1, V2, X),
stdio::write(X), stdio::nl.
end implement main
goal
mainExe::run(main::run).

Figure 6.2: Zip in Prolog

68 Console Applications

6.4 String operations

You will find a few examples of string operations below. These examples
should be enough to show you how to deal with such data structures. You
can find more operations using the help of Visual Prolog.
implement main
open core, string
class predicates
tokenize:(string, string_list) procedure (i, o).
clauses
classInfo("main", "string_example").

tokenize(S, [T|Ts]) :-
frontToken(S, T, R), !, tokenize(R, Ts).
tokenize(_, []).

run():- console::init(),
L= ["it ", "was ", "in ", "the ",
"bleak ", "December!"],
S= concatList(L), UCase= toUpperCase(S),
RR= string::concat("case: ", Ucase),
R1= string::concat("It is ", "upper ", RR),
stdio::write(R1), stdio::nl, tokenize(R1, Us),
stdio::write(Us), stdio::nl.
end implement main

goal
mainExe::run(main::run).
Try to understand how frontToken works, because it is a very useful pred-
icate. One uses it to break a string into tokens, in a process called lexical
analysis. For instance, if you execute
frontToken("break December", T, R)
you will get T="break", and R=" December".
One often has to transform a string representation of a term into an oper-
ational representation. In this case, s/he can use the function toTerm. In the
example below, Sx= stdio::readLine() reads a string from the command
line into Sx; then toTerm transforms the string representation into a real
number; a call to hasdomain(real, IX) makes sure that toTerm will yield
a real number.
6.4 String operations 69

implement main
clauses
classinfo("main", "toTerm_example").
run():- console::init(),
Sx= stdio::readLine(),
hasdomain(real, IX),
IX= toTerm(Sx),
stdio::write(IX^2).
end implement main
goal mainExe::run(main::run).

This scheme works even for lists, and other complex data structures, as you
can see below. Unfortunately, hasdomain does not accept the star notation
for list types; therefore, you must declare a list domain, or use a pre-defined
domain, like core::integer_list.

implement main
clauses
classInfo("main", "hasdomain_example").

run():- console::init(),
hasdomain(core::integer_list, Xs),
Xs= toTerm(stdio::readLine()),
stdio::write(Xs), stdio::nl.
end implement main
goal mainExe::run(main::run).

You have learned how to transform a string representation of a data type

to a Prolog term. Now, let us see how to do the inverse operation, i.e.,
to transform a term to its string representation. If you need the standard
representation of a term, then you can use the predicate toString.

implement main
clauses
classInfo("main", "toString_example").

run():- console::init(),
Xs= [3.4, 2, 5, 8],
Str= toString(Xs),
Msg= string::concat("String representation: ", Str),
stdio::write(Msg), stdio::nl.
end implement main
goal mainExe::run(main::run).
70 Console Applications

If you want to format numbers and other simple terms nicely into a string,
you can use the string::format format function; an example is given below.
implement main
clauses
classInfo("main", "formatting").
run():- console::init(),
Str1= string::format("%8.3f\n %10.1e\n", 3.45678, 35678.0),
Str2= string::format("%d\n %10d\n", 456, 18),
Str3= string::format("%-10d\n %010d\n", 18, 18),
stdio::write(Str1, Str2, Str3).
end implement main
goal
mainExe::run(main::run).

In the above example, the format "%8.3f\n means that I want to repre-
sent a real number, and a carriage return; the field width is 8, and the number
must be represented with three digits of precision. The format "%010d\n"
requires an integer right justified on a field of width 10; the field empty po-
sitions must be filled in with zeroes. The format "%-10d\n" specifies the
representation of a left justified integer on a field of width 10; the minus
sign assures left justification; right justification is the default. The format
"%10.1e\n" specifies scientific notation for real numbers. Below, you can see
the output of the example.
3.457
3.6e+004
456
18
18
0000000018
Below, you will find a listing of the data types accepted by the format string.
Note that when converting real values into a text representation they are
truncated and rounded with 17 digits, unless another precision is specified.
f Format reals in fixed-decimal notation (such as 123.4).

e Format reals in exponential notation (such as 1.234e+002).

g Format reals in the shortest of f and e format.

d Format as a signed decimal number.

u Format as an unsigned integer.

6.4 String operations 71

x Format as a hexadecimal number.

o Format as an octal number.

c Format as a char.

B Format as the Visual Prolog binary type.

R Format as a database reference number.

P Format as a procedure parameter.

s Format as a string.

6.4.1 Useful String Predicates

In this section, you will find a listing of string predicates that you may find
useful in the development of applications; adjustBehaviour, adjustSide,
and caseSensitivity can be considered as having the following definitions:

domains
adjustBehaviour =
expand();
cutRear();
cutOpposite().
adjustSide = left(); right().
caseSensitivity =
caseSensitive();
caseInsensitive();
casePreserve().

adjust : (string Src, charCount FieldSize, adjustSide Side) ->

string AdjustedString.

adjust : (string Src, charCount FieldSize, string Padding,

adjustSide Side) -> string AdjustedString.

adjustLeft : (string Src, charCount FieldSize) ->

string AdjustedString.

adjustLeft : (string Src, charCount FieldSize,

string Padding) -> string AdjustedString.
72 Console Applications

adjustRight : (string Src, charCount FieldSize) ->

string AdjustedString.

adjustRight : (string Src, charCount FieldSize,

string Padding) -> string AdjustedString.

adjustRight : (string Src, charCount FieldSize, string Padding,

adjustBehaviour Behaviour) -> string AdjustedString.

/*********************************
Project Name: adjust_example
UI Strategy: console
***********************************/

implement main
clauses
classInfo("main", "adjust_example").
run():- console::init(),
FstLn=
"Rage -- Goddess, sing the rage of Peleus’ son Achilles,",
Str= string::adjust(FstLn, 60, "*", string::right),
stdio::write(Str), stdio::nl.
end implement main
goal mainExe::run(main::run).

*****Rage -- Goddess, sing the rage of Peleus’ son Achilles,

concat : (string First, string Last) ->

string Output procedure (i,i).

concatList : (core::string_list Source) ->

string Output procedure (i).

concatWithDelimiter : (core::string_list Source,

string Delimiter) ->
string Output procedure (i,i).

create : (charCount Length) -> string Output procedure (i).

6.4 String operations 73

create : (charCount Length, string Fill) ->

string Output procedure (i,i).

createFromCharList : (core::char_list CharList) ->

string String.

/*********************************
Project Name: stringops
UI Strategy: console
***********************************/
implement main
clauses
classInfo("main", "stringops").

run():-
console::init(),
SndLn= [ "murderous", "doomed",
"that cost the Achaeans countless losses"],
Str= string::concatWithDelimiter(SndLn, ", "),
Rule= string::create(20, "-"),
stdio::write(Str), stdio::nl,
stdio::write(Rule), stdio::nl.
end implement main

goal mainExe::run(main::run).

murderous, doomed, that cost the Achaeans countless losses

--------------------

equalIgnoreCase : (string First, string Second) determ (i,i).

front : (string Source, charCount Position,

string First, string Last) procedure (i,i,o,o).
74 Console Applications

frontChar : (string Source, char First, string Last) determ (i,o,o).

frontToken : (string Source, string Token,

string Remainder) determ (i,o,o).

getCharFromValue : (core::unsigned16 Value) -> char Char.

getCharValue : (char Char) -> core::unsigned16 Value.

hasAlpha : (string Source) determ (i).

hasDecimalDigits : (string Source) determ (i).

hasPrefix : (string Source, string Prefix,

string Rest) determ (i,i,o).

hasSuffix : (string Source, string Suffix,

string Rest) determ (i,i,o).

isLowerCase : (string Source) determ (i).

isUpperCase : (string Source) determ (i).

isWhiteSpace : (string Source) determ.

lastChar : (string Source, string First,

char Last) determ (i,o,o).

length : (string Source) ->

charCount Length procedure (i).

replace : (string Source, string ReplaceWhat,

string ReplaceWith, caseSensitivity Case) ->
string Output procedure

replaceAll : (string Source, string ReplaceWhat,

string ReplaceWith) -> string Output.

replaceAll : (string Source, string ReplaceWhat,

string ReplaceWith,
caseSensitivity Case) ->
string Output.
6.4 String operations 75

search : (string Source, string LookFor) ->

charCount Position determ (i,i).

search : (string Source, string LookFor,

caseSensitivity Case) ->
charCount Position determ (i,i,i).

split : (string Input, string Separators) -> string_list.

split_delimiter : (string Source, string Delimiter) ->

core::string_list List procedure (i,i).

subString : (string Source, charCount Position,

charCount HowLong) ->
string Output procedure (i,i,i).

toLowerCase : (string Source) ->

string Output procedure (i).
/* Convert string characters to lowercase */

toUpperCase : (string Source) ->

string Output procedure (i).
/* Convert string character to uppercase */

trim : (string Source) ->

string Output procedure (i).
/* Removes both leading and trailing
whitespaces from the string. */

trimFront : (string Source) ->

string Output procedure (i).
/* Removes leading whitespaces from the string. */

trimInner : (string Source) ->

string Output procedure (i).
/* Removes groups of whitespaces from the Source line. */

trimRear : (string Source) ->

string Output procedure (i).
/*Removes trailing whitespaces from the string. */
76 Console Applications

/*********************************
Project Name: trim_example
UI Strategy: console
***********************************/
implement main
clauses
classInfo("main", "trim_example").

run():-
console::init(),
Str= " murderous, doomed ",
T= string::trim(Str),
stdio::write(T), stdio::nl,
T1= string::trimInner(T),
stdio::write(T1), stdio::nl.
end implement main

goal mainExe::run(main::run).

murderous, doomed
murderous, doomed

6.5 Notions of Logics: Grammar for Predicates

Let us define informal grammar rules for the Predicate Calculus.

• The symbol p(t1 , t2 , . . . tn−1 , tn ) is called literal. In a literal, p/n is a

n-place symbol distinct from all other symbols of the language, and ti
are terms; a term can be

– Functional symbols, i.e., symbols of the form f (x1 , x2 . . . xn . A

functional symbol may have arity zero, which means that a, x, b,
and rosa can be functional symbols.
– Numerals, strings, and characters can also be terms.

A formula consisting of a sole literal is called an atomic formula.

• If t1 and t2 are terms, then t1 = t2 , t1 > t2 , t1 < t2 , t1 ≥ t2 , and t1 ≤ t2

are literals, and one can use them to create formulæ.
6.5 Notions of Logics: Grammar for Predicates 77

• If P and Q are formulae, then ¬P , P ∧ Q, P ∨ Q, P → Q, and P ≡ Q

are formulæ. Besides this, if P is a formula, and X is a variable, ∀X(P )
(for any X, P is true) and ∃X(P ) (there exists X, that makes P true)
are both formulæ.

• A variable introduced by the symbol ∃ is called existential variable.

Therefore, the variable X in the formula

∃X(sin(X) = 0)

is an existential variable. One can remove all existential variables of a

formular, by substituting constants for them. In the example, one can
substitute π for X:
sin(X) = 0

Clausal form Any sentence of the predicate can be re-written into the
clausal form, i.e., as a disjunction of conjunctions, with no existential quantifier(∃),
and no universal quantifier inside parentheses. Since A ∨ B is a disjunction,
and A ∧ B is a conjunction, any formula can be re-written as

∀X∀Y ∀Z . . . ( (L11 ∨ L12 ∨ L13 . . .) ∧

(L21 ∨ L12 ∨ L13 . . .) ∧
(L31 ∨ L12 ∨ L13 . . .) ∧ . . .)

where Lij are literals, or negations of literals. We have already seen how to
remove existential quantifiers from a formula. Let us examine the remaining
necessary steps to obtain the clausal formula.
1. Remove implications, and equivalences.

α→β ¬α ∨ β
α≡β (α ∧ β) ∨ (¬α ∧ ¬β)

2. Distribute negations. Use truth tables to prove that

¬(α ∨ β) ≡ ¬α ∧ ¬β

and that
¬(α ∧ β) ≡ ¬α ∨ ¬β
Then, use these results to make the following re-writing operations:

¬(α ∨ β) ¬α ∧ ¬β
¬(α ∧ β) ¬α ∨ ¬β
78 Console Applications

3. Thoraf Albert Skolem (1887-1963) is a Norwegian mathematician, who

invented the method of removing existential variables named after him.
Skolemization substitutes constants for existential variables. E. g.

∃X(p(X)) p(s0 )
∀X(man(X) → ∃W (loves(W, X))) ∀X(man(X) → loves(w(X), X))

4. Distribute conjunctions using the following re-write rules:

(α ∧ β) ∨ γ (α ∨ β) ∧ (β ∨ γ)
¬(α ∧ β) ¬α ∨ ¬β
Chapter 7

Grammars

The philosopher Wittgenstein, in his Tratactus Logicus Philosophicus, ex-

pressed the opinion that the World is a state of things. Things and objects
have attributes, like position, integrity1 , color, etc. The set of values that
these attributes take is called state. For example, according to Euclid, a
point has no attribute other than position. Then, the state of a point can be
given by its coordinates.
The first sentence of the Tratactus is: The world is all, that is the case.
The last one is: Whereof one cannot speak, thereof one must be silent. It is
practically impossible to create a data structure to represent Wittgenstein’s
ideas on a computer. However, if ever devised by a linguistic genius, that
data structure will certainly have three constructors: the world node, the
case node, and the silence node. Here is a semantic class whose sole use is
to export the semantic domain of the Tratactus:

class semantic
open core

domains
category= art; nom; cop; rel.
tree= case(category, string); world(tree, tree); silence.

predicates
classInfo : core::classInfo.
end class semantic

There is a novelty in this listing: The use of constructors of data structures.

A constructor has the shape of a function, i.e., it has a functor and arguments.
1
Is it whole, or broken?

79
80 Grammars

For instance, the constructor case/2 has two arguments, a grammatical cat-
egory and a string representing a token.

7.1 Parsing grammar

Acting is to change the attributes of the objects. Thus, the result of an act is
a change of state. For instance, if a predicate parses a list of words, the world
will undergo a change of state. In the final state, the parser has consumed
part of the list. To understand this point, let us create a german class to
parse the first sentence of the Tratactus. Using it as a model, you can write
a more powerful class to parse the whole book. It is a cinch, because the
Tratactus is very short.
%File: german.cl
class german
open core, semantic
predicates
classInfo : core::classInfo.
article:(tree, string*, string*) nondeterm (o, i, o).
noun:(tree, string*, string*) nondeterm (o, i, o).
nounphr:(tree, string*, string*) nondeterm (o, i, o).
copula:(tree, string*, string*) nondeterm (o, i, o).
phr:(tree, string*, string*) nondeterm (o, i, o).
end class german

%File: german.pro
implement german
open core, semantic
clauses
classInfo("german", "1.0").

article(case(art, ""), ["die"|Rest], Rest).

article(case(art, ""), ["der"|Rest], Rest).

noun(case(nom, "Welt"), ["Welt"|R], R).

noun(case(nom, "Fall"), ["Fall"|R], R).

copula(world(case(cop, "ist"), T), ["ist"|R1], R) :-

nounphr(T, R1, R).
7.2 Generating grammar 81

nounphr( world(T1, T2), S, End) :-

article(T1, S, S1), noun(T2, S1, End).
nounphr( case(nom, "alles"), ["alles"|R], R).

phr(world(T1, T2), Start, End) :-

nounphr(T1, Start, S1), copula(T2, S1, End).
end implement german

If you pay attention to the declaration of article, you will notice that its
flow pattern is (o, i, o), id est, it inputs a list of words, and outputs a parse
tree2 and the partially consumed input list. The definition of noun works
exactly like article. An example is worth one thousand words; let us feed
["die", "Welt", "ist", "alles"] to article/3.

article(T1, ["die", "Welt", "ist", "alles"], S1).

This call will match the first clause of article/3, with

Rest=["Welt", "ist", "alles"].

The output will be T1= case(art, ""), and S1= ["Welt", "ist", "alles"].
Next, let us feed S1= ["Welt", "ist", "alles"] to noun/3.

noun(T2, ["Welt", "ist", "alles"], End)

The above call will match the first clause of noun/3, with R=["ist", "alles"].
The output will be T2=case(nom, "Welt"), and End=["ist", "alles"].
The definition nounphr makes these two calls in succession, parsing a noun
phrase like ["die", "Welt"]; copula/3, and phr/3 behaves like nounphr.

7.2 Generating grammar

While the German parse takes a sentence, and outputs a node, the English
parse, that has been given below, takes a node, in order to generate a sen-
tence. One could say that the node is the meaning of the sentence. The
German parse consumes a German sentence, and outputs its meaning. The
English parser inputs the meaning, and produces an English translation.
2
If you do not know these linguistic terms, take a look at a book about natural language
processing, or compiler construction, whichever pleases you most.
82 Grammars

%File: english.cl
class english
open core, semantic
predicates
classInfo : core::classInfo.
article:(tree, string*) nondeterm (i,o)(o,o).
noun:(tree, string*) nondeterm (i,o)(o,o).
nounphr:(tree, string*) nondeterm (i,o)(o,o).
copula:(tree, string*) nondeterm (o,o)(i, o).
phr:(tree, string*) nondeterm (i,o).
end class english

% File: english.pro
implement english
open core, semantic
clauses
classInfo("english", "1.0").

article(case(art, ""), ["the"]).

noun(case(nom, "Welt"), ["world"]).

noun(case(nom, "Fall"), ["case"]).
noun(case(nom, "alles"), ["all"]).

nounphr( world(T1, T2), list::append(A,N)) :-

article(T1, A), noun(T2, N).
nounphr( case(nom, "alles"), ["all"]).

copula(world(case(cop, "ist"), T), list::append(["is"], R1)) :-

nounphr(T, R1).

phr(world(T1, T2), list::append(Start, S1)) :-

nounphr(T1, Start), copula(T2, S1).
end implement english

Horn clauses that yields a phrase according to the grammar of a given

language are called production rules. In the english-class, article/2, noun/2
and nounphr are examples of production rules.
Each English production rule outputs a list of words. Keeping this in
mind, let us take a look at the production rule for phr/2.
phr(world(T1, T2), list::append(Start, S1)) :-
nounphr(T1, Start), copula(T2, S1).
7.3 Why Prolog? 83

The second argument of the head uses the list::append(Start, S1) func-
tion to concatenate a noun phrase with a copula, yielding a phrase.
To implement and test this program, create the tratactus console project.
Insert the classes german, english and semantic into the project tree. Do
not forget to untick the Create Objects box. Add the given code to the
appropriate class files. The testing program is given in figure 7.1. After
compiling it, if you want to test it from inside VIP IDE, use the option
Build/Run in Window. By the way, if you do not want to see a lot of errors,
introduce the semantic class first of all; build the application; then introduce
the german class, and build the application again; the next step is to add the
english class, and rebuild the application. Finally, you can insert the code
for the main class, and build the application for the last time.

7.3 Why Prolog?

One of the goals of a high level computer language is to to increase the
reasoning abilities of programmers, in order to achieve higher productivity,
and artificial intelligence. Although there is a consensus that Prolog meets
these goals better than any other language, there are several controversies
surrounding these terms. Therefore, let us see what we mean by higher
productivity, and artificial intelligence.

Higher productivity. A productive programmer delivers more function-

ality in shorter time. In a software project, there are parts that take most
of the programming effort. For instance, data structures like lists and trees
are needed everywhere, but are difficult to code and maintain in languages
like C, or Java. In Prolog, there isn’t anything easier than dealing with lists,
or trees. Parsing is another difficult subject that one cannot live without;
by the way, parsing is what you have done in order to implement the small
English grammar that you have learned in this chapter. The design of the
application logic is also very unwieldy, but I do not need to elaborate on this
subject, because Prolog has Logic even in its name.

Artificial Intelligence makes your programs highly adaptable to change,

and more user friendly. Prolog, alongside with Lisp and Scheme, is very
popular among people who work with Artificial Intelligence. This means
that you will find Prolog libraries for almost any AI algorithm.
84 Grammars

7.4 Examples
The example of this chapter (tratactus) shows how one could write a program
that translates philosophical German into English.

/***********************************

Project Name: tratactus

UI Strategy: console

************************************/

implement main
open core, console, semantic
class predicates
test:().
clauses
classInfo("main", "tratactus").

test() :- german::phr(T, ["die", "Welt", "ist", "alles"], _X),

write(T), nl, english::phr(T, Translation), !,
write(Translation), nl.
test().

run():- console::init(), test().

end implement main
goal mainExe::run(main::run).

Figure 7.1: Die Welt ist alles, was der Fall ist.

7.5 Notions of Logics: Natural deduction

Both [Gries] and [Dijkstra] propose that one uses Logics to build correct
programs, or to prove that a given program is correct. Many people say
that these authors are dated, because the languages (Pascal-like structured
languages) used in their books are dated. This is a stupid thing to say,
because the methods they propose are based on Predicate Calculus, that is
unlike to become dated; besides this, one can use the Dijkstra/Gries system
to program in a modern language, like Visual Prolog, Clean, or Mercury. In
fact, since [Dijkstra] method is based on Predicate Calculus, one can apply
7.5 Notions of Logics: Natural deduction 85

it more easily to a Prolog dialect than to a Pascal dialect. In fact, [Cohn]

and [Yamaki] have devised a very simple procedure to do it:

• Write the specification of your algorithm in clausal form.

• Transform the specification into a Prolog programming by introducing

control mechanisms, like cuts, and if-then-else structures.

• Prove that the introduction of control mechanisms does not change the
semantic of the algorithm.

To prove that a program is correct, [Gries] proposes that one specifies the
program with Predicate Calculus, and prove its correctness using a method
known as Natural Deduction. In Logics, natural deduction is an approach to
proof theory that tries to provide a more natural model for logical reasoning
than the one proposed by Frege and Russel.
Epistemology says that a judgment is something that one can know. An
object is evident if one in fact knows it. Something is evident if one can
observe it; for instance, global warming is evident because scientists have
records that shows it happening. Many times, evidence is not directly ob-
servable, but rather deduced from basic judgments. The steps of deduction
is what constitutes a proof. There are many kinds of judgement:

Aristotelic judgement: A is true, A is false and A is a proposition.

Temporal logic: A is true at time t

Modal logic: A is necessarily true or A is possibly true

We have seen how to proffer a judgement of the kind A is a proposition: The

grammar rules given on pages 30 and 76 provide inference rules for deciding
whether a string of symbols is a valid formular of Propositional Calculus, or
Predicate Calculus. However, the given grammar for Propositional Calculus
is informal. In a formal treatment,
V grammar rules are called formation rules,
and are represented by F . Then the inference rules to recognize a well
formed formula of the propositional calculus can be written thus:
A prop ^
F (7.1)
¬A prop
A prop B prop ^
F (7.2)
A ∧ B prop , A ∨ B prop , A → B prop , A ≡ B prop
In the Natural Deduction system, one has formation rules for the appropri-
ate calculus, and one rules of insertion, and one rule of elimination for each
86 Grammars

one of the logic connectives present in the calculus one is working with. If
one is working with Predicate Calculus, there are insertion, and elimination
rules for negation (¬P ), conjunction (P ∧ Q), disjunction (P ∨ Q), equiva-
lence (P ≡ Q), implication (P → Q), universal quantification(∀X(P )), and
existential quantification (∃X(P )). In an inference rule, one calls prosthesis
the expressions above the line, and apodosis what comes below the line. The
turnstile symbol ` is not a part of a well-formed formula: but a meta-symbol.
One can see P ` Q as meaning P entails Q, and ` Q as Q is a theorem. Here
are a few inference rules of the Natural Deduction system that one can use
with Predicate Calculus sentences in the clausal form:
E 1 , E2 , . . . E n
and Insertion ∧ /I :
E1 ∧ E2 . . . En
E1 ∧ E2 . . . En
and Elimination ∧ /E :
Ei
Ei
or Insertion ∨ /I :
E1 ∨ E2 . . . En−1 ∨ En
E1 ∨ E2 , E1 → E, E2 → E
or Elimination ∨ /E :
E
E 1 , E2 ` E
implication Insertion → /I :
E1 ∧ E2 → E
E1 → E2 , E1
Modus Ponens → /E :
E2
E ` E1 ∧ ¬E1
not Insertion ¬/I :
¬E
¬E ` E1 ∧ ¬E1
Modus Ponens ¬/E :
E
Since we intend to work with the clausal form, you will not need the rules
corresponding to the quantifiers. Let us play a little with our new toy. Prove
that p ∧ q ` p ∧ (r ∨ q):

p ∧ q ` p ∧ (r ∨ q)
1 p∧q premise 1
2 p ∧/E, 1
3 q ∧/E, 1
4 r∨q ∨/I, 3
5 p ∧ (r ∨ q) ∧/I, 2, 4
In the proof shown above, each step always contains the inference rule,
and the lines on which it was used. For instance, line p ∧ (r ∨ q) because of
the ∧/I inference rule, and of lines 2 and 4.
Chapter 8

Painting

In this chapter, you will learn how to draw pictures using the onPaint event
handler. Start by creating a project with a form, where you will draw things.

• Create a project:

Project Name: painting

Object-oriented GUI (pfc/gui)

• Create paintPack package attached to the root of the Project Tree.

• Put canvas.frm inside paintPack. Build/Build the application.

• Enable the option File/New of the application task menu, and add

clauses
onFileNew(S, _MenuTag) :-
X= canvas::new(S), X:show().

to TaskWindow/Code Expert/Menu/TaskMenu/id_file/id_file_new.
After this step, if you compile and run the program, whenever you
choose the File/New option of the application task menu, the above
predicate will be executed. The command X= canvas::new(S) creates
a new X object of class window. This window will be the child of the
task window S. The command X:show() sends a message to the object
X to show the window.

What will you do next? If I were you, I would draw something on canvas,
to create an interesting background.

87
88 Painting

8.1 onPainting
Whenever a window needs painting, it calls the event handler onPainting.
Therefore, if you add instructions to onPainting, they will be executed.

• Create a class dopaint inside paintPack. Untick Creates Objects.

• Add the code below to dopaint.cl and dopaint.pro.

% File dopaint.cl
class dopaint
open core
predicates
bkg:(windowGDI).
end class dopaint

%File: dopaint.pro
implement dopaint
open core, vpiDomains
clauses
bkg(W) :-
P= [pnt(0,0), pnt(10,80), pnt(150, 100)],
W:drawPolygon(P).
end implement dopaint

• Build/Build the application to add the class dopaint to the project.

• Go to the Properties-dialog of the canvas.frm, press the Events-button,

and add the following code to the PaintResponder:

onPaint(_Source, _Rectangle, GDIObj) :-

dopaint::bkg(GDIObj).

If the canvas.frm is closed, click on its branch in the Project Tree.

As I said before, whenever the window needs painting, it calls onPaint,

that calls dopaint::bkg(GDIObj). The class dopaint has a bkg(GDIObj)
method, that draws things on the window pointed to by GDIObj.
Let us understand what the method bkg(W) does. As you can see, P
holds a list of points. Every point is defined by its coordinates. For instance,
pnt(0, 0) is a point on (0, 0). When bkg(W) calls W:drawPolygon(P), it
sends a message to the object W, asking it to draw a polygon whose vertices
8.1 onPainting 89

are given by the list P. Compile the program, and check whether it will act
exactly as said.
Let us improve method bkg(W). It draws a white triangle on the pane of
canvas. To make the triangle red, one must change the painter’s brush. A
brush is a two argument object.

brush = brush( patStyle PatStyle, color Color).

Colors are represented by integers that specify the amount of red, green,
and blue. As you probably know, you can get a lot of colors by combining
red, green, and blue palettes. Let us represent the integers in hexadecimal
basis. In hexadecimal, numerals have fifteen digits: 0, 1, 2, 3, 4, 5, 6, 7,
8, 9, A, B, C, D, E, F. The red color is given by the integer 0x000000FF,
where 0x shows that we are dealing with the hexadecimal basis. Representing
colors by integers is fine, but you have the option of using identifiers. For
instance, color_Red represents the red color, as you can guess. There are
also identifiers for patterns:

• pat_Solid: Solid brush.

• pat_Horz: Horizontal hatch.

• pat_Vert: Vertical hatch.

• pat_FDiag: 45-degree downward hatch.

Here is a modification of bkg(W) that provides a red triangle:

bkg(W) :- Brush= brush(pat_Solid, color_Red),

P= [pnt(0,0), pnt(10,80), pnt(150, 100)],
W:setBrush(Brush), W:drawPolygon(P).

Finally, here is a version of bkg(W) that draws an ellipse, and clears a green
rectangle inside it.

bkg(W) :- R= rct(40, 60, 150, 200),

W:drawEllipse(R),
R1= rct( 60, 90, 140, 130), W:clear(R1, color_Green).

The last thing that you will learn in this chapter is how to add a bmp picture
to your window. Here is how to modify method bkg(W):

bkg(W) :- P= vpi::pictLoad("frogs.bmp"),
W:pictDraw(P, pnt(10, 10), rop_SrcCopy).
90 Painting

The first step is to load the picture from

a file. This is done by using the predicate
pictLoad/1. The next step is to draw the pic-
ture. In the example, the picture P is put at
the point pnt(10, 10).
One often needs to draw a small picture on
top of another, removing the background of the
small picture. For instance, you may want to
put an eagle among the frogs of the previous
examples.
The first step is to draw a mask of the eagle using
rop_SrcAnd. The mask shows a black shadow of the eagle
on a white back ground. The next step is to draw the eagle
itself using rop_SrcInvert. The eagle must fit perfectly well
onto its mask.

The program that put the eagle among the frogs is given
below, and is implemented in directory paintEagle, of the
examples.
The paintEagle project is exactly alike painting, except for
the definition of bkw(W).
% File dopaint.cl
class dopaint
open core
predicates
bkg:(windowGDI).
end class dopaint

%File: dopaint.pro
implement dopaint
open core, vpiDomains
clauses
bkg(W) :- P= vpi::pictLoad("frogs.bmp"),
W:pictDraw(P, pnt(10, 10), rop_SrcCopy),
Mask= vpi::pictLoad("a3Mask.bmp"),
Eagle= vpi::pictLoad("a3.bmp"),
W:pictDraw(Mask,pnt(50, 50),rop_SrcAnd),
W:pictDraw(Eagle,pnt(50, 50),rop_SrcInvert).
end implement dopaint
8.2 Custom Control 91

8.2 Custom Control

We have been painting directly on the canvas of a form. However, it is a
good idea to create a custom control, and use it for painting.

• Create a new project.

Project Name: customcontrol

Object-oriented GUI (pfc/gui)

• Choose the option File/New in New Package from the IDE task menu.
At the Create Project Item dialog, select the Control tag from the left
hand side pane. The name of the new custom control will be canvas.

Name: canvas
New Package
Parent Directory [ ]

Leave the Parent Directory field blank. Press the Create button.

• A prototype of a canvas control will appear on the screen. You can

close it to avoid cluttering the IDE. Build the application to include
the new control.

• Add a new form to the project tree. Choose File/New in New Package
from the task menu, and select the Form tag from the left hand side
pane of the Create Project Item dialog. Let the new form’s name be
greetings. After pressing the Create-button, you will obtain a proto-
type of the greetings-form.

• Choose the key-symbol that appears on the Controls-dialog. Click on

the pane of the greetings-form. The system will show you a menu
containing a list of custom controls; the canvas-control is the first one.

• Remove the Ok-button from the greetings-form, and replace a hi_ctl-

button for it, as in the figure 8.1.
92 Painting

• Build the application.

• Go to the Properties-dialog of the hi_ctl-button, and add the following

snippet to the ClickResponder:

clauses
onHiClick(_Source) = button::defaultAction :-
W= custom_ctl:getVPIWindow(),
X= convert(integer, math::random(100)),
Y= convert(integer, math::random(100)),
vpi::drawText(W, X , Y, "Hello!").

• Build the application again. Enable the TaskMenu.mnu → File/New

choice. Add

clauses
onFileNew(S, _MenuTag) :-
F= greetings::new(S), F:show().

to TaskWindow.win/Code Expert/Menu/TaskMenu/id_file/id_file_new.
Build and execute the application. Choose the option File/New, and
a new form will appear. Whenever you click on the canvas control of
the new form, it will print warm greetings.

Figure 8.1: A form with a custom control

8.3 Notions of Logics: Resolution principle 93

8.3 Notions of Logics: Resolution principle

The most important inference rule for Prolog programmers is the resolution
principle, proposed in 1963 by [Robinson]:

P ∨ Q, ¬P ∨ R ` Q ∨ R

To prove this principle, we will need two syllogisms. One of them is the first
distributive law, that says

(a ∨ b) ∧ (a ∨ c) ≡ a ∨ (b ∧ c)

There is also a second distributive law, that we are not going to use in the
proof of Robinson’s resolution principle.

(a ∧ b) ∨ (a ∧ c) ≡ a ∧ (b ∨ c)

Let us take a concrete case of these laws. Lysias was an Athenian lawyer
that lived in the time of Socrates. In fact, he was the son of that Kephalos
who entertains Socrates in Plato’s Republic. He was a metic, or resident
alien; as such, he did not have civil rights; nevertheless his family was very
rich, as Plato explains in the Republic. His defences and prosecutions are
very interesting. For instance, during the dictatorship of the thirty tyrants,
the rulers of country decided to take 8 rich metics, and two poor ones, kill
them, and rob them; Lysias and his brother Polemarchus were among those
condemned to death. However, Lysias managed to flee, and returned to his
country to help restore the democracy, and to prosecute one of the tyrants
that killed his brother.
In another occasion, Lysias defended a man who killed his wife’s lover. To
understand his defence, you need to know a few facts about the Greek society.
Man and woman lived in different quarters: Man lived in the androecium, and
women lived in the gynoecium. You probably remember that angiosperms
also have androecia (housing for males) and gynoecia (housing for females);
the reason is that Carl von Linné named these parts of the vegetal physiology
after the housing arrangements of the Ancient Greeks. The other Greek
custom was the dowry. When the woman left her father’s house, she would
take a handsome amount of money to the house of her husband. Therefore,
no man was inclined to divorce his wive, because this would entail returning
the money to her family.
When Eratosthenes discovered that his wife was cheating him, he decided
to kill her lover, for the divorce would mean losing her rich dowry. He feigned
a trip to his farm in the outskirts of the city, but returned on time to kill the
94 Painting

villain on the unfaithful woman’s bed. However, the prosecutor claimed that
the ravenous husband killed the rival on the domestic altar, to where he fled
seeking the protection of Zeus. If Eratosthenes killed the man on the altar,
the jury would condemn him to death; if he killed him on the bed, he would
be acquitted of all charges. Lysias’s defense of his client was based on the
following argument:

He was on the altar or he was on the bed, and he was on the altar,
or he was killed.

The prosecutor accepted the argument. Then Lysias simplified it, using the
first distributive law:

He was on the altar, or he was on the bed and he was killed. Since
you know that he was killed, he was on the bed.

Eratosthenes was acquitted. After this long digression, let us prove the res-
olution principle.
p ∨ q, ¬p ∨ r ` q ∨ r
1 p∨q∨r ∨/I premise 1
2 ¬p ∨ q ∨ r ∨/I premise 2
3 (p ∨ q ∨ r) ∧ (¬p ∨ q ∨ r) ∧/I, 1, 2
4 (q ∨ r) ∨ (p ∧ ¬p) first distributive law, 3
5 q∨r Modus tollendo ponens
Chapter 9

Data types

Most modern languages work with strongly typed data. This

means that the compiler checks whether the data fed to a pred-
icate or a procedure belong to the right type. For instance, the
arithmetic operations (x + y, x × y, a − b, p ÷ q) work on integer,
real or complex numbers. Thus, the compiler makes sure that your program
will feed numbers, and nothing else, to these operations. If there is a logical
mistake in your code, and you try to divide a string by an integer, the com-
piler balks. I am sure that you are saying to yourself that this is the only
reasonable thing to do. I agree. However, not all language designers had the
common sense to hardwire type checking into their compilers. For instance,
Standard Prolog does not check type until the faulty code is executed. The
program fails when it is already in the computer of the final user.

9.1 Primitive data types

Visual Prolog has a lot of primitive data types:

integer : 3, 45, 65, 34, 0x0000FA1B, 845. By the way, 0x0000FA1B is an

hexadecimal integer, i.e., a numeral expressed in basis 16, which has
the following digits: 1, 2, 3, 4, 5, 6, 7, 8, 9, A, B, C, D, E, F.

real : 3.45, 3.1416, 2.18E-18, 1.6E-19, etc.

string : "pencil", "John McKnight", "Cornell University", etc.

symbol : "Na", "Natrium", "K", "Kalium", etc.

You have seen that symbols look a lot like strings: Both are a sequence
of UNICODE characters. However, they are stored differently. Symbols

95
96 Data types

are kept in a look-up table and Prolog uses their addresses in the table to
represent them internally. Therefore, if a symbol occurs many times in your
program, it will use less storage than a string. Below, there is a very simple
example of using real numbers, symbols, and strings. This example is taken
from chemistry.

/********************************
Project Name: primtype1
UI Strategy: console
*********************************/
% File main.pro
implement main
open core
constants
className = "main".
classVersion = "primtype1".
clauses
classInfo(className, classVersion).
class predicates
energy:(real, real, real) procedure (i, i, o).
family:(symbol, string) procedure (i, o).
clauses
energy(N, Z, E) :- E= -2.18E-19*Z*Z/(N*N).

family("Na", "alkaline metal") :- !.

family(_, "unknown").

run():- console::init(),
energy(4, 1, E1),
energy(2, 1, E2),
DeltaE= E1-E2,
stdio::write(DeltaE), stdio::nl,
family("Na", F),
stdio::write("Na", " is an ", F), stdio::nl,
succeed().
end implement main

goal
mainExe::run(main::run).
9.2 Sets 97

9.2 Sets
A set is a collection of things. You certainly know that collections do not
have repeated items. I mean, if a guy or gall has a collection of stickers, s/he
does not want to have two copies of the same sticker in his/her collection. If
s/he has a repeated item, s/he will trade it for another item that s/he lacks
in his/her collection.
It is possible that if your father has a collection of Greek coins, he will
willingly accept another drachma in his set. However, the two drachmas are
not exactly equal; one of them may be older than the other.

9.3 Sets of numbers

Mathematicians collect other things, besides coins, stamps, and slide rules.
They collect numbers, for instance; therefore you are supposed to learn a lot
about sets of numbers.

N is the set of natural integers. Here is how mathematicians write the

elements of N: {0, 1, 2, 3, 4 . . .}.

Z is the set of integers, i.e., Z = {. . . − 3, −2, −1, 0, 1, 2, 3, 4 . . .}.

Why is the set of integers represented by the letter Z? I do not know,

but I can make an educated guess. The set theory was discovered by Georg
Ferdinand Ludwig Philipp Cantor, a Russian whose parents were Danish,
but who wrote his Mengenlehre in German! In this language, integers may
have some strange name like Zahlen.
You may think that set theory is boring; however, many people think that
it is quite interesting. For instance, there is an Argentinean that scholars con-
sider to be the greatest writer that lived after the fall of Greek civilization. In
few words, only the Greeks could put forward a better author. You probably
heard Chileans saying that Argentineans are somewhat conceited. You know
what is the best possible deal? It is to pay a fair price for Argentineans, and
resell them at what they think is their worth. However, notwithstanding the
opinion of the Chileans, Jorge Luiz Borges is the greatest writer who wrote in
a language different from Greek. Do you know what was his favorite subject?
It was the Set Theory, or Der Mengenlehre, as he liked to call it.
Please, my Argentinian friends, no hard feelings. After all, your coun-
try ranks fourth in my list of favorite countries, after Greece, France, and
Paraguay. The high position of Paraguay in my list is solely due to the fact
that José Asunción Flores was Paraguayan; for those who do not know José
98 Data types

Asunción Flores, listen to his song India, and I am sure that you will put
Paraguay in your map. As for Argentina, it is the fatherland of Bernardo
Houssay, Luis Federico Leloir, César Milstein, Adolfo Pérez Esquivel, Carlos
Saavedra Lamas, Jorge Luiz Borges, Alberto Calderon, Alberto Ginastera,
José Cura, Dario Volonté, Verónica Dahl, Carolina Monard, Che Guevara,
Diego Maradona, Juan Manuel Fangio, Osvaldo Golijov, Eva Peron, Hector
Panizza, José Cura etc. Before I forget, José Asunción Flores lived in Buenos
Aires. But let us return to sets.
When a mathematician wants to say that an element is a member of a
set, he writes something like
3∈Z
If he wants to say that something is not an element of a set, for instance, if
he wants to state that −3 is not an element of N, he writes:

−3 6∈ N

Let us summarize the notation that Algebra teachers use, when they explain
set theory to their students.
Double backslash. The weird notation {x2 ||x ∈ N} represents the set of
x2 , such that x is a member of N, or else, {0, 1, 4, 9, 16, 25 . . .}
Guard. If you want to say that x is a member of N on the condition that
x > 10, you can write {x||x ∈ N ∧ x > 10}.
Conjunction. In Mathematics, you can use a symbol ∧ to say and; therefore
x > 2 ∧ x < 5 means that x > 2 and x < 5.
Disjunction. The expression

(x < 2) ∨ (x > 5)

means x < 2 or x > 5.

Using the above notation, you can define the set of rational numbers:
p
Q = { ||p ∈ Z ∧ q ∈ N ∧ q 6= 0}
q
In informal English, this expression means that a rational number is a fraction
p
q
such that p is a member of Z and q is also a member of N, submitted to the
condition that q is not equal to 0.
9.3 Sets of numbers 99

Visual Prolog does not have a notation for sets, but you can use lists to
represent sets. For instance, choose the option Project/New from the task
menu, and fill the Project Settings dialog thus:

General
Project Name: zermelo
UI Strategy: console

Pay attention that we are going to use the console strategy, not GUI. Choose
the option Build/Build from the task menu to put a prototype of class zermelo
into the project tree. Edit zermelo.pro as shown below. Build the project
again, and execute it using Build/Run in Window.

implement main
open core

clauses
classInfo("main", "zermelo").

run():-
console::init(),
Q= [tuple(X, Y) || X= std::fromTo(1, 4),
Y= std::fromTo(1, 5)],
stdio::nl,
stdio::write(Q), stdio::nl, stdio::nl,
foreach tuple(Num, Den)= list::getMember_nd(Q) do
stdio::write(Num, "/", Den,", ")
end foreach,
stdio::nl.
end implement main

goal
mainExe::run(main::run).

You may wonder why I named the above program after the German
mathematician Ernst Friedrich Ferdinand Zermelo. One reason could be
that he was appointed to an honorary chair at Freiburg im Breisgau in 1926,
which he resigned in 1935 because he disapproved of Hitler’s regime. Another
reason was that he invented the notation for set that we have been using.

Q= [tuple(X, Y) || X= std::fromTo(1, 4),

Y= std::fromTo(1, 5)],
100 Data types

In Zermelo’s notation, this expression is written thus:

Q = {(X, Y )||X ∈ [1 . . . 4] ∧ Y ∈ [1 . . . 5]}

In plain English, Q is a list of pairs (X, Y ) such that X belongs to [1, 2, 3, 4],
and Y belongs to [1, 2, 3, 4, 5]. The snippet

foreach tuple(Num, Den)= list::getMember_nd(Q) do

stdio::write(Num, "/", Den,", ")
end foreach

tells the computer to write(Num, "/", Den,", ") for each tuple (N um, Den)
that is member of the list Q.
The set of rational numbers is so named because its elements can be
represented as a fraction like
p
q
where p ∈ N and p ∈ N.
The next section is somewhat hard to digest. Therefore, if you want
to skip it, feel free to do so. In the section about Real Numbers, I will
summarize the important results, so you wont’t miss anything worth the
effort of understanding a whole page of nasty mathematics.

9.4 Irrational Numbers

At Pythagora’s time, Ancient Greeks
claimed that any pair of line segments
is commensurable, i.e., you can always
find a meter, such that the lengths of
any two segments are given by integers.
The following example will help you un-
derstand the Greek theory of commen-
surable lengths at work. Consider the
square of figure 9.1.
If the Greeks were right, I would be
able to find a meter, possibly a very small
one, that produces an integer measure for
the diagonal of the square, and another
integer measure for the side. Suppose
that p is the result of measuring the side
9.4 Irrational Numbers 101

of the square, and q is the result of measuring the diagonal. The Pythagorean
2 2 2
theorem states that AC + CB = AB , i.e.,

p2 + p2 = q 2 ∴ 2p2 = q 2 (9.1)

You can also choose the meter so that p and q have no common factors.
For instance, if both p and q were divisible by 2, you could double the length
of the meter, getting values no longer divisible by 2. E.g. if p = 20 and
q = 18, and you double the length of the meter, you get p = 10, and q = 9.
Thus let us assume that one has chosen a meter so that p and q are not
simultaneously even. But from equation 9.1, one has that q 2 is even. But if
q 2 is even, q is even too. You can check that the square of an odd number
is always an odd number. Since q is even, you can substitute 2 × n for it in
equation 9.1.

2 × p2 = q 2 = (2 × n)2 = 4 × n2 ∴ 2 × p2 = 4 × q 2 ∴ p2 = 2 × n2 (9.2)

Equation 9.1 shows that q is even; equa-

tion 9.2 proves that p is even too. But this
is against our assumption that p and q are
not both even. Therefore, p and q cannot
be integers in equation 9.1, which you can
rewrite as
p √
= 2
q
√
The number 2, that gives the ratio
between the side and the diagonal of any
square, is not an element of Q, or else,
√
2 6∈ Q. It was Hypasus of Metapontum, a
Greek philosopher, who proved this.

Figure 9.1: A thinking Greek

The Greeks were a people of wise men and women. Nevertheless they
had the strange habit of consulting with an illiterate peasant girl at Delphi,
before doing anything useful. Keeping with this tradition, Hyppasus asked
the Delphian priestess— the illiterate girl— what he should do to please
Apollo. She told him to measure the side and the diagonal of the god’s square
altar using the same meter. By proving that the problem was impossible,
Hypasus discovered a type of number that can not be written as a fraction.
This kind of number is called irrational.
102 Data types

9.5 Real Numbers

The set of all numbers, integer, irrational, and rational is called R, or the
set of real numbers. In Visual Prolog, Z is called integer, although the set
integer does not cover all integers, but enough of them to satisfy your needs.
A Visual Prolog real number belongs to the set real, a subset of R.
If x ∈ integer, Visual Prolog programmers say that x has type integer.
They also say that r has type real if r ∈ real. There are other types besides
integer, and real. Here is a list of primitive types.

integer — Integer numbers between −2147483648 and 2147483647.

real — Reals must be written with a decimal point: 3.4, 3.1416, etc.

string — A quoted string of characters: "3.12", "Hippasus", "pen", etc.

char — Characters between single quotes: ’A’, ’b’, ’3’, ’ ’, ’.’, etc.

9.6 Mathematics
I always thought it very hard to satisfy mathematicians, since they are so
picky, and I have an appalling dificulty in keeping even a pretense of being
formal. In particular, my definitions of the different numerical sets are not
rigorous as most mathematicians would require. However, I have no intention
of writing for an AMS journal, or making major contributions to number
theory; I will be content if I am able to give non-mathematicians a working
knowledge of the concepts involved. By working knowledge, I mean that I
hope that my readers will be able to understand what a type is, and how
type are related to Set Theory. As a bonus, I hope that students of literature
will find Borges’ books easier to digest, after reading my text; hence my
Argentinian friends may forgive me for telling jokes about them.

9.7 Format
The method format creates nice string representations of primitive data
types for output. For instance, it is nice to represent colors using hexadecimal
integers. In this representation, the fundamental colors become 0x00FF0000
(red), 0x0000FF00 (green), and 0x000000FF (blue). However, if you try to
print these numerals using stdio::write/n, you will get them in the decimal
basis. Using string::format/n, as shown below, you will get a printout of
the primitive colors in hexadecimal.
9.7 Format 103

/*******************************
Project Name: primtype2
UI Strategy: console
********************************/
% File main.pro
implement main
open core
constants
className = "main".
classVersion = "primtype2".
clauses
classInfo(className, classVersion).
class predicates
color:(symbol, integer) procedure (i, o).
clauses

color("red", 0x00FF0000) :- !.
color("blue", 0x000000FF) :- !.
color("green", 0x0000FF00) :- !.
color(_, 0x0).

run():- console::init(),
color("red", C),
S= string::format("%x\n", C),
stdio::write(S), stdio::nl,
succeed().
end implement main

goal
mainExe::run(main::run).

The first argument of function format specifies how you want the printout.
In the example, the argument is "%x\n": You want a hexadecimal repre-
sentation of a number, followed by a carriage return (\n). The arguments
that follow the first show the data that you want to print. Here are a few
examples of formatting:

S= string::format("Pi=%4.3f\n", 3.14159) S= 3.142

S= string::format("%4.3e\n", 33578.3) 3.358e+004
S= string::format("Fac(%d)=%d", 5, 120) S= "Fac(5)=120"
S= string::format("%s(%d)=%d", "f" 5, 120) S= "f(5)=120"
104 Data types

The format fields specification is %[-][0][width][.precision][type],

where the hyphen (-) indicates that the field is to be left justified, right jus-
tification being the default; the zero before width pads the formatted string
with zeros until the minimum width is reached; width specifies a minimum
field size; precision specify the precision of a floating-point numeral; finally
type can be f (fixed point numeral), e (scientific notation for reals), d (dec-
imal integer), x (hexadecimal integer), o (octal). Strings are represented as
"%s". You can add any message to the format specification; you can also
add carriage returns ("\n").

9.8 domains
You may need other data types besides the primitives provided by Visual
Prolog. In this case, you need to specify the new data types that you will
insert in your code. In previous chapters you have already learned how to
declare domains; you also learned how to create complex data structures, like
lists. Therefore, all we have to do in this section is to recall concepts already
acquired, and to make a synthesis of what we have learned.

9.8.1 Lists
You have learned that a list is an ordered sequence of elements. You also
know how to create list domains, and declare predicates on lists. For instance,
here are a few list domains:
domains
reals= real*. % a list of reals
integers= integer*. % a list of integers
strings= string*. % a list of strings
rr= reals*. % a list of lists of reals

Once you have a domain declaration, you can use it like any primitive type.
In the case of list, domain declaration is not necessary, since one can use list
types like real*, string* or integer* directly in the predicate declaration,
as in the program of figure 9.2.

9.8.2 Functors
Predicate calculus, a branch of Logics, has a data type called functional
symbol. A functional symbol has a name, or functor, followed by arguments.
9.8 domains 105

/*******************************
Project Name: newdomains
UI Strategy: console
********************************/
% File main.pro
implement main
open core

class predicates
sum:(real*, real) procedure (i, o).

clauses
classInfo("main", "newdomains").

sum([], 0) :- !.
sum([X|Xs], X+S) :- sum(Xs, S).

run():- console::init(),
sum([1, 2, 3, 4, 5, 6], A),
stdio::writef("The sum is %-4.0f", A).
end implement main

goal
mainExe::run(main::run).

Figure 9.2: List declaration

The arguments are written between parentheses, and separated from each
other by commas. To make a long story short, they have exactly the same
notation as functions. Here are a few example of functional symbols:
• author("Wellesley", "Barros", 1970)

• book( author("Peter", "Novig", 1960),

"Artificial Intelligence")

• date("July", 15, 1875)

• entry("Prolog", "A functional language")

• index("Functors", 165)

• sunday
106 Data types

The last example shows that a functional symbol can lack arguments. This
said, let us see how to declare data types for a few functors. The example
calculates the day of the week for any date between the years 2000 and 2099.
It has the following domain declarations:

DOMAINS Comments
year= integer. year is synonym of integer. The use of
synonyms makes things clearer.
day= sun; mon;tue; wed; This domain has many functional sym-
thu; fri; sat; err. bols without arguments. Each func-
tional symbol is separated from the
next by semicolon, and represents a day
of the week.
month= jan;feb;mar;apr; Another domain (type) with more than
may;jun;jul;aug;sep; one functional symbol.
oct;nov;dec.
month code= Here, one has a functional symbol with
m(month, integer). two arguments.
numberOfDays= Another type with two functional sym-
nd(integer); bols; the first functional symbol has one
february. argument; the second one has no argu-
ment at all.

/*******************************
Project Name: dayOfWeek
UI Strategy: console
********************************/
% File: main.pro
% This program calculates the day of the week
% for the years between 2000 and 2099.
implement main
open core
constants
className = "main".
classVersion = "dayOfWeek".
domains
year= integer.
day= sun; mon;tue; wed; thu; fri; sat; err.
month= jan;feb;mar;apr;may;jun;jul;aug;sep;oct;nov;dec.
month_code= m(month, integer).
numberOfDays= nd(integer); february.
9.8 domains 107

class predicates

dayOfWeek:(integer, day) procedure (i, o).

calculateDay:( string Month,

integer Day_of_the_month,
year Y,
day D) procedure (i, i, i, o).

class facts
monthCode:(string, month_code, numberOfDays).

clauses
classInfo(className, classVersion).

monthCode("January", m(jan, 6), nd(31)).

monthCode("February", m(feb, 2), february ).
monthCode("March", m(mar, 2), nd(31)).
monthCode("April", m(apr, 5), nd(30)).
monthCode("May", m(may, 0), nd(31)).
monthCode("June", m(jun, 3), nd(30)).
monthCode("July", m(jul, 5), nd(31)).
monthCode("August", m(aug, 1), nd(31)).
monthCode("September", m(sep, 4), nd(30)).
monthCode("October", m(oct, 6), nd(31)).
monthCode("November", m(nov, 2), nd(30)).
monthCode("December", m(dec, 4), nd(31)).

dayOfWeek(0, sun) :- !.
dayOfWeek(1, mon) :- !.
dayOfWeek(2, tue) :- !.
dayOfWeek(3, wed) :- !.
dayOfWeek(4, thu) :- !.
dayOfWeek(5, fri) :- !.
dayOfWeek(6, sat) :- !.
dayOfWeek(_, err).
108 Data types

calculateDay(Month, MD, Year, D) :-

Y= Year-2000, % Last digits of Year.
monthCode(Month, m(_M, C), _), % Get month code.
!,
S= Y+Y div 4+C+MD,
R= S mod 7, /* Day of the week as an integer
between 0 (sunday) and 6 (saturday) */
dayOfWeek(R, D). % Get functional symbol of the day
calculateDay(_, _, _, err).

run():-
console::init(),
calculateDay("May", 3, 2005, R),
stdio::write("Day of the week: ", R), stdio::nl,
succeed(). % place your own code here
end implement main
goal
mainExe::run(main::run).

9.9 Notions of Logics: Horn clauses

A Horn clause has at most one positive literal. For mathematicians, Horn
clauses have interesting properties. However, their appeal is even greater for
computer scientists, because they can be put in the following form:
H1 ← +L11 ∧ +L12 ∧ +L13 . . . ,
H2 ← +L21 ∧ +L12 ∧ +L13 . . . ,
H3 ← +L31 ∧ +L12 ∧ +L13 . . . , . . .
where ← is the implication symbol, and +Lij is the negation of the negative
literals. The proof that a clausal clause(page 77) H ∨ ¬T can indeed be put
in the form H ← T was given on page 30. The form H ← T1 , T2 , T3 . . .
is appealing because it resembles informal rules that people use to express
knowledge:
The rain is certain if swallows fly close to the ground.
From what you have seen, it is no wonder that the first trial to use Logic as
programming language was focused on Horn clauses. However, the design-
ers of Prolog faced the fact that Horn clauses are not complete enough for
most applications. This difficulty was solved with the introduction of control
mechanisms, like the cut, into the Horn clauses.
Chapter 10

How to solve it in Prolog

The title of this chapter is a homage to Helder Coelho, the first author
of the best book on Prolog ever written: [HowToSolveItWithProlog]. This
book was printed by the Laboratorio Nacional de Engenharia Civil, where
Helder Coelho works in Portugal. Later, Coelho published Prolog by Example
(see [Coelho/Cotta]), that is good too, but not as good as his first book.
Coelho’s book is a collection of short problems, proposed and solved by
great programmers and computer scientists. The whole thing is organized as
a kind of FAQ. The problems are interesting, and the solutions are illumi-
nating. More important, the book is also very amusing. I hope that it will
see a new edition soon, since it shows the history of the creators of Logic
Programming, their discoveries, their explorations.
All programs in this chapter will be console applications. Besides this,
I will only provide the listing of the main class implementation. It is up to
you to complete the red tape. For instance, if I give you the listing

implement main
open core
clauses
classInfo("main", "hello").
clauses
run():- console::init(), stdio::write("Hello, world!\n").
end implement main
goal
mainExe::run(main::run).

you must create a console project, whose name is hello.

109
110 How to solve it in Prolog

10.1 Utilities
Verbal statement: Search all members of a list L. Logic program:

implement main
open core
class predicates
member:(integer, integer*) nondeterm anyflow.
member:(string, string*) nondeterm anyflow.
test:(string*) procedure (i).
test:(integer*) procedure (i).

clauses
classInfo("main", "utilities").

member(H, [H|_]).
member(H, [_|T]) :- member(H, T).

test(L) :- member(H, L),

stdio::write(H), stdio::nl, fail or succeed().

run():- console::init(), L= [2,3,4,5], test(L),

S= ["a", "b", "c"], test(S).
end implement main

goal
mainExe::run(main::run).

The example shows that you can define a predicate with different domains.
For instance, there is a definition of member/2 for string*, and another for
integer*. The same thing happens to test/1.
The definition of test/1 uses the predicate fail to backtrack, and print all
solutions of the nondeterm predicate member. As a result, the program will
print all elements of the list.
10.1 Utilities 111

Verbal statement: Check whether U is the intersection of lists L1 and L2;

search the intersection of lists L1 and L2.

implement main
open core, stdio
class predicates
isec:(integer*, integer*, integer*) nondeterm anyFlow.
memb:(integer, integer*) nondeterm anyFlow.
tst:(integer*, integer*, integer*, string) procedure(i, i, i, o).
findsec:(integer*, integer*, integer*) procedure (i, i, o).
length:(integer*, integer) procedure (i, o).
clauses
classInfo("main", "intersection").

memb(H, [H1|T]) :- H=H1; memb(H, T).

isec(L1, L2, [H|U]) :- memb(H, L1), memb(H, L2), !,
isec(L1, L2, U).
isec(_, _, []).
length([], 0) :- !.
length([_|R], L+1) :- length(R, L).

tst(L1, L2, U, R) :-findsec(L1, L2, S),

length(U, LU), length(S, LS), LU= LS,
isec(L1, L2, U), !, R= "yes"; R= "no".

findsec([H|T], L, [H|U]) :- memb(H, L), !, findsec(T, L, U).

findsec([_|T], L, U) :- findsec(T, L, U), !.
findsec(_L1, _L2, []).

run():- console::init(), L1= [3, 6, 4, 5],

L2= [4, 5, 6], U1= [4, 6, 5], U2= [4, 3],
tst(L1, L2, U2, Resp2), tst(L1, L2, U1, Resp1),
write(Resp1, ", ", Resp2), nl,
findsec(L1, L2, I), write(I), nl.
end implement main

goal
mainExe::run(main::run).
112 How to solve it in Prolog

Verbal Statement: Specify the relationship append/3 between three lists

which holds if the last one is the result of appending the first two.

implement main
open core
class predicates
app:(Elem* L1, Elem* L2, Elem* L3)
nondeterm anyFlow.
test1:(string*) procedure (i).
test2:(integer*, integer*) procedure (i, i).
clauses
classInfo("main", "append").

app([], L, L).
app([H|T], L, [H|U]) :- app(T, L, U).

test1(U) :- app(L1, L2, U),

stdio::write(L1, " ++ ", L2, "= ", U), stdio::nl, fail.
test1(_).

test2(L1, L2) :- app(L1, L2, U),

stdio::write(L1, " ++ ", L2, ": ", U), stdio::nl, fail.
test2(_, _).

clauses
run():- console::init(),
test1(["a", "b", "c", "d"]), stdio::nl,
test2([1, 2, 3], [4, 5, 6]).
end implement main

goal
mainExe::run(main::run).

This problem is an all time favorite. If you run the program, test1/1 will
find and print all ways of cutting the list ["a", "b", "c", "d"]. On the
other hand, test2/2 will concatenate two lists. Both predicates use app/3,
which perform two operations, concatenation, and cutting down lists.
10.1 Utilities 113

Verbal Statement: Write a program that reverses a list.

/*******************************
Project Name: reverse
UI Strategy: console
********************************/

implement main
open core

class predicates
rev:(integer*, integer*) procedure (i, o).
rev1:(integer*, integer*, integer*) procedure (i, i, o).

clauses
classInfo("main", "reverse").

rev1([], L, L) :- !.
rev1([H|T], L1, L) :- rev1(T, [H|L1], L).

rev(L, R) :- rev1(L, [], R).

run():- console::init(), L= [1, 2, 3, 4, 5],

rev(L, R), stdio::write("Reverse of", L, "= ", R),
stdio::nl.

end implement main

goal
mainExe::run(main::run).

This program is useful and illuminating. It is illuminating because it shows

how to use accumulating parameters in a very efficient way. Assume that
you feed the goal

rev([1, 2, 3], R)

to Prolog. Then the inference engine will try to make rev([1, 2, 3], R)
equal to the head of one of the Horn clauses that are in the program. In the
present case, it will succeed if it makes L=[1,2,3] in the clause

rev(L, R) :- rev1(L, [], R)

114 How to solve it in Prolog

But if L=[1,2,3], this clause becomes

rev([1,2,3], R) :- rev1([1,2,3], [], R)
This process of making a goal equal to the head of a clause is called unifica-
tion, and one says that rev([1,2,3], R) unifies with
rev(L, R) :- rev1(L, [], R), for L=[1,2,3]
producing rev1(L, [], R), that is equal to rev1([1,2,3], [], R), since
L=[1,2,3]. The product of unification is the tail of a Horn clause, because
you must prove the tail, in order to prove the head. For example, consider
the following Horn clause in English notation:
X is grandfather of Y if X is father of Z and Z is father of Y.
If you want to prove that X is grandfather of Y, you must prove that Y is
father of Z, and Z is father of Y. In fewer words, one must prove the tail, in
order to prove the head of a Horn clause. All this long discussion may be
summarized thus
• rev([1,2,3], R) unifies with
rev(L, R) :- rev1(L, [], R), for L=[1,2,3]

yielding rev1([1,2,3], [], R).

• rev1([1,2,3], [ ], R) unifies with

rev1([H|T],L1,L) :- rev1(T,[H|L1],L)/ H=1, L1=[], T=[2,3]

producing rev1([2, 3], [1], L).

• rev1([2, 3], [1], L) unifies with

rev1([H|T],L1,L) :- rev1(T,[H|L1],L)/ H=2, L1=[1], T=[3]

yielding rev1([3], [2,1], L).

• rev1([3], [2,1], L) unifies with

rev1([H|T],L1,L) :- rev1(T,[H|L1],L)/ H=3, L1=[2,1], T=[]

yielding rev1([], [3,2,1], L).

• rev1([], [3,2,1], L) unifies with rev1([], L, L) :- ! for L=[3,2,1],

producing the result, that is, L=[3,2,1].
10.1 Utilities 115

Laplace used to say that Newton was not only the greatest scientist of all
times and lands, but also a very lucky man, for he was born before Laplace.
Of course, Le Marquis would discover the Celestial Mechanics before Newton,
if he had the chance of doing so. I suppose that you would write the quick
sort algorithm before Van Emden, if he was not lucky enough to put his hands
on a Prolog compiler before you. The quick sort algorithm was invented by
Tony Hoare, that wanted a fast method for sorting lists and vectors. This
was a time when B+Trees did not exist, and the only way to perform a fast
search in a list of customers was to have it sorted.
Let us see how Van Emden explained his algorithm to his friend Coelho.
First of all, he implemented a split/4 algorithm, that splits a list L in two
sublists: S(mall) and B(ig). The S sublist contains all of L that are smaller of
equal to P(ivot). The B sublist contains the elements that are larger than P.
Here is how Van Emden implemented the split/4:

implement main
open core
clauses
classInfo("main", "quicksort").

class predicates
split:(E, E*, E*, E*) procedure (i, i, o, o).
clauses

split(P, [H|T], S, [H|B]) :- H>P, !, split(P, T, S, B).

split(P, [H|T], [H|S], B) :- !, split(P, T, S, B).
split(_, [], [], []).

run():- console::init(),
split(5, [3, 7, 4, 5, 8, 2], Small, Big),
stdio::write("Small=", Small, ", Big= ", Big), stdio::nl,
split("c", ["b", "a", "d", "c", "f"], S1, B1),
stdio::write("Small=", S1, ", Big= ", B1), stdio::nl.

end implement main

goal
mainExe::run(main::run).

If you run this program, the computer will print

Small=[3,4,5,2], Big=[7,8]
116 How to solve it in Prolog

which is what it should do. Now, let us understand the idea behind the quick-
sort algorithm. Let us assume that you have the list [5, 3, 7, 4, 5, 8, 2],
and you want to sort it.

• Take the first element of L= [5, 3, 7, 4, 5, 8, 2] as pivot, and

split it in Small=[3,4,5,2] and Big=[7,8].

• Sort Small using the same algorithm; you will get

Sorted_Small= [2, 3, 4, 5]

• Sort Big and get Sorted_Big= [7,8].

• Append Sorted_Small to [Pivot|Sorted_Big]. You will get the

sorted list: [2,3,4,5,5,7,8].

This algorithm needs a predicate that appends one list to another. You
can use app/3, that you learned on page 112. However, that algorithm is
nondeterm, which we do not need, since we do not need the functionality of
cutting a list in two; all we need is a vanilla concatenation. Therefore, you
can insert a cut after the first clause of app/3.

app([], L, L) :- !.
app([H|T], L, [H|U]) :- app(T, L, U).

Declare app as a procedure with two input arguments, and one output.

app:( integer_list, integer_list,

integer_list) procedure (i, i, o).

To test the algorithm, we are going to use a list of integers. Of course, it will
be more useful with a lists of strings.
10.1 Utilities 117

implement main
open core
clauses
classInfo("main", "quicksort").
class predicates
split:(E, E*, E*, E*) procedure (i, i, o, o).

app:(E*, E*, E*) procedure (i, i, o).

sort:(E*, E*) procedure (i, o).

clauses
split(P, [H|T], S, [H|B]) :- H>P, !, split(P, T, S, B).
split(P, [H|T], [H|S], B) :- !, split(P, T, S, B).
split(_, [], [], []).

app([], L, L) :- !.
app([H|T], L, [H|U]) :- app(T, L, U).

sort([P|L], S) :- !, split(P, L, Small, Big),

sort(Small, QSmall), sort(Big, QBig),
app(QSmall, [P|QBig], S).
sort([], []).

run():- console::init(), L= [5, 3, 7, 4, 5, 8, 2],

sort(L, S),
stdio::write("L=", L , ", S= ", S), stdio::nl,
Strs= ["b", "a", "c"], sort(Strs, Qs),
stdio::write("Qs= ", Qs), stdio::nl.
end implement main

goal
mainExe::run(main::run).
118 How to solve it in Prolog

Prolog was invented by Alain Colmerauer, a French linguist. His student

Philippe Roussel implemented an interpreter. It seems that Roussel’s wife
gave the name Prolog to the computer language. Warren was the man who
implemented the first Prolog compiler. The problem below was proposed by
this very same Warren.

Verbal Statement: The number of days in a year is 366 each four years;
otherwise it is 365. How many days have years 1975, 1976, and 2000?

implement main
open core
class predicates
no_of_days_in:(integer, integer) procedure (i, o).
member:(integer, integer_list) nondeterm (o, i).
test:(integer_list) procedure (i).
clauses
classInfo("numdays", "1.0").

member(H, [H|_T]).
member(H, [_|T]) :- member(H, T).

no_of_days_in(Y, 366) :- 0= Y mod 400, !.

no_of_days_in(Y, 366) :- 0= Y mod 4,
not(0= Y mod 100), !.
no_of_days_in(_Y, 365).

test(L) :- member(X, L),

no_of_days_in(X, N),
stdio::write("Year ", X, " has ", N, " days."),
stdio::nl, fail.
test(_L).

run():- console::init(),
test([1975, 1976, 2000]).
end implement main

goal
mainExe::run(main::run).
10.1 Utilities 119

Luis Moniz Pereira is a Portuguese who works with Artificial Intelligence.

He was one of the first persons to use Prolog for Artificial Intelligence. He is
co-author of the book How to Solve it in Prolog.

Verbal statement: Write a program for coloring any planar map with at
most four colors, such that no two adjacent regions have the same color.

The problem is a classical generate and test program. In test(L), the pred-
icate calls

generateColor(A), generateColor(B),
generateColor(C), generateColor(D),
generateColor(E), generateColor(F),

generate colors for the six regions of the map. Then, test(L) builds a the
map as a list of neighboring countries:

L= [ nb(A, B), nb(A, C), nb(A, E), nb(A, F),

nb(B, C), nb(B, D), nb(B, E), nb(B, F),
nb(C, D), nb(C, F), nb(C, F)],

Finally, the predicate aMap(L) checks whether L is an admissible map. It

will be admissible if there is no two neighboring countries painted with the
same color. If predicate aMap(L) fails, the proposed colors are not correct.
Therefore, the program backtracks to generateColor(X) in order to get a
new set of colors.

The four color problem is interesting for two reasons.

1. The generate and test scheme was one of the first Artificial Intelligence
techniques.

2. There is a very famous theorem, that states:

Every map can be colored with four colors so that adjacent

countries are assigned different colors.

This theorem was proved in 1976 by Kenneth Appel and Wolfgang

Haken, two mathematicians at the University of Illinois.
120 How to solve it in Prolog

implement main
open core
domains
colors= blue;yellow;red;green.
neighbors= nb(colors, colors).
map= neighbors*.
class predicates
aMap:(map) nondeterm anyFlow.
test:(map) procedure anyFlow.
generateColor:(colors) multi (o).
clauses
classInfo("main", "fourcolors").

generateColor(R) :- R= blue; R= yellow;

R= green; R= red.

aMap([]).
aMap([X|Xs]) :- X= nb(C1, C2), not(C1 = C2),
aMap(Xs).

test(L) :- generateColor(A), generateColor(B),

generateColor(C), generateColor(D),
generateColor(E), generateColor(F),
L= [ nb(A, B), nb(A, C), nb(A, E), nb(A, F),
nb(B, C), nb(B, D), nb(B, E), nb(B, F),
nb(C, D), nb(C, F), nb(C, F)],
aMap(L), !; L= [].

run():- console::init(), test(L),

stdio::write(L), stdio::nl.
end implement main

goal
mainExe::run(main::run).
10.1 Utilities 121

Verbal statement: Write a program for the game of Nim. A position of

the game of Nim can be visualized as a set of heaps of matches; we will
represent each heap as a list of integers.

[1,1,1,1,1]

Therefore, the set of heaps will be a list of lists; e.g.

[ [1,1,1,1,1],
[1, 1],
[1,1,1,1]]

Two players, you and the computer, alternate making moves. As soon as a
player is unable to make a move, the game is over, and that player has lost.
A move consists of taking at least one match from exactly one heap. In the
example, if you take three matches from the third heap, you make a valid
move, and the board becomes

[ [1,1,1,1,1],
[1, 1],
[1]]

where you have removed three matches from the third heap. To implement
this project, we will use a programming technique called incremental de-
velopment of systems. First, you will implement and test a program that
appends two lists. Since you are going to use this program both to split and
to concatenate lists, you need to test both possibilities. In fact, if you run
the first program (next page), you will get

[[1],[1],[1],[1],[1],[1],[1],[1]]
[][[1],[1],[1,1]]
[[1]][[1],[1,1]]
[[1],[1]][[1,1]]
[[1],[1],[1,1]][]

The first line is the result of concatenating two sets of heaps; the lines that
follow shows the many possibilities of splitting a list of heaps. Then the first
program seems to be working properly.
122 How to solve it in Prolog

implement main
open core

domains
ms= integer*.
hs= ms*.

class predicates
append:(E*, E*, E*) nondeterm anyFlow.

clauses
classInfo("main", "nimgame").

append([], Y, Y).
append([U|X], Y, [U|Z]) :- append(X, Y, Z).

clauses
run():-
console::init(),
append([[1],[1],[1],[1],[1]], [[1],[1],[1]], L), !,
stdio::write(L), stdio::nl; succeed().
end implement main

goal
mainExe::run(main::run).

The next step is to write a predicate that takes at least one match from
a heap; of course, it can take more than one match. After removing one or
more matches from the heap, takesome/3 inserts the modified heap into a
set of heaps. If you test the program, it will print the following result:

[[1,1,1,1],[1,1]]
[[1,1,1],[1,1]]
[[1,1],[1,1]]
[[1],[1,1]]
[[1,1]]

NB: the first clause of takesome/3 makes sure that the predicate will not
insert an empty heap into the set of heaps.
10.1 Utilities 123

implement main
open core
domains
ms= integer*.
hs= ms*.
class predicates
append:(E*, E*, E*) nondeterm anyFlow.
test:() procedure.
takesome:(ms, hs, hs) nondeterm anyFlow.
clauses
classInfo("main", "nimgame").

append([], Y, Y).
append([U|X], Y, [U|Z]) :- append(X, Y, Z).

takesome([_X], V, V).
takesome([_X, X1|Y], V, [[X1|Y]|V]).
takesome([_X|T], V, Y) :- takesome(T, V, Y).

test() :- L= [1, 1, 1, 1, 1],

V= [[1,1]],
takesome(L, V, Resp),
stdio::write(Resp), stdio::nl,
fail; succeed().

clauses
run():- console::init(),
test().
end implement main

goal
mainExe::run(main::run).
Now, you are ready to generate all possible moves from a given position. If
you run the test program, it will produce the following screen:
Possible moves from [[1,1,1],[1,1]]:
[[1,1],[1,1]]
[[1],[1,1]]
[[1,1]]
[[1,1,1],[1]]
[[1,1,1]]
124 How to solve it in Prolog

implement main
open core
domains
ms= integer*.
hs= ms*.
class predicates
append:(E*, E*, E*) nondeterm anyFlow.
takesome:(ms, hs, hs) nondeterm anyFlow.
move:(hs, hs) nondeterm anyFlow.
clauses
classInfo("main", "nimgame").

append([], Y, Y).
append([U|X], Y, [U|Z]) :- append(X, Y, Z).

takesome([_X], V, V).
takesome([_X, X1|Y], V, [[X1|Y]|V]).
takesome([_X|T], V, Y) :- takesome(T, V, Y).

move(X, Y) :- append(U, [X1|V], X),

takesome(X1, V, R), append(U, R, Y).

run():- console::init(), L= [[1, 1, 1], [1, 1]],

stdio::write("Possible moves from ", L, ": "),
stdio::nl, move(L, Resp),
stdio::write(Resp), stdio::nl, fail; succeed().
end implement main

goal mainExe::run(main::run).
10.1 Utilities 125

implement main
open core
domains
ms= integer*.
hs= ms*.
class predicates
append:(E*, E*, E*) nondeterm anyFlow.
takesome:(ms, hs, hs) nondeterm anyFlow.
move:(hs, hs) nondeterm anyFlow.
winningMove:(hs, hs) nondeterm (i, o).
clauses
classInfo("main", "nimgame").

append([], Y, Y).
append([U|X], Y, [U|Z]) :- append(X, Y, Z).

takesome([_X], V, V).
takesome([_X, X1|Y], V, [[X1|Y]|V]).
takesome([_X|T], V, Y) :- takesome(T, V, Y).

move(X, Y) :- append(U, [X1|V], X),

takesome(X1, V, R), append(U, R, Y).

winningMove(X, Y) :- move(X, Y), not(winningMove(Y, _)).

run():- console::init(), L= [[1, 1, 1], [1, 1]],

winningMove(L, S),
stdio::write("Winning move: ", S),
stdio::nl, fail; succeed().
end implement main

goal mainExe::run(main::run).

The heart of this program is the predicate

winningMove(X, Y) :- move(X, Y), not(winningMove(Y, _)).

If there is a winning strategy, this amazing one line predicate will find it!
You will find below an example of a game against the computer. As you
can see, if there is a chance of winning, the machine will take it.
126 How to solve it in Prolog

[[1,1,1],[1,1]]
Your move: [[1], [1,1]]
My move: [[1],[1]]
Your move: [[1]]
My move: []
I win

To implement the final program, you will create a class for the game itself.
Let the name of this class be nim. Then, the main program is given below.

% File: main.pro
implement main
open core, stdio

clauses
classInfo("main", "nimgame-1.0").

run():- console::init(), L= [[1, 1, 1], [1, 1]],

write(L), nl, nim::game(L), ! ; succeed().
end implement main

goal mainExe::run(main::run).

The class interface exports the method game/1, and the domains that
you need to describe a game position. It is defined below.

% File: nim.cl
class nim
open core
domains
ms= integer*.
hs= ms*.
predicates
classInfo : core::classInfo.
game:(hs) determ (i).
end class nim
10.1 Utilities 127

% File:nim.pro
implement nim
open core, stdio

class predicates
append:(hs, hs, hs) nondeterm anyFlow.
takesome:(ms, hs, hs) nondeterm anyFlow.
move:(hs, hs) nondeterm anyFlow.
winningMove:(hs, hs) nondeterm (i, o).
iwin:(hs) determ (i).
youwin:(hs, hs) determ (i, o).

clauses
classInfo("nim", "1.0").

append([], Y, Y).
append([U|X], Y, [U|Z]) :- append(X, Y, Z).

takesome([_X], V, V).
takesome([_X, X1|Y], V, [[X1|Y]|V]).
takesome([_X|T], V, Y) :- takesome(T, V, Y).

move(X, Y) :- append(U, [X1|V], X),

takesome(X1, V, R), append(U, R, Y).

winningMove(X, Y) :- move(X, Y), not(winningMove(Y, _)).

iwin([]) :- !, write("I win"), nl.

youwin(L, L2) :- winningMove(L, L2), !.
youwin(_L, _L2) :- write("You win"), nl, fail.

game(L1) :- write("Your move: "),

L= console::read(), move(L1, LTest),
L= LTest, youwin(L, L2), !,
write("My move: "), write(L2), nl,
not(iwin(L2)), game(L2).
end implement nim
128 How to solve it in Prolog

The strategy used in this very simple program is called minmax algorithm.
The algorithm can be applied to any game of two contenders, where both
have perfect knowledge of the game state. Examples of such games: Chess,
Go, Checkers, Tic-Tac-Toe, and Nim. It can also be used to control robots;
in fact, there is a kind of filter based on the minmax algorithm that is similar
to Kalman’s. In the minmax strategy, the states of the game are organized in

Figure 10.1: Minmax tree

a tree, that is called search tree. When it is the computer’s turn, it generates
states of minimum value for the opponent. On the other hand, it tries to
follow tree branches that leads to a winning position. In a simple game like
Nim, the computer is able to find a path to sure victory. In more complex
games, like chess, this is not always possible; one must interrupt the search
before being sure of winning the game. In this case, game strategists design
functions that assign values to a position even when it is not a sure win, or
defeat.
Chapter 11

Facts

A fact is a Horn clause without a tail. Facts can be asserted, modified, or

retracted dynamically, at runtime. An example will help you to understand
why facts are necessary in Prolog. Let us assume that you want to create
a small database of publications. When you have a new book, you want to
add it to the database.

addItem(journal("AI in focus", "MIT Press")),

The predicate addItem/1 can be defined thus:

class predicates
addItem:(volume).
clauses
addItem(V) :-
num := num+1,
assert(item(V, num)).

The execution of addItem(journal("AI", "AldinePress")) adds the clause

item(journal("AI", "AldinePress")).

to the database, and increments the variable num. The declaration

class facts - bib

num:integer := 0.
item:(volume, integer) nondeterm.

creates the fact database bib, with a variable num and a predicate item/2;
the predicate is nondeterm, and the variable is single.

129
130 Facts

The domain volume, that will provide typing to the records of this small
database, is defined as

domains
name= string.
author= n1(name); n2(name, name); etal(name).
publisher= string.
title= string.
volume= journal(title, publisher); book(title, author).

The program below shows how to define facts, and how to save the fact
database to a file.

% File main.pro
implement main
open core
domains
name= string.
author= n1(name); n2(name, name); etal(name).
publisher= string.
title= string.
volume= journal(title, publisher); book(title, author).

class facts - bib

num:integer := 0.
item:(volume, integer) nondeterm.

class predicates
addItem:(volume).
prtDataBase:().

clauses
classInfo("main", "facttest").

addItem(V) :- num := num+1,

assert(item(V, num)).

prtDataBase() :- item(V, I),

stdio::write(I, "=", V), stdio::nl,
fail.
prtDataBase().
 131

clauses
run():-
console::init(),
addItem(journal("AI in focus", "MIT Press")),
addItem(book( "Databases in Prolog",
n1("Wellesley Barros"))),
file::save("bibliography.fac", bib),
prtDataBase().
end implement main

goal
mainExe::run(main::run).
After creating the database and saving it to a file, you can use it in another
program. To test this possibility, create a console project factread, and add
the following program to it:
% File: main.pro
implement main
open core
domains
name= string.
author= n1(name); n2(name, name); etal(name).
publisher= string.
title= string.
volume= journal(title, publisher); book(title, author).

class facts - bib

num:integer := 0.
item:(volume, integer) nondeterm.

class predicates
prtDataBase:().

clauses
classInfo("main", "factread").

prtDataBase() :- item(V, I),

stdio::write(I, "=", V), stdio::nl,
fail.
prtDataBase().
132 Facts

clauses
run():- console::init(),
file::consult("bibliography.fac", bib),
prtDataBase().
end implement main
goal
mainExe::run(main::run).
You need to transfer the file
bibliography.fac
created by facttest to factread\exe. Then, run the program
factread.exe.
You will see that the program will read the database, and print it.

11.1 Class file

You have used class file to save facts to a file. There are many other useful
predicates in class file, that you will learn in this section.

11.1.1 Reading and writing a string

You often need to read a text file, do something to it, and write it back. In
the past, when computers did not have much memory, it was necessary to
read streams of characters. Nowadays, even a personal computer memory
capacity is on the order of gigabytes. Therefore, the best approach to the
problem is to read the whole file into a string, and use the powerful string
class to do the processing. Prolog has two predicates to do this kind of job:
readString:(string FileName, boolean IsUnicodeFile)
-> string String procedure (i,o).
writeString:( string Filename, string Source,
boolean IsUnicodeFile) procedure (i,i,i).
Both predicates have a version that does not require specifying whether the
file is unicode.
readString:(string FileName)
-> string String procedure (i).
writeString:(string Filename, string Source) procedure (i,i).
11.2 Constants 133

%File: main.pro
implement main
open core

constants
className = "main".
classVersion = "filetest".

clauses
classInfo(className, classVersion).

clauses
run():-
console::init(),
Str= file::readString("test.txt", Unicode),
stdio::write(Str),
S_Upper= string::toUpperCase(Str),
file::writeString("upper.txt", S_Upper, Unicode),
succeed(). % place your own code here
end implement main

goal
mainExe::run(main::run).

Figure 11.1: Reading a string from file

Figure 11.1 shows an example of using these predicates. In the example,

one reads the string from a file, converts it to uppercase, and write the
uppercase version to another file.
I believe that you will be able to figure out all the functionality of class
file from the help menu. Therefore, I will not linger on this topic.

11.2 Constants
In order to learn how to use constants, let us build a form with a task menu;
Visual Prolog, and most other languages as well, represent menu options by
integer numbers; this representation is fine for the computer, but not as good
for people, who will analyze the program. Therefore, Visual Prolog represent
these these handles to menu options by constants with meaningful names.
134 Facts

• Create a new project.

Project Name: janua

UI Strategy: Object-oriented GUI (pfc/gui)

By the way, janua is the Latin word for door; window is a small door,
or janella; the door of the year is mensis januarius, or january.

• Create a New Form in New Package; let the form be named txtWin.
From the Properties-dialog, add a TaskMenu to txtWin.

• Build the project. Use the Events-tab of the Properties-dialog to intro-

duce the following snippet into the MenuItemListener:

clauses
onMenuItem(_Source, M) :-
Item= resourceIdentifiers::id_help_contents,
M= Item, !,
stdio::write("Hello").
onMenuItem(_Source, _M) .

Enable the File/New option of the Taskmenu, and add the snippet

clauses
onFileNew(S, _MenuTag) :-
X= txtWin::new(S), X:show().

to TaskWindow.win/CodeExpert/Menu/TaskMenu/id_file/id_file_new.
Chapter 12

Classes and Objects

Let us create an exploratory console project, in order to get a better under-

standing on how classes and objects work.
• Create a new console project.
Project Name: classexample
UI Strategy: console
Build/Build the application. Create a new class: account. Do not
untick the Creates Objects box.
• Modify the file account.i as shown:

% File:account.i
interface account
open core
predicates
ssN:(string SocialSecurity) procedure (o).
setSocialSecurity:(string SSN) procedure(i).
deposit:(real).
end interface account

Class account will produce a new account object whenever you call the
method
A= account::new()
You can send messages to this object. For example, you may set the
SocialSecurity Number of the customer:
A:setSocialSecurity("AA345"),

135
136 Classes and Objects

Afterwards, you may retrieve this very same Social Security Number.

A:ssN(SocialSecurity),

Both ssN/1 and setSocialSecurity/1 must be declared in the in-

terface, which is in file account.i. An account needs other methods,
besides the ssN/1 and setSocialSecurity/1. For instance, you will
need methods for deposits, for passwords, etc. I will leave to your
immagination the creation of these methods. An interesting addition
would be methods for cryptography.

• Modify account.pro as shown below.

% File: account.pro
implement account
open core
facts - customer
funds:real := 3.0.
personalData:(string Name,
string SocialSecurity) single.
clauses
classInfo("account", "1.0").

personalData("", "").

ssN(SS) :- personalData(_, SS).

setSocialSecurity(SSN) :- personalData(N, _),

assert(personalData( N, SSN)).

deposit(Value) :- funds := funds+Value.

end implement account

• Modify run/0 in file main.pro, in order to test the new class.

run():- console::init(),
A= account::new(), A:setSocialSecurity("AA345"),
A:ssN(SocialSecurity),
stdio::write(SocialSecurity), stdio::nl.
12.1 Object facts 137

12.1 Object facts

Objects have facts, that can be used to hold their state. The philosopher
Wittgenstein was very fond of facts, and state of things. Let us take a look
at the first few sentences of his Tractatus Logico-Philosophicus.
1. The world is everything that is the case. In German, the
philosopher used the word Fall: Die Welt ist alles, was der Fall ist. The
Latin word casus means fall, which explains the translations. However,
what is falling in the world? The dice of Fortune are falling. There are
a lot of possible events. The Fortune chooses with her dice what will
happen. At least, this is what the Romans believed. Do you remember
that Julius Cesar said Alea jacta est? I know! He said it in Greek, and
Suetonius had it translated to Latin, but the point remains.

• The world is the totality of facts, not of things.= Die

Welt ist die Gesamtheit der Tatsachen, nicht der Dinge. Did you
remarked how Wittgenstein liked facts? Later on, you will see
that he thought that facts must have states.
• The world divides into facts. From now on, I will spare you
the German original.
• The world is determined by the facts, and by these begin
all the facts.
• For the totality of facts determines both what is the
case, and also all that is not the case. Pay attention
that Wittgenstein takes into consideration both facts that hap-
pen (what is the case), and also those that are mere possibilities
(all that is not the case).

2. What is the case, the fact, is the existence of atomic facts.

• An atomic fact is a combination of objects.

• The object is simple.
• Objects contain the possibility of all states of affairs.
In Visual Prolog, object facts, and variables are used to represent
the state of an object.
• The possibility of its occurrence in atomic facts is the
form of the object.
One can see that the idea of classes and objects is an old idea in Western
Philosophy. The same goes for states. In the example, the state objects
138 Classes and Objects

are individual accounts. The state of the account (our object) specifies the
funds, the name of the account holder, his/her name, his/her Social Security
Number, the password, etc. This state is represented by a fact, and a variable:

facts - customer
funds:real := 3.0.
personalData:(string Name,string SocialSecurity) single.

Interface predicates are used to access the facts that give the state of the
object.

% File:account.i
interface account
open core
predicates
ssN:(string SocialSecurity) procedure (o).
setSocialSecurity:(string SSN) procedure(i).
deposit:(real).
end interface account

About eight persons wrote me saying the name of Wittgensteing book

should be “Tratactus Logicus-Philosophicus”. Dr. Mary Rose Shapiro went
into a long discussion on the differences between Classical and Philosophical
Latin. She claims that the name as I wrote it was a mistake commited by
Ogden, when he translated the book into English. My knowledge of the
subject does not go to such a depth, but I decided to keep the title as it is
in every bookshelf.
Chapter 13

Giuseppe Peano

Giuseppe Peano was one of the greatest mathematicians of all time. He wrote
a small book whose title is
Arithmetices Principia Novo Methodo Exposita
From this, you can figure out how good he was. Not many people write things
in Latin at this time and age, and find readers. Gauss was one of these few
Mathematicians that could afford writing in Latin. Peano was another one.
Of course, Gauss wrote mainly in Latin. Peano wrote mostly in French,
the language of science at the beginning of the last century, and in Italian,
his mother tongue. The only thing he wrote in Latin was the Arithmetices
Principia, a very short book of 29 pages. There, he proposes to the world
the modern notation for formal logics, a branch of Mathematics that gave
origin to Prolog. He also talks about other things, like number theory, and
recursion. I strongly suggest that you learn Latin, and read Peano’s book.

13.1 Turtle graphics

Before diving into Peano’s contributions to science (and there are many),
let us implement turtle graphics in Prolog. To this end, create an Object-
oriented GUI, with the following project settings:
General
Project name: peano
UI Strategy: Object-oriented GUI (pfc/gui)
Target type: Exe

• Create a New in New Package form: canvas (section 2.1). The name
of the form is canvas, and it is at the root of the peano project.

139
140 Giuseppe Peano

• Create a class named curve: Go to the project window, select the folder
canvas, choose the option File/New in Existing Package from the task
window, and create the class. Untick the Create objects box. Build the
application.

• Create a package called turtleGraphics in the directory C:\vispro, and

insert a class named turtle into it. To accomplish this step, choose the
option File/New in New Package from the task menu; when in Create
project Item dialog, click the Browse button for the Parent Directory,
and choose the C:\vispro\ directory from the Set New Directory dia-
log. Do not forget to check off the Create objects box. Be sure to put
the package turtleGraphics outside the project peano. This way, it will
be easier to reuse it. Build the application.

• Enable the File/New option of the application’s TaskMenu.mnu.

• Add the following code

clauses
onFileNew(S, _MenuTag) :-
X= canvas::new(S),
X:show().

to TaskWindow.win/CodeExpert/Menu/TaskMnu/id_file/id_file_new.

• Edit curve.cl and curve.pro as in figure 13.1. Edit files turtle.cl and
turtle.pro, as shown in figures 13.2 and 13.3. Build the application.

• Add the following code to the PaintResponder of canvas.frm:

clauses
onPaint(S, _Rectangle, _GDIObject) :-
W= S:getVPIWindow(),
curve::drawCurve(W).

Build and execute the project. Whenever you choose the option File/New
from the application menu, a window with a star will pop up on the screen.

13.2 Turtle states

The class turtle implements a drawing turtle, like in Logo, a language
created by Papert in an effort to forward the ideas of Jean Piaget. The
turtle is an object that moves on a window, leaving a track behind itself.
13.2 Turtle states 141

% File: curve.cl
class curve
open core, vpiDomains

predicates
classInfo : core::classInfo.
drawCurve:(windowHandle).
end class curve

%File: curve.pro
implement curve
open core, vpi, vpiDomains, math

class predicates
star:(windowHandle, integer, real, integer) procedure (i, i, i, i).
clauses
classInfo("plotter/curve", "1.0").

star(_Win, 0, _A, _L) :- !.

star(Win, N, A, L) :- turtle::right(A),
turtle::forward(Win, L),
star(Win, N-1, A, L).

drawCurve(Win) :- star(Win, 10, 3*pi/5, 40).

end implement curve

Figure 13.1: Files curve.cl and curve.pro

A turtle has two state variables, i.e., its position and the direction of its
movement. These state variables are given by a fact:

class facts
turtle:(pnt, real) single.
clauses
turtle(pnt(200, 200), -pi/2).
...

Consider the predicate:

forward(Win, L) :- turtle(P1, Facing), P1= pnt(X1, Y1),

X2= math::round( X1+ L*cos(Facing)),
Y2= math::round(Y1+L*sin(Facing)),
P2= pnt(X2, Y2), drawline(Win, P1, P2),
assert(turtle(P2, Facing)).
142 Giuseppe Peano

It implements the forward movement of the turtle. If the turtle is at position

P1=pnt(X1, Y1), and moves a distance L in a direction that it is Facing,
one can find its new position by projecting L on the X and Y axis:

X2 = X1 + L × cos(F ) (13.1)
Y2 = Y1 + L × sin(F ) (13.2)

After getting the new position P2= pnt(X2, Y2), forward/2 uses
drawline(Win, P1, P2)
to connect point P1 to P2. Finally, information about the new position of
the turtle is asserted into the database:
assert(turtle(P2, Facing))
The predicate move/1 is similar to forward/2, but does not connect the new
position of the turtle to the old one. Predicates right/1 and left/1 turn
the turtle to the right and to the left, respectively. Their implementations
are straightforward.

13.3 Recursion
Peano’s recursion postulate can be stated as follows: In order to prove that
a predicate is true for any natural number,
• Prove that it is true for 0.

• Prove that it is true for N , if it is true for N − 1.

Of course, this version of the postulate runs towards 0, and it is written in
English. The original runs towards ∞ and is in Latin. Cute, isn’t it? In any
case, let us examine the star/4 predicate(see figure 13.1) under the light of
Peano’s postulate. The first clause
star(_Win, 0, _A, _L) :- !.
says: You draw 0 points to a star if you do nothing. The second clause
star(Win, N, A, L) :- turtle::right(A),
turtle::forward(Win, L),
star(Win, N-1, A, L).
says: You draw N points to a star if you draw a point (turn A radians to the
right, and go L pixels forward), and proceed to draw the other N − 1 points.
13.4 Peano’s curve 143

% File: turtle.cl
class turtle
open core, vpiDomains

predicates
classInfo : core::classInfo.
forward:(windowHandle, integer) procedure.
move:(integer) procedure.
right:(real) procedure.
left:(real) procedure.
end class turtle

Figure 13.2: File turtle.cl

13.4 Peano’s curve

Peano designed a recursive curve that fills the space, which the mathemati-
cians consider to be a very interesting property. What I really wanted to
draw when I created the peano project was the Peano’s curve, then I de-
cided to start with an easier example. In any case, to get Peano’s curve, all
you have to do is to replace curve.pro given in figure 13.4 for the file given
in 13.1.

13.5 Latino Sine Flexione

You have seen that Peano created the modern notation for Logic. He also
came out with the idea of recursion. Since recursion is a basic Prolog pro-
gramming scheme, you could say that Peano brought two important contri-
butions to Logic Programming. However, he did not stop here. Prolog was
invented by Colmerauer for Natural Language Processing:

The programming language, Prolog, was born of a project aimed

not at producing a programming language but at processing nat-
ural languages; in this case, French (Colmerauer).

In the case of Natural Language, Peano had an interesting proposal. Since

most European languages have borrowed a large number of Latin words, why
not use Latin for scientific communication? All you need to do is select Latin
words that are common to the major European languages and you have a
vocabulary large enough to express almost any idea. The problem is that
Latin is a difficult language, with all those declensions and verb forms. Peano
took care of this too: In his brand of Latin, he eliminated all inflexions. Let
144 Giuseppe Peano

% File: turtle.pro
implement turtle
open core, vpiDomains, vpi, math
class facts
turtle:(pnt, real) single.

clauses
classInfo("turtleGraphics/turtle", "1.0").

turtle(pnt(80, 80), -pi/2).

forward(Win, L) :- turtle(P1, Facing), P1= pnt(X1, Y1),

X2= math::round( X1+ L*cos(Facing)),
Y2= math::round(Y1+L*sin(Facing)),
P2= pnt(X2, Y2), drawline(Win, P1, P2),
assert(turtle(P2, Facing)).

move(L) :- turtle(P1, Facing), P1= pnt(X1, Y1),

X2= round( X1+ L*cos(Facing)),
Y2= round(Y1+L*sin(Facing)),
P2= pnt(X2, Y2), assert(turtle(P2, Facing)).

right(A) :- turtle(P1, Facing), NewAngle= Facing+A,

assert(turtle(P1, NewAngle)).

left(A) :- turtle(P1, Facing), NewAngle= Facing-A,

assert(turtle(P1, NewAngle)).
end implement turtle

Figure 13.3: turtle.pro

us take a look at a few quotations from Key to Interlingua, where Interlingua

is the official name for Latin without inflexions.

13.6 Quotations from Key to Interlingua

1. Interlingua adopts every word common to English, German, French,
Spanish, Italian, Portuguese, Russian, and every Anglo-Latin word;

2. Every word has the form of the Latin stem or root or radical.

The vocabulary of Interlingua is composed of a great many Latin words which

have come into English, and therefore are easily comprehended by an English-
speaking person. It includes all terms used in the scientific nomenclatures of
botany, chemistry, medicine, zoology, and other sciences. Many of these
13.6 Quotations from Key to Interlingua 145

implement curve
open core, vpi, vpiDomains, math
class predicates
peano:(windowHandle, integer, real, integer) procedure (i, i, i, i).

clauses
classInfo("plotter/curve", "1.0").

peano(_Win, N, _A, _H) :- N<1, !.

peano(Win, N, A, H) :- turtle::right(A),
peano(Win, N-1, -A, H),
turtle::forward(Win, H),
peano(Win, N-1, A, H),
turtle::forward(Win, H),
peano(Win, N-1, -A, H),
turtle::left(A).

drawCurve(Win) :- turtle::move(-60), peano(Win, 6, pi/2, 5).

end implement curve

Figure 13.4: Peano’s curve

terms are either Greek or Greco-Latin. A few necessary classical Latin words
without international equivalents are a part of the vocabulary, to wit:
Latino English Example

que that Interlingua adopta vocabulo que existe

in Anglo, Germano, Franco et Russo.
de of Latino habe praecisione de expressione.
et and et cetera, et al.
sed but Vocabulario de Latino non es formato ad
arbitrio, sed consiste de vocabulo internationale.
isto this, that pro isto (because of this)
illo he, it Illo es Americano.
illa she Illa es Americana
me I, me Me programma in Prolog.
te you Te programma in Prolog.
nos we Nos programma in Prolog.
vos you guys Vos programma in Prolog.
id it i.e., id est
qui? who? Qui programma in Prolog?
que? what, which Te programma in que?
omne all Illo es omnepotente= He is omnipotent.
146 Giuseppe Peano

Latino should not have any kind of inflection. The plural is formed by using
the word plure as in:
Plure vocabulo in Latino es in usu in Anglo
The words of Latino are in use in English.
However, Peano was a democrat, and since most people wanted to form the
plural with s, he gave his approval to an optional s-plural. Adjectives can be
created from nouns with help of preposition de:
• de fratre = fraternal

• de amico = friendly, amicable

Of course, there are many words that are adjectives by nature: internationale,
nationale, breve (short), magno (large, big), commune (common), etc. You
can use these adjectives as adverbs, without modification. You can also form
adverbs using the word modo: in modo de fratre (fraternally), in modo de
patre (paternally), in modo rapido (rapidly, speedily). I guess that this is
enough. Here is an example of text in Latino sine flexione:
Interlingua es lingua universale que persona in modo facile scribe
et intellige sine usu de speciale studio. Interlingua adopta vo-
cabulo que existe in Anglo, Germano, Franco et Russo. Usu de
Interlingua es indicato pro scientifico communicatione et pro com-
merciale correspondentia. Usu de Interlingua in internationale
congressu facilita intelligentia et economiza tempore. In addi-
tione Interlingua es de valore educationale nam (because) illo es
de logico et naturale formatione.
Why do we need an international language, like Peano’s Interlingua, if we
already have English? I think that we don’t. However, people who propose
Interlingua have a few interesting arguments to defend their project. For
instance, humankind cannot let the international language change as eco-
nomic hegemony and cultural prestige shift from one nation to the other. In
the XIX century, French was the language of diplomacy. Then, the Amer-
ican economy surpassed the French, and English became the international
language. In this century, China will be the dominant superpower, and will
impose written Chinese to other nations. There is an interesting story about
this point. People still use Latin to describe plants. Thus, if you discover a
new plant, you must describe it in Latin. This was also the case of zoologists
and anatomists, but they shifted to English not long ago. Therefore, a few
13.6 Quotations from Key to Interlingua 147

botanists proposed to follow suit, and adopt a modern language in Botany

too. The problem is that the language they wanted was Chinese, not English!
There is another argument in favor of Latin. Languages like French,
English and Chinese are difficult to learn even by people that speak them
as their mother tongues. Not many Frenchmen can write acceptable French.
Very few Americans can come close to English written communication. My
son is at the eighth grade in Mount Logan Middle School, thus I often go
there to check his progress. Most of his fellow students cannot read, let
alone write. If this happens to Americans, what can I say about a Japanese
who tries to learn English? Summary: Ten years is not enough to teach an
American to read and write English prose, but in ten days he can write more
than acceptable Latino. But let us leave this discussion to the wise, and
build a project that takes a text in LsF (Latino sine Flexione), and explains
difficult words and expressions.

• Create a new GUI project: LsF.

• Add a new package to the project tree: LsF/editor. Add the

editorControl

package to the project root. To this end, choose the option File/Add
from the task menu, browse for the Visual Prolog installation directory,
and select the file

\pfc\gui\controls\editorControl\editorcontrol.pack

Build the application.

• Create a new form: editor/txtWin. Resize the form window, since

the editorControl must be quite large. Add a custom control to the
form. The custom control is the one represented by the icon of the
key. The system will offer you a list of custom control names; choose
editorControl. Resize the editor pane.
148 Giuseppe Peano

• Build/Build the application.

• Create an editor/grammar class with the Creates Objects box unticked.

If you do not remember how to do it, take a look at section 4.4. Build
the application once more.

Modify the grammar.cl and grammar.pro files as shown below.

%File: grammar.cl
class grammar
open core
predicates
classInfo : core::classInfo.
dic:(string).
end class grammar

%File: grammar.pro
implement grammar
open core
clauses
classInfo( "editor/grammar", "1.0").
dic(S) :- stdio::write(S), stdio::nl.
end implement grammar

Enable the File/New option, of the application task menu. Add the snippet

clauses
onFileNew(S, _MenuTag) :- X= txtWin::new(S), X:show().

to ProjWin/TaskWindow.win/Code Expert/Menu/TaskMenu/id_file/id_file_new.

Add the snippet

clauses
onHelpClick(_Source) = button::defaultAction :-
editorControl_ctl:getSelection(Start, End),
Txt=editorControl_ctl:getText(Start, End),
grammar::dic(Txt).

to the ClickResponder of the help_ctl-button of txtWin.frm. In the same

file you added the last code for onHelpClick (txtWin.pro), modify the defi-
nition of predicate new as shown in figure 13.5. Finally, create a file LsF.txt
in the \exe directory with the text below.
13.7 Examples 149

Interlingua es lingua universale que persona in modo

facile scribe et intellige sine usu de speciale studio.
Interlingua es indicato pro scientifico communicatione
et pro commerciale correspondentia. Usu de Interlingua
in internationale congressu economiza tempore.

class predicates
tryGetFileContent:(
string FileName,
string InputStr,
string UpdatedFileName) procedure (i, o, o).
clauses
tryGetFileContent( "", "", "LsF.txt") :- !.
tryGetFileContent(FileName, InputStr, FileName) :-
trap(file::existFile(FileName), _TraceId, fail), !,
InputStr= file::readString(FileName, _).
tryGetFileContent(FileName, "", FileName).

new(Parent):- formWindow::new(Parent),
generatedInitialize(),
PossibleName= "LsF.txt",
tryGetFileContent(PossibleName, InputStr, _FileName), !,
XX=string::concat("Here: ", InputStr),
editorControl_ctl:pasteStr(XX).

Figure 13.5: Inserting a file into an editor

If you open a new txtWin from the application, the above text will appear
in the window. If you select part of the text, and press the button Help from
the TaskMenu of the application, you will echo the selected text on the
message window. The next step is to add dictionary entries to the selected
text, before writing it on the message window (see figure 13.6). To this end,
we start by breaking down the selection into a list of words. In figure 13.6,
this is done by the following predicate:
strTok:(string, string_list) procedure (i, o).
Natural Language Processing (NLP) is the most difficult branch of Computer
Science, and outside the scope of this doc. If you want to learn the state of
the art, you can start by adding more entries to the program of figure 13.6.

13.7 Examples
The examples of this chapter are in directories peano, and LsF.
150 Giuseppe Peano

%File: grammar.pro
implement grammar
open core, string
class predicates
strTok:(string, string_list) procedure (i, o).
addExplanation:(string, string) procedure (i, o).
class facts
entry:(string, string).
clauses
classInfo( "editor/grammar", "1.0").

entry("que", "that").
entry("sine", "without").
entry("es", "am, are, is, was, were, to be").
entry("de", "of").
entry("et", "and").

addExplanation(T, R) :- entry(T, Meaning), !,

R= format("%s= %s\n", T, Meaning).
addExplanation(_T, "").

strTok(S, [T1|L]) :- frontToken(S, T, Rest), !,

addExplanation(T, T1), strTok(Rest, L).
strTok(_S, []).

dic(S) :- strTok(S, SList),

Resp= concatList(SList),
stdio::write(Resp), stdio::nl.
end implement grammar

Figure 13.6: A dictionary of “difficult” LsF words

Chapter 14

L-Systems

Aristid Lindenmayer was a Dutch botanist who invented a clever way to

describe plant shapes. As you know, botanists spend a great deal of their time
describing plant shapes. Their descriptions are usually made in a simplified
version of Latin, invented by Carl Linaeus.
Lindenmayer’s method of describing plant shapes is based on the plotting
turtle that you learned in section 13.1 on page 139 . On the internet, you will
find a vast quantity of material on the subject. Therefore, I will limit myself
to an example, that you can use as a starting point to your own projects.

• Create a GUI project: Lsys

• Add a package to the root of the project tree: aristid

• Add a form to package aristid: canvas.

• Build the application.

• Add a class to aristid: draw. Untick the Create Objects box. Build
the project to insert the class into the project tree.

• Enable the option File/New of the TaskMenu. Add

clauses
onFileNew(S, _MenuTag) :-
X= canvas::new(S), X:show().

to TaskWindow.win/CodeExpert/Menu/TaskMenu/id_file/id_file_new.

151
152 L-Systems

14.1 Class draw

Then, edit the draw class as shown below. Build the application.
%File: draw.cl
class draw
open core, vpiDomains
predicates
classInfo : core::classInfo.
tree:(windowHandle).
end class draw

% File: draw.pro
implement draw
open core, vpiDomains, vpi, math

clauses
classInfo("plotter/draw", "1.0").
domains
command = t(commandList); f(integer);
r(integer); m.
commandList= command*.
class facts
pos:(real Delta, real Turn) single.
grammar:(commandList) single.
class predicates
move:(windowHandle, pnt, real, commandList)
procedure (i, i, i, i).
derive:(commandList, commandList)
procedure (i, o).
mv:(windowHandle, pnt, real, command, pnt, real)
procedure (i,i, i, i, o, o).
iter:(integer, commandList, commandList) procedure (i, i, o).

clauses
pos(4.0, 0.1745329).

grammar([f(1)]).

iter(0, S, S) :- !.
iter(I, InTree, OutTree) :- derive(InTree, S),
iter(I-1, S, OutTree).
14.1 Class draw 153

derive([], []) :- !.
derive([f(0)|Rest], [f(0),f(0)|S]) :- !,
derive(Rest, S).
derive([f(_)|Rest],
[f(0), t([r(1),f(1)]), f(0),
t([r(-1),f(1)]), r(1), f(1)|S]) :- !,
derive(Rest, S).
derive([t(Branches)|Rest], [t(B)|S]) :- !,
derive(Branches, B),
derive(Rest, S).
derive([X|Rest], [X|S]) :- derive(Rest, S).

mv(Win, P1, Facing, f(_), P2, Facing) :- !,

pos(Displacement, _Turn),
P1= pnt(X1, Y1),
X2= round( X1+ Displacement*cos(Facing)),
Y2= round( Y1+Displacement*sin(Facing)),
P2= pnt(X2, Y2),
drawline(Win, P1, P2).
mv(_Win, P1, Facing, m, P2, Facing) :- !,
pos(Displacement, _Turn),
P1= pnt(X1, Y1),
X2= round( X1+ Displacement*cos(Facing)),
Y2= round(Y1+Displacement*sin(Facing)),
P2= pnt(X2, Y2).
mv(_Win, P1, Facing, r(Direction), P1, F) :- !,
pos(_Displacement, Turn),
F= Facing+ Direction*Turn.
mv(Win, P1, Facing, t(B), P1, Facing) :-
move(Win, P1, Facing, B).

move(_Win, _P1, _Facing, []) :- !.

move(Win, P1, Facing, [X|Rest]) :-
mv(Win, P1, Facing, X, P, F),
move(Win, P, F, Rest).

tree(Win) :- grammar(Commands),
iter(5, Commands, C),
Facing= -pi/2,
Point= pnt(100, 250),
move(Win, Point, Facing, C).
end implement draw
154 L-Systems

Finally, add the snippet

clauses
onPaint(S, _Rectangle, _GDIObject) :-
draw::tree(S:getVPIWindow()).

to the PaintResponder. Build and execute the application.

Open a canvas form by choosing the option File/New from
the application menu. It will present a tree as shown in the
figure on the left hand side.

14.2 Examples
The example of this chapter is in directory Lsys.
Chapter 15

Games

In this chapter, you will learn how to implement games in Visual Prolog. Of
course, a good game is like a good poem. It is not enough to learn Russian,
and versification rules in order to write like Pushkin. A poet, besides knowing
the language in which he will write, needs imagination, creativity, etc. The
same goes for a good game writer. You will need a lot of creativity to come
up with something like Tetris or Space Invaders. I used these examples to
show you that good games do not depend on fancy graphics or sound effects.
The important features are suspense, continued tension and a sense of reward
for good performance.
Once, a young man asked Lope de Vega how to write a poem. The answer
was: It is simple: Uppercase in the beginning of each verse, and a rhyme at
the end. “What do I put in the middle?”, asked the boy. In the middle, “hay
que poner talento”, stated Lope de Vega.
This lesson has to do with a few techniques, like uppercase letters and
rhymes. To get a good game, you will need to add talent to the recipe.
The game you will implement is simple. There is a worm moving fast on
the screen. The player must change its direction in order to prevent it from
hitting obstacles and walls. You will need to implement the following items:

1. A window, where the worm will move. The window must have four
buttons, to control the movement of the worm.

2. A state class, that describes the direction of the movement, and the
position of the worm. It exports the predicates mov/0, to change the
position of the worm, and turn/2, that can be used to turn it around.

3. A draw class, that draws a snapshot of the state of the worm. It

exports the snapshot(Win) predicate. Pay attention to this detail:
The predicate snapshot/1 cannot draw the worm directly onto the

155
156 Games

Figure 15.1: Lawn form

window. This would make the screen flicker. It must build a picture,
and then draw it onto the screen.

4. A timer, that calls mov/0 and snapshot/1, at fixed intervals, in order

to create an illusion of movement.

We know what to do; let us do it. I have avoided creating objects until
now. In fact, every time that you created a class, you were recommended
to untick the Creates Objects radio button. However, if we rigidly maintain
this state of mind, our game will have a very nasty property. If you lose,
quit the current game, and try to start another one, you will find out that
the new game is not in the initial state. In fact, it is in the state where you
lost the old one. The problem is that the variables used to describe a state
have memory that lasts from one form to another. Objects were invented to
prevent this kind of nasty behavior. Therefore, let us take this very good
opportunity to learn how to create objects, and deal with object states.

• Create a project: game.

Project Name: game

UI Strategy: Object-Oriented GUI (pfc/vpi)
TargetType: Exe
Base Directory: C\vip70
Sub-Directory: game
 157

• Create a package: playground.

• Create a form: lawn. You will add four buttons to the lawn form; refer
to figure 15.1 for the design. In the Properties dialog, it is a good idea
to change the names of the buttons from pushButton_ctl to up_ctl,
down_ctl, left_ctl, and right_ctl. Build the application in order
to include the form. Then enable File/New, and add

clauses
onFileNew(S, _MenuTag) :- X= lawn::new(S), X:show().

to ProjWin/Code Expert/Menu/TaskMenu/id_file/id_file_new.

• Create class objstate inside playground. Do not untick Create Objects

for the class objstate. Insert code shown in figure 15.3; NB: Class
objstate must have an interface in file objstate.i. Build the application.

• Create class draw, inside playground. Untick Create Objects. Insert

code of figure 15.2. Build the application.

• Create a class click to hold the clock. Untick Creates Objects.

%File: click.cl
class click
open core, vpiDomains
predicates
bip:(window).
kill:().
end class click

% File:click.pro
implement click
open core
class facts
tm:vpiDomains::timerId := null.
clauses
bip(W) :- tm := W:timerSet(500).
kill() :- vpi::timerKill(tm).
end implement click

Build the application to include click into the project.

158 Games

• Add

clauses
onDestroy(_Source) :- click::kill().

to the DestroyerListener from the Properties-dialog of lawn.frm.

• Add

class facts
stt:objstate := objstate::new().
clauses
onShow(Parent, _CreationData) :- stt := objstate::new(),
stt:init(), click::bip(Parent).

to ShowListener from the Properties-dialog of lawn.frm.

• Add code for the ClickResponder corresponding to each of the buttons.

clauses
onDownClick(_S) = button::defaultAction() :-
stt:turn(0, 10).

clauses
onLeftClick(_S) = button::defaultAction() :-
stt:turn(-10, 0).

clauses
onRightClick(_S) = button::defaultAction() :-
stt:turn(10, 0).

clauses
onUpClick(_S) = button::defaultAction() :-
stt:turn(0, -10).

• clauses
onTimer(_Source, _TimerID) :- stt:mov(),
R= rct(10, 10, 210, 210),
invalidate(R).
to TimeListener from the Properties-dialog of lawn.frm.
 159

• Add

clauses
onPaint(_Source, _Rectangle, GDIObject) :-
draw::snapshot(GDIObject, stt).

to the PaintResponder frm the Properties-dialog of lawn.frm.

Build and run the program. After choosing the option File/New, a win-
dow with a worm pops up. You can control the worm using the four buttons.
It is up to you to improve this game. You can add obstacles, like walls.
You can also put little apples for the worm to eat. Finally, you can write
a 3D game. The 3D game is very simple indeed. The segments have three
coordinates, and you must find their projections on the plane before drawing
them. If you info-fish using the Internet, you certainly will find formulae to
calculate the perspective projection of a point. Alex Soares wrote the original
game on which I based this tutorial. His game is in 3D, in Visual Prolog 5.2,
using Direct X, a library that has a lot of resources for 3D games.
Let us keep to our study of drawing predicates. If you draw figures at
every tick of the clock, you will have created an animation. However, if
you perform the drawings directly on a visible window, the animation will
blink, which is not very nice. The solution for this problem is to perform the
drawing on a hidden window, and transfer it to the visible window only after
it is ready for showing. The predicate
W= pictOpen(Rectangle)
opens a hidden window for you to draw things without worrying about a
blinking animation. The predicate
Pict= pictClose(W)
creates a picture from the contents of a hidden window; at the same time, it
closes the window. Finally, the predicate
pictDraw(Win, Pict, pnt(10, 10), rop_SrcCopy)
transfers the picture to a visible window. Here is an example of code con-
taining these three important predicates:
snapshot(Win, S) :- S:mov(), !,
Rectangle= rct(0, 0, 200, 200),
W= pictOpen(Rectangle), draw(W, S),
Pict= pictClose(W),
Win:pictDraw(Pict, pnt(10, 10), rop_SrcCopy).
160 Games

The argument rop_SrcCopy specifies the mode of the raster operations. It

says that the system should copy the figure from the source to the destiny.
There are other modes of transferring figures from one window to another:

• rop_SrcAnd, performs a logical-AND between the bits of the figure,

and the background. You can use it for creating sprites.

• rop_PatInvert, inverts the colors.

• rop_SrcInvert inverts the colors of the source.

%File draw.cl
class draw
open core, vpiDomains
predicates
classInfo : core::classInfo.
snapshot:(windowGDI Win, objstate).
end class draw

%File draw.pro
implement draw
open core, vpiDomains, vpi
class predicates
draw:(windowHandle, objstate) procedure.

clauses
classInfo("playground/draw", "1.0").

draw(W, S) :- S:segm(Rectangle),
vpi::drawEllipse(W, Rectangle), fail.
draw(_W, _S).

snapshot(Win, S) :- S:mov(), !,
Rectangle= rct(0, 0, 200, 200),
W= pictOpen(Rectangle),
draw(W, S),
Pict= pictClose(W),
Win:pictDraw(Pict, pnt(10, 10), rop_SrcCopy).
end implement draw

Figure 15.2: draw.cl and draw.pro

 161

%File:objstate.cl
class objstate : objstate
end class objstate

%File:objstate.i
interface objstate
open core, vpiDomains
predicates
init:().
turn:(integer, integer).
mov:().
segm:(rct) nondeterm (o).
end interface objstate

%File:objstate.pro
implement objstate
open core, vpiDomains
facts
w:(integer, pnt) nondeterm. % worm
d:(integer, integer) single. % direction

clauses
init() :- assert(w(1, pnt(110, 100))),
assert(w(2, pnt(120, 100))), assert(w(3, pnt(130, 100))),
assert(w(4, pnt(140, 100))), assert(w(5, pnt(140, 100))).

d(10, 0).

mov() :- retract(w(1, P)), P= pnt(X1, Y1),

d(DX, DY), P1= pnt(X1+DX, Y1+DY), assert(w(1, P1)),
retract(w(2, P2)), assert(w(2, P)),
retract(w(3, P3)), assert(w(3, P2)),
retract(w(4, P4)), assert(w(4, P3)),
retract(w(5, _)), assert(w(5, P4)), !.
mov().

segm(rct(X, Y, X+10, Y+10)) :- w(_, pnt(X, Y)).

turn(DX, DY) :- assert(d(DX, DY)).
end implement objstate

Figure 15.3: objstate

162 Games

15.1 Object Facts

Pay careful attention to one particular detail: When dealing with classes that
do not create objects, the state of a class is established by class facts; here
is how a class fact is declared:
class facts
city:(string Name, pnt Position).
conn:(pnt, pnt).
clauses
city("Salt Lake", pnt(30, 40)).
city("Logan", pnt(100, 120)).
city("Provo", pnt(100, 160)).

conn(pnt(30, 40) , pnt(100, 120)).

conn(pnt(100, 120), pnt(100, 160)).

A class fact is shared by all objects of the class. This means that, if an object
predicate (a predicate declared in the interface) changes a fact, it is changed
for all objects of the class. Therefore, if you had kept the state of the worm
in class facts, the game would have the following nasty feature:

Nasty Game Property — Whenever the player loses a worm, quits the
current game, and plays another one, he will find that the new worm is not
in the initial state. In fact, it is in the state where he lost the old one.

The problem is that the class facts used to describe the state of the worm
have class memory that lasts from one form to another, from one worm
object to another. The solution is simple: Declare the facts without adding
the keyword class to the head of the declaration; then there will be a private
set of facts for each object. Here is how to declare object facts:
facts
w:(integer, pnt) nondeterm. % worm
d:(integer, integer) single. % direction
clauses
init() :- assert(w(1, pnt(110, 100))),
assert(w(2, pnt(120, 100))), assert(w(3, pnt(130, 100))),
assert(w(4, pnt(140, 100))), assert(w(5, pnt(140, 100))).

d(10, 0).

This scheme of keeping an object state is implemented in the program of

figure 15.3, that appears on the next page.
Chapter 16

Animation

Please, take a look at chapter 8, page 87, if you do not remember how to use
the event painting. You will also need to use masks to create figures with a
transparent background; this theme is also explained in chapter 8. Finally,
you may want to recall to memory the use of GDIObject, which is dealt with
in section 3.2, page 28, where you will find material about invalidating a
rectangle for painting as well.

• Create a project.

Project Name: rattlesnake

UI Strategy: Object-oriented GUI (pfc/gui)

In this project, you will animate the famous rattlesnake that appears
on the John Paul Jones’ Don’t tread on me flag.

• Create a new package, and let its name be snake.

• Create a new canvas form. Put it inside the snake package.

• Create class click to hold clock events. Untick Creates Objects. Use
the code given in figure 16.1 to define and implement click.

• Build the application. Enable File/New. Then, add

onFileNew(S, _MenuTag) :-
F= canvas::new(S), F:show(), click::bip(F).

to TaskWindow.win/CodeExpert/Menu/TaskMenu/id_file/id_file_new.

163
164 Animation

%File: click.cl
class click
open core

predicates
bip:(window).
kill:(window).
end class click

%File: click.pro
implement click
open core
class facts
tm:vpiDomains::timerId := null.
clauses
bip(W) :- tm := W:timerSet(1000).
kill(W) :- W:timerKill(tm).
end implement click

Figure 16.1: click class handles the timer

16.1 dopaint
Class dopaint will handle painting, and works more or less like the dopaint
method in Java. Create a class, name it dopaint, and untick the Creates
Objects option. Here is the class definition:

%File: dopaint.cl
class dopaint
open core
predicates
classInfo : core::classInfo.
draw:(windowGDI).
invalidrectangle:(vpiDomains::rct) procedure (o).
end class dopaint

Figure 16.2 shows the implementation of dopaint.pro. Build the application.

16.1 dopaint 165

implement dopaint
open core, vpiDomains

constants
className = "snake/snakestate".
classVersion = "".

class facts
yesno:integer := 0.

class predicates
flipflop:(picture Picture,
picture Mask) determ (o, o).

clauses
classInfo(className, classVersion).

flipflop(Pict, Mask) :- yesno= 0,

yesno := 1,
Pict= vpi::pictLoad("figs\\n0.bmp"),
Mask= vpi::pictLoad("figs\\n0Mask.bmp"), !.
flipflop(Pict, Mask) :- yesno= 1,
yesno := 0,
Pict= vpi::pictLoad("figs\\n1.bmp"),
Mask= vpi::pictLoad("figs\\n1Mask.bmp").

draw(W) :-
P= vpi::pictLoad("figs\\frogs.bmp"),
W:pictDraw(P, pnt(10, 10), rop_SrcCopy),
flipflop(Snake, Mask), !,
W:pictDraw(Mask, pnt(40, 50), rop_SrcAnd),
W:pictDraw(Snake, pnt(40, 50), rop_SrcInvert).
draw(_).

invalidRectangle(rct(40, 50, 100, 100)).

end implement dopaint

Figure 16.2: dopaint.pro

166 Animation

16.2 Handling the timer

When you close canvas, you must kill the timer. To achieve this goal, add

onDestroy(W) :- click::kill(W).

to the DestroyListener of the Properties-dialog of canvas.frm. Then, add

onTimer(_Source, _TimerID) :-
dopaint::invalidRectangle(R),
invalidate(R).

to the TimeListener of the Properties-dialog of canvas.frm. Finally, add the

event handler

onPaint(_Source, _Rectangle, GDIObject) :-

dopaint::draw(GDIObject).

to the PaintResponder of the Properties-dialog of canvas.frm.

16.3 How the program works

The first thing that predicate draw(W) does is to paint a few frogs onto the
background. The frogs are loaded by the following predicate:

P= vpi::pictLoad("figs\\frogs.bmp"),

Of course, you must have the file frogs.bmp in directory figs.

Then, draw(W) obtains the picture and the mask of a rattlesnake.
The flipflop(Snake, Mask) predicate is
designed to alternate two snake paintings, in
order to get the illusion of motion. To achieve
the insertion of the snake onto the background,
you must use masks. If you do not remember
how to do it, read chapter 8 again.
Chapter 17

Text Editor

In this chapter, you will learn how to use the editControl class to create a
text editor.

• Create a new project.

Project Name: editor

UI Strategy: Object-oriented GUI(pfc/gui)

• Add the editorControl package, that is found in the installation direc-

tory of Visual Prolog. Choose the task menu option File/Add and look
for the package in

Visual Prolog 7.x\pfc\gui\controls\editorControl.

• Build the application. Then create a new form edform.frm at the

project root, as shown in figure 17.3. To insert the editorControl into
the form, choose the custom control option in the form prototype; it
is the button that has a Yale key1 (figure 17.1); and then choose edi-
torControl in the Dialog shown in figure 17.2. Change the name of the
control to editorControl_ctl as shown in figure 17.4. Build/Build
the application.

• Enable File/New and add the snippet below to it.

onFileNew(S, _MenuTag) :- F= edform::new(S), F:show().

1
Yale invented this kind of key, that is very popular nowadays. Legend has it that soon
after his invention, Houdini was able to break through the new device.

167
168 Text Editor

Figure 17.1: The insert custom control button

Figure 17.2: Choose Custom Control dialog

17.1 Saving and Loading files

Go to the project tree, and open edform.frm, if it is not already open. Use
the Properties-dialog to add the snippet of figure 17.5 to the ClickResponder
of button:save_ctl; then add the code shown in figure 17.6 to the ClickRe-
sponder of button:load_ctl. This is about all you have to do in order to
create an operational text editor.
17.1 Saving and Loading files 169

Figure 17.3: A form with Edit Control

Figure 17.4: User control properties

170 Text Editor

predicates
onSaveClick : button::clickResponder.
clauses
onSaveClick(_Source) = button::defaultAction() :-
Txt= editorControl_ctl:getEntireText(),
FName= vpiCommonDialogs::getFileName("*.*",
["Text", "*.txt"],
"Save", [], ".", _X), !,
file::writeString(FName, Txt).
onSaveClick(_Source) = button::defaultAction().

Figure 17.5: Code for onSave

predicates
onLoadClick : button::clickResponder.
clauses
onLoadClick(_Source) = button::defaultAction() :-
FName= vpiCommonDialogs::getFileName("*.*",
["Text", "*.txt"],
"Load", [], ".", _X),
file::existFile(FName),
!,
Str= file::readString(FName, _IsUnicodeFile),
editorControl_ctl:pasteStr(1, Str).
onLoadClick(_Source) = button::defaultAction().

Figure 17.6: Code for onLoad

Chapter 18

Printing

In the previous chapter, you learned how to build a text editor. In this
chapter, you will acquire introductory notions of printing.
• Create a project.

Project Name: testPrt

UI Strategy: Object-oriented GUI (pfc/GUI)
Target Type: Exe
Base Directory: C:\vip\codeForTyros

• Build the application, in order to insert prototypes of the Task Window

onto the Project Tree.

• On the Project Tree, double click on the branch TaskMenu.mnu, and

enable the &File/&New/tF7 option from the menu prototype.

• On the Project Tree, right click on the TaskWindow.win, and choose the
option Code Expert from the floating menu, as in figure 2.6, page 21.
The Dialog and Window Expert will pop up. Open the folders Menu,
TaskMenu, and id_file, as shown below.

Select the branch id_file_new, and press the button Add. Then, dou-
ble click on the newly created id_file_new->OnFileNew branch; finally,
replace the onFileNew(_Source, _MenuTag) prototype with the snip-
pet given below.

171
172 Printing

clauses
onFileNew(_Source, _MenuTag) :-
PW=vpi::printStartJob("Recoreco"),
_HRES = vpi::winGetAttrVal(PW,attr_printer_hres),
VRES = vpi::winGetAttrVal(PW,attr_printer_vres),
V_SCR_RES=vpi::winGetAttrVal(PW,attr_screen_vres),
FNT=vpi::fontCreate(ff_Fixed,[], VRES*40 div V_SCR_RES),
vpi::winSetFont(PW,FNT),
vpi::printStartPage(PW),
vpi::drawText(PW, 100, 200, "Before the sunset!"),
vpi::printEndPage(PW),
vpi::printEndJob(PW).

The beauty of Visual Prolog printing scheme is that you deal with the printer
as if it were a normal graphic window. In fact, the first thing you do, in order
to print something, is to open a printer job window.

PW=vpi::printStartJob("Recoreco")

Once you have a window, you can use it to obtain information about the
printer resolution.

_HRES= vpi::winGetAttrVal(PW,attr_printer_hres),
VRES= vpi::winGetAttrVal(PW,attr_printer_vres),
V_SCR_RES=vpi::winGetAttrVal(PW,attr_screen_vres),

By trial and error, and using information about the printer resolution, you
can define a font, that looks good on print.

FNT=vpi::fontCreate(ff_Fixed,[], VRES*40 div V_SCR_RES),

vpi::winSetFont(PW,FNT),

Finally, you can use any drawing predicate to produce a fine printed page.

vpi::drawText(PW, 100, 200, "Before the sunset!"),

Chapter 19

Tab forms and other goodies

• Create a new project:

Name: tab_example
UI Stategy: Object GUI

• New in New Package: Create a package with the name forms

• From the installation directory, File/Add:

pfc\gui\controls\tabControl\tabControl.pack

• New in Existing Package (forms): Create a Control with the name

helloTab; to achieve this goal, select the option Control on the left
hand side pane. Rename the button to button:yourname_ctl.

• Build the application, in order to insert snippets for events. Add the fol-
lowing snippet to the ClickResponder belonging to button:yourname_ctl:

clauses
onYournameClick(_Source) = button::defaultAction :-
Name= edit_ctl:getText(),
Ans= string::concat("Hello, ", Name, "!\n"),
stdio::write(Ans).

173
174 Tab forms and other goodies

19.1 Famous programs

There are two programs that one can call infamous: "Hello, world", and
recursive Fibonacci. In this chapter, you will learn how to place both pro-
grams in a control with tabs. You already created the form to hold the
hello-application. Now, let us create a form for the Fibonacci sequence.
• New in Existing Package (forms): Control with name fibTab; rename
the button to button:fib_ctl.

• Build the application, in order to insert code for events. Add the
following snippet to the ClickResponder belonging to button:fib_ctl:

class predicates
fibo:(integer, integer) procedure (i, o).

clauses
fibo(N, F) :-
if N<2 then F=1
else
fibo(N-1, F1), fibo(N-2, F2),
F= F1+F2
end if.

predicates
onFibClick : button::clickResponder.
clauses
onFibClick(_Source) = button::defaultAction :-
Num= edit_ctl:getText(),
I= toTerm(Num), fibo(I, F),
Ans= string::format("fibo(%d)= %d", I, F),
edit_ctl:setText(Ans).

Build the application again, to make sure that everything is working

properly.
19.1 Famous programs 175

• Create a New form in Existing Package (the existing package is forms)

— Name it forms/tabs. Use the Yale key to insert a tabControl into
the form. Build the application.

• Go to the file tabs.pro, which you can access by clicking the corre-
sponding branch of the project tree, and replace the clause

clauses
new(Parent):-
formWindow::new(Parent),
generatedInitialize().

with the following snippet:

clauses
new(Parent):-
formWindow::new(Parent),
generatedInitialize(),
%
Page1 = tabPage::new(),
Tab1 = helloTab::new(Page1:getContainerControl()),
Page1:setText(Tab1:getText()),
tabControl_ctl:addPage(Page1),
Page2 = tabPage::new(),
Tab2 = fibTab::new(Page2:getContainerControl()),
Page2:setText(Tab2:getText()),
tabControl_ctl:addPage(Page2),
succeed.

• Enable the option File/New of the application file menu.

176 Tab forms and other goodies

• Add the snippet

clauses
onFileNew(S, _MenuTag) :-
W= tabs::new(S), W:show().

to TaskWindow.win/CodeExpert/Menu/TaskMenu/id_file/id_file_new.

• Build the application one last time. Cute, isn’t it?

19.2 Botany
Your interest in Botany may be none. However, there are many reasons for
cultivating this important branch of science. Let us enumerate them.

Saving the world. Global warming is the most serious challenge facing us
today. To protect the planet for future generations, we must reduce
19.2 Botany 177

the concentration of heat-trapping gases by using the technology, and

practical solutions already at our disposal. It turns out that plants of-
fer poetically the only viable way out of this trap that we set ourselves.
Therefore, we need a climate policy iniative to reduce emissions from
tropical deforestation, understand the dynamics of plant interdepen-
dency, in order to check the decrease of bio-diversity, and select rapidly
growing new vegetable species to replace forests that our lack of vision
and direction destroyed in the past. Therefore, if saving the world is
your vocation, here is a good place to start. I believe that saving the
world is also good business, since there will be plenty of money for re-
searchers and specialists in global warming. I am proud to say that my
family is very active in protecting nature, and bio-diversity. I want to
pay homage to my grandmother, an Indian woman (Native American,
not from India) who with my father created an ecological sanctuary
that we maintain until today. I remember that when I was a child, the
most cherished book in our house was Maeterlinck’s L’Intelligence des
fleurs. My father used to make me repeat plusieurs fois Maenterlinck’s
famous saying: When the last tree has been thrown down, the last man
will die clinging to its trunk. To tell you the truth, I never found any
evidence that Maeterlinck said such a sentence; however, my grand-
mother was sure that he did. I hope that a Belgian reader may shed
light on this point.

Computer graphics. In the chapter about Aristid Lindenmeyer you have

seen that Botanists invented clever methods to create computer graph-
ics depicting flowers. You probably remember Bizet’s song:

Votre toast, je peux vous le rendre, señors, señors, car avec

les soldats oui, les toreros peuvent s’entendre.

I could say that Computer Scientist and Botanists can understand each
other, at least in the realms of computer graphics.

Orchids. These exotic plants are among the most beautiful creations of
Nature. They are objects of a cult that owes nothing to wine, or French
cuisine. By the way, the Serbian Nero Wold was fond of wine, French
cuisine, and orchids. If you read Wolf’s books, you will notice that
the suspense is not created by how Wolf solves mysteries, but by how
he will get the money necessary to maintain his expensive cults. He
hires Fritz Brenner, an exceptionally talented Swiss cook who prepares
and serves all his meals, and speaks French when addressing him; he
also hires Theodore Horstmann, an orchid expert who takes care of
178 Tab forms and other goodies

the plant rooms; finally, the most important of his employers is Archer
Goodwin, who works as a detective to earn the money necessary to
keep the expensive habits of his overweight boss.
One of the most interesting features of Botany, as compared with other sci-
ences, it that there is a special kind of Latin internationally used by botanists
for the description, and naming of plants. Since many Botanists do not know
enough Latin, a program to help them deal with this difficult language would
give a head start to the goals of saving the world, cultivating orchids, or cre-
ating cool computer graphics. In this section, I will see how such a program
can be written. Besides this, you will learn how to work with tab-controls,
and listEdit-controls.
The program to teach Botanical Latin is given in the skinner directory
of the folder containing the examples that accompany this document. The
program contains three tab-controls:
• Pronunciation tab. In order to teach you how to pronounce Latin, I
used excerpts from a very famous book called Philosophiae Naturalis
Principia Mathematica; the author of the book is a certain Isaac New-
ton; I am not sure whether you heard about this scientist, but English
scholars consider him to be the greatest scientist who ever lived. They
certainly are exaggerating; notwithstanding Newton was good enough
to write in Latin, and find readers; as you remember, when a scientist
is very good, one says that he could have written in Latin, that he
would find readers; Newton, Peano, Gauss, and others took this saying
literally, and wrote in Latin. In this tab, my son, who speaks fluent
Latin, English, Ancient Greek, and Chinese, will read Newton for you.
• Declensiontab. There are languages where nouns change to show their
function in a sentence. Latin, Sanskrit, German, Russian, Polish are
among these languages. Therefore, even if you do not think that Botan-
ical Latin is useful enough to merit your attention, you can use the
underlying structure of the Latin program to write programs to teach
Russian, or German.
• Verbtab. Declension is not the only feature that Latin shares with
German, Russian, and Ancient Greek. Latin has also a very complex
verbal system, that it passed down to the so called Romance Languages
(Spanish, Italian, Portuguese, Rumanian, French, Galician, and Cata-
lan).
If you want to learn one of the Romance languages, you can adapt the sup-
port provided for Latin, in order to study Spanish, or French verbs. By the
19.2 Botany 179

way, Spanish and Portuguese are much closer to Latin than other Romance
languages. The Latin verbal system was transmitted to Spanish and Por-
tuguese almost without alteration. Consider the verb amare:

Latin Spanish
Present Imperfect Present Imperfect
amo amabam amo amaba
amas amabas amas amabas
amat amabat ama amaba
amamus amabamus amamos amábamos
Now, let us concentrate our efforts on the design of the program. I suggest
that you place all tab controls in the same package; in the example, the pack-
age is called tabPackage. In order to create a tab control, choose the option
New in New Package from the IDE task menu; select the option Control from
the left hand pane; press the button Create, and mount the following layout:

The verbTab control contains a listEdit control, that will hold the verb drilling
list; two read-only edit controls, to show the tense, and mood that the student
will be practicing; and a button, that will compare the student’s answer with
a template. The main reason behind offering you this application is to show
180 Tab forms and other goodies

how to handle the listEdit control; therefore, let us take a close look at the
methods associated with this important kind of control. From the Properties-
dialog, you must press the Events-button, and add the following snippet

predicates
onShow : window::showListener.
clauses
onShow(_Source, _Data) :-
VerbList= listVerb_ctl : tryGetVpiWindow(),
verbs::allVerbs(L),
vpi::lboxAdd(VerbList, L),
vpi::lboxSetSel(VerbList, 0, 1), !,
verbs:: get_verb_state(Tense, Mood),
mood_ctl:setText(Mood),
tense_ctl:setText(Tense).
onShow(_Source, _Data).

to the showListener. The snippet shows you how to add a list of options to
the control; it shows also how to select an entry:

VerbList= listVerb_ctl : tryGetVpiWindow(), ...

vpi::lboxAdd(VerbList, L),
vpi::lboxSetSel(VerbList, 0, 1),...

From the Properties-dialog, you must also add the code from figure 19.1 to
Events-onCheckClick; then you will learn a few other listEdit methods:

Get the window associated with the control :

VerbList= listVerb_ctl : tryGetVpiWindow()

Get the selected index :

IV= vpi::lboxGetSelIndex(VerbList)

Delete entry, by giving its index :

vpi::lboxDelete (VerbList, IV)

Obtain the number of entries :

C= vpi::lboxCountAll(VerbList)
19.2 Botany 181

The declensionTab control works exactly like the verbTab-control. The

pronunciationTab control has something new. It triggers the Windows’ playSound
in order to play a wave file. Since we are dealing with an external function,
it is necessary to go through the red tape of linking a Prolog program with a
Windows’ API; you must begin by declaring that playSound is an external
routine. Then you must introduce a Prolog predicate to represent the API
entry. Below one finds the declaration that playSound is an external routine:

resolve playSound externally

onCheckClick(_Source) = button::defaultAction :-
Txt= editorControl_ctl:getEntireText(),
Verb= listVerb_ctl:getText(),
VerbList= listVerb_ctl : tryGetVpiWindow(),
IV= vpi::lboxGetSelIndex(VerbList), !,
Mood= mood_ctl:getText(),
Tense= tense_ctl:getText(),
verbs::pres_root_conj(Verb, Tense, Mood, L),
if template::check_it(Txt, L, Ans) then
editorControl_ctl:pasteStr(" "),
vpi::lboxDelete (VerbList, IV) ,
C= vpi::lboxCountAll(VerbList),
if C= 0 then
verbs::next_verb_state(),
verbs::get_verb_state(NewTense, NewMood),
mood_ctl:setText(NewMood),
tense_ctl:setText(NewTense),
verbs::allVerbs(NewList),
vpi::lboxAdd(VerbList, NewList)
else
succeed()
end if,
vpi::lboxSetSel(VerbList, 0, 1)
else
template::check_present(Txt, L, Ans),
editorControl_ctl:pasteStr(Ans)
end if.
onCheckClick(_Source) = button::defaultAction.

Figure 19.1: listEdict methods

182 Tab forms and other goodies

Now, let us declare a Prolog predicate that will be called whenever you want
to trigger playSound, and a few necessary constants.

class predicates
playSound : (string Sound,
pointer HModule,
soundFlag SoundFlag)
-> booleanInt Success
language apicall.

constants
snd_sync : soundFlag = 0x0000.
/* play synchronously (default) */
snd_async : soundFlag = 0x0001.
/* play asynchronously */
snd_filename : soundFlag = 0x00020000.
/* name is file name */
snd_purge : soundFlag = 0x0040.
/* purge non-static events for task */
snd_application : soundFlag = 0x0080.
/* look for application specific association */

Finally, introduce the snippet given below

clauses
onPlayClick(_Source) = button::defaultAction :-
File= listEdit_ctl:getText(),
_ = playSound(File,
null, /*snd_nodefault*/ snd_async+snd_filename).
into the clickResponder of the Play button.

19.3 Handling Chinese

After this incursion into Latin, let us see how to deal with Chinese. A Chinese
dialect known in the West as Mandarin is the language with the largest
number of speakers; besides this, the writing system used in Mandarin was
adopted by other Chinese languages, by the Japanese, and by the cultivated
Koreans. As a result, at least one third of the World’s population has some
knowledge of written Chinese.
Mandarin itself has three writing system: Pinyin, that uses the Latin
alphabet with Greek accents, Tradition Ideograms, and Simplified Ideograms.
19.3 Handling Chinese 183

When using a computer to write Chinese, one usually express his/her ideas
in Pinyin, and a Word Processing Editor translates Pinyin into Simplified,
or Traditional Ideograms. Of course, I do not know Chinese, but my son
knows. Therefore, at my request, he has used the multiMedia_native class
to add sound to a couple of menu entries. You will find the snippet that he
has used below; browse the multimedia class to uncover additional resources.

predicates
onFilePronunciationHowAreYou : window::menuItemListener.
clauses
onFilePronunciationHowAreYou(_Source, _MenuTag) :-
_ = multiMedia_native::sndPlaySound("howareu.wav",
multiMedia_native::snd_ASync ).

In the case of Chinese, my son used the multiMedia_native class. Therefore

it was not necessary to go through the red tape required to call an external
routine, as he has done in the case of Latin. You may ask why he did not
adopt the same solution for Latin and Chinese. The point is that he wanted
to show you how to call external routines from Prolog, and the multimedia
API provides an easy example of such an operation.

• Create a project: chinese.

• Add the editControl to its root.

• Create a package chineseforms.

• Add a chineseForm to chineseforms. The chineseForm has

editControl, and
a Pinyin2Ideogram button.

• Create an ideograms class; un-tick the Creates Objects radio button.

You will find the class file below. The implementation is listed in
figure 19.2.

% File: ideograms.cl
class ideograms
open core
predicates
classInfo : core::classInfo.
ideograms:(string, string) procedure (i, o).
end class ideograms
184 Tab forms and other goodies

implement ideograms
open core
clauses
classInfo("chineseforms/ideograms", "1.0").
class facts - pideograms
pinyin:(string, string).
class predicates
pinyin2ideogram:(string, string) procedure (i, o).
converList:(string*, string) procedure (i, o).
clauses
pinyin("ni^", "\u4f60").
pinyin("ha^o", "\u597d").
pinyin("ma", "\u55ce").
pinyin("wo^", "\u6211").
pinyin("Ni^", "\u4f60").

pinyin2ideogram(P, Idea) :- pinyin(P, Idea), !.

pinyin2ideogram(P, P).

ideograms(S, Idea) :- L= string::split(S, " \n"),

converList(L, Idea).

converList([], "") :- !.
converList([P|Ps], Acc) :- pinyin2ideogram(P, Idea),
converList(Ps, Rest),
Acc= string::concat(Idea, Rest).
end implement ideograms

Figure 19.2: File: ideograms.pro

From the Properties-dialog, add the following snippet

clauses
onPinyinClick(_Source) = button::defaultAction :-
Txt= editorControl_ctl:getEntireText(),
ideograms::ideograms(Txt, Ans),
editorControl_ctl:pasteStr(Ans).

to the ClickResponder belonging to Pinyin2Ideogram-button.

19.3 Handling Chinese 185

The last step is the addition of the following snippet

predicates
onShow : window::showListener.
clauses
onShow(_Source, _Data) :-
editorControl_ctl:setFontDialog(), !.
onShow(_Source, _Data).

to the ShowListener of the Pinyin2Ideogram-button. Of course this program

will work only if you have a Unicode font installed; whenever a new form
is created, you must choose a Unicode font. If you live in China, Japan,
or Korea, this is not a huge problem. However, if you live in the United
States, or in Europe, your Unicode Font may not have the Chinese characters.
Windows assumes that you speak the most weird languages, except Chinese.
I have told you before that Paraguay is my favorite country. Therefore I often
download Paraguayan songs; the result is that my computer often assumes
that I speak one of the two official languages of that country. Anyway, I am
pretty sure that Arial Unicode MS does have Chinese characters.
186 Tab forms and other goodies

19.4 Regular expressions

Whenever one is dealing with languages, regular expressions are useful tools.
In fact, if the programmer wants to search a string for a substring, s/he can
use traditional search methods, like the Boyer-Moore algorithm. However,
s/he may want to find a pattern that describes an integer, for instance. In this
case, the regular expression would be "[0-9]+", which means: A sequence
of one or more digits. The plus sign is called positive Kleen closure, and
indicates one or more repetitions. There is also a reflexive Kleen closure,
that accepts the empty string: "[0-9]*". These are the two instances of
regular expressions that I use; for anything more complex I prefer the full
power of Prolog.
/******************************************
Project Name: regexpr.prj
UI Strategy: console
******************************************/
implement main
open core
clauses
classInfo("main", "regexpr").

class predicates
test:(string) procedure (i).

clauses
test(Str) :-
regexp::search("-[0-9]+|[0-9]", Str, Pos, Len), !,
stdio::write(Pos), stdio::nl,
stdio::write(Len), stdio::nl.
test(Str) :- stdio::write("No integer found in ", Str),
stdio::nl.

clauses
run():-
console::init(),
test( "There were 99 small Romans"),
succeed(). % place your own code here
end implement main

goal
mainExe::run(main::run).
19.4 Regular expressions 187

For a deeper understanding of regular expressions, it is a good idea to learn

what is a Chomsky-Schutzenberger hierarchy. Noam Chomsky (an Amer-
ican Linguist) and Marcel Paul Schutzenberger (a French Medical Doctor)
proposed that languages should be described by grammar rules very simi-
lar to logical expressions in the clausal form; if you do not remember what
a clausal form is, go back to page 77, and you will find something about
the subject. In Chomsky-Schutzenberger formalism, a simple English noun
phrase is described by the following set of rules:

article → [the]
noun → [iguana]
adjective → [pretty]
npr → article, adjective, noun

Any symbol that appears on the right hand side of an arrow is a non-terminal
symbol. Therefore, article, npr, noun, and adjective are non-terminal
symbols, i.e., symbols that the grammar defines. Symbols that appears only
on the right hand side of the rules are terminal symbols; e.g. [iguana]. The
rule for npr can be understood thus:

npr → A noun-phrase is defined as

article an article followed by
adjective an adjective followed by
noun a noun

A language is regular if there is at most one non-terminal symbol at the

right hand side of any one of the rules that describe it; besides this, the non-
terminal symbol must be the rightmost one. Regular languages are good to
describe tokens, like numerals, lexical entries, etc.
Regular expressions were introduced in the 1950s by Stephen Kleene as a
method of describing regular languages. In modern notation, a member of a
set of terminal symbols is represented by brackets around a description of the
set; for instance, the regular expression "[0-9]" is a pattern that matches
any digit; the programmer may uses the + sign (positive Kleene closure),
in order to indicate a sequence of one or more members of the set; e. g.
"[0-9]" matches any numeral representing a positive integer. A sequence
of zero or more members of the set is indicated by an asterisk (the reflexive
188 Tab forms and other goodies

Kleene closure). Alternation is indicated by a vertical | bar; for instance,

"-[0-9]+|[0-9]+" is a pattern that matches a negative numeral, or an un-
signed integer. Finally, one can use the hat character to represent characters
that do not belong to a set; then "[^a-z]" is the set of all characters that
are not lowercase letters.

"[0-9]" a member the set of all digits

"[0-9]+" a sequence of one or more digits
"[0-9]*" a sequence of zero or more digits
"-[0-9]+|[0-9]+" a sequence of digits prefixed with
minus, or a sequence of digits
without prefix
"[^A-Z]" a character that is not an upper-
case letter
In replacements, one can add back references by enclosing the pattern in
parentheses, and using "\\1" to insert the matched string into the replace-
ment. An example is given below.

implement main
open core
class predicates
test:(string) procedure (i).
clauses
classInfo("main", "regexpr").

test(Str) :-
Ans= regexp::replace(Str, "<H1\\b[^>]*>(.*?)</H1>", "--\\1--"),
stdio::write(Ans), !.

run():- console::init(),
test( "<H1> Viatores in Roma </H1> etc.").
end implement main
goal
mainExe::run(main::run).

The example replaces an HTML tag with its contents. Anything between
the tags is captured into the first back reference. The question mark in the
regular expression makes sure that the star stops before the first closing tag
rather than before the last one.
19.5 A MIDI player 189

19.5 A MIDI player

Computers made it possible for a musician to create interesting songs, explore
new artistic venues, and perform his/her work single-handedly. In this sec-
tion, you will learn a few facts about MIDI, a protocol that brings computer
music to the concert hall. Let us start by writing a program that executes
MIDI files.
Project Name: midiplay
UI Strategy: Object-oriented GUI(pfc/gui)

• Add a New form in a new package. Let the form name be midiPlayer.
Add a listEdit box to it, and two buttons: play_ctl and stop_ctl.

• From the Properties-dialog of the midiPlayer form, press the Events-

button, and add the following snippet
predicates
onShow : window::showListener.
clauses
onShow(_Source, _Data) :-
SongList= listSongs_ctl : tryGetVpiWindow(),
L= ["Wagner", "Beethoven"],
vpi::lboxAdd(SongList, L),
vpi::lboxSetSel(SongList, 0, 1), !.
onShow(_Source, _Data).

to the showListener.
190 Tab forms and other goodies

• Enable the File/New option of the application task menu. Then, add
the snippet

clauses
onFileNew(S, _MenuTag) :-
W= midiPlayer::new(S), W:show().

to it. Build the application.

• The next step is to add the snippet below to the ClickResponder of

the play_ctl-button.

clauses
onPlayClick(_Source) = button::defaultAction :-
Null = null, Len = 300,
_= multiMedia_native::mciSendString(
string::createCopy("Close mid"),
string::create(300, " "), Len, Null),
Buffer =string::create(300, " "),
File= listSongs_ctl:getText(),
C= "open %s.mid type sequencer alias mid",
Command= string::format(C, File),
_= multiMedia_native::mciSendString(
Command, Buffer, Len, Null),
_= multiMedia_native::mciSendString(
"Play mid from 0",
Buffer, Len, Null),
_= vpi::processEvents().

• The final step is to add code to the button that will interrupt the music,
if you decide to do something else.

clauses
onStopClick(_Source) = button::defaultAction :-
Null= null,
_Z= multiMedia_native::mciSendString(
string::createCopy("Close mid"),
string::create(300, " "), 300, Null).
19.6 Midi format 191

19.6 Midi format

My father, and mother didn’t like music very much. Notwithstanding, I was
lucky enough to know many great musicians. For instance, my son, who is a
pianist, and singer, is the best friend of José Bernal, who happens to be the
son of Sergio Bernal, the conductor of the Utah State University Symphony
Orchestra. What is more, I was raised next door to the house of a great
composer, Maestro Joubert de Carvalho; the beautiful sonata that I offer
you in the skinner project was composed by my neighbor; it may interest
you to know that there exists a city named after that haunting piece of music.
I also knew Maestro Sukorski who created a very interesting work of art;
he wrote a program in Prolog (yes, it was in Prolog!) that performes an anal-
ysis of the mood of a person, by measuring skin resistance, electromyograph
signals, heartbeat increases, etc. Basically, Sukorski program was something
like a lie detector. Using these data, the program was able to compose a
music suited to the occasion.
I studied Space Science, and Classical Literature in Cornell; my office
there had been occupied by Moog, before my arrival; in fact, Moog scribbled
his name on the very same table that I used while in Cornell; it seems that
I have inherited Moog’s table. I also know well Maestro Dovicchi, whose
adviser was no other than the famous Hungarian composer Gyorgy Ligetti.
Maestro Dovicchi researches methods of producing high quality computer
music in the midi format; using his computers, he was able to play Mascagni’s
Cavalleria Rusticana single handedly; I happened to be present during this
memorable performance.
Besides Maestro Dovicchi, another great expert in computer music is the
composer Luciano Lima. His programs analyze the work of great composers
like Tchaikovsky, or Beethoven, and write music according to the same mu-
sical style. I included one of Lima’s compositions in the examples. I used
the Internet to contact Maestro Lima, who is an accomplished Prolog pro-
grammer, and asked him to show you how to create a simple MIDI file.
Project Name: midiplay
UI Strategy: Object Oriented GUI (pfc/gui)

• Build the application in order to construct the Project Tree.

• From the Project Tree, open the folders
$(Prodir)/pfc/gui/Controls/
Right click on the Controls folder, and add the editorControl pack-
age to the Project Tree. Rebuild the application.
192 Tab forms and other goodies

• Add a New in New Package form with the name midiform.

• Rename the Save-button to save-ctl. Rebuild the application. Enable

the File/New option of the application task menu, and add the snippet
below to it.

clauses
onFileNew(S, _MenuTag) :-
W= midiform::new(S), W:show().

• Create a class with the name midi. You will find the class declaration
below, and the implementation on figure 19.3.

class midi
open core
predicates
classInfo : core::classInfo.
generate:(string, string*) procedure (i, i).
end class midi

• Add the snippet below to the ClickResponder belonging to save_ctl.

clauses
onSaveClick(_Source) = button::defaultAction :-
Txt= editorControl_ctl:getEntireText(),
S= list::removeAll(string::split(Txt, "\n "), ""),
FName= vpiCommonDialogs::getFileName("*.mid",
["Midi", "*.mid"],
"Save in MIDI format", [], ".", _X), !,
midi::generate(FName, S).
onSaveClick(_Source) = button::defaultAction.
19.6 Midi format 193

In the examples, you will find a more refined and complete version of the
program listed in figure 19.3. The program of figure 19.3 represents a good
opportunity to learn how to write data to a file. The first step is to create
an output stream file:
P= outputStream_file::create(File, stream::binary())

Then you can write a list of bytes to the file by using the foreach construc-
tion, as you can see below.
foreach X= list::getMember_nd (MidiCode) do
P:write(X)
end foreach,
P:close().

From the point of view of a musician, the programs given in the examples are
grossly oversimplified. However, you can easily complete the missing steps.
For instance, you can add multiple tracks, improve the resolution, change
the instruments, etc.

By the way, the example in the figure is an oversimplified rendering of India,

the beautiful Guarania of José Asunción Flores. The lyrics says:

India, bella mezcla de diosa y pantera, India, pretty mestiza of goddess and panther,
doncella desnuda que habita el Guairá. naked girl that lives in Guaira.
Arisca romanza curvó sus caderas Dance has bent her hips into the shape
copiando un recodo de azul Paraná. of the blue curves of the Parana river.

Bravea en las sienes su orgullo de plumas, On her temples shine the pride of feathers,
su lengua es salvaje panal de eirusu. her language is a wild honeycomb.
Collar de colmillos de tigres y pumas A necklace from teeth of tigers and pumas
enjoya a la musa de Ybytyruzú. is a jewel for the muse of Ybytyruz.
194 Tab forms and other goodies

implement midi
open core
class predicates
note:(string, unsigned8) procedure (i, o).
duration:(string, unsigned8) procedure (i, o).
tomidi:(string*, unsigned8*) procedure (i, o).
mSz:(integer, unsigned8, unsigned8, unsigned8, unsigned8)
procedure (i, o, o, o, o).
class facts
nt:(string, unsigned8).
clauses
classInfo("midi/midi", "1.0").

duration(X, Y) :- Y= toTerm(X), Y < 128, !.

duration(_X, 24).

nt("C", 60). nt("D", 62). nt("E", 64). nt("F", 65). nt("G", 67).
nt("A", 69). nt("B", 71). nt("C5", 72). nt("D5", 74).
nt("C#", 61). nt("D#", 63). nt("F#", 66).
nt("G#", 68). nt("A#", 70). nt("Bb", 70).

note(X, Y) :- nt(X, Y), !.

note(_, 69).

tomidi([N, D|Rs], [0,192, 24, 0,144,N1, 127,F1,128,N1,0|R] ) :-

/* guitar=24, violin= 40 voice= 52 */
note(N, N1), duration(D, F1), !, tomidi(Rs, R).
tomidi(_, []).

mSz(S, 0, 0, X3, X4) :- I4= S mod 256, X4= convert(unsigned8, I4),

S1= S div 256, I3= S1 mod 256, X3= convert(unsigned8, I3).

generate(File, Music) :- tomidi(Music, L),

Events= [0, 255, 89, 2, 0, 0, 0, 255, 81, 3, 7, 161,
32, 0, 255, 88, 4, 2, 2, 24, 8],
NC= list::length(Events), N1= list::length(L),
Cnt= NC+ N1 + 4, mSz(Cnt, X1, X2, X3, X4),
MainHeader= [ 77, 84, 104, 100, 0, 0, 0, 6, 0, 0, 0, 1, 0, 48,
77, 84, 114, 107, X1, X2, X3, X4],
H= list::append(MainHeader, Events, L, [0,255,47,0]),
P= outputStream_file::create(File, stream::binary()),
foreach X= list::getMember_nd (H) do
P:write(X)
end foreach,
P:close().
end implement midi

Figure 19.3: Midi generator

Chapter 20

Bugs

Old computers were based on re-

lays, that are bulky electrical de-
vices, typically incorporating an
electromagnet, which is activated
by a current in one circuit to turn
on or off another circuit. Comput-
ers made of such a thing were enor-
mous, slow, and unreliable. There-
fore, on September 9th, 1945, a
moth flew into one of the relays of
Harvard Mark II computer and jammed it. From that time on, bug became
the standard word to indicate an error that prevents a computer from work-
ing as intended.

Due to bugs, the Visual Prolog compiler frequently returns error mes-
sages, instead of generating code and running the corresponding programs.
In fact, the Visual Prolog compiler gives more bug alerts than any other
language, except perhaps Clean. Students don’t appreciate this constant
nagging about errors, and seldom try to analyze the error messages, but be-
lieve me, you should be thankful to the designers of such a compiler that
gives you a chance to clean up all the stains that can mar your program. As
I have said, Visual Prolog is good at pinpointing bugs. Notwithstanding, if
you do not learn how to deal with its error messages, you will find debugging
a random walk that almost never leads to success. Therefore, in this chapter,
you will learn how to interpret messages from the compiler.

195
196 Bugs

20.1 Type error

The program below has no bugs; however, if you need to use the unsigned
integer which results from the factorial function in order to, say, calculate a
binomial number, the compiler may refuse to comply.

implement main
open core

clauses
classInfo("main", "bugs").

class predicates
fact:(unsigned) -> unsigned.

clauses
fact(N) = F :- if N < 1 then F=1
else F=N*fact(N-1) end if.

run():- console::init(),
stdio::write("N: "),
N= stdio::read(),
stdio::write(fact(N)), stdio::nl.
end implement main

goal mainExe::run(main::run).

For instance, if you try to compile the program of figure 20.1, you will get
the following error message:

error c504: The expression has type ’::integer’,

which is incompatible with the type ’::unsigned’

The error will disappear if you explicitly convert each input argument of
binum/3 to an unsigned integer; you must also convert the output of factorial
to an integer.

binum(N, P, R) :-
N1= convert(unsigned, N),
P1= convert(unsigned, P),
R = convert(integer, fact(N1) div (fact(P1)*fact(N1-P1))).
20.2 Non-procedure inside a procedure 197

implement main
open core
clauses
classInfo("main", "bugs").

class predicates
fact:(unsigned) -> unsigned.
binum:(integer, integer, integer) procedure (i, i, o).

clauses
fact(N) = F :- if N < 1 then F=1 else F=N*fact(N-1) end if.

binum(N, P, R) :- R = fact(N) div (fact(P)*fact(N-P)).

run():- console::init(),
stdio::write("N: "), N= stdio::read(),
stdio::write("P: "), P= stdio::read(),
binum(N, P, R), stdio::write(R), stdio::nl.
end implement main

goal mainExe::run(main::run).

Figure 20.1: A buggy program

20.2 Non-procedure inside a procedure

There is another bug that poses many difficulties to beginners. Visual Prolog
declares certain predicates as procedures. This is the case of the run()
predicate that appears in all console programs. This means that you cannot
call a nondeterm predicate from run(). Therefore, if you try to compile the
program of listing 20.2, you will get the following error message:

error c631: The predicate ’nonprocedure::run/0’,

which is declared as ’procedure’, is actually ’nondeterm’

It is true that the weight/2 fact predicate is indexed by its integer argument.
This means that, given its integer argument, weight/2 should act like a
procedure. However Prolog is impervious to wishful thinking; then you need
to define a getweight/2 procedure, that will retrieve the weight values.
Below, you can see how getweight can be defined. Use it as a model
for getting around a situation, where you need to call a nondeterm predicate
from a procedure.
198 Bugs

implement main
class facts
weight:(integer, real).
class predicates
getweight:(integer, real) procedure (i, o).
clauses
classInfo("main", "nonprocedure").

weight(0, -1.0).
weight(1, 2.0).
weight(2, 3.5).

getweight(I, R) :- weight(I, R), ! or R=0.

run():- console::init(), getweight(1, X),

stdio::write(X), stdio::nl.
end implement main
goal mainExe::run(main::run).

20.3 Non-determ condition

Prolog has constructions that do not accept nondeterm predicates. For in-
stance, the condition of an if-then-else construction must be a determ
predicate. Therefore, the program in figure 20.3 will issue an error message:
error c634: The condition part of if-then-else may not
have a backtrack point (almost determ)
The error will disappear if you define member as a determ predicate:
class predicates
member:(string, string*) determ.
clauses
member(X, [X|_]) :- !.
member(X, [_|Xs]) :- member(X, Xs).

20.4 Impossible to determine the type

There are occasions when one needs to specify the type before the instanti-
ation of a variable; type specification can be carried out by the procedure
hasdomain(type, Var)
20.4 Impossible to determine the type 199

implement main
class facts
weight:(integer, real).

clauses
classInfo("main", "nonprocedure").

weight(0, -1.0).
weight(1, 2.0).
weight(2, 3.5).

run():- console::init(),
weight(1, X), stdio::write(X), stdio::nl.
end implement main

goal mainExe::run(main::run).

Figure 20.2: A bug caused by a non deterministic predicate

implement main
class predicates
vowel:(string, string) procedure (i, o).
member:(string, string*) nondeterm.

clauses
classInfo("main", "ifexample-1.0").

member(X, [X|_]).
member(X, [_|Xs]) :- member(X, Xs).

vowel(X, Ans) :- if member(X, ["a", "e", "i", "o", "u"])

then Ans= "yes" else Ans="no" end if.
run():- console::init(), vowel("e", Ans), stdio::write(Ans).
end implement main

goal mainExe::run(main::run).

Figure 20.3: Deterministic predicate required.

200 Bugs

In the useless program listed below, if one removes hasdomain(integer, X),

s/he will get an error message stating that it is impossible to determine the
type of the term X.
implement main
clauses
classInfo("main", "vartest-1.0").
run() :- console::init(),
hasdomain(integer, X), X= stdio::read(),
stdio::write(X, X, X), stdio::nl.
end implement main
goal mainExe::run(main::run).

20.5 Flow pattern

When you do not declare a flow pattern, Prolog assumes that all arguments
are input variables. Since this simplified assumption is more often than not
wrong, you will get an error message, as in the example below.
implement main
class predicates
factorial:(integer, integer).
clauses
classInfo("main", "dataflow-1.0").
factorial(0, 1) :- !.
factorial(N, N*F1) :- factorial(N-1, F1).
run():- console::init(), N= toTerm(stdio::readLine()),
factorial(N, F), stdio::write(F).
end implement main
goal mainExe::run(main::run).

You can fix the bug by declaring factorial as an input/output procedure.

class predicates
factorial:(integer, integer) procedure (i, o).
Chapter 21

A Data Base Manager

This chapter shows how to create a form with edit fields, and use its contents
to feed a data base. However, the most important thing that you will learn is
how to access the fields from the form. Edit field windows are stored as fact
variables, whose names you must discover, and use to handle the window, as
in the example below.
onSave(_Source) = button::defaultAction() :- Key= keyvalue_ctl:getText(),
Email= email_ctl:getText(), T= listEdit_ctl:getText(),
emails::insert_info(Key, Email, T).

21.1 Database manager

Databases are very useful in any commercial application. If you do not know
much about databases, I strongly recommend that you browse the Internet
for a book on the subject. The main concept of a B+Tree database are:

Indexes that are stored in a special data structure called B+Tree. All you
need to know about B+Trees is that they store things in order (for
instance, in alphabetical order). When things are in order, like words
in a dictionary, it is easy to search for them.
Domains give the register structure. In our example, the contact domain
is defined as

domains
contact= email(string, string, string).

References are pointers to the places where registers are stored.

After this brief introduction, let us describe the program.

201
202 A Data Base Manager

• Create a project whose name is emailBook with Object-oriented GUI.

• Using the option File/Add from the task menu, add the package
chainDB. This package is inside the folder pfc, in the installation di-
rectory. Build the application.

• Create a class emails at the root of the Project Tree. Untick the Creates
Objects option. In section 21.2, you will find the listing of emails.cl and
emails.pro, that you should insert in the appropriate files. Build the
application.

• Add a new item to the application task menu. In order to accomplish

this task, go to the Project Tree window, left click on TaskMenu.mnu.
On the TaskMenu dialog, right click the mouse, and choose New from
the pop up menu. On the appropriate field, write Create, that is the
name of the menu item. Build the application.

• Add the following code

class predicates
createNewDB:(string FNAME).
predicates
oncreate : window::menuItemListener.
clauses
oncreate(_Source, _MenuTag) :-
FName= vpiCommonDialogs::getFileName("*.*",
["Email", "*.dbs"],
"Create", [], ".", _X), !,
createNewDB(FName).
oncreate(_, _).

createNewDB(Fname) :- file::existfile(Fname), !.
createNewDB(Fname) :- emails::data_base_create(Fname).

to Project Window\TaskMenu.win\Code Expert\Menu\id_create.

You must develop your programs step by step, checking whether each new
step is correct before going forward. In the present step, you have added
a new Create menu item to TaskMenu.mnu. You have also inserted code
into onCreate, in order to read a file name, and create a corresponding
database. Build the application, run it, and click on the Create option of
the application menu. When the file selection dialog asks you for a file name,
21.1 Database manager 203

type mails.dbs. If everything follows the script, the program will create
a mails.dbs file. Whenever you want to use a data base, you must open
it. After using it, you must close it, before turning off the computer. To
make sure that we will not forget to close the data base before exiting the
application, let us add the snippet

clauses
onDestroy(_) :- emails::data_base_close().

to Project Tree/TaskWindow.win/Code Expert/Window/onDestroy.

Now that you know for sure that nobody will forget to close the data
base, you can proceed to open it. From the Project Window, click on
TaskMenu.mnu, and enable the &Open\tF8 menu item. Add the snippet

class predicates
openDB:(string FNAME).
predicates
onFileOpen : window::menuItemListener.
clauses
onFileOpen(_Source, _MenuTag) :-
FName= vpiCommonDialogs::getFileName("*.*",
["Email", "*.dbs"],
"Open", [], ".", _X), !,
openDB(FName).
onFileOpen(_, _).

openDB(Fname) :- file::existfile(Fname), !,
emails::data_base_open(Fname).
openDB(_).

to ProjWin/TaskWindow.win/Code Expert/Menu/TaskMenu/id_file/id_file_open.
Build the application. Then, from the VDE task menu, choose the option
File/New in New Package, and insert the form below onto the root of the
project.
204 A Data Base Manager

The form has a list box. Therefore, it is necessary to add a list of options
into it. In order to achieve this goal, add the snippet
clauses
onListEditShow(_Source, _CreationData) :-
WlboxHandle=listEdit_ctl:tryGetVpiWindow(), !,
vpi::lboxAdd(WlboxHandle, ["Person", "Commerce"]).
onListEditShow(_, _).
to the ShowListener of listEdit:listEdit_ctl.

The last step of this project is to add code to the Save, and List buttons.
Therefore, add
onSaveClick(_Source) = button::defaultAction() :-
Key= key_ctl:getText(),
Email= address_ctl:getText(),
T= listEdit_ctl:getText(),
emails::insert_info(Key, Email, T).
to ClickResponder of button:save_ctl. Finally, add
clauses
onListClick(_Source) = button::defaultAction :-
emails::db_list().
to the ClickResponder of button:list_ctl. Enable the File/New option of
the application task menu; add the snippet
onFileNew(S, _MenuTag) :-
X= insertForm::new(S), X:show().
to TaskWindow/Code Expert/Menu/TaskMenu/id_file/id_file_new. Build
the application once more, and that is it.
21.2 btr class 205

21.2 btr class

%File: emails.cl
class emails
open core
predicates
classInfo : core::classInfo.
data_base_create:(string FileName) procedure (i).
data_base_open:(string FileName).
data_base_close:().
insert_info:(string Name, string Contact, string Tipo).
data_base_search:(string Name, string Contact, string Tipo)
procedure (i, o, o).
db_list:().
del:(string).
end class emails

%File: emails.pro
implement emails
open core

domains
contact= email(string, string, string).

class facts - dbState

dbase:(chainDB, chainDB::bt_selector) determ.

clauses
classInfo("db/emails/emails", "1.0").

data_base_create(FileName) :-
ChainDB= chainDB::db_create(FileName, chainDB::in_file),
ChainDB:bt_create("email", ContactName, 40, 21, 1),
ChainDB:bt_close(ContactName),
ChainDB:db_close(), !.
data_base_create(_).

data_base_open(FileName) :-
ChainDB= chainDB::db_open(FileName, chainDB::in_file),
ChainDB:bt_open("email", ContactName),
assert(dbase(ChainDB, ContactName)).

data_base_close() :-
retract(dbase(ChainDB, BTREE)), !,
ChainDB:bt_close(BTREE),
ChainDB:db_close().
data_base_close().
206 A Data Base Manager

insert_info(Key, Email, Tipo) :-

dbase(ChainDB, BTREE), !,
ChainDB:chain_insertz( "Contacts",
email(Key, Email, Tipo),
RefNumber),
ChainDB:key_insert(BTREE, Key, RefNumber).
insert_info( _, _, _).

data_base_search(Name, Contact, Tipo) :- dbase(ChainDB, BTREE),

ChainDB:key_search(BTREE, Name, Ref),
ChainDB:ref_term(Ref, Term),
Term= email(Name, Contact, Tipo), !.
data_base_search(_Name, "none", "none").

class predicates
getInfo:(chainDB, chainDB::ref, string, string)
procedure (i, i, o, o).
dbLoop:().
clauses
getInfo(ChainDB, Ref, Name, Email) :-
ChainDB:ref_term(Ref, Term),
Term= email(Name, Email, _), !.
getInfo(_, _, "none", "none").

db_list() :-
dbase(ChainDB, BTREE),
ChainDB:key_first(BTREE, Ref), !,
getInfo(ChainDB, Ref, Name, Email),
stdio::write(Name, ", ", Email), stdio::nl,
dbLoop().
db_list() :- stdio::write("None"), stdio::nl.

dbLoop() :- dbase(ChainDB, BTREE),

ChainDB:key_next(BTREE, Ref), !,
getInfo(ChainDB, Ref, Name, Email),
stdio::write(Name, ", ", Email), stdio::nl,
dbLoop().
dbLoop().

del(Key) :-
dbase(ChainDB, BTREE),
ChainDB:key_search(BTREE, Key, Ref), !,
ChainDB:key_delete(BTREE, Key, Ref),
ChainDB:term_delete("Contacts", Ref).
del(_Key).

end implement emails

21.3 The database manager 207

21.3 The database manager

The function that creates the database is very simple:
data_base_create(FileName) :-
ChainDB= chainDB::db_create(FileName, chainDB::in_file),
ChainDB:bt_create("email", ContactName, 40, 21, 1),
ChainDB:bt_close(ContactName),
ChainDB:db_close(), !.
data_base_create(_).
The method db_create/2 does what it means, i.e., it creates a database
inside FileName, and returns an object to manipulate it. The object is
stored in ChainDB. From ChainDB, one calls bt_create in order to obtain
a BTree; teh BTree package instantiates the variable ContactName with a
pointer to the BTree. Finally, one closes both the BTree, and the database.
One opens the database with the help of a predicate that is nearly as
simple as the one used to create it.
data_base_open(FileName) :-
ChainDB= chainDB::db_open(FileName, chainDB::in_file),
ChainDB:bt_open("email", ContactName),
assert(dbase(ChainDB, ContactName)).

data_base_close() :- retract(dbase(ChainDB, BTREE)), !,

ChainDB:bt_close(BTREE), ChainDB:db_close().
data_base_close().
The predicate db_open opens the database stored inside FileName, and re-
turns an object that can be used to handle it; the object is stored in ChanDB,
which is used to open the email-Btree. Both ChainDB and the BTree pointer
are stored in a fact predicate, for future references.
assert(dbase(ChainDB, ContactName))
When the user kills the application, the data_base_close predicate retracts
the database manager, and the Btree pointer from the fact database, and
closes both the BTree, and ChainDB. Up to this point, we have been describing
predicates that administrate the database; now we will describe the method
that one can use to insert data into the database.
insert_info(Key, Email, Tipo) :- dbase(ChainDB, BTREE), !,
ChainDB:chain_insertz( "Contacts", email(Key, Email, Tipo),
RefNumber),
ChainDB:key_insert(BTREE, Key, RefNumber).
insert_info( _, _, _).
208 A Data Base Manager

The database is a filing system. As any filing system, it has two com-
ponents: a chain of portfolios, each one containing data that one wants to
store; and an alphabetical index, that points to the different portfolios, and
allows a potential user to discover where the desired information is stored.
In the example database, the chain of portfolios is named "Contacts"; each
portfolio has the following form: email(string, string, string). The
first argument of email holds the name of a contact, the second argument
contains an email, and the third argument informs the type of contact one
has with the entity associated with the database entry.
The chain_insertz predicate stores the portfolio onto the "Contacts"
chain, and returns a RefNumber to its location. The RefNumber will be stored
in alphabetical order into the BTree.
To search a Contact-email, corresponding to a given Name, one searches
the BTree, in order to find the reference; once in possession of the reference
number, it is straightforward to obtain the portfolio.
data_base_search(Name, Contact, Tipo) :- dbase(ChainDB, BTREE),
ChainDB:key_search(BTREE, Name, Ref),
ChainDB:ref_term(Ref, Term),
Term= email(Name, Contact, Tipo), !.
data_base_search(_Name, "none", "none").

The predicate used to delete an entry is very easy to understand, and I

will not linger on it. The predicate that lists the contents of the database
in alphabetical order is much more complex, but not difficult to understand,
once you know the basic principle of database management.
Chapter 22

Books and articles

In this chapter, I will talk about a few books on Prolog, and programming
techniques. You will notice that the books are somewhat old, and out of
print. One reason for this is that people do not buy technical books as often
as before, since most students hope to get material like this tutorial from
the Internet. Another reason is that great computer scientists were very
enthusiastic about Prolog in the eighties and nineties. These scientists wrote
books that are hard to beat. For instance, I cannot find any modern book
that comes close to [Coelho/Cotta].

22.1 Grammars
Prolog excels in language processing, compiler writing, and document prepa-
ration. In fact, the language was invented with language processing in mind.
Most books deal with standard Prolog, that is slightly different from Visual
Prolog. While standard Prolog was designed for fast prototype of small ap-
plications and for testing ideas, Visual Prolog is used in large commercial
and industrial critical systems. Wellesley Barros, for instance, has designed
and programmed a large system for Hospital administration in Visual Prolog.
You bet that one cannot afford a failure in such a system due to type errors.
In critical situations, type checking and type declarations are very impor-
tant. In any case, although programming in Visual Prolog, you can profit
from ideas presented in books based on standard Prolog. An old book, that
you may find in your local library, was written by the Argentinean scientist
Veronica Dahl(see [Abramson/Dahl]), one of the greatest modern linguists.
Dr. Cedric de Carvalho wrote a Prolog compiler that generates Java byte-
code. This compiler allows you to write applets in Prolog. His compiler is
not a fully fledged tool, but you can easily improve it. In the mean time, you

209
210 Books and articles

will learn a lot about compilation, parsing, Java, etc. You can start reading
Cedric’s paper in [Cedric et all]. You can also port the compiler, written in
VIP 5.2, to VIP 6.2. The address is

http://netprolog.pdc.dk

Another book on Natural Language Processing with many interesting ap-

proaches to grammar formalism is [Pereira/Shieber]. This book is more di-
dactic than [Abramson/Dahl]. In fact, [Abramson/Dahl] wrote it for special-
ists, and [Pereira/Shieber] talk to students that are learning the difficult art
of Language Processing. A book that is not specialized in Natural Language
Processing, but gives a good introduction to it is [Sterling and Shapiro]. If
you want to read an advanced book on the subject, a good option is the book
by [Gazdar/Mellish].

22.2 Databases
David Warren, a computer scientist, is the main proponent of using Prolog
as a database engine. The book [Warren] is out of print, but you should get
a used copy, if you need to learn about databases. It is a classic.
David Warren leads a team that develops XSB, a database oriented Pro-
log, at Stony Brook University. Although I do not think that his version
of Prolog is usable for serious applications, since it relies on other tools for
building GUI, it is not compiled (it is not easy to write a beautiful applica-
tion in XSB), and it is not typed, I still strongly recommend a visit to the
XSB page. There you will find many good ideas on databases and Prolog,
and tools to prototype them. The address is

http://www.cs.sunysb.edu/~sbprolog/xsb-page.html

By the way, XSB provides a set of tools to export its functionality to other
languages. Therefore, if you like its features, you can call the XSB DLL from
Visual Prolog! The XSB page shows how to call the DLL from Visual Basic.
You can use the same approach to call it from Visual Prolog.

22.3 Programming Techniques

If you are having a hard time with concepts such as Recursive Programming,
List Operations, Cuts, Non Deterministic Programming, and Parsing, you
should read the book The Art of Prolog (see [Sterling and Shapiro]). The
ideas presented in this book can be easily ported to Visual Prolog. In Visual
22.3 Programming Techniques 211

Prolog, you can even improve the original programs. For instance, one of my
students was able to create diagrams for the circuits described in Chapter 2
of [Sterling and Shapiro].
My favorite book on Prolog by far is Prolog by Example, by [Coelho/Cotta].
It is a kind of FAQ: The authors state a problem, and present the solution. It
is also an anthology. The problems and solutions were proposed and solved
by great computer scientists. This book deserves a new edition. Try, and try
hard, to get a used copy. If you are not able to get a copy, ask permission
from the authors, and make a xerox copy.
212 Books and articles
 Part II

Artificial Intelligence

213
Chapter 23

Search

Patrick Henry Winston, author of a very famous book of LISP, starts his
exploration of Artificial Intelligence with search, claiming that search is ubiq-
uitous. Nothing could be closer to the truth. In fact, everything, from robot
motion planning to neural network, can be reduced to search. Therefore, like
Winston, let us start our study of Artificial Intelligence from search.

23.1 States
In artificial intelligence, the expression problem solving refers to the analysis
of how computers can use search to solve problems in well-circumscribed
domains. One question arises from this definition: What does an intelligent
program inspect in its search for the solution to a problem? It inspects a
maze of states connected by operations that cause feature transitions.
An object has states, i.e., it has a set of features that determines its
situation at a given instant. For example, Euclid defined point as an entity
whose sole feature is its position. Therefore, the state of a punctual object
is given by its coordinates. However, in general, one can use other features
besides position, in order to define the state of an object:

Color: Red, green, blue, etc.

Integrity informs whether the object is whole, or unmounted.

Attitude gives the direction the object is facing.

Safety says whether the object is protected against destruction.

Each feature has one or more attributes, and values for the attributes. For
instance, the position has three attributes, wich are the coordinates of a

215
216 Search

point; in a 2D space, the position attributes can have values in the Cartesian
product R × R; therefore, the tuple (50, 200) gives valid values for the
position. To make a long story short, one can say that state is a snapshot in
time. Each attribute that one can change to solve a problem is called degree
of freedom. Then a point in a 3D space has three degrees of freedom.
Operator is a procedure that can be used to make the transition from one
state to another. The artificially intelligent program starts at an initial state
and uses the allowable operators to move towards a goal state. A set of all
states that the objects under consideration can take, and all transitions that
lead from one state to another is called state space.

23.2 Search tree

Search tree is a popular diagram, where each state is represented by a labeled
node (identifier inside a circle), and each possible transition from one state
to another is described by a branch of an inverted tree. In figure 23.1, you
can see a picture of a search tree.

Figure 23.1: A search tree

In figure 23.1, the nodes b, c, d. e, f, g, h, i, j, k, and m are the states that

the object may go through. The branches that connect two nodes are the
operators that cause the transition from one state to another. For instance,
the branch that is labeled 3 in figure 23.1 is the operator that changes state b
into e. When a node X can be obtained from another node Y using only one
23.2 Search tree 217

Figure 23.2: Puzzle

operator, one says that X is daughter of Y, or that Y is mother of X. Therefore,

the node c is daughter of b, since one can obtain c using the operator 1. When
a node has no daughter, it is called a leaf. A node that has no mother is the
root of the search tree.
A more concrete example will help you to under-
stand the concepts that we have studied above. Two
black pieces, like the ones used in the game of check-
ers, and two white pieces are arranged as shown in
the figure. The black pieces are separated from the
white ones by a space. One wants to place the black
pieces between the white pieces. Two operations are
available: (1) Slide a piece to an empty space, and
(2) Jump over a piece, landing on an empty space.
The figure on the right hand side shows a step by
step solution of this problem.
In the problem that we want to solve, the objects
of interest are restricted to the four pieces, and the
empty space. The relative positions of the pieces,
and of the empty space, define a state. The initial
state E1 is shown at the top of the figure. The goal E5
is at the bottom of the figure.
In figure 23.2, you can see a search tree for the puzzle that was described
above. One could say that the states are like cities on a map, and the
operators are similar to roads connecting the cities. The problem is solved
when one finds a path that leads from the root to the goal, and the most
obvious way to find the path is to go from city to city until finding one that
matches the goal. This process is called search.
218 Search

23.3 Breadth First Strategy

The search is said to be exhaustive if it is guaranteed to generate all reachable
states before it terminates in failure. One way to perform an exhaustive
search is by generating all the nodes at a particular level before proceeding to
the next level of the tree; this method is called the Breadth First Strategy, and
ensures that the space of possible operations will be examined systematically.
In the tree of figure 23.1, one starts by examining section (c-d-e), con-
taining the nodes that are one branch away from the root. Then one goes to
section (f-g-h-i-i), wich are two nodes away from the root. Finally, the
search engine visits section (k-m). The program that performs the search op-
eration has a parameter called Queue, containing a list of candidate paths.
Each path is represented by a list of nodes, where a node has the form
r(integer, string).

implement main
domains
node= r(integer, string).
path= node*.
queue= path*.

class facts
operator:(integer, string, string).

class predicates
bSearch:(queue, path) determ (i, o).
nextLevel:(path, path) nondeterm (i, o).
solved:(path) determ.
prtSolution:(path).

clauses
classInfo("main", "breath").

solved(L) :- L= [r(_, "k")|_].

prtSolution(L) :-
foreach P= list::getMember_nd(list::reverse(L)) do
stdio::write(P), stdio::nl
end foreach.
23.3 Breadth First Strategy 219

operator(1, "b", "c").

operator(2, "b", "d").
operator(3, "b", "e").
operator(4, "c", "f").
operator(5, "c", "g").
operator(6, "c", "h").
operator(7, "e", "i").
operator(8, "e", "j").
operator(9, "g", "k").
operator(10, "g", "m").

bSearch([T|Queue],Solution) :-
if solved(T) then Solution= T
else Extentions= [Daughter || nextLevel(T, Daughter)],
ExtendedQueue= list::append(Queue, Extentions),
bSearch(ExtendedQueue, Solution) end if.

nextLevel([r(Branch, N)|Path], [r(Op, Daughter),r(Branch, N)|Path]):-

operator(Op, N, Daughter),
not(list::isMember(r(Op, Daughter),Path)).

run():- console::init(),
if bSearch([[r(0, "b")]], L) then prtSolution(L)
else stdio::write("No solution!"), stdio::nl end if.
end implement main /* breath*/

goal mainExe::run(main::run).

Initially, the Queue has a single path, that starts and ends at the root:

bSearch([[r(0, "b")]], L)

Since this path does not lead to the goal, it is extended to every daughter of
"b", and the extensions are appended to the end of the Queue:

bSearch([ [r(1, "c"),r(0, "b")],

[r(2, "d"),r(0, "b")],
[r(3, "e"),r(0, "b")]], L)

At this point, the robot has visited only one node, to wit, "b". Then the
extensions of node "c" are appended to the end of the Queue.
220 Search

bSearch([ [r(2, "d"),r(0, "b")],

[r(3, "e"),r(0, "b")],
[r(4, "f"),r(3, "c"),r(0, "b")],
[r(5, "g"),r(3, "c"),r(0, "b")],
[r(6, "h"),r(3, "c"),r(0, "b")]], L)
Node "d" is visited, but since it has no daughter, it is simply removed from
the Queue.
bSearch([ [r(3, "e"),r(0, "b")],
[r(4, "f"),r(3, "c"),r(0, "b")],
[r(5, "g"),r(3, "c"),r(0, "b")],
[r(6, "h"),r(3, "c"),r(0, "b")]], L)
The daughters of "e" are appended to the end of the Queue:
bSearch([ [r(4, "f"),r(3, "c"),r(0, "b")],
[r(5, "g"),r(3, "c"),r(0, "b")],
[r(6, "h"),r(3, "c"),r(0, "b")],
[r(7, "i"), r(3, "e"),r(0, "b")],
[r(8, "j"), r(3, "e"),r(0, "b")]], L)
Node "f" is removed from the Queue, without leaving offspring. The daugh-
ters of "g" are added to the end of the Queue.
bSearch([ [r(6, "h"),r(3, "c"),r(0, "b")],
[r(7, "i"), r(3, "e"),r(0, "b")],
[r(8, "j"), r(3, "e"),r(0, "b")],
[r(9, "k"), r(5, "g"),r(3, "c"),r(0, "b")],
[r(10, "m"), r(5, "g"),r(3, "c"),r(0, "b")]], L)
Nodes "h", "i", and "j" are visited in turns, and removed from the Queue.
The search ends when the algorithm finds the path
[r(9, "k"), r(5, "g"),r(3, "c"),r(0, "b")]
The procedure
prtSolution(L) :-
foreach P= list::getMember_nd(list::reverse(L)) do
stdio::write(P), stdio::nl
end foreach.
reverse this path, and prints each one of its members. The final result which
comes on the screen is shown below.
23.3 Breadth First Strategy 221

r(0,"b")
r(1,"c")
r(5,"g")
r(9,"k")

D:\vipro\aityros\aiprogs\breath\Exe>pause
Press any key to continue...

At this point, you are ready to try the algorithm on a real problem, to wit,
the slide puzzle.

implement slide
domains
node= r(string, stt).
path= node*.
queue= path*.
piece= e;b;w.
stt= st(piece, piece, piece, piece, piece).

class predicates
bSearch:(queue, path) determ (i, o).
nextLevel:(path, path) nondeterm (i, o).
solved:(path) determ.
prtSolution:(path).
operator:(string, stt, stt) nondeterm (o, i, o).

clauses
classInfo("breath", "1.0").

solved(L) :- L= [r(_, st(w,b,b,w,e))|_].

prtSolution(L) :-
foreach P= list::getMember_nd(list::reverse(L)) do
stdio::write(P), stdio::nl
end foreach.

operator("slide", st(e,A,B,C,D), st(A,e,B,C,D)).

operator("jump", st(e,A,B,C,D), st(B,A,e,C,D)).
operator("slide", st(A,e,B,C,D), st(A,B,e,C,D)).
operator("slide", st(A,e,B,C,D), st(e,A,B,C,D)).
operator("jump", st(A,e,B,C,D), st(A,C,B,e,D)).
operator("slide",st(A,B,e,C,D),st(A,e,B,C,D)).
operator("slide",st(A,B,e,C,D),st(A,B,C,e,D)).
222 Search

operator("jump",st(A,B,e,C,D),st(e,B,A,C,D)).
operator("jump",st(A,B,e,C,D),st(A,B,D,C,e)).
operator("slide",st(A,B,C,e,D),st(A,B,e,C,D)).
operator("jump",st(A,B,C,e,D),st(A,e,C,B,D)).
operator("slide",st(A,B,C,e,D),st(A,B,C,D,e)).
operator("slide",st(A,B,C,D,e),st(A,B,C,e,D)).
operator("jump",st(A,B,C,D,e),st(A,B,e,D,C)).

bSearch([T|Queue],Solution) :-
if solved(T) then Solution= T
else Extentions= [Daughter || nextLevel(T, Daughter)],
ExtendedQueue= list::append(Queue, Extentions),
bSearch(ExtendedQueue, Solution) end if.

nextLevel([r(Branch, N)|Path], [r(Op, Daughter),r(Branch, N)|Path]) :-

operator(Op, N, Daughter),
not(list::isMember(r(Op, Daughter),Path)).

run():- console::init(),
if bSearch([[r("0", st(w,w,e,b,b))]], L) then prtSolution(L)
else stdio::write("No solution!"), stdio::nl end if.
end implement slide
goal mainExe::run(slide::run).

The solution of the puzzle, as printed by the computer, is presented below.

r("0",st(w(),w(),e(),b(),b()))
r("slide",st(w(),e(),w(),b(),b()))
r("jump",st(w(),b(),w(),e(),b()))
r("slide",st(w(),b(),e(),w(),b()))
r("jump",st(w(),b(),b(),w(),e()))

D:\vipro\aityros\aiprogs\slide\Exe>pause
Pressione qualquer tecla para continuar. . .

23.4 Depth First Strategy

Analyze1 figure 23.1. The depth first strategy, after visiting "b", goes to its
leftmost daughter, which is "c". After visiting "c", it goes to "f". Only
1
Yes, I know that analyse should be written with an S, since it comes from ἀνάλυσις, a
Greek word. The problem is that the Americans insist on writing it with a Z. That makes
me remember Professor Wolfgang Fuchs, who has an Erdös number equal to one, since
he worked directly with Paul Erdös. That makes my own Erdös number be as low as 2,
23.5 Heuristic search 223

after visiting all descendants of "c", the depth first strategy visits "g". To
implement this kind of strategy, all you need to do is to append the children
of a given node to the front of the Queue, instead of appending them to the
tail. Here is the only necessary modification made to the previous program:

dSearch([T|Queue],Solution) :-
if solved(T) then Solution= T
else Extentions= [Daughter || nextLevel(T, Daughter)],
ExtendedQueue= list::append(Extentions, Queue),
dSearch(ExtendedQueue, Solution) end if.

I also changed the name of the predicate to dSearch for obvious reasons.

23.5 Heuristic search

The search strategies that we have examined until now do not make any
use of domain specific knowledge to choose a branch to investigate. This
weakness is corrected by the heuristic strategy, that evaluates the cost of
going from the root to the goal passing through the node N , before jumping
into N. The evaluation function is given by the following formula:

f (N ) = g(N ) + h(N )

In this formula, g(N ) is the known cost of going from the root until the N
node; the function h(N ) gives the estimated cost of going from N to the goal
node. The heuristic strategy sorts the Queue in such a way that paths with
lower values of f (X) are visited first. Let us use the heuristic strategy to
find the path between two cities on the map below.

because Dr. Fuchs was my minor adviser. However, it is not to brag that I have a low
Erdös number that I am writing this footnote, but to tell the opinion that Dr. Fuchs had
about the French language. He said that French is certainly the most beautiful and sweet
sounding language in the world; the only problem is the awful pronunciation that French
people insist on maintaining.
224 Search

However, before starting, I would like to tell a story. I like obscure lan-
guages, and have learned quite a few of them. For instance, I know Ancient
Egyptian, Homeric Greek, Esperanto, Latin, etc. When I was a teenager, I
even studied Guarani, a second rate language spoken only in Paraguay, an
underpopulated country.
A guy or gal has translated a book by a certain V. N. Pushkin from
Russian into one of those second rate languages that I learned. The book’s
title was: Heuristics, the Science of Creative Thought. When I read Pushkin’s
book in translation, I understood how difficult is to find a good heuristic
function, and how much more difficult it is to translate a text from a language
to another, specially if you do not know the subject treated in the text well.
Natural language processing is difficult not only for computers, but for people
too. For instance, Dr. Pushkin tried to explain the rules of chess, in order to
use the game as a test bed for Artificial Intelligence. In his explanation, he
has said that the knight moves along a Γ-shaped path, where Γ is a letter of
the Russian alphabet. The translator came out with the following rendering
of the original: The knight moves along an upsidedown L-shaped path. After
that long digression into the realms of translation, let us return to heuristic
search. The cities will be represented by their coordinates:
xy("a", 2, 4). xy("b", 5, 6).
Two place facts describe the one way roads that connect the cities:
op("a", "b" ). op("b", "m"). op("m", "f").
23.5 Heuristic search 225

You have learned that, in the heuristic search, the queue must be sorted so
that the most promising paths come forward. The sorting algorithm that
you will learn is very interesting, and was invented by Tony Hoare. Each
path is represented by the structure shown below.
t(real, real, it)
where the first argument gives the value of the heuristic function F for the
node; the second argument gives the value of g(N). The sorting algorithm has
a comparing predicate as its first argument, and pushes the most promising
branches to the front of the queue.
search([T|Queue],S):-
if goalReached(T) then S= T
else Extension= [E || toDaughter(T,E)],
NewQueue= list::append(Queue,Extension),
BestFirst= list::sortBy(cmp, NewQueue),
search(BestFirst, S)
end if.
The complete program is implemented below.

implement main /*heureka*/

domains
item=r(string, string).
it= item*.
node=t(real, real, it).
tree=node*.
class facts
xy: (string, integer, integer).
op: (string, string).
class predicates
getXY:(string, integer, integer) determ (i, o, o).
cost: (string, string) -> real.
hn: (string) -> real.
not_in_circle: (string, it) determ (i,i).
theGoal: (string) procedure (o).
toDaughter: (node, node) nondeterm (i,o).
init:(string) procedure (o).
goalReached:(node) determ.
search: (tree, node) determ (i,o).
prtSolution: (node) procedure (i).
solve: () procedure.
cmp:(node, node) -> core::compareResult.
226 Search

clauses
classInfo("main", "heureka-1.0").

cmp(t(A, _, _), t(B, _, _)) = core::greater() :- A > B, !.

cmp(t(A, _, _), t(B, _, _)) = core::equal() :- A=B, !.
cmp(_, _) = core::less().

op("a", "b" ). op("b", "m"). op("m", "f").

op("f", "q"). op("q", "p"). op("p", "n").
op("p", "s"). op("b", "c"). op("c", "d").
op("d", "q"). op("d", "n"). op("d", "g").
op("n", "h"). op("n", "s"). op("h", "g").

init("a").

goalReached(t(_,_,[r(_,M)|_])):- theGoal(R), R=M.

theGoal("s").

not_in_circle(Stt, Path):-not(list::isMember(r("", Stt), Path)).

xy("a", 2, 4). xy("b", 5, 6).

xy("c", 4, 2). xy("d", 7, 4).
xy("f", 7, 8). xy("g", 8, 2).
xy("h", 10, 1). xy("m", 9,6).
xy("n", 11, 3). xy("p", 12, 6).
xy("q", 11, 7). xy("s", 13, 2).
getXY(M, X, Y) :- xy(M, X, Y), !.

cost(No, NoFilho) = C:-

if getXY(No, XN, YN), getXY(NoFilho, XF, YF)
then
C = math::sqrt(((XN-XF)*(XN-XF)) + ((YN - YF)*(YN - YF)))
else
C= 0.0
end if.

hn(N) = HN :- theGoal(S),
if getXY(S, XS, YS), getXY(N, XN, YN)
then HN= math::sqrt(((XN-XS)*(XN-XS)) +
((YN - YS)*(YN - YS)))
else HN= 0.0 end if.
23.5 Heuristic search 227

search([T|Queue],S):-
if goalReached(T) then S= T
else Extension= [E || toDaughter(T,E)],
NewQueue= list::append(Queue,Extension),
BestFirst= list::sortBy(cmp, NewQueue),
search(BestFirst, S)
end if.

toDaughter(t(_F,G,[r(B,N)|Path]),t(F1,G1,[r(Op, Child),r(B,N)|Path])):-
op(N, Child),
Op= string::format("%s to %s", N, Child),
not_in_circle(Child, Path),
G1 = G + cost(N, Child), F1 = G1 + hn(Child).

prtSolution( t(_,_,T)):-
foreach X= list::getMember_nd(list::reverse(T)) do
stdio::write(X), stdio::nl
end foreach.

solve():- if init(E), search([t(hn(E),0,[r("root",E)])],S)

then prtSolution(S)
else stdio::write("No solution") end if.

run():- console::init(), solve().

end implement main /*heureka*/

goal mainExe::run(main::run).
228 Search
Chapter 24

Neural Networks

Spain is not distinguished for its science. In fact, there are families in France
that have twice as many Nobel Laureates in science than the whole of Spain.
In Europe, there are very few countries with less Nobel Laureates than Spain;
one of them is Servia, and one of the Serbians who did not receive the Nobel
prize is Nikola Tesla, and a country that has Nikola Tesla, does not need
Nobel laureates. Notwithstanding, the only two Spanish Nobel laureates are
counted among the greatest scientists of all time, together with Newton, and
Archimedes. In this chapter, you will learn about one of them.
The Spanish scientist Don Santiago Ramón y Ca-
...
jal studied the physiology of the nervous system and ...
...
...
...
concluded that it was driven by cells that he called ....
......
...
....
.. .
....
.......... . ......
... .......... . .....................
neurons. Nowadays, scientists believe that a given .....
...
.............. ..
...
.
... .. ...
..... ........ ...................................
neuron receives inputs from other neurons through ........
.. ...
...
.
.
......
...........
................... ......................
links called synapses. By the way, synapsis is the .....
.
.....
.... .
..
. .......
.....
.....
.....
Greek word for link. Typically, a neuron collects .....
.....
.....
.....
.....
signals from others through a set of structures called .....
.....
.....
.....
.....
dendrites, which means tree shoots in Greek; if the .....
...

intensity of the incoming signals goes beyond a certain threshold, the neuron
is fired, and sends out electrical stimuli through a stem known as an axon,
which splits into the branch-like dendrites, whose synapses will activate other
neurons. So, to sum up. . . at the end of each dendrite, a synapse converts
neuron activity into electrical stimulus that inhibits or excites another neu-
ron.
In 1947, Warren McCullogh and Walter Pitts proposed the first mathe-
matical model for a neuron. Their neuron is a binary device, with a fixed
threshold; it receives multiple inputs from excitatory synapses, which gives
to each input source a weight that is effectively a positive integer; inhibitory
inputs have an absolute veto power over any excitatory inputs; at each pro-

229
230 Neural Networks

cessing step the neurons are synchronously updated by summing the weighted
excitatory inputs and setting the output to 1 if and only if the sum is greater
than or equal to the threshold.
The next major advance in the theory of neural networks was the percep-
tron, introduced by Frank Rosenblatt in his 1958 paper. Rosenblatt was a
professor at Cornell University, that happens to be my alma mater. However,
I did not have the good fortune of knowing him, since he passed away many
years before my arrival at that university.
'&%$
Ã!"#TTT
TTT
3.0 TTTT
'&%$
Ã!"# TT'&%$
*4
2.5
jjjj/Ã!"#
O / Output
jj
jjj 2.0
Ã!"#jjj
'&%$ −1.0

'&%$
Ã!"#

Figure 24.1: A three input perceptron

The perceptron adds up the input signals

1.0 ................
.............
........
.......
multiplied by a weight (a different weight for
.
......
0.8
..
...
...
...
each input signal), and if the result is greater
..

0.6 ..
..
.
... than a threshold, it produces a non zero out-
..
...
.
..
put. This scheme was improved by feeding the
0.4 ..
..
.
.
..
...
..
weighted sum to a sigmoid function, and taking
0.2 ..
..
.
...
....
...
the value returned as the output of the percep-
.......
........
0.0 .............................
tron. If you plot a sigmoid, it takes the shape
of the Greek letter ς, that is known as final
−6 −4 −2 0 2 4 6 sigma; hence the name sigmoid. The sigmoid
is defined as
1
ς(x) =
1 + e−x
In Prolog, this definition becomes:
sigmoid(X)= 1.0/(1.0+math::exp(-X)).
Figure 24.1 above shows a perceptron with three inputs. Let the inputs be
x1 , x2 , and x3 . In this case, the output will be the result of the expression
ς(3x1 + 2.5x2 + 2.0x1 − 1)
One interesting feature of the perceptron is its learning abilities: You can
train the perceptron to recognize input patterns. In fact, given enough ex-
amples, there is an algorithm that modifies the weights of the perceptron, so
 231

that it can discriminate between two or more patterns. For instance, a two
input perceptron can learn to tell when a and-gate will output 1, and when
it will output 0.
The learning algorithm of the perceptron must have a set of training
examples. Let us assume that it is learning the workings of an and-gate.
Here are the examples:

facts
eg:(unsigned, real, real, real).
clauses
eg(0, 1, 1, 1).
eg(1, 1, 0, 0).
eg(2, 0, 1, 0).
eg(3, 0, 0, 0).
P
The perceptron has a predicting procedure, that uses the formula ς( wi xi )
to predict the output value, given a set of inputs. In Prolog, the predicting
procedure can be coded as shown below.

predicted_out(E, V) :-
eg(E, I_1, I_2, _),
getWgt(0, W_0),
getWgt(1, W_1),
getWgt(2, W_2),
X = W_0+W_1* I_1 + W_2* I_2,
V= sigmoid(X).

The weights are stored in a database.

facts
weight:(unsigned, real).
clauses
weight(0, 1.5). weight(1, 2.0). weight(2, 1.8).

The initial weights are fictitious. The true weights must be found by the
learning algorithm. The procedure getWgt has a very simple definition:

getWgt(I, W) :- weight(I, W), !.

getWgt(_I, 0.0).
232 Neural Networks

You may ask why one needs to define getWgt at all; it would be easier to use
weight directly; the problem is that weight defines a nondeterm predicate,
and we need a procedural weight retrieving predicate. By the way, one needs
a procedure to retrieve the examples too:
egratia(Eg, I1, I2, Output) :- eg(Eg, I1, I2, Output), !.
egratia(Eg, I1, I2, Output) :- Ex = Eg rem 4,
eg(Ex, I1, I2, Output), !.
egratia(_Eg, 1, 1, 0).
The learning algorithm is based on the calculation and correction of the error
made by the predicting function. The error of a given example is calculated
by the following predicate:
evalError(Obj, E) :- nn:predicted_out(Obj, VC),
nn:egratia(Obj, _I1, _I2, V),
E= (VC-V)*(VC-V).
What we really need is not the error of a given example, but the sum of the
errors for each example. One can calculate this total error as shown below.
errSum(Exs, _Partial_err, Total_err) :- acc := 0.0,
foreach Eg= list::getMember_nd(Exs) do
evalError(Eg, Err), acc := acc + Err
end foreach,
Total_err= acc.
The computational scheme foreach will pick each example from the Exs list,
calculate the corresponding error, and add the result of the calculation to
the accumulator acc, which is stored in a global variable.
class facts
acc:real := 0.0.
The weights will be updated through the gradient descent method in such a
way as to reduce the error to a minimum.
updateWeight(DX, LC, Exs) :- errSum(Exs, 0.0, Err0),
PN= [tuple(I, W) || nn:getWgt_nd(I, W)],
nn:setWeight([ tuple(P, NV) ||
tuple(P, V) = list::getMember_nd(PN),
V1= V+DX,
nn:assertWgt(P, V1),
errSum(Exs, 0.0, Nerr),
nn:popweight(P, V1),
NV= V+LC*(Err0-Nerr)/DX]).
 233

Gradient descent is an optimization algorithm, that finds a local minimum of

a function by taking steps proportional to the negative of the gradient of the
function at the current point. At each step of the gradient descent method,
one updates the weight according to the following formula:
ωn+1 = ωn − γOerror(ω)
In this formula, if γ > 0 is a small enough number, error(ωn+1 ) < error(ωn ).
If one starts with ω0 , the sequence ω0 , ω1 , ω2 . . . will converge to a minimum.
The minimum will be reached when the gradient becomes zero (or close
to zero, in practical situations). You can think that there is a terrain, with
mountains and valleys; the position on this terrain is given by the coordinates
ω1 and ω2 ; the heights of mountains, and the depths of valleys represent
errors. The training algorithm moves the network from an initial position to
a valley, i.e., to a position where the error reaches a local minimum.
The ideal situation happens when the net-
work reaches a global minimum. The Ameri-
cans who serve in counter-espionage like to say
that they are the best of the best. That is what
one wants for the solution of any optimization
problem: The best of the best. Notwithstand-
ing, gradient methods more often than not be-
come trapped in local minimum. This happens
because when the network reaches a minimum,
it doesn’t matter whether the minimum is lo-
cal or global, its gradient comes close to zero;
in fact, the algorithm knows that it reached a
minimum, because the gradient reaches a value close to zero. The problem
is that this criterium does not say whether the minimum is local or global.
The dictionary says that a gradient gives the inclination of a road or railway.
At the top of a mountain, or at the bottom of a valley, the gradient is zero.
If one uses a level at the top of a mountain, or at the bottom of a valley, s/he
will not detect any inclination at all.
A neural network training algorithm updates the weights of a neuron,
until the error stays in some small neighbourhood; then it stops.
train(DX, LC, Exs) :- errSum(Exs, 0, Err0),
if (Err0 < 0.1) then
stdio::write("Total Err: ", Err0), stdio::nl
else
updateWeight(DX, LC, Exs),
train(DX, LC, Exs)
end if.
234 Neural Networks

24.1 Neuron description

A neuron like the one shown in figure 24.2 can be described by the following
equation:
ω1 x1 + ω2 x2 + ω0 = 0
This is the equation of a straight line. If one uses more entries, s/he will
get straight lines in high dimensions, but they will remain straight lines
nevertheless. Let us plot a logical function known as or-gate. It has two
inputs, x1 , and x2 .
'&%$
Ã!"#TTT
TTT
3.0 TTTT
TT'&%$
*4
jjjjjÃ!"#
O / Output
j
jjj 2.0
Ã!"#jjj
'&%$ −1.0

'&%$
Ã!"#

Figure 24.2: A three input perceptron

The or-gate function produces 1, if either

x2 x1 = 1, or x2 = 2; it also produces 1, if both
x1 = 1, and x2 = 1. The graphics of the figure
1 ....... ....... ....... ....... ....... ...........1 shows the behavior of the or-gate. What the
....

.....
..
..
neural network of figure 24.2 does is to draw a
.....
.....
.....
.....
.....
..
.
.
..
straight lines separating the points where or-
.....
.....
.....
.....
.....
..
....
gate is 1 from the point where it is 0, as you
.....
.....
..... x 1 can see from the figure. Then, all that a per-
0 .
1
ceptron can do is to learn how to draw straight
lines in a state space, in order to separate clusters of points with a given
property. The problem is that one often cannot draw such straight lines in a
multidimensional space.
The xor-gate is much more useful than the or-gate for two reasons. First,
the two inputs of the xor-gate can represent logic propositions like:
x1 = the machine is on; x2 = the machine is off
x1 = the door is open; x2 = the door is closed

Since the machine cannot be on, and off at the same time, x1 = 1 and x2 = 1
cannot be true; x1 = 0 and x2 = 0 cannot be true either. To describe logical
gates, one often uses truth tables, that give the output value for all values of
the input variables; below, you can see a truth table for the xor-gate.
24.2 Implementation of a multi-layer network 235

x1 x2 output
1 1 0
1 0 1
0 1 1
0 0 0

You can see from the figure that a single

straight line cannot discriminate between the x2
values of a xor-gate. However, two straight .....
.....
....... ....... ....... ....... ............... ....... 0
.....
.

lines can separate the values that happen to be 1 ..... ...

..... .
.....
..........
.. .......
zero, from values that happen to be 1, as you . ....
.....
.
.....
can see in the figure. Likewise, a single percep- .....
.....
.....
.....
.
.
.
.
..... ....
tron cannot learn how to calculate the output .....
.....
.....
.....
.....
..
..
values of a xor-gate, but a network with more .....
.....
..... x1
0 1
than one perceptron, like the one that you see
in figure 24.3, can perform the task. In the next paragraphs, you will learn
how to use objects to implement a multi-layer neuron network in Prolog.

'&%$
Ã!"#OO
OOO
OOO
OOO
OOO
'&%$
Ã!"#? O/'&%$
'
Ã!"#
Ä? OOOOO
−3.1
??? Ä
Input 1
?? ÄÄ OOO3.0
OOO
?? ÄÄÄ OOO
??ÄÄ 1.7 O'&%$
'7
Ã!"#
Ä? o / XOR
Ä
Ä ?? ? oo O gate
ÄÄ 1.5
?? 2.0 ooo
Ä oo
Ä ?? oo
ÄÄ ? ooooo
Ã!"#Ä
'&%$ 2.7 o7/ '&%$
Ã!"# −1.0
Input 2
o oooo
ooo
oo ooo
Ã!"#o
'&%$ '&%$
Ã!"#

Figure 24.3: A neural network that can learn how to calculate a xor-gate

24.2 Implementation of a multi-layer network

• Create a new console project.

Project Name: neurons

UI Strategy: console
236 Neural Networks

• Select the root of the Project Tree. Choose the option File/New from
the VDE Task Menu. This will open the Create Project Item dialog.
Select the Class tag from the left hand side pane; create a network
class, that Creates Objects. Press the Create-button. Add the class
specification to network.cl, network.i, and network.pro.

class network : network %File: network.cl

predicates
sigmoid:(real) -> real.
end class network

Add the class interface to network.i:

%File: network.i
interface network
open core
domains
example= e(real, real, real).
examples= example*.
predicates
setWeight:(tuple{unsigned, real}*) procedure.
predicted_out:(unsigned, real) procedure (i, o).
egratia:(unsigned, real, real, real) procedure (i, o, o, o).
getWgt_nd:(unsigned, real) nondeterm (o, o).
assertWgt:(unsigned I, real W) procedure (i, i).
popweight:(unsigned I, real W) procedure (i, i).
usit:( real*, real) procedure ( i, o).
setExamples:(examples) procedure (i).
getWgt:(unsigned, real) procedure (i, o).
hidden:(unsigned, unsigned*, real) procedure (i, i, o).
setEx:(unsigned, examples).
end interface network

Add the listing below to network.pro. Build the application.

%File network.pro
implement network
open core
constants
className = "network/network".
classVersion = "".
facts
weight:(unsigned, real).
eg:(unsigned, real, real, real).
24.2 Implementation of a multi-layer network 237

clauses

sigmoid(X)= 1.0/(1.0+math::exp(-X)).

getWgt_nd(I, R) :- weight(I, R).

assertWgt(I, R) :- asserta(weight(I, R)).

popweight(I, W) :- retract(weight(I, W)), !.

popweight(_, _).

setExamples(Es) :- retractall(eg(_,_,_,_)),
setEx(0, Es).

setEx(_N, []) :- !.
setEx(N, [e(I1, I2,R)|Es]) :-
assertz(eg(N, I1, I2, R)),
setEx(N+1, Es).

egratia(Eg, I1, I2, Output) :- eg(Eg, I1, I2, Output), !.

egratia(Eg, I1, I2, Output) :- Ex = Eg rem 4,
eg(Ex, I1, I2, Output), !.
egratia(_Eg, 1, 1, 0).

setWeight(Qs) :-
retractall(weight(_,_)),
foreach tuple(X, WX)= list::getMember_nd(Qs) do
assert(weight(X, WX))
end foreach.

getWgt(I, W) :- weight(I, W), !.

getWgt(_I, 0.0).

predicted_out(Obj, V) :-
hidden(Obj, [3,4,5], I_1),
hidden(Obj, [6,7,8], I_2),
getWgt(0, W_0),
getWgt(1, W_1),
getWgt(2, W_2),
X = W_0+W_1* I_1 + W_2* I_2,
V= sigmoid(X).
238 Neural Networks

hidden(Obj,[A,B,C],V) :- eg(Obj, I_1, I_2, _), !,

getWgt(A, WA),
getWgt(B, WB),
getWgt(C, WC),
XX = WA+WB* I_1+WC* I_2,
V=sigmoid(XX).
hidden(_, _, 0).

usit( [I1, I2], V) :-

getWgt(0, W_0), getWgt(1, W_1), getWgt(2, W_2),
getWgt(3, W_3), getWgt(4, W_4), getWgt(5, W_5),
getWgt(6, W_6), getWgt(7, W_7), getWgt(8, W_8), !,
X1 = W_3 + W_4*I1 + W_5*I2,
V1=sigmoid(X1),
X2 = W_6 + W_7*I1 + W_8*I2,
V2= sigmoid(X2),
VX = W_0 + W_1*V1 + W_2*V2,
V= sigmoid(VX).
usit( _, 0).
end implement network

Objects of the network class can play the role of a neuron. In the main
module, you will find the training program, and a test run() predicate for
the and-gate, and also for the xor-gate. Build the application, and Run in
Window.
%File: main.pro
implement main
open core
constants
className = "main".
classVersion = "nnxor".
clauses
classInfo(className, classVersion).

class facts
acc:real := 0.0.
nn:network := erroneous.

class predicates
evalError:(unsigned Obj, real Err) procedure (i, o).
errSum:(unsigned* Exs,
real Partial_err, real Total_err) procedure (i, i, o).
updateWeight:(real DX, real LC, unsigned* Exs) procedure (i, i, i).
24.2 Implementation of a multi-layer network 239

clauses
evalError(Obj, E) :- nn:predicted_out(Obj, VC),
nn:egratia(Obj, _I1, _I2, V),
E= (VC-V)*(VC-V).

errSum(Exs, _Partial_err, Total_err) :- acc := 0.0,

foreach Eg= list::getMember_nd(Exs) do
evalError(Eg, Err),
acc := acc + Err
end foreach,
Total_err= acc.

updateWeight(DX, LC, Exs) :- errSum(Exs, 0.0, Err0),

PN= [tuple(I, W) || nn:getWgt_nd(I, W)],
nn:setWeight([tuple(P, NV) || tuple(P, V)= list::getMember_nd(PN),
V1= V+DX,
nn:assertWgt(P, V1),
errSum(Exs, 0.0, Nerr),
nn:popweight(P, V1),
NV= V+LC*(Err0-Nerr)/DX]).

class predicates
train:(real DX, real LC, unsigned* Exs) procedure (i, i, i).
clauses
train(DX, LC, Exs) :- errSum(Exs, 0, Err0),
if (Err0 < 0.1) then
stdio::write("Total Err: ", Err0), stdio::nl
else
updateWeight(DX, LC, Exs),
train(DX, LC, Exs)
end if.

class predicates
training:(network, real, real) procedure (i, i, i).
clauses
training(NN, _DX, _LC) :- nn := NN,
WWW= [ tuple(0,0),tuple(1,2),tuple(2,0),
tuple(3,0),tuple(4,1),tuple(5,0),
tuple(6,0),tuple(7,0),tuple(8,0)],
nn:setWeight(WWW),
train(0.001,0.1,[0,1,2,3, 1, 2, 3, 0, 3, 2, 1, 0]).
240 Neural Networks

run():- console::init(),
XOR = network::new(), AND = network::new(),
XOR:setExamples([ network::e(1,1,0), network::e(1,0,1),
network::e(0,1,1), network::e(0,0,0)]),
AND:setExamples([ network::e(1,1,1), network::e(1,0,0),
network::e(0,1,0), network::e(0,0,0)]),
training(XOR, 0.001, 0.5),
training(AND, 0.001, 0.5),
XOR:usit( [1,1], XOR11), stdio::nl,
stdio::write("xor(1,1)=", XOR11), stdio::nl,
XOR:usit( [1,0], XOR10), stdio::nl,
stdio::write("xor(1,0)=", XOR10), stdio::nl,
XOR:usit( [0,1], XOR01), stdio::nl,
stdio::write("xor(0,1)=", XOR01), stdio::nl,
XOR:usit( [0,0], XOR00), stdio::nl,
stdio::write("xor(0,0)=", XOR00), stdio::nl, stdio::nl,
AND:usit( [1,1], AND11), stdio::nl,
stdio::write("1&&1=", AND11), stdio::nl,
AND:usit( [1,0], AND10), stdio::nl,
stdio::write("1&&0=", AND10), stdio::nl,
AND:usit( [0,1], AND01), stdio::nl,
stdio::write("0&&1=", AND01), stdio::nl,
AND:usit( [0,0], AND00), stdio::nl,
stdio::write("0&&0=", AND00), stdio::nl.

end implement main

goal
mainExe::run(main::run).

24.3 Running the two layer neural network

After building the neurons application, and running it, you will get the
following result:

Total Err: 0.09983168220091619

Total Err: 0.09968265082000113

xor(1,1)=0.08040475166406771

xor(1,0)=0.9194034557462105
24.4 A Historical Perspective 241

xor(0,1)=0.8913291572885467

xor(0,0)=0.0922342035736988

1&&1=0.8712088309977442

1&&0=0.08891005182518963

0&&1=0.09297532089700676

0&&0=0.009538209942478315

D:\vipro\aityros\aiprogs\neurons\Exe>pause
Press any key to continue. . .

As you can see, the learning algorithm converged to a behavior that is very
close to the desired behavior of the xor-gate.

24.4 A Historical Perspective

There are people who claim that, when Rosenblatt introduced the concept
of perceptron, Minsky crusaded against him on a very nasty and personal
level, including contacting every group who funded Rosenblatt’s research to
denounce him as a charlatan, hoping to ruin Rosenblatt professionally and
to cut off all funding for his research in neural nets. As I told you, I was
not lucky enough to know Rosenblatt, though I know Minsky well (I inter-
viewed him when I used to double as a newspaper reporter), and I am sure
that such slanders against Minsky simply aren’t true. What Minsky and his
collaborator Papert did was to point out the weakness of the theory that
the intelligence can emerge purely from the learning activities of a neural
network. In fact, Minsky pratically demonstrated a theorem that neural net-
work cannot learn how to do certain things; this limitation is not exclusive
to artificial neural networks; there are things that even our natural neural
network cannot learn: In the cover of Minsk’s book (Perceptrons), there is
a labyrinth like drawing; the reader should discover whether the labyrinth
interlaced walls are connected or not. Minsk claims that it is impossible
to train a neural network to perform such a task, and it does not matter
whether the neural network is natural, or artificial.
242 Neural Networks
Chapter 25

Alpha Beta Pruning

Mancala is a family of board games that are very popular in Africa and
Asia; it is a count and capture game. The game of this family best known
in the Western world is Kalah. Mancala games play a role in many African
and some Asian societies comparable to that of chess in the West. It is also
popular in India, specially among the Tamil women.

In the variant of Mancala that we are going to

learn, the board has a series of holes arranged in two
rows, one for each player. The holes may be referred
to as pits, pots, or houses. Two large holes at two
ends of the board, called stores, are used for holding
captured pieces. Playing pieces are seeds, beans, or
stones, that are placed in and transferred between
the holes during play. At the beginning of a player’s
turn, s/he selects a hole from his/her side, and sows
its seeds around the board.

In the process of sowing, all the seeds from a hole are dropped one-by-one
into subsequent holes in a motion wrapping around the board. Depending
on the contents of each hole sown in a lap, a player may capture seeds from
the board. If the hole contains less than 3 seeds, the player can capture its
contents, and also the seed that s/he would drop in that hole. The captured
seeds go to the players store. The game ends if one player captures at least
10 seeds more than his/her opponent. The game also ends if one player has
no seeds on his side of the board. In any case, the player that captures more
seeds wins the game.

243
244 Alpha Beta Pruning

25.1 Move generation

The most cherished games are those where an oponent is pitted against the
other, in a combat of brains. However, I would like to point out that even
in this kind o game, Erast Fandorin always win, as you can check by reading
The Jack of Spades. However, computers are not as lucky as the state coun-
sellor; therefore programmers need a tool to analyze the game; this tool is
the minmax tree (see figure 10.1). In such a diagram, a player is called the
Maximizer, while the other is the Minimizer. In the analysis, the Maximizer
is supposed to achieve the maximum value for an evaluation funcion, hence
the name; the minimizer tries to force the evaluation function to a minimum
value. To tell the whole story, one does not use a tree, like the one in fig-
ure 10.1. To understand why, let us suppose that the maximizer is analyzing
the move B, of the tree below. She knows that the Minimizer will play F,
since F is the counter-move of least value. Therefore she concludes that B is
worth 5 from her point of view; now she proceeds to the analysis of C; when
she discovers that H=3, she can suspend the search, because a move to C
will be worth at most 3, and she already has 5 in B. This process of skip-
ping branches like I and J is called Alpha-pruning. When reasoning from the
point of view of the Minimizer, all one needs to do is to invert the signal of
the evaluation function, and apply the same algorithm that was used in the
case of the Maximizer; in this case, skipping branches is called Beta-pruning.
Finally, if the node is a leaf, evaluation is performed by a static function, i.e.,
a function that does not descend to the branches.
A

B=5 C D

E=6 F=5 G=7 H=3 I J

Implementing a game of strategy is a difficult task. In order to accomplish

it, we are going to keep the move generation as a separate class, so the
complexities of the game itself will not interfere with our strategic reasoning.
Create a new console project.

Project Name: kalahGame

UI Strategy: console
25.1 Move generation 245

Build the application in order to insert the main class into the Project Tree.
Then create a class mv by selecting the option New in New Package from
the IDE task menu.

%File: mv.cl
class mv
open core
domains
side= a;b.
board= none ; t( side, integer LastMove,
integer Pot, integer Pot,
integer* Brd).

predicates
prtBoard:(board) procedure (i).
newBoard:(side, board) procedure (i, o).
addPot:(side, integer, integer, integer,
integer, integer) procedure (i,i,i,i, o,o).
changeSides:(side, side) procedure (i, o).
sowing:(side, integer, integer, positive,
integer, integer, integer*,
board) procedure (i, i, i, i, i, i, i, o).
sow:(integer, board, board) procedure (i,i,o).

play:(board, board) nondeterm (i, o).

stateval:(board, integer) procedure (i, o).
end class mv

%File: mv.pro
implement mv
open core

clauses
prtBoard(t(S, LM, P1, P2, [I0, I1, I2, I3, I4, I5, I6, I7])) :- !,
stdio::write("Move: ", LM), stdio::nl,
if S= a then T= "a" else T= "b" end if,
stdio::writef(" %4d %4d %4d %4d\n", I3, I2, I1,I0),
stdio::writef("%4d %4d %s\n", P1, P2, T),
stdio::writef(" %4d %4d %4d %4d\n", I4, I5, I6, I7).
prtBoard(_) :- stdio::write("None\n").
246 Alpha Beta Pruning

newBoard(Side, t(Side, -1, 0, 0, [6,6,6,6, 6,6,6,6])).

addPot(a, C, P1, P2, P1+C+1, P2) :- !.

addPot(b, C, P1, P2, P1, P2+C+1).

changeSides(a, b) :- !.
changeSides(b, a).

sow(I, t(Side, _LastMove, P1, P2, B), NewBoard) :-

J= I rem 8,
Counts= list::nth(J, B), !,
list::setNth(J, B, 0, BA),
K= (J+1) rem 8,
sowing(Side, J, Counts, K, P1, P2, BA, NewBoard).
sow(_, T, T).

sowing(S, LM, N, K, P1, P2, BA, NuB) :- N>0,

C= list::nth(K, BA),
C > 0, C < 3, !,
list::setNth(K, BA, 0, BB),
J= (K+1) rem 8,
addPot(S, C, P1, P2, PP1, PP2),
sowing(S, LM, N-1, J, PP1, PP2, BB, NuB).
sowing(S, LM, N, K, P1, P2, BA, NuB) :- N > 0, !,
list::setNth(K, BA, list::nth(K, BA) + 1, BB),
J= (K+1) rem 8,
sowing(S, LM, N-1, J, P1, P2, BB, NuB).
sowing(S, LM, _, _, P1, P2, BB, t(S1, LM, P1, P2, BB)) :-
changeSides(S, S1).

stateval(t(_, _, P1, P2, _), P2-P1) :- !.

stateval(_, 0).

play(t(a, LM, P1, P2, L), X) :- list::nth(0, L) > 0,

sow(0, t(a, LM, P1, P2, L), X).
play(t(a, LM, P1, P2, L), X) :- list::nth(1, L) > 0,
sow(1, t(a, LM, P1, P2, L), X).
play(t(a, LM, P1, P2, L), X) :- list::nth(2, L) > 0,
sow(2, t(a, LM, P1, P2, L), X).
25.1 Move generation 247

play(t(a, LM, P1, P2, L), X) :- list::nth(3, L) > 0,

sow(3, t(a, LM, P1, P2, L), X).
play(t(b, LM, P1, P2, L), X) :- list::nth(4, L) > 0,
sow(4, t(b, LM, P1, P2, L), X).
play(t(b, LM, P1, P2, L), X) :- list::nth(5, L) > 0,
sow(5, t(b, LM, P1, P2, L), X).
play(t(b, LM, P1, P2, L), X) :- list::nth(6, L) > 0,
sow(6, t(b, LM, P1, P2, L), X).
play(t(b, LM, P1, P2, L), X) :- list::nth(7, L) > 0,
sow(7, t(b, LM, P1, P2, L), X).
end implement mv

Now you are ready to implement the main class, that will search the game-
tree for the best move.

implement main /* kalahGame */

open mv
clauses
classInfo("kalahGame", "1.0").
class predicates
maxeval:( mv::board, integer, integer,
integer, mv::board, integer) procedure (i, i, i, i, o, o).
mineval:( mv::board, integer, integer,
integer, mv::board, integer) procedure (i, i, i, i, o, o).
bestMax:( mv::board*, integer, integer, integer,
mv::board, integer) determ (i, i, i, i, o, o).
bestMin:( mv::board*, integer, integer, integer,
mv::board, integer) determ (i, i, i, i, o, o).
betaPruning:( mv::board*, integer, integer, integer,
mv::board, integer, mv::board, integer)
procedure (i, i, i, i, i, i, o, o).
alphaPruning:( mv::board*, integer, integer, integer,
mv::board, integer, mv::board, integer)
procedure (i, i, i, i, i, i, o, o).
clauses
maxeval(Pos, LookAhead, Alpha, Beta, Mov, Val) :-
LookAhead > 0,
Lcs = [P || play(Pos, P)],
bestMax(Lcs, LookAhead - 1, Alpha, Beta, Mov, Val), ! ;
stateval(Pos, Val), Mov= none.
248 Alpha Beta Pruning

bestMax([Lnc|Lncs], LookAhead, Alpha,

Beta, BestMove, BestVal) :- !,
mineval(Lnc, LookAhead, Alpha, Beta, _, Val),
betaPruning(Lncs, LookAhead, Alpha, Beta,
Lnc, Val, BestMove, BestVal).

betaPruning([], _, _, _, Lnc, Val, Lnc, Val) :- !.

betaPruning(_, _, _Alpha, Beta,
Lnc, Val, Lnc, Val) :- Val > Beta, !.
betaPruning(Lncs, LookAhead, Alpha, Beta,
Lnc, Val, MLnc, MVal) :-
if Val > Alpha then NewAlpha= Val
else NewAlpha= Alpha end if,
if bestMax(Lncs, LookAhead, NewAlpha, Beta, Lnc1, Val1),
Val1 > Val then MLnc= Lnc1, MVal= Val1
else MLnc= Lnc, MVal= Val end if.

mineval(Pos, LookAhead, Alpha, Beta, Mov, Val) :-

LookAhead > 0,
Lcs = [P || play(Pos, P)],
bestMin(Lcs, LookAhead - 1, Alpha, Beta, Mov, Val), ! ;
stateval(Pos, Val), Mov= none.

bestMin([Lnc|Lncs], LookAhead,
Alpha, Beta, BestMove, BestVal) :- !,
maxeval(Lnc, LookAhead, Alpha, Beta, _, Val),
alphaPruning(Lncs, LookAhead,
Alpha, Beta, Lnc, Val, BestMove, BestVal).

alphaPruning([], _, _, _, Lnc, Val, Lnc, Val) :- !.

alphaPruning(_, _, Alpha, _Beta,
Lnc, Val, Lnc, Val) :- Val < Alpha, !.
alphaPruning(Lncs, LookAhead, Alpha, Beta,
Lnc, Val, MLnc, MVal) :-
if Val < Beta then NewBeta= Val
else NewBeta= Beta end if,
if bestMin(Lncs, LookAhead, Alpha, NewBeta, Lnc1, Val1),
Val1 < Val then MLnc= Lnc1, MVal= Val1
else MLnc= Lnc, MVal= Val end if.
25.1 Move generation 249

class predicates
readMove:(board, board) procedure (i, o).
legalMove:(core::positive, board) determ.
game:(board, integer) procedure (i, i).
endGame:(board) determ (i).
announceWinner:(integer) procedure (i).
clauses
legalMove(I, t(a, _LM, _P1, _P2, X)) :-
I >= 0, I < 4,
list::nth(I, X) > 0.

readMove(Pos, ReadPos) :-
stdio::write("Present situation: "), stdio::nl,
prtBoard(Pos),
%stdio::write("Your move: "),
SR= stdio::readLine(),
trap(I1= toTerm(SR), _, fail),
stdio::nl,
legalMove(I1, Pos),
sow(I1, Pos, ReadPos),
prtBoard(ReadPos), !.
readMove(Pos, ReadPos) :-
stdio::write("Something failed"), stdio::nl,
readMove(Pos, ReadPos).

endGame(X) :- stateval(X, V), math::abs(V) > 10, !.

endGame(t(a, _LM, _P1, _P2, [I0, I1, I2, I3|_])) :-
I0 < 1, I1 < 1, I2 < 1, I3 < 1, !.
endGame(t(b, _LM, _P1, _P2, [_I0, _I1, _I2, _I3, I4, I5, I6, I7])) :-
I4 < 1, I5 < 1, I6 < 1, I7 < 1, !.

game(X, _Val) :- endGame(X), !, stateval(X, V),

prtBoard(X),
announceWinner(V).
game(none, Val) :- stdio::write("End game: ", Val),
announceWinner(Val), !.
game(Pos, _Val) :-readMove(Pos, NewPos),
maxeval(NewPos, 10, -10, 10, BestMove, BestVal), !,
game(BestMove, BestVal).
250 Alpha Beta Pruning

announceWinner(Val) :- Val > 0,

stdio::write("\n I won!\n", Val), !.
announceWinner(Val) :- Val < 0,
stdio::write("\n You won!\n", Val), !.
announceWinner(_Val) :- stdio::write("\n It is a draw!\n").

run():- console::init(),
newboard(a, B),
stateval(B, Val),
game(B, Val).

end implement main /* kalahGame */

goal mainExe::run(main::run).
Like Erast Fandorin, I don’t like games; besides this, I don’t have his luck,
which means that I always loose. All the same, I pitted my wits against the
computer, and inevitably lost. Here is the beginning of the game:
Present situation:
Move: -1
6 6 6 6
0 0 a
6 6 6 6
0

Move: 0
7 7 7 0
0 0 b
7 7 7 6
Present situation:
Move: 7
8 8 8 1
0 0 a
8 8 7 0
1

Move: 1
9 9 1 0
2 0 b
9 9 8 1
Present situation:
Move: 4
10 10 0 1
2 4 a
1 11 9 0
Index

:-, 49 Compiling a program, 15

conjunction, 48
Accumulator parameters, 63 console
and, 48 Run in Window, 58
animation, 163 testing, 58
append, 112 console applications, 57
average, 60 cut, 57
B+Tree, 201 days in a year, 118
handling, 205 determ, 52
background domains
transparent, 90 list declaration, 104
backtrack, 52 doPaint
Blinking animation, 159 Java, 28
Botany, 151 draw
brush, 89 GDIObject, 88
setBrush(B), 89 Drawing predicates, 53
Building, 15 pictClose/1, 159
pictDraw/4, 159
Class pictOpen/1, 159
creating, 39 ropSrcCopy, 160
interface, 135, 157 drawing predicates
with objects, 156 drawEllipse/2, 53
class getVPIWindow(), 53
open, 52
userControl, 167 Edit Control
class facts, 51 getText(), 201
class predicates, 51 Edit control
Code Expert, 18, 35 fact variable, 201
Coelho Edit field
How to solve it, 109 accessing, 39
Prolog by Example, 109 contents, 39
Colmerauer, 118, 143 editor, 167
color, 89 editorControl, 167

251
252 INDEX

loading, 168 books, 209

saving, 168 grammar
Enable File/New, 17 English parser, 81
Euclid, 79 German parser, 80
event Grammars, 79
onPainting, 87 Graphics
Executing a program, 15 turtle, 139
GUI
factorial, 57 Object-oriented, 33
facts
class facts, 51 Hoare, 115
file, 132 Horn clauses, 49
readString, 132 facts, 49
writeString, 132 head, 49
file (class), 132 tail, 49
Flickering
implication, 49
elimination, 159
Interlingua, 144
Flickering animation, 159
International language, 146
Form
iteraction, 62
create a new one, 36
creating, 17 Java
prototype, 17 doPaint, 28
form junctors, 48
new(Window), 87
onPainting, 88 L-Systems, 151
show(), 87 Laplace, 115
Format, 102 Latino Sine Flexione, 143
Forms, 17 Linaeus, 151
four colors, 119 Lindenmayer, 151
Appel and Haken, 119 L-Systems, 151
Functions List
definition, 41 domains, 104
of string, 68
game list
minimax algorithm, 125 append, 112
move generation, 123 average, 60
Nim, 121 domains, 60
Games, 155 intersection, 111
GDIObject, 88 member, 110
generate and test, 119 reverse, 113
Grammar sort, 115
INDEX 253

Lists, 58 onPaiting, 88
iterative reduction, 64 open class, 52
recusive reduction, 64
reduction, 64 painting
schemes, 64 event handling, 87
zip, 64 paiting
ZIP scheme, 66 drawPolygon, 88
lists Papert, 140
[X:Xs], 59 pattern style, 89
head, 59 Peano, 139
patterns, 59 curve, 143
tail, 59 Latino Sine Flexione, 143
logical and, 48 recursion postulate, 142
logical if, 49 Pereira
logical implication, 49 Luis Moniz, 119
logical junctors, 48 Piaget, 140
Logo, 140 pictDraw, 90
Lope de Vega, 155 pictLoad, 90
LsF picture
Latino sine Flexione, 147 mask, 90
pictDraw, 89
map pictLoad, 89
four colors, 119 transparent, 90
mask, 90 polymorphism, 110
member, 110 predicates
Menu class predicates, 51
Enable option, 17 declarations, 51
task, 37 definition of, 42
Mouse, 27 determ, 52
add code, 27 multi, 52
event, 27 nondeterm, 52
MouseDown, 27 procedure, 52
multi, 52 type/flow patterns, 51
multi-solutions, 46 procedure, 52
Natural Language Processing, 143 Project
Newton, 115 building, 15
Nim, 121 Create item, 36
nondeterm, 52 creating, 13
settings, 14
Object Project Tree, 34
interface, 157 Project tree
254 INDEX

add code, 18 editorControl, 167

Prolog
How to solve it, 109 Van Emden, 115
Pushkin, 155 variables, 43

quicksort, 115 Warren, 118

Hoare, 115 window
GDIObject, 88
readString, 132 Wittgenstein, 79
real*, 60 writeString, 132
Recursion
postulate, 142
recursion
iteraction, 62
Roussel, 118
setBrush, 89
solutions, 42
Multi solutions, 46
sort, 115
String, 68, 102
frontToken, 68
string
format, 102
from file, 132
Sundback
Gideon, 66
systems
Incremental development, 121
Task menu, 37
timer
handling, 166
transparent background, 90
Tratactus, 79
Turtle graphics, 139
Type, 102
Char, 102
Int, 102
Real, 102
String, 102
user control, 167
Bibliography

[Abramson/Dahl] Harvey Abramson and Veronica Dahl, ”Logic

Grammars”, Springer Verlag, 1989.

[Cedric et all] de Carvalho, C. L., Costa Pereira, A. E. and

da Silva Julia, R. M. ”Data-flow Synthesis for
Logic Programs”. In: System Analysis Model-
ing Simulation, Vol 36, pp. 349-366, 1999.

[Warren] David S. Warren. ”Computing With Logic:

Logic Programming With Prolog”. Addison-
Wesley / January 1988.

[Lewis Carroll] Lewis Carroll. The Game of Logic. Hard Press,

2006.

[Sterling and Shapiro] Sterling and Shapiro. ”The Art of Prolog”.

MIT Press / January 1986.

[Coelho/Cotta] Coelho, H. and Cotta, J. ”Prolog by Example”.

(1988) Berlin: Springer-Verlag.

[HowToSolveItWithProlog] Helder Coelho, Carlos Cotta, Luis Moniz

Pereira. ”How To Solve It With Prolog”. Min-
isterio do Equipamento Social. Laboratorio
Nacional de Engenharia Civil.

[Pereira/Shieber] F.C.N. Pereira and S.M. Shieber. ”Prolog and

Natural-Language Analysis”. CSLI Publica-
tions, 1987.

[Gazdar/Mellish] Gazdar, G., and C.S. Mellish. ”Natural Lan-

guage Processing in Prolog”. Addison-Wesley.

255
256 BIBLIOGRAPHY

[John Hughes] John Hughes. Why Functional Programming

Matters. Computer Journal, vol. 32, number
2, pages 98-107, 1989.

[Wadler & Bird] Richard Bird and Philip Wadler. Introduc-

tion to Functional Programming. Prentice Hall
(1988).

[John Backus] John Backus. Can Programming be Liberated

from the Von Newman Style? Communica-
tions of the ACM, 21(8):613-641, August 1978.

[Gries] David Gries. The Science of programming

— New York : Springer-Verlag, cop.1981.-
X,366p.

[Dijkstra] E. Dijkstra, W. Feijen: A Method of Program-

ming, Addison-Wesley Publishing Company.

[Cohn] Cohn, P. G. Logic Programming, Expressiv-

ity versus Efficiency. State University of Camp-
inas. 1991.

[Yamaki] Kimie Yamaki. Combining partial evaluation

and automatic control insertion to increase the
efficiency of Prolog programs. State University
of Campinas. 1990.

[Gentzen] Gerhard Karl Erich Gentzen. ”Die Wider-

spruchsfreiheit der reinen Zahlentheorie”,
Mathematische Annalen, 112: 493-565.(1936)

[Robinson] Robinson, J. A. Theorem Proving on the Com-

puter. Journal of the Association for Comput-
ing Machinery. Vol. 10, N. 12, pp 163-174.

[William Stearn] William T Stearn. Botanical Latin. Timber

Press. Paperback edition, 2004.