Visual Prolog Version 5.

Language Tutorial

(c) Copyright 1986-1997

Prolog Development Center A/S
H.J. Holst Vej 3A-5A, Copenhagen
DK - 2605 Broendby
Denmark
Copyright
The documentation for this software is copyrighted, and all rights are reserved. It
may not be reproduced, transmitted, stored in a retrieval system, or translated,
either by electronic, mechanical or any other means, without the prior written
consent of Prolog Development Center A/S.
The software products described in these manuals are also copyrighted, and are
licensed to the End User only for use in accordance with the End User License
Agreement, which is printed on the diskette packaging. The prospective user
should read this agreement carefully prior to use of the software.
Visual Prolog is a registered trademark of Prolog Development Center A/S.
Portions of the Software and Documentation under license from:
 Borland International; Copyright (c) 1986, 1988 Borland International.

Other brand and product names are trademarks or registered trademarks of their
respective holders.
Table of Contents
Using This Manual..............................................................................................1
PART 1 4
CHAPTER 1....................................................................................................5
What Needs to be Installed for This Book ?.......................................................5
Starting the Visual Prolog Environment.............................................................6
Opening an Editor Window................................................................................6
Running and Testing a Program.........................................................................6
Loading Examples From Disk............................................................................7
Handling Errors...................................................................................................7
PART 2 9
CHAPTER 2..................................................................................................10
PROgramming in LOGic..................................................................................10
Sentences: Facts and Rules...........................................................................11
Queries..........................................................................................................13
Variables: General Sentences.......................................................................16
Overview.......................................................................................................17
From Natural Language to Prolog Programs....................................................18
Clauses (Facts and Rules).............................................................................18
Predicates (Relations)...................................................................................22
Variables (General Clauses).........................................................................22
Goals (Queries).............................................................................................26
Comments.....................................................................................................29
What Is a Match?..............................................................................................30
Summary...........................................................................................................32
CHAPTER 3..................................................................................................35
Visual Prolog's Basic Program Sections...........................................................35
The Clauses Section......................................................................................36
The Predicates Section..................................................................................36
The Domains Section....................................................................................39
The Goal Section..........................................................................................43
A Closer Look at Declarations and Rules.........................................................43
Other Program Sections....................................................................................53
The Facts Section..........................................................................................53
The Constants Section..................................................................................53
The Global Sections......................................................................................55
The Compiler Directives...............................................................................55
Summary...........................................................................................................56

i
CHAPTER 4..................................................................................................59
Matching Things Up: Unification.....................................................................59
Backtracking.....................................................................................................62
Visual Prolog's Relentless Search for Solutions...........................................65
Controlling the Search for Solutions............................................................75
Using the fail Predicate.................................................................................76
Preventing Backtracking: The Cut................................................................78
Prolog from a Procedural Perspective..............................................................87
How Rules and Facts Are Like Procedures..................................................87
Summary...........................................................................................................93
CHAPTER 5..................................................................................................95
Simple Data Objects.........................................................................................95
Variables as Data Objects.............................................................................95
Constants as Data Objects............................................................................96
Compound Data Objects and Functors.............................................................98
Unification of Compound Objects................................................................99
Treating Several Items as One....................................................................100
Declaring Domains of Compound Objects.................................................105
Compound Mixed-Domain Declarations....................................................110
Summary.........................................................................................................113
CHAPTER 6................................................................................................114
Repetitive Processes.......................................................................................114
Backtracking Revisited...............................................................................114
Implementing Backtracking with Loops....................................................118
Recursive Procedures..................................................................................120
Tail Recursion Optimisation.......................................................................122
Using Arguments as Loop Variables..........................................................128
Recursive Data Structures...............................................................................132
Trees as a Data Type...................................................................................133
Binary Search Trees....................................................................................139
Summary.........................................................................................................145
CHAPTER 7................................................................................................146
What Is a List?................................................................................................146
Declaring Lists............................................................................................147
List Processing................................................................................................149
Using Lists......................................................................................................150
Writing Lists...............................................................................................150
Counting List Elements..............................................................................151
Tail Recursion Revisited.............................................................................153
List Membership.........................................................................................157
Appending One List to Another: Declarative and Procedural Programming
....................................................................................................................158

ii
 Finding All the Solutions at Once...................................................................162
Compound Lists..............................................................................................163
Summary.........................................................................................................169
CHAPTER 8................................................................................................171
Declaring the facts-sections............................................................................171
Using the facts sections..............................................................................173
Accessing the facts sections........................................................................173
Updating the facts section...........................................................................174
Facts determiner-keywords.........................................................................179
Saving a database of facts at runtime.........................................................181
Examples.........................................................................................................182
Summary.........................................................................................................186
CHAPTER 9................................................................................................187
Arithmetic Expressions...................................................................................187
Operations...................................................................................................188
Order of Evaluation....................................................................................188
Functions and Predicates................................................................................189
Generating Random Numbers....................................................................190
Integer and Real Arithmetic........................................................................192
Comparisons...................................................................................................196
Equality and the equal (=) Predicate...........................................................197
Comparing Characters, Strings, and Symbols............................................201
CHAPTER 10..............................................................................................203
Encapsulation..............................................................................................203
Objects and classes.....................................................................................203
Inheritance......................................................................................................204
Identity............................................................................................................204
Visual Prolog Classes.....................................................................................204
Class declarations...........................................................................................205
Class implementation......................................................................................205
Class instances................................................................................................206
Destroying Objects.........................................................................................207
Class Domains................................................................................................207
Sub-classing and inheritance..........................................................................208
Virtual Predicates............................................................................................210
Static facts and predicates...............................................................................211
Class Scopes...................................................................................................212
Constructors and Destructors..........................................................................214
Reference to the Object Itself (This)...............................................................215
Abstract Classes..............................................................................................216
Protected facts and predicates.........................................................................217
Derived class access control...........................................................................217

iii
Formal definition for classes...............................................................................219
CHAPTER 11..............................................................................................220
The Flow Analysis..........................................................................................220
Specifying Flowpatterns for Predicates......................................................222
Controlling the Flow Analysis....................................................................223
Reference Variables....................................................................................224
Declaring Domains as Reference................................................................224
Reference Domains and the Trail Array.....................................................225
Using Reference Domains..........................................................................226
Flow Patterns Revisited..............................................................................228
Using Binary Trees with Reference Domains............................................228
Sorting with Reference Domains................................................................230
Functions and Return Values..........................................................................231
Determinism Monitoring in Visual Prolog.....................................................233
Visual Prologs determinism checking system............................................235
Predicates as Arguments.................................................................................237
Predicate Domains......................................................................................237
Examples.....................................................................................................238
The Binary Domain........................................................................................242
Implementation of binary terms..................................................................243
Text syntax of Binary Terms......................................................................243
Creating Binary Terms...............................................................................244
Accessing Binary Terms.............................................................................245
Unifying Binary Terms...............................................................................245
Example......................................................................................................246
Converting Terms to Binary Terms............................................................247
Errors and Exception Handling.......................................................................249
Exception Handling and Error Trapping....................................................249
Error reporting............................................................................................252
Handling Errors from the Term Reader......................................................254
Break Control (Textmode Only).....................................................................255
Manual Break and Signal Checking in UNIX............................................257
Critical Error Handling under DOS Textmode...........................................261
Dynamic Cutting.............................................................................................264
Free Type Conversions...................................................................................265
Programming Style.........................................................................................266
Rules for Efficient Programming................................................................266
Using the fail Predicate...............................................................................269
Determinism vs. Non-determinism: Setting the Cut...................................270
PART 3 271
CHAPTER 12..............................................................................................272
Writing and Reading.......................................................................................272

iv
 Writing........................................................................................................272
Reading.......................................................................................................280
Binary Block Transfer................................................................................284
Visual Prolog's File System............................................................................285
Opening and Closing Files..........................................................................286
Redirecting Standard I/O............................................................................290
Working with Files.....................................................................................290
File Attributes.............................................................................................296
File and Path Names.......................................................................................300
Directory Searching........................................................................................301
Manipulating File Attributes...........................................................................306
Handling terms in text files.............................................................................307
Manipulating Facts Like Terms..................................................................308
Summary.........................................................................................................310
CHAPTER 13..............................................................................................313
String Processing............................................................................................313
Basic String-Handling Predicates...............................................................313
Type Conversion.............................................................................................320
Summary.........................................................................................................325
CHAPTER 14..............................................................................................327
External Databases in Visual Prolog..............................................................327
An Overview: What's in an External Database?.........................................328
Chains.........................................................................................................329
External Database Domains........................................................................331
Manipulating Whole External Databases...................................................333
Manipulating Chains...................................................................................339
Manipulating Terms....................................................................................341
A Complete Program Example...................................................................342
B+ Trees..........................................................................................................346
Pages, Order, and Keylength......................................................................347
Duplicate Keys............................................................................................347
Multiple Scans............................................................................................347
The B+ Tree Standard Predicates...............................................................348
Example: Accessing a Database via B+ Trees...........................................351
External Database Programming....................................................................353
Scanning through a Database......................................................................353
Displaying the Contents of a Database.......................................................355
Implementing a Database That Won't Break Down...................................357
Updating the Database................................................................................358
Using Internal B+ Tree Pointers.................................................................362
Changing the Structure of a Database.........................................................363
Filesharing and the External Database........................................................365

v
 Filesharing Domains...................................................................................366
Opening the Database in Sharemode..........................................................366
Programming with Filesharing...................................................................369
Implementing highlevel locking.................................................................372
A Complete Filesharing Example...............................................................372
Implementation Aspects of Visual Prolog Filesharing...............................378
Miscellaneous.................................................................................................379
Summary.........................................................................................................379
CHAPTER 15..............................................................................................381
Access to the operating system.......................................................................381
Timing Services..........................................................................................385
Bit-Level Operations.......................................................................................389
Access to the Hardware: Low-Level Support.................................................391
Summary.........................................................................................................395
CHAPTER 16..............................................................................................397
Building a Small Expert System.....................................................................397
Prototyping: A Simple Routing Problem........................................................402
Adventures in a Dangerous Cave....................................................................404
Hardware Simulation......................................................................................406
Towers of Hanoi.............................................................................................408
Dividing Words into Syllables........................................................................409
The N Queens Problem...................................................................................413
PART 4 417
CHAPTER 17..............................................................................................418
Names.............................................................................................................418
Program Sections............................................................................................419
The Domains Section..................................................................................421
The Predicates Section................................................................................426
The Facts Database Section........................................................................428
The Clauses Section....................................................................................428
The Constants Section................................................................................436
Conditional Compilation............................................................................439
Including Files in Your Program....................................................................439
Compiler Directives........................................................................................440
Visual Prolog Memory Management..............................................................447
Releasing Spare Memory Resources..........................................................447
Modular Programming....................................................................................448
Global Declarations....................................................................................448
CHAPTER 18..............................................................................................452
Using DLL’s...................................................................................................452
Calling Other Languages from Visual Prolog................................................453

vi
 Declaring External Predicates.....................................................................453
Calling Conventions and Parameter Passing..............................................453
Domain Implementation.................................................................................457
Simple Domains..............................................................................................458
Complex Domains......................................................................................458
Memory Considerations..................................................................................460
Memory Alignment....................................................................................461
Memory Allocation.....................................................................................462
Examples.........................................................................................................466
List Handling..............................................................................................466
Calling Prolog from Foreign Languages....................................................469
Calling an Assembler Routine from Visual Prolog....................................471
Index...................................................................................................................474

vii
Using This Manual

If you have never programmed in Prolog before, you should read all of this
manual. Chapters 1-2 cover Prolog fundamentals, and you should read them
before attempting any serious application development. The later chapters
become very important as soon as you want to do serious programming. If you
program in a procedural programming language such as C, Pascal, or Basic, pay
close attention to the procedural discussions. At the end of Chapter 3, you will
find a procedural overview of the material covered in the first three tutorial
chapters. We also provide procedural discussions of recursion in Chapter 4.
If you have programmed in other Prologs and have a good understanding of
Prolog fundamentals, you won't need much review. However, Visual Prolog has
several extensions and is different from interpreted Prologs. We recommend that
you read the release notes and Chapters 5 as an introduction. Chapter 6 explains
the structure of a Visual Prolog program and Chapter 7 introduces the
declarations. We also recommend that you read Chapter 8 on Visual Prolog's
facts section, and Chapter 9, on the external database.
Chapters 10 through 11 provide valuable information that you will need if you
plan to do serious programming.
If you think you could use a review of Visual Prolog programming, we
recommend that you read from Chapter 12 on.
This user's guide is divided into four parts: a short introduction to the Visual
Prolog environment; then the first ten tutorial chapters--which teach you how to
program in Visual Prolog; then six chapters--which gives an overview of the
predefined features of Visual Prolog - the standard predicates, the last part gives a
complete systematic overview of the language, modular programming and
interfacing to other languages.
Here's a summary of each chapter in this book:

Part 1: Introduction to Visual Prolog

Chapter 1: Getting Started describes how to run Visual Prolog on your system,
provides a quick guide to the menus and the editor, and takes you through the
steps involved in creating, running, and saving your first Visual Prolog program.

Part 2: Tutorial Chapters 2--10: Learning Visual Prolog

1
Chapter 2: Fundamentals of Prolog provides a general introduction to Prolog
from a natural language perspective and discusses how to convert natural
language statements and questions into Prolog facts, rules, and queries.
Chapter 3: Visual Prolog Programs covers Visual Prolog syntax, discusses the
sections of a Visual Prolog program, and introduces programming in Visual
Prolog.
Chapter 4: Unification and Backtracking describes how Visual Prolog solves
problems and assigns values to variables.
Chapter 5: Simple and Compound Objects discusses declaring and building
structures in Visual Prolog.
Chapter 6: Repetition and Recursion explains how to write repetitive
procedures using backtracking and recursion; also introduces recursive structures
and trees.
Chapter 7: Lists and Recursion introduces lists and their use through recursion,
as well as covers general list manipulation.
Chapter 8: The facts section discusses Visual Prolog's facts section for adding
facts to your program at run time and for storing global information.
Chapter 9: Classes and Objects gives a short introduction to object oriented
programming and introduces the object mechanism in Visual Prolog.
Chapter 10: Arithmetic and Comparison introduces the full range of arithmetic
and comparison functions built into Visual Prolog and gives examples that
demonstrate how to use them.
Chapter 11: Advanced Techniques controlling the flow analysis, using
reference variables, pointers to predicates, the binary domain, term conversions,
using the dynamic cut, tools and techniques for error and signal handling, and
programming style for efficient programs.
Part 3: Tutorial Chapters 13--14: Using Visual Prolog
Chapter 12: Writing, Reading, and Files introduces I/O in Visual Prolog;
covers reading and writing, and file- and directory-handling.
Chapter 13: String-Handling in Visual Prolog covers string manipulation,
including string comparison and string conversion, plus constructing and parsing
strings.
Chapter 14: The External Database System covers Visual Prolog's external
database system: chained data, B+ trees, storing data (in EMS, conventional
memory, and hard disk), and sorting data. Includes examples of constructing real
database applications.

2 Visual Prolog
Chapter 15: System-Level Programming introduces the low-level control
supported within Visual Prolog: system calls, BIOS, low-level memory
addressing, and bit manipulation.
Chapter 16: Example Prolog Programs provides a diverse group of Prolog
program examples demonstrating some of the elegant ways Prolog solves
complex problems.
Part 4: Reference Chapters 15--16: An overview

Chapter 17 Elements of the Language gives a systematic overview of all the

features in the Visual Prolog language. The chapter also introduces modular
programming.
Chapter 18 Interfacing with Other Languages gives a describtion on how to
interface with C and other languages

3
PA RT 1
Using the Environment

4 Visual Prolog
CHAPTER 17

Using the Visual Prolog Environment

18This chapter describes the basic operation of the Visual Prolog system focusing
on running the examples described in this book.
We assume, that you have experience using the Graphical User Interface system,
the windowing system. This might be either Windows 3.x, Windows 95,
Windows NT, Win-OS/2 or OS/2 PM. You should thus know about using menus,
closing and resizing windows, loading a file in the File Open dialog etc. If you do
not have this knowledge, you should not start off trying to create an application
that runs under this environment. You must first learn to use the environment.
If you are a beginner to Prolog, you don’t want to mix learning the Prolog
language with the complexity of creating Windows applications with event
handling and all the Windows options and possibilities. The code for the
examples in this book are platform independent: They can run in DOS text mode,
under UNIX, or with the adding of a little extra code for the User Interface
handling, in a Windowing environment like MS Windows or OS/2 PM.
We do suggest, that you at an early stage try compiling some of the examples in
the VPI subdirectory, and work your way through the Guided Tour in the Getting
Started book. This gives you an impression what can be done with Visual Prolog -
just so you know what you should be able to do, when you have learned to master
Visual Prolog.
However, if you are going to seriously use the Visual Prolog system, you need to
learn the basic concepts of Prolog properly. You will not be able to build a
skyscraper without having a solid foundation. In Visual Prolog, the foundation is
understanding the Prolog language and the VPI layer.

What Needs to be Installed for This Book ?

To Run and test the examples in this book you need during installation to:
Install the Visual Development Environment. You can choose the Win16
version or the Win32 version if you own the Professional version of
In the Dialog for choosing Documentation; install "Answers to exercises" and
"Examples".

Chapter 19, Using the Visual Prolog Environment 5

Starting the Visual Prolog Environment

The installation program will install a program group with an Icon which are
normally used to start the Visual Prolog Environment. However, there are many
ways to start an application in the GUI World, if you prefer another method, you
can just start the VIP.EXE down in the BIN\WIN\16 or the BIN\WIN\32
directories under the main VIP directory.
If Visual Prolog had an open project (a .VPR file) last time it was closed on your
computer, it will automatically reopen this project next time it starts. For the
examples in this manual, you do not need to have an open project.

Opening an Editor Window

To create a new edit window, you can use the menu command File | New. This
will bring up a new editor window with the title "NONAME".
The editor is a fairly standard text editor. It is documented in the VDE manual,
but you should be able to use cursor keys and the mouse as you are used to in
other editors. It supports cut, copy and Paste and Undo / Redo, which you can all
be activated from the Edit menu. Also the Edit menu shows the accelerator keys
associated for these actions.

Running and Testing a Program

To check, that your system is set up properly, you should try to type in the
following text in the window:
GOAL write("Hello world"),nl.

This is what is called a GOAL in the Prolog terminology, and this is enough to be
a program that can be executed. To execute the GOAL, you should activate the
menu item Project | Test Goal, or just press the accelerator key Ctrl+G. If your
system is installed properly, your screen will look like the following:

6 Visual Prolog
The result of the execution will come up in a separate window, which you must
close before you can test another GOAL.
(The Visual Prolog system treats the GOAL as a program which it compiles, links
and generates a Windows executable from. It uses the EASYWIN strategy, which
is described in the VDE manual.).

Loading Examples From Disk

The examples in this manual are provided on the disk. You can find them in the
subdirectory: DOC\EXAMPLES. Also there are some answers to exercises in the
subdirectory DOC\ANSWERS.
They're named after the chapter they appear in: chCCeNN.pro, where CC will be
02, 03, 04, etc. according to chapter, and NN is the example number within that
chapter (01, 02, 03, etc.).

Chapter 19, Using the Visual Prolog Environment 7

You should now try to open one of these example, and test one. You should just
use the File | Open command to open the file, and the press Ctrl+G to test the
GOAL.

Handling Errors

If you, like all programmers do, happen to make some errors in your program, the
Visual Prolog system will display an error window, which contains a list of
errors. You can double click on one of these errors to come to the position of the
error in the source text.

8 Visual Prolog
PA RT 2
Tutorial Chapters 20 -- 21: Learning
Prolog2223Visual Prolog

Chapter 24, Prolog Fundamentals 9

CHAPTER 25

Prolog Fundamentals
26This is the first in a sequence of chapters giving a step-by-step tutorial
introduction to the Visual Prolog language. We begin this chapter with an
introduction to programming in logic. After that we discuss some of Prolog's
basic concepts, including clauses, predicates, variables, goals, and matching.

PROgramming in LOGic

In Prolog, you arrive at solutions by logically inferring one thing from something
already known. Typically, a Prolog program isn't a sequence of actions--it's a
collection of facts together with rules for drawing conclusions from those facts.
Prolog is therefore what is known as a declarative language.
Prolog is based on Horn clauses, which are a subset of a formal system called
predicate logic. Don't let this name scare you. Predicate logic is simply a way of
making it clear how reasoning is done. It's simpler than arithmetic once you get
used to it.
Prolog uses a simplified variation of predicate logic syntax because it provides an
easy-to-understand syntax very similar to natural language, and because
computers are not as fast, large, or as inexpensive as we would like. If Prolog
were to accept English statements, the compiler would need to know every
possible way something could be worded in English. In many cases, it would take
many times longer to translate the program into something the computer
understands than it would to run the program. The computer hardware needed to
run such a system would be monstrous.
Prolog includes an inference engine, which is a process for reasoning logically
about information. The inference engine includes a pattern matcher, which
retrieves stored (known) information by matching answers to questions. Prolog
tries to infer that a hypothesis is true (in other words, answer a question) by
questioning the set of information already known to be true. Prolog's known
world is the finite set of facts (and rules) that are given in the program.
One important feature of Prolog is that, in addition to logically finding answers to
the questions you pose, it can deal with alternatives and find all possible solutions

10 Visual Prolog
rather than only one. Instead of just proceeding from the beginning of the
program to the end, Prolog can actually back up and look for more than one way
of solving each part of the problem.
Predicate logic was developed to easily convey logic-based ideas into a written
form. Prolog takes advantage of this syntax to develop a programming language
based on logic. In predicate logic, you first eliminate all unnecessary words from
your sentences. You then transform the sentence, placing the relationship first and
grouping the objects after the relationship. The objects then become arguments
that the relationship acts upon. For example, the following sentences are
transformed into predicate logic syntax:

Natural Language: Predicate Logic:

A car is fun. fun(car).
red(rose).
A rose is red.
Bill likes a car if the car is fun. likes(bill, Car) if fun(Car).

Sentences: Facts and Rules

A Prolog programmer defines objects and relations, then defines rules about
when these relations are true. For example, the sentence
Bill likes dogs.

shows a relation between the objects Bill and dogs; the relation is likes. Here is a
rule that defines when the sentence Bill likes dogs. is true:
Bill likes dogs i f the dogs are nice.

Facts: What Is Known

In Prolog, a relation between objects is called a predicate. In natural language, a
relation is symbolized by a sentence. In the predicate logic that Prolog uses, a
relation is summarized in a simple phrase--a fact--that consists of the relation
name followed by the object or objects (enclosed in parentheses). As with a
sentence, the fact ends with a period (.).
Here are some more facts expressing "likes" relations in natural language:
Bill likes Cindy.
Cindy likes Bill.
Bill likes dogs.

Here are the same facts, written in Prolog syntax:

Chapter 24, Prolog Fundamentals 11

 likes(bill, cindy).
likes(cindy, bill).
likes(bill, dogs).

Facts can also express properties of objects as well as relations; in natural

language "Kermit is green" and "Caitlin is a girl." Here are some Prolog facts that
express these same properties:
green(kermit).
girl(caitlin).

Rules: What You Can Infer from Given Facts

Rules enable you to infer facts from other facts. Another way to say this is that a
rule, as conclusions is a conclusion that is known to be true if one or more other
conclusions or facts are found to be true. Here are some rules concerning a "likes"
relation:
Cindy likes everything that Bill likes.
Caitlin likes everything that is green.

Given these rules, you can infer from the previous facts some of the things that
Cindy and Caitlin like:
Cindy likes Cindy.
Caitlin likes Kermit.

To encode these same rules into Prolog, you only need to change the syntax a
little, like this:
likes(cindy, Something):- likes(bill, Something).
likes(caitlin, Something):- green(Something).

The :- symbol is simply pronounced "if", and serves to separate the two parts of a
rule: the head and the body.
You can also think of a rule as a procedure. In other words, these rules
likes(cindy, Something):- likes(bill, Something)
likes(caitlin, Something):- green(Something).

also mean "To prove that Cindy likes something, prove that Bill likes that same
thing" and "To prove that Caitlin likes something, prove that it is green." In the
sameside effects procedural way, a rule can ask Prolog to perform actions other
than proving things--such as writing something or creating a file.

12 Visual Prolog
Queries
Once we give Prolog a set of facts, we can proceed to ask questions concerning
these facts; this is known as querying the Prolog system. We can ask Prolog the
same type of questions that we would ask you about these relations. Based upon
the known facts and rules given earlier, you can answer questions about these
relations, just as Prolog can.
In natural language, we ask you:
Does Bill like Cindy?

In Prolog syntax, we ask Prolog:

likes(bill, cindy).

Given this query, Prolog would answer

yes

because Prolog has a fact that says so. As a little more complicated and general
question, we could ask you in natural language:
What does Bill like?

In Prolog syntax, we ask Prolog:

likes(bill, What).

Notice that Prolog syntax does not change when you ask a question: this query
looks very similar to a fact. However, it is important to notice that the second
object--What--begins with a capital letter, while the first object--bill--does not.
This is because bill is a fixed, constant object--aconstants known value--but What
is a variable. Variables always begin with an upper-case letter or an underscore.
Prolog always looks for an answer to a query by starting at the top of the facts. It
looks at each fact until it reaches the bottom, where there are no more. Given the
query about what Bill likes, Prolog will return
What=cindy
What=dogs
2 Solutions

This is because Prolog knows

likes(bill, cindy).

and

Chapter 24, Prolog Fundamentals 13

 likes(bill, dogs).

We hope that you draw the same conclusion.

If we were to ask you (and Prolog):
What does Cindy like?

likes(cindy, What).

Prolog would answer

What = bill
What = cindy
What = dogs
3 solutions

This is because Prolog knows that Cindy likes Bill, and that Cindy likes what Bill
likes, and that Bill likes Cindy and dogs.
We could ask Prolog other questions that we might ask a person; however, a
question such as "What girl does Bill like?" will yield no solution because Prolog,
in this case, knows no facts about girls, and it can't draw any conclusions based
on material not known (supplied to it). In this example, we have not given Prolog
any relation or property to determine if any of the objects are girls.

Putting Facts, Rules, and Queries Together

Suppose you have the following facts and rules:
A fast car is fun.
A big car is nice.
A little car is practical.
Bill likes a car if the car is fun.

When you read these facts, you can deduce that Bill likes a fast car. In much
the same way, Prolog will come to the same conclusion. If no fact were given
about fast cars, then you would not be able to logically deduce what kind of a
car Bill likes. You could take a guess at what kind of a car might be fun, but
Prolog only knows what you tell it; Prolog does not guess.
Here's an example demonstrating how Prolog uses rules to answer queries. Look
at the facts and rules in this portion of Program 1:

14 Visual Prolog
 likes(ellen, tennis).
likes(john, football).
likes(tom, baseball).
likes(eric, swimming).
likes(mark, tennis).
likes(bill, Activity):- likes(tom, Activity).

The last line in Program 1 is a rule:

likes(bill, Activity):- likes(tom, Activity).

This rule corresponds to the natural language statement

Bill likes an activity if Tom likes that activity.

In this rule, the head is likes(bill, Activity), and the body is likes(tom,
Activity). Notice that there is no fact in this example about Bill liking
baseball. For Prolog to discover if Bill likes baseball, you can give the query
likes(bill, baseball).

When attempting to find a solution to this query, Prolog will use the rule:
likes(bill, Activity):- likes(tom, Activity).

Load Program ch02e01.pro into the Prolog System and run it.
PREDICATES
nondeterm likes(symbol,symbol)

CLAUSES
likes(ellen,tennis).
likes(john,football).
likes(tom,baseball).
likes(eric,swimming).
likes(mark,tennis).
likes(bill,Activity):-
likes(tom, Activity).

GOAL
likes(bill, baseball).

The system replies in the Dialog window

yes

It has combined the rule

likes(bill, Activity):- likes(tom, Activity).

with the fact

Chapter 24, Prolog Fundamentals 15

 likes(tom, baseball).

to decide that
likes(bill, baseball).

Try also this query:

likes(bill, tennis).

The system replies

no

Visual Prolog replies no to the latest query ("Does Bill like tennis?") because
There is no fact that says Bill likes tennis.
Bill's relationship with tennis can't be inferred using the given rule and the
available facts.
Of course, it may be that Bill absolutely adores tennis in real life, but Visual
Prolog's response is based only upon the facts and the rules you have given it
in the program.

Variables: General Sentences

In Prolog, variables enable you to write general facts and rules and ask general
questions. In natural language, you use variables in sentences all the time. A
typical general statement in English could be
Bill likes the same thing as Kim.

As we mentioned earlier in this chapter, to represent a variable in Prolog, the first

character of the name must be an upper-case letter or an underscore. For example,
in the following line, Thing is a variable.
likes(bill, Thing):- likes(kim, Thing).

In the preceding discussion of rules, you saw this line:

likes(cindy, Something):- likes(bill, Something).

The object Something begins with a capital letter because it is a variable; it must
be able to match anything that Bill likes. It could equally well have been called X
or Zorro.
The objects bill and cindy begin with lower-case letters because they are not
variables--instead, they are symbols, having a constant value. Visual Prolog can

16 Visual Prolog
also handle arbitrary text strings, much like we've been handling symbols above,
if the text is surrounded by double quotes. Hence, the token bill could have been
written as "Bill", if you wanted it to begin with an upper-case letter.

Overview
1. A Prolog program is made up of two types of phrases (also known as clauses):
facts and rules.
Facts are relations or properties that you, the programmer, know to be true.
Rules are dependent relations; they allow Prolog to infer one piece of
information from another. A rule becomes true if a given set of conditions
is proven to be true. Each rule depends upon proving its conditions to be
true.
2. In Prolog, all rules have two parts: a head and a body separated by the special
:- token.

The head is the fact that would be true if some number of conditions were
true. This is also known as the conclusion or the dependent relation.
The body is the set of conditions that must be true so that Prolog can prove
that the head of the rule is true.
3. As you may have already guessed, facts and rules are really the same, except
that a fact has no explicit body. The fact simply behaves as if it had a body
that was always true.
4. Once you give Prolog a set of facts and/or rules, you can proceed to ask
questions concerning these; this is known as querying the Prolog system.
Prolog always looks for a solution by starting at the top of the facts and/or
rules, and keeps looking until it reaches the bottom.
5. Prolog's inference engine takes the conditions of a rule (the body of the rule)
and looks through its list of known facts and rules, trying to satisfy the
conditions. Once all the conditions have been met, the dependent relation (the
head of the rule) is found to be true. If all the conditions can't be matched with
known facts, the rule doesn't conclude anything.

Exercises
Write natural language sentences that represent what these Prolog facts might
convey to a human reader. (Remember that, to the computer, these facts are
simple pieces of information that can be used for matching answers to questions.)
1. likes(jeff, painting).

Chapter 24, Prolog Fundamentals 17

2. male(john).

3. building("Empire State Building", new_york).

4. person(roslin, jeanie, "1429 East Sutter St.",

"Scotts Valley", "CA", 95066).

Write Visual Prolog facts that represent the following natural language
statements:
Helen likes pizza.
San Francisco is in California.
Amy's telephone number is 476-0299.
Len's father is Alphonso Grenaldi.

From Natural Language to Prolog Programs

In the first section of this chapter we talked about facts and rules, relations,
general sentences, and queries. Those words are all part of a discussion of logic
and natural language. Now we're going to discuss the same ideas, but we're going
to use more Prolog-ish words, like clauses, predicates, variables, and goals.

Clauses (Facts and Rules)

Basically, there are only two types of phrases that make up the Prolog language; a
phrase can be either a fact or a rule. These phrases are known in Prolog as
clauses. The heart of a Prolog program is made up of clauses.

More About Facts

A fact represents one single instance of either a property of an object or a relation
between objects. A fact is self-standing; Prolog doesn't need to look any further
for confirmation of the fact, and the fact can be used as a basis for inferences.

More About Rules

In Prolog, as in ordinary life, it is often possible to find out that something is true
by inferring it from other facts. The Prolog construct that describes what you can
infer from other information is a rule. A rule is a property or relation known to be
true when some set of other relations is known. Syntactically, these relations are
separated by commas, as we illustrate in example 1 below.

18 Visual Prolog
Examples of Rules
1. This first example shows a rule that can be used to conclude whether a menu
item is suitable for Diane.
Diane is a vegetarian and eats only what her doctor tells her to eat.

Given a menu and the preceding rule, you can conclude if Diane can order a
particular item on the menu. To do this, you must check to see if the item on
the menu matches the constraints given.
a. Is Food_on_menu a vegetable?
b. Is Food_on_menu on the doctor's list?
c. Conclusion: If both answers are yes, Diane can order Food_on_menu.
In Prolog, a relationship like this must be represented by a rule because the
conclusion is based on facts. Here's one way of writing the rule:
diane_can_eat(Food_on_menu):-
vegetable(Food_on_menu),
on_doctor_list(Food_on_menu).

Notice here the comma after vegetable(Food_on_menu). The comma introduces

a conjunction of several goals, and is simply read as "and"; both
vegetable(Food_on_menu) and on_doctor_list(Food_on_menu) must be true, for
diane_can_eat(Food_on_menu) to be true.

Suppose you want to make a Prolog fact that is true if Person1 is the parent of
Person2. This is easy enough; simply state the Prolog fact
parent(paul, samantha).

This shows that Paul is the parent of Samantha. But, suppose your Prolog
database already has facts stating father relationships. For example, "Paul is
the father of Samantha":
father(paul, samantha).

And you also have facts stating mother relationships; "Julie is the mother of
Samantha":
mother(julie, samantha).

If you already had a collection of facts stating these father/mother

relationships, it would be a waste of time to write parent facts into the
database for each parent relationship.

Chapter 24, Prolog Fundamentals 19

 Since you know that Person1 is the parent of Person2 if Person1 is the father
of Person2 or if Person1 is the mother of Person2, then why not write a rule to
convey these constraints? After stating these conditions in natural language, it
should be fairly simple to code this into a Prolog rule by writing a rule that
states the relationships.
parent(Person1, Person2):- father(Person1, Person2).
parent(Person1, Person2):- mother(Person1, Person2).

These Prolog rules simply state that

Person1 is the parent of Person2 if Person1 is the father of Person2.
Person1 is the parent of Person2 if Person1 is the mother of Person2.

3. Here's another example:

A person can buy a car if the person likes the car and the car is for
sale.

This natural language relationship can be conveyed in Prolog with the

following rule:
can_buy(Name, Model):-
person(Name),
car(Model),
likes(Name, Model),
for_sale(Model).

This rule shows the following relationship:

Name can_buy Model if
Name is a person and
Model is a car and
Name likes Model and
Model is for sale.

This Prolog rule will succeed if all four conditions in the body of the rule
succeed.
4. Here is a program designed to find solutions to this car-buying problem:
/* Program ch027e02.pro */

PREDICATES
nondeterm can_buy(symbol, symbol)
nondeterm person(symbol)
nondeterm car(symbol)
likes(symbol, symbol)
for_sale(symbol)

20 Visual Prolog
CLAUSES
can_buy(X,Y):-
person(X),
car(Y),
likes(X,Y),
for_sale(Y).

person(kelly).
person(judy).
person(ellen).
person(mark).

car(lemon).
car(hot_rod).

likes(kelly, hot_rod).
likes(judy, pizza).
likes(ellen, tennis).
likes(mark, tennis).

for_sale(pizza).
for_sale(lemon).
for_sale(hot_rod).

What can Judy and Kelly buy? Who can buy the hot rod? You can try the
following goals:
can_buy(Who, What).
can_buy(judy, What).
can_buy(kelly, What).
can_buy(Who, hot_rod).

Experiment! Add other facts and maybe even a rule or two to this Prolog
program. Test the new program with queries that you make up. Does Prolog
respond in a way you would expect it to?

Exercises
Write natural-language sentences corresponding to the following Visual Prolog
rules:
eats(Who, What):- food(What), likes(Who, What).

pass_class(Who):- did_homework(Who), good_attendance(Who).

does_not_eat(toby, Stuff):- food(Stuff), greasy(Stuff).

owns(Who, What):- bought(Who, What).

Chapter 24, Prolog Fundamentals 21

Write Visual Prolog rules that convey the meaning of these natural-language
sentences:
a. A person is hungry if that person's stomach is empty.
b. Everybody likes a job if it's fun and it pays well.
c. Sally likes french fries if they're cooked.
d. Everybody owns a car who buys one, pays for it, and keeps it.

Predicates (Relations)
The symbolic name of a relation is called the predicate name. The objects that it
relates are called its arguments; in the fact likes(bill, cindy)., the relation likes
is the predicate and the objects bill and cindy are the arguments.
Here are some examples of Prolog predicates with zero or more arguments:
pred(integer, symbol)
person(last, first, gender)
run
insert_mode
birthday(firstName, lastName, date)

As we've shown here, a predicate might not have any arguments at all, but the use
of such a predicate is limited. You can use a query such as
person(rosemont,Name,male). to find out Mr. Rosemont's first name. But what can
you do with the zero-argument query run? You can find out whether the clause
run is in the program, or--if run is the head of a rule, you can evaluate that rule.
This can be useful in a few cases--for instance, you might want to make a
program behave differently depending on whether the clause insert_mode. is
present.

Variables (General Clauses)

In a simple query, you can use variables to ask Prolog to find who likes tennis.
For example:
likes(X, tennis).

This query uses the letter X as a variable to indicate an unknown person. Variable
names in Visual Prolog must begin with a capital letter (or an underscore), after
which any number of letters (upper-case or lower-case), digits, or underline
characters (_) can be used. For example, the following are valid variable names:

22 Visual Prolog
 My_first_correct_variable_name
Sales_10_11_86

while the next three are invalid:

1stattempt
second_attempt
"disaster"

(Careful choice of variable names makes programs more readable. For example,
likes(Person, tennis).

is better than
likes(X, tennis).

because Person makes more sense than X.) Now try the goal
GOAL likes(Person, tennis).

Visual Prolog replies

Person=ellen
Person=mark
2 Solutions

because the goal can be solved in just two ways; namely, by taking the variable
Person and successively matching it with the values ellen and mark.
In variable names, except for the first character (which must be an upper-case
letter or an underscore), Visual Prolog allows lower-case or upper-case letters in
any position. One way to make variable names more readable is by using mixed
upper-case and lower-case letters, as in
IncomeAndExpenditureAccount

How Variables Get Their Values

You may have noticed that Prolog has no assignment statement; this is a
significant distinction between Prolog and other programming languages.
Variables in Prolog get their values by being matched to constants in facts or
rules.
Until it gets a value, a variable is said to be free; when it gets a value, it becomes
bound. But it only stays bound for the time needed to obtain one solution to the
query; then Prolog unbinds it, backs up, and looks for alternative solutions.

Chapter 24, Prolog Fundamentals 23

This is a very important point: You can't store information by giving a value to a
variable. Variables are used as part of the pattern-matching, process, not as a kind
of information storage.
Take a look at the following example, which uses program 3 to demonstrate how
and when variables get their values.
/* Program ch028e03.pro */

PREDICATES
nondeterm likes(symbol,symbol)

CLAUSES
likes(ellen,reading).
likes(john,computers).
likes(john,badminton).
likes(leonard,badminton).
likes(eric,swimming).
likes(eric,reading).

Consider this query: Is there a person who likes both reading and swimming?
likes(Person, reading), likes(Person, swimming).

Prolog will solve the two parts of this query by searching the program's clauses
from top to bottom. In the first part of the query
likes(Person, reading)

the variable Person is free; its value is unknown before Prolog attempts to find a
solution. On the other hand, the second argument, reading, is known. Prolog
searches for a fact that matches the first part of the query. The first fact in the
program
likes(ellen, reading)

is a match (reading in the fact matches reading in the query), so Prolog binds the
free variable Person to the value ellen, the relevant value in the fact. At the same
time, Prolog places a pointer in the list of facts indicating how far down the
search procedure has reached.
Next, in order for the query to be fully satisfied (find a person who likes both
reading and swimming), the second part must also be fulfilled. Since Person is
now bound to ellen, Prolog must search for the fact
likes(ellen, swimming)

24 Visual Prolog
Prolog searches for this fact from the beginning of the program, but no match
occurs (because there is no such fact in the program). The second part of the
query is not true when Person is ellen.
Prolog now "unbinds" Person and attempts another solution of the first part of the
query with Person once again a free variable. The search for another fact that
fulfills the first part of the query starts from the pointer in the list of facts. (This
returning to the place last marked is known as backtracking, which we'll cover in
chapter 29.)
Prolog looks for the next person who likes reading and finds the fact likes(eric,
reading). Person is now bound to eric, and Prolog tries once again to satisfy the
second part of the query, this time by looking in the program for the fact
likes(eric, swimming)

This time it finds a match (the last clause in the program), and the query is fully
satisfied. Prolog returns
Person=eric
1 Solution

Anonymous Variables
Anonymous variables enable you to unclutter your programs. If you only need
certain information from a query, you can use anonymous variables to ignore the
values you don't need. In Prolog, the anonymous variable is represented by a lone
underscore ("_").
The following parents example demonstrates how the anonymous variable is
used. Load Program 4.
/* Program ch030e04.pro */

PREDICATES
male(symbol)
female(symbol)
nondeterm parent(symbol, symbol)

CLAUSES
male(bill).
male(joe).

female(sue).
female(tammy).

Chapter 24, Prolog Fundamentals 25

 parent(bill,joe).
parent(sue,joe).
parent(joe,tammy).

The anonymous variable can be used in place of any other variable. The
difference is that the anonymous variable will never get set to a value.
For example, in the following query, you need to know which people are parents,
but you don't need to know who their children are. Prolog realizes that each time
you use the underscore symbol in the query, you don't need information about
what value is represented in that variable's place.
GOAL
parent(Parent, _).

Given this query, Prolog replies

Parent=bill
Parent=sue
Parent=joe
3 Solutions

In this case, because of the anonymous variable, Prolog finds and reports three
parents, but it does not report the values associated with the second argument in
the parent clause.
Anonymous variables can also be used in facts. The following Prolog facts
owns(_, shoes).
eats(_).

could be used to express the natural language statements

Everyone owns shoes.
Everyone eats.

The anonymous variable matches anything. A named variable would work

equally well in most cases, but its name would serve no useful purpose.

Goals (Queries)
Up to now, we've been mixing the word query when talking about the questions
you ask Prolog, with the more common name goal, which we'll use from now on.
Referring to queries as goals should make sense: when you query Prolog, you are
actually giving it a goal to accomplish ("Find an answer to this question, if one
exists: ...").

26 Visual Prolog
Goals can be simple, such as these two:
likes(ellen, swimming).

likes(bill, What).

or they can be more complex. In the "Variables" section of this chapter, you saw a
goal made up of two parts:
likes(Person, reading), likes(Person, swimming).

A goal made up of two or more parts is known as a compound goal, and each part
of the compound goal is called a subgoal.
Often you need to know the intersection of two goals. For instance, in the
previous parents example, you might also need to know which persons are male
parents. You can get Prolog to search for the solutions to such a query by setting a
compound goal. Load the Program 4 and enter the following compound goal:
Goal parent(Person, _), male(Person).

Prolog will first try to solve the subgoal

parent(Person, _)

by searching the clauses for a match, then binding the variable Person to a value
returned by parent (Person is a parent). The value that parent returns will then
provide the second subgoal with the value on which to search (Is Person--now
bound--a male?).
male(Person)

If you entered the goal correctly, Prolog will answer

Person=bill
Person=joe
2 Solutions

Compound Goals: Conjunctions and Disjunctions

As you have seen, you can use a compound goal to find a solution where both
subgoal A and subgoal B are true (a conjunction), by separating the subgoals with
a comma, but this is not all. You can also find a solution where subgoal A or
subgoal B is true (a disjunction), by separating the subgoals with a semicolon.
Here's an example program illustrating this idea:

Chapter 24, Prolog Fundamentals 27

 /* Program ch031e05.pro */

PREDICATES
car(symbol,long,integer,symbol,long)
truck(symbol,long,integer,symbol,long)
nondeterm vehicle(symbol,long,integer,symbol,long)

CLAUSES
car(chrysler,130000,3,red,12000).
car(ford,90000,4,gray,25000).
car(datsun,8000,1,red,30000).

truck(ford,80000,6,blue,8000).
truck(datsun,50000,5,orange,20000).
truck(toyota,25000,2,black,25000).

vehicle(Make,Odometer,Age,Color,Price):-
car(Make,Odometer,Age,Color,Price);
truck(Make,Odometer,Age,Color,Price).

Load and run this program, then try the goal

GOAL car(Make, Odometer, Years_on_road, Body, 25000).

This goal attempts to find a car described in the clauses that costs exactly
$25,000. Prolog replies
Make=ford, Odometer=90000, Years_on_road=4, Body=gray
1 Solution

But this goal is slightly unnatural, since you'd probably rather ask a question like:
Is there a car listed that costs less than $25,000?

You can get Visual Prolog to search for a solution by setting this compound goal:
car(Make, Odometer, Years_on_road, Body, Cost), /*subgoal A and*/
Cost < 25000. /*subgoal B */

This is known as a conjunction. To fulfill this compound goal, Prolog will try to
solve the subgoals in order. First, it will try to solve
car(Make, Odometer, Years_on_road, Body, Cost).

and then
Cost < 25000.

28 Visual Prolog
with the variable Cost referring to the same value in both subgoals. Try it out
now.
Note: The subgoal Cost < 25000 involves the relation less than, which is built
into the Visual Prolog system. The less than relation is no different from any
other relation involving two numeric objects, but it is more natural to place the
symbol for it between the two objects.
Now we will try to see if the following, expressed in natural language, is
true:goals, disjunctivedisjunctive goals queries, disjunctive
Is there a car listed that costs less than $25,000?, or is there a truck
listed that costs less than $20,000?

Prolog will search for a solution if you set this compound goal:
car(Make,Odometer,Years_on_road,Body,Cost), Cost<25000 ;
/* subgoal A or */
truck(Make,Odometer,Years_on_road,Body,Cost), Cost < 20000.
/* subgoal B */

This kind of compound goal is known as a disjunction. This one sets up the two
subgoals as alternatives, much as though they were two clauses for the same rule.
Prolog will then find any solution that satisfies either of the subgoals.
To fulfill this compound goal, Prolog will try to solve the first subgoal ("find a
car ..."), which is composed of these subgoals:
car(Make, Odometer, Years_on_road, Body, Cost.)

and
Cost < 25000.

If a car is found, the goal will succeed; if not, Prolog will try to fulfill the second
compound goal ("find a truck ..."), made up of the subgoals
truck(Make, Odometer, Years_on_road, Body, Cost),

and
Cost < 20000.

Comments
It's good programming style to include comments in your program to explain
things that might not be obvious to someone else (or to you in six months). This

Chapter 24, Prolog Fundamentals 29

makes the program easy for you and others to understand. If you choose
appropriate names for variables, predicates, and domains, you'll need fewer
comments, since the program will be more self-explanatory.
Multiple-line comments must begin with the characters /* (slash, asterisk) and
end with the characters */ (asterisk, slash). To set off single-line comments, you
can use these same characters, or you can begin the comment with a percent sign
(%).
/* This is an example of a comment */

% This is also a comment

/***************************************/
/* and so are these three lines */
/***************************************/

/*You can also nest a Visual Prolog comment /*within a comment*/ like
this */

In Visual Prolog 5.0 you can also use a comment after de decalratition of a
domain.
DOMAINS
articles = book(STRING title, STRING author); horse(STRING name)

PREDICATES
conv(STRING uppercase,STRING lowercase)

The words title, author, name, uppercase and lowercase will be ignored by the
compiler, but makes the program much more readable.

What Is a Match?

In the previous sections of this chapter, we've talked about Prolog "matching
answers to questions", "finding a match", "matching conditions with facts",
"matching variables with constants", and so on. In this section we explain what
we mean when we use the term "match."
There are several ways Prolog can match one thing to another. Obviously,
identical structures match each other;
parent(joe,tammy) matches parent(joe,tammy).

30 Visual Prolog
However, a match usually involves one or more free variables. For example,
with X free,
parent(joe,X) matches parent(joe,tammy)

and X takes on (is bound to) the value tammy.

If X is already bound, it acts exactly like a constant. Thus, if X is bound to the
value tammy, then
parent(joe,X) matches parent(joe,tammy) but

parent(joe,X) would not match parent(joe,millie)

The second instance doesn't match because, once a variable becomes bound, its
value can't change.
How could a variable, bindings already be bound when Prolog tries to match it
with something? Remember that variables don't store values--they only stay
bound for the length of time needed to find (or try to find) one solution to one
goal. So the only way a variable could be bound before trying a match is that the
goal involves more than one step, and the variable became bound in a previous
step. For example,
parent(joe,X), parent(X,jenny)

is a legitimate goal; it means, "Find someone who is a child of Joe and a parent of
Jenny." Here X will already be bound when the subgoal parent(X,jenny) is
reached. If there is no solution to parent(X,jenny), Prolog will unbind X and go
back and try to find another solution to parent(joe,X), then see if parent(X,jenny)
will work with the new value of X.
Two free variables can even match each other. For example,
parent(joe,X) matches parent(joe,Y)

binding the variables X and Y to each other. As long as the binding lasts, X and Y
are treated as a single variable, and if one of them gets a value the other one will
immediately have the same value. When free variables are bound to each other
like this, they're called pointers, shared free sharing variables. Some really
powerful programming techniques involve binding together variables that were
originally separate.
In Prolog, variable bindings (values) are passed in two ways: in and out. The
direction in which a value is passed is referred to as its flow pattern. When a
variable is passed into a clause, it is an input , argument, signified by (i); when
passed out of a clause, a variable is an output argument, signified by (o).

Chapter 24, Prolog Fundamentals 31

Summary

These are the ideas we've introduced in this chapter:

1. A Prolog program is made up of clauses, which conceptually are two types of
phrases: facts and rules.
Facts are relations or properties that you, the programmer, know to be true.
Rules are dependent relations; they allow Prolog to infer one piece of
information from another.
2. Facts have the general form:
property(object1, object2, ..., objectN)

or
relation(object1, object2, ..., objectN)

where a property is a property of the objects and a relation is a relation

between the objects. As far as Prolog programming goes, the distinction
doesn't exist and we will refer to both as relations in this book.
3. Each fact given in a program consists of either a relation that affects one or
more objects or a property of one or more objects. For example, in the Prolog
fact
likes(tom, baseball).

the relation is likes, and the objects are tom and baseball; Tom likes baseball.
Also, in the fact
left_handed(benjamin)

the property is left_handed and the object is benjamin; in other words,

Benjamin is left-handed.
4. Rules have the general form Head:- Body, which looks like this in a program:
relation(object,object,...,object):-
relation(object,...,object),
.
.
relation(object,...,object).

5. You are free to choose names for the relations and objects in your programs,
subject to the following constraints:

32 Visual Prolog
 Object names must begin with a lower-case letter, followed by any number
of characters; characters are upper-case or lower-case letters, digits, and
underscores.
Properties and relation names must start with a lower-case letter, followed
by any combination of letters, digits, and underscore characters.
6. A predicate is the symbolic name (identifier) for a relation and a sequence of
arguments. A Prolog program is a sequence of clauses and directives, and a
procedure is a sequence of clauses defining a predicate. Clauses that belong to
the same predicate must follow one another.
7. Variables enable you to write general facts and rules and ask general
questions.
Variable names in Visual Prolog must begin with a capital letter or an
underscore character (_), after which you can use any number of letters
(upper-case or lower-case), digits, or underscores.
Variables in Prolog get their values by being matched to constants in facts
or rules. Until it gets a value, a variable is said to be free; when it gets a
value, it becomes bound.
You can't store information globally by binding a value to a variable,
because a variable is only bound within a clause.
8. If you only need certain information from a query, you can use anonymous
variables to ignore the values you don't need. In Prolog, the anonymous
variable is represented by a lone underscore (_).
The anonymous variable can be used in place of any other variable; it matches
anything. The anonymous variable will never get set to a value.
9. Asking Prolog questions about the facts in your program is known as
querying the Prolog system; the query is commonly called a goal. Prolog tries
to satisfy a goal (answer the query) by starting at the top of the facts, looking
at each fact until it reaches the bottom.
10. A compound goal is a goal made up of two or more parts; each part of the
compound goal is called a subgoal. Compound goals can be conjunctions
(subgoal A and subgoal B) or disjunctions (subgoal A or subgoal B).
11. Comments make your programs easier to read; you can enclose a comment
with delimiters /* like this */ or precede a single-line comment with a percent
sign, % like this.
12. There are several ways Prolog can match one thing to another:
Identical structures match each other.

Chapter 24, Prolog Fundamentals 33

 A free variable matches a constant or a previously-bound variable (and
becomes bound to that value).
Two free variables can match (and be bound to) each other. As long as the
binding lasts, they are treated as a single variable; if one gets a value the
other will immediately have the same value.

34 Visual Prolog
CHAPTER 32

Visual Prolog Programs

33Visual PrologThe syntax of Visual Prolog is designed to express knowledge
about properties and relationships. You've already seen the basics of how this is
done; in Chapter 34 you learned about clauses (facts and rules), predicates,
variables, and goals.
Unlike other versions of Prolog, Visual Prolog is a typed Prolog compiler; you
declare the types of the objects that each predicate applies to. The type
declarations allow Visual Prolog programs to be compiled right down to native
machine code, giving execution speeds similar to those of compiled C and pascal.
We discuss the four basic sections of a Visual Prolog program--where you declare
and define the predicates and arguments, define rules, and specify the program's
goal--in the first part of this chapter. In the second part of this chapter we take a
closer look at declarations and rule syntax. Then, at the end of this chapter, we
briefly introduce the other sections of a Visual Prolog program, including the
facts, constants, and various global sections, and compiler directives.

Visual Prolog's Basic Program Sections

Visual PrologGenerally, a Visual Prolog program includes four basic program

sections. These are the clauses section, the predicates section, the domains
section, and the goal section.
The clauses section is the heart of a Visual Prolog program; this is where you
put the facts and rules that Visual Prolog will operate on when trying to satisfy
the program's goal.
The predicates section is where you declare your predicates and the domains
(types) of the arguments to your predicates. (You don't need to declare Visual
Prolog's built-in predicates.)
The domains section is where you declare any domains you're using that
aren't Visual Prolog's standard domains. (You don't need to declare standard
domains.)
The goal section is where you put the starting goal for a Visual Prolog
program.

Chapter 35, Visual Prolog Programs 35

The Clauses Section
The clauses section is where you put all the facts and rules that make up your
program. Most of the discussion in Chapter 36 was centered around the clauses
(facts and rules) in your programs; what they convey, how to write them, and so
on.
If you understand what facts and rules are and how to write them in Prolog, you
know what goes in the clauses section. Clauses for a given predicate must be
placed together in the clauses section; a sequence of clauses defining a predicate
is called a procedure.
When attempting to satisfy a goal, Visual Prolog will start at the top of the
clauses section, looking at each fact and rule as it searches for a match. As Visual
Prolog proceeds down through the clauses section, it places internal pointers next
to each clause that matches the current subgoal. If that clause is not part of a
logical path that leads to a solution, Visual Prolog returns to the set pointer and
looks for another match (this is backtracking, which we mentioned in Chapter
37).

The Predicates Section

If you define your own predicate in the clauses section of a Visual Prolog
program, you must declare it in a predicates section, or Visual Prolog won't
know what you're talking about. When you declare a predicate, you tell Visual
Prolog which domains the arguments of that predicate belong to.
Visual Prolog comes with a wealth of built-in predicates. You don't need to
declare any of Visual Prolog's built-in predicates that you use in your program.
The Visual Prolog on-line help gives a full explanation of the built-in predicates.
Facts and rules define predicates. The predicates section of the program simply
lists each predicate, showing the types (domains) of its arguments. Although the
clauses section is the heart of your program, Visual Prolog gets much of its
efficiency from the fact that you also declare the types of objects (arguments) that
your facts and rules refer to.

How to Declare User-Defined Predicates

A predicate declaration begins with the predicate name, followed by an open
(left) parenthesis. After the predicate name and the open parenthesis come zero or
more arguments to the predicate.
predicateName(argument_type1, argument_type2, ..., argument_typeN)

36 Visual Prolog
Each argument type is followed by a comma, and the last argument type is
followed by the closing (right) parenthesis. Note that, unlike the clauses in the
clauses section of your program, a predicate declaration is not followed by a
period. The argument types are either standard domains or domains that you've
declared in the domains section.

Predicate Names
The name of a predicate must begin with a letter, followed by a sequence of
letters, digits, and underscores. The case of the letters does not matter, but we
strongly recommend using only a lower-case letter as the first letter in the
predicate name. (Other versions of Prolog don't allow predicate names to begin
with upper-case letters or underscores, and future versions of Visual Prolog might
not, either.) Predicate names can be up to 250 characters long.
You can't use spaces, the minus sign, asterisks, slashes, or other non-
alphanumeric characters in predicate names. Valid naming charactersnaming
characters in Visual Prolog consist of the following:
Upper-case Letters : A, B, ... , Z
Lower-case Letters : a, b, ... , z
Digits : 0, 1, ... , 9
Underscore character : _

All predicate names and arguments can consist of combinations of these

characters, as long as you obey the rules for forming both predicate and argument
names.
Below are a few examples of legal and illegal predicate names.

Legal Predicate Names Illegal Predicate Names

fact [fact]
is_a *is_a*
has_a has/a
patternCheckList pattern-Check-List
choose_Menu_Item choose Menu Item
predicateName predicate<Name>
first_in_10 >first_in_10

Chapter 35, Visual Prolog Programs 37

Predicate Arguments
The arguments to the predicates must belong to known Visual Prolog domains. A
domain can be a standard domain, or it can be one you declare in the domains
section.

Examples
If you declare a predicate my_predicate(symbol, integer) in the predicates
section, like this:
PREDICATES
my_predicate(symbol, integer)

you don't need to declare its arguments' domains in a domains section,

because symbol and integer are standard domains. But if you declare a
predicate my_predicate(name, number) in the predicates section, like this:
PREDICATES
my_predicate(name, number)

you will need to declare suitable domains for name and number. Assuming
you want these to be symbol and integer respectively, the domain declaration
looks like this:
DOMAINS
name = symbol
number = integer

PREDICATES
my_predicate(name, number)

This program excerpt shows some more predicate and domain declarations:
DOMAINS
person, activity = symbol
car, make, color = symbol
mileage, years_on_road, cost = integer

PREDICATES
likes(person, activity)
parent(person, person)
can_buy(person, car)
car(make, mileage, years_on_road, color, cost)
green(symbol)
ranking(symbol, integer)

This excerpt specifies the following information about these predicates and
their arguments:

38 Visual Prolog
 The predicate likes takes two arguments (person and activity), both of
which belong to unique symbol domains (which means that their values are
names rather than numbers).
The predicate parent takes two person arguments, where person is a
symbol type.
The predicate can_buy takes two arguments, person and car, which are
also both symbol types.
The predicate car takes five arguments: make and color are of unique
symbol domains, while mileage, years_on_road, and cost are of unique
integer domains.
The predicate green takes one argument, a symbol: there is no need to
declare the argument's type, because it's of the standard domain symbol.
The predicate ranking takes two arguments, both of which belong to
standard domains (symbol and integer), so there is no need to declare the
argument types.
Chapter 38, "Simple and Compound Objects," gives more detail about domain
declarations.

The Domains Section

In traditional Prolog there is only one type; the term. We have the same in Visual
Prolog, but we are declaring what the domains of the arguments to the predicates
actually are.
Domains enable you to give distinctive names to different kinds of data that
would otherwise look alike. In a Visual Prolog program, objects in a relation (the
arguments to a predicate) belong to domains; these can be pre-defined domains,
or special domains that you specify.
The domains section serves two very useful purposes. First, you can give
meaningful names to domains even if, internally, they are the same as domains
that already exist. Second, special domain declarations are used to declare data
structures that are not defined by the standard domains.
It is sometimes useful to declare a domain when you want to clarify portions of
the predicates section. Declaring your own domains helps document the
predicates that you define by giving a useful name to the argument type.

Chapter 35, Visual Prolog Programs 39

Examples
Here's an example to illustrate how declaring domains helps to document your
predicates:
Frank is a male who is 45 years old.

With the pre-defined domains, you come up with the following predicate
declaration:
person(symbol, symbol, integer)

This declaration will work fine for most purposes. But suppose you want to
maintain your code months after you've finished writing it. The preceding
predicate declaration won't mean much to you in six months. Instead, the
following declarations will help you understand what the arguments in the
predicate declaration stand for:
DOMAINS
name, sex = symbol
age = integer

PREDICATES
person(name, sex, age)

One of the main advantages of this declarations, is that Visual Prolog can
catch type errors, like the following obvious mistake:
same_sex(X, Y) :-
person(X, Sex, _),
person(Sex, Y, _).

Even though name and sex are both defined as symbol, they are not equivalent
to each other. This enables Visual Prolog to detect an error if you accidentally
swap them. This feature is very useful when your programs get large and
complex.
You might be wondering why we don't use special domains for all argument
declarations, since special domains communicate the meaning of the argument
so much better. The answer is that once an argument is typed to a specific
domain, that domain can't be mixed with another domain you have declared,
even if the domains are the same! So, even though name and sex are of the
same domain (symbol), they can't be mixed. However, all user-defined
domains can be matched with the pre-defined domains.
This next example program will yield a type error when run:
/* Program ch039e01.pro */

DOMAINS
product,sum = integer

40 Visual Prolog
PREDICATES
add_em_up(sum,sum,sum)
multiply_em(product,product,product)

CLAUSES
add_em_up(X,Y,Sum):-
Sum=X+Y.
multiply_em(X,Y,Product):-
Product=X*Y.

This program does two things: It adds and it multiplies. Given the goal
add_em_up(32, 54, Sum).

Visual Prolog will come up with

Sum=86
1 Solution

which is the sum of the two integers you supplied the program.
On the other hand, this program will also multiply two arguments with the
multiply_em predicate. Now experiment with this program. If you need to
figure out what the product of 31 and 13 is, you could enter the goal:
multiply_em(31, 13, Product).

Visual Prolog would then respond with the correct answer.

Product=403
1 Solution

But suppose you need the sum of 42 and 17; the goal for this would be
add_em_up(42, 17, Sum).

Now you need to double the product of 31 and 17, so you write the following
goal:
multiply_em(31, 17, Sum), add_em_up(Sum, Sum, Answer).

You might expect Visual Prolog to return

Sum=527, Answer=1054
1 Solution

But, instead, you get a type error. What happened is that you tried to pass the
resulting value of multiply_em (that is, of domain product), into the first and
second arguments in add_em_up, which have domains of sum. This yields a
type error because product is a different domain than sum. Even though both
domains are really of type integer, they are different domains, and are treated
as such.

Chapter 35, Visual Prolog Programs 41

 So, if a variable is used in more than one predicate within a clause, it must be
declared the same in each predicate. Be sure that you fully understand the
concept behind the type error given here; knowing the concept will avoid
frustrating compiler error messages. Later in this chapter we will describe the
different automatic and explicit type-conversions Visual Prolog offers.
To further understand how you can use domain declarations to catch type errors,
consider the following program example:
/* Program ch040e02.pro */

DOMAINS
brand,color = symbol
age = byte
price, mileage = ulong

PREDICATES
nondeterm car(brand,mileage,age,color,price)

CLAUSES
car(chrysler,130000,3,red,12000).
car(ford,90000,4,gray,25000).
car(datsun,8000,1,black,30000).

Here, the car predicate declared in the predicates section takes five
arguments. One belongs to the age domain, which is of byte type. On the 'x86
family of CPUs, a byte is an 8-bit unsigned integer, which can take on values
between 0 and 255, both inclusive. Similarly, the domains mileage and price
are of type ulong, which is a 32-bit unsigned integer, and the domains brand
and color are of type symbol.
We'll discuss the built-in domains in greater detail in a moment. For now, load
and run this program and try each of the following goals in turn.
car(renault, 13, 40000, red, 12000).
car(ford, 90000, gray, 4, 25000).
car(1, red, 30000, 80000, datsun).

Each goal produces a domain error. In the first case, for example, it's because
age must be a byte. Hence, Visual Prolog can easily detect if someone typing
in this goal has reversed the mileage and age objects in predicate car. In the
second case, age and color have been swapped, and in the third case you get to
find out for yourself where the mixups are.

42 Visual Prolog
The Goal Section
Essentially, the goal section is the same as the body of a rule: it's simply a list of
subgoals. There are two differences between the goal section and a rule:
The goal keyword is not followed by :-.
Visual Prolog automatically executes the goal when the program runs.
It's as if Visual Prolog makes a call to goal, and the program runs, trying to
satisfy the body of the goal rule. If the subgoals in the goal section all succeed,
then the program terminates successfully. If, while the program is running, a
subgoal in the goal section fails, then the program is said to have failed.
(Although, from an external point of view, there isn't necessarily any difference;
the program simply terminates.)

A Closer Look at Declarations and Rules

Visual Prolog has several built-in standard domains. You can use standard
domains when declaring the types of a predicate's arguments. Standard domains
are already known to Visual Prolog and should not be defined in the domains
section.
We'll first look at all the integral ones, shown in Table 41.1.

Table 42.2:Integral Standard Domains433

Domain Description and implementation

short A small, signed, quantity.

All platforms 16 bits,2s comp 32768 .. 32767

ushort A small, unsigned, quantity.

All platforms 16 bits 0 .. 65535

long A large signed quantity

All platforms 32 bits,2s comp -2147483648 ..

2147483647

ulong A large, unsigned quantity

All platforms 32 bits 0 .. 4294967295

Chapter 35, Visual Prolog Programs 43

 integer A signed quantity, having the natural size for the
machine/platform architecture in question.

16bit platforms 16 bits,2s comp -32768 .. 32767

32bit platforms 32 bits,2s comp -2147483648 ..

2147483647

unsigned An unsigned quantity, having the natural size for the

machine/platform architecture in question.

16bit platforms 16 bits 0 .. 65535

32bit platforms 32 bits 0 .. 4294967295

byte
All platforms ³ 8 bits 0 .. 255

word
All platforms 16 bits 0 .. 65535

dword
All platforms 32 bits 0 .. 4294967295

Syntactically, a value belonging in one of the integral domains is written as a

sequence of digits, optionally preceded by a minus-sign for the signed domains,
with no white-space. There are also octal and hexadecimal syntaces for the
integral domains; these will be illustrated in chapter 44.
The byte, word, and dword domains are most useful when dealing with machine-
related quantities, except perhaps for the byte; an 8-bit integral quantity can prove
quite relevant, as we have already seen. For general use, the integer and unsigned
quantities are the ones to use, augmented by the short and long (and their
unsigned counterparts) for slightly more specialized applications. Generally, the
most efficient code results from using what's natural for the machine; a short is
not as efficient on a '386 platform as a long, and a long is not as efficient on a
'286 platform as a short, hence the different implementations of integer and
unsigned.
In domain declarations, the signed, and unsigned keywords may be used in
conjunction with the byte, word, and dword built-in domains to construct new
integral domains, as in
DOMAINS
i8 = signed byte

44 Visual Prolog
creating a new integral domain having a range of -128 to +127.
The other basic domains are shown in table 45.4. Visual Prolog recognizes
several other standard domains, but we cover them in other chapters, after you've
got a good grasp of the basics.

Table 46.5: Basic Standard Domains476

Domain Description and implementation

char A character, implemented as an unsigned byte.
Syntactically, it is written as a character surrounded
by single quotation marks: 'a'.

real A floating-point number, implemented as 8 bytes in

accordance with IEEE conventions; equivalent to C's
double. Syntactically, a real is written with an
optional sign (+ or -) followed by some digits DDDDDDD,
then an optional decimal point (.) followed by more
digits DDDDDDD, and an optional exponential part (e(+
or -)DDD):
<+|-> DDDDD <.> DDDDDDD <e <+|-> DDD>
Examples of real numbers:
42705 9999 86.72
9111.929437521e238 79.83e+21

Here 79.83e+21 means 79.83 x 10^21, just as in

other languages.
The permitted number range is 1 ) 10-307 to 1 )
10308 (1e-307 to 1e+308). Values from the integral
domains are automatically converted to real numbers
when necessary.

string A sequnce of characters, implemented as a pointer to a zero-

terminated byte array, as in C. Two formats are permitted for
strings:
1. a sequence of letters, numbers and underscores,
provided the first character is lower-case; or
2. a character sequence surrounded by a pair of
double quotation marks.

Examples of strings:
telephone_number "railway ticket" "Dorid Inc"

Strings that you write in the program can be up to 255

Chapter 35, Visual Prolog Programs 45

 characters long. Strings that the Visual Prolog system
reads from a file or builds up internally can be up to 64K
characters long on 16-bit platforms, and (theoretically)
up to 4G long on 32-bit platforms.

symbol A sequence of characters, implemented as a pointer to an

entry in a hashed symbol-table, containing strings. The
syntax is the same as for strings.

Symbols and strings are largely interchangeable as far as your program is

concerned, but Visual Prolog stores them differently. Symbols are kept in a look-
up table, and their addresses, rather than the symbols themselves, are stored to
represent your objects. This means that symbols can be matched very quickly,
and if a symbol occurs repeatedly in a program, it can be stored very compactly.
Strings are not kept in a look-up table; Visual Prolog examines them character-
by-character whenever they are to be matched. You must determine which
domain will give better performance in a particular program.
The following table gives some examples of simple objects that belong to the
basic standard domains.

Table 48.7: Simple Objects49

"&&", caitlin, "animal lover", b_l_t (symbol or string)

-1, 3, 5, 0 (integer)
3.45, 0.01, -30.5, 123.4e+5 (real)
'a', 'b', 'c' '/', '&' (char)

Typing Arguments in Predicate Declarations

Declaring the domain of an argument in the predicates section is called typing
the argument. For example, suppose you have the following relationship and
objects:
Frank is a male who is 45 years old.

The Prolog fact that corresponds to this natural language relation might be
person(frank, male, 45).

In order to declare person as a predicate with these three arguments, you could
place the following declaration in the predicates section:

46 Visual Prolog
 person(symbol, symbol, unsigned)

Here, you have used standard domains for all three arguments. Now, whenever
you use the predicate person, you must supply three arguments to the predicate;
the first two must be of type symbol, while the third argument must be an integer.
If your program only uses standard domains, it does not need a domains section;
you have seen several programs of this type already.
Or, suppose you want to define a predicate that will tell you the position of a
letter in the alphabet. That is,
alphabet_position(Letter, Position)

will have Position = 1 if Letter = a, Position = 2 if Letter = b, and so on. The

clauses for this predicate would look like this:
alphabet_position(A_character, N).

If standard domains are the only domains in the predicate declarations, the
program does not need a domains section. Suppose you want to define a
predicate so that the goal will be true if A_character is the Nth letter in the
alphabet. The clauses for this predicate would look like this:
alphabet_position('a', 1).
alphabet_position('b', 2).
alphabet_position('c', 3).
...
alphabet_position('z', 26).

You can declare the predicate as follows:

PREDICATES
alphabet_position(char, unsigned)

and there is no need for a domains section. If you put the whole program
together, you get
PREDICATES
alphabet_position(char, integer)

CLAUSES
alphabet_position('a', 1).
alphabet_position('b', 2).
alphabet_position('c', 3).
/* ... other letters go here ... */
alphabet_position('z', 26).

Chapter 35, Visual Prolog Programs 47

Here are a few sample goals you could enter:
alphabet_position('a', 1).

alphabet_position(X, 3).

alphabet_position('z', What).

Exercises
Program 4 is a complete Visual Prolog program that functions as a mini telephone
directorytelephone directory. The domains section is not needed here, since
only standard domains are used.
/* Program ch050e04.pro */

PREDICATES
nondeterm phone_number(symbol,symbol)

CLAUSES
phone_number("Albert","EZY-3665").
phone_number("Betty","555-5233").
phone_number("Carol","909-1010").
phone_number("Dorothy","438-8400").

Load and run the program 4, then try each of these goals in turn:
a . phone_number("Carol", Number).

b . phone_number(Who, "438-8400").

c . phone_number("Albert", Number).

d . phone_number(Who, Number).

Now update the clauses. Suppose that Kim shares a condo with Dorothy and
so has the same phone number. Add this fact to the clauses section and try the
goal
phone_number(Who, "438-8400").

You should get two solutions to this query:

Who=Dorothy
Who=Kim
2 Solutions

To illustrate the char domain, program 5 defines isletter, which, when given the
goals

48 Visual Prolog
 isletter('%').
isletter('Q').

will return No and Yes, respectively.

/* Program ch051e05.pro */

PREDICATES
nondeterm isletter(char)

CLAUSES

/* When applied to characters, '<=' means */

/* "alphabetically precedes or is the same as" */

isletter(Ch):-
'a' <= Ch,
Ch <= 'z'.
isletter(Ch):-
'A' <= Ch,
Ch <= 'Z'.

Load and run Program 5 and try each of these goals in turn:
a . isletter('x').

b . isletter('2').

c . isletter("hello").

d . isletter(a).

e . isletter(X).

Goals (c) and (d) will result in a type error message, and (e) will return an Free
variable message, because you can't test whether an unidentified object
follows a or precedes z.

Multiple Arity
The arity of a predicate is the number of arguments that it takes. You can have
two predicates with the same name but different arity. You must group different
arity versions of a given predicate name together in both the predicates and
clauses sections of your program; apart from this restriction, the different arities
are treated as completely different predicates.
/* Program ch052e06.pro */

DOMAINS
person = symbol

Chapter 35, Visual Prolog Programs 49

PREDICATES
father(person) % This person is a father
father(person, person) % One person is the father of the other person

CLAUSES
father(Man):-
father(Man,_).
father(adam,seth).
father(abraham,isaac).

Rule Syntax
Rules are used in Prolog when a fact depends upon the success (truth) of another
fact or group of facts. As we explained in Chapter 53, a Prolog rule has two parts:
the head and the body. This is the generic syntax for a Visual Prolog rule:
HEAD :- <Subgoal>, <Subgoal>, ..., <Subgoal>.

The body of the rule consists of one or more subgoals. Subgoals are separated by
commas, specifying conjunction, and the last subgoal in a rule is terminated by a
period.
Each subgoal is a call to another Prolog predicate, which may succeed or fail. In
effect, calling another predicate amounts to evaluating its subgoals, and,
depending on their success or failure, the call will succeed or fail. If the current
subgoal can be satisfied (proven true), the call returns, and processing continues
on to the next subgoal. Once the final subgoal in a rule succeeds, the call returns
successfully; if any of the subgoals fail, the rule immediately fails.
To use a rule successfully, Prolog must satisfy all of the subgoals in it, creating a
consistent set of variable bindings as it does so. If one subgoal fails, Prolog will
back up and look for alternatives to earlier subgoals, then proceed forward again
with different variable values. This is called backtracking. A full discussion of
backtracking and how Prolog finds solutions is covered in Chapter 54.

Prolog if Symbol vs. IF in Other Languages

As we have mentioned earlier, the :- separating the head and the body of a rule, is
read "if". However, a Prolog if differs from the IF written in other languages,
such as Pascal.
In Pascal, for instance, the condition contained in the IF statement must be met
before the body of the statement can be executed; in other words,
"if HEAD is true, then BODY is true (or: then do BODY)"

50 Visual Prolog
This type of statement is known as an if/then conditional. Prolog, on the other
hand, uses a different form of logic in its rules. The head of a Prolog rule is
concluded to be true if (after) the body of the rule succeeds; in other words,
"HEAD is true if BODY is true (or: if BODY can be done)"
Seen in this manner, a Prolog rule is in the form of a then/if conditional.

Automatic Type Conversions

When Visual Prolog matches two variables, it's not always necessary that they
belong to the same domain. Also, variables can sometimes be bound to constants
from other domains. This (selective) mixing is allowed because Visual Prolog
performs automatic type conversion (from one domain to another) in the
following circumstances:
· Between strings and symbols.
· Between all the integral domains, and also real. When a character is
converted to a numeric value, the number is the ASCII value for that
character.
An argument from a domain my_dom declared in this form
DOMAINS
my_dom = <base domain> /*<base domain> is a standard domain */

can mix freely with arguments from that base domain and all other standard
domains that are compatible with that base domain. (If the base domain is string,
arguments from the symbol domain are compatible; if the base domain is integer,
arguments from the real, char, word, etc., domains are compatible.
These type conversions mean, for example, that you can
call a predicate that handles strings with a symbol argument, and vice versa
call a predicate that handles reals with an integer argument
call a predicate that handles characters with integer values
use characters in expressions and comparisons without needing to look up
their ASCII values.
There are a number of rules deciding what domain the result of the expression
belongs to, when different domains are mixed. These will be detailed in chapter
55.

Chapter 35, Visual Prolog Programs 51

Other Program Sections

Now that you're reasonably familiar with the clauses, predicates, domains, and
goal sections of a Visual Prolog program, we'll tell you a little bit about some
other commonly-used program sections: the facts section, the constants section,
and the various global sections. This is just an introduction; as you work through
the rest of the tutorials in this book you'll learn more about these sections and
how to use them in your programs.

The Facts Section

A Visual Prolog program is a collection of facts and rules. Sometimes, while the
program is running, you might want to update (change, remove, or add) some of
the facts the program operates on. In such a case, the facts constitute a dynamic or
internal database; it can change while the program is running. Visual Prolog
includes a special section for declaring the facts in the program that are to be a
part of the dynamic (or changing) database; this is the facts section.
The keyword facts declares the facts section. It is here that you declare the facts
to be included in the dynamic facts section. Visual Prolog includes a number of
built-in predicates that allow easy use of the dynamic facts section. The keyword
facts is synonymous with database.
Chapter 56 provides a complete discussion of the facts section and the predicates
used along with it.

The Constants Section

You can declare and use symbolic constants in your Visual Prolog programs. A
constant declaration section is indicated by the keyword constants, followed by
the declarations themselves, using the following syntax:
<Id> = <Macro definition>

<Id> is the name of your symbolic constant, and <Macro definition> is what
you're assigning to that constant. Each <Macro definition> is terminated by a
newline character, so there can only be one constant declaration per line.
Constants declared in this way can then be referred to later in the program.
Consider the following program fragment:

52 Visual Prolog
 CONSTANTS
zero = 0
one = 1
two = 2
hundred = (10*(10-1)+10)
pi = 3.141592653
ega = 3
slash_fill = 4
red = 4

Before compiling your program, Visual Prolog will replace each constant with the
actual string to which it corresponds. For instance:
...,
A = hundred*34, delay(A),
setfillstyle(slash_fill, red),
Circumf = pi*Diam,
...

will be handled by the compiler in exactly the same way as

...,
A = (10*(10-1)+10)*34, delay(A),
setfillstyle(4, 4),
Circumf = 3.141592653*Diam,
...

There are a few restrictions on the use of symbolic constants:

The definition of a constant can't refer to itself. For example:
my_number = 2*my_number/2 /* Is not allowed */

will generate the error message Recursion in constant definition.

The system does not distinguish between upper-case and lower-case in a
constants declaration. Consequently, when a constants identifier is used in
the clauses section of a program, the first letter must be lower-case to avoid
confusing constants with variables. So, for example, the following is a valid
construction:
CONSTANTS
Two = 2

GOAL
A=two, write(A).

There can be several constants declaration sections in a program, but

constants must be declared before they are used.

Chapter 35, Visual Prolog Programs 53

 Declared constants are effective from their point of declaration to the end of
the source file, and in any files included after the declaration. Constant
identifiers can only be declared once. Multiple declarations of the same
identifier will result in the error message This constant is already defined.

The Global Sections

Visual Prolog allows you to declare some domains, predicates, and clauses in
your program to be global (rather than local); you do this by setting aside
separate global domains, global predicates, and global facts sections at the top
of your program. These global sections are discussed in the chapter 57.

The Compiler Directives

Visual Prolog provides several compiler directives you can add to your program
to tell the compiler to treat your code in specified ways when compiling. You can
also set most of the compiler directives from the Options | Project | Compiler
Options menu item in the Visual Prolog system. Compiler directives are covered
in detail in the chapter 58, but you'll want to know how to use a couple of them
before you get to that chapter, so we introduce the basic ones here.

The include Directive

As you get more familiar with using Visual Prolog, you'll probably find that you
use certain procedures over and over again in your programs. You can use the
include directive to save yourself from having to type those procedures in again
and again.
Here's an example of how you could use it:
You create a file (such as MYSTUFF.PRO) in which you declare your
frequently-used predicates (using domains and predicates sections) and give
the procedures defining those predicates in a clauses section.
You write the source text for the program that will make use of these
procedures.
At a natural boundary in your source text, you place the line
include "mystuff.pro"

(A natural boundary is anywhere in your program that you can place a

domains, facts, predicates, clauses, or goal section.)

54 Visual Prolog
 When you compile your source text, Visual Prolog will compile the contents
of MYSTUFF.PRO right into the final compiled product of your source text.
You can use the include directive to include practically any often-used text into
your source text, and one included file can in turn include another (but a given
file can only be included once in your program). The include directive can appear
at any natural boundary in your source text. However, you must observe the
restrictions on program structure when you include a file into your source text.

Summary

These are the ideas we've introduced in this chapter:

A Visual Prolog program has the following basic structure:
DOMAINS
/* ...
domain declarations
... */
PREDICATES
/* ...
predicate declarations
... */
CLAUSES
/* ...
clauses (rules and facts)
... */
GOAL
/* ...
subgoal_1,
subgoal_2,
etc. */

The clauses section is where you put the facts and rules that Visual Prolog will
operate on when trying to satisfy the program's goal.
The predicates section is where you declare your predicates and the domains
(types) of the arguments to your predicates. Predicate names must begin with a
letter (preferably lower-case), followed by a sequence of letters, digits, and
underscores, up to 250 characters long. You can't use spaces, the minus sign,
asterisks, or slashes in predicate names. Predicate declarations are of the form
PREDICATES
predicateName(argument_type1, argument_type2, ..., argument_typeN)

Chapter 35, Visual Prolog Programs 55

 argument_type1, ..., argument_typeN are either standard domains or domains
that you've declared in the domains section. Declaring the domain of an
argument and defining the argument's type are the same thing.
The domains section is where you declare any nonstandard domains you're
using for the arguments to your predicates. Domains in Prolog are like types in
other languages. Visual Prolog's basic standard domains are char, byte, short,
ushort, word, integer, unsigned, long, ulong, dword, real, string, and symbol;
the more specialized standard domains are covered in other chapters. The
basic domain declarations are of the form
DOMAINS
argument_type1, ..., argument_typeN = <standard domain>
argument_1, ..., argument_N) = <compound domain 1>;
<compound domain 2>;
< ... >;
<compound domain N>;

Compound domains haven't been covered in this chapter; you'll see them in
Chapter 59.
The goal section is where you put your program's internal goal; this allows the
program to run independent of the development environment. With an internal
goal, Visual Prolog only searches for the first solution, and the values to which
variables are bound are not displayed.
If you don't use an internal goal, you'll enter an external goal in the Dialog
window at run time. With an external goal, Visual Prolog searches for all
solutions, and displays the values to which variables are bound.
The arity of a predicate is the number of arguments that it takes; two
predicates can have the same name but different arity. You must group a
predicate's different arity versions together in both the predicates and clauses
sections, but different arities are treated as completely different predicates.
Rules are of the form
HEAD :- <Subgoal1>, <Subgoal2>, ..., <SubgoalN>.

For a rule to succeed, Prolog must satisfy all of its subgoals, creating a
consistent set of variable bindings. If one subgoal fails, Prolog backs up and
looks for alternatives to earlier subgoals, then proceeds forward with different
variable values. This is called backtracking.
The :- ("if") in Prolog should not be confused with the IF used in other
languages; a Prolog rule is in the form of a then/if conditional, while IF
statements in other languages are in the form of an if/then conditional.

56 Visual Prolog
CHAPTER 60

Unification and Backtracking

61This chapter is divided into four main parts. In the first part, we examine in
detail the process Visual Prolog uses when trying to match a call (from a subgoal)
with a clause (in the clauses section of the program). This search process includes
a procedure known as unification, which attempts to match up the data-structures
embodied in the call with those found in a given clause. In Prolog, unification
implements several of the procedures you might know from other, more
traditional languages--procedures such as parameter passing, case selection,
structure building, structure access, and assignment.
In the second part, we show you how Visual Prolog searches for solutions to a
goal (through backtracking) and how to control a search. This includes techniques
that make it possible for a program to carry out a task that would otherwise be
impossible, either because the search would take too long (which is less likely
with Visual Prolog than with other Prologs) or because the system would run out
of free memory.
In the third part of this chapter, we introduce a predicate you can use to encourage
backtracking, and go into more detail about how you can control backtracking.
We also introduce a predicate you can use to verify that a certain constraint in
your program is (or is not) met.
To shed more light on the subject, in the fourth part of this chapter we review the
more important tutorial material (presented so far) from a procedural perspective.
We show how you can understand the basic aspects of Prolog, a declarative
language, by also looking at them as procedures.

Matching Things Up: Unification

Consider Program 1 in terms of the external goal

written_by(X, Y).

When Visual Prolog tries to fulfill the goal written_by(X, Y)., it must test each
written_by clause in the program for a match. In the attempt to match the
arguments X and Y with the arguments found in each written_by clause, Visual
Prolog will search from the top of the program to the bottom. When it finds a

Chapter 62, Unification and Backtracking 57

clause that matches the goal, it binds values to free variables so that the goal and
the clause are identical; the goal is said to unify with the clause. This matching
operation is called unification.
/* Program ch063e01.pro */

/* Copyright (c) 1986, '95 by Prolog Development Center */

DOMAINS
title,author = symbol
pages = unsigned

PREDICATES
book(title, pages)
nondeterm written_by(author, title)
nondeterm long_novel(title)

CLAUSES
written_by(fleming, "DR NO").
written_by(melville, "MOBY DICK").

book("MOBY DICK", 250).

book("DR NO", 310).

long_novel(Title):-
written_by(_, Title),
book(Title, Length),
Length > 300.

Since X and Y are free variables in the goal, and a free variable can be unified
with any other argument (even another free variable), the call (goal) can be
unified with the first written_by clause in the program, as shown here:
written_by( X , Y ).
| |
written_by(fleming, "DR NO").

Visual Prolog makes a match, X becomes bound to fleming, and Y becomes bound
to "DR NO." At this point, Visual Prolog displays
X=fleming, Y=DR NO

Since Visual Prolog looks for all solutions when you use an external goal, the
goal is also unified with the second written_by clause
written_by(melville,"MOBY DICK").

and Visual Prolog displays the second solution:

58 Visual Prolog
 X=melville, Y=MOBY DICK
2 Solutions

If, on the other hand, you give the program the goal
written_by(X, "MOBY DICK").

Visual Prolog will attempt a match with the first clause for written_by:
written_by( X ,"MOBY DICK").
| |
written_by(fleming,"DR NO").

Since "MOBY DICK" and "DR NO" do not match, the attempt at unification fails.
Visual Prolog then tries the next fact in the program:
written_by(melville, "MOBY DICK").

This does unify, and X becomes bound to melville.

Consider how Visual Prolog executes the following:
long_novel(X).

When Visual Prolog tries to fulfill a goal, it investigates whether or not the call
can match a fact or the head of a rule. In this case, the match is with
long_novel(Title)

Visual Prolog looks at the clause for long_novel, trying to complete the match by
unifying the arguments. Since X is not bound in the goal, the free variable X can
be unified with any other argument. Title is also unbound in the head of the
long_novel clause. The goal matches the head of the rule and unification is made.
Visual Prolog will subsequently attempt to satisfy the subgoals to the rule.
long_novel(Title):-
written_by(_, Title),
book(Title, Length),
Length>300.

In attempting to satisfy the body of the rule, Visual Prolog will call the first
subgoal in the body of the rule, written_by(_, Title). Notice that, since who
authored the book is immaterial, the anonymous variable (_) appears in the
position of the author argument. The call written_by(_, Title) becomes the
current subgoal, and Prolog searches for a solution to this call.

Chapter 62, Unification and Backtracking 59

Prolog searches for a match with this subgoal from the top of the program to the
bottom. In doing so, it achieves unification with the first fact for written_by as
follows:
written_by(_, Title),
| |
written_by(fleming,"DR NO").

The variable Title becomes bound to "DR NO" and the next subgoal, book(Title,
Length), is called with this binding.

Visual Prolog now begins its next search, trying to find a match with the call to
book. Since Title is bound to "DR NO", the actual call resembles book("DR NO",
Length). Again, the search starts from the top of the program. Notice that the first
attempt to match with the clause book("MOBY DICK", 250) will fail, and Visual
Prolog will go on to the second clause of book in search of a match. Here, the
book title matches the subgoal and Visual Prolog binds the variable Length with
the value 310.
The third clause in the body of long_novel now becomes the current subgoal:
Length > 300.

Visual Prolog makes the comparison and succeeds; 310 is greater than 300. At
this point, all the subgoals in the body of the rule have succeeded and therefore
the call long_novel(X) succeeds. Since the X in the call was unified with the
variable Title in the rule, the value to which Title is bound when the rule succeeds
is returned to the call and unified with the variable X. Title has the value "DR NO"
when the rule succeeds, so Visual Prolog will output:
X=DR NO
1 Solution

In the following chapters, we will show several advanced examples of unification.

However, there are still a few basics that need to be introduced first, such as
complex structures. In the next section of this chapter, we'll discuss how Prolog
searches for its solutions.

Backtracking

Often, when solving real problems, you must pursue a path to its logical
conclusion. If this conclusion does not give the answer you were looking for, you
must choose an alternate path. For instance, you might have played maze games
when you were a child. One sure way to find the end of the maze was to turn left

60 Visual Prolog
at every fork in the maze until you hit a dead end. At that point you would back
up to the last fork, and try the right-hand path, once again turning left at each
branch encountered. By methodically trying each alternate path, you would
eventually find the right path and win the game.
Visual Prolog uses this same backing-up-and-trying-again method, called
backtracking, to find a solution to a given problem. As Visual Prolog begins to
look for a solution to a problem (or goal), it might have to decide between two
possible cases. It sets a marker at the branching spot (known as a backtracking
point) and selects the first subgoal to pursue. If that subgoal fails (equivalent to
reaching a dead end), Visual Prolog will backtrack to the back-tracking point and
try an alternate subgoal.
Here is a simple example:
/* Program ch064e02.pro */

/* Copyright (c) 1986, '95 by Prolog Development Center */

PREDICATES
nondeterm likes(symbol,symbol)
tastes(symbol,symbol)
nondeterm food(symbol)

CLAUSES
likes(bill,X):-
food(X),
tastes(X,good).

tastes(pizza,good).
tastes(brussels_sprouts,bad).

food(brussels_sprouts).
food(pizza).

This small program is made up of two sets of facts and one rule. The rule,
represented by the relationship likes, simply states that Bill likes good-tasting
food.
To see how backtracking works, give the program the following goal to solve:
likes(bill, What).

When Prolog begins an attempt to satisfy a goal, it starts at the top of

the program in search of a match.

Chapter 62, Unification and Backtracking 61

In this case, it will begin the search for a solution by looking from the top for a
match to the subgoal likes(bill, What).
It finds a match with the first clause in the program, and the variable What is
unified with the variable X. Matching with the head of the rule causes Visual
Prolog to attempt to satisfy that rule. In doing so, it moves on to the body of the
rule, and calls the first subgoal located there: food(X).

When a new call is made, a search for a match to that call also begins
at the top of the program.

In the search to satisfy the first subgoal, Visual Prolog starts at the top, attempting
a match with each fact or head of a rule encountered as processing goes down into
the program.
It finds a match with the call at the first fact representing the food relationship.
Here, the variable X is bound to the value brussels_sprouts. Since there is more
than one possible answer to the call food(X), Visual Prolog sets a backtracking
point next to the fact food(brussels_sprouts). This backtracking point keeps track
of where Prolog will start searching for the next possible match for food(X).

When a call has found a successful match, the call is said to succeed,
and the next subgoal in turn may be tried.

With X bound to brussels_sprouts, the next call made is

tastes(brussels_sprouts, good)

and Visual Prolog begins a search to attempt to satisfy this call, again starting
from the top of the program. Since no clause is found to match, the call fails and
Visual Prolog kicks in its automatic backtracking mechanism. When backtracking
begins, Prolog retreats to the last backtracking point set. In this case, Prolog
returns to the fact food(brussels_sprouts).

62 Visual Prolog
 Once a variable has been bound in a clause, the only way to free that
binding is through backtracking.

When Prolog retreats to a backtracking point, it frees all the variables set after
that point, and sets out to find another solution to the original call.
The call was food(X), so the binding of brussels_sprouts for X is released. Prolog
now tries to resolve this call, beginning from the place where it left off. It finds a
match with the fact food(pizza)] and returns, this time with the variable X bound
to the value pizza.
Prolog now moves on to the next subgoal in the rule, with the new variable
binding. A new call is made, tastes(pizza, good)], and the search begins at the
top of the program. This time, a match is found and the goal returns successfully.
Since the variable What in the goal is unified with the variable X in the likes rule,
and the variable X is bound to the value pizza, the variable What is now bound to
the value pizza and Visual Prolog reports the solution
What=pizza
1 Solution

Visual Prolog's Relentless Search for Solutions

As we've described earlier, with the aid of backtracking, Visual Prolog will not
only find the first solution to a problem, but is actually capable of finding all
possible solutions.
Consider Program 3, which contains facts about the names and ages of some
players in a racquet club.
/* Program ch065e03.pro */

/* Copyright (c) 1986, '95 by Prolog Development Center */

DOMAINS
child = symbol
age = integer

PREDICATES
nondeterm player(child, age)

Chapter 62, Unification and Backtracking 63

CLAUSES
player(peter,9).
player(paul,10).
player(chris,9).
player(susan,9).

You'll use Visual Prolog to arrange a ping-pong tournament between the nine-
year-olds in a racquet club. There will be two games for each pair of club players.
Your aim is to find all possible pairs of club players who are nine years old. This
can be achieved with the compound goal:
player(Person1, 9),
player(Person2, 9),
Person1 <> Person2.

In natural language: Find Person1 (age 9) and Person2 (age 9) so that Person1 is
different from Person2.
Visual Prolog will try to find a solution to the first subgoal player(Person1, 9)
and continue to the next subgoal only after the first subgoal is reached. The
first subgoal is satisfied by matching Person1 with peter. Now Visual Prolog
can attempt to satisfy the next subgoal:
player(Person2, 9)

by also matching Person2 with peter. Now Prolog comes to the third and final
subgoal
Person1 <> Person2

Since Person1 and Person2 are both bound to peter, this subgoal fails.
Because of this, Visual Prolog backtracks to the previous subgoal, and
searches for another solution to the second subgoal:
player(Person2, 9)

This subgoal is fulfilled by matching Person2 with chris.

Now, the third subgoal:
Person1 <> Person2

can succeed, since peter and chris are different. Here, the entire goal is
satisfied by creating a tournament between the two players, chris and peter.
However, since Visual Prolog must find all possible solutions to a goal, it
backtracks to the previous goal--hoping to succeed again. Since
player(Person2, 9)

64 Visual Prolog
 can also be satisfied by taking Person2 to be susan, Visual Prolog tries the
third subgoal once again. It succeeds (since peter and susan are different), so
another solution to the entire goal has been found.
Searching for more solutions, Visual Prolog once again backtracks to the
second subgoal, but all possibilities for this subgoal have been exhausted.
Because of this, backtracking now continues back to the first subgoal. This can
be satisfied again by matching Person1 with chris. The second subgoal now
succeeds by matching Person2 with peter, so the third subgoal is satisfied,
again fulfilling the entire goal. Here, another tournament has been scheduled,
this time between chris and peter.
Searching for yet another solution to the goal, Visual Prolog backtracks to the
second subgoal in the rule. Here, Person2 is matched to chris and again the
third subgoal is tried with these bindings. The third subgoal fails, since
Person1 and Person2 are equal, so backtracking regresses to the second
subgoal in search of another solution. Person2 is now matched with susan, and
the third subgoal succeeds, providing another tournament for the racket club
(chris vs. susan).
Once again, searching for all solutions, Prolog backtracks to the second
subgoal, but this time without success. When the second subgoal fails,
backtracking goes back to the first subgoal, this time finding a match for
Person1 with susan. In an attempt to fulfill the second subgoal, Prolog
matches Person2 with peter, and subsequently the third subgoal succeeds with
these bindings. A fifth tournament has been scheduled for the players.
Backtracking again goes to the second subgoal, where Person2 is matched
with chris. A sixth solution is found for the racquet club, producing a full set
of tournaments.
The final solution tried is with both Person1 and Person2 bound to susan.
Since this causes the final subgoal to fail, Visual Prolog must backtrack to the
second subgoal, but there are no new possibilities. Visual Prolog then
backtracks to the first subgoal, but the possibilities for Person1 have also been
exhausted. No more solutions can be found for the goal, so the program
terminates.

Type in this compound goal for the program:

player(Person1, 9),
player(Person2, 9),
Person1 <> Person2.

Verify that Visual Prolog responds with

Chapter 62, Unification and Backtracking 65

 Person1=peter, Person2=chris
Person1=peter, Person2=susan
Person1=chris, Person2=peter
Person1=chris, Person2=susan
Person1=susan, Person2=peter
Person1=susan, Person2=chris
6 Solutions

Notice how backtracking might cause Visual Prolog to come up with redundant
solutions. In this example, Visual Prolog does not distinguish that Person1 =
peter is the same thing as Person2 = peter. We will show you later in this chapter
how to control the search Visual Prolog generates.

Exercise in Backtracking
Using Program 3, decide what Visual Prolog will reply to the following goal:
player(Person1, 9), player(Person2, 10).

Check your answer by typing in the exercise and the given goal when you run the
program.

A Detailed Look at Backtracking

With this simple example under your belt, you can take a more detailed look at
how Visual Prolog's backtracking mechanism works. Start by looking at Program
4 in light of the following goal, which consists of two subgoals:
likes(X, wine) , likes(X, books)

When evaluating the goal, Visual Prolog notes which subgoals have been
satisfied and which have not. This search can be represented by a goal tree:

Before the goal evaluation begins, the goal tree consists of two unsatisfied
subgoals. In the following goal tree diagrams, a subgoal satisfied in the goal tree
is marked with an underline, and the corresponding clause is shown beneath that
subgoal.
/* Program ch066e04.pro */

66 Visual Prolog
 /* Copyright (c) 1986, '95 by Prolog Development Center */

DOMAINS
name,thing = symbol

PREDICATES
likes(name, thing)
reads(name)
is_inquisitive(name)

CLAUSES
likes(john,wine):-!.
likes(lance,skiing):-!.
likes(lance,books):-!.
likes(lance,films):-!.
likes(Z,books):-
reads(Z),
is_inquisitive(Z).

reads(john).

is_inquisitive(john).

The Four Basic Principles of Backtracking

In this example, the goal tree shows that two subgoals must be satisfied. To do so,
Visual Prolog follows the first basic principle of backtracking:

Subgoals must be satisfied in order, from top to bottom.

Visual Prolog determines which subgoal it will use when trying to satisfy the
clause according to the second basic principle of backtracking:

Predicate clauses are tested in the order they appear in the program,
from top to bottom.

When executing Program 4, Visual Prolog finds a matching clause with the first
fact defining the likes predicate. Take a look at the goal tree now.

Chapter 62, Unification and Backtracking 67

The subgoal likes(X, wine) matches the fact likes(john, wine) and binds X to
the value john. Visual Prolog tries to satisfy the next subgoal to the right.
The call to the second subgoal begins a completely new search with the binding X
= john. The first clause
likes(john, wine)

does not match the subgoal

likes(X, books)

since wine is not the same as books. Visual Prolog must therefore try the next
clause, but lance does not match the value X (because, in this case, X is bound to
john), so the search continues to the third clause defining the predicate likes:
likes(Z, books):- reads(Z), is_inquisitive(Z).

The argument Z is a variable, so it is able to match with X. The second arguments

agree, so the call matches the head of the rule. When X matches Z, the arguments
are unified. With the arguments unified, Visual Prolog will equate the value X has
(which is john) with the variable Z. Because of this, now the variable Z also has
the value john.
The subgoal now matches the left side (head) of a rule. Continued searching is
determined by the third basic principle of backtracking:

When a subgoal matches the head of a rule, the body of that rule must
be satisfied next. The body of the rule then constitutes a new set of
subgoals to be satisfied.

This yields the following goal tree:

68 Visual Prolog
The goal tree now includes the subgoals
reads(Z) and is_inquisitive(Z)

where Z is bound to the value john. Visual Prolog will now search for facts that
match both subgoals. This is the resulting final goal tree:

According to the fourth basic principle of backtracking:

A goal has been satisfied when a matching fact is found for each of the
extremities (leaves) of the goal tree.

So now the initial goal is satisfied.

Chapter 62, Unification and Backtracking 69

Visual Prolog uses the result of the search procedure in different ways, depending
on how the search was initiated. If the goal is a call from a subgoal in the body of
a rule, Visual Prolog attempts to satisfy the next subgoal in the rule after the call
has returned. If the goal is a query from the user, Visual Prolog replies directly:
X=john
1 Solution

As you saw in Program 4, having once satisfied an external goal, Visual Prolog
backtracks to find all alternate solutions. It also backtracks if a subgoal fails,
hoping to re-satisfy a previous subgoal in such a way that the failed subgoal is
satisfied by other clauses.
To fulfill a subgoal, Visual Prolog begins a search with the first clause that
defines the predicate. One of two things can then happen:
It finds a matching clause, in which case the following occurs:
If there is another clause that can possibly re-satisfy the subgoal, Visual
Prolog places a pointer (to indicate a backtracking point) next to the
matching clause.
All free variables in the subgoal that match values in the clause are bound to
the corresponding values.
If the matching clause is the head of a rule, that rule's body is then evaluated;
the body's subgoals must succeed for the call to succeed.
It can't find a matching clause, so the goal fails. Visual Prolog backtracks as it
attempts to re-satisfy a previous subgoal. When processing reaches the last
backtracking point, Visual Prolog frees all variables that had been assigned
new values since the backtracking point was set, then attempts to re-satisfy the
original call.
Visual Prolog begins a search from the top of the program. When it backtracks to
a call, the new search begins from the last backtracking point set. If the search is
unsuccessful, it backtracks again. If backtracking exhausts all clauses for all
subgoals, the goal fails.

Backtracking
Here is another, slightly more complex, example, illustrating how backtracking
takes place in Prolog.
/* Program ch067e05.pro */

/* Copyright (c) 1986, '95 by Prolog Development Center */

70 Visual Prolog
PREDICATES
nondeterm type(symbol, symbol)
nondeterm is_a(symbol, symbol)
lives(symbol, symbol)
nondeterm can_swim(symbol)

CLAUSES
type(ungulate,animal).
type(fish,animal).

is_a(zebra,ungulate).
is_a(herring,fish).
is_a(shark,fish).

lives(zebra,on_land).
lives(frog,on_land).
lives(frog,in_water).
lives(shark,in_water).

can_swim(Y):-
type(X,animal),
is_a(Y,X),
lives(Y,in_water).

This example program uses an internal goal to illustrate how backtracking works.
When the program is compiled and run, Visual Prolog will automatically begin
executing the goal, attempting to satisfy all the subgoals in the goal section.
Visual Prolog calls the can_swim predicate with a free variable, What. In
trying to solve this call, Visual Prolog searches the program looking for a
match. It finds a match with the clause defining can_swim, and the variable
What is unified with the variable Y.
Next, Visual Prolog attempts to satisfy the body of the rule. In doing so,
Visual Prolog calls the first subgoal in the body of the rule, type(X, animal),
and searches for a match to this call. It finds a match with the first fact
defining the type relationship.
At this point, X is bound to ungulate. Since there is more than one possible
solution, Visual Prolog sets a backtracking point at the fact type(ungulate,
animal).

With X bound to ungulate, Visual Prolog makes a call to the second subgoal in
the rule (is_a(Y, ungulate)), and again searches for a match. It finds one with
the first fact, is_a(zebra, ungulate). Y is bound to zebra and Prolog sets a
backtracking point at is_a(zebra, ungulate).
Now, with X bound to ungulate and Y bound to zebra, Prolog tries to satisfy
the last subgoal, lives(zebra, in_water). Prolog tries each lives clause, but

Chapter 62, Unification and Backtracking 71

 there is no lives(zebra, in_water) clause in the program, so the call fails and
Prolog begins to backtrack in search of another solution.
When Visual Prolog backtracks, processing returns to the last point where a
backtracking point was placed. In this case, the last backtracking point was
placed at the second subgoal in the rule, on the fact is_a(zebra, ungulate).
When Visual Prolog reaches a backtracking point, it frees the variables that
were assigned new values after the last backtracking point and attempts to find
another solution to the call it made at that time. In this case, the call was
is_a(Y, ungulate).

Visual Prolog continues down into the clauses in search of another clause that
will match with this one, starting from the point where it previously left off.
Since there are no other clauses in the program that can match this one, the
call fails and Visual Prolog backtracks again in an attempt to solve the original
goal.
From this position, the last backtracking point was set at type(ungulate,
animal).

Visual Prolog frees the variables set in the original call and tries to find
another solution to the call type(X, animal). The search begins after the
backtracking point. Visual Prolog finds a match with the next type fact in the
program (type(fish, animal)); X is bound to fish, and a new backtracking
point is set at that fact.
Visual Prolog now moves down to the next subgoal in the rule; since this is a
new call, the search begins at the top of the program with is_a(Y, fish).
Visual Prolog finds a match to this call and Y is bound to herring.
Since Y is now bound to herring, the next subgoal called is lives(herring,
in_water). Again, this is a new call, and the search begins from the top of the
program.
Visual Prolog tries each lives fact, but fails to find a match and the subgoal
fails.
Visual Prolog now returns to the last backtracking point, is_a(herring, fish).

The variables that were bound by this matching are now freed. Starting at the
point where it last left off, Visual Prolog now searches for a new solution to
the call is_a(Y, fish).
Visual Prolog finds a match with the next is_a clause, and Y becomes bound to
the symbol shark.
Visual Prolog tries the last subgoal again, with the variable Y bound to shark.
It calls lives(shark, in_water); the search begins at the top of the program,

72 Visual Prolog
 since this is a new call. It finds a match and the last subgoal to the rule
succeeds.
At this point, the body of the can_swim(Y) rule is satisfied. Visual Prolog
returns Y to the call can_swim(What). Since What is bound to Y, and Y is bound
to shark, What is now bound to shark in the goal.
Visual Prolog continues processing where it left off in the goal section, and
calls the second subgoal in the goal.
Visual Prolog completes the program by outputting
A shark can swim.

and the program terminates successfully.

Chapter 62, Unification and Backtracking 73

 Figure 68.1: How the can_swim Program Works69

Controlling the Search for Solutions

Prolog's built-in backtracking mechanism can result in unnecessary searching;
because of this, inefficiencies can arise. For instance, there may be times when
you want to find unique solutions to a given question. In other cases, it may be
necessary to force Visual Prolog to continue looking for additional solutions even
though a particular goal has been satisfied. In cases such as these, you must

74 Visual Prolog
control the backtracking process. In this section, we'll show you some techniques
you can use to control Visual Prolog's search for the solutions to your goals.
Visual Prolog provides two tools that allow you to control the backtracking
mechanism: the fail predicate, which is used to force backtracking, and the cut
(signified by !), which is used to prevent backtracking.

Using the fail Predicate

Visual Prolog begins backtracking when a call fails. In certain situations, it's
necessary to force backtracking in order to find alternate solutions. Visual Prolog
provides a special predicate, fail, to force failure and thereby encourage
backtracking. The effect of the fail predicate corresponds to the effect of the
comparison 2 = 3 or any other impossible subgoal. Program 6 illustrates the use
of this special predicate.
/* Program ch070e06.pro */

/* Copyright (c) 1986, '95 by Prolog Development Center */

DOMAINS
name = symbol

PREDICATES
nondeterm father(name, name)
everybody

CLAUSES
father(leonard,katherine).
father(carl,jason).
father(carl,marilyn).
everybody:-
father(X,Y),
write(X," is ",Y,"'s father\n"),
fail.
everybody.

Once an internal goal has completely succeeded, there is nothing that tells Visual
Prolog to backtrack. Because of this, an internal call to father will come up with
only one solution. However, the predicate everybody in Program 6 uses fail to
force backtracking, and therefore finds all possible solutions.
The object of the predicate everybody is to produce a cleaner response from
program runs. Compare the answers to the two preceding goals:
Goal father(X, Y).

Chapter 62, Unification and Backtracking 75

 X=leonard, Y=katherine
X=carl, Y=jason
X=carl, Y=marilyn

3 Solutions

and
Goal everybody.

leonard is katherine's father

carl is jason's father
carl is marilyn's father
yes

The predicate everybody uses backtracking to generate more solutions for

father(X, Y) by forcing Prolog to backtrack through the body of the everybody
rule:
father(X, Y), write(X," is ",Y,"'s father\n"), fail.

fail can never be satisfied (it always fails), so Visual Prolog is forced to
backtrack. When backtracking takes place, Prolog backtracks to the last call that
can produce multiple solutions. Such a call is labeled non-deterministic. A non-
deterministic call contrasts with a call that can produce only one solution,
which is a deterministic call.
The write predicate can't be re-satisfied (it can't offer new solutions), so Visual
Prolog must backtrack again, this time to the first subgoal in the rule.
Notice that it's useless to place a subgoal after fail in the body of a rule. Since the
predicate fail always fails, there would be no way of reaching a subgoal located
after fail.

Exercises
Load and run Program 6 and evaluate the following goals:
a. father(X, Y).

b. everybody.

2. Edit the body of the rule defining everybody so that the rule ends with the call
to the write predicate (delete the call to fail). Now compile and run the
program, giving everybody. as the goal. Why doesn't Visual Prolog find all the
solutions as it does with the query father(X, Y)?

76 Visual Prolog
3. Relocate the call to fail at the end of the everybody rule. Again, give the
query everybody as the goal. Why are the solutions to everybody terminated
by no? For a clue, append everybody. as a second clause to the definition of
predicate everybody and re-evaluate the goal.

Preventing Backtracking: The Cut

Visual Prolog contains the cut, which is used to prevent backtracking; it's written
as an exclamation mark (!). The effect of the cut is simple: It is impossible to
backtrack across a cut.
You place the cut in your program the same way you place a subgoal in the body
of a rule. When processing comes across the cut, the call to cut immediately
succeeds, and the next subgoal (if there is one) is called. Once a cut has been
passed, it is not possible to backtrack to subgoals placed before the cut in the
clause being processed, and it is not possible to backtrack to other predicates
defining the predicate currently in process (the predicate containing the cut).
There are two main uses of the cut:
When you know in advance that certain possibilities will never give rise to
meaningful solutions, it's a waste of time and storage space to look for
alternate solutions. If you use a cut in this situation, your resulting program
will run quicker and use less memory. This is called a green cut.
When the logic of a program demands the cut, to prevent consideration of
alternate subgoals. This is a red cut.

How to Use the Cut

In this section, we give examples that show how you can use the cut in your
programs. In these examples, we use several schematic Visual Prolog rules (r1,
r2, and r3), which all describe the same predicate r, plus several subgoals (a, b, c,
etc.).

Prevent Backtracking to a Previous Subgoal in a Rule

r1 :- a, b, !, c.

This is a way of telling Visual Prolog that you are satisfied with the first solution
it finds to the subgoals a and b. Although Visual Prolog is able to find multiple
solutions to the call to c through backtracking, it is not allowed to backtrack
across the cut to find an alternate solution to the calls a or b. It is also not allowed
to backtrack to another clause that defines the predicate r1.
As a concrete example, consider Program 7.

Chapter 62, Unification and Backtracking 77

 /* Program ch071e07.pro */

/* Copyright (c) 1986, '95 by Prolog Development Center */

PREDICATES
buy_car(symbol,symbol)
nondeterm car(symbol,symbol,integer)
colors(symbol,symbol)

CLAUSES
buy_car(Model,Color):-
car(Model,Color,Price),
colors(Color,sexy),!,
Price > 25000.

car(maserati,green,25000).
car(corvette,black,24000).
car(corvette,red,26000).
car(porsche,red,24000).

colors(red,sexy).
colors(black,mean).
colors(green,preppy).

In this example, the goal is to find a Corvette with a sexy color and a price that's
ostensibly affordable. The cut in the buy_car rule means that, since there is only
one Corvette with a sexy color in the database, if its price is too high there's no
need to search for another car.
Given the goal
buy_car(corvette, Y)

Visual Prolog calls car, the first subgoal to the buy_car predicate.
It makes a test on the first car, the Maserati, which fails.
It then tests the next car clauses and finds a match, binding the variable Color
with the value black.
It proceeds to the next call and tests to see whether the car chosen has a sexy
color. Black is not a sexy color in the program, so the test fails.
Visual Prolog backtracks to the call to car and once again looks for a Corvette
to meet the criteria.
It finds a match and again tests the color. This time the color is sexy, and
Visual Prolog proceeds to the next subgoal in the rule: the cut. The cut
immediately succeeds and effectively "freezes into place" the variable
bindings previously made in this clause.

78 Visual Prolog
 Visual Prolog now proceeds to the next (and final) subgoal in the rule: the
comparison
Price < 25000.

This test fails, and Visual Prolog attempts to backtrack in order to find another
car to test. Since the cut prevents backtracking, there is no other way to solve
the final subgoal, and the goal terminates in failure.

Prevent Backtracking to the Next Clause

The cut can be used as a way to tell Visual Prolog that it has chosen the correct
clause for a particular predicate. For example, consider the following code:
r(1):- ! , a , b , c.
r(2):- ! , d.
r(3):- ! , c.
r(_):- write("This is a catchall clause.").

Using the cut makes the predicate r deterministic. Here, Visual Prolog calls r with
a single integer argument. Assume that the call is r(1). Visual Prolog searches the
program, looking for a match to the call; it finds one with the first clause defining
r. Since there is more than one possible solution to the call, Visual Prolog places a
backtracking point next to this clause.
Now the rule fires and Visual Prolog begins to process the body of the rule. The
first thing that happens is that it passes the cut; doing so eliminates the possibility
of backtracking to another r clause. This eliminates backtracking points,
increasing the run-time efficiency. It also ensures that the error-trapping clause is
executed only if none of the other conditions match the call to r.
Note that this type of structure is much like a "case" structure written in other
programming languages. Also notice that the test condition is coded into the head
of the rules. You could just as easily write the clauses like this:
r(X) :- X = 1 , ! , a , b , c.
r(X) :- X = 2 , ! , d.
r(X) :- X = 3 , ! , c.
r(_) :- write("This is a catchall clause.").

However, you should place the testing condition in the head of the rule as much
as possible, as doing this adds efficiency to the program and makes for easier
reading.
As another example, consider the following program. Run this program and give
the query friend(bill, Who) as the goal.

Chapter 62, Unification and Backtracking 79

 /* Program ch072e08.pro */

/* Copyright (c) 1986, '95 by Prolog Development Center */

PREDICATES
friend(symbol,symbol)
girl(symbol)
likes(symbol,symbol)

CLAUSES
friend(bill,jane):-
girl(jane),
likes(bill,jane),!.
friend(bill,jim):-
likes(jim,baseball),!.
friend(bill,sue):-
girl(sue).

girl(mary).
girl(jane).
girl(sue).

likes(jim,baseball).
likes(bill,sue).

Without cuts in the program, Visual Prolog would come up with two solutions:
Bill is a friend of both Jane and Sue. However, the cut in the first clause defining
friend tells Visual Prolog that, if this clause is satisfied, it has found a friend of
Bill and there's no need to continue searching for more friends. A cut of this type
says, in effect, that you are satisfied with the solution found and that there is no
reason to continue searching for another friend.
Backtracking can take place inside the clauses, in an attempt to satisfy the call,
but once a solution is found, Visual Prolog passes a cut. The friend clauses,
written as such, will return one and only one friend of Bill's (given that a friend
can be found).

Determinism and the Cut

If the friend predicate (defined in the previous program) were coded without the
cuts, it would be a non-deterministic predicate (one capable of generating
multiple solutions through backtracking). In many implementations of Prolog,
programmers must take special care with non-deterministic clauses because of the
attendant demands made on memory resources at run time. However, Visual
Prolog makes internal checks for non-deterministic clauses, reducing the burden
on you, the programmer.

80 Visual Prolog
However, for debugging (and other) purposes, it can still be necessary for you to
intercede; the check_determ compiler directive is provided for this reason. If
check_determ is inserted at the very beginning of a program, Visual Prolog will
display a warning if it encounters any non-deterministic clauses during
compilation.
You can make non-deterministic clauses into deterministic clauses by inserting
cuts into the body of the rules defining the predicate. For example, placing cuts in
the clauses defining the friend predicate causes that predicate to be deterministic
because, with the cuts in place, a call to friend can return one, and only one,
solution.

The not Predicate

This program demonstrates how you can use the not predicate to identify an
honor student: one whose grade point average (GPA) is at least 3.5 and who is not
on probation.
/* Program ch073e10.pro */

/* Copyright (c) 1986, '95 by Prolog Development Center */

DOMAINS
name = symbol
gpa = real

PREDICATES
nondeterm honor_student(name)
nondeterm student(name, gpa)
probation(name)

CLAUSES
honor_student(Name):-
student(Name, GPA),
GPA>=3.5,
not(probation(Name)).

student("Betty Blue", 3.5).

student("David Smith", 2.0).
student("John Johnson", 3.7).

probation("Betty Blue").
probation("David Smith").

There is one thing to note when using not: The not predicate succeeds when the
subgoal can't be proven true. This results in a situation that prevents unbound
variables from being bound within a not. When a subgoal with free variables is
called from within not, Visual Prolog will return the error message Free

Chapter 62, Unification and Backtracking 81

variables not allowed in 'not' or 'retractall'. This happens because, for
Prolog to bind the free variables in a subgoal, that subgoal must unify with some
other clause and the subgoal must succeed. The correct way to handle unbound
variables within a not subgoal is with anonymous variables.
Here are some examples of correct clauses and incorrect clauses.
likes(bill, Anyone):- /* 'Anyone' is an output argument */
likes(sue, Anyone),
not(hates(bill, Anyone).

In this example, Anyone is bound by likes(sue, Anyone) before Visual Prolog

finds out that hates(bill, Anyone) is not true. This clause works just as it should.
If you rewrite this so that it calls not first, you will get an error message to the
effect that free variables are not allowed in not.
likes(bill, Anyone):- /* This won't work right */
not(hates(bill, Anyone)),
likes(sue, Anyone).

Even if you correct this (by replacing Anyone in not(hates(bill, Anyone)) with
an anonymous variable) so that the clause does not return the error, it will still
return the wrong result.
likes(bill, Anyone):- /* This won't work right */
not(hates(bill, _)),
likes(sue, Anyone).

This clause states that Bill likes Anyone if nothing that Bill hates is known and if
Sue likes Anyone. The original clause stated that Bill likes Anyone if there is some
Anyone that Sue likes and that Bill does not hate.

Example
Always be sure that you think twice when using the not predicate. Incorrect use
will result in an error message or errors in your program's logic. The following is
an example of the proper way to use the not predicate.
/* Program ch074e11.pro */

/* Copyright (c) 1986, '95 by Prolog Development Center */

PREDICATES
nondeterm likes_shopping(symbol)
nondeterm has_credit_card(symbol,symbol)
bottomed_out(symbol,symbol)

82 Visual Prolog
CLAUSES
likes_shopping(Who):-
has_credit_card(Who,Card),
not(bottomed_out(Who,Card)),
write(Who," can shop with the ",Card, " credit card.\n").

has_credit_card(chris,visa).
has_credit_card(chris,diners).
has_credit_card(joe,shell).
has_credit_card(sam,mastercard).
has_credit_card(sam,citibank).

bottomed_out(chris,diners).
bottomed_out(sam,mastercard).
bottomed_out(chris,visa).

Give the following at the goal

likes_shopping(Who).

Exercises
Suppose an average taxpayer in the USA is a married US citizen with two
children who earns no less than $500 a month and no more than $2,000 per
month. Define a special_taxpayer predicate that, given the goal
special_taxpayer(fred)., will succeed only if fred fails one of the conditions
for an average taxpayer. Use the cut to ensure that there is no unnecessary
backtracking.
Players in a certain squash club are divided into three leagues, and players
may only challenge members in their own league or the league below (if there
is one).
Write a Visual Prolog program that will display all possible matches between
club players in the form:
tom versus bill
marjory versus annette

Use the cut to ensure, for example, that

tom versus bill

and
bill versus tom

are not both displayed.

Chapter 62, Unification and Backtracking 83

 This is an exercise in backtracking, not a test of your ability to solve murder
mysteries. Load and run the following program, then enter the goal killer(X).
(Note: Bert is guilty because he has a motive and is smeared in the same stuff
as the victim.)
/* Program ch075e12.pro */

/* Copyright (c) 1986, '95 by Prolog Development Center */

DOMAINS
name,sex,occupation,object,vice,substance = symbol
age=integer

PREDICATES
nondeterm person(name, age, sex, occupation)
nondeterm had_affair(name, name)
killed_with(name, object)
killed(name)
nondeterm killer(name)
motive(vice)
smeared_in(name, substance)
owns(name, object)
nondeterm operates_identically(object, object)
nondeterm owns_probably(name, object)
nondeterm suspect(name)

/* * * Facts about the murder * * */

CLAUSES
person(bert,55,m,carpenter).
person(allan,25,m,football_player).
person(allan,25,m,butcher).
person(john,25,m,pickpocket).

had_affair(barbara,john).
had_affair(barbara,bert).
had_affair(susan,john).

killed_with(susan,club).
killed(susan).

motive(money).
motive(jealousy).
motive(righteousness).

84 Visual Prolog
 smeared_in(bert, blood).
smeared_in(susan, blood).
smeared_in(allan, mud).
smeared_in(john, chocolate).
smeared_in(barbara,chocolate).

owns(bert,wooden_leg).
owns(john,pistol).

/* * * Background knowledge * * */

operates_identically(wooden_leg, club).
operates_identically(bar, club).
operates_identically(pair_of_scissors, knife).
operates_identically(football_boot, club).

owns_probably(X,football_boot):-
person(X,_,_,football_player).
owns_probably(X,pair_of_scissors):-
person(X,_,_,hairdresser).
owns_probably(X,Object):-
owns(X,Object).

/* * * * * * * * * * * * * * * * * * * * * * *
* Suspect all those who own a weapon with *
* which Susan could have been killed. *
* * * * * * * * * * * * * * * * * * * * * * */

suspect(X):-
killed_with(susan,Weapon) ,
operates_identically(Object,Weapon) ,
owns_probably(X,Object).

/* * * * * * * * * * * * * * * * * * * * * * * * * *
* Suspect men who have had an affair with Susan. *
* * * * * * * * * * * * * * * * * * * * * * * * * */

suspect(X):-
motive(jealousy),
person(X,_,m,_),
had_affair(susan,X).

/* * * * * * * * * * * * * * * * * * * * *
* Suspect females who have had an *
* affair with someone that Susan knew. *
* * * * * * * * * * * * * * * * * * * * */

Chapter 62, Unification and Backtracking 85

 suspect(X):-
motive(jealousy),
person(X,_,f,_),
had_affair(X,Man),
had_affair(susan,Man).

/* * * * * * * * * * * * * * * * * * * * * * * * * * *
* Suspect pickpockets whose motive could be money. *
* * * * * * * * * * * * * * * * * * * * * * * * * * */

suspect(X):-
motive(money),
person(X,_,_,pickpocket).

killer(Killer):-
person(Killer,_,_,_),
killed(Killed),
Killed <> Killer, /* It is not a suicide */
suspect(Killer),
smeared_in(Killer,Goo),
smeared_in(Killed,Goo).

Prolog from a Procedural Perspective

Now that you've read chapters 76, 77, and the first three parts of this chapter, you
should have a pretty good understanding of the basics of Prolog programming and
using Visual Prolog. Remember, Prolog is a declarative language, which means
that you describe a problem in terms of facts and rules and let the computer figure
out how to find a solution. Other programming languages--such as Pascal,
BASIC, and C--are procedural, which means that you must write subroutines and
functions that tell the computer exactly what steps to go through in order to solve
the problem.
We're going to back up now and review of some of the material you've just
learned about Prolog, but this time we're going to present it from a procedural
perspective.

How Rules and Facts Are Like Procedures

It's easy to think of a Prolog rule as a procedure definition. For instance, the rule
likes(bill,Something):- likes(cindy,Something).

means,

86 Visual Prolog
 "To prove that Bill likes something, prove that Cindy likes it."
With this in mind, you can see how procedures like
say_hello:- write("Hello"), nl.

and
greet:-
write("Hello, Earthlings!"),
nl.

correspond to subroutines and functions in other programming languages.

You can even think of Prolog facts of as procedures; for instance, the fact
likes(bill, pasta).

means
"To prove that Bill likes pasta, do nothing--and by the way, if the arguments
Who and What in your query likes(Who, What) are free variables, you can bind
them to bill and pasta, respectively."
Some programming procedures that you might be familiar with from other
languages are case statements, boolean tests, goto statements, and computational
returns. In the next sections, by reiterating what we've already covered from a
different (procedural) point of view, we'll show you how Prolog rules can
perform these same functions.

Using Rules Like Case Statements

One big difference between rules in Prolog and procedures in other languages is
that Prolog allows you to give multiple alternative definitions of the same
procedure. This came up with the "parent" program earlier on page 25; a person
can be a parent by being a father or by being a mother, so the definition of
"parent" is made up of two rules.
You can use multiple definitions like you use a Pascal case statement by writing a
different definition for each argument value (or set of argument values). Prolog
will try one rule after another until it finds a rule that matches, then perform the
actions that rule specifies, as in Program 13.
/* Program ch078e13.pro */

/* Copyright (c) 1986, '95 by Prolog Development Center */

PREDICATES
nondeterm action(integer)

Chapter 62, Unification and Backtracking 87

CLAUSES
action(1):-
nl,
write("You typed 1."),nl.
action(2):-
nl,
write("You typed two."),nl.
action(3):-
nl,
write("Three was what you typed."),nl.
action(N):-
nl,
N<>1, N<>2, N<>3,
write("I don't know that number!").

GOAL
write("Type a number from 1 to 3: "),
readint(Num),
action(Num).

If the user types 1, 2, or 3, action will be called with its argument bound to the
appropriate value, and it will match only one of the first three rules.

Performing Tests within the Rule

Look more closely at the fourth clause for action. It will match whatever
argument it's called with, binding X to that value. So you have to make sure that it
doesn't print I don't know that number unless the number is indeed out of range.
That's the purpose of the subgoals
X<>1, X<>2, X<>3

where <> means not equal. In order to print I don't know that number, Prolog
must first prove that X is not 1, 2, or 3. If any of these subgoals fail, Prolog will
try to back up and find alternatives--but there aren't any alternatives, so the rest of
the clause will never be executed.
Notice that action relies on Choice being bound. If you call action with a free
variable as an argument, the goal would match all of the clauses. The first three
would return alternative solutions, and then the last one would raise an error
because you can't test whether an unbound variable is not equal to a number.

88 Visual Prolog
The Cut as a GoTo
Program 13 is somewhat wasteful because, after choosing and executing the
correct rule, Prolog still keeps looking for alternatives and has to find out the hard
way that the last rule doesn't apply.
It would save time and memory if you could tell Prolog to stop looking for
alternatives. And you can, by using the cut, which means,
"If you get this far, don't do any backtracking within this rule, and don't look
for any alternatives to this rule."
In other words, "Burn your bridges behind you." Backtracking is still possible,
but only at a higher level. If the current rule was called by another rule, and the
higher rule has alternatives, they can still be tried. But the cut rules out
alternatives within, and alternatives to, the present rule.
Using cuts, the program can be rewritten as follows:
/* Program ch079e14.pro */

/* Copyright (c) 1986, '95 by Prolog Development Center */

PREDICATES
action(integer)

CLAUSES
action(1):-!,
nl,
write("You typed 1.").
action(2):-!,
nl,
write("You typed two.").
action(3):-!,
nl,
write("Three was what you typed.").
action(_):-
write("I don't know that number!").

GOAL
write("Type a number from 1 to 3: "),
readint(Num),
action(Num),nl.

The cut has no effect unless it is actually executed. That is, in order to perform a
cut, Prolog must actually get into the rule containing the cut and reach the point
where the cut is located.
The cut can be preceded by other tests, like this:

Chapter 62, Unification and Backtracking 89

 action(X) :- X>3, !, write("Too high.").

In this rule, the cut won't have any effect unless the subgoal X>3 succeeds first.
Notice that the order of the rules is now significant. In 13, you could have written
the rules in any order; only one of them will match any particular number. But in
Program 14 you must make sure that the computer doesn't even try the rule that
prints I don't know that number unless all of the preceding rules have been tried
(and have not executed their cuts).
The cuts in 14 are what some people call red cuts--cuts that change the logic of
the program. If you had kept the tests X<>1, X<>2, and X<>3, changing the program
only by inserting a cut in each clause, you would have been using green cuts--
cuts that save time in a program that would be equally correct without them. The
efficiency gained is not as great, but there is less risk of making an error in the
program.
The cut is a powerful, but messy, Prolog operation. In this respect it resembles
the goto statement in other programming languages--you can do many things with
it, but it can make your program really hard to understand.

Returning Computed Values

As we have seen, a Prolog rule or fact can return information to the goal that
called it. This is done by binding arguments that were previously unbound. The
fact
likes(bill, cindy).

returns information to the goal

likes(bill, Who).

by binding Who to cindy.

A rule can return the results of a computation the same way. Here's a simple
example:
/* Program ch080e15.pro */

/* Copyright (c) 1986, '95 by Prolog Development Center */

PREDICATES
nondeterm classify(integer,symbol)

90 Visual Prolog
CLAUSES
classify(0,zero).
classify(X,negative):-
X < 0.
classify(X,positive):-
X > 0.

The first argument of classify must always be either a constant or a bound

variable. The second argument can be either bound or unbound; it gets matched
with the symbol zero, negative, or positive, depending on the value of the first
argument.
Here are some examples of how rules can return values:
You can ask whether 45 is positive by giving the goal:
Goal classify(45, positive).

yes

Because 45 is greater than 0, only the third clause of classify can succeed. In
doing so, it matches the second argument with positive. But the second
argument is already positive, so the match succeeds, and you get the answer
yes.

Conversely, if the match fails, you get no:

Goal classify(45, negative).

no

What happens is this:

Prolog tries the first clause, but the first argument won't match 0 (nor does
the second argument match zero).
Then it tries the second clause, binding X to 45, but the test X<0 fails.
So it backs out and tries the third clause, but this time the second
arguments don't match.
To get an actual answer, rather than just yes or no, you must call classify with
the second argument free:
Goal classify(45, What).

What=positive
1 Solution

Here's what really takes place in this case:

Chapter 62, Unification and Backtracking 91

 The goal classify(45, What) won't match the head of the first clause,
classify(0, zero), because 45 doesn't match 0. So the first clause can't be
used.
Again, the goal classify(45, What) matches the head of the second clause,
classify(X, negative), binding X to 45 and negative to What. But then the
text X<0 fails, because X is 45 and it is not true that 45<0. So Prolog backs
out of this clause, undoing the variable bindings just created.
Finally, classify(45, What) matches classify(X, positive), binding X to 45
and What to positive. The test X>0 succeeds. Since this is a successful
solution, Prolog doesn't backtrack; it returns to the calling procedure
(which in this case is the goal that you typed). And since the variable X
belongs to the calling procedure, that procedure can use its binding--in this
case, to print out the value automatically.

Summary

In this chapter we've introduced unification, backtracking, determinism, the

predicates not and fail, and the cut (!), and we've reviewed the important parts of
the tutorial information up to this point from a procedural perspective.
Prolog facts and rules receive information by being called with arguments that are
constants or bound variables; they return information to the calling procedure
by binding variable arguments that were unbound.
Unification is the process of matching two predicates and assigning free variables
to make the predicates identical. This mechanism is necessary so Prolog can
identify which clauses to call and bind values to variables. These are the major
points about matching (unification) presented in this chapter:
When Prolog begins an attempt to satisfy a goal, it starts at the top of the
program in search of a match.
When a new call is made, a search for a match to that call also begins at the
top of the program.
When a call has found a successful match, the call is said to return, and the
next subgoal in turn can be tried.
Once a variable has been bound in a clause, the only way to free that binding
is through backtracking.
Backtracking is the mechanism that instructs Prolog where to go to look for
solutions to the program. This process gives Prolog the ability to search

92 Visual Prolog
 through all known facts and rules for a solution. These are the four basic
principles of backtracking given in this chapter:
Subgoals must be satisfied in order, from top to bottom.
Predicate clauses are tested in the order they appear in the program, from top
to bottom.
When a subgoal matches the head of a rule, the body of that rule must be
satisfied next. The body of the rule then constitutes a new set of subgoals to
be satisfied.
A goal has been satisfied when a matching fact is found for each of the
extremities (leaves) of the goal tree.
A call that can produce multiple solutions is non-deterministic, while a call that
can produce one and only one solution is deterministic.
Visual Prolog provides three tools for controlling the course of your program's
logical search for solutions: these are the two predicates fail and not, and the
cut.
The fail predicate always fails; it forces backtracking in order to find alternate
solutions.
The not predicate succeeds when its associated subgoal can't be proven true.
The cut prevents backtracking.
It's easy to think of a Prolog rule as a procedure definition. From a procedural
perspective, rules can function as case statements, perform boolean tests, act
like goto statements (using the cut), and return computed values.

Chapter 62, Unification and Backtracking 93

CHAPTER 81

Simple and Compound Objects

82So far we've only shown you a few kinds of Visual Prolog data objects, such as
numbers, symbols, and strings. In this chapter we discuss the whole range of data
objects that Visual Prolog can create, from simple to compound objects.
We also show the different types of data structures and data objects that a Visual
Prolog program can contain. Because the standard domains do not cover some of
the compound data structures, we explain how to declare these compound data
structures in both the domains section and the predicates section of your
programs.

Simple Data Objects

A simple data object is either a variable or a constant. Don't confuse this use of
the word "constant" with the symbolic constants you define in the constants
section of a program. What we mean here by a constant, is anything identifying
an object not subject to variation, such as a character (a char), a number (an
integral value or a real), or an atom (a symbol or string).

Variables as Data Objects

Variables, which we've discussed in chapter 83, must begin with an upper-case
letter (A-Z) or an underscore (_). A single underscore represents an anonymous
variable, which stands for a "don't care what it is" situation. In Prolog, a variable
can bind with any legal Prolog argument or data object.
Prolog variables are local, not global. That is, if two clauses each contain a
variable called X, these Xs are two distinct variables. They may get bound to each
other if they happen to be brought together during unification, but ordinarily they
have no effect on each other.

Constants as Data Objects

Constants include characters, numbers, and atoms. Again, don't confuse constants
in this context with the symbolic constants defined in the constants section of a

94 Visual Prolog
program. A constant's value is its name. That is, the constant 2 can only stand for
the number 2, and the constant abracadabra can only stand for the symbol
abracadabra.

Characters
Characters are char type. The printable characters (ASCII 32-127) are the digits
0-9, upper-case letters A-Z, lower-case letters a-z, and the punctuation and
familiar TTY characters. Characters outside this range may not be portable
between different platforms; in particular, characters less than ASCII 32 (space)
are control characters, traditionally used by terminals and communication
equipment.
A character constant is simply written as the character you want, enclosed by
single quotes:
'a' '3'
'*' '{'
'W' 'A'

If, however, you want to specify a backslash or a single quote itself as the
character, precede it by a backslash (\):
'\\' backslash '\'' single quote.
There are a few characters that perform a special function, when preceded by the
escape character:

'\n' Newline (linefeed)

'\r' Carriage return.
'\t' Tab (horizontal)

Character constants can also be written as their ASCII codes, preceded by the
escape character, like this:
'\225' ß
'\3' %]

but the exact character displayed by more exotic ASCII values will vary
depending on your video-card/terminal.

Chapter 84, Repetition and recursion 95

Numbers
Numbers are either from one of the integral domains (see Table 85.8 on page 43),
or the real domain. Real numbers are stored in the IEEE standard format and
range from 1e-308 to 1e308 (10-308 to 10+308). Examples are:

Integers Real Numbers

3 3.
-77 34.96
32034 -32769
-10 4e27
0 -7.4e-296

Atoms
An atom is either a symbol or a string. The distinction between these is largely a
question about machine-representation and implementation, and is generally not
syntactically visible. When an atom is used as an argument in a predicate call, it is
the declaration for the predicate that determines if that argument should be
implemented as a string or a symbol.
Visual Prolog performs an automatic type conversion between the string domain
and the symbol domain, so you can use symbol atoms for string domains and
string atoms for the symbol domains. However, there is a loose convention
stating that anything in double quotes should be considered a string, while
anything not needing to be quoted to be syntactically valid is a symbol:
Symbol atoms are names starting with a lower-case letter, and containing only
letters, digits, and underscores.
String atoms are bound within double quotes and can contain any combination
of characters, except ASCII NULL (0, binary zero), which marks the end of
the string.

Symbol Atoms String Atoms

food "Jesse James"
rick_Jones_2nd "123 Pike street"
fred_Flintstone_1000_Bc_Rd_B "jon"
edrock
a "a"
new_york "New York"

96 Visual Prolog
 pdcProlog "Visual Prolog, by Prolog
Development Center"

As far as the string/symbol domain interchangeability goes, this distinction is not

important. However, things such as predicate names and functors for compound
objects, introduced below, must follow the syntactic conventions for symbols.

Compound Data Objects and Functors

Compound data objects allow you to treat several pieces of information as a

single item in such a way that you can easily pick them apart again. Consider, for
instance, the date April 2, 1988. It consists of three pieces of information--the
month, day, and year--but it's useful to treat the whole thing as a single object
with a treelike structure:
DATE
/ | \
October 15 1991

You can do this by declaring a domain containing the compound object date:
DOMAINS
date_cmp = date(string,unsigned,unsigned)

and then simply writing e.g.

..., D = date("October",15,1991), ...

This looks like a Prolog fact, but it isn't here--it's just a data object, which you can
handle in much the same way as a symbol or number. It begins with a name,
usually called a functor (in this case date), followed by three arguments.
Note carefully that a functor in Visual Prolog has nothing to do with a function in
other programming languages. A functor does not stand for some computation to
be performed. It's just a name that identifies a kind of compound data object and
holds its arguments together.
The arguments of a compound data object can themselves be compound. For
instance, you might think of someone's birthday as an information structure like
this:

Chapter 84, Repetition and recursion 97

 BIRTHDAY
/ \
/ \
person date
/ \ / | \
"Per" "Bilse" "Apr" 14 1960

In Prolog you would write this as:

birthday(person("Per","Bilse"),date("Apr",14,1960))

In this example, there are two parts to the compound object birthday: the object
person("Per", "Bilse") and the object date("Apr", 14, 1960). The functors of
these data objects are person and date.

Unification of Compound Objects

A compound object can unify either with a simple variable or with a compound
object that matches it (perhaps containing variables as parts of its internal
structure). This means you can use a compound object to pass a whole collection
of items as a single object, and then use unification to pick them apart. For
example,
date("April",14,1960)

matches X and binds X to date("April",14,1960).

Also
date("April",14,1960)

matches date(Mo,Da,Yr) and binds Mo to "April", Da to 14, and Yr to 1960.

Some examples of programming with compound objects follow in the next
sections.

Using the Equal Sign to Unify Compound Objects

Visual Prolog performs unification in two places. The first is when a call or goal
matches the head of a clause. The second is the across the equal (=) sign, which is
actually an infix predicate (a predicate that is located between its arguments rather
than before them).
Visual Prolog will make the necessary bindings to unify the objects on both sides
of the equal sign. This is useful for finding the values of arguments within a
compound object. For example, the following code excerpt tests if two people

98 Visual Prolog
have the same last name, then gives the second person the same address as the
first.
/* Program ch086e01.pro */

/* Copyright (c) 1986, '95 by Prolog Development Center */

DOMAINS
person = person(name,address)
name = name(first,last)
address = addr(street,city,state)
street = street(number,street_name)
city,state,street_name = string
first,last = string
number = integer

GOAL
P1 = person(name(jim,mos),addr(street(5,"1st st"),igo,"CA")),
P1 = person(name(_,mos),Address),
P2 = person(name(jane,mos),Address),
write("P1=",P1),nl,
write("P2=",P2),nl.

Treating Several Items as One

Compound objects can be regarded and treated as single objects in your Prolog
clauses, which greatly simplifies programming. Consider, for example, the fact
owns(john, book("From Here to Eternity", "James Jones")).

in which you state that John owns the book From Here to Eternity, written by
James Jones. Likewise, you could write
owns(john, horse(blacky)).

which can be interpreted as

John owns a horse named blacky.

The compound objects in these two examples are

book("From Here to Eternity", "James Jones")

and
horse(blacky)

If you had instead written two facts:

Chapter 84, Repetition and recursion 99

 owns(john, "From Here to Eternity").
owns(john, blacky ).

you would not have been able to decide whether blacky was the title of a book or
the name of a horse. On the other hand, you can use the first component of a
compound object--the functor--to distinguish between different objects. This
example used the functors book and horse to indicate the difference between the
objects.
Remember: Compound objects consist of a functor and the objects belonging to
that functor, as follows:
functor(object1, object2, ..., objectN)

An Example Using Compound Objects

A important feature of compound objects allows you to easily pass a group of
values as one argument. Consider a case where you are keeping a telephone
database. In your database, you want to include your friends' and family members'
birthdays. Here is a section of code you might have come up with:
PREDICATES
phone_list(symbol, symbol, symbol, symbol, integer, integer)
/* ( First, Last, Phone, Month, Day, Year) */

CLAUSES
phone_list(ed, willis, 422-0208, aug, 3, 1955).
phone_list(chris, grahm, 433-9906, may, 12, 1962).

Examine the data, noticing the six arguments in the fact phone_list; five of these
arguments can be broken down into two compound objects, like this:
person birthday
/ \ / | \
First Name Last Name Month Day Year

It might be more useful to represent your facts so that they reflect these
compound data objects. Going back a step, you can see that person is a
relationship, and the first and last names are the objects. Also, birthday is a
relationship with three arguments: month, day, and year. The Prolog
representation of these relationships is
person(First_name, Last_name)
birthday(Month, Day, Year)

100 Visual Prolog

You can now rewrite your small database to include these compound objects as
part of your database.
DOMAINS
name = person(symbol, symbol) /* (First, Last) */
birthday = b_date(symbol, integer, integer) /* (Month, Day, Year) */
ph_num = symbo /* Phone_number */

PREDICATES
phone_list(name, ph_num, birthday)

CLAUSES
phone_list(person(ed, willis), "422-0208", b_date(aug, 3, 1955)).
phone_list(person(chris, grahm), "433-9906", b_date(may, 12, 1962)).

In this program, two compound domains declarations were introduced. We go

into more detail about these compound data structures later in this chapter. For
now, we'll concentrate on the benefits of using such compound objects.
The phone_list predicate now contains three arguments, as opposed to the
previous six. Sometimes breaking up your data into compound objects will clarify
your program and might help process the data.
Now add some rules to your small program. Suppose you want to create a list of
people whose birthdays are in the current month. Here's the program code to
accomplish this task; this program uses the standard predicate date to get the
current date from the computer's internal clock. The date predicate is discussed
later in chapter 87. For now, all you need to know is that it will return the current
year, month, and day from your computer's clock.
/* Program ch088e03.pro */

/* Copyright (c) 1986, '95 by Prolog Development Center */

DOMAINS
name = person(symbol,symbol) /* (First, Last) */
birthday = b_date(symbol,integer,integer) /* (Month, Day, Year) */
ph_num = symbol /* Phone_number */

PREDICATES
nondeterm phone_list(name,symbol,birthday)
get_months_birthdays
convert_month(symbol,integer)
check_birthday_month(integer,birthday)
write_person(name)

Chapter 84, Repetition and recursion 101

CLAUSES
get_months_birthdays:-
write("************ This Month's Birthday List *************"),nl,
write(" First name\t\t Last Name\n"),
write("*****************************************************"),nl,
date(_, This_month, _), /* Get month from system clock */
phone_list(Person, _, Date),
check_birthday_month(This_month, Date),
write_person(Person),
fail.

get_months_birthdays:-
write("\n\n Press any key to continue: "),nl,
readchar(_).

write_person(person(First_name,Last_name)):-
write(" ",First_name,"\t\t ",Last_name),nl.

check_birthday_month(Mon,b_date(Month,_,_)):-
convert_month(Month,Month1),
Mon = Month1.

phone_list(person(ed, willis), "767-8463", b_date(jan, 3, 1955)).

phone_list(person(benjamin, thomas), "438-8400", b_date(feb, 5, 1985)).
phone_list(person(ray, william), "555-5653", b_date(mar, 3, 1935)).
phone_list(person(thomas, alfred), "767-2223", b_date(apr, 29, 1951)).
phone_list(person(chris, grahm), "555-1212", b_date(may, 12, 1962)).
phone_list(person(dustin, robert), "438-8400", b_date(jun, 17, 1980)).
phone_list(person(anna, friend), "767-8463", b_date(jun, 20, 1986)).
phone_list(person(brandy, rae), "555-5653", b_date(jul, 16, 1981)).
phone_list(person(naomi, friend), "767-2223", b_date(aug, 10, 1981)).
phone_list(person(christina, lynn), "438-8400", b_date(sep, 25, 1981)).
phone_list(person(kathy, ann), "438-8400", b_date(oct, 20, 1952)).
phone_list(person(elizabeth, ann), "555-1212", b_date(nov, 9, 1984)).
phone_list(person(aaron, friend), "767-2223", b_date(nov, 15, 1987)).
phone_list(person(jennifer, caitlin), "438-8400", b_date(dec, 31,
1981)).

102 Visual Prolog

 convert_month(jan, 1).
convert_month(feb, 2).
convert_month(mar, 3).
convert_month(apr, 4).
convert_month(may, 5).
convert_month(jun, 6).
convert_month(jul, 7).
convert_month(aug, 8).
convert_month(sep, 9).
convert_month(oct, 10).
convert_month(nov, 11).
convert_month(dec, 12).

GOAL
get_months_birthdays.

Load and run this program.

How do compound data objects help in this program? This should be easy to see
when you examine the code. Most of the processing goes on in the
get_months_birthdays predicate.
First, the program makes a window to display the results.
After this, it writes a header in the window to help interpret the results.
Next, in get_months_birthdays, the program uses the built-in predicate date to
obtain the current month.
After this, the program is all set to search the database and list the people who
were born in the current month. The first thing to do is find the first person in
the database. The call phone_list(Person, _, Date) binds the person's first and
last names to the variable Person by binding the entire functor person to
Person. It also binds the person's birthday to the variable Date.
Notice that you only need to use one variable to store a person's complete
name, and one variable to hold the birthday. This is the power of using
compound data objects.
Your program can now pass around a person's birthday simply by passing on the
variable Date. This happens in the next subgoal, where the program passes the
current month (represented by an integer) and the birthday (of the person it's
processing) to the predicate check_birthday_month.
Look closely at what happens. Visual Prolog calls the predicate
check_birthday_month with two variables: The first variable is bound to an
integer, and the second is bound to a birthday term. In the head of the rule that
defines check_birthday_month, the first argument, This_month, is matched

Chapter 84, Repetition and recursion 103

 with the variable Mon. The second argument, Date, is matched against
b_date(Month, _,_).
Since all you're concerned with is the month of a person's birthday, you have used
the anonymous variable for both the day and the year of birth.
The predicate check_birthday_month first converts the symbol for the month into
an integer value. Once this is done, Visual Prolog can compare the value of the
current month with the value of the person's birthday month. If this
comparison succeeds, then the subgoal check_birthday_month succeeds, and
processing can continue. If the comparison fails (the person currently being
processed was not born in the current month), Visual Prolog begins to
backtrack to look for another solution to the problem.
The next subgoal to process is write_person. The person currently being
processed has a birthday this month, so it's OK to print that person's name in
the report. After printing the information, the clause fails, which forces
backtracking.
Backtracking always goes up to the most recent non-deterministic call and tries
to re-satisfy that call. In this program, the last non-deterministic call
processed is the call to phone_list. It is here that the program looks up another
person to be processed. If there are no more people in the database to process,
the current clause fails; Visual Prolog then attempts to satisfy this call by
looking further down in the database. Since there is another clause that defines
get_months_birthdays, Visual Prolog tries to satisfy the call to
get_months_birthdays by satisfying the subgoals to this other clause.

Exercise
Modify the previous program so that it will also print the birth dates of the people
listed. Next, add telephone numbers to the report.

Declaring Domains of Compound Objects

In this section, we show you how domains for compound objects are defined.
After compiling a program that contains the following relationships:
owns(john, book("From Here to Eternity", "James Jones")).

and
owns(john, horse(blacky)).

you could query the system with this goal:

104 Visual Prolog

 owns(john, X)

The variable X can be bound to different types of objects: a book, a horse, or

perhaps other objects you define. Because of your definition of the owns
predicate, you can no longer employ the old predicate declaration of owns:
owns(symbol, symbol)

The second argument no longer refers to objects belonging to the domain symbol.
Instead, you must formulate a new declaration to the predicate, such as
owns(name, articles)

You can describe the articles domain in the domains section as shown here:
DOMAINS
articles = book(title,author); horse(name)
/* Articles are books or horses */
title, author, name = symbol

The semicolon is read as or. In this case, two alternatives are possible: A book
can be identified by its title and author, or a horse can be identified by its name.
The domains title, author, and name are all of the standard domain symbol.
More alternatives can easily be added to the domains declaration. For example,
articles could also include a boat, a house, or a bankbook. For a boat, you can
make do with a functor that has no arguments attached to it. On the other hand,
you might want to give a bank balance as a figure within the bankbook. The
domains declaration of articles is therefore extended to:
articles = book(title, author) ; horse(name) ;
boat ; bankbook(balance)
title, author, name = symbol
balance = real

Here is a full program that shows how compound objects from the domain
articles can be used in facts that define the predicate owns.
/* Program ch089e04.pro */

/* Copyright (c) 1986, '95 by Prolog Development Center */

DOMAINS
articles = book(title, author) ;
horse(name) ; boat ;
bankbook(balance)
title, author, name = symbol
balance = real

Chapter 84, Repetition and recursion 105

PREDICATES
nondeterm owns(name,articles)

CLAUSES
owns(john, book("A friend of the family", "Irwin Shaw")).
owns(john, horse(blacky)).
owns(john, boat).
owns(john, bankbook(1000)).

Now compile and run the program with the following goal:
owns(john, Thing).

Visual Prolog responds with:

Thing=book("A friend of the family","Irwin Shaw")
Thing=horse("blacky")
Thing=boat
Thing=bankbook(1000)
4 Solutions

Writing Domain Declarations: a Summary

This is a generic representation of how to write domain declarations for
compound objects:
domain =alternative1(D, D, ...);
alternative2(D, D, ...);
...

Here, alternative1 and alternative2 are arbitrary (but different) functors. The
notation (D, D, ...) represents a list of domain names that are either declared
elsewhere or are one of the standard domain types (such as symbol, integer, real,
etc).
Note:
The alternatives are separated by semicolons.
Every alternative consists of a functor and, possibly, a list of domains for the
corresponding arguments.
If the functor has no arguments, you can write it as alternativeN or alternativeN(
) in your programs. In this book, we use the former syntax.

106 Visual Prolog

Multi-Level Compound Objects
Visual Prolog allows you to construct compound objects on several levels. For
example, in
book("The Ugly Duckling", "Andersen")

instead of using the author's last name, you could use a new structure that
describes the author in more detail, including both the author's first and last
names. By calling the functor for the resulting new compound object author, you
can change the description of the book to
book("The Ugly Duckling", author("Hans Christian", "Andersen"))

In the old domain declaration

book(title, author)

the second argument of the book functor is author. But the old declaration
author = symbol

can only include a single name, so it's no longer sufficient. You must now specify
that an author is also a compound object made up of the author`s first and last
name. You do this with the domain statement:
author = author(first_name, last_name)

which leads to the following declarations:

DOMAINS
articles = book(title, author); .. /* First level */
author = author(first_name, last_name)/* Second level */
title, first_name, last_name = symbol /* Third level */

When using compound objects on different levels in this way, it's often helpful to
draw a "tree":
book
/ \
title author
/ \
/ \
firstname lastname

A domain declaration describes only one level of the tree at a time, and not the
whole tree. For instance, a book can't be defined with the following domain
declaration:

Chapter 84, Repetition and recursion 107

 book = book(title,author(first_name,last_name)) /* Not allowed */

An Example That Illustrates Sentence Structure

As another example, consider how to represent the grammatical structure of the
sentence
ellen owns the book.

using a compound object. The most simple sentence structure consists of a noun
and a verb phrase:
sentence = sentence(noun, verbphrase)

A noun is just a simple word:

noun = noun(word)

A verb phrase consists of either a verb with a noun phrase or a single verb.
verbphrase = verbphrase(verb, noun); verb(word)
verb = verb(word)

Using these domain declarations (sentence, noun, verbphrase, and verb), the
sentence ellen owns the book. becomes
sentence(noun(ellen), verbphrase(verb(owns), noun(book)))

The corresponding tree is

sentence
/ \
/ \
noun verbphrase
| / \
| verb noun
| | |
ellen owns the book

A data structure like this might be the output of a parser, which is a program that
determines the grammatical structure of a sentence. Parsing is not built into
Visual Prolog, but we have included a parser implementing simple sentence
analysis with your Visual Prolog package. (Try to run the project VPI\
PROGRAMS\SEN_AN when you're ready to tackle this subject.)

108 Visual Prolog

Exercises
Write a suitable domains declaration using compound objects that could be used
in a Visual Prolog catalog of musical shows. A typical entry in the catalog
might be
Show: West Side Story
Lyrics: Stephen Sondheim
Music: Leonard Bernstein

Using compound objects wherever possible, write a Visual Prolog program to

keep a database of United States senators. Entries should include the senator's
first and last name, affiliation (state and party), size of constituency, date of
election, and voting record on ten bills. Or, if you're not familiar with United
States senators, use any political (or other) organization that you're familiar
with.

Compound Mixed-Domain Declarations

In this section, we discuss three different types of domain declarations you can
add to your programs. These declarations allow you to use predicates that
take an argument, more than one type of more than one possible type
take a variable number of arguments, each of a specified type
take a variable number of arguments, some of which might be of more than
one possible type

Multiple-Type Arguments
To allow a Visual Prolog predicate to accept an argument that gives information
of different types, you must add a functor declaration. In the following example,
the your_age clause will accept an argument of type age, which can be a string, a
real, or an integer.
DOMAINS
age = i(integer); r(real); s(string)

PREDICATES
your_age(age)

CLAUSES
your_age(i(AGE)) :- write(Age).
your_age(r(AGE)) :- write(Age).
your_age(s(AGE)) :- write(Age).

Chapter 84, Repetition and recursion 109

Visual Prolog does not allow the following domain declaration:
DOMAINS
age = integer; real; string /* Not permitted. */

Lists
Suppose you are keeping track of the different classes a professor might teach.
You might produce the following code:
PREDICATES
teacher(symbol, symbol, symbol) /* First_name, Last_name, Class) */

CLAUSES
teacher(ed, willis, english1).
teacher(ed, willis, math1).
teacher(ed, willis, history1).
teacher(mary, maker, history2).
teacher(mary, maker, math2).
teacher(chris, grahm, geometry).

Here, you need to repeat the teacher's name for each class he or she teaches. For
each class, you need to add another fact to the database. Although this is perfectly
OK in this situation, you might find a school where there are hundreds of classes;
this type of data structure would get a little tedious. Here, it would be helpful if
you could create an argument to a predicate that could take on one or more
values.
A list in Prolog does just that. In the following code, the argument class is
declared to be of a list type. We show here how a list is represented in Prolog, but
list-handling predicates are covered in chapter 90.
DOMAINS
classes = symbol* /* declare a list domain */

PREDICATES
teacher(symbol, symbol, classes) /* (First, Last, Classes) */

CLAUSES
teacher(ed, willis, [english1, math1, history1]).
teacher(mary, maker, [history2, math2]).
teacher(chris, grahm, [geometry]).

In this example, the code is more concise and easier to read than in the preceding
one. Notice the domains declaration:
DOMAINS
classes = symbol*

110 Visual Prolog

The asterisk (*) means that classes is a list of symbols. You can just as easily
declare a list of integers:
DOMAINS
integer_list = integer*

Once you declare a domain, it's easy to use it; just place it as an argument to a
predicate declared in the predicates section. Here's an example of using an
integer list:
DOMAINS
integer_list = integer*

PREDICATES
test_scores(symbol, symbol, integer_list)/* (First, Last, Test Scores)
*/

CLAUSES
test_scores(lisa, lavender, [86, 91, 75]).
test_scores(libby, dazzner, [79, 75]).
test_scores(jeff, zheutlin, []).

In the case of Jeff Zheutlin, notice that a list doesn't need to contain any elements
at all.
Lists are discussed in greater detail in chapter 91.

Summary

These are the important points covered in this chapter:

A Visual Prolog program can contain many types of data objects: simple and
compound, standard and user-defined. A simple data object is one of the
following:
a variable; such as X, MyVariable, _another_variable, or a single
underscore ( _ ) for an anonymous variable
a constant; a char, an integer or real number, or a symbol or string atom
Compound data objects allow you to treat several pieces of information as a
single item. A compound data object consists of a name (known as a functor)
and one or more arguments. You can define a domain with several alternative
functors.
A functor in Visual Prolog is not the same thing as a function in other
programming languages. A functor does not stand for some computation to

Chapter 84, Repetition and recursion 111

 be performed. It's just a name that identifies a kind of compound data object
and holds its arguments together.
Compound objects can be regarded and treated as single objects; you use the
functor to distinguish between different objects. Visual Prolog allows you to
construct compound objects on several levels; the arguments of a compound
data object can also be compound objects. With compound mixed domain
declarations, you can use predicates that:
take an argument of more than one possible type (functor declaration).
take a variable number of arguments, each of a specified type (list
declaration).
take a variable number of arguments, some of which might be of more than
one possible type.

112 Visual Prolog

CHAPTER 92

Repetition and Recursion

93Much of the usefulness of computers comes from the fact that they are good at
doing the same thing over and over again. Prolog can express repetition both in its
procedures and in its data structures. The idea of a repetitive data structure may
sound strange, but Prolog allows you to create data structures whose ultimate size
is not known at the time you create them. In this chapter, we discuss repetitive
processes first (as loops and recursive procedures), then cover recursive data
structures.

Repetitive Processes

Pascal, BASIC, or C programmers who start using Visual Prolog are often
dismayed to find that the language has no FOR, WHILE, or REPEAT statements.
There is no direct way to express iteration. Prolog allows only two kinds of
repetition--backtracking, in which it searches for multiple solutions in a single
query, and recursion, in which a procedure calls itself.
As it turns out, this lack doesn't restrict the power of the Prolog language. In fact,
Visual Prolog recognizes a special case of recursion--called tail recursion --and
compiles it into an iterative loop in machine language. This means that although
the program logic is expressed recursively, the compiled code is as efficient as it
would be in Pascal or BASIC.
In this section, we explore the art of writing repetitive processes in Prolog. As
you'll see, recursion is--in most cases--clearer, more logical, and less error-prone
than the loops that conventional languages use. Before delving into recursion,
however, take another look at backtracking.

Backtracking Revisited
When a procedure backtracks, it looks for another solution to a goal that has
already been satisfied. It does this by retreating to the most recent subgoal that
has an untried alternative, using that alternative, then moving forward again. You
can exploit backtracking as a way to perform repetitive processes.

Chapter 94, Repetition and recursion 113

Example
Program ch06e01.pro demonstrates how to use backtracking to perform repetitive
processes--it prints all solutions to a query.
/* Program ch095e01.pro */

PREDICATES
nondeterm country(symbol)
print_countries

CLAUSES
country("England").
country("France").
country("Germany").
country("Denmark").

print_countries:-
country(X),
write(X), /* write the value of X */
nl, /* start a new line */
fail.
print_countries.

The predicate country simply lists the names of various countries, so that a goal
such as
country(X)

has multiple solutions. The predicate print_countries then prints out all of these
solutions. It is defined as follows:
print_countries :-
country(X), write(X), nl, fail.

print_countries.

The first clause says:

"To print countries, find a solution to country(X), then write X and start a new
line, then fail."
In this case, "fail" means:
"assume that a solution to the original goal has not been reached, so back up
and look for an alternative."
The built-in predicate fail always fails, but you could equally well force
backtracking by using any other goal that would always fail, such as 5=2+2 or
country(shangri_la).

114 Visual Prolog

The first time through, X is bound to england, which is printed. Then, when it hits
fail, the computer backs up. There are no alternative ways to satisfy nl or
write(X), so the computer looks for a different solution to country(X).

The last time country(X) was executed, it bound a value to the previously free
variable X. So, before retrying this step, the computer unbinds X (frees it). Then it
can look for an alternative solution for country(X) and bind X to a different value.
If it succeeds, processing goes forward again and the name of another country is
printed.
Eventually, the first clause runs out of alternatives. The only hope then is to try
another clause for the same predicate. Sure enough, execution falls through to the
second clause, which succeeds without doing anything further. In this way the
goal print_countries terminates with success. Its complete output is
england
france
germany
denmark
yes

If the second clause were not there, the print_countries goal would terminate
with failure, and the final message would be no. Apart from that, the output
would be the same.

Exercise
Modify ch06e01.pro so that
country has two arguments, name and population, and
only those countries with populations greater than 10 million (1e+7) are printed

Pre- and Post-Actions

Typically, a program that retrieves all the solutions to a goal will also want to do
something beforehand and afterward. For instance, your program could
Print Some delightful places to live are....
Print all solutions to country(X).
Close by printing And maybe others.

Note that print_countries, as defined in the preceding example, already includes

clauses that print all solutions to country(X) and close by (potentially) printing a
final message.

Chapter 94, Repetition and recursion 115

The first clause for print_countries corresponds to step 2 and prints all the
solutions; its second clause corresponds to step 3 and simply terminates the goal
successfully (because the first clause always fails).
You could change the second clause in ch06e01.pro to
print_countries :- write("And maybe others."), nl.

which would implement step 3 as specified.

What about step 1? There's no reason why print_countries should have only two
clauses. It can have three, like this:
print_countries :-
write("Some delightful places to live are"),
nl,
fail.

print_countries :-
country(X),
write(X),
nl,
fail.

print_countries :-
write("And maybe others."), nl.

The fail in the first clause is important--it ensures that, after executing the first
clause, the computer backs up and tries the second clause. It's also important that
the predicates write and nl do not generate alternatives; strictly speaking, the first
clause tries all possible solutions before failing.
This three-clause structure is more of a trick than an established programming
technique. A more fastidious programmer might try to do things this way:
print_countries_with_captions :-
write("Some delightful places to live are"),
nl,
print_countries,
write("And maybe others."),
nl.

print_countries :-
country(X),
write(X),
nl,
fail.

116 Visual Prolog

There's nothing essentially wrong here, but this hypothetical fastidious
programmer has made a mistake.

Exercise
Don't look ahead--figure out what's wrong with this program, and fix it!
You're right--the problem is that, as written in the latest example, print_countries
will always fail, and print_countries_with_captions will never get to execute any
of the subgoals that follow it. As a result, And maybe others. will never be printed.
To fix this, all you need to do is restore the original second clause for
print_countries.
print_countries.

to its original position. If you want the goal print_countries_with_captions to

succeed, it must have at least one clause that does not contain fail.

Implementing Backtracking with Loops

Backtracking is a good way to get all the alternative solutions to a goal. But even
if your goal doesn't have multiple solutions, you can still use backtracking to
introduce repetition. Simply define the two-clause predicate
repeat.
repeat :- repeat.

This tricks Prolog's control structure into thinking it has an infinite number of
different solutions. (Never mind how--after reading about tail recursion, you'll
see how this works.) The purpose of repeat is to allow backtracking ad infinitum.
/* Program ch096e02.pro */

/* Uses repeat to keep accepting characters and printing them

until the user presses Enter. */

PREDICATES
repeat
typewriter

CLAUSES
repeat.
repeat:-repeat.

Chapter 94, Repetition and recursion 117

 typewriter:-
repeat,
readchar(C), /* Read a char, bind C to it */
write(C),
C = '\r',!. /* Is it a carriage return? fail if not */

Program 2 shows how repeat works. The rule typewriter :- ... describes a
procedure that accepts characters from the keyboard and prints them on the screen
until the user presses the Enter (Return) key.
typewriter works as follows:
Execute repeat (which does nothing).
Then read a character into the variable C.
Then write C.
Then check if C is a carriage return.
If so, you're finished. If not, backtrack and look for alternatives. Neither write nor
readchar generates alternative solutions, so backtrack all the way to repeat,
which always has alternative solutions.
Now processing can go forward again, reading another character, printing it, and
checking whether it's a carriage return.
Note, by the way, that C looses its binding when you backtrack past readchar(C),
which bound it. This kind of unbinding is vital when you use backtracking to
obtain alternative solutions to a goal, but it makes it hard to use backtracking for
any other purpose. The reason is that, although a backtracking process can repeat
operations any number of times, it can't "remember" anything from one repetition
to the next. All variables loose their values when execution backtracks over the
steps that established those values. There is no simple way for a repeat loop to
keep a counter, a total, or any other record of its progress.

Exercises
Modify 2 so that, if the user types lower-case letters, they will be displayed as
upper-case.
If you'd like to play with file I/O now, look up the appropriate built-in predicates
and write a program that uses a repeat loop to copy a file character-by-
character. (Refer to chapter 97.)

118 Visual Prolog

Recursive Procedures
The other way to express repetition is through recursion. A recursive procedure is
one that calls itself. Recursive procedures have no trouble keeping records of their
progress because counters, totals, and intermediate results can be passed from
each iteration to the next as arguments.
The logic of recursion is easy to follow if you forget, for the moment, how
computers work. (Prolog is so different from machine language that ignorance of
computers is often an asset to the Prolog programmer.) Forget for the moment
that the computer is trekking through memory addresses one by one, and imagine
a machine that can follow recipes like this one:
To find the factorial of a number N:

If N is 1, the factorial is 1.

Otherwise, find the factorial of N-1, then multiply it by N.

This recipe says: To find the factorial of 3, you must find the factorial of 2, and,
to find the factorial of 2, you must find the factorial of 1. Fortunately, you can
find the factorial of 1 without referring to any other factorials, so the repetition
doesn't go on forever. When you have the factorial of 1, you multiply it by 2 to
get the factorial of 2, then multiply that by 3 to get the factorial of 3, and you're
done.
In Visual Prolog:
factorial(1, 1) :- !.

factorial(X, FactX) :-
Y = X-1,
factorial(Y, FactY),
FactX = X*FactY.

A complete program is as follows:

/* Program ch098e03.pro */

/* Recursive program to compute factorials.

Ordinary recursion, not tail recursion. */

PREDICATES
factorial(unsigned,real)

CLAUSES
factorial(1,1):-!.

Chapter 94, Repetition and recursion 119

 factorial(X,FactX):-
Y=X-1,
factorial(Y,FactY),
FactX = X*FactY.

What the Computer is Really Doing

But wait a minute, you say. How does the computer execute factorial while it's in
the middle of executing factorial? If you call factorial with X=3, factorial will
then call itself with X=2. Will X then have two values, or will the second value
just wipe out the first, or what?
The answer is that the computer creates a new copy of factorial so that factorial
can call itself as if it were a completely separate procedure. The executable code
doesn't have to be duplicated, of course, but the arguments and internal variables
do.
This information is stored in an area called a stack frame, which is created every
time a rule is called. When the rule terminates, the stack is reset (unless it was a
non-deterministic return) and execution continues in the stack frame for the
parent.

Advantages of Recursion
Recursion has three main advantages:
It can express algorithms that can't conveniently be expressed any other way.
It is logically simpler than iteration.
It is used extensively in list processing.
Recursion is the natural way to describe any problem that contains within itself
another problem of the same kind. Examples include tree search (a tree is made
up of smaller trees) and recursive sorting (to sort a list, partition it, sort the parts,
and then put them together).
Logically, recursive algorithms have the structure of an inductive mathematical
proof. The preceding recursive factorial algorithm, in Program 3, describes an
infinite number of different computations by means of just two clauses. This
makes it easy to see that the clauses are correct. Further, the correctness of each
clause can be judged independently of the other.

120 Visual Prolog

Tail Recursion Optimisation
Recursion has one big drawback: It eats memory. Whenever one procedure calls
another, the calling procedure's state of execution must be saved so that it (the
calling procedure) can resume where it left off after the called procedure has
finished. This means that, if a procedure calls itself 100 times, 100 different states
of execution must be stored at once. (The saved state of execution is known as a
stack frame.) The maximum stack size on 16bit platforms, such as the IBM PC
running DOS, is 64K, which will accommodate, at most, 3000 or 4000 stack
frames. On 32bit platforms, the stack may theoretically grow to several
GigaBytes; here, other system limitations will set in before the stack overflows.
Anyway, what can be done to avoid using so much stack space?
It turns out that there's a special case in which a procedure can call itself without
storing its state of execution. What if the calling procedure isn't going to resume
after the called procedure finishes?
Suppose the calling procedure calls a procedure as its very last step. When the
called procedure finishes, the calling procedure won't have anything else to do.
This means the calling procedure doesn't need to save its state of execution,
because that information isn't needed any more. As soon as the called procedure
finishes, control can go directly to wherever it would have gone when the calling
procedure finished.
For example, suppose that procedure A calls procedure B, and B calls procedure
C as its very last step. When B calls C, B isn't going to do anything more. So,
instead of storing the current state of execution for C under B, you can replace B's
old stored state (which isn't needed any more) with C's current state, making
appropriate changes in the stored information. When C finishes, it thinks it was
called by A directly.
Now suppose that, instead of calling C, procedure B calls itself as its very last
step. The recipe says that, when B calls B, the stack frame for the calling B
should be replaced by a stack frame for the called B. This is a particularly simple
operation; only the arguments need to be set to new values, and then processing
jumps back to the beginning of the procedure. So, from a procedural point of
view, what happens is very similar to updating the control variables in a loop.
This is called tail recursion optimization, or last-call optimization. Note that for
technical reasons, recursive functions (predicates returning a value, described in
chapter 99) cannot be tail recursive.

Making Tail Recursion Work

What does it mean to say that one procedure calls another "as its very last step?"
In Prolog, this means that

Chapter 94, Repetition and recursion 121

The call is the very last subgoal of the clause.
There are no backtrack points earlier in the clause.
Here's an example that satisfies both conditions:
count(N) :-
write(N), nl,
NewN = N+1,
count(NewN).

This procedure is tail recursive; it calls itself without allocating a new stack
frame, so it never runs out of memory. As program 4 shows, if you give it the
goal
count(0)

count will print integers starting with 0 and never ending. Eventually, rounding
errors will make it print inaccurate numbers, but it will never stop.
/* Program ch0100e04.pro */

/* Tail recursive program that never runs out of memory */

PREDICATES
count(ulong)

CLAUSES
count(N):-
write('\r',N),
NewN = N+1,
count(NewN).

GOAL
nl,
count(0).

Exercise
Without looking ahead, modify 4 so that it is no longer tail recursive. How many
iterations can it execute before running out of memory? Try it and see. (On 32bit
platforms, this will take a considerable length of time, and the program will most
likely not run out of stack space; it, or the system, will run out of memory in
general. On 16bit platforms, the number of possible iterations is directly related to
the stack size.

122 Visual Prolog

How Not to Do Tail Recursion
Now that you've seen how to do tail recursion right, program 5 shows you three
ways to do it wrong.
If the recursive call isn't the very last step, the procedure isn't tail recursive. For
example:
badcount1(X) :-
write('\r',X),
NewX = X+1,
badcount1(NewX),
nl.

Every time badcount1 calls itself, a stack frame has to be saved so that control
can return to the calling procedure, which has yet to execute its final nl. So
only a few thousand recursive calls can take place before the program runs out
of memory.
Another way to lose tail recursion is to leave an alternative untried at the time
the recursive call is made. Then a stack frame must be saved so that, if the
recursive call fails, the calling procedure can go back and try the alternative.
For example:
badcount2(X) :-
write('\r',X),
NewX = X+1,
badcount2(NewX).

badcount2(X) :-
X < 0,
write("X is negative.").

Here, the first clause of badcount2 calls itself before the second clause has
been tried. Again, the program runs out of memory after a certain number of
calls.
The untried alternative doesn't need to be a separate clause for the recursive
procedure itself. It can equally well be an alternative in some other clause that
it calls. For example:
badcount3(X) :-
write('\r',X),
NewX = X+1,
check(NewX),
badcount3(NewX).

check(Z) :- Z >= 0.
check(Z) :- Z < 0.

Chapter 94, Repetition and recursion 123

 Suppose X is positive, as it normally is. Then, when badcount3 calls itself, the
first clause of check has succeeded, but the second clause of check has not yet
been tried. So badcount3 has to preserve a copy of its stack frame in order to
go back and try the other clause of check if the recursive call fails.
/* Program ch0101e05.pro */

/* In 32bit memory architectures, such as '386 UNIX, the examples here

will run for a considerable length of time, occupying large amounts
of memory and possibly reducing system performance significantly.
*/

PREDICATES
badcount1(long)
badcount2(long)
badcount3(long)
check(long)

CLAUSES
/* badcount1:
The recursive call is not the last step. */

badcount1(X):-
write('\r',X),
NewX = X+1,
badcount1(NewX),
nl.

/* badcount2:
There is a clause that has not been tried
at the time the recursive call is made. */

badcount2(X):-
write('\r',X),
NewX = X+1,
badcount2(NewX).

badcount2(X):-
X < 0,
write("X is negative.").

/* badcount3:
There is an untried alternative in a
predicate called before the recursive call. */

124 Visual Prolog

 badcount3(X):-
write('\r',X),
NewX = X+1,
check(NewX),
badcount3(NewX).

check(Z):-
Z >= 0.
check(Z):-
Z < 0.

Cuts to the Rescue

By now, you may think it's impossible to guarantee that a procedure is tail
recursive. After all, it's easy enough to put the recursive call in the last subgoal of
the last clause, but how do you guarantee there are no alternatives in any of the
other procedures that it calls?
Fortunately, you don't have to. The cut (!) allows you to discard whatever
alternatives may exist. You'll need to use the check_determ compiler directive to
guide you through setting the cuts. (Compiler directives are described in the
chapter 102.)
You can fix up badcount3 as follows (changing its name in the process):
cutcount3(X) :-
write('\r',X),
NewX = X+1,
check(NewX),
!,
cutcount3(NewX).

leaving check as it was.

The cut means "burn your bridges behind you" or, more precisely, "once you
reach this point, disregard alternative clauses for this predicate and alternative
solutions to earlier subgoals within this clause." That's precisely what you need.
Because alternatives are ruled out, no stack frame is needed and the recursive call
can go inexorably ahead.
A cut is equally effective in badcount2, by negating and moving the test from the
second clause to the first:

Chapter 94, Repetition and recursion 125

 cutcount2(X) :-
X >= 0, !,
write('\r',X),
NewX = X+1,
cutcount2(NewX).

cutcount2(X) :-
write("X is negative.").

A cut is really all about making up ones mind. You set a cut whenever you can
look at non-deterministic code, and say "Yes! Go ahead!" -- whenever it's obvious
that alternatives are of no interest. In the original version of the above example,
which tries to illustrate a situation where you have to decide something about X
(the test X < 0 in the second clause), the second clause had to remain an option as
the code in the first clause didn't test X. By moving the test to the first clause and
negating it, a decision can be reached already there and a cut set in accordance:
"Now I know I don't want to write that X is negative.".
The same applies to cutcount3. The predicate check illustrates a situation where
you want to do some additional processing of X, based on its sign. However, the
code for check is, in this case for illustration, non-deterministic, and the cut after
the call to it is all about you having made up your mind. After the call to check,
you can say "Yes! Go ahead!". However, the above is slightly artificial -- it would
probably be more correct for check to be deterministic:
check(Z) :- Z >= 0, !, ... % processing using Z
check(Z) :- Z < 0, ... %processing using Z

And, since the test in the second clause of check is the perfect negation of the test
in the first, check can be further rewritten as:
check(Z) :- Z >= 0, !, % processing using Z
check(Z) :- ... % processing using Z

When a cut is executed, the computer assumes there are no untried alternatives
and does not create a stack frame. Program 6 contains modified versions of
badcount2 and badcount3:
/* Program ch0103e06.pro */

/* Shows how badcount2 and badcount3 can be fixed by adding cuts to

rule out the untried clauses. These versions are tail recursive. */

PREDICATES
cutcount2(long)
cutcount3(long)
nondeterm check(long)

126 Visual Prolog

CLAUSES
/* cutcount2:
There is a clause that has not been tried
at the time the recursive call is made. */

cutcount2(X):-
X>=0,!,
write('\r',X),
NewX = X + 1,
cutcount2(NewX).

cutcount2(_):-
write("X is negative.").

/* cutcount3:
There is an untried alternative in a
clause called before the recursive call. */

cutcount3(X):-
write('\r',X),
NewX = X+1,
check(NewX),
!,
cutcount3(NewX).

check(Z):-Z >= 0.
check(Z):-Z < 0.

Unfortunately, cuts won't help with badcount1, whose need for stack frames has
nothing to do with untried alternatives. The only way to improve badcount1
would be to rearrange the computation so that the recursive call comes at the end
of the clause.

Using Arguments as Loop Variables

Now that you've mastered tail recursion, what can you do about loop variables
and counters? To answer that question, we'll do a bit of Pascal-to-Prolog
translation, assuming that you're familiar with Pascal. Generally, the results of
direct translations between two languages, whether natural or programming, are
poor. The following isn't too bad and serves as a reasonable illustration of strictly
imperative programming in Prolog, but you should never write Prolog programs
by blind translation from another language. Prolog is a very powerful and
expressive language, and properly written Prolog programs will display a
programming style and problem focus quite different from what programs in
other languages do.

Chapter 94, Repetition and recursion 127

In the "Recursion" section, we developed a recursive procedure to compute
factorials; in this section we'll develop an iterative one. In Pascal, this would be:
P := 1;
for I := 1 to N do P := P*I;
FactN := P;

If you're unfamiliar with Pascal, the :- is the assignment, read as "becomes".

There are four variables here. N is the number whose factorial will be calculated;
FactN is the result of the calculation; I is the loop variable, counting from 1 to N;
and P is the variable in which the product accumulates. A more efficient Pascal
programmer might combine FactN and P, but in Prolog it pays to be fastidiously
tidy.
The first step in translating this into Prolog is to replace for with a simpler loop
statement, making what happens to I in each step more explicit. Here is the
algorithm recast as a while loop:
P := 1; /* Initialize P and I */
I := 1;
while I <= N do /* Loop test */
begin
P := P*I; /* Update P and I */
I := I+1
end;
FactN := P; /* Return result */

shows the Prolog translation constructed from this Pascal while loop.
/* Program ch0104e07.pro */

PREDICATES
factorial(unsigned,long)
factorial_aux(unsigned,long,unsigned,long)

/* Numbers likely to become large are declared as longs. */

CLAUSES
factorial(N, FactN):-
factorial_aux(N,FactN,1,1).

factorial_aux(N,FactN,I,P):-
I <= N,!,
NewP = P * I,
NewI = I + 1,
factorial_aux(N, FactN, NewI, NewP).

128 Visual Prolog

 factorial_aux(N, FactN, I, P) :-
I > N,
FactN = P.

Let's look at this in greater detail.

The factorial clause has only N and FactN as arguments; they are its input and
output, from the viewpoint of someone who is using it to find a factorial. A
second clause, factorial_aux(N, FactN, I, P), actually performs the recursion;
its four arguments are the four variables that need to be passed along from each
step to the next. So factorial simply invokes factorial_aux, passing to it N and
FactN, along with the initial values for I and P, like so:
factorial(N, FactN) :-
factorial_aux(N, FactN, 1, 1).

That's how I and P get initialized.

How can factorial "pass along" FactN? It doesn't even have a value yet! The
answer is that, conceptually, all Visual Prolog's doing here is unifying a variable
called FactN in one clause with a variable called FactN in another clause. The
same thing will happen whenever factorial_aux passes FactN to itself as an
argument in a recursive call. Eventually, the last FactN will get a value, and,
when this happens, all the other FactN's, having been unified with it, will get the
same value. We said "conceptually" above, because in reality there is only one
FactN. Visual Prolog can determine from the source code that FactN is never
really used before the second clause for factorial_aux, and just shuffles the same
FactN around all the time.
Now for factorial_aux. Ordinarily, this predicate will check that I is less than or
equal to N--the condition for continuing the loop--and then call itself recursively
with new values for I and P. Here another peculiarity of Prolog asserts itself. In
Prolog there is no assignment statement such as
P = P + 1

which is found in most other programming languages. You can't change the
value of a Prolog variable. In Prolog, the above is as absurd as in algebra, and
will fail. Instead, you have to create a new variable and say something like
NewP = P + 1

So here's the first clause:

Chapter 94, Repetition and recursion 129

 factorial_aux(N, FactN, I, P) :-
I <= N, !,
NewP = P*I,
NewI = I+1,
factorial_aux(N, FactN, NewI, NewP).

As in cutcount2, the cut enables last-call optimization to take effect, even though
the clause isn't the last in the predicate.
Eventually I will exceed N. When it does, processing should unify the current
value of P with FactN and stop the recursion. This is done in the second clause,
which will be reached when the test I <= N in the first clause fails:
factorial_aux(N, FactN, I, P) :-
I > N,
FactN = P.

But there is no need for FactN = P to be a separate step; the unification can be
performed in the argument list. Putting the same variable name in the positions
occupied by FactN and P requires the arguments in these positions to be matched
with each other. Moreover, the test I > N is redundant since the opposite has been
tested for in the first clause. This gives the final clause:
factorial_aux(_, FactN, _, FactN).

Exercises
The following is a more elegant version of factorial.
/* Program ch0105e08.pro */

PREDICATES
factorial(unsigned,real)
factorial(unsigned,real,unsigned,real)

/* Numbers likely to become large are declared as reals. */

CLAUSES
factorial(N,FactN):-
factorial(N,FactN,1,1).

factorial(N,FactN,N,FactN):-!.
factorial(N,FactN,I,P):-
NewI = I+1,
NewP = P*NewI,
factorial(N, FactN, NewI, NewP).

130 Visual Prolog

 Load and run this program. Carefully look at the code in the second clause of
factorial/4. It takes advantage of the fact that the first time it's called the
counter variable I always has the value 1. This allows the multiplication step
to be carried out with the incremented counter variable NewI rather than I,
saving one recursion/iteration. This is reflected in the first clause.
Write a tail recursive program that behaves like 2 but doesn't use backtracking.
Write a tail recursive program that prints a table of powers of 2, like this:
N 2^N
-- -----
1 2
2 4
3 8
4 16
... ...
10 1024

Make it stop at 10 as shown here.

Write a tail recursive program that accepts a number as input and can end in
either of two ways. It will start multiplying the number by itself over and over
until it either reaches 81 or reaches a number greater than 100. If it reaches 81,
it will print yes; if it exceeds 100, it will print no.

Recursive Data Structures

Not only can rules be recursive; so can data structures. Prolog is the only widely
used programming language that allows you to define recursive data types. A data
type is recursive if it allows structures to contain other structures like themselves.
The most basic recursive data type is the list, although it doesn't immediately look
recursively constructed. A lot of list-processing power is built into Prolog, but we
won't discuss it here; lists are such an important part of Prolog that there is a
whole chapter devoted to them, chapter 106.
In this chapter, we invent a recursive data type, implement it, and use it to write a
very fast sorting program. The structure of this invented recursive data type is a
tree (Figure 107.1). Crucially, each branch of the tree is itself a tree; that's why
the structure is recursive.

Chapter 94, Repetition and recursion 131

 Figure 108.2: Part of a Family Tree1093

Trees as a Data Type

Recursive types were popularized by Niklaus Wirth in Algorithms + Data
Structures = Programs. Wirth derived Pascal from ALGOL60 and published this
work in the early 70's. He didn't implement recursive types in Pascal, but he did
discuss what it would be like to have them. If Pascal had recursive types, you
would be able to define a tree as something like this:
tree = record /* Not correct Pascal! */
name: string[80];
left, right: tree
end;

This code, translated into natural language, means "A tree consists of a name,
which is a string, and the left and right subtrees, which are trees."
The nearest approach to this in Pascal is to use pointers and say
treeptr = ^tree;

tree = record
name: string[80];
left, right: treeptr
end;

132 Visual Prolog

But notice a subtle difference: This code deals with the memory representation of
a tree, not the structure of the tree itself. It treats the tree as consisting of cells,
each containing some data plus pointers to two more cells.
Visual Prolog allows truly recursive type definitions in which the pointers are
created and maintained automatically. For example, you can define a tree as
follows:
DOMAINS
treetype = tree(string, treetype, treetype)

This declaration says that a tree will be written as the functor, tree, whose
arguments are a string and two more trees.
But this isn't quite right; it provides no way to end the recursion, and, in real life,
the tree does not go on forever. Some cells don't have links to further trees. In
Pascal, you could express this by setting some pointers equal to the special value
nil, but pointers are an implementation issue that ordinarily doesn't surface in
Prolog source code. Rather, in Prolog we define two kinds of trees: ordinary ones
and empty ones. This is done by allowing a tree to have either of two functors:
tree, with three arguments, or empty, with no arguments.
DOMAINS
treetype = tree(string, treetype, treetype) ; empty

Notice that the names tree (a functor that takes three arguments) and empty (a
functor taking no arguments) are created by the programmer; neither of them has
any pre-defined meaning in Prolog. You could equally well have used xxx and
yyy.
This is how the tree in Figure 110.4 could appear in a Prolog program:
tree("Cathy",
tree("Michael"
tree("Charles", empty, empty)
tree("Hazel", empty, empty))
tree("Melody"
tree("Jim", empty, empty)
tree("Eleanor", empty, empty)))

This is indented here for readability, but Prolog does not require indentation, nor
are trees indented when you print them out normally. Another way of setting up
this same data structure is:

Chapter 94, Repetition and recursion 133

 tree("Cathy"
tree("Michael", tree("Charles", empty, empty), tree("Hazel", empty,
empty))
tree("Melody", tree("Jim", empty, empty), tree("Eleanor", empty,
empty)))

Note that this is not a Prolog clause; it is just a complex data structure.

Traversing a Tree
Before going on to the discussion of how to create trees, first consider what you'll
do with a tree once you have it. One of the most frequent tree operations is to
examine all the cells and process them in some way, either searching for a
particular value or collecting all the values. This is known as traversing the tree.
One basic algorithm for doing so is the following:
If the tree is empty, do nothing.
Otherwise, process the current node, then traverse the left subtree, then traverse
the right subtree.
Like the tree itself, the algorithm is recursive: it treats the left and right subtrees
exactly like the original tree. Prolog expresses it with two clauses, one for empty
and one for nonempty trees:
traverse(empty). /* do nothing */

traverse(tree(X, Y, Z)) :-
do something with X,
traverse(Y),
traverse(Z).

134 Visual Prolog

 Figure 111.5: Depth-First Traversal of the Tree in Figure 112.6113114

This tree traversal algorithm is known as depth-first search because it goes as far
as possible down each branch before backing up and trying another branch
(Figure 115.7). To see it in action, look at program 9, which traverses a tree and
prints all the elements as it encounters them. Given the tree in Figures 116.8 and
117.9, 9 prints
Cathy
Michael
Charles
Hazel
Melody
Jim
Eleanor

Of course, you could easily adapt the program to perform some other operation
on the elements, rather than printing them.
/* Program ch0118e09.pro */

/* Traversing a tree by depth-first search

and printing each element as it is encountered */

DOMAINS
treetype = tree(string, treetype, treetype) ; empty()

Chapter 94, Repetition and recursion 135

PREDICATES
traverse(treetype)

CLAUSES
traverse(empty).

traverse(tree(Name,Left,Right)):-
write(Name,'\n'),
traverse(Left),
traverse(Right).

GOAL
traverse(tree("Cathy",
tree("Michael",
tree("Charles", empty, empty),
tree("Hazel", empty, empty)),
tree("Melody",
tree("Jim", empty, empty),
tree("Eleanor", empty, empty)))).

Depth-first search is strikingly similar to the way Prolog searches a knowledge

base, arranging the clauses into a tree and pursuing each branch until a query
fails. If you wanted to, you could describe the tree by means of a set of Prolog
clauses such as:
father_of("Cathy", "Michael").
mother_of("Cathy", "Melody").
father_of("Michael", "Charles").
mother_of("Michael", "Hazel").
...

This is preferable if the only purpose of the tree is to express relationships

between individuals. But this kind of description makes it impossible to treat the
whole tree as a single complex data structure; as you'll see, complex data
structures are very useful because they simplify difficult computational tasks.

Creating a Tree
One way to create a tree is to write down a nested structure of functors and
arguments, as in the preceding example (Program 9). Ordinarily, however, Prolog
creates trees by computation. In each step, an empty subtree is replaced by a
nonempty one through Prolog's process of unification (argument matching).
Creating a one-cell tree from an ordinary data item is trivial:
create_tree(N, tree(N, empty, empty)).

136 Visual Prolog

This says: "If N is a data item, then tree(N, empty, empty) is a one-cell tree
containing it."
Building a tree structure is almost as easy. The following procedure takes three
trees as arguments. It inserts the first tree as the left subtree of the second tree,
giving the third tree as the result:
insert_left(X, tree(A, _, B), tree(A, X, B)).

Notice that this rule has no body--there are no explicit steps in executing it. All
the computer has to do is match the arguments with each other in the proper
positions, and the work is done.
Suppose, for example, you want to insert tree("Michael", empty, empty) as the
left subtree of tree("Cathy", empty, empty). To do this, just execute the goal
insert_left(tree("Michael", empty, empty),
tree("Cathy", empty, empty),
T)

and T immediately takes on the value

tree("Cathy", tree("Michael", empty, empty), empty).

This gives a way to build up trees step-by-step. Program 10 demonstrates this

technique. In real life, the items to be inserted into the tree could come from
external input.
/* Program ch0119e10.pro */

/* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
* Simple tree-building procedures *
* create_tree(A, B) puts A in the data field of a one-cell tree *
* giving B insert_left(A, B, C) inserts A as left subtree of B *
* giving C insert_right(A, B, C) inserts A as right subtree of B *
* giving C *
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * */

DOMAINS
treetype = tree(string,treetype,treetype) ; empty()

PREDICATES
create_tree(string,treetype)
insert_left(treetype,treetype,treetype)
insert_right(treetype, treetype, treetype)

Chapter 94, Repetition and recursion 137

CLAUSES
create_tree(A,tree(A,empty,empty)).
insert_left(X,tree(A,_,B),tree(A,X,B)).
insert_right(X,tree(A,B,_),tree(A,B,X)).

GOAL
/* First create some one-cell trees... */

create_tree("Charles",Ch),
create_tree("Hazel",H),
create_tree("Michael",Mi),
create_tree("Jim",J),
create_tree("Eleanor",E),
create_tree("Melody",Me),
create_tree("Cathy",Ca),

/* ...then link them up... */

insert_left(Ch, Mi, Mi2),

insert_right(H, Mi2, Mi3),
insert_left(J, Me, Me2),
insert_right(E, Me2, Me3),
insert_left(Mi3, Ca, Ca2),
insert_right(Me3, Ca2, Ca3),

/* ...and print the result. */

write(Ca3,'\n').

Notice that there is no way to change the value of a Prolog variable once it is
bound. That's why 10 uses so many variable names; every time you create a new
value, you need a new variable. The large number of variable names here is
unusual; more commonly, repetitive procedures obtain new variables by invoking
themselves recursively, since each invocation has a distinct set of variables.

Binary Search Trees

So far, we have been using the tree to represent relationships between its
elements. Of course, this is not the best use for trees, since Prolog clauses can do
the same job. But trees have other uses.
Trees provide a good way to store data items so that they can be found quickly. A
tree built for this purpose is called a search tree; from the user's point of view, the
tree structure carries no information--the tree is merely a faster alternative to a list
or array. Recall that, to traverse an ordinary tree, you look at the current cell and
then at both of its subtrees. To find a particular item, you might have to look at
every cell in the whole tree.

138 Visual Prolog

 The time taken to search an ordinary tree with N elements is, on the
average, proportional to N.

A binary search tree is constructed so that you can predict, upon looking at any
cell, which of its subtrees a given item will be in. This is done by defining an
ordering relation on the data items, such as alphabetical or numerical order. Items
in the left subtree precede the item in the current cell and, in the right subtree,
they follow it. Figure 120.10 shows an example. Note that the same names, added
in a different order, would produce a somewhat different tree. Notice also that,
although there are ten names in the tree, you can find any of them in--at most--
five steps.

Figure 121.11: Binary Search Tree12212

Every time you look at a cell in a binary search tree during a search, you
eliminate half the remaining cells from consideration, and the search proceeds
very quickly. If the size of the tree were doubled, then, typically, only one extra
step would be needed to search it.
The time taken to find an item in a binary search tree is, on the average,
proportional to log2 N (or, in fact, proportional to log N with logarithms to
any base).

Chapter 94, Repetition and recursion 139

To build the tree, you start with an empty tree and add items one by one. The
procedure for adding an item is the same as for finding one: you simply search for
the place where it ought to be, and insert it there. The algorithm is as follows:
If the current node is an empty tree, insert the item there.
Otherwise, compare the item to be inserted and the item stored in the current
node. Insert the item into the left subtree or the right subtree, depending on the
result of the comparison.
In Prolog, this requires three clauses, one for each situation. The first clause is
insert(NewItem, empty, tree(NewItem, empty, empty) :- !.

Translated to natural language, this code says "The result of inserting NewItem
into empty is tree(NewItem, empty, empty)." The cut ensures that, if this clause
can be used successfully, no other clauses will be tried.
The second and third clauses take care of insertion into nonempty trees:
insert(NewItem, tree(Element, Left, Right), tree(Element, NewLeft,
Right) :-
NewItem < Element,
!,
insert(NewItem, Left, NewLeft).

insert(NewItem, tree(Element, Left, Right), tree(Element, Left,

NewRight) :-
insert(NewItem, Right, NewRight).

If NewItem < Element, you insert it into the left subtree; otherwise, you insert it
into the right subtree. Notice that, because of the cuts, you get to the third clause
only if neither of the preceding clauses has succeeded. Also notice how much of
the work is done by matching arguments in the head of the rule.

Tree-Based Sorting
Once you have built the tree, it is easy to retrieve all the items in alphabetical
order. The algorithm is again a variant of depth-first search:
If the tree is empty, do nothing.
Otherwise, retrieve all the items in the left subtree, then the current element, then
all the items in the right subtree.
Or, in Prolog:
retrieve_all(empty). /* Do nothing */

140 Visual Prolog

 retrieve_all(tree(Item, Left, Right)) :-
retrieve_all(Left),
do_something_to(Item),
retrieve_all(Right).

You can sort a sequence of items by inserting them into a tree and then retrieving
them in order. For N items, this takes time proportional to N log N, because both
insertion and retrieval take time proportional to log N, and each of them has to be
done N times. This is the fastest known sorting algorithm.

Example
Program 11 uses this technique to alphabetize character input. In this example we
use some of Visual Prolog's standard predicates we haven't introduced before.
These predicates will be discussed in detail in later chapters.
/* Program ch0123e11.pro */

DOMAINS
chartree = tree(char, chartree, chartree); end

PREDICATES
nondeterm do(chartree)
action(char, chartree, chartree)
create_tree(chartree, chartree)
insert(char, chartree, chartree)
write_tree(chartree)
nondeterm repeat

CLAUSES
do(Tree):-
repeat,nl,
write("*****************************************************"),nl,
write("Enter 1 to update tree\n"),
write("Enter 2 to show tree\n"),
write("Enter 7 to exit\n"),
write("*****************************************************"),nl,
write("Enter number - "),
readchar(X),nl,
action(X, Tree, NewTree),
do(NewTree).

Chapter 94, Repetition and recursion 141

 action('1',Tree,NewTree):-
write("Enter characters or # to end: "),
create_Tree(Tree, NewTree).
action('2',Tree,Tree):-
write_Tree(Tree),
write("\nPress a key to continue"),
readchar(_),nl.
action('7', _, end):-
exit.

create_Tree(Tree, NewTree):-
readchar(C),
C<>'#',!,
write(C, " "),
insert(C, Tree, TempTree),
create_Tree(TempTree, NewTree).
create_Tree(Tree, Tree).

insert(New,end,tree(New,end,end)):-!.
insert(New,tree(Element,Left,Right),tree(Element,NewLeft,Right)):-
New<Element,!,
insert(New,Left,NewLeft).
insert(New,tree(Element,Left,Right),tree(Element,Left,NewRight)):-
insert(New,Right,NewRight).

write_Tree(end).
write_Tree(tree(Item,Left,Right)):-
write_Tree(Left),
write(Item, " "),
write_Tree(Right).

repeat.
repeat:-repeat.

GOAL
write("*************** Character tree sort *******************"),nl,
do(end).

Load and run Program 11 and watch how Visual Prolog does tree-based sorting
on a sequence of characters.

Exercises
Program 12 is similar to 11, but more complex. It uses the same sorting technique
to alphabetize any standard text file, line by line. Typically it's more than five
times faster than "SORT.EXE", the sort program provided by DOS and OS/2,

142 Visual Prolog

 but it's beaten by the highly optimized "sort" on UNIX. Nevertheless, tree-
based sorting is remarkably efficient.
In this example we use some of the predicates from Visual Prolog's file
system, to give you a taste of file redirection. To redirect input or output to a
file, you must tell the system about the file; you use openread to read from the
file or openwrite to write to it. Once files are open, you can switch I/O
between an open file and the screen with writedevice, and between an open
file and the keyboard with readdevice. These predicates are discussed in detail
later in chapter 124.
Load and run Program 12. When it prompts File to read type in the name of an
existing text file; the program will then alphabetize that file, line by line.

/* Program ch0125e12.pro */

DOMAINS
treetype = tree(string, treetype, treetype) ; empty
file = infile ; outfile

PREDICATES
main
read_input(treetype)
read_input_aux(treetype, treetype)
insert(string, treetype, treetype)
write_output(treetype)

CLAUSES
main :-
write("PDC Prolog Treesort"),nl,
write("File to read: "),
readln(In),nl,
openread(infile, In), /* open the specified file for reading */
write("File to write: "),
readln(Out),nl,
openwrite(outfile, Out),
readdevice(infile),
/* redirect all read operations to the opened file */
read_input(Tree),
writedevice(outfile),
/* redirect all write operations to the opened file */
write_output(Tree),
closefile(infile), /* close the file opened for reading */
closefile(outfile).

Chapter 94, Repetition and recursion 143

 /* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
* read_input(Tree) *
* reads lines from the current input device until EOF, then *
* instantiates Tree to the binary search tree built *
* therefrom *
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * */

read_input(Tree):-
read_input_aux(empty,Tree).

/* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
* read_input_aux(Tree, NewTree) *
* reads a line, inserts it into Tree giving NewTree, *
* and calls itself recursively unless at EOF. *
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * */

read_input_aux(Tree, NewTree):-
readln(S),
!,
insert(S, Tree, Tree1),
read_input_aux(Tree1, NewTree).

read_input_aux(Tree, Tree). /* The first clause fails at EOF. */

/* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
* insert(Element, Tree, NewTree) *
* inserts Element into Tree giving NewTree. *
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * */

insert(NewItem, empty, tree(NewItem,empty,empty)):-!.

insert(NewItem,tree(Element,Left,Right),tree(Element,NewLeft, Right)):-
NewItem < Element,
!,
insert(NewItem, Left, NewLeft).

insert(NewItem,tree(Element,Left,Right),tree(Element,Left,NewRight)):-
insert(NewItem, Right, NewRight).

/* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
* write_output(Tree) *
* writes out the elements of Tree in alphabetical order. *
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * */

write_output(empty). /* Do nothing */

write_output(tree(Item,Left,Right)):-
write_output(Left),
write(Item), nl,
write_output(Right).

144 Visual Prolog

GOAL
main,nl.

Use recursive data structures to implement hypertext. A hypertext is a structure in

which each entry, made up of several lines of text, is accompanied by pointers
to several other entries. Any entry can be connected to any other entry; for
instance, you could get to an entry about Abraham Lincoln either from
"Presidents" or from "Civil War."
To keep things simple, use one-line entries (strings) and let each of them
contain a pointer to only one other entry.
Hint: Start with
DOMAINS
entrytype = empty() ; entry(string, entry)

Build a linked structure in which most of the entries have a nonempty second
argument.
Now, take your hypertext implementation and redo it using Prolog clauses. That
is, use clauses (rather than recursive data structures) to record which entry
follows which.

Summary

These are the major points covered in this chapter:

In Prolog there are two ways to repeat the same clause; through backtracking and
recursion. By failing, Prolog will backtrack to find a new piece of data and
repeat the clause until there are no more options. Recursion is the process of a
clause calling itself.
Backtracking is very powerful and memory efficient, but variables are freed after
each iteration, so their values are lost. Recursion allows variables to be
incremented, but it is not memory efficient.
However, Visual Prolog does tail recursion elimination, which relieves the
memory demands of recursion. For Visual Prolog to achieve tail recursion
elimination, the recursive call must be the last subgoal in the clause body.

Chapter 94, Repetition and recursion 145

CHAPTER 126

Lists and Recursion

127List processing--handling objects that contain an arbitrary number of
elements--is a powerful technique in Prolog. In this chapter, we explain what lists
are and how to declare them, then give several examples that show how you
might use list processing in your own applications. We also define two well-
known Prolog predicates--member and append--while looking at list processing
from both a recursive and a procedural standpoint.
After that, we introduce findall, a Visual Prolog standard predicate that enables
you to find and collect all solutions to a single internal goal. We round out this
chapter with a discussion of compound lists--combinations of different types of
elements--and an example of parsing by difference lists.

What Is a List?

In Prolog, a list is an object that contains an arbitrary number of other objects

within it. Lists correspond roughly to arrays in other languages, but, unlike an
array, a list does not require you to declare how big it will be before you use it.
There are other ways to combine several objects into one, of course. If the
number of objects to be combined is known in advance, you can make them the
arguments of a single compound data structure. And even if the number of objects
is unpredictable, you can use a recursive compound data structure, such as a tree.
But lists are usually easier to use because the language provides a concise
notation for them.
A list that contains the numbers 1, 2, and 3 is written as
[ 1, 2, 3 ]

Each item contained in the list is known as an element. To form a list data
structure, you separate the elements of a list with commas and then enclose them
in square brackets. Here are some examples:
[dog, cat, canary]
["valerie ann", "jennifer caitlin", "benjamin thomas"]

146 Visual Prolog

Declaring Lists
To declare the domain for a list of integers, you use the domains declaration, like
this:
DOMAINS
integerlist = integer*

The asterisk means "list of"; that is, integer* means "list of integers."
Note that the word list has no special meaning in Visual Prolog. You could
equally well have called your list domain zanzibar. It's the asterisk, not the name,
that signifies a list domain.
The elements in a list can be anything, including other lists. However, all
elements in a list must belong to the same domain, and in addition to the
declaration of the list domain there must be a domains declaration for the
elements:
DOMAINS
elementlist = elements*
elements = ....

Here elements must be equated to a single domain type (for example: integer,
real, or symbol) or to a set of alternatives marked with different functors. Visual
Prolog does not allow you to mix standard types in a list. For example, the
following declarations would not properly indicate a list made up of integers,
reals, and symbols:
elementlist = elements*
elements = integer; real; symbol /* Incorrect */

The way to declare a list made up of integers, reals, and symbols is to define a
single domain comprising all three types, with functors to show which type a
particular element belongs to. For example:
elementlist = elements*
elements = i(integer); r(real); s(symbol)
/* the functors are i, r, and s */

(For more information about this, refer to "Compound Lists" later in this chapter.)

Heads and Tails

A list is really a recursive compound object. It consists of two parts: the head, of
list which is the first element, and the tail, which is a list comprising all the

Chapter 128, Lists and Recursion 147

subsequent elements. The tail of a list is always a list; the head of a list is an
element. For example,
the head of [a, b, c] is a
the tail of [a, b, c] is [b, c]

What happens when you get down to a one-element list? The answer is that
the head of [c] is c
the tail of [c] is []
If you take the first element from the tail of a list enough times, you'll
eventually get down to the empty list ([ ]).
The empty list can't be broken into head and tail.
This means that, conceptually, lists have a tree structure just like other compound
objects. The tree structure of [a, b, c, d] is
list
/ \
a list
/ \
b list
/ \
c list
/ \
d []

Further, a one-element list such as [a] is not the same as the element that it
contains because, simple as it looks, [a] is really the compound data structure
shown here:
list
/ \
a []

List Processing

Prolog provides a way to make the head and tail of a list explicit. Instead of
separating elements with commas, you can separate the head and tail with a
vertical bar (|). For instance,
[a, b, c] is equivalent to [a|[b, c]]
and, continuing the process,

148 Visual Prolog

 [a|[b, c]] is equivalent to [a|[b|[c]]]
which is equivalent to [a|[b|[c|[]]]]
You can even use both kinds of separators in the same list, provided the vertical
bar is the last separator. So, if you really want to, you can write [a, b, c, d] as
[a, b|[c, d]]. Table 129.1 gives more examples.

Table 130.2: Heads and Tails of Lists1313

List Head Tail

['a', 'b', 'c'] 'a' ['b', 'c']
[ 'a' ] 'a' [] /* an empty list */
[ ] undefined undefined
[[1, 2, 3], [2, 3, 4], []] [1, 2, 3] [[2, 3, 4], []]

Table 132.4 gives several examples of list unification.

Table 133.5: Unification of Lists1346

List 1 List 2 Variable Binding

[X, Y, Z] [egbert, eats, icecream] X=egbert, Y=eats, Z=icecream
[7] [X | Y] X=7, Y=[]
[1, 2, 3, 4] [X, Y | Z] X=1, Y=2, Z=[3,4]
[1, 2] [3 | X] fail

Using Lists

Because a list is really a recursive compound data structure, you need recursive
algorithms to process it. The most basic way to process a list is to work through
it, doing something to each element until you reach the end.
An algorithm of this kind usually needs two clauses. One of them says what to do
with an ordinary list (one that can be divided into a head and a tail). The other
says what to do with an empty list.

Chapter 128, Lists and Recursion 149

Writing Lists
For example, if you just want to print out the elements of the list, here's what you
do:
/* Program ch0135e01.pro */

DOMAINS
list = integer* /* or whatever type you wish to use */

PREDICATES
write_a_list(list)

CLAUSES
write_a_list([]). /* If the list is empty, do nothing more. */

write_a_list([H|T]):-
/* Match the head to H and the tail to T, then... */
write(H),nl,
write_a_list(T).

GOAL
write_a_list([1, 2, 3]).

Here are the two write_a_list clauses described in natural language:

To write an empty list, do nothing.

Otherwise, to write a list, write its head (which

is a single element), then write its tail (a list).

The first time through, the goal is:

write_a_list([1, 2, 3]).

This matches the second clause, with H=1 and T=[2, 3]; this writes 1 and then
calls write_a_list recursively with the tail of the list:
write_a_list([2, 3]). /* This is write_a_list(T). */

This recursive call matches the second clause, this time with H=2 and T=[3], so it
writes 2 and again calls write_a_list recursively:
write_a_list([3]).

150 Visual Prolog

Now, which clause will this goal match? Recall that, even though the list [3] has
only one element, it does have a head and tail; the head is 3 and the tail is []. So
again the goal matches the second clause, with H=3 and T=[]. Hence, 3 is written
and write_a_list is called recursively like this:
write_a_list([]).

Now you see why this program needs the first clause. The second clause won't
match this goal because [] can't be divided into head and tail. So, if the first
clause weren't there, the goal would fail. As it is, the first clause matches and the
goal succeeds without doing anything further.

Exercise
Is write_a_list tail-recursive? Would it be if the two clauses were written in the
opposite order?

Counting List Elements

Now consider how you might find out how many elements are in a list. What is
the length of a list, anyway? Here's a simple logical definition:
The length of [] is 0.
The length of any other list is 1 plus the length of its tail.
Can you implement this? In Prolog it's very easy. It takes just two clauses:
/* Program ch0136e02.pro */

DOMAINS
list = integer* /* or whatever type you want to use */

PREDICATES
length_of(list,integer)

CLAUSES
length_of([], 0).
length_of([_|T],L):-
length_of(T,TailLength),
L = TailLength + 1.

Take a look at the second clause first. Crucially, [_|T] will match any nonempty
list, binding T to the tail of the list. The value of the head is unimportant; as long
as it exists, it can be counted it as one element.
So the goal:

Chapter 128, Lists and Recursion 151

 length_of([1, 2, 3], L)

will match the second clause, with T=[2, 3]. The next step is to compute the
length of T. When this is done (never mind how), TailLength will get the value 2,
and the computer can then add 1 to it and bind L to 3.
So how is the middle step executed? That step was to find the length of [2, 3] by
satisfying the goal
length_of([2, 3], TailLength).

In other words, length_of calls itself recursively. This goal matches the second
clause, binding
[3] in the goal to T in the clause and
TailLength in the goal to L in the clause.
Recall that TailLength in the goal will not interfere with TailLength in the clause,
because each recursive invocation of a clause has its own set of variables. If this
is unclear, review the section on recursion in chapter 137.
So now the problem is to find the length of [3], which will be 1, and then add 1 to
that to get the length of [2, 3], which will be 2. So far, so good.
Likewise, length_of will call itself recursively again to get the length of [3]. The
tail of [3] is [], so T is bound to [], and the problem is to get the length of [],
then add 1 to it, giving the length of [3].
This time it's easy. The goal
length_of([], TailLength)

matches the first clause, binding TailLength to 0. So now the computer can add 1
to that, giving the length of [3], and return to the calling clause. This, in turn, will
add 1 again, giving the length of [2, 3], and return to the clause that called it; this
original clause will add 1 again, giving the length of [1, 2, 3].
Confused yet? We hope not. In the following brief illustration we'll summarize
the calls. We've used subscripts to indicate that similarly-named variables in
different clauses--or different invocations of the same clause--are distinct.
length_of([1, 2, 3], L1).
length_of([2, 3], L2).
length_of([3], L3).
length_of([], 0).
L3 = 0+1 = 1.
L2 = L3+1 = 2.

152 Visual Prolog

 L1 = L2+1 = 3.

Exercises
What happens when you satisfy the following goal?
length_of(X, 3), !.

Does the goal succeed, and if so, what is bound to X? Why? (Work through
carefully by hand to see how this works.)
Write a predicate called sum_of that works exactly like length_of, except that it
takes a list of numbers and adds them up. For example, the goal:
sum_of([1, 2, 3, 4], S).

should bind S to 10.

What happens if you execute this goal?
sum_of(List, 10).

This goal says, "Give me a list whose elements add up to 10." Can Visual
Prolog do this? If not, why not? (Hint: It's not possible to do arithmetic on
unbound variables in Prolog.)

Tail Recursion Revisited

You probably noticed that length_of is not, and can't be, tail-recursive, because
the recursive call is not the last step in its clause. Can you create a tail-recursive
list-length predicate? Yes, but it will take some effort.
The problem with length_of is that you can't compute the length of a list until
you've already computed the length of the tail. It turns out there's a way around
this. You'll need a list-length predicate with three arguments.
One is the list, which the computer will whittle away on each call until it
eventually becomes empty, just as before.
Another is a free argument that will ultimately contain the result (the length).
The third is a counter that starts out as 0 and increments on each call.
When the list is finally empty, you'll unify the counter with the (up to then)
unbound result.
/* Program ch0138e03.pro */

DOMAINS
list = integer* /* or whatever type you want to use */

Chapter 128, Lists and Recursion 153

PREDICATES
length_of(list,integer,integer)

CLAUSES
length_of([], Result, Result).
length_of([_|T],Result,Counter):-
NewCounter = Counter + 1,
length_of(T, Result, NewCounter).

GOAL
length_of([1, 2, 3], L, 0), /* start with Counter = 0 */
write("L=",L), nl.

This version of the length_of predicate is more complicated, and in many ways
less logical, than the previous one. We've presented it merely to show you that, by
devious means, you can often find a tail-recursive algorithm for a problem that
seems to demand a different type of recursion.

Exercises
Try both versions of length_of on enormous lists (lists with perhaps 200 to 500
elements). What happens? On long lists, how do they compare in speed?
What happens with the new version of length_of if you give the following goal?
length_of(MyList, 5, 0).

Hint: You are discovering a very important property of Prolog called

interchangeability of unknowns. Not all Prolog predicates have it.
Rewrite sum_of to work like the new version of length_of.

Another Example -- Modifying the List

Sometimes you want to take a list and create another list from it. You do this by
working through the list element by element, replacing each element with a
computed value. For example, here is a program that takes a list of numbers and
adds 1 to each of them:
/* Program ch0139e04.pro */

DOMAINS
list = integer*

PREDICATES
add1(list,list)

154 Visual Prolog

CLAUSES
add1([], []). /* boundary condition */
add1([Head|Tail],[Head1|Tail1]):- /* separate the head */
/* from the rest of the list */
Head1= Head+1 /* add 1 to the first element */
add1(Tail,Tail1). /* call element with the rest of the list */

To paraphrase this in natural language:

To add 1 to all the elements of the empty list,
just produce another empty list.

To add 1 to all the elements of any other list,

add 1 to the head and make it the head of the result, and then
add 1 to each element of the tail and make that the tail of the
result.

Load the program, and enter the goal add1([1,2,3,4], NewList).

Visual Prolog will return

NewList=[2,3,4,5]
1 Solution

Tail Recursion Again

Is add1 tail-recursive? If you're accustomed to using Lisp or Pascal, you might
think it isn't, because you think of it as performing the following operations:
Split the list into Head and Tail.
Add 1 to Head, giving Head1.
Recursively add 1 to all the elements of Tail, giving Tail1.
Combine Head1 and Tail1, giving the resulting list.
This isn't tail-recursive, because the recursive call is not the last step.
But--and this is important--that is not how Prolog does it. In Visual Prolog, add1
is tail-recursive, because its steps are really the following:
Bind the head and tail of the original list to Head and Tail.
Bind the head and tail of the result to Head1 and Tail1. (Head1 and Tail1 do not
have values yet.)
Add 1 to Head, giving Head1.
Recursively add 1 to all the elements of Tail, giving Tail1.

Chapter 128, Lists and Recursion 155

When this is done, Head1 and Tail1 are already the head and tail of the result;
there is no separate operation of combining them. So the recursive call really is
the last step.

More on Modifying Lists

Of course you don't actually need to put in a replacement for every element.
Here's a program that scans a list of numbers and copies it, leaving out the
negative numbers:
/* Program ch0140e05.pro */

DOMAINS
list = integer*

PREDICATES
discard_negatives(list, list)

CLAUSES
discard_negatives([], []).

discard_negatives([H|T],ProcessedTail):-
H < 0, /* If H is negative, just skip it */
!,
discard_negatives(T, ProcessedTail).

discard_negatives([H|T],[H|ProcessedTail]):-
discard_negatives(T, ProcessedTail).

For example, the goal

discard_negatives([2, -45, 3, 468], X)

gives X=[2, 3, 468].

And here's a predicate that copies the elements of a list, making each element
occur twice:
doubletalk([], []).

doubletalk([H|T], [H, H|DoubledTail]) :-

doubletalk(T, DoubledTail).

List Membership
Suppose you have a list with the names John, Leonard, Eric, and Frank and
would like to use Visual Prolog to investigate if a given name is in this list. In

156 Visual Prolog

other words, you must express the relation "membership" between two
arguments: a name and a list of names. This corresponds to the predicate
member(name, namelist). /* "name" is a member of "namelist" */

In Program 6, the first clause investigates the head of the list. If the head of the
list is equal to the name you're searching for, then you can conclude that Name is
a member of the list. Since the tail of the list is of no interest, it is indicated by the
anonymous variable. Thanks to this first clause, the goal
member(john, [john, leonard, eric, frank])

is satisfied.
/* Program ch0141e06.pro */

DOMAINS
namelist = name*
name = symbol

PREDICATES
nondeterm member(name, namelist)

CLAUSES
member(Name, [Name|_]).
member(Name, [_|Tail]):-
member(Name,Tail).

If the head of the list is not equal to Name, you need to investigate whether Name
can be found in the tail of the list.
In English:
Name is a member of the list if Name is the first element
of the list, or
Name is a member of the list if Name is a member of the tail.

The second clause of member relates to this relationship. In Visual Prolog:

member(Name, [_|Tail]) :- member(Name, Tail).

Exercises
Load Program 6 and try the following goal:
member(susan, [ian, susan, john]).

Chapter 128, Lists and Recursion 157

Add domain and predicate statements so you can use member to investigate if a
number is a member of a list of numbers. Try several goals, including
member(X, [1, 2, 3, 4]).

to test your new program.

Does the order of the two clauses for the member predicate have any
significance? Test the behavior of the program when the two rules are
swapped. The difference appears if you test the goal
member(X, [1, 2, 3, 4, 5])

in both situations.

Appending One List to Another: Declarative and

Procedural Programming
As given, the member predicate of Program 6 works in two ways. Consider its
clauses once again:
member(Name, [Name|_]).
member(Name, [_|Tail]) :- member(Name, Tail).

You can look at these clauses from two different points of view: declarative and
procedural.
From a declarative viewpoint, the clauses say
Name is a member of a list if the head is equal to Name;
if not, Name is a member of the list if it is a member of the tail.

From a procedural viewpoint, the two clauses could be interpreted as saying:

To find a member of a list, find its head;
otherwise, find a member of its tail.

These two points of view correspond to the goals

member(2, [1, 2, 3, 4]).

and
member(X, [1, 2, 3, 4]).

In effect, the first goal asks Visual Prolog to check whether something is true; the
second asks Visual Prolog to find all members of the list [1,2,3,4]. Don't be
confused by this. The member predicate is the same in both cases, but its
behavior may be viewed from different angles.

158 Visual Prolog

Recursion from a Procedural Viewpoint
The beauty of Prolog is that, often, when you construct the clauses for a predicate
from one point of view, they'll work from the other. To see this duality, in this
next example you'll construct a predicate to append one list to another. You'll
define the predicate append with three arguments:
append(List1, List2, List3)

This combines List1 and List2 to form List3. Once again you are using recursion
(this time from a procedural point of view).
If List1 is empty, the result of appending List1 and List2 will be the same as List2.
In Prolog:
append([], List2, List2).

If List1 is not empty, you can combine List1 and List2 to form List3 by making
the head of List1 the head of List3. (In the following code, the variable H is used
as the head of both List1 and List3.) The tail of List3 is L3, which is composed of
the rest of List1 (namely, L1) and all of List2. In Prolog:
append([H|L1], List2, [H|L3]) :-
append(L1, List2, L3).

The append predicate operates as follows: While List1 is not empty, the recursive
rule transfers one element at a time to List3. When List1 is empty, the first clause
ensures that List2 hooks onto the back of List3.

Exercise
The predicate append is defined in Program 7. Load the program.
/* Program ch0142e07.pro */

DOMAINS
integerlist = integer*

PREDICATES
append(integerlist,integerlist,integerlist)

CLAUSES
append([],List,List).
append([H|L1],List2,[H|L3]):-
append(L1,List2,L3).

Now run it with the following goal:

append([1, 2, 3], [5, 6], L).

Chapter 128, Lists and Recursion 159

Now try this goal:
append([1, 2], [3], L), append(L, L, LL).

One Predicate Can Have Different Uses

Looking at append from a declarative point of view, you have defined a relation
between three lists. This relation also holds if List1 and List3 are known but List2
isn't. However, it also holds true if only List3 is known. For example, to find
which two lists could be appended to form a known list, you could use a goal of
the form
append(L1, L2, [1, 2, 4]).

With this goal, Visual Prolog will find these solutions:

L1=[], L2=[1,2,4]
L1=[1], L2=[2,4]
L1=[1,2], L2=[4]
L1=[1,2,4], L2=[]
4 Solutions

You can also use append to find which list you could append to [3,4] to form the
list [1,2,3,4]. Try giving the goal
append(L1, [3,4], [1,2,3,4]).

Visual Prolog finds the solution

L1=[1,2].

This append predicate has defined a relation between an input set and an output
set in such a way that the relation applies both ways. Given that relation, you can
ask
Which output corresponds to this given input?

or
Which input corresponds to this given output?

The status of the arguments to a given predicate when you call that predicate is
referred to as a flow pattern. An argument that is bound or instantiated at the time
of the call is an input argument, signified by (i); a free argument is an output
argument, signified by (o).

160 Visual Prolog

The append predicate has the ability to handle any flow pattern you provide.
However, not all predicates have the capability of being called with different flow
patterns. When a Prolog clause is able to handle multiple flow patterns, it is
known as an invertible clause. When writing your own Visual Prolog clauses,
keep in mind that an invertible clause has this extra advantage and that creating
invertible clauses adds power to the predicates you write.

Exercise
Amend the clauses defining member in Program 6 and construct the clauses for a
predicate even_member that will succeed if you give the goal
even_member(2, [1, 2, 3, 4, 5, 6]).

The predicate should also display the following result:

X=2
X=4
X=6
3 Solutions

given the goal

even_member(X, [1, 2, 3, 4, 5, 6]).

Finding All the Solutions at Once

In chapter 143, we compared backtracking and recursion as ways to perform

repetitive processes. Recursion won out because, unlike backtracking, it can pass
information (through arguments) from one recursive call to the next. Because of
this, a recursive procedure can keep track of partial results or counters as it goes
along.
But there's one thing backtracking can do that recursion can't do--namely, find all
the alternative solutions to a goal. So you may find yourself in a quandary: You
need all the solutions to a goal, but you need them all at once, as part of a single
compound data structure. What do you do?
Fortunately, Visual Prolog provides a way out of this impasse. The built-in
predicate findall takes a goal as one of its arguments and collects all of the
solutions to that goal into a single list. findall takes three arguments:
The first argument, VarName, specifies which argument in the specified
predicate is to be collected into a list.

Chapter 128, Lists and Recursion 161

 The second, mypredicate, indicates the predicate from which the values will
be collected.
The third argument, ListParam, is a variable that holds the list of values
collected through backtracking. Note that there must be a user-defined domain
to which the values of ListParam belong.
Program 8 uses findall to print the average age of a group of people.
/* Program ch0144e08.pro */

DOMAINS
name,address = string
age = integer
list = age*

PREDICATES
nondeterm person(name, address, age)
sumlist(list, age, integer)

CLAUSES
sumlist([],0,0).
sumlist([H|T],Sum,N):-
sumlist(T,S1,N1),
Sum=H+S1, N=1+N1.

person("Sherlock Holmes", "22B Baker Street", 42).

person("Pete Spiers", "Apt. 22, 21st Street", 36).
person("Mary Darrow", "Suite 2, Omega Home", 51).

GOAL
findall(Age,person(_, _, Age),L),
sumlist(L,Sum,N),
Ave = Sum/N,
write("Average=", Ave),nl.

The findall clause in this program creates a list, L, that is a collection of all the
ages obtained from the predicate person. If you wanted to collect a list of all the
people who are 42 years old, you could give the following subgoal:
findall(Who, person(Who, _, 42), List)

Before trying this, please note that it requires the program to contain a domain
declaration for the resulting list:
slist = string*

162 Visual Prolog

Compound Lists

A list of integers can be simply declared as

integerlist = integer*

The same is true for a list of real numbers, a list of symbols, or a list of strings.
However, it is often valuable to store a combination of different types of elements
within a list, such as:
[2, 3, 5.12, ["food", "goo"], "new"] /* Not correct Visual Prolog*/

Compound lists are lists that contain more than one type of element. You need
special declarations to handle lists of multiple-type elements, because Visual
Prolog requires that all elements in a list belong to the same domain. The way
to create a list in Prolog that stores these different types of elements is to use
functors, because a domain can contain more than one data type as arguments
to functors.
The following is an example of a domain declaration for a list that can contain an
integer, a character, a string, or a list of any of these:
DOMAINS /* the functors are l, i, c, and s */
llist = l(list); i(integer); c(char); s(string)
list = llist*

The list
[ 2, 9, ["food", "goo"], "new" ] /* Not correct Visual Prolog */

would be written in Visual Prolog as

[i(2), i(9), l([s("food"), s("goo")]), s("new")]
/* Correct Visual Prolog */

The following example of append shows how to use this domain declaration in a
typical list-manipulation program.
/* Program ch0145e09.pro */

DOMAINS
llist = l(list); i(integer); c(char); s(string)
list = llist*

PREDICATES
append(list,list,list)

Chapter 128, Lists and Recursion 163

CLAUSES
append([],L,L).
append([X|L1],L2,[X|L3]):-
append(L1, L2, L3).

GOAL
append([s(likes), l([s(bill), s(mary)])],[s(bill), s(sue)],Ans),
write("FIRST LIST: ", Ans,"\n\n"),
append([l([s("This"),s("is"),s("a"),s("list")]),s(bee)],
[c('c')],Ans2),
write("SECOND LIST: ", Ans2, '\n').

Exercises
Write a predicate, oddlist, that takes two arguments. The first argument is a list of
integers, while the second argument returns a list of all the odd numbers
found in the first list.
Write a predicate, real_average, that calculates the average value of all the
elements in a list of reals.
Write a predicate that takes a compound list as its first argument and returns a
second argument that is the list with all the sub-lists removed. This predicate
is commonly known as flatten, as it flattens a list of lists into a single list. For
example, the call
flatten([s(ed), i(3), l([r(3.9), l([s(sally)])])], r(4.21), X)

returns
X = [s(ed), i(3), r(3.9), s(sally), r(4.21)]
1 Solution

which is the original list, flattened.

Parsing by Difference Lists

Program 10 demonstrates parsing by difference lists. The process of parsing by
difference lists works by reducing the problem; in this example we transform a
string of input into a Prolog structure that can be used or evaluated later.
The parser in this example is for a very primitive computer language. Although
this example is very advanced for this point in the tutorial, we decided to put it
here because parsing is one of the areas where Visual Prolog is very powerful. If
you do not feel ready for this topic, you can skip this example and continue on in
the tutorial without any loss of continuity.

164 Visual Prolog

 /* Program ch0146e10.pro */

DOMAINS
toklist = string*

PREDICATES
tokl(string,toklist)

CLAUSES
tokl(Str,[H|T]):-
fronttoken(Str,H,Str1),!,
tokl(Str1,T).
tokl(_,[]).

/* * * * * * * * * * * * * * * * * * * * * * * * *
* This second part of the program is the parser *
* * * * * * * * * * * * * * * * * * * * * * * * */

DOMAINS
program = program(statementlist)
statementlist = statement*

/* * * * * * * * * * * * * * * * * * * * * * * *
* Definition of what constitutes a statement *
* * * * * * * * * * * * * * * * * * * * * * * */

statement = if_Then_Else(exp,statement,statement);
if_Then(exp,statement);
while(exp,statement);
assign(id,exp)

/* * * * * * * * * * * * * * *
* Definition of expression *
* * * * * * * * * * * * * * */

exp = plus(exp,exp);
minus(exp,exp);
var(id);
int(integer)

id = string

PREDICATES
s_program(toklist,program)
s_statement(toklist,toklist,statement)
s_statementlist(toklist,toklist,statementlist)
s_exp(toklist,toklist,exp)
s_exp1(toklist,toklist,exp,exp)
s_exp2(toklist,toklist,exp)

Chapter 128, Lists and Recursion 165

CLAUSES
s_program(List1,program(StatementList)):-
s_statementlist(List1,List2,StatementList),
List2=[].

s_statementlist([],[],[]):-!.
s_statementlist(List1,List4,[Statement|Program]):-
s_statement(List1,List2,Statement),
List2=[";"|List3],
s_statementlist(List3,List4,Program).

s_statement(["if"|List1],List7,if_then_else(Exp,Statement1,
Statement2)):-
s_exp(List1,List2,Exp),
List2=["then"|List3],
s_statement(List3,List4,Statement1),
List4=["else"|List5],!,
s_statement(List5,List6,Statement2),
List6=["fi"|List7].
s_statement(["if"|List1],List5,if_then(Exp,Statement)):-!,
s_exp(List1,List2,Exp),
List2=["then"|List3],
s_statement(List3,List4,Statement),
List4=["fi"|List5].
s_statement(["do"|List1],List4,while(Exp,Statement)):-!,
s_statement(List1,List2,Statement),
List2=["while"|List3],
s_exp(List3,List4,Exp).
s_statement([ID|List1],List3,assign(Id,Exp)):-
isname(ID),
List1=["="|List2],
s_exp(List2,List3,Exp).

s_exp(LIST1,List3,Exp):-
s_exp2(List1,List2,Exp1),
s_exp1(List2,List3,Exp1,Exp).

s_exp1(["+"|List1],List3,Exp1,Exp):-!,
s_exp2(List1,List2,Exp2),
s_exp1(List2,List3,plus(Exp1,Exp2),Exp).
s_exp1(["-"|List1],List3,Exp1,Exp):-!,
s_exp2(List1,List2,Exp2),
s_exp1(List2,List3,minus(Exp1,Exp2),Exp).
s_exp1(List,List,Exp,Exp).

166 Visual Prolog

 s_exp2([Int|Rest],Rest,int(I)):-
str_int(Int,I),!.
s_exp2([Id|Rest],Rest,var(Id)):-
isname(Id).

Load and run this program, then enter the following goal:
Goal tokl("b=2; if b then a=1 else a=2 fi; do a=a-1 while a;",Ans),
s_program(Ans,Res).

Visual Prolog will return the program structure:

Ans=["b","=","2",";","if","b","then","a","=","1",
"else","a","=","2","fi",";","do","a","=","a",
"-","1","while","a",";"
],
Res=program([assign("b",int(2)),
if_then_else(var("b"),assign("a",int(1)),
assign("a",int(2))),
while(var("a"),assign("a",minus(var("a"),int(1))))
])
1 Solution

The transformation in this example is divided into two stages: scanning and
parsing. The tokl predicate is the scanner; it accepts a string and converts it into a
list of tokens. All the predicates with names beginning in s_ are parser predicates.
In this example the input text is a Pascal-like program made up of Pascal-like
statements. This programming language only understands certain statements: IF
THEN ELSE, IF THEN, DO WHILE, and ASSIGNMENT. Statements are made
up of expressions and other statements. Expressions are addition, subtraction,
variables, and integers.
Here's how this example works:
The first scanner clause, s_program, takes a list of tokens and tests if it can be
transformed into a list of statements.
The predicate s_statementlist takes this same list of tokens and tests if the tokens
can be divided up into individual statements, each ending with a semicolon.
The predicate s_statement tests if the first tokens of the token list make up a legal
statement. If so, the statement is returned in a structure and the remaining
tokens are returned back to s_statementlist.

Chapter 128, Lists and Recursion 167

 The four s_statement clauses correspond to the four types of statements the
parser understands. If the first s_statement clause is unable to transform
the list of tokens into an IF THEN ELSE statement, the clause fails and
backtracks to the next s_statement clause, which tries to transform the list
of tokens into an IF THEN statement. If that clause fails, the next one
tries to transform the list of tokens into a DO WHILE statement.
If the first three s_statement clauses fail, the last clause for that predicate
tests if the statement does assignment. This clause tests for assignment by
testing if the first term is a symbol, the second term is "=", and the next
terms make up a simple math expression.
The s_exp, s_exp1, and s_exp2 predicates work the same way, by testing if the
first terms are expressions and--if so--returning the remainder of the terms and
an expression structure back to s_statement.
See the Sentence Analyzer VPI\PROGRAMS\SEN_AN program on your disk for
a more detailed example of parsing natural-language.

Summary

These are the important points covered in this chapter:

Lists are objects that can contain an arbitrary number of elements; you declare
them by adding an asterisk at the end of a previously defined domain.
A list is a recursive compound object that consists of a head and a tail. The head
is the first element and the tail is the rest of the list (without the first element).
The tail of a list is always a list; the head of a list is an element. A list can
contain zero or more elements; the empty list is written [].
The elements in a list can be anything, including other lists; all elements in a list
must belong to the same domain. The domain declaration for the elements
must be of this form:
DOMAINS
elementlist = elements*
elements = ....

where elements = one of the standard domains (integer, real, etc.) or a set of
alternatives marked with different functors (int(integer); rl(real); smb(symbol);
etc.). You can only mix types in a list in Visual Prolog by enclosing them in
compund objects/functors.
You can use separators (commas, [, and |) to make the head and tail of a list
explicit; for example, the list

168 Visual Prolog

 [a, b, c, d]

can be written as
[a|[b, c, d]] or
[a, b|[c, d]] or
[a, b, c|[d]] or
[a|[b|[c, d]]] or
[a|[b|[c|[d]]]] or even
[a|[b|[c|[d|[]]]]]

List processing consists of recursively removing the head of the list (and usually
doing something with it) until the list is an empty list.
The classic Prolog list-handling predicates member and append enable you to
check if an element is in a list and check if one list is in another (or append
one list to another), respectively.
A predicate's flow pattern is the status of its arguments when you call it; they can
be input parameters (i)--which are bound or instantiated--or output
parameters (o), which are free.
Visual Prolog provides a built-in predicate, findall, which takes a goal as one of
its arguments and collects all of the solutions to that goal into a single list.
Because Visual Prolog requires that all elements in a list belong to the same
domain, you use functors to create a list that stores different types of elements.
The process of parsing by difference lists works by reducing the problem; the
example in this chapter transforms a string of input into a Prolog structure that
can be used or evaluated later.

Chapter 128, Lists and Recursion 169

CHAPTER 147

Visual Prologs fact sections

148Visual PrologVisual PrologIn this chapter, we describe how you declare facts
sections and how you can modify the contents of your fact section.
A facts-section is composed of facts that you can add directly into and remove
from your Visual Prolog program at run time. You declare the predicates
describing the facts section in the facts section of your program, and you use
these predicates the same way you use the ones declared in the predicates section.
In Visual Prolog, you use the the predicates assert, asserta, assertz to add new
facts to the facts section, and the predicates retract and retractall to remove
existing facts. You can modify the contents of your facts section by first
retracting a fact and then asserting the new version of that fact (or a different fact
altogether). The predicate consult reads facts from a file and asserts them into the
internal facts, and save saves the contents of an internal facts section to a file.
Visual Prolog treats facts belonging to facts sections differently from the way it
treats normal predicates. Facts for the facts section predicates are kept in tables,
which are easy to modify, while the normal predicates are compiled to binary
code for maximum speed.

Declaring the facts-sections

The keyword facts or (database) marks the beginning of a sequence of

declarations for predicates describing an facts-section. You can add facts--but not
rules--to a facts-section from the keyboard at run time with asserta and assertz.
Or, by calling the standard predicate consult, you can retrieve the added facts
from a disk file. The facts section looks something like the following example.
DOMAINS
name, address = string
age = integer
gender = male ; female

FACTS
person(name, address, age, gender)

170 Visual Prolog Language Tutorial

PREDICATES
male(name, address, age)
female(name, address, age)
child(name, age, gender)

CLAUSES
male(Name, Address, Age) :-
person(Name, Address, Age, male).

In this example, you can use the predicate person the same way you use the other
predicates, (male, female, child); the only difference is that you can insert and
remove facts for person while the program is running.
There are two restrictions on using predicates in facts sections:
1. You can add them into the factssection as facts only--not as rules.
2. Facts in factssections may not have free variables.
It is possible to have several facts sections, but in order to do this, you must
explicitly name each facts section.
FACTS - mydatabase
myFirstRelation(integer)
mySecondRelation(real, string)
myThirdRelation(string)
/* etc. */

This declaration creates a factssection with the name mydatabase. If you don't
supply a name for an facts database, it defaults to the standard name dbasedom.
The names of predicates in a facts section must be unique within a module
(source file), you can't use the same predicate name in two different facts
sections. However, the predicates in the named facts sections are private to the
module in which they're declared, and won't interfere with predicates in other
modules. Modules are explained in the chapter 149.

Using the facts sections

Because Visual Prolog represents a relational facts sections as a collection of
facts, you can use it as a powerful query language for databases. Visual Prolog's
unification algorithm automatically selects facts with the correct values for the
known arguments and assigns values to any unknown arguments, while its
backtracking algorithm can give all the solutions to a given query.

Chapter 150,Visual Prologs facts sections 171

Accessing the facts sections
Predicates belonging to a facts section are accessed in exactly the same way as
other predicates. The only visible difference in your program is that the
declarations for the predicates are in a facts section instead of a predicates
section. Given for instance the following:
DOMAINS
name = string
sex = char

FACTS
person(name,sex)

CLAUSES
person("Helen",'F').
person("Maggie",'F').
person("Suzanne",'F').
person("Per",'M').

you can call person with the goal person(Name,'F') to find all women, or
person("Maggie",'F') to verify that there is a woman called Maggie in your facts
section.
You should be aware that, by their very nature, predicates in factssections are
always nondeterministic. Because facts can be added anytime at run time, the
compiler must always assume that it's possible to find alternative solutions during
backtracking. If you have a predicate in a factssection for which you'll never have
more than one fact, you can override this by prefacing the declaration with the
compiler directive determ to the declaration:
FACTS
determ daylight_saving(integer)

You will get an error if you try to add a fact for a deterministic database predicate
which already has a fact.

Updating the facts section

Facts for database predicates can be specified at compile time in the clauses
section, as in the example above. At run time, facts can be added and removed by
using the predicates described below. Note that facts specified at compile time in
the clauses section can be removed too, they're not different from facts added at
run time.

172 Visual Prolog Language Tutorial

Visual Prolog's standard database predicates assert, asserta, assertz, retract,
retractall, consult, and save will all take one or two arguments. The optional
second argument is the name of an facts section. We describe these predicates in
the following pages. The notation "/1" and "/2" after each predicate name refers
to the number of arguments that arity version of the predicate takes. The
comments after the formats (such as /* (i) */ and /* (o,i) */ show the flow
pattern(s) for that predicate.

Adding Facts at Run Time

At run time, facts can be added to the facts sections with the predicates: assert,
asserta and assertz, or by loading a file of facts with consult.
There are three predicates to add a single fact at runtime:
asserta(< the fact > ) /* (i) */

asserta(< the fact > , facts_sectionName) /* (i, i) */

assertz(< the fact > ) /* (i) */

assertz(< the fac t > , facts_sectionName) /* (i, i) */

assert(< the fact > ) /* (i) */

assert(< the fact > , facts_sectionName) /* (i, i) */

asserta asserts a new fact into the facts section before the existing facts for the
given predicate, while assertz asserts a new fact after the existing facts for that
predicate. assert behaves like assertz.
The assertion predicates always know which facts section to insert the fact in,
because the names of the facts section predicates are unique within a program or
module. However, you can use the optional second argument for type-checking
purposes in order to ensure that you are working on the correct facts section.
The first of the following goals inserts a fact about Suzanne for the person
predicate, after all the facts for person currently stored in the facts section. The
second inserts a fact about Michael before all the currently-stored facts for
person. The third inserts a fact about John after all the other likes facts in the facts
section likesDatabase, while the fourth inserts a fact about Shannon in the same
facts section, before all the other likes facts.
assertz(person("Suzanne", "New Haven", 35)).
asserta(person("Michael", "New York", 26)).
assertz(likes("John", "money"), likesDatabase).
asserta(likes("Shannon", "hard work"), likesDatabase).

Chapter 150,Visual Prologs facts sections 173

After these calls the facts sections look as if you'd started with the following
facts:
/* Facts section -- dbasedom */
person("Michael", "New York", 26).
/* ... other person facts ... */
person("Suzanne", "New Haven", 35).

/* Facts section -- likesDatabase */

likes("Shannon", "hard work").
/* ... other likes facts ... */
likes("John", "money").

Be careful that you don't accidentally write code asserting the same fact twice.
The facts sections don't impose any kind of uniqueness, and the same fact may
appear many times in a facts section. However, a uniqueness-testing version of
assert is very easy to write:
FACTS - people
person(string,string)

PREDICATES
uassert(people)

CLAUSES
uassert(person(Name,Address)):-
person(Name,Address), ! ; % OR
assert(person(Name,Address)).

Loading Facts from a File at Run Time

consult reads in a file, fileName, containing facts declared in a facts section and
asserts them at the end) of the appropriate facts section. consult takes one or two
arguments:
consult(fileName) /* (i) */
consult(fileName, databaseName) /* (i, i) */

Unlike assertz, if you call consult with only one argument (no facts section
name), it will only read facts that were declared in the default dbasedom facts
section section.
If you call consult with two arguments (the file name and a facts section name), it
will only consult facts from that named facts section. If the file contains anything
other than a fact belonging to the specified facts section, consult will return an
error when it reaches that part of the file.

174 Visual Prolog Language Tutorial

Keep in mind that the consult predicate reads one fact at a time; if the file has ten
facts, and the seventh fact has some syntax error, consult will insert the first six
facts into the facts section--then issue an error message.
Note that consult is only able to read a file in exactly the same format as save
generates (in order to insert facts as fast as possible). There can be
no upper-case characters except in double-quoted strings
no spaces except in double-quoted strings
no comments
no empty lines
no symbols without double quotes
You should be careful when modifying or creating such a file of facts with an
editor.

Removing Facts at Run Time

retract unifies facts and removes them from the facts sections. It's of the
following form:
retract(<the fact> [, databaseName]) /* (i, i) */

retract will remove the first fact in your facts section that matches <the fact>,
instantiating any free variables in <the fact> in the process. Retracting facts from
a facts section is exactly like accessing it, with the side-effect that the matched
fact is removed. Unless the facts section predicate accessed by retract was
declared to be deterministic, retract is nondeterministic and will, during
backtracking, remove and return the remaining matching facts, one at a time.
When all matching facts have been removed, retract fails.
Suppose you have the following facts sections in your program:
DATABASE
person(string, string, integer)

FACTS - likesDatabase
likes(string, string)
dislikes(string, string)

CLAUSES
person("Fred", "Capitola", 35).
person("Fred", "Omaha", 37).
person("Michael", "Brooklyn", 26).

Chapter 150,Visual Prologs facts sections 175

 likes("John", "money").
likes("Jane", "money").
likes("Chris", "chocolate").
likes("John", "broccoli").

dislikes("Fred", "broccoli").
dislikes("Michael", "beer").

Armed with these facts sections, you give Visual Prolog the following goals:
retract(person("Fred", _, _)). /* 1 */
retract(likes(_, "broccoli")). /* 2 */
retract(likes(_, "money"), likesDatabase). /* 3 */
retract(person("Fred", _, _), likesDatabase) /* 4 */

The first goal retracts the first fact for person about Fred from the default
dbasedom facts section. The second goal retracts the first fact matching likes(X,
"broccoli") from the facts section likesDatabase. With both of these goals,
Visual Prolog knows which facts section to retract from because the names of the
facts section predicates are unique: person is only in the default facts section, and
likes is only in the facts section likesDatabase.
The third and fourth goals illustrate how you can use the optional second
argument for type-checking. The third goal succeeds, retracting the first fact that
matches likes(_, "money") from likesDatabase, but the fourth cannot be
compiled because there are (and can be) no person facts in the facts section
likesDatabase. The error message given by the compiler is:
506 Type error: The functor does not belong to the domain.

The following goal illustrates how you can obtain values from retract:
GOAL
retract(person(Name, Age)),
write(Name, ", ", Age),
fail.

If you supply the name of a facts section section as the second argument to
retract, you don't have to specify the name of the facts section predicate you're
retracting from. In this case, retract will find and remove all facts in the specified
facts section. Here's an example:
GOAL
retract(X, mydatabase),
write(X),
fail.

176 Visual Prolog Language Tutorial

Removing Several Facts at Once
retractall removes all facts that match <the fact> from your facts section, and is
of the following form:
retractall(<the fact> [, databaseName])

retractall behaves as if defined by

retractall(X):- retract(X), fail.
retractall(_).

but it's considerably faster than the above.

As you can gather, retractall always succeeds exactly once, and you can't obtain
output values from retractall. This means that, as with not, you must use
underscores for free variables.
As with assert and retract, you can use the optional second argument for type-
checking. And, as with retract, if you call retractall with an underscore, it can
remove all the facts from a given facts section.
The following goal removes all the facts about males from a database of person
facts:
retractall(person(_, _, _, male)).

The next goal removes all the facts from the facts section mydatabase.
retractall(_, mydatabase).

Facts determiner-keywords
Facts can be declared with several optional keywords:
NONDETERM determines that any number of instances of a fact fact_N can
exist. This is default.
DETERM determines that only one instance of a fact fact_N can exist at any
time.
GLOBAL determines, that the facts section is global in the project.
SINGLE determines, only one instance of a fact fact_N should always
exist. Fact_N is the functors for the facts (predicates) belonging
to this facts section.
NOCOPY determins, that dates are not copied from the heap to the Visual
Prolog Global Stack (GStack), when the fact is referenced.
Chapter 150,Visual Prologs facts sections 177
 Normally, when calling a fact to bind a variable to a string or a
compound object, the string or object is copied to the Gstack.

Discussions

Facts declared with the keyword nondeterm.

The keyword nondeterm is the default type for facts (database predicates)
declared in facts sections. If none of the keywords determ or single are used in a
fact declaration, the compiler applies nondeterm keyword. Normally, by their
very nature, database predicates are non-deterministic. Because facts can be
added anytime at runtime, the compiler must always assume that it's possible to
find alternative solutions during backtracking.
If you have a database predicate of which you'll never have more than one fact,
you can override this by adding the keyword determ or single to the declaration.

Facts declared with the keyword determ.

The keyword determ determins that the facts database can only contain one
instance of a fact (database predicate) fact_N(...) declared with this keyword. So
if you try to assert one and then a second such fact into the database, the Visual
Prolog engine will generate runtime error. (1041 Assert to a fact declared as
determ, but fact already exists). In such cases, programmer must take special care
about this.
Preceding a fact with determ enables the compiler to produce better code, and you
will not get non-deterministic warnings for calling such a predicate. This is useful
for flags, counters, and other things that are essentially global variables.
Particularly note that when retracting a fact that is declared to be determ, the call
to non-deterministic predicates retract/1 and retract/2 will be deterministic. So if
you know that at any moment the facts section contains no more then one fact
counter() then you can write:
FACTS
determ counter(integer CounterValue)
GOAL
...
retract(counter(CurrentCount)),
/* here Prolog will not set backtracking point */
Count= CurrentCount + 1,
assert(counter(Count)),
...

178 Visual Prolog Language Tutorial

instead of
FACTS
counter(integer CounterValue)
PREDICATES
determ my_retract(dbasedom)
CLAUSES
my_retract(X): - retract(X),!. % deterministic predicate
GOAL
...
my_retract(counter(CurrentCount)),
/* here Prolog will not set backtracking point */
Count= CurrentCount + 1,
asserta(counter(Count)),
...

Facts declared with the keyword single.

The keyword single before a fact fact_N declaration determines that one and only
one instance of a fact must always exist:
Since single facts must be already known when the program calls Goal; therefore,
single facts must be initialized in a clauses section in the program's source code.
For example:

FACTS
single singleFact(STRING, STRING)
CLAUSES
singleFact("","").

·Single facts cannot be retracted. If one try to apply any retract predicate to a
single fact then the compiler will generates the error 249 "Attempt to retract a fact
declared as single".·Since one instance of a single fact always exists, a single fact
never fails if it is called with free arguments. For example, a following call
singleFact(X,Y),

never fails if X and Y are free variables. Therefore, it is convenient to use single
facts in procedures.assert, asserta, assertz, and consult predicates applied to a
single fact act similarly to couples of retract and assert predicates. That is, assert
(consult) predicates change the existing instance of a fact to the specified one.
Preceding a fact with single enables the compiler to produce optimized code for
accessing and updating of a fact. For example, for assert predicate applied to a
single fact the compiler generates a code that works more effectively that a couple
Chapter 150,Visual Prologs facts sections 179
of retract and assert predicates applied to a determ fact (and all the more so then
retract and assert predicates applied to a nondeterm fact).

Saving a database of facts at runtime

save saves facts from a given facts section to a file. save takes one or two
arguments:
save(fileName) /* (i) */
save(fileName, databaseName) /* (i, i) */

If you call save with only one argument (no facts section name), it will save the
facts from the default dbasedom database to the file fileName.
If you call save with two arguments (the file name and a facts section name), it
will save all facts of the facts section databaseName to the named file.

Examples

1. This is a simple example of how to write a classification expert system using

the facts section. The important advantage of using the facts section in this
example is that you can add knowledge to (and delete it from) the program at
run time.
/* Program ch0151e01.pro */

/* Copyright (c) 1986, '95 by Prolog Development Center */

DOMAINS
thing = string
conds = cond*
cond = string

FACTS
is_a(thing, thing, conds)
type_of(thing, thing, conds)
false(cond)

PREDICATES
nondeterm run(thing)
nondeterm ask(conds)
update

180 Visual Prolog Language Tutorial

CLAUSES
run(Item):-
is_a(X, Item, List),
ask(List),
type_of(ANS, X, List2),
ask(List2),
write("The ", Item," you need is a/an ", Ans),nl.

run(_):-
write("This program does not have enough "),
write("data to draw any conclusions."),
nl.

ask([]).
ask([H|T]):-
not(false(H)),
write("Does this thing help you to "),
write(H," (enter y/n)"),
readchar(Ans), nl, Ans='y',
ask(T).

ask([H|_]):-
assertz(false(H)), fail.

is_a(language, tool, ["communicate"]).

is_a(hammer, tool, ["build a house", "fix a fender", "crack a nut"]).
is_a(sewing_machine, tool, ["make clothing", "repair sails"]).
is_a(plow, tool, ["prepare fields", "farm"]).

type_of(english, language, ["communicate with people"]).

type_of(prolog, language, ["communicate with a computer"]).

update:-
retractall(type_of(prolog, language, ["communicate with a
computer"])),
asserta(type_of("PDC Prolog", language,
["communicate with a personal computer"])),
asserta(type_of(prolog, language,
["communicate with a mainframe computer"])).

The following database facts could have been asserted using asserta or
assertz, or consulted from a file using consult. In this example, however,
they're listed in the clauses section.
is_a(language, tool, ["communicate"]).
is_a(hammer, tool, ["build a house", "fix a fender", "crack a nut"]).
is_a(sewing_machine, tool, ["make clothing", "repair sails"]).
is_a(plow, tool, ["prepare fields", "farm"]).

Chapter 150,Visual Prologs facts sections 181

 type_of(english, language, ["communicate with people"]).
type_of(prolog, language, ["communicate with a computer"]).

As the goal enter:

run(tool).

Respond to each question as if you were looking for some tool to

communicate with a personal computer.
Now enter the following goal:
update, run(tool).

The update predicate is included in the source code for the program, to save
you a lot of typing, and will remove the fact
type_of(prolog, language, ["communicate with a computer"])

from the facts section and add two new facts into it:
type_of(prolog, language,
["communicate with a mainframe computer"]).
type_of("Visual Prolog", language,
["communicate with a personal computer"]).

Now respond to each question once again as if you were looking for some tool
to communicate with a personal computer.
You can save the whole facts database in a text file by calling the predicate
save with the name of the text file as its argument. For example, after the call
to
save("mydata.dba")

the file mydata.dba will resemble the clauses section of an ordinary Visual
Prolog program, with a fact on each line. You can read this file into memory
later using the consult predicate:
consult("mydata.dba")

2. You can manipulate facts describing database predicates (facts declared in the
facts section of your program) as though they were terms.
When you declare a facts section, Visual Prolog will internally generate a
domain definition corresponding to the facts declaration. As an example,
consider the declarations
FACTS - dba1 /* dba1 is the domain for these predicates */
person(name, telno)
city(cno, cname)

182 Visual Prolog Language Tutorial

 Given these declarations, the Visual Prolog system internally generates the
corresponding dba1 domain:
DOMAINS
dba1 = person(name, telno) ; city(cno, cname)

This dba1 domain can be used like any other predefined domain. For example,
you could use the standard predicate readterm (which is covered in chapter
152) to construct a predicate my_consult, similar to the standard predicate
consult.
Program 2 illustrates one practical way you might use the facts section in an
application. This example uses a screen handler, which places text on the
screen in predefined locations. A screen layout for the current screen display
can be stored in the field and textfield facts that are defined in the screen facts
section. Several screen names can be stored in the screens facts section. At run
time, the shiftscreen predicate can copy one of these stored screens to the
screen facts section by first retracting all current data from the screen facts
section, calling the screen predicate to get the layout information for the
upcoming screen, then asserting the new screen's form into the screen facts
section.
/* Program ch0153e02.pro */

/* Copyright (c) 1986, '95 by Prolog Development Center */

DOMAINS
screenname, fname, type = symbol
row,col,len = integer

FACTS
screen(symbol,dbasedom) /* Saving different screens */

FACTS - screen
field(fname, type, row, col, len) /* Definitions of field on screen */
txtfield(row, col, len, string) /* Showing textfields */
windowsize(row,col)

PREDICATES
shiftscreen(symbol)

CLAUSES
shiftscreen(_):-
retract(field(_,_,_,_,_)),
fail.

shiftscreen(_):-
retract(txtfield(_,_,_,_)),
fail.

Chapter 150,Visual Prologs facts sections 183

 shiftscreen(_):-
retract(windowsize(_,_)),
fail.

shiftscreen(Name):-
screen(Name,Term),
assert(Term),
fail.

shiftscreen(_).

GOAL
shiftscreen(person).

Summary

1. Visual Prolog's facts section is composed of the facts in your program that are
grouped into facts sections. You declare the user-defined predicates used in
these groups of facts with the keyword facts.
2. You can name facts sections (which creates a corresponding internal
domain); the default domain for (unnamed) facts sections is dbasedom. Your
program can have multiple facts sections, but each one must have a unique
name. You can declare a given facts predicate in only one facts section.
3. With the standard predicates assert, asserta, assertz, and consult, you can add
facts to the facts section at run time. You can remove such facts at run time
with the standard predicates retract and retractall.
4. The save predicate saves facts from a facts section to a file (in a specific
format). You can create or edit such a fact file with an editor, then insert facts
from the file into your running program with consult.
5. You can call database predicates in your program just like you call other
predicates.
6. You can handle facts as terms when using the domain internally generated for
a database section.

184 Visual Prolog Language Tutorial

CHAPTER 154

Arithmetic and Comparison

155Visual PrologVisual Prolog's arithmetic and comparison capabilities are
similar to those provided in programming languages such as BASIC, C, and
Pascal. Visual Prolog includes a full range of arithmetic functions; you have
already seen some simple examples of Visual Prolog's arithmetic capabilities.
In this chapter we summarize Visual Prolog's built-in predicates and functions for
performing arithmetic and comparisons, as well as a two arity versions of a
standard predicate used for random number generation. We'll also discuss
comparison of strings and characters.

Arithmetic Expressions

Arithmetic expressions consist of operands (numbers and variables), operators

(+, -, *, /, div, and mod), and parentheses. The symbols on the right side of the
equal sign (which is the = predicate) in the following make up an arithmetic
expression.
A = 1 + 6 / (11 + 3) * Z

Leading "0x" or "0o" signify hexadecimal and octal numbers, respectively, e.g.
0xFFF = 4095
86 = 0o112 + 12

The value of an expression can only be calculated if all variables are bound at the
time of evaluation. The calculation then occurs in a certain order, determined by
the priority of the arithmetic operators; operators with the highest priority are
evaluated first.

Operations
Visual Prolog can perform all four basic arithmetic operations (addition,
subtraction, multiplication, and division) between integral and real values; the
type of the result is determined according to Table 156.1.

Chapter 157 Advanced topics

 Table 158.1 Arithmetic Operations1592

Operand 1 Operator Operand 2 Result

integral +, -, * integral integral
real +, -, * integral real
integral +, -, * real real
real +, -, * real real
integral or real / integral or real real
integral div integral integral
integral mod integral integral

In case of mixed integral arithmetic, involving both signed and unsigned

quantities, the result is signed. The size of the result will be that of the larger of
the two operands. Hence, if a ushort and a long are involved the result is long; if
a ushort and a ulong are involved the result is ulong.

Order of Evaluation
Arithmetic expressions are evaluated in this order:
If the expression contains sub-expressions in parentheses, the sub-expressions are
evaluated first.
If the expression contains multiplication (*) or division (/, div or mod), these
operations are carried out next, working from left to right through the
expression.
Finally, addition (+) and subtraction (-) are carried out, again working from left to
right.
Hence, these are the operator precedence:

Table 160.3 Operator

Precedence161

Operator Priority
+- 1
* / mod div 2
186 Visual Prolog Language Tutorial
 - + (unary 3

In the expression A = 1 + 6/(11+3)*Z, assume that Z has the value 4, since

variables must be bound before evaluation.
(11 + 3) is the first sub-expression evaluated, because it's in parentheses; it
evaluates to 14.
Then 6/14 is evaluated, because / and * are evaluated left to right; this gives
0.428571.
Next, 0.428571 * 4 gives 1.714285.
Finally, evaluating 1 + 1.714285 gives the value of the expression as 2.714285.
A will then be bound to 2.714285 which makes it a real value.
However, you should exercise some care when handling floating point (real)
quantities. In most cases they're not represented accurately and small errors can
accumulate, giving unpredictable results. An example follows later in the chapter.

Functions and Predicates

Visual Prolog has a full range of built-in mathematical functions and predicates
that operate on integral and real values. The complete list is given in Table 162.4

Table 163.5: Visual Prolog Arithmetic Predicates and Functions1646Visual

Prolog

Name Description
X mod Y Returns the remainder (modulos) of X divided by Y.

X div Y Returns the quotient of X divided by Y.

abs(X) If X is bound to a positive value val, abs(X) returns that
value; otherwise, it returns -1 * val.
cos(X) The trigonometric functions require that X be bound to
sin(X) a value representing an angle in radians.
tan(X) Returns the tangent of its argument.

Chapter 157 Advanced topics

 arctan(X) Returns the arc tangent of the real value to which X is
bound.
exp(X) e raised to the value to which X is bound.
ln(X) Logarithm of X, base e.
log(X) Logarithm of X, base 10.
sqrt(X) Square root of X.
random(X) Binds X to a random real; 0 <= X < 1.
random(X, Y) Binds Y to a random integer; 0 <= Y < X.
round(X) Returns the rounded value of X. The result still being a
real
trunc(X) Truncates X. The result still being a real
val(domain,X) Explicit conversion between numeric domains.

Generating Random Numbers

Visual Prolog provides two standard predicates for generating random numbers.
One returns a random real between 0 and 1, while the other returns a random
integer between 0 and a given integer. Additionally, the random numbering
sequence may be re-initialized.

random/1
This version of random returns a random real number that satisfies the constraints
0 <= RandomReal < 1.

random/1 takes this format:

random(RandomReal) /* (o) */

random/2
This version of random takes two arguments, in this format:
random(MaxValue, RandomInt) /* (i, o) */

It binds RandomInt to a random integer that satisfies the constraints

188 Visual Prolog Language Tutorial

 0 <= RandomInt < MaxValue

random/2 is much faster than random/1 because random/2 only uses integer
arithmetic.

randominit/1
randominit will initialize the random number generator and is of the following
form:
randominit(Seed) /* (i) */

The default random number seed value is genarated as function from system time,
and the Seed argument to randominit sets this seed value. The main use for
randominit is to provide repeatable sequences of pseudo random numbers for
statistical testing. Note that both the integer and floating point versions of
random use the same seed and basic number generator.

Example
Program 1 uses random/1 to select three names from five at random.
/* Program ch165e01.pro */

/* Copyright (c) 1986, '95 by Prolog Development Center */

PREDICATES
person(integer, symbol)
rand_int_1_5(integer)
rand_person(integer)

CLAUSES
person(1,fred).
person(2,tom).
person(3,mary).
person(4,dick).
person(5,george).

rand_int_1_5(X):-
random(Y),
X=Y*4+1.

Chapter 157 Advanced topics

 rand_person(0):-!.
rand_person(Count):-
rand_int_1_5(N),
person(N,Name),
write(Name),nl,
NewCount=Count-1,
rand_person(NewCount).

GOAL
rand_person(3).

Integer and Real Arithmetic

Visual Prolog provides predicates and functions for modulos function, integer
division, square roots and absolute values, trigonometry, transcendental functions,
rounding (up or down), and truncation. They are summarized in Table 166.7 and
explained in the following pages.

mod/2
mod performs the function X modulo Y (where X and Y are integers).
X mod Y /* (i, i) */

The expression Z = X mod Y binds Z to the result. For example,

Z = 7 mod 4 /* Z will equal 3 */
Y = 4 mod 7 /* Y will equal 4 */

div/2
div performs the integer division X/Y (where X and Y are integers).
X div Y /* (i, i) */

The expression Z = X div Y binds Z to the integer part of the result. For example,
Z = 7 div 4 /* Z will equal 1 */
Y = 4 div 7 /* Y will equal 0 */

abs/1
abs returns the absolute value of its argument.
abs(X) /* (i) */

190 Visual Prolog Language Tutorial

The expression Z = abs(X) binds Z (if it's free) to the result, or succeed/fail if Z is
already bound. For example,
Z = abs(-7) /* Z will equal 7 */

cos/1
cos returns the cosine of its argument.
cos(X) /* (i) */

The expression Z = cos(X) binds Z (if it's free) to the result, or succeed/fail if Z is
already bound. For example,
Pi = 3.141592653,
Z = cos(Pi) /* Z will equal -1 */

sin/1
sin returns the sine of its argument.
sin(X) /* (i) */

The expression Z = sin(X) binds Z (if it's free) to the result, or succeed/fail if Z is
already bound. For example:
Pi = 3.141592653,
Z = sin(Pi) /* Z will almost equal 0 */

tan/1
tan returns the tangent of its argument.
tan(X) /* (i) */

The expression Z = tan(X) binds Z (if it's free) to the result, or succeed/fail if Z is
already bound. For example,
Pi = 3.141592653,
Z = tan(Pi) /* Z will almost equal 0 */

arctan/1
arctan returns the arc tangent of the real value to which X is bound.
arctan(X) /* (i) */

Chapter 157 Advanced topics

The expression Z = arctan(X) binds Z (if it's free) to the result, or succeed/fail if
Z is already bound. For example,
Pi = 3.141592653,
Z = arctan(Pi) /* Z will equal 1.2626272556 */

exp/1
exp returns e raised to the value to which X is bound.
exp(X) /* (i) */

The expression Z = exp(X) binds Z (if it's free) to the result, or succeed/fail if Z is
already bound. For example,
Z = exp(2.5) /* Z will equal 12.182493961 */

ln/1
ln returns the natural logarithm of X (base e).
ln(X /* (i) */

The expression Z = ln(X) binds Z (if it's free) to the result, or succeed/fail if Z is
already bound. For example,
Z = ln(12.182493961) /* Z will equal 2.5 */

log/1
log returns the base 10 logarithm of X.
log(X) /* (i) */

The expression Z = log(X) binds Z (if it's free) to the result, or succeed/fail if Z is
already bound. For example,
Z = log(2.5) /* Z will equal 0.39794000867 */

sqrt/1
sqrt returns the positive square root of X.
sqrt(X) /* (i) */

The expression Z = sqrt(X) binds Z (if it's free) to the result, or succeed/fail if Z
is already bound. For example,
192 Visual Prolog Language Tutorial
 Z = sqrt(25) /* Z will equal 5 */

round/1
round returns the rounded value of X.
round(X) /* (i) */

round rounds X up or down to the nearest integral value of X, but performs no

type conversion. For example,
Z1 = round(4.51) /* Z1 will equal 5 */
Z2 = round(3.40) /* Z2 will equal 3 */

Both Z1 and Z2 are floating point values following the above; only the fractional
parts of the arguments to round have been rounded up or down.

trunc/1
trunc truncates X to the right of the decimal point, discarding any fractional part.
Just like round, trunc performs no type conversion.
trunc(X) /* (i) */

For example,
Z = trunc(4.7) /* Z will equal 4 */

Again, Z is a floating point number.

val/2
val provides general purpose conversion between the numeric domains, in cases
where you want full control over the result of the operation. val observes any
possible overflow condition. The format is
Result = val(returndom,Expr)

where Expr will be evaluated (if it's an expression), the result converted to
returndom and unified with Result. Visual Prolog also has a cast function that
will convert uncritically between any domains; this is described in chapter 167.

Exercise
Use the trigonometric functions in Visual Prolog to display a table of sine, cosine,
and tangent values on the screen. The left column of the table should contain

Chapter 157 Advanced topics

angle values in degrees, starting at 0 degrees and continuing to 360 degrees in
steps of 15 degrees.
Note: Because the trigonometric functions take values expressed in radians, you
must convert radians to angles to obtain entries for the left column.
Degrees = Radians * 180/3.14159265...

Comparisons

Visual Prolog can compare arithmetic expressions as well as characters, strings,

and symbols. The following statement is the Visual Prolog equivalent of "The
total of X and 4 is less than 9 minus Y."
X + 4 < 9 - Y

The less than (<) relational operator indicates the relation between the two
expressions, X + 4 and 9 - Y.
Visual Prolog uses infix notation, which means that operators are placed between
the operands (like this: X+4) instead of preceding them (like this: +(X,4)).
The complete range of relational operators allowed in Visual Prolog is shown in
Table 168.8.

Table 169.9: Relational Operators17010

Symbol Relation
< less than
<= less than or equal to
= equal
> greater than
>= greater than or equal to
<> or >< not equal

Equality and the equal (=) Predicate

In Visual Prolog, statements like N = N1 - 2 indicate a relation between three
objects (N, N1, and 2), or a relation between two objects (N and the value of N1 -
194 Visual Prolog Language Tutorial
2). If N is still free, the statement can be satisfied by binding N to the value of the
expression N1 - 2. This corresponds roughly to what other programming
languages call an assignment statement. Note that N1 must always be bound to a
value, since it is part of an expression to be evaluated.
When using the equal predicate (=) to compare real values, you must take care to
ensure that the necessarily approximate representation of real numbers does not
lead to unexpected results. For example, the goal
7/3 * 3 = 7

will frequently fail (the exact outcome depends on the accuracy of the floating
point calculations in use on your particular platform). Program 2 illustrates
another example:
/* Program ch171e02.pro */
/* Copyright (c) 1986, '95 by Prolog Development Center */

PREDICATES
test(real,real)
run

CLAUSES
test(X,X):-!,
write("ok\n").
test(X,Y):-
Diff = X-Y,
write(X,"<>",Y,"\nX-Y = ",Diff,'\n').

GOAL
X=47,
Y=4.7*10,
test(X,Y).

Except when running Prolog on the UNIX platform, where it behaves as one
might expect, this prints:
47<>47
X-Y = 7.1054273576E-15

Therefore, when comparing two real values for equality you should always check
that the two are within a certain range of one another.

Chapter 157 Advanced topics

Example
Program 3 shows how to handle approximate equality; this is an iterative
procedure for finding the square root in order to calculate the solutions to the
quadratic equation:
A*X*X + B*X + C = 0

The existence of solutions depends on the value of the discriminant D, defined as

follows:
D = B*B - 4*A*C.

D > 0 implies that there are two unique solutions.

D = 0 implies there is only one solution.
D < 0 implies that there are no solutions if X is to take a real value (there can
be one or two complex solutions).
/* Program ch172e03.pro */

/* Copyright (c) 1986, '95 by Prolog Development Center */

PREDICATES
solve(real, real, real)
reply(real, real, real)
mysqrt(real, real, real)
equal(real, real)

CLAUSES
solve(A,B,C):-
D=B*B-4*A*C,
reply(A, B, D), nl.

reply(_,_,D):-
D < 0,
write("No solution"),!.
reply(A,B,D):-
D=0,
X=-B/(2*A),write("x=", X),!.
reply(A,B,D):-
mysqrt(D,D,SqrtD),
X1=(-B+SqrtD)/(2*A),
X2 = (-B - SqrtD)/(2*A),
write("x1 = ", X1," and x2 = ", X2).

196 Visual Prolog Language Tutorial

 mysqrt(X,Guess,Root):-
NewGuess = Guess-(Guess*Guess-X)/2/Guess,
not(equal(NewGuess,Guess)), !,
mysqrt(X,NewGuess,Root).

mysqrt(_,Guess,Guess).

equal(X,Y):-
X/Y >0.99999,
X/Y < 1.00001.

To solve the quadratic equation, this program calculates the square root of the
discriminant, D. The program calculates square roots with an iterative formula
where a better guess (NewGuess) for the square root of X can be obtained from
the previous guess (Guess):
NewGuess = Guess-(Guess*Guess-X)/2/Guess

Each iteration gets a little closer to the square root of X. Once the condition
equal(X, Y) is satisfied, no further progress can be made, and the calculation
stops. Once this calculation stops, the program can solve the quadratic using the
values X1 and X2, where
X1 = (-B + sqrtD)/(2*A)
X2 = (-B - sqrtD)/(2*A)

Exercises
Load Program 3 and try the following goals:
solve(1, 2, 1).
solve(1, 1, 4).
solve(1, -3, 2).

The solutions should be

x = -1
No solution
x1 = 2 and x2 = 1

respectively.
The object of this exercise is to experiment with the mysqrt predicate in Program
3. To ensure that temporary calculations are monitored, add the following as
the first subgoal in the first mysqrt clause:
write(Guess).

To see the effect of this amendment, try this goal:

Chapter 157 Advanced topics
 mysqrt(8, 1, Result).

Next, replace the equal clause with this fact:

equal(X, X).

and retry the goal. Experiment a little more with the properties of equal. For
instance, try
equal(X, Y) :-
X/Y < 1.1 , X/Y > 0.9.

Visual Prolog has a built-in square root function, sqrt. For example,
X = sqrt(D)

will bind X to the square root of the value to which D is bound. Rewrite
Program 3 using sqrt and compare the answers with those from the original
version.

Comparing Characters, Strings, and Symbols

Besides numeric expressions, you can also compare single characters, strings and
symbols. Consider the following comparisons:
'a' < 'b' /* Characters */
"antony" > "antonia" /* Strings */
P1 = peter, P2 = sally, P1 > P2 /* Symbols */

Characters
Visual Prolog converts the 'a' < 'b' to the equivalent arithmetic expression 97 <
98, using the corresponding ASCII code value for each character. You should be
aware that only 7 bit ASCII comparisons should be relied upon (i.e. upper and
lower case letters a-z, digits, etc.). 8 bit characters, used for a number of national
characters, are not necessarily portable between the different platforms.

Strings
When two strings or symbols are compared, the outcome depends on a character-
by-character comparison of the corresponding positions. The result is the same as
you'd get from comparing the initial characters, unless those two characters are
the same. If they are, Visual Prolog compares the next corresponding pair of
characters and returns that result, unless those characters are also equal, in which
case it examines a third pair, and so on. Comparison stops when two differing
characters are found or the end of one of the strings is reached. If the end is

198 Visual Prolog Language Tutorial

reached without finding a differing pair of characters, the shorter string is
considered smaller.
The comparison "antony" > "antonia" evaluates to true, since the two symbols
first differ at the position where one contains the letter y (ASCII value 79) and the
other the letter i (ASCII value 69). In the same vein, the character comparison
"aa" > "a" is true.

Similarly, the expression "peter" > "sally" would be false--as determined by

comparing the ASCII values for the characters that make up peter and sally,
respectively. The character p comes before s in the alphabet, so p has the lower
ASCII value. Because of this, the expression evaluates to false.

Symbols
Symbols can't be compared directly because of syntax. In the preceding P1 =
peter, P2 ... example, the symbol peter can't be compared directly to the
symbol sally; they must be bound to variables to be compared, or written as
strings

Chapter 157 Advanced topics

CHAPTER 173

Classes and objects

Visual Prolog contains a powerful object-mechanism, that melts together the logic
programming and object oriented programming (OOP) paradigms.
You will find some small examples of Visual Prolog programs, that uses OOP-
technlogy on the CD in the directory \OOP\EXAMPLES.
Four criterias have to be fulfilled, before a system can be considered to be object
orientated: encapsulation, classes, inheritance, and identity.

Encapsulation
The importance of encapsulation and modularity are well known. Encapsulated
objects can help building more structured and readable programs because objects
are treated like blackboxes. Look at complex problems, find a part, which you can
declare and describe. Encapsulate it in an object, construct an interface and
continue so, until you have declared all the sub problems. When you have
encapsulated the objects of the problem, and ensured that they work correctly,
you can abstract from them.
OOP is also sometimes known as data-driven programming. You can actually let
the objects themselves do the work for you. They contain methods, which are
invoked, when they are created, deleted and whenever you call them. Methods
can call methods in other objects.

Objects and classes

The way data is stored in traditional programming languages is usually hard to
grasp for humans and not suited for modelling. Objects are much easier to work
with, because it is closer to the way humans understand real-world objects and in
fact a tool for modeling in it self.
Object is a far more complex data structure than lists. An object is at the basic
level a declaration of coherent data. This declaration can also contain predicates,
which work on these data. In OOP-terminology these are called methods. Each
class type represents a unique set of objects and the operations (methods)
available to create, manipulate, and destroy such objects.

200 Visual Prolog Language Tutorial

A class is a definition of an object. An instance is an actual occurrence of this
object. Normally you can define as many instances as you like of a class.
Example
class automobile
owner string
brand string
endclass

Actual instance 1
Owner Beatrice
brand Morris Mascot
End

Actual instance 2
Owner John
brand Rolls Royce
End

Inheritance
OOP is a powerful modeling tool. Objects can be defined on the abstraction level
that suits best. From this level child-objects can be defined on lower levels, or
parent-objects on higher levels. An object can inherit data and methods from
objects at higher levels. Objects are thus an easy way to make very modular
programs.

Identity
A very important characteristic of objects is, that it identity remains, even though
it attributes changes.

Visual Prolog Classes

Defining a Class in Visual Prolog requires two things: a class declaration, and a
class implementation. The class declaration specifies the interface to the class,
what can be seen from the outside. The class implementation contains the Prolog
clauses for defining the actual functionallity for the class.
The declaration of the interface to a class and the actual definition of the clauses
for a class are separated. The Class declarations will often be placed in header
files which can be included in places which uses the class.

Chapter 174 Classes and objects 201

Class declarations
A simplified "first-look" syntax for a class declaration is:
CLASS class-name [: parentclass-list ]
PREDICATES
predicatedeclaration-list

FACTS
factdeclaration-list

ENDCLASS

The optional parentclass-list specifies the parent class or classes from which the
class class-name will derive (or inherit) predicates and facts (methods and
objects). If any parent classes are specified, the class class-name is called a
derived class.

Class implementation
A simplified "first-look" syntax for a class declaration is:
IMPLEMENT class-name [: parentclass-list ]
PREDICATES
predicatedeclaration-list

FACTS
factdeclaration-list

CLAUSES
Clause-list

ENDCLASS

The definition of the clauses for a class is done in a section starting with the
IMPLEMENT keyword and ending with the ENDCLASS keyword. Inside the
class implementation can be multiple Predicates, Facts and Clauses sections.
Unless the Predicates and Facts sections are preceded with the keyword STATIC,
the declarations work like they where given in the class declaration, meaning that
the facts will belong to the instance, and the predicates will carry the invisible
object pointer. However things declared in the implementation will be entirely
private to the class implementation.
Note also, that it is possible to inherit classes down in the implementation, thus
encapsulating details on the class implementation.

202 Visual Prolog Language Tutorial

Class instances
When a class declaration and definition has been made a multiple number of
instances of this class can be created. A new instance for a class is made with a
call to new for that class. New will return a reference to the instance, which can
then be used to perform operations on the object.
Example:
CLASS counter
PREDICATES
inc()
dec()
INTEGER getval()
ENDCLASS

IMPLEMENT counter
FACTS
single count(INTEGER)

CLAUSES
count(0).

inc:-
count(X),
X1=X+1,
assert(count(X1)).

dec:-
count(X),
X1=X-1,
assert(count(X1)).

getval(VAL):-
count(Val).
ENDCLASS

GOAL
O = counter::new,
Initial = O:getval(),
O:inc,
NewVal = O:getval(),
O:delete.

The output from the program in goal mode will be:

Chapter 174 Classes and objects 203

 O=145464, Initial=0, NewVal=1
1 Solution

Objects will survive a fail. Each instance of the class will have its own copy of
the facts database. Such a database can be manipulated as usual by retract, assert,
save, consult etc.
Note that specific new-predicates can be defined. These are called constructors
and will be explained later.

Destroying Objects
Error! Bookmark not defined.Objects have to be deleted explicitly by calling
delete for the object.
O = customers::new,
O:change_account,
O:delete.

Deleting an object will cause an automatic retracting of all facts in the database
for that instance.
Note that specific delete-predicates can be defined. These are called destructors
and will be explained later.

Class Domains
The declaration of a class generates a domain with the name of the class. This
domain can be used to declare parameters for predicates that should handle a
reference to the object.

CLASS customer
.....
ENDCLASS

PREDICATES
p(customer)

Passing an object in a parameter means just passing a pointer to the object as in

normal OOP-programmingstyle.

204 Visual Prolog Language Tutorial

Sub-classing and inheritance
In a class, both parent class predicates and global predicates can be redefined. For
example can the global predicates beep, concat etc. be overridden inside a class.
When using the child class, it is possible to use predicates and facts from both the
parent class and from the child class, however, the child class might choose to
redefine some of the predicates or facts.
CLASS person
FACTS
name( STRING )
father( person )
mother( person )

PREDICATES
write_info()

ENDCLASS

CLASS employe : person

FACTS
company(STRING Name)

PREDICATES
write_info()

ENDCLASS

IMPLEMENT person
CLAUSES
write_info():-
name(X),write("Name=",X),nl,fail.
write_info():-
father(F),write("Father:\n"),
F:person::write_info(),fail.
write_info():-
mother(M),write("Mother:\n"),
M:person::write_info(),fail.
write_info().
ENDCLASS

Chapter 174 Classes and objects 205

 IMPLEMENT employe
CLAUSES
write_info():-
this(O),
O:person::write_info(),fail.
write_info():-
company(X),write("Company=",X),nl,fail.
write_info().
ENDCLASS

GOAL
F = person::new(),
assert(F:name("Arne")),
O = employe::new(),
assert(O:name("Leo")),
assert(O:father(F)),
assert(O:company("PDC")),
O:write_info(),
O:delete().

The formal syntax for using the members of an object is:

[ObjectVariable:] [name_of_class:]
name_of_member[( list_of_arguments ) ]

The object can be omitted inside the implementation of a class or for calling the
members, which were declared as static. It will be considered as a call to the
corresponding member of that class (or its parent) in case it exists. Otherwise, if
there is no member with given name, it will be considered as a call to the
predicate with same name which must be previously declared in some
PREDICATES- or DATABASE section.
Names for members may be redefined in the hierarchy of classes. So, in order to
refer to names of members in the previous scope, the name of class defined in call
to some member may be used for explicit qualification of class in use.

Virtual Predicates
In Visual Prolog all Class predicates are what is in the C++ terminology called
Virtual methods. Virtual methods allow derived classes to provide different
versions of a parent class method. You can declare a method in a parent class and
then redefine it in any derived class.

206 Visual Prolog Language Tutorial

Assume, that a parent class P contains a who_am_i, and class D, derived from P,
has definitions for the predicate who_am_i. If who_am_i is called for an object of
D, the call made is D: who_am_i, even if the access is via a reference to P. For
example:
CLASS P
PREDICATES
test
who_am_i()
ENDCLASS

CLASS D : P
PREDICATES
who_am_i()
ENDCLASS

IMPLEMENT P
CLAUSES
test:-who_am_i().
who_am_i():-
write("I am of class P\n").
ENDCLASS

IMPLEMENT D
CLAUSES
who_am_i():-
write("I am of class D\n").
ENDCLASS

GOAL
O = D::new,
O:test,
O:delete.

The output from the above program would be:

I am of class D

Note, that if you define a predicate in a subclass with different domains or

number of arguments, Prolog will treat this as a different declaration, and it will
not work as a virtual predicate.

Chapter 174 Classes and objects 207

Static facts and predicates
It is possible to declare predicates or facts as being static, which for facts means
that the facts are not generated for each instance but there exists only one version
for the class. This is useful for example to count the number of instances for a
class. Preceding a predicate with the keyword static means that it will not carry
the invisible extra argument, which is a pointer to the actual instance.
Example:
CLASS Count
PREDICATES
procedure new( )
ENDCLASS

IMPLEMENT Count
STATIC FACTS
single countInstance( INTEGER )

CLAUSES
CountInstance( 0 ).

new( ):-
countInstance( Num ),
NumNext = Num +1,
assert( countInstance( NumNext ) ),
writef( "Count = %d\n", NumNext ).

ENDCLASS

GOAL
NewObject = count::new(),
NewObject1 = count::new().

The output of this program will be:

Count = 1
Count = 2
1 Solution

Class Scopes
A predicates scope is defined as the area, in which you can access it. Predicate
and fact names may be redefined in the class hierarchy. In order to refer to names
208 Visual Prolog Language Tutorial
in the previous scope the class-name::name() notation can be used to do an
explicit naming.
CLASS parent
PREDICATES
p(INTEGER)
ENDCLASS

CLASS child : parent

PREDICATES
p(STRING, INTEGER)
ENDCLASS

% IMPLEMENTATION not shown for clarity

GOAL
O = child::new,
O : parent:p(99) % Access the definition in parent

Another usage of the explicit scoping is in using classes with static predicates and
static facts as packages, like a module system:
Example:
ILIST = INTEGER*

CLASS list
static PREDICATES
append(ILIST, ILIST, ILIST) - (i,i,o)
ILIST gen(INTEGER)
ENDCLASS

IMPLEMENT list
CLAUSES
append([],L,L).
append([H|L1],L2,[H|L3]):-
append(L1,L2,L3).
gen(0,[]):-!.
gen(N,[N|L]):-
N1=N-1,
L = gen(N1).
ENDCLASS

Chapter 174 Classes and objects 209

 GOAL
L1 = list::gen(3),
L2 = list::gen(5),
list::append(L1,L2,L3).

Constructors and Destructors

The Visual Prolog system will itself allocate and initialize the memory during
creation of an object. However there might still be the desire to specify how an
object is created, for instance initialize facts to special values, create a window on
the screen or open a file. In the same way, a programmer may wish to control,
how an object is deleted, for instance closing windows or files. User defined
predicates for creating of deleting objects are called constructors and destructors.
A constructor is made by giving a definition for the predicate new for an object,
and a destructor is made by giving a definition for the predicate delete for an
object. In the clauses for the new predicate, It is possible to refer to the
constructors of the base class by baseclass::new.

CLASS mywind : wind

PREDICATES
new(INFO,Color)
ENDCLASS

IMPLEMENT mywind
CLAUSES
new(Info,Color):-
wind::new(... , ...,Color),
assert(info(INFO)).
ENDCLASS

GOAL
O = mywind::new(“info”,blue),
.....
O:delete.

Note! Any base class constructors must be called before doing any references to
the object, the compiler will check this.
The constructors and destructors are not allowed to fail, they are implicitly
declared as procedures.
If a constructor or destructor exits with a runtime error, the state of the object is
undefined.
new will automatically return a reference to the created instance.
210 Visual Prolog Language Tutorial
Reference to the Object Itself (This)
All the non-static predicates of an object have an invisible (to the programmer)
extra parameter, which is a pointer to the object.
In a clause like:
IMPLEMENT x
CLAUSES
inc:-
count(X),
X1=X+1,
assert(count(X1)).
ENDCLASS

The object is entirely invisible. If it is necessary to refer to the object itself for
instance to access a predicate in an inherited class, it is possible to use the built-in
predicate this. The predicate allows for an instance to get access to any
member-predicate, which is defined, in corresponding class or in its parents. The
syntax for making call to this predicate is:
this ( name_of_variable )

The usage of the predicate is allowed only in predicates, which were declared as
not static. This predicate has the single output-parameter.
For example:
IMPLEMENT x

CLAUSES
inc:-
this(Object),
Object:x::count(X),
X1=X+1,
assert(count(X1)).
ENDCLASS

This piece of code is functionally identical with the piece of code just above, with
the only difference, that you create a pointer to this object. This can be passed on
as a parameter to other predicates.

Chapter 174 Classes and objects 211

Abstract Classes
An abstract class is a class definition without an implementation. It is only
inherited by subclasses. The purpose of an abstract class is to have a declaration
of some predicates in order to have some other predicates working on a different
specialization of the more general class. An abstract class is defined by the
keyword ABSTRACT. In case an abstract class inherits some base classes, these
must also be declared abstract.

As example if you want to create a browser that can work on many different kinds
of data, you will opening the browser by passing it in an object which it can make
call to to get the data and move forward or backward. By using an abstract class,
the browser knows which predicates it can make calls to.
ABSTRACT CLASS browseinterface
PREDICATES
STRING get_Current()
next()
prev()
ENDCLASS

CLASS dbinterface : browseinterface

PREDICATES
new(DB_SELECTOR,CHAIN)
STRING get_Current()
next()
prev()
ENDCLASS

CLASS fileinterface : browseinterface

STRING get_Current()
next()
prev()
ENDCLASS

CLASS browser
PREDICATES
new(browseinterface)
ENDCLASS

212 Visual Prolog Language Tutorial

Protected facts and predicates
It is possible to declare whether you can access the facts or predicates from
outside of the class. By default all the facts and predicates are public which means
that they can be called from all other predicates.
The default access rights can be changed by preceding a fact or predicates
declaration with the keywords: protected. Protected means that all classes derived
from the class are allowed to access the fact or predicate, but the fact or predicate
can not be accessed from outside the class.
An example of the usage of protected predicates is call back event handlers,
where subclasses might redefine some handling predicates, but it makes no sense
to call these from the outside:
CLASS window
PROTECTED PREDICATES
onUpdate(RCT)
onCreate(LONG)
ENDCLASS

Derived class access control

An important issue in building the hierarchies of objects correct, so that you can
reuse as much code as possible, is inheritance. You can define methods on one
level, and these can then be reused on lower levels. If a class inherits from other
classes, we say, that this class is a derived class. When you declare a derived class
D, you list the parent classes P1, P2 .. in a comma-delimited parent class-list:
CLASS D : P1, P2 ...

D inherits all the facts and predicates of the parent class. Redefined names can be
accessed using scope overrides, if needed.
In case the name of some member can be achieved by several ways with the
directed acyclic graph (DAG), the longest way is accepted.
If 2 or more predicates in a class hierarchy are named the same, we say, that this
name is overloaded. We then have to consider the scope of the name, here defined
as the area in which each predicate is valid. Every usage of name for any member
of some class must be unambiguous (to an approximation of overloading). The
access to the member of base class is ambiguous if the expression of access
names more than one member. The test for unambiguous is performed before the
access control.

Chapter 174 Classes and objects 213

In case the synonymously defined name is the name of any overloaded member,
then the scope of overloading is performed after the ambiguous control (but
before the access control). The ambiguity can be scoped with explicit qualifying
name of the member by the name of the corresponding class.
All predicates from the class declaration, except new and delete, are virtual.
Opposite, all predicates declared in the implementation become non-virtual.

214 Visual Prolog Language Tutorial

Formal definition for classes
<class_definition> ::= <definition_begin> <definition_body> [
<class_section_end> ]
<definition_begin> ::= [ ABSTRACT ] CLASS <class_name> [ :
<parent_class_list> ]
<class_name> ::= identifier
<parent_class_list> ::= <parent_class_list_item> [ ,
<parent_class_list_item> . . . ]
<parent_class_list_item> ::= [ <parent_mode> ] [ <parent_access> ]
<class_name>
<definition_body> ::= { <definition_body_item> }
<definition_body_item> ::= <domains_section> |
[ <access_mode> ] [ STATIC ] <facts_section> |
[ <access_mode> ] [ <predicate_mode> ] <predicates_section>
<access_mode> ::= PROTECTED
<facts_section> ::= FACTS [ - <name_of_facts_section> ] {
<fact_item> }
<name_of_facts_section> ::= identifier
<fact_item> ::= [ <fact_kind> ] declaration_of_fact
<fact_kind> ::= SINGLE | DETERM | NODETERM
<predicate_mode> ::= STATIC
<predicates_section> ::= PREDICATES { declaration_of_predicate }
<class_section_end> ::= ENDCLASS [ <class_name> ]
<class_implementation> ::= <implement_begin> <implement_body>
<class_section_end>
<implement_begin> ::= IMPLEMENT <class_name>
<implement_body> ::= { <implement_body_item> }
<implement_body_item> ::= <domains_section> | <predicates_section> |
<clauses_section>
<clauses_section> ::= CLAUSES [ clause_section_body ]

Chapter 174 Classes and objects 215

CHAPTER 175

Advanced Topics
176This is an advanced chapter; we expect that you have been working with the
various examples earlier in this book and are now beginning to be an experienced
Visual Prolog user. In this chapter, we illustrate how you can control the flow
analysis by using the standard predicates free and bound, reference domains, how
to use them and how to separate them from the other domains. We also discuss
more advanced topics about domains, including the binary domain, pointers to
predicates and functions, and return values from functions. Finally, we look at
error-handling, dynamic cutting, free type conversions, and a discussion of some
programming style issues that will improve your programs' efficiency.

The Flow Analysis

In a given predicate call, the known arguments are called input arguments (i), and
the unknown arguments are called output arguments (o). The pattern of the input
and output arguments in a given predicate call is called the flow pattern.
For example, if a predicate is to be called with two arguments, there are four
possibilities for its flow pattern:
(i, i) (i, o) (o, i) (o, o)

When compiling programs, Visual Prolog carries out a global flow analysis of the
predicates. It starts with the main goal and then performs a pseudo-evaluation of
the entire program, where it binds flow patterns to all the predicate calls in the
program.
The flow analysis is quite simple; you are actually carrying it out yourself when
you write your program. Here are some examples:
GOAL
cursor(R, C), R1 = R+1, cursor(R1, C).

In the first call to the cursor, the two variables R and C are free; this means that
the cursor predicate will be called with the flow pattern cursor(o,o). You know
that the variables are free because this is the first time they've been encountered.

216 Visual Prolog Language Tutorial

In the expression R1=R+1, the flow analyzer knows that the variable R is bound
because it comes from the cursor predicate. If it were free, an error message
would have been issued. R1 will be a known argument after this call.
In the last call to cursor, both of the variables R1 and C have been encountered
before, so they will be treated as input arguments; the call will have the flow
pattern cursor(i,i).
For each flow pattern that a user-defined predicate is called with, the flow
analyzer goes through that predicate's clauses with the variables from the head set
to either input or output (depending on the flow pattern being analyzed).
Here's an example illustrating this:
% This example will only run for DOS Textmode Target
PREDICATES
changeattrib(Integer, Integer)

CLAUSES
changeattrib(NewAttrib, OldAttrib) :-
attribute(OldAttrib), attribute(NewAttrib).

GOAL
changeattrib(112, Old), write("Hello"),
attribute(Old), write(" there").

In the goal section, the first call to the predicate changeattrib is made with the
flow pattern changeattrib(i, o) (because 112 is known, and Old is not). This
implies that, in the clause for changeattrib, the variable NewAttrib will be an
input argument, and OldAttrib will be an output argument. Therefore, when the
flow analyzer encounters the first subgoal attribute(OldAttrib), the predicate
attribute will be called with the flow pattern attribute(o), while the second call
to attribute will have the flow pattern attribute(i). Finally, the call to attribute
in the goal will have an input flow pattern, because Old came out of changeattrib.

Compound Flow
If a predicate argument is a compound term it's also possible to have a compound
flow pattern, where the same argument has both input and output flow. Suppose
for instance that you have a database of information about countries. To enable
easy expansion with new data, it may well be desirable to contain each piece of
information in its own domain alternative:

Chapter 177 Advanced topics 217

 /* Program ch178e01.pro */

diagnostics

DOMAINS
cinfo = area(string,ulong); popn(string,ulong); apital(string,string)

PREDICATES
nondeterm country(cinfo)

CLAUSES
country(area("Denmark",16633)).
country(popn("Denmark",5097000)).
country(capital("Denmark","Copenhagen")).
country(area("Singapore",224)).
country(popn("Singapore",2584000)).
country(capital("Singapore","Singapore")).

The following depicts some of the different flow patterns country can be called
with:
country(C) (o)
country(area(Name,Area))area (o,o)
country(popn("Denmark",Pop))popn (i,o)
country(capital("Singapore","Singapore")) (i)

Note that as all elements of the term are known in the last example, the flow
pattern defaults to plain input.
Load 1 and try the examples above. When you look at the diagnostics output,
don't be confused by the presence of several flow variants as you progress. The
compiler keeps previously compiled code between executions of external goals,
unless the source code is changed.
When the domains involved in a compound flow pattern are reference domains,
the distinction between known and unknown arguments becomes blurred. We'll
return to this example in the reference domain section later.

Specifying Flowpatterns for Predicates

It is sometimes convenient to specify flowpatterns for your predicates. If you
know, that your predicates will only be valid for special flow patterns, it is a good
idea to specify flowpatterns for your predicates because the flow analyzer will
then catch any wrong usage of these predicates. After specifying the domains, a
dash and the possible flowpatterns can be given like in:

218 Visual Prolog Language Tutorial

PREDICATES
frame_text_mask(STRING,STRING,SLIST) - (i,o,o)(o,i,o)

Controlling the Flow Analysis

When the flow analyzer recognizes that a standard predicate is called with a
nonexistent flow pattern, it issues an error message. This can help you identify
meaningless flow patterns when you're creating user-defined predicates that call
standard predicates.
For example, if you use:
Z = X + Y

where the variable X or Y is not bound, the flow analyzer will give an error
message saying that the flow pattern doesn't exist for that predicate. To control
this situation, you can use the standard predicates free and bound.
Suppose you want to create a predicate for addition, plus, which can be called
with all possible flow patterns. Program 2 gives the code for such a predicate.
/* Program ch179e02.pro */

PREDICATES
nondeterm plus(integer, integer, integer)
nondeterm num(integer)

CLAUSES
plus(X,Y,Z):-
bound(X),bound(Y),Z=X+Y. /* (i,i,o) */
plus(X,Y,Z):-
bound(Y),bound(Z),X=Z-Y. /* (o,i,i) */
plus(X,Y,Z):-
bound(X),bound(Z),Y=Z-X. /* (i,o,i) */
plus(X,Y,Z):-
free(X),free(Y),bound(Z),num(X),Y=Z-X. /* (o,o,i) */
plus(X,Y,Z):-
free(X),free(Z),bound(Y),num(X),Z=X+Y. /* (o,i,o) */
plus(X,Y,Z):-
free(Y),free(Z),bound(X),num(Y),Z=X+Y. /* (i,o,o) */
plus(X,Y,Z):-
free(X),free(Y),free(Z),num(X),num(Y),Z=X+Y. /* (o,o,o) */

Chapter 177 Advanced topics 219

 % Generator of numbers starting from 0
num(0).
num(X):-
num(A),
X = A+1.

Reference Variables
When the flow analyzer has been through a clause, it checks that all output
variables in the clause head have been bound in the clause body. If a variable is
not bound in a clause, it needs to be treated as a reference variable. Here's an
example demonstrating this dilemma:
PREDICATES
p(integer)

CLAUSES
p(X):- !.

GOAL
p(V), V = 99, write(V).

In the goal, the predicate p is called with an output pattern but, in the clause for p,
the argument X is not bound. When the flow analyzer recognizes this, it will take
a look at the domain corresponding to the variable. If the domain is already
declared as a reference domain, there's no problem; if it's not, Visual Prolog gives
a warning.
When a variable is not bound in a clause, the clause can't return a value. Instead,
it will return a pointer to a reference record where the actual value can be inserted
at a later time. This requires that the whole domain be treated equally; instead of
just passing the values directly for some of the variables of that type, pointers to
records will be passed through arguments belonging to the reference domain.
When a compound domain becomes a reference domain, all of its subdomains
must also become reference domains, because they must also be capable of
containing free variables. If you just declare a compound domain to be a
reference domain, the compiler will automatically know that all the subdomains
are also reference domains.

Declaring Domains as Reference

When the flow analyzer encounters an unbound variable, it will only give a
warning if the variable is not bound on return from a clause. If you accept this,
the domain will automatically be treated as a reference domain. However, you

220 Visual Prolog Language Tutorial

should always explicitly declare the domains intended to be reference domains in
the domains section. This is also required in projects (programs consisting of
several modules); when global domains should handle unbound values, the
compiler will not allow automatic conversion of the domains. Global domains
and predicates are covered in the chapter 180.
Notice that the following special predefined domains are not allowed to become
reference domains: file, reg, db_selector, bt_selector, and place.

Reference Domains and the Trail Array

Because coercion’s and some extra unification are needed, reference domains will
in general give a reduction in execution speed. However, some problems can be
solved far more elegant and efficiently when you use reference domains, and
Visual Prolog has facilities to limit their effect.
When you use reference domains, Visual Prolog uses the trail array. The trail
array is used to remember when reference variables become instantiated. This is
necessary because if you backtrack to a point between the creation and the
instantiation of a reference variable, it must be uninstantiated. This problem
doesn't exist with ordinary variables, as their points of creation and instantiation
are the same. Each instantiation recorded in the trail uses 4 bytes (the size of a 32-
bit pointer). However, the trail usage is heavily optimized and no record will be
placed there if there are no backtrack points between the variable's creation and
instantiation.
The trail is automatically increased in size when necessary. The maximum size is
64K in the 16-bit versions of Visual Prolog, and practically unbounded in the 32-
bit versions.
Because the code is treated the same for the whole domain, it is usually not a
good idea to treat the basic domains as reference domains. Instead, you should
declare a domain as being a reference domain to the desired base domain. For
instance, in the following code excerpt, the user-defined domain refinteger is
declared to be a reference domain to the integer domain. All occurrences of
refinteger types will be handled as reference domains, but any occurrence of
other integers will still be treated as integers.
DOMAINS
refinteger = reference integer

PREDICATES
p(refinteger)

CLAUSES
p(_).

Chapter 177 Advanced topics 221

Using Reference Domains
The right way to use reference domains is to use them only in the few places
where they are needed and to use non-reference domains for all the rest. Visual
Prolog allows you to convert reference domains to non-reference domains
whenever needed. For example, you can create a predicate that converts a
reference integer to a non-reference integer with a single fact:
DOMAINS
refint = reference integer

PREDICATES
conv(refint,integer)

CLAUSES
conv(X, X).

Visual Prolog does the conversion automatically when the same variable is used
with both a reference domain and a non-reference domain, as it does in the clause
when converting X from a refint to an integer. The above is only an explicit
example, you don't need to write any special code to convert from reference to
non-reference domains. Note that the reference variable needs to be instantiated
to a value before it can be converted to the non-reference value. In the same way,
if you try to convert a variable from one reference domain to another (such as
from reference integers to reference characters), you should make sure the value
is bound. Otherwise, Visual Prolog will issue an error message to the effect that
free variables are not allowed in the context.
Pay attention to the automatic type conversions when you're creating a new free
reference variable through a call to free, like so:
free(X), Y = X, bind_integer(X), ...

or creating a free variable with the = predicate (equal), like this:

Y = X, bind_integer(X), ...

In these examples, free and the = predicate have difficulty finding the correct
domain. The type-checker will try to find a suitable domain for the variable
during backtracking by means of successive attempts to carry out the flow
analysis. The type-checker starts with the character domain and (because char
types can be converted to integer types) will choose the character domain instead
of proceeding to the integer domain.
With reference domains you can return variables that will receive values at a later
point. You can also create structures where some places are left uninstantiated
until later.
222 Visual Prolog Language Tutorial
Example
To get a feel for how reference domains work, you should try some goals with the
well-known predicates member and append:
/* Program ch181e03.pro */

DOMAINS
refinteger = integer
reflist = reference refinteger*

PREDICATES
nondeterm member(refinteger, reflist)
append(reflist, reflist, reflist)

CLAUSES
member(X,[X|_]).
member(X,[_|L]):-
member(X,L).

append([],L,L).
append([X|L1],L2,[X|L3]):-
append(L1, L2, L3).

Load and run this example program, and try the following goals:
member(1,L). /* Give all lists where 1 is a member */
member(X,L), X=1. /* Same as before */
member(1,L), member(2,L). /* Lists where both 1 and 2 are members */
X=Y,member(X,L),member(Y,L), X=3.% Both X and Y are members in the list
member(1,L), append(L,[2,3],L1).
append(L,L,L1), member(1,L). /* All lists where 1 is A member twice */

You will discover that the answers are what you logically expect them to be.

Flow Patterns Revisited

A reference variable may well be unbound and yet exist at the time it's used in a
predicate call. In example 1, this will happen if for instance you want to find all
countries having the same name as their capital, using e.g.
samecaps:- country(capital(C,C)), write(C,'\n'), fail.

Here the variable C is used twice with output flow, but what the code really says
is that the two variables in capital should share the same value once one of them
becomes instantiated. Therefore, both variables are created and unified before the
call. In order to do this their domain is converted to a reference domain, and both

Chapter 177 Advanced topics 223

variables are in effect known at the time of call, giving a straight input flow
pattern.
Note that, as said before, it's bad practice to let the standard domains become
reference domains. If you want to use the above call, you should declare a
suitable reference domain. However, this would create an overhead in all accesses
of the country predicate, and it would probably be more efficient to use
backtracking to find the special case where name and capital are identical, by
using.
country(capital(Co,Ca)), Co = Ca, !, ...

Whether this is true or not depends on the size of the database, how many times
you perform the call, how many other calls you have, how the arguments are used
after the calls, etc.

Using Binary Trees with Reference Domains

In chapter 182, you saw how binary trees could be used for fast and efficient
sorting. However, sorting can actually be done in a more elegant fashion with
reference domains. Because there is no way to change the leaves of a tree when
they get new values, a lot of node copying occurs when the tree is created. When
you are sorting large amounts of data, this copying can result in a memory
overflow error. A reference domain can handle this by letting the leaves of the
tree remain as free variables (where the subtrees will later be inserted). By using a
reference domain this way, you don't need to copy the tree above the place where
the new node is to be inserted.
Consider the predicate insert in ch183e04.pro during the evaluation of the
following goal:
GOAL
insert("tom", Tree),
insert("dick", Tree),
insert("harry", Tree).

In this program, the insert predicate creates a binary tree using the reference
domain tree.
/* Program ch184e04.pro */

DOMAINS
tree = reference t(val, tree, tree)
val = string

224 Visual Prolog Language Tutorial

PREDICATES
insert(val, tree)

CLAUSES
insert(ID,t(ID,_,_)):-!.
insert(ID,t(ID1,Tree,_)):-
ID<ID1,
!,insert(ID,Tree).
insert(ID,t(_,_,Tree)):-
insert(ID,Tree).

GOAL
insert("tom",Tree),
insert("dick",Tree),
insert("harry",Tree),
write("Tree=",Tree),nl

The first subgoal, insert("tom",Tree), will match with the first rule, and the
compound object to which Tree is bound takes this form:
t("tom", _, _)

Even though the last two arguments in t are not bound, t carried is forward to the
next subgoal evaluation:
insert("dick", Tree)

This, in turn, binds Tree to

t("tom", t("dick", _, _), _)

Finally, the subgoal

insert("harry", Tree)

binds Tree to
t("tom", t("dick", _, t("harry", _, _)), _)

which is the result returned by the goal.

Sorting with Reference Domains

In this section, we add onto the preceding binary tree example (ch185e04.pro) to
show how you can isolate the use of reference domains and convert between
reference and non-reference domains. The next example defines a predicate that
is able to sort a list of values.
Chapter 177 Advanced topics 225
 /* Program ch186e05.pro */

DOMAINS
tree = reference t(val, tree, tree)
val = integer
list = integer*

PREDICATES
insert(integer,tree)
instree(list,tree)
nondeterm treemembers(integer,tree)
sort(list,list)

CLAUSES
insert(Val,t(Val,_,_)):-!.
insert(Val,t(Val1,Tree,_)):-
Val<Val1,!,
insert(Val,Tree).
insert(Val,t(_,_,Tree)):-
insert(Val,Tree).

instree([],_).
instree([H|T],Tree):-
insert(H,Tree),
instree(T,Tree).

treemembers(_,T):-
free(T),!,fail.
treemembers(X,t(_,L,_)):-
treemembers(X,L).
treemembers(X,t(Refstr,_,_)):-
X = Refstr.
treemembers(X,t(_,_,R)):-
treemembers(X,R).

sort(L,L1):-
instree(L,Tree),
findall(X,treemembers(X,Tree),L1).

GOAL
sort([3,6,1,4,5],L),
write("L=",L),nl.

In this example, note that reference domains are only used in the tree. All the
other arguments use non-reference domains.

226 Visual Prolog Language Tutorial

Functions and Return Values

Visual Prolog includes syntax for letting predicates be considered functions

having a return value, rather than plain predicates using an output argument. The
difference is a bit more than syntactic, however. Because return values are stored
in registers, Prolog functions can return values to, and get return values from,
foreign languages, but that's an issue covered in the chapter 187.
A function declaration looks like an ordinary predicate declaration, except that
the name is prefixed by the domain it's returning:
PREDICATES
unsigned triple(unsigned)

However, the clauses for a function should have an extra last argument, to be
unified with the return value upon success:
CLAUSES
triple(N,Tpl):- Tpl = N*3.

GOAL
TVal = triple(6), write(TVal).

The return value need not be one of the standard domains, it can be any domain.
If you declare a function that doesn't take any arguments, you must supply an
empty pair of brackets when calling it, in order to distinguish it from a string
symbol. Given for instance a function to return the hour of the day
PREDICATES
unsigned hour()

CLAUSES
hour(H):- time(H,_,_,_).

you must call it like this:

..., Hour = hour(), ...

and not like this

..., Hour = hour, ...

as this will simply consider hour to be the text string "hour", following which the
compiler will complain about type errors once you try to use Hour.

Chapter 177 Advanced topics 227

It is also recommended to supply an empty pair of brackets in the declaration of
functions and predicates having no arguments. If not, confusing syntax errors may
result from misunderstandings between predicate names and domain names, if
they clash. If for instance you have a domain named key and you also have a
predicate named key, then the declaration:
PREDICATES
key
mypred

can be interpreted in two ways: 1) a predicate named key and a predicate named
mypred; 2) a predicate name mypred returning a key. If instead you write:
PREDICATES
key()
mypred()

all ambiguity is resolved.

Note that when a predicate is declared as a function, having a return value, it
cannot be called as an ordinary Prolog predicate using the extra argument as an
output argument; it must be called as a function. The reason for this is that
functions store return values in registers, meaning that the code compiled before
and in particular after a function call is different from the code around a call of an
ordinary predicate. For the same reason, functions calling themselves are
currently not tail recursive but this may change in future versions of Visual
Prolog.
For instance, if you write a function neg to negate each element in a list, like this:
DOMAINS
ilist = integer*

PREDICATES
ilist neg(ilist)

CLAUSES
neg([],[]).
neg([Head|Tail],[NHead|NTail]):-
NHead = -Head,
NTail = neg(Tail).

it is not tail-recursive, while neg as a predicate:

DOMAINS
ilist = integer*

228 Visual Prolog Language Tutorial

PREDICATES
neg(ilist,ilist)

CLAUSES
neg([],[]).
neg([Head|Tail],[NHead|NTail]):-
NHead = -Head,
neg(Tail,NTail).

is tail-recursive. Therefore, don't overdo the use of functions. Their primary aim
is to enable you to get returned values from, and return values to, foreign
language routines.
As a final note, you should be aware that functions with arithmetic return values
must be deterministic if they take part in arithmetic expressions.

Determinism Monitoring in Visual Prolog

Most programming languages are deterministic in nature. That is, any set of input
values leads to a single set of instructions used to produce output values.
Furthermore in most languages, for example in C, a called function can produce
only a single set of output values. On the contrary, Visual Prolog naturally
supports non-deterministic inference based on non-deterministic predicates.
The object behind the determinism monitoring is to save run-time storage space.
In fact, when a deterministic clause succeeds, the corresponding run-time stack
space can be dispensed with at once, thus freeing the storage it occupied. There
are a number of reasons why determinism should also concern programmers,
most of them involving programming optimization.
Visual Prolog has a strongly typed determinism system. Visual Prolog's
determinism checking system enforces the programmer to declare the following
two behavior aspects of predicates:
whether a call to a predicate can fail;
number of solutions a predicate can produce.
According to these aspects of determinism the following Types of Predicates
(rules) are supported in Visual Prolog:

Chapter 177 Advanced topics 229

 Predicate Types
Number of Solutions, that can be produced
0 1 >1
Can fail:| failure determ nondeterm
Never
fails: erroneous procedure multi

Using keywords from the above table in declarations of predicates and predicate
domains the programmer can declare the six different types of predicates.
Applied these aspects of determinism to declarations of facts we can obtain the
following table:

Facts Types
Number of Solutions can be Produced
0 1 >1
Can fail:| determ nondeterm
Never
fails: single

In this table Never fails means that as less one instance of a fact always exists,
and therefore such a fact never fails if it is called with free arguments. Using
keywords from the above table in declarations of facts the programmer can
declare the three different types of database predicates (facts).

Visual Prologs determinism checking system

Visual Prolog offers unique determinism monitoring facilities based on
declarations of types of predicates and facts. All Visual Prolog's standard
predicates are internally defined as nondeterm, multi, determ, procedure, failure
or erroneous. For user-defined predicates declared with the keywords determ,
procedure, failure or erroneous, the compiler always checks and gives warnings
for each program clause that results in a non-deterministic predicate. There are
two kinds of non-deterministic clauses:
If a clause does not contain a cut, and there are one or more clauses that can
match with the same input arguments for that flow pattern.

230 Visual Prolog Language Tutorial

 If a clause calls a non-deterministic predicate, and that predicate call is not
followed by a cut.
Because of the second reason above, non-determinism has a tendency to spread
like wildfire throughout a program unless (literally) cut off by one or more cuts.
By default, the compiler checks clauses and gives a warning (595 or 596) if it
cannot guarantee that a predicate corresponds to the declared type. For example,
if the compiler is unable to guarantee that a predicate declared with the keyword
multi, procedure or erroneous never fails.
Take into account that the compiler is able to verify only necessary conditions for
fail (not necessary and sufficient). Therefore, the compiler can sometimes
generate warnings 595 and 596 for predicates (declared with the keyword multi,
procedure or erroneous) that, in fact, will never fail. For example,
PREDICATES
procedure str_chrlist(STRING,CHARLIST) - (i,o)

CLAUSES
str_chrlist("",[]):-!.
str_chrlist(Str,[H|T]):-
frontchar(Str,H,Str1),
str_chrlist(Str1,T).

The frontchar predicate can fail if the first parameter is an empty string. The
compiler is not sophisticated enough to detect that Str in the second clause cannot
be empty. For this example the compiler will generate warning 595 "Non-
procedure clause in the procedure predicate".
Of course, non-procedure predicates checking can be switched off if the
programmer specifies the compiler option -upro- (), but it is not good
programming style. Instead you should reorder the clauses like:
str_chrlist(Str,[H|T]):-
frontchar(Str,H,Str1),!,
str_chrlist(Str1,T).
str_chrlist(_,[]):-!.

The declaration of procedures catches many small mistakes, like forgetting a

catchall clause.
There are two rules that you must use when writing predicates declared with the
keyword multi, procedure or erroneous:
If anyone predicate clause can fail than the final catchall clause must be
defined (see the str_chrlist example above).

Chapter 177 Advanced topics 231

 For any possible (according to declared domains) set of input arguments, a
clause, having a head which matches this set, must exist. Otherwise, the
compiler will generate the warning 596.
For instance, in the following example the third clause p(_) can be redundant if
the predicate is declared without procedure keyword, but is required to satisfy this
rule if the predicate is declared as procedure.
DOMAINS
BOOLEAN = INTEGER % b_True = 1, b_False = 0

PREDICATES
procedure p(BOOLEAN)

CLAUSES
p(b_False):- !, ... .
p(b_True): - !, ... .
p(_): - dlg_error("An illegal argument value").

Notice that the compiler handles erroneous predicates in a special way providing
possibility to use them in the final catchall clauses (for handling error situations)
in predicates of other types. For instance, the catchall clause in the previous
example can be rewritten as the following:
p(_): - errorexit(error_vpi_package_bad_data).

Predicates as Arguments

So far we have only seen predicate calls of a static nature. That is, the predicates
being called as subgoals are specified satirically in the source code. However, in
many cases it may be desirable to call different predicates, depending on previous
events and evaluations, from the same place, to avoid large-scale duplication of
code. To this end, you can declare a predicate domain, and pass pointers to
predicates of that domain as variables.
The main usage of this feature in Visual Prolog is to pass eventhandler predicates
to the VPI layer.

Predicate Domains
The declaration for a predicate domain is of the form

232 Visual Prolog Language Tutorial

 pdom = { determ | nondeterm | multi | procedure | failure | erroneous }
[ domain ] arglist [ - flowpattern ] [ language ]
(curly braces indicate "choose one", square brackets indicate optional items)
where
domain is the return domain, if you're declaring a function
arglist is of the form
( [ domain [ , domain ]* ] )

flowpattern is of the form

( flow [ , flow ]* )
where flow is
{ i | o | functor flowpattern | listflow }
where listflow is
'[' flow [ , flow ]* [ '|' { i | o | listflow } ] ']'

language is of the form

language { c | asm | pascal | stdcall | syscall }

The language specification tells the compiler which calling convention to use,
and is only required when declaring domains for routines written in other
languages (see the chapter 188).
The flowpattern specifies how each argument is to be used. It should be the letter
i for an argument with input flow, the letter o for one with output flow, a functor
and flowpattern for a compound term (e.g. (i,o,myfunc(i,i),o)), or a listflow
(eg [i,myfunc(i,o),o], [o,o|i]).

You can have no more than one flowpattern declaration for a predicate pointer
domain, and it must be given unless the argument list is empty.
Hence, the declaration for a group of deterministic predicates taking an integer as
argument and returning an integer, would be
DOMAINS
list_process = determ integer (integer) - (i)

This group, or class of predicates, is now known as list_process. To declare a

predicate square as belonging to this group, the syntax is:
PREDICATES
square: list_process

The clause for square is just like an ordinary clause, but as it's declared as a
function it needs a return argument:

Chapter 177 Advanced topics 233

 CLAUSES
square(E,ES):- ES = E*E.

Elaborating on the above, a domain declaration for a group of deterministic

predicates, to be called ilist_p, taking an integer list and a pointer to a list_process
predicate as input arguments, and an integer list as output argument, would hence
be
DOMAINS
ilist = integer*
list_process = determ integer (integer) - (i)
ilist_p = determ (ilist,list_process,ilist) - (i,i,o)

Examples
Now look at the following program:
/* Program ch189e06.pro */

DOMAINS
ilist = integer*
list_process = determ integer (integer) - (i)
ilist_p = determ (ilist,list_process,ilist) - (i,i,o)

PREDICATES
list_square: list_process
list_cube: list_process
il_process: ilist_p

CLAUSES
list_square(E,ES):- ES = E*E.
list_cube(E,EC):- EC = E*E*E.

il_process([],_,[]).
il_process([Head|Tail],L_Process,[P_Head|P_Tail]):-
P_Head = L_Process(Head),
il_process(Tail,L_Process,P_Tail).

GOAL
List = [-12,6,24,14,-3],
il_process(List,list_square,P_List1),
write("P_List1 = ",P_List1,'\n'),
il_process(List,list_cube,P_List2),
write("P_List2 = ",P_List2,'\n').

This declares two functions, list_square and list_cube, belonging to the

list_process group, and a predicate il_process creating a new integer list by
234 Visual Prolog Language Tutorial
applying the listelement-processing predicate L_Process to each element of the
original list. Note that the domain declaration ilist_p is only included for
illustration; il_process could equally well have been declared using:
PREDICATES
il_process(ilist,list_process,ilist)

since it isn't referred to as a variable.

With the goal shown, il_process is called twice, first creating a list of squares by
applying the list_square function, and then a list of cubes by applying the
list_cube function. Compile and run this program, and you will get:
P_List1 = [144,36,576,196,9]
P_List2 = [-1728,216,13824,2744,-27]

Make sure you understand the complexities of this, and, when you do, make sure
you don't abuse it. It's all too easy to create totally unreadable programs. Program
ch190e07, which is a somewhat elaborated version of ch191e06, illustrates the
concept taken to a resonable limit:
/* Program ch192e07.pro */

DOMAINS
ilist = integer*
list_process = determ integer (integer) - (i)
list_p_list = list_process*
elem_process = determ (integer,integer,integer) - (i,i,o)
elem_p_list = elem_process*

PREDICATES
list_same: list_process
list_square: list_process
list_cube: list_process

elem_add: elem_process
elem_max: elem_process
elem_min: elem_process

il_process(ilist,list_process,ilist)
il_post_process(ilist,elem_process,integer)

apply_elemprocess(ilist,elem_p_list)
apply_listprocess(ilist,list_p_list,elem_p_list)

string lpname(list_process)
string epname(elem_process)

Chapter 177 Advanced topics 235

CLAUSES
lpname(list_same,list_same). % Map predicate pointer to its name
lpname(list_square,list_square).
lpname(list_cube,list_cube).

epname(elem_add,elem_add).
epname(elem_min,elem_min).
epname(elem_max,elem_max).

elem_add(E1,E2,E3):- E3 = E1+E2.
elem_max(E1,E2,E1):- E1 >= E2, !.
elem_max(_,E2,E2).
elem_min(E1,E2,E1):- E1 <= E2, !.
elem_min(_,E2,E2).

list_same(E,E).
list_square(E,ES):- ES = E*E.
list_cube(E,EC):- EC = E*E*E.

il_process([],_,[]).
il_process([Head|Tail],E_Process,[P_Head|P_Tail]):-
P_Head = E_Process(Head),
il_process(Tail,E_Process,P_Tail).

il_post_process([E],_,E):-!.
il_post_process([H|T],L_Process,Result):-
il_post_process(T,L_Process,R1),
L_Process(H,R1,Result).

apply_elemprocess(_,[]).
apply_elemprocess(P_List,[E_Process|E_Tail]):-
il_post_process(P_List,E_Process,PostProcess),
NE_Process = epname(E_Process),
write(NE_Process,": Result = ",PostProcess,'\n'),
apply_elemprocess(P_List,E_Tail).

apply_listprocess(_,[],_).
apply_listprocess(I_List,[L_Process|L_Tail],E_List):-
il_process(I_List,L_Process,P_List),
NL_Process = lpname(L_Process),
write('\n',NL_Process,":\nProcessed list = ",P_List,
"\nPost-processing with:\n"),
apply_elemprocess(P_List,E_List),
apply_listprocess(I_List,L_Tail,E_List).

236 Visual Prolog Language Tutorial

GOAL
List = [-12,6,24,14,-3],
write("Processing ",List," using:\n"),nl,
apply_listprocess(List,[list_same,list_square,list_cube],
[elem_add,elem_max,elem_min]).

Among other things, this program illustrates the use of lists of predicate pointers.
If you run it, you'll get the following output:
Processing [-12,6,24,14,-3] using:

list_same:
Processed list = [-12,6,24,14,-3]
Post-processing with:
elem_add: Result = 29
elem_max: Result = 24
elem_min: Result = -12

list_square:
Processed list = [144,36,576,196,9]
Post-processing with:
elem_add: Result = 961
elem_max: Result = 576
elem_min: Result = 9

list_cube:
Processed list = [-1728,216,13824,2744,-27]
Post-processing with:
elem_add: Result = 15029
elem_max: Result = 13824
elem_min: Result = -1728

Predicate pointers may be used like almost any other object in a program. In
particular, they can appear as parts of compound terms, creating object oriented
possibilities where each object carries with it a series of routines for its own
management. You should take note, though, that predicate pointers is a fairly low-
level mechanism. The actual value of such a pointer is simply a code-address, and
it's therefore only valid in the particular program where it was created. Hence,
although you can store and retrieve predicate pointers via the databases, highly
unexpected and quite possibly disastrous results will occur if you try to use a
predicate pointer not originating in the current program.

Chapter 177 Advanced topics 237

The Binary Domain

Visual Prolog has a special binary domain for holding binary data, as well as
special predicates for accessing individual elements of binary terms. The main
use for binary terms is to hold data that has no reasonable representation
otherwise, such as screen bitmaps and other arbitrary memory blocks. There are
separate predicates for reading binary terms from, and writing them to, files.
These will be discussed in chapter 193. With the help of the built-in conversion
predicate term_bin, conversion from things such as binary file-headers to Prolog
terms is a snap, and binary items going into or coming out of foreign language
routines are easily handled. Finally arrays may also be implemented easily and
efficiently.
Binary terms is a low-level mechanism, whose primary aim is to allow easy and
efficient interfacing to other, non-logical, objects, and foreign languages. To this
end, binary terms do not behave like other Prolog terms with respect to
backtracking. Binary terms will be released if you backtrack to a point previous to
their creation, but if you don't backtrack that far any changes done to the term will
not be undone. We will illustrate this in the example program at the end of this
section.

Implementation of binary terms

Pointer

A binary term is simply a sequence of

bytes, Size bytes preceded by a word (16bit platforms)
or ^ dword (32bit platforms), holding its
size. |

When interfacing to other languages,

you should be aware that a term of binary
type (the variable passed in the foreign
language function call) points to the actual contents, not the size. The Size field
includes the space taken up by the field itself. Binary terms are subject to the
usual 64K size restriction on 16bit platforms.

Text syntax of Binary Terms

Binary terms can be read and written in text format, and also specified in source
form in Visual Prolog source code. The syntax is:
$[b1,b2,...,bn]
238 Visual Prolog Language Tutorial
where b1, b2, etc. are the individual bytes of the term. When a binary term is
specified in source form in a program, the bytes may be written using any suitable
unsigned integral format: decimal, hexadecimal, octal, or as a character.
However, the text-representation of binary terms created and converted at run-
time is fixed hexadecimal, with no leading "0x" on the individual bytes. Program
8 illustrates this:
/* Program ch194e08.pro */

GOAL
write("Text form of binary term: ",$['B',105,0o154,0x73,'e',0],'\n').

Load and run this program, and Visual Prolog will respond
Text form of binary term: $[42,69,6C,73,65,00]

You should hence be careful if you use e.g. readterm to read a binary term at
runtime.

Creating Binary Terms

Below we discuss the standard predicates Visual Prolog includes, for creation of
binary terms.

makebinary/1
makebinary creates and returns a binary term with the number of bytes specified,
and sets its contents to binary zero.
..., Bin = makebinary(10), ...

The number of bytes should be the net size, excluding the size of the size field.

makebinary/2
makebinary is also available in a two-arity version, allowing specification of an
element size.
..., USize = sizeof(unsigned), Bin = makebinary(10,USize), ...

This creates a binary term with a size given by the number of elements (10 in the
above example) multiplied by the elementsize (sizeof(unsigned) in the above),
and sets its contents to zero.

Chapter 177 Advanced topics 239

composebinary/2
composebinary creates a binary term from an existing pointer and a length. It's
useful in converting pointers to arbitrary blocks of memory returned by foreign
language functions. composebinary takes two arguments, and returns a binary.
..., Bin = composebinary(StringVar,Size), ...

composebinary takes a copy of the StringVar given as input, so changes to the

binary term will not affect StringVar, and vice versa.

getbinarysize/1
getbinarysize returns the net size (in bytes) of the binary term, excluding the size
field in front of the data.
..., Size = getbinarysize(Bin), ...

Accessing Binary Terms

There are eight predicates for accessing binary terms, four for setting entries and
four for getting entries. Both groups perform range checking based on the size of
the binary term, the index specified, and the size of the desired item (byte, word,
dword, or real). It's an error to try to get or set entries outside the range of the
binary term.
Take special note that indices (element numbers) are 0-relative; the first element
of a binary term has index 0, and the last element of an N-element binary term has
index N-1.

getentry/2
getentry is either getbyteentry, getwordentry, getdwordentry, or getrealentry,
accessing and returning the specified entry as a byte, word, dword, or real,
respectively.
..., SomeByte = getbyteentry(Bin,3), ...

setentry/3
setentry is the counterpart to getentry, setting the specified byte, word, dword, or
real entry.
..., setbyteentry(Bin,3,SomeByte), ...

240 Visual Prolog Language Tutorial

Unifying Binary Terms
Binary terms may be unified just like any other term, in clause heads or using the
= predicate:
..., Bin1 = Bin2, ...

If either of the terms is free at the time of unification, they will be unified and
point to the same binary object. If both are bound at the time of unification, they
will be compared for equality.

Comparing Binary Terms

The result of comparing two binary terms is as follows:
If they are of different sizes, the bigger is considered larger; otherwise,
they're compared byte by byte, as unsigned values; comparison stops when
two differing bytes are found, and the result of their comparison is also the
result of the comparison of the binary terms.
For instance, $[1,2] is bigger than $[100], and smaller than $[1,3].

Example
Program 9 demonstrates a number of aspects of binary terms.
/* Program ch195e09.pro */

PREDICATES
comp_unify_bin
comp_unify(binary,binary)
access(binary)

CLAUSES
comp_unify_bin:-
Bin = makebinary(5),
comp_unify(Bin,_),
comp_unify($[1,2],$[100]),
comp_unify($[0],Bin),
comp_unify($[1,2,3],$[1,2,4]).

Chapter 177 Advanced topics 241

 comp_unify(B,B):-!,
write(B," = ",B,'\n').
comp_unify(B1,B2):-
B1 > B2,!,
write(B1," > ",B2,'\n').
comp_unify(B1,B2):-
write(B1," < ",B2,'\n').

access(Bin):-
setwordentry(Bin,3,255),
fail. % Changes are not undone when backtracking!
access(Bin):-
Size = getbinarysize(Bin),
X = getwordentry(Bin,3),
write("\nSize=",Size," X=",X," Bin=",Bin,'\n').

GOAL
% Illustrate comparison and unification of binary terms
comp_unify_bin,

% Allocate a binary chunk of 4 words

WordSize = sizeof(word),
Bin = makebinary(4,WordSize),
access(Bin),

% Illustrate range checking; element numbers are 0-relative

write("Run-time error due to wrong index:\n"),
Index = 4,
trap(setwordentry(Bin,Index,0),E,
write("Error ",E," setting word index ",Index," of ",Bin,'\n')).

This example uses the trap predicate, which will be discussed in the section about
error handling below.

Converting Terms to Binary Terms

A compound term may have its arguments scattered all over memory, depending
on what domains they belong to. Simple types are stored directly in the term
record itself, while complex types (those accessed via a pointer, and allocated
separately on the global stack) will not necessarily be anywhere near the term
they appear in. This is a problem if a term has to be sent out of a program, so to
speak, as there is no way make an explicit copy of its contents. Unifying a term
variable with another variable will only take a copy of the pointer to the term.

242 Visual Prolog Language Tutorial

Using term_str (discussed in chapter 196), it is possible to convert the term to a
string and back again, but this is rather inefficient when all that's needed is a copy
of the term's contents.
term_bin solves this problem.

term_bin/3
term_bin will convert between a term of any domain and a block of binary data,
holding the term's contents as well as pointer fixup information. The pointer fixup
information will be applied to the binary data when converted back to a term,
allowing recreation of any pointers to complex terms the term contains.
term_bin looks like this:
term_bin(domain,Term,Bin) /* (i,i,o) (i,_,i) */

The domain is the domain the Term belongs, or should belong, to, and Bin is a
binary term holding the Term's contents.

Example
Program 11 demonstrates conversion between a term and its binary
representation. The domains and alignment have been explicitly chosen to ease
description, as they would otherwise differ between 16- and 32-bit platforms.
Alignment of terms is usually only relevant when interfacing to foreign
languages, and is fully described in the chapter 197.
/* Program ch198e11.pro */

DOMAINS
dom = align byte cmp(string,short)

GOAL
T = cmp("Bilse",31),
term_bin(dom,T,B),
write("Binary form of ",T,":\n",B),
term_bin(dom,T1,B),
write("\nConverted back: ",T1,'\n').

If you run this, you'll get:

Binary form of cmp("Bilse",31):
$[01,07,00,00,00,1F,00,42,69,6C,73,65,00,01,00,00,00,01,00,00,00]
Converted back: cmp("Bilse",31)

Chapter 177 Advanced topics 243

You shouldn't be too concerned about the actual format of this, in particular as
we're dealing with implementation details, which may change. Nevertheless, we'll
briefly describe the contents of the binary information:
$[01,07,00,00,00,1F,00,42,69,6C,73,65,00,01,00,00,00,01,00,00,00]
| |_________| |___| |_______________| |_________| |_________|
| | | | | |
functor | 31 "Bilse"\0 offset of # of ptrs
| ptr to fix in fixup
0-relative (array, but
ptr to string only one element here)

The offset of ptr to fix array will be 16-bit quantities on 16-bit platforms, as will
the # of ptrs in fixup.
If the term contains elements from the symbol domain, the binary term will
contain additional information to insert the symbols in the symboltable when the
term is recreated.
Visual Prolog uses term_bin itself when storing things in the database system and
when sending terms over a message pipe to another program. If several programs
share external databases or communicate over pipes, it's hence crucial that the
domains involved use the same alignment.

Errors and Exception Handling

As software quality improves, error handling becomes increasingly important in

providing safe and trustworthy programs that users feel they can rely on. In this
section we look at the standard predicates Visual Prolog provides, giving you
control over the errors and exceptions that may occur when your application is
running. This includes trapping run-time errors and controlling user interruption.
If you look in Visual Prolog's error-message file (PROLOG.ERR on the DOS-
related platforms, "PDCProlog.err" in UNIX), you'll see all the error numbers
applicable to both compile-time and run-time problems. All numbers above and
including 10000 are reserved for user program exit codes, and you may modify
and distribute the error message file if required. Additionally, in the include
directory you'll find the ERROR.CON include file, containing constant
declarations for all error codes. To guard against future changes, use this file for
error codes rather than hard-coding numbers into your application.

244 Visual Prolog Language Tutorial

Exception Handling and Error Trapping
The cornerstone of error and exception handling is the trap predicate, which can
catch run-time errors as well as exceptions activated by the exit predicate. You
can also use this mechanism to catch signals, such as that generated by Ctrl-
Break in the textmode platforms, as well as a kind of "block exit" mechanism.

exit/0 and exit/1

A call to exit has an effect identical to a run-time error.
exit /* (no arguments) */
exit(ExitCode) /* (i) */

exit without an argument is equivalent to exit(0). If the call to exit is executed in

a direct or indirect subgoal of a trap, the ExitCode will be passed to the trap.
The behavior of an untrapped exit depends on the platform. The VPI event
handlers do their own trapping, and an exit will be caught here resulting in an
error message.
An untrapped call of the exit predicate on the textmode platforms results in
program termination, and the OS return code ('ErrorLevel' in the DOS-related
operating systems, '$?' in UNIX sh) will be set to the value used in the call. The
maximum value a process can exit with is 254; 255 is reserved for Visual Prolog's
system call, but no checks are performed.

trap/3
trap, which takes three arguments, carries out error trapping and exception
handling. The first and the last arguments to trap are predicate calls, and the
second argument is a variable; it takes this format:
trap(PredicateCall, ExitCode, PredicateToCallOnError)

For example, consider the call:

trap(menuact(P1, P2, P3), ExitCode, error(ExitCode, P1)), ...

If an error occurs during execution of menuact--including all further called

subgoals--an error code will be returned in the variable ExitCode, and the error-
handling predicate error will be called. trap will then fail on return from error. If
menuact returns successfully, evaluation will continue after the trap, which will
no longer be effective.
Before calling the error predicate, the system resets the stack, global stack, and
trail to the values they had before the goal specified in the trap (menuact in the
Chapter 177 Advanced topics 245
example above) was called. This means that you can use a trap to catch memory
overflows, but you shouldn't rely on big memory consuming operations such as
database updates to be in either a complete or unaltered state - a heap-full error
may occur anytime.
If Break is enabled on the textmode platforms, and a Break occurs (because the
user pressed Ctrl-Break during execution of a predicate with a surrounding trap),
the trap will catch the Break and return 0 in the ExitCode variable.

Example: catching file-not-open

/* Program ch199e12.pro */

include "error.con"

DOMAINS
file = inpfile

PREDICATES
ioehand(integer,file)
getline(file,string)

CLAUSES
ioehand(err_notopen,File):-!,
write(File," isn't open\n"),
exit(1).
ioehand(Err,File):-
write("Error ",Err," on ",File,'\n'),
exit(1).

getline(File,Line):-
readdevice(Old),
readdevice(File),
readln(Line),
readdevice(Old).

GOAL
trap(getline(inpfile,First),Err,ioehand(Err,inpfile)),
write(First).

errormsg/4
You can use the errormsg predicate to access files that are structured the same as
Visual Prolog's error-message file.
errormsg(File name, ErrorNo, ErrorMsg, ExtraHelpMsg) /* (i,i,o,o) */

246 Visual Prolog Language Tutorial

A typical use of errormsg is in error-trapping predicates to obtain an explanation
of an error code, as illustrated below.
PREDICATES
error(integer)
main
/*....*/

CLAUSES
error(0) :- !. % discard break.
error(E) :-
errormsg("prolog.err", E, ErrorMsg, _),
write("\nSorry; the error\n", E, " : ", ErrorMsg),
write("\nhas occurred in your program."),
write("\nYour database will be saved in the file error.sav"),
save("error.sav").

GOAL
trap(main, ExitCode, error(Exitcode)).

Error reporting
Visual Prolog includes several compiler directives that you can use to control run-
time error reporting in your programs. These directives allow you to select the
following:
whether code should be generated to check for integer overflows.
the level of detail in reporting run-time errors.
whether code should be generated for stack overflow checking.
You can place these compiler directives at the top of your program, or choose
them from the Compiler Options dialog.

errorlevel
Visual Prolog has a mechanism to locate the source position where a run-time
error occurs. To do this, it generates code before predicate calls to save the source
code position where executions are actually performed. The level of error
reporting and the storing of source positions are selected by the errorlevel
compiler directive. The syntax is:
errorlevel = d

where d is one of 0, 1, or 2, representing the following levels:

Chapter 177 Advanced topics 247

 0 This level generates the smallest and most efficient code. No source
positions are saved. When an error occurs, just the error number is
reported.
1 This is the default level. When an error occurs, the Visual Prolog system
displays its origin (module name and include file, if applicable). The place
where the error was detected within the relevant source file is also
displayed, expressed in terms of the number of bytes from the beginning
of the file.
2 At this level, certain errors not reported at level 1, including stack
overflow, heap overflow, trail overflow, etc., are also reported. Before
each predicate call code is generated to store the source position.

When a source position are reported, the source program can be loaded into the
editor, and you can activate the Edit | Go To Line Number menu item, where
you can enter the position number and the cursor will move to the place where the
error occurred.
In a project, the errorlevel directive in each module controls that module's detail
of error reporting. However, if the errorlevel directive in the main module is
higher than that of the other modules, the system might generate misleading error
information.
If, for example, an error occurs in a module compiled with errorlevel = 0, which
is linked with a main module compiled with errorlevel set to 1 or 2, the system
will be unable to show the correct location of the error--instead, it will indicate
the position of some previously executed code.
For more information about projects, refer to "Modular Programming" in the
chapter 200.

lasterror/4
Hand in hand with trap and errormsg goes lasterror. It returns all relevant
information about the most recent error, and looks like this:
lasterror(ErrNo,Module,IncFile,Pos) /* (i,i,i,i) */

where ErrNo is the error number, Module is the source file name, IncFile is the
include file name, and Pos is the position in the source code where the error
occurred. However, the program must be compiled with an errorlevel greater than
1 in order for the information to be relevant in case of memory overflow. For
ordinary errors, an errorlevel of 1 is sufficient.

248 Visual Prolog Language Tutorial

The primary aim of lasterror is to ease debugging when subgoals are trapped, but
it may equally well form the basis of a cause-of-death indicator in commercially
distributed software. Using lasterror, your code can provide the user with a quite
sober error message in addition to exact information about what happened.

Handling Errors from the Term Reader

When you call consult or readterm and a syntax error occurs in the line read, the
predicates will exit with an error. The syntax error could be any one of the
following:
A string is not terminated.
A symbol is placed where an integer is expected.
Upper-case letters are used for the predicate name.
A symbol is not surrounded by double quotes.
Etc.
When consult was originally introduced in Visual Prolog, it was not meant to be
used for reading user-edited files: It was designed to read back files that were
saved by the save predicate. In order to make it easier to consult user-created
files, we have introduced the two predicates readtermerror and consulterror.
You can call these to obtain information about what went wrong in readterm or
consult, respectively.
If the errors from consult and readterm are caught by the trap predicate,
consulterror and readtermerror allow you to inspect and possibly edit the cause
of the syntax error.

consulterror/3
consulterror returns information about the line containing a syntax error.
consulterror(Line, LinePos, Filepos), /* (o,o,o) */

Line is bound to the line that has the syntax error, LinePos is bound to the
position in the line where the syntax error was found, and FilePos is bound to the
position in the file where the line was read.
/* Program ch201e13.pro */

CONSTANTS
helpfile = "prolog.hlp"
errorfile = "prolog.err"

Chapter 177 Advanced topics 249

DOMAINS
dom = f(integer)
list = integer*

DATABASE - mydba
p1(integer, string, char, real, dom, list)

PREDICATES
handleconsulterr(string, integer)

CLAUSES
handleconsulterr(File, Err):-
Err>1400, Err<1410, !,
retractall(_, mydba),
consulterror(Line, LinePos, _),
errormsg(errorfile, Err, Msg, _),
str_len(Blanks,LinePos),
write("Syntax error in ",File,'\n',Line,'\n',Blanks,"^\n",Msg,'\n'),
exit(1).
handleconsulterr(File,Err):-
errormsg(errorfile,Err,Msg,_),
write("Error while trying to consult ",File,":\n",Msg,'\n'),
exit(2).

GOAL
File="faulty.dba",
trap(consult(File, mydba), Err, handleconsulterr(File,Err)),
write("\nSUCCESS\n").

readtermerror/2
readtermerror returns information about the readterm-read line containing a
syntax error.
readtermerror(Line, LinePos), /* (o,o) */

Line is bound to the line that has the syntax error, and LinePos is bound to the
position in the line where the syntax error was found.

Break Control (Textmode Only)

It is important to understand how the break/signal mechanism is implemented in

Visual Prolog. Generally, a break does not immediately abort the current
execution. Rather, Visual Prolog has an exception handler installed, which sets a

250 Visual Prolog Language Tutorial

flag when activated by the signal. Visual Prolog checks this flag in two different
cases:
If the code is compiled with break-checking enabled, the status of the break-
flag is examined each time a predicate is entered. Break-checking may be
disabled by using the nobreak directive in the source code, through the
Options/Compiler Directives/Run-time check menu item, or from the
commandline.
Several of the library routines check the break-flag.

If the break-flag is set, the outcome depends on the break-status, set by the
predicate break: If break-status is Off, the signal is ignored for the time being,
otherwise the code will exit with an appropriate exitcode (discussed below). This
exit will of course be caught by a trap, if any is set.

break/1
break enables and disables the sensing of the break-flag during execution. break
takes one of the following forms:
break(on) /* (i); enables the BREAK key */
break(off) /* (i); disables the BREAK key */
break(BreakStatus) /* (o); returns the current BREAK status */

You can read the current break status by calling break with an output variable.
This means that, during critical operations, you can disable break and then return
to the original break state afterwards. For example:
update :-
break(OldBreak),
break(off),
/* .... do the updating, */
break(OldBreak).

For the DOS-related versions, the exitcode resulting from a break will always be
0, as the only signal recognized is the user interrupt. For the UNIX version,
SIGINT also results in an exit value of 0, for backwards compatibility with the
large base of installed DOS programs written in Visual Prolog. For other signals
which the process has been set up to catch, the exitcode is the signal number plus
the constant err_signaloffset, defined in the include file ERROR.CON.

Chapter 177 Advanced topics 251

breakpressed/0
breakpressed succeeds if the break-flag is set, even when the break-state has been
turned off by break(off) or the program compiled with the nobreak option. If
successful, breakpressed returns the exitcode generated by the most recently
caught signal, and clears the break-flag. For the DOS-related versions of Visual
Prolog, this will always be 0; for UNIX, it will be the same value as would
otherwise be passed through the exit/trap mechanism, as described above. This
too will be 0 if SIGINT is received.

Manual Break and Signal Checking in UNIX

This section, down to page 255, only applies to UNIX and may be skipped by
users of the DOS-related versions.
A Visual Prolog program may be configured to catch any of the many different
signals a UNIX process can receive (see signal(S)). However, as signals may
arrive at any time, quite asynchronously from the running process, it's important
that they don't interrupt the process while in the middle of something critical,
such as memory allocation. The reason for this is that, due to Prolog's modularity,
the only means of communication between different predicates is through
arguments or through databases. Obviously, an asynchronously executed signal-
handler predicate can't communicate to the rest of the program through
arguments, leaving only the database. And since databases rely on memory
allocation, which invariably is in use by the rest of the program, an
asynchronously executed signal-handling predicate could create havoc, if trying
to e.g. assert something to indicate that a signal was received, while the
interrupted process was in the middle of allocating memory. It really all boils
down to Prolog not having global variables, leaving asynchronously executed
predicates with no means of communication with the rest of the program.
Therefore, rather than invoking a signal-handling predicate the instant the signal
is received, signals are routed through the exit/trap mechanism.

signal/2
Signal-handling in Visual Prolog programs is controlled by the signal predicate,
defined in the include file ERROR.PRE:
GLOBAL DOMAINS
sighand = determ (integer) - (i) language C

GLOBAL PREDICATES
sighand signal(integer,integer) - (i,i) language C as "_BRK_Signal"
sighand signal(integer,sighand) - (i,i) language C as "_BRK_Signal"

252 Visual Prolog Language Tutorial

 CONSTANTS
sig_default = 0
sig_ignore = 1
sig_catch = 2

To modify the handling of a specific signal, call signal with the signal exitcode
you want to catch, such as err_sigalrm, defined in ERROR.PRE, specifying in the
second argument what to do:
sig_default to reset the handling of the signal to the default for the process
sig_ignore to ignore the signal completely
sig_catch to have the signal routed through the exit/trap mechanism
anything else is taken to be the address of a function to be invoked when the
signal occurs
The return value of signal is the previous handling of the signal in question,
which will be one of the values outlined above. The only cases where you may
use the fourth alternative (address of function) is when this value was returned by
a previous call to signal, or when the function is one you have written yourself in
C, exercising the usual precautions when writing signal handlers. In particular,
SIGINT is masked out during the execution of the signal handler, so if you intend
to do a longjump from a signal handler you're written in C, SIGINT must be
enabled first (see sigprocmask(S)). The validity of the function address is not
verified at the time signal is called and results may be highly erratic if it's an
invalid address; see signal(S).
Although the name and argument profile of signal matches that of signal(S), it is
implemented using sigaction(S) and SIGINT is ignored during execution of
installed signal handlers.
By default, Visual Prolog catches the following signals:
SIGINT (user interrupt); results in exit of 0 when detected.
SIGFPE (floating point exception); results in an exit of err_realoverflow
immediately after the erroneous calculation.
SIGBUS and SIGSEGV (memory fault); these signals result from attempting
to access memory not belonging to the process, typically due to a faulty
pointer. A short message, indicating where in the program the error happened,
will be printed if possible (see the errorlevel compiler directive), and the
process is terminated, leaving a core dump. Unless you have made a mistake
in modules you have written yourself in C, this invariably indicates an internal
error.

Chapter 177 Advanced topics 253

 SIGILL (illegal instruction); the processor encountered an unrecognized or
illegal instruction. Other details as for SIGBUS and SIGSEGV.
Any signals caught will be routed through the same function as SIGINT. Note
that once catching has been enabled for a signal, it remains in effect until
explicitly reset by another call to signal. Receiving and catching a signal will not
reset the signal handling.
Needless to say, signal catching should be used very carefully, and the break-state
should always be turned off if you intend to receive and test for signals without
interrupting the program. In particular, a number of operating system calls will be
terminated prematurely if a signal is caught while they're executing. When the
break-state is off, the reception of the signal will be noted in the break-flag and
the interrupted system call called again, meaning the program should work as
expected. However, while every care has been taken to ensure the integrity of this
scheme, no guarantees can be given. For instance, some versions of SCO UNIX
and SCO Open Desktop will allow interrupts of certain terminal I/O functions,
without giving any indication that such an interrupt occurred.
Below are two examples, using the alarm clock signal. Both use the breakpressed
predicate, which will be described later.
The first example will print the message "Do something!" every three seconds,
until the user enters a character. It doesn't turn the break-state off during the
central parts of the program, as the whole purpose is to interrupt a system call.
/* Program ch202e14.pro */

/* For UNIX platform only */

include error.con"

GLOBAL PREDICATES
alarm(integer) - (i) language C % See alarm(S)

PREDICATES
brkclear
nondeterm repeat
ehand(integer)
getchar(char)

CLAUSES
brkclear:-breakpressed,!. % Clear break-flag, if set
brkclear.

repeat.
repeat:-repeat.

254 Visual Prolog Language Tutorial

 ehand(2214):-!,
write("Do something!\n").
ehand(E):-
write("\nUnknown exit ",E,'\n'),
exit(2).

getchar(C):-
write("Enter char: "),
alarm(3), % Alarm to go off in 3 seconds
readchar(C),
% This will exit with err_sigalrm when receiving SIGALRM
alarm(0), % Cancel pending alarm signal
break(off),
brkclear, % Clear break-flag, in case alarm went off
break(on). % just before cancellation above.

GOAL
Old=signal(err_sigalrm,sig_catch), % Declared in error.con
repeat,
trap(getchar(C),Err,ehand(Err)),
!,
signal(err_sigalrm,Old),
write("\nYou entered '",C,"'\n").

The next example, which has been deliberately written to be somewhat

inefficient, displays program progress during lengthy processing. Break-status is
turned off in this program, and the detection of any signals is handled manually,
using the breakpressed predicate.
/* Program ch203e15.pro */

/* For UNIX platform only */

include "error.con"

GLOBAL PREDICATES
alarm(integer) - (i) language C% See alarm(S)

DATABASE
rcount(unsigned)
dba(real,real,real)

PREDICATES
nondeterm repeat
process_dba
bcheck
bcheck1(integer)

Chapter 177 Advanced topics 255

CLAUSES
repeat.
repeat:- repeat.

rcount(0).

dba(1,1,1).

process_dba:-
retract(dba(F1,F2,F3)), !, F = F1 * F2 * F3, assert(dba(F,F,F)),
retract(rcount(N)), !, NN = N+1, assert(rcount(NN)),
NN = 25000. % fail back to repeat in goal

bcheck:-
Break = breakpressed(),!,
bcheck1(Break).
bcheck.

bcheck1(err_sigalrm):-!,
rcount(N),!,
time(H,M,S,_),
writef("\r%:%:% % records ",H,M,S,N),
alarm(1). % Next alarm in 1 second
bcheck1(0):-!,
write("\nInterrupt\n"),
exit(1).
bcheck1(Exit):-
write("\nUnknown exit ",Exit,"; runtime error?\n"),
exit(2).

GOAL
break(off),
Old = signal(err_sigalrm,sig_catch), % Declared in error.pre
alarm(1), % First alarm in 1 second
repeat,
bcheck, process_dba,
!,
alarm(0), % Cancel pending alarm
signal(err_sigalrm,Old),
dba(F1,F2,F3), !,
write('\n',F1,' ',F2,' ',F3,'\n').

The writef predicate is covered in chapter 204.

256 Visual Prolog Language Tutorial

Critical Error Handling under DOS Textmode
This section applies only to the DOS textmode platform, and are not relevant for
VPI programs.
The DOS-version of Visual Prolog's library contains some default routines for
handling error situations, but you can actually substitute the default code with
your own clauses. In this section, we describe two routines -- criticalerror and
fileerror. DOS will call criticalerror when a DOS error occurs. The Visual
Prolog system calls fileerror when it gets a file error in the run-time editor. If you
define these predicates as global and supply your own clauses for them, the linker
will take your code instead of the code from the library. The result is that you
gain better control over error situations. Your .EXE program's size might also
decrease (because the code for the default routines draw in window routines).
Global declarations for criticalerror and fileerror are given in the include file
ERROR.PRE shipped with the Visual Prolog system in the include directory.

criticalerror/4
Visual Prolog defines this routine for handling DOS critical errorscritical errors
(DOS interrupt 24H)DOS, interrupt 24H, but not for OS/2. If you want to use
your own version of criticalerror, you should include ERROR.PRE which gives a
global declaration as follows:
GLOBAL PREDICATES
criticalerror(ErrNo, ErrType, DiskNo, Action) - (i, i, i, o) language
c as "_CriticalError_0"

Refer to the chapter 205 for information on how to use global declarations.
The criticalerror predicate must never fail, and it works only from an .EXE file
application. The criticalerror predicate replaces the DOS critical error interrupt
handler and has the same restriction as the original interrupt handler. (Refer to the
DOS Technical Reference for details.) You can only use DOS function calls 01h
to 0Ch and 59h ("Get extended error")--that means console I/O and nothing else.
If your application uses any other DOS function calls, the operating system is left
in an unpredictable state.

Chapter 177 Advanced topics 257

 Table 206.1: Argument Values for the criticalerror Predicate207

Argument Value Meaning

ErrNo = 0 Attempt to write on write-protected disk
= 1 Unknown unit
= 2 Drive not ready
= 3 Unknown command
= 4 CRC error in data
= 5 Bad drive request structure length
= 6 Seek error
= 7 Unknown media type
= 8 Sector not found
= 9 = 12
= 10 Printer out of paper
= 11 Write fault
Read fault
General failure

ErrType = 0 Character device error

= 1 Disk read error
= 2 Disk write error

DiskNo = 0-25 Means device A to Z

Action = 0 Abort current operation

= 1 Retry current operation
= 2 Ignore current operation (this could be very
dangerous and is not recommended)

fileerror/2
Visual Prolog will activate the predicate fileerror when a file in the textmode
editor action fails.
If you define your own fileerror predicate, it is not allowed to fail, and it works
only from an .EXE file application.
The declaration for fileerror in the ERROR.PRE file is:

258 Visual Prolog Language Tutorial

 GLOBAL PREDICATES
fileerror(integer, string) - (i, i) language c as "_MNU_FileError"

Note that this declaration is correct -- you must specify language c even though
the source code will be in Prolog.

Dynamic Cutting

The traditional cut in Prolog is static. One problem with this is that the effect of
the cut happens when execution passes the ! symbol, and it affects only those
clauses in which it was placed (in the source text). There is no way of passing the
effect of a cut in an argument to another predicate, where the cut might only be
evaluated if some conditions were fulfilled. Another problem with the traditional
cut is that it is impossible to cut away further solutions to a subgoal in a clause,
without also cutting away the backtrack point to the following clauses in the
predicate.
Visual Prolog has a dynamic cutting mechanism, which is implemented by the
two standard predicates getbacktrack and cutbacktrack. This mechanism allows
you to handle both of these problems. The predicate getbacktrack returns the
current pointer to the top of the stack of backtrack points. You can remove all
backtrack points above this place, at some later time, by giving the pointer thus
found to the cutbacktrack predicate.

Examples
Here are some examples that illustrate the use of these two predicates.
Suppose you have a database of people and their incomes, and you have
registered who their friends are.
DATABASE
person(symbol, income)
friends(symbol, symbol)

If you define a happy person as one who either has some friends or pays little
tax, the clauses that return happy persons could be as follows:
happy_person(has_friends(P)) :- person(P, _), friends(P, _).
happy_person(is_rich(P)) :- person(P, Income), not(rich(Income)).

If a person has more than one friend, the first clause will return a multiple
number of solutions for the same person. You could, of course, add another
predicate have_friends(P,P) that has a cut, or you could use the dynamic cut
instead.
Chapter 177 Advanced topics 259
 happy_person(has_friends(P)) :-
person(P, _),
getbacktrack(BTOP),
friends(P, _),
cutbacktrack(BTOP).

Although the friends predicate might return many solutions if backtracked

into, that possibility is cut away with the call to cutbacktrack. A subsequent
failure would backtrack into the person predicate.
The more important use of a dynamic cut is when you pass the backtrack pointer
to another predicate and execute the cut conditionally. The pointer is of
unsigned type and can be passed in arguments of unsigned type.
As an illustration of this, let's say you want a predicate to return numbers until
the user presses a key.
PREDICATES
number(integer)
return_numbers(integer)
checkuser(unsigned)

CLAUSES
number(0).
number(N) :- number(N1), N = N1+1.

return_numbers(N) :- getbacktrack(BTOP), number(N),

checkuser(BTOP).
checkuser(BTOP) :- keypressed, cutbacktrack(BTOP).
checkuser(_).

The compiler does not recognize the cutbacktrack predicate in the pass that
analyzes the clauses for determinism. This means you could get the warning
Non-deterministic clause when using the check_determ directive, even if
you called cutbacktrack.
You should use dynamic cutting very judiciously. It's all too easy to destroy
program structure with dynamic cutting, and careless use will invariably lead
to problems that are very hard to track down.

Free Type Conversions

In most cases there is little need to start mixing wildly differing types. However,
from time to time, in particular when dealing with system level programming or
when interfacing to foreign languages, rather reckless conversions have to be
dealt with. To this end the cast function will convert from anything to anything.
260 Visual Prolog Language Tutorial
No checks are performed on the supplied values, and quite disastrous results will
occur if you try to use incorrectly cast variables.
The format of cast is
Result = cast(returndomain,Expr)

where Expr is evaluated (if it's a numerical expression), converted to

returndomain type, and unified with Result.
For instance, a null string pointer (a character pointer with a value of 0; not an
empty string, which is a pointer to a byte with a value of 0) can be created using:
NullPtr = cast(string,0)

Don't try to write the resulting string, you'd most probably get a protection
violation, a hung system, or at best garbage characters.
If you don't see any obvious use for cast, don't worry. It plays no part in ordinary
Prolog programs.

Programming Style

In this section, we provide some comprehensive guidelines for writing good

Visual Prolog programs. After summarizing a few rules of thumb about
programming style, we give you some tips about when and how to use the fail
predicate and the cut.

Rules for Efficient Programming

Rule 1. Use more variables rather than more predicates.

This rule is often in direct conflict with program readability. To achieve programs
that are efficient (both in their demands upon relatively cheap machines and upon
relatively expensive human resources) requires a careful matching of objectives.
Often, the purely declarative style of Prolog leads to code that is significantly less
efficient than other (non-declarative) approaches. For instance, if you're writing a
predicate to reverse the elements of a list, this code fragment:
reverse(X, Y) :- reverse1([], X, Y). /* More efficient */
reverse1(Y, [], Y).
reverse1(X1, [U|X2], Y) :- reverse1([U|X1], X2, Y).

Chapter 177 Advanced topics 261

makes less demands upon the stack than the next one, which uses the extra
predicate append:
reverse([], []). /* Less efficient */
reverse([U|X], Y) :- reverse(X, Y1), append(Y1, [U], Y).

append([], Y, Y).
append([U|X], Y, [U|Z]) :- append(X, Y, Z).

Rule 2. Try to ensure that execution fails efficiently when no solutions

exist.

Suppose you want to write a predicate singlepeak that checks the integers in a list
to see if, in the order given, they ascend to a single maximum and then descend
again. With this predicate, the call:
singlepeak([1, 2, 5, 7, 11, 8, 6, 4]).

would succeed, while the call:

singlepeak([1, 2, 3, 9, 6, 8, 5, 4, 3]).

would fail.
The following definition for singlepeak breaks Rule 3, since the failure of a list to
have a single peak is only recognized when append has split the list into every
possible decomposition:
/* Definition 1 - Breaks Rule 2 */

singlepeak(X) :- append(X1, X2, X), up(X1), down(X2).

up[_].
up([U, V|Y]) :- U<V, up([V|Y]).

down([]).
down([U]).
down([U, V|Y]) :- U>V, down([V|Y]).

append([], Y, Y).
append([U|X], Y, [U|Z]) :- append(X, Y, Z).

On the other hand, the next definition recognizes failure at the earliest possible
moment:
/* Definition 2 - Follows Rule 2 */

262 Visual Prolog Language Tutorial

 singlepeak([]).
singlepeak([U, V|Y]) :- U<V, singlepeak([V|Y]).
singlepeak([U, V|Y]) :- U>V, down([V|Y]).

down([]).
down([U]).
down([U, V|Y]) :- U>V, down([V|Y]).

The third and final definition shortens singlepeak even further by observing Rule
1.
/* Definition 3 - Follows Rule 1 */

singlepeak([], _).
singlepeak([U, V|W], up) :- U<V, singlepeak([V|W], up).
singlepeak([U, V|W], _) :- U>V, singlepeak([V|W], down).

Using Definition 3, this call to singlepeak

singlepeak(Y, up)

succeeds if Y is bound to a single peaked list appended to an ascending list. This

call
singlepeak(Y, down)

succeeds if Y is bound to a descending list.

Rule 3. Let Visual Prolog's unification mechanism do as much of the

work as possible.

At first thought, you might define a predicate equal to test two lists from the same
domain for equality as follows:
equal([], []).
equal([U|X], [U|Y]) :- equal(X, Y).

This is unnecessary. Using the definition

equal(X, X).

or, even simpler, unification by means of =, Visual Prolog's unification

mechanism does all the work!

Chapter 177 Advanced topics 263

 Rule 4. Use backtracking--instead of recursion--for repetition.

Backtracking decreases stack requirements. The idea is to use the repeat ... fail
combination repeat--fail combination instead of recursion. This is so important
that the next section is dedicated to the technique.

Using the fail Predicate

To evaluate a particular sequence of subgoals repeatedly, it is often necessary to
define a predicate like run with a clause of the form subgoals, evaluating
repeatedly
run :-
readln(X),
process(X, Y),
write(Y),
run.

This kind of definition incurs unnecessary recursion overheads that can't be

automatically eliminated by the system if process(X,Y) is non-deterministic.
In this case, the repeat ... fail combination avoids the need for the final recursive
call. Given
repeat.
repeat :- repeat.

you can redefine run without recursion as follows:

run :-
repeat,
readln(X),
process(X, Y),
write(Y),
fail.

fail causes Visual Prolog to backtrack to process and eventually to repeat, which
always succeeds. But how do you break out of a repeat ... fail combination? Well,
in the cases where you want infinite execution (the run:- ..., ..., run variety, you
will usually only want to quit if some exceptional condition arises. To this end,
you can use the exit predicate in non-interactive programs, or just press break in
interactive ones. In other cases, where you have a clear condition of completion,
replace the fail with a test for completion:

264 Visual Prolog Language Tutorial

 run:-
repeat,
getstuff(X),
process(X,Y),
putstuff(Y),
test_for_completion(Y),
!.

Determinism vs. Non-determinism: Setting the Cut

The compiler directive check_determ is useful when you need to decide where to
place the cut, since it marks those clauses that give rise to non-deterministic
predicates. If you want to make these predicates deterministic, you must insert the
cut to stop the backtracking (which causes the non-determinism).
As a general rule, in such cases, the cut should always be inserted as far to the left
as possible (close to the head of a rule) without destroying the underlying logic of
the program.
Keep in mind these two rules used by the compiler to decide that a clause is non-
deterministic:
There is no cut in the clause, and there is another clause that can match the input
arguments in the clause head.
There is a call to another non-deterministic predicate in the clause body, and this
non-deterministic call is not followed by a cut.

Chapter 177 Advanced topics 265

PA RT 3
Tutorial Chapters 208 -- 209: Using Visual
Prolog.210211Visual Prolog

266 Visual Prolog Language Tutorial

CHAPTER 212

Writing, Reading, and Files

213In this chapter, we first cover the basic set of built-in predicates for writing
and reading. Next we describe how the file system works in Visual Prolog and
show how you can redirect both input and output to files. We also discuss the file
domain and some predefined files.

Writing and Reading

In these tutorials, most of the Input/Output has been interactive via screen and
keyboard. In this section, we provide formal covererage of the standard predicates
you use for I/O, including predicates for file operations.

Writing
Visual Prolog includes three standard predicates for writing. These predicates are
write, nl and writef.

write/* and nl
The predicate write can be called with an arbitrary number of arguments:
write(Param1, Param2, Param3, ..., ParamN)
/* (i, i, i, ..., i) */

These arguments can either be constants from standard domains or they can be
variables. If they're variables, they must be input parameters.
The standard predicate nl (for new line) is often used in conjunction with write; it
generates a newline on the display screen. For example, the following subgoals:
pupil(PUPIL, CL),
write(PUPIL," is in the ",CL," class"),
nl,
write("-----------------------------------").

could result in this display:

Chapter 214, Writing, Reading, and Files 267

 Helen Smith is in the fourth class
----------------------------------

while this goal:

....,
write("List1= ", L1, ", List2= ", L2 ).

could give:
List1= [cow,pig,rooster], List2= [1,2,3]

Also, if My_sentence is bound to

sentence(subject(john),sentence_verb(sleeps))

in the following program

DOMAINS
sentence = sentence(subject, sentence_verb)
subject = subject(symbol) ; ......
sentence_verb = sentence_verb(verb) ; ......
verb = symbol

CLAUSES
....
write( " SENTENCE= ", My_sentence ).

you would obtain this display:

SENTENCE= sentence(subject(john),sentence_verb(sleeps))

Note that with respect to strings, the backslash (\) is an escape character. To print
the backslash character verbatim, you must type two backslashes. For example, to
designate the DOS directory path name A:\PROLOG\MYPROJS\MYFILE.PRO
in a Visual Prolog program, you would type a:\\prolog\\myprojs\\myfile.pro.
If a backslash is followed by one of a few specially recognized characters, it will
be converted to a print control character. These are
'n' newline and carriage return
't' tab
'r' carriage return

Alternatively, the backslash may be followed by up to three decimal digits,

specifying a particular ASCII code. However, avoid putting \0 into a string unless
you know what you're doing. Visual Prolog uses the C convention with 0-
terminated strings.
Be very careful with the '\r' option. It sets the current write position back to the
start of the current line, but if you accidentally do that in between writing
different things, it may happen so quickly that the first thing you write becomes
overwritten before you even notice it's there. Also, if you write something which
is too long for a single line, causing the output to wrap, the '\r' will set the cursor
back to the beginning of the last line, not the beginning of the line where the
writing started.
Often write does not, by itself, give you as much control as you'd like over the
printing of compound objects such as lists, but it's easy to write programs that
give better control. The following four small examples illustrate the possibilities.

Examples Demonstrating the write Predicate

These examples show how you can use write to customize your own predicates
for writing such things as lists and compound data structures.
1. Program ch11e01.pro prints out lists without the opening bracket ([) and
closing bracket (]).
/* Program ch215e01.pro */

/* Copyright (c) 1986, '95 by Prolog Development Center */

DOMAINS
integerlist = integer*
namelist = symbol*

PREDICATES
writelist(integerlist)
writelist(namelist).

CLAUSES
writelist([]).
writelist([H|T]):-
write(H, " "),
writelist(T).

Notice how this program uses recursion to process a list. Load the program
and try this goal:
writelist([1, 2, 3, 4]).

2. The next example, Program ch11e02.pro, writes out the elements in a list with
no more than five elements per line.

Chapter 214, Writing, Reading, and Files 269

 /* Program ch216e02.pro */

/* Copyright (c) 1986, '95 by Prolog Development Center */

DOMAINS
integerlist = integer*

PREDICATES
writelist(integerlist)
write5(integerlist,integer)

CLAUSES
writelist(NL):-
nl,
write5(NL,0),nl.
write5(TL,5):-!,
nl,
write5(TL, 0).
write5([H|T],N):-!,
write(H," "),
N1=N+1,
write5(T,N1).
write5([],_).

If you give the program this goal:

writelist([2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22]).

Visual Prolog responds with:

2 4 6 8 10
12 14 16 18 20
22

3. Frequently, you may want a predicate that displays compound data structures
in a more readable form. Program ch11e03.pro displays a compound object
like:
plus(mult(x, number(99)), mult(number(3), x))

in the form:
x*99+3*x

(This is known as infix notation.)

/* Program ch217e03.pro */

/* Copyright (c) 1986, '95 by Prolog Development Center */

DOMAINS
expr = number(integer); x; log(expr);
plus(expr, expr); mult(expr, expr)

PREDICATES
writeExp(expr)

CLAUSES
writeExp(x):-write('x').
writeExp(number(No)):-write(No).
writeExp(log(Expr)):-
write("log("), writeExp(Expr), write(')').
writeExp(plus(U1, U2)):-
writeExp(U1),write('+'),writeExp(U2).
writeExp(mult(U1,U2)):-
writeExp(U1),write('*'),writeExp(U2).

4. Program ch11e04.pro is similar to Program ch11e03.pro.

/* Program ch218e04.pro */

/* Copyright (c) 1986, '95 by Prolog Development Center */

DOMAINS
sentence = sentence(nounphrase, verbphrase)
nounphrase = nounp(article, noun); name(name)
verbphrase = verb(verb); verbphrase(verb, nounphrase)
article, noun, name, verb = symbol

PREDICATES
write_sentence(sentence)
write_nounphrase(nounphrase)
write_verbphrase(verbphrase)

CLAUSES
write_sentence(sentence(S,V)):-
write_nounphrase(S),write_verbphrase(V).
write_nounphrase(nounp(A,N)):-
write(A,' ',N,' ').
write_nounphrase(name(N)):-write(N,' ').
write_verbphrase(verb(V)):-write(V,' ').
write_verbphrase(verbphrase(V,N)):-
write(V,' '),write_nounphrase(N).

Try Program ch11e04.pro with this goal:

write_sentence(sentence(name(bill), verb(jumps))).

Chapter 214, Writing, Reading, and Files 271

Exercise
Write a Visual Prolog program that, when given a list of addresses contained in
the program as clauses of the form:
address("Sylvia Dickson", "14 Railway Boulevard","Any Town", 27240).

displays the addresses in a form suitable for mailing labels, such as:
Sylvia Dickson
14 Railway Boulevard
Any Town
27240

writef/*
The writef predicate allows you to produce formatted output; it uses this format:
writef(FormatString, Arg1, Arg2, Arg3, ...,ArgN)
/* (i, i, i, i, ..., i) */

Arg1 to ArgN must be constants or bound variables belonging to standard

domains; it is not possible to format compound domains. The format string
contains ordinary characters and format specifiers; ordinary characters are
printed without modification, and format specifiers take the following form:
%-m.pf

The characters in the format specifiers following the % sign are optional and have
these meanings:

hyphen (-) Indicates that the field is to be left-justified (right-justified

is the default).
m field Decimal number specifying a minimum field length.
.p field Decimal number specifying either the precision of a
floating-point number or the maximum number of
characters to be printed from a string.
f field Specifies formats other than the default format for the
given object. For example, the f field can specify that
integers should be printed as unsigned or hexadecimal
numbers.

Visual Prolog recognizes the following format specifiers in the f field:

 f reals in fixed-decimal notation (such as 123.4 or 0.004321)
e reals in exponential notation (such as 1.234e2 or 4.321e-3)
g reals in the shorter format of f or e (this is the default for reals)
d,D integral domains as a signed decimal number
u,U integral domains as an unsigned decimal integer
o,O integral domains as an octal number
x,X integral domains as a hexadecimal number
c integral domains as a char
s as a string (symbols and strings)
R as a database reference number (ref domain only)
B as a binary (binary domain only)
P as a function pointer

The ref domain will be described in chapter 219, and the binary and function
pointer domains in chapter 220.
For the integral domain specifiers, an uppercase format letter denotes that the
associated object is a long type. If no format letter is given, Visual Prolog will
automatically select a suitable format.

Examples of Formatted Output

1. The following example program illustrates the effect of different format
specifiers on output formatted with writef.
/* Program ch221e05.pro */

/* Copyright (c) 1986, '95 by Prolog Development Center */

Chapter 214, Writing, Reading, and Files 273

% Note that the format strings in this example specify 16bit integers

GOAL
A = one,
B = 330.12,
C = 4.3333375,
D = "one two three",
writef("A = '%-7' \nB = '%8.1e'\n",A,B),
writef("A = '%' \nB = '%8.4e'\n",A,B),nl,
writef("C = '%-7.7g' \nD = '%7.7'\n",C,D),
writef("C = '%-7.0f' \nD = '%0'\n",C,D),
writef("char: %c, decimal: %d, octal: %o, hex: %x",'a','a','a','a').

When run, this program will produce the following output:

A = 'one '
B = ' 3.3E+02'
A = 'one'
B = '3.3012E+02'
C = '4.3333375'
D = 'one two'
C = '4 '
D = 'one two three'
char: a, decimal: 97, octal: 141, hex: 61

2. Here's another example, showing how you can use writef to format your
output.
/* Program ch222e06.pro */

/* Copyright (c) 1986, '95 by Prolog Development Center */

DATABASE
person(string,integer,real)

CLAUSES
person("Pete Ashton",20,11111.111).
person("Marc Spiers",32,33333.333).
person("Kim Clark",28,66666.666).

GOAL
% Name is left-justified, at least 15 characters wide
% Age is right-justified, 2 characters wide
% Income is right-justified, 9 characters wide, with 2
% decimal places, printed in fixed-decimal notation
 person(N, A, I),
writef("Name= %-15, Age= %2, Income= $%9.2f \n",N,A,I),
fail
;
true.

This produces the following:

Name= Pete Ashton , Age= 20, Income= $ 11111.11
Name= Marc Spiers , Age= 32, Income= $ 33333.33
Name= Kim Clark , Age= 28, Income= $ 66666.67

Reading
Visual Prolog includes several standard predicates for reading. The four basic
ones are readln for reading whole lines of characters, readchar for reading single
characters/keystrokes, readint for reading integers, and readreal for reading
floating point numbers. Additionally, readterm will read any term, including
compound objects. These predicates can all be redirected to read from files.
Another, more specialized, predicate that belong in the reading category is
file_str for reading a whole text file into a string.

readln/1
readln reads a line of text; it uses this format:
readln(Line) /* (o) */

The domain for the variable Line will be a string. Before you call readln, the
variable Line must be free. readln reads up to 254 characters (plus a carriage
return) from the keyboard, up to 64K from other devices. If Esc is pressed during
input from the keyboard, readln will fail.

readint/1, readreal/1, and readchar/1

readint reads an integer value, using this format:
readint(X) /* (o) */

The domain for the variable X must be of integer type, and X must be free prior
to the call. readint will read an integer value from the current input device
(probably the keyboard) until the Enter key is pressed. If the line read does not
correspond to the usual syntax for integers, readint fails and Visual Prolog

Chapter 214, Writing, Reading, and Files 275

invokes its backtracking mechanism. If Esc is pressed during input from the
keyboard, readint will fail.
readreal does as its name conveys: it reads a real number (as opposed to readint,
which reads an integer). readreal uses the following format:
readreal(X) /* (o) */

The domain for the variable X must be real, and X must be free prior to the call.
readreal will read a real value from the current input device until the Enter key is
pressed. If the input does not correspond to the usual syntax for a real number,
readreal fails. If Esc is pressed during input from the keyboard, readreal will fail.
readchar reads a single character from the current input device, using this format:
readchar(CharParam) /* (o) */

CharParam must be a free variable before you call readchar and must belong to
a domain of char type. If the current input stream is the keyboard, readchar will
wait for a single character to be typed before it returns. If Esc is pressed during
input from the keyboard, readchar will fail.

readterm/2
readterm reads a compound term and converts it to an object; it takes this format:
readterm(DomainName, Term) /* (i, i) */

You call readterm with two arguments: a domain name and a term. readterm
reads a line and converts it to an object of the given domain. If the line does not
look like write would have formatted the object, readterm gives an error. The
standard predicate readtermerror may be used in connection with a trap to
produce customized error handling for readterm. See chapter 223.
readterm is useful for handling terms in text files. For example, you can
implement you own version of consult.

file_str/2
file_str reads characters from a file and transfers them to a variable, or creates a
file and writes the string into the file. It uses this format:
file_str(Filename, Text) /* (i, o), (i, i) */

If, before your program calls file_str, the variable Text is free, file_str reads the
entire contents of the file Filename into Text. In the DOS-related versions of
Visual Prolog, an eof character (Ctrl+Z) will terminate reading when
encountered and will not be included in the string.
For example, the call
file_str("t.dat", My_text)

binds My_text to the contents of the file t.dat. The file size can't exceed the
maximum length of a string, which is 64 Kbytes on the 16-bit platforms. If the
file exceeds the maximum size, file_str will return an error message.
With My_text bound to the text in "t.dat", file_str("t.bak", My_text) will
create a file called t.bak that contains the text from "t.dat". If t.bak already exists
it will be overwritten.

Examples
These examples demonstrate how you can use the standard reading predicates to
handle compound data structures and lists as input.
1. Program ch11e07.pro illustrates assembling of compound objects from
individually read pieces of information.
/* Program ch224e07.pro */

/* Copyright (c) 1986, '95 by Prolog Development Center */

DOMAINS
person = p(name, age, telno, job)
age = integer
telno, name, job = string

PREDICATES
readperson(person)
run

CLAUSES
readperson(p(Name,Age,Telno,Job)):-
write("Which name ? "), readln(Name),
write("Job ? "), readln(Job),
write("Age ? "), readint(Age),
write("Telephone no ? "), readln(Telno).

run :-
readperson(P),nl,write(P),nl,nl,
write("Is this compound object OK (y/n)"),
readchar(Ch),Ch='y', !.

Chapter 214, Writing, Reading, and Files 277

 run :-
nl,write("Alright, try again"),nl,nl,run.

GOAL
run.

2. This next example reads one integer per line until you type a non-integer (such
as the X key); then readint will fail and Visual Prolog displays the list.
/* Program ch225e08.pro */

/* Copyright (c) 1986, '95 by Prolog Development Center */

DOMAINS
list = integer*

PREDICATES
readlist(list)

CLAUSES
readlist([H|T]):-
write("> "),
readint(H),!,
readlist(T).
readlist([]).

GOAL
write("*************** Integer List *****************"),nl,
write(" Type in a column of integers, like this:"),nl,
write(" integer (press ENTER)\n integer (press ENTER)\n"),
write(" etc.\n\n Type X (and press ENTER) to end the list.\n\n"),
readlist(TheList),nl,
write("The list is: ",TheList).

Load Program ch12e08.pro and run it. At the prompt, enter a column of integers
(such as 1234 Enter 567 Enter 89 Enter 0 Enter), then press X Enter to end the
list.

Exercise
Write and test clauses for a predicate readbnumb, which, in the call:
readbnumb(IntVar)

converts a user-input, 16-bit binary number like "1111 0110 0011 0001" to a
corresponding integer value to which IntVar is bound. Check your work by
writing a program that contains readbnumb.
Binary Block Transfer
Three standard predicates allow reading and writing of binary blocks, or byte
sequences of a given length. They all use the binary standard domain. This
emulates an array, with a word (dword on the 32-bit versions of Visual Prolog) in
front, holding the size. For a complete description of the binary domain, see
chapter 226.

readblock/2
readblock has the following format:
readblock(Length,BTerm) /* (i, o) */

where Length is the number of bytes to read and BTerm is the returned binary
term. As described in chapter 11, there are no restrictions on the contents of a
binary term, and it will contain whatever was read from the file including zeros
and DOS eof-markers (Ctrl+Z).
The current input device must be assigned to a file (see readdevice).
If Length = 0 is specified, the readblock attempts to read maximum possible
number of bytes from an input device. (Remember that BinBlock < 64 K on 16-bit
platforms).
If Length is larger than the actual remaining number of bytes in the file - then the
readblock generates the run-time error 1111: "Wrong number of bytes read from
file".

writeblock/2
writeblock complements readblock:
writeblock(Length,BTerm) /* (i, i) */

As with readblock, there are no restrictions on the contents of BTerm. The Length
specifies how many bytes to write; a length of 0 will write the whole term.
For compatibility with previous versions of Visual Prolog, where binary blocks
were disguised as strings, writeblock may be called with a string argument instead
of a binary term. In this case, it is imperative that Length contains the correct
number of bytes to write.

file_bin/2
file_bin will read a whole file into a binary term, and vice versa. It takes two
arguments, the filename and the binary term:
Chapter 214, Writing, Reading, and Files 279
 file_bin(FileName,BinTerm) /* (i, o) (i, i) */

Visual Prolog's File System

In this section, we give you a look at Visual Prolog's file system and the standard
predicates relevant to files. We also introduce I/O redirection, an efficient method
for routing input and output to various devices. With a few exceptions, the file
system works identically in the different versions of Visual Prolog.
Visual Prolog uses a current_read_device, from which it reads input, and a
current_write_device, to which it sends output. Normally, the keyboard is the
current read device, and the screen is the current write device. However, you can
specify other devices. For instance, input could be read from an externally stored
file (on disk perhaps). Not only can you specify other devices, you can even
reassign the current input and output devices while a program is running.
Regardless of what read and write devices you use, reading and writing are
handled identically within a Visual Prolog program. To access a file, you must
first open it. A file can be opened in one of four ways:
for reading
for writing
for appending
for modification
A file opened for any activity other than reading must be closed when that
activity is finished, or the changes to the file might be lost. You can open several
files simultaneously, and input and output can be quickly redirected between open
files. In contrast, it takes much longer to open and close a file than to redirect data
between files.
When Visual Prolog opens a file, it connects a symbolic name to the actual file
name. Visual Prolog uses this symbolic name when directing input and output.
Symbolic file names must start with a lower-case letter and must be declared in
the file domain declaration like this:
file = file1; source; auxiliary; somethingelse

Only one file domain is allowed in any program. Visual Prolog recognizes five
predefined file alternatives:

keyboard reading from the keyboard (default)

 screen writing to the screen
stdin reading from standard input
stdout writing to standard output
stderr writing to standard error

These predefined alternatives must not appear in the file domain declaration; they
don't need to be opened and they should not be closed. Note, that when using the
VPI strategy, only the screen alternative can be used.

Opening and Closing Files

The following sections describe the standard predicates for opening and closing
files.
Note: In the DOS-related versions of Visual Prolog, remember that the backslash
character, used to separate subdirectories, is an escape character. You must
always use two backslash characters when you give a path in the program; for
example, the string
"c:\\prolog\\include\\iodecl.con"

represents the path name

c:\prolog\include\iodecl.con

openread/2
openread opens the file OSFileName for reading, using this format:
openread(SymbolicFileName, OSFileName) /* (i, i) */

Visual Prolog refers to the opened file by the symbolic name SymbolicFileName
declared in the file domain. If the file can't be opened, Visual Prolog returns an
error message.

openwrite/2
openwrite opens the file OSFileName for writing; it takes this format:
openwrite(SymbolicFileName, OSFileName) /* (i,i) */

If the file already exists, it is erased. Otherwise, Visual Prolog creates a new file
and makes an entry in the appropriate directory. If the file can't be created, the
predicate exits with an error message.
Chapter 214, Writing, Reading, and Files 281
openappend/2
openappend opens the file OSFileName for writing at the end, using this format:
openappend(SymbolicFileName, OSFileName) /* (i, i) */

If the file can't be opened for write access, Visual Prolog issues an error message.

openmodify/2
openmodify opens the file OSFileName for both reading and writing; if the file
already exists, it won't be overwritten. openmodify takes this format:
openmodify(SymbolicFileName, OSFileName) /* (i, i) */

If the system can't open OSFileName, it issues an error message. openmodify can
be used in conjunction with the filepos standard predicate to update a random-
access file.

filemode/2
When a file has been opened, filemode sets the specified file to text mode or
binary mode, using this format:
filemode(SymbolicFileName, FileMode) /* (i, i) */

If FileMode = 0, the file specified by SymbolicFileName is set to binary mode; if

FileMode = 1, it's set to text mode.
In text mode, newlines are expanded to carriage- return/line-feed pairs during
writes, and carriage-return/line-feed pairs are converted to newlines during reads.
Carriage return = ASCII 13
Line feed = ASCII 10
In binary mode, no expansions or conversions occur. To read a binary file, you
can only use readchar or the binary file-access predicates discussed in chapter
227.
filemode is only relevant in the DOS-related versions of Visual Prolog. In the
UNIX version it has no effect.
Don't confuse filemode with the peculiarly named DosQFileMode and
DosSetFileMode OS/2 primitives - they get or change the attributes of physical
disk files.

closefile/1
closefile closes the indicated file; it takes this format:
 closefile(SymbolicFileName) /* (i) */

This predicate always succeeds, even if the file has not been opened.

readdevice/1
readdevice either reassigns the current_read_device or gets its name; the
predicate takes this format:
readdevice(SymbolicFileName) /* (i), (o) */

readdevice reassigns the current read device if SymbolicFileName is bound and

has been opened for reading. If SymbolicFileName is free, readdevice binds it to
the name of the current active read device.

writedevice/1
writedevice either reassigns or gets the name of the current_write_device; it takes
this format:
writedevice(SymbolicFileName) /* (i), (o) */

writedevice reassigns the current write device if the indicated file has been
opened for either writing or appending. If SymbolicFileName is free, writedevice
binds it to the name of the current active write device.

Examples
1. The following sequence opens the file mydata.fil for writing, then directs all
output produced by clauses between the two occurrences of writedevice to that
file. The file is associated with the symbolic file name destination appearing in
the declaration of the file domain.
DOMAINS
file = destination

GOAL
openwrite(destination, "mydata.fil"),
writedevice(OldOut), /* gets current output device */
writedevice(destination), /* redirects output to the file */
:
:
writedevice(OldOut), /* resets output device */

2. Program ch11e09.pro uses some standard read and write predicates to

construct a program that stores characters typed at the keyboard in the file
Chapter 214, Writing, Reading, and Files 283
 tryfile.one. Characters typed are not echoed to the display; it would be a good
exercise for you to change the program so that characters are echoed. The file
is closed when you press the # key.
/* Program ch228e09.pro */

/* Copyright (c) 1986, '95 by Prolog Development Center */

DOMAINS
file = myfile

PREDICATES
readloop

CLAUSES
readloop:-
readchar(X),
X<>'#',!,
write(X),
readloop.
readloop.

GOAL
write("This program reads your input and writes it to"),nl,
write("tryfile.one. For stop press #"),nl,
openwrite(myfile,"tryfile.one"),
writedevice(myfile),
readloop,
closefile(myfile),
writedevice(screen),
write("Your input has been transferred to the file tryfile.one"),nl.

Redirecting Standard I/O

The file domain has three additional options: stdin, stdout, and stderr. The
advantage of these file streams is that you can redirect I/O at the command line.

stdin Standard input is a read-only file; the keyboard, by default.

readdevice(stdin) directs the input device to stdin.

stdout Standard output is a write-only file that defaults to the screen.

writedevice(stdout) directs the output device to stdout.

stderr Standard error is a write-only file that defaults to the screen.

writedevice(stderr) directs the output device to stderr.
Working with Files
In this section, we describe several other predicates used for working with files;
these are filepos, eof, flush, existfile, searchfile, deletefile, renamefile, disk, and
copyfile.

filepos/3
filepos can control the position where reading or writing takes place; it takes the
form
filepos(SymbolicFileName, FilePosition, Mode) % (i, i, i), (i, o, i)

With FilePosition bound, this predicate can change the read and write position for
the file identified by SymbolicFileName. It can return the current file position if
called with FilePosition free. FilePosition is a long value.
Mode is an integer and specifies how the value of FilePosition is to be
interpreted, as shown in Table 229.1.

Table 230.2: Mode and FilePosition231

Mode FilePosition
0 Relative to the beginning of the file.
1 Relative to current position.
2 Relative to the end of the file. (The end of the file is position
0.)

When returning FilePosition, filepos will return the position relative to the
beginning of the file irrespective of the value of Mode. Note: In the DOS-related
versions of Visual Prolog, filepos does not consider files in text mode to be
different from files in binary mode. No translation of DOS newline conventions
takes place, and a newline in a file following DOS newline conventions consists
of two characters.

Example
1. The following sequence writes the value of Text into the file somefile.pro
(referred to by Prolog as myfile), starting at position 100 (relative to the
beginning of the file).
Chapter 214, Writing, Reading, and Files 285
 Text = "A text to be written in the file",
openmodify(myfile, "somefile.pro"),
writedevice(myfile),
filepos(myfile, 100, 0),
write(Text),
closefile(myfile).

2. Using filepos, you can inspect the contents of a file on a byte-by-byte basis, as
outlined in Program ch11e10.pro. This program requests a file name, then
displays the contents of positions in the file as their position numbers are
entered at the keyboard.
/* Program ch232e10.pro */

/* Copyright (c) 1986, '95 by Prolog Development Center */

DOMAINS
file = input

PREDICATES
inspect_positions(file)

CLAUSES
inspect_positions(UserInput):-
readdevice(UserInput),
nl,write("Position No? "),
readln(X),
term_str(ulong,Posn,X),
readdevice(input),
filepos(input,Posn,0),
readchar(Y),nl,
write("Char is: ",Y),
inspect_positions(UserInput).

GOAL
write("Which file do you want to work with ?"),nl,
readln(FileName),
openread(input, FileName),
readdevice(UserInput),
inspect_positions(UserInput).

eof/1
eof checks whether the file position is at the end of the file, in which case eof
succeeds; otherwise, it fails. eof has the form
eof(SymbolicFileName) /* (i) */
eof gives a run-time error if the file has been opened with write-only access. Note
that it doesn't consider a DOS eof character (Ctrl+Z) to have any particular
meaning.

Example
eof can be used to define a predicate repfile that's handy when operating with
files. repfile generates backtrack points as long as the end of the file has not been
reached.
PREDICATES
repfile(FILE)

CLAUSES
repfile(_).
repfile(F):- not(eof(F)), repfile(F).

The following program converts one file to another where all the characters are
upper-case.
/* Program ch233e11.pro */

/* Copyright (c) 1986, '95 by Prolog Development Center */

DOMAINS
file = input; output

PREDICATES
convert_file
nondeterm repfile(FILE)

CLAUSES
convert_file :-
repfile(input),
readln(Ln),
upper_lower(LnInUpper,Ln), /* converts the string to uppercase */
write(LnInUpper),nl,
fail.
convert_file.

repfile(_).
repfile(F):-
not(eof(F)),
repfile(F).

Chapter 214, Writing, Reading, and Files 287

GOAL
write("Which file do you want convert ?"),
readln(InputFileName),nl,
write("What is the name of the output file ?"),
readln(OutputFileName),nl,
openread(input, InputFileName),
readdevice(input),
openwrite(output, OutputFileName),
writedevice(output),
convert_file,
closefile(input),
closefile(output).

flush/1
flush forces the contents of the internal buffer to be written to the named file. It
takes this format:
flush(SymbolicFileName) /* (i) */

flush also requests the operating system to flush its buffers. For versions of DOS
previous to 3.30, this entails closing and re-opening the file. For newer versions
of DOS, as well as the other platforms, the appropriate operating system function
is called.

existfile/1
existfile succeeds if OSFileName exists. It takes this format:
existfile(OSFileName) /* (i) */

where OSFileName may contain a directory path and the name itself may contain
wildcards, e.g. c:\psys\*.cfg. existfile fails if the name does not appear in the
directory. However, note that although existfile finds all files, including those
with the 'system' and 'hidden' attribute set, it doesn't find directories. This may be
accomplished using the directory search predicates described later on.
You can use the following sequence to verify that a file exists before attempting
to open it.
open(File, Name) :-
existfile(Name), !,
openread(File, Name).
open(_, Name) :-
write("Error: the file ", Name," is not found").
existfile/2
In UNIX, existfile is also available in a two- arity version:
existfile(OSFileName,AccessMode) /* (i, i) */

with AccessMode specifying the type of access desired. This should be one of the
following constants:
f_ok to test for existence
x_ok to test for execute permission
w_ok to test for write permission
r_ok to test for read permission
These constants are declared in the include file iodecl.con.
existfile with only one argument tests for file-existence only.

searchfile/3
searchfile is used to locate a file along a path list, and is a kind of automated
existfile. It takes three arguments, as follows:
searchfile(PathList,Name,FoundName) /* (i,i,o) */

The PathList is a string containing one or more paths, separated by semicolons

(or colons, for UNIX), and Name is the name of the file to locate. If found,
FoundName will be bound to the fully qualified name, otherwise searchfile will
fail. For instance, for DOS
SearchFile(".;..;C:\\","autoexec.bat",FoundName),

will - provided autoexec.bat is located in the root of drive C - set FoundName to

C:\AUTOEXEC.BAT.

The file name may contain wildcards. In that case, FoundName is bound to the
fully qualified wildcard name, which may subsequently be used as input to the
directory matching predicates described later on. For instance, if the name is
specified as *.bat instead of autoexec.bat in the above example, FoundName will
be bound to C:\*.BAT.

deletefile/1
deletefile removes the file specified by its argument:
deletefile(OSFileName) /* (i) */

Chapter 214, Writing, Reading, and Files 289

deletefile gives an error if it can't remove the file. The OSFileName can not
contain wildcards.

renamefile/1
renamefile renames the file OldOSFileName to NewOSFileName. It takes this
format:
renamefile(OldOSFileName, NewOSFileName) /* (i, i) */

renamefile succeeds if a file called NewOSFileName doesn't already exist and

both names are valid file names; otherwise, it gives an error.

disk/1
disk is used to change the current disk and/or directory; it takes this format:
disk(Path) /* (i) (o) */

Called with a free variable, disk will return the current directory. In the DOS-
related versions, to change to another disk without changing the existing current
directory on that disk, use D:. where D is the drive letter.

copyfile/2
copyfile is used to copy a file. It takes two file names as follows:
copyfile(SourceName,DestinationName) /* (i,i)*/

The names may be partly or fully qualified file names, including disks and
directories. However, no wildcards are allowed. The copied file will have the
same attributes (and permissions) as those of the source.

File Attributes
Although the standard file open predicates described previously cover all general
cases, there may be a need to open or create files with specialized attributes and
non-obvious sharing modes. To this end Visual Prolog incorporates a general
purpose open predicate, but before discussing that we need to look at file
attributes and sharing modes.
The attributes and access modes used by Visual Prolog use the same values as
your operating system, with the exception of the default ('normal') attribute in the
NonUNIX-related versions of Visual Prolog. However, for easy porting to other
environments, you should avoid coding inherently non-portable constructs such
as file attributes (and even the fact that files have attributes) all over an
application. Rather, wrap things up nicely and write your own intermediate level
of predicates, getting and setting information in transparent ways.
The attributes and sharing modes are found in the include file IODECL.CON.

Opening and creating files

When opening or creating a file, the OS needs to know the file's attributes (e.g.
'hidden'), the type of use or access (e.g. 'read'), and how the file may be shared
with other programs while open (e.g. 'deny write'). Don't confuse these - they are
three different pieces of information, only partly related:

Attributes
The file attributes are the permanent attributes relating to the physical file on
disk, whether currently in use by a program or not. In DOS and OS/2 there's only
a few attributes, such as 'read only' and 'hidden'. These attributes inform the
operating system about how it may handle the file. Network and multiuser
operating systems, such as UNIX, typically have a much wider range of
attributes. These may include access allowed by other users (e.g. 'execute-only',
no read or write, giving copy-protection) and direct instructions to the OS ('this is
an executable program').
The attributes have no effect when opening an existing file, as files are unique
based on names only. They only apply when creating a new file.
The standard predicates described in the previous section all reference 'normal'
files. However, when a file has been modified the archive bit will automatically
be set by the operating system when the file is closed.

Access Modes
The access modes indicate how the file will be used. The OS will combine this
information with the files physical attributes, to determine if the access requested
is acceptable. For instance, opening a file having the read-only physical attribute
set, with either fm_access_wo or fm_access_rw will not be accepted.

Sharing Modes
The sharing modes indicate how this process views sharing with other processes.
The OS will combine the sharing and access modes with the sharing and access
modes specified by other processes, if the file is already in use, to determine if the
open call should succeed. If successful, the modes will restrict future open
attempts.

Chapter 214, Writing, Reading, and Files 291

Note that conceptually the sharing and access modes work both ways to form a
combined set of restrictions on the file: they specify both what the process wants
from a file and what it will allow from other processes. For instance, if a file has
been opened with 'deny write' and 'read only', an open attempt with 'deny none'
and 'write only' will fail because the first process has specified 'deny write' - in
this case it is the existing restriction on the file that rejects the open attempt. On
the other hand, an open attempt with 'deny read' and 'read only' will fail because
the file is already open with read access - in this case it is the current requirement
that rejects the open attempt.
Note that the fm_sh_denyrw denies all modes from other processes; it doesn't
mean 'deny read- write, but allow read-only or write-only'.
All the standard predicates described in the previous section specify the sharing
mode as 'deny write'.

Special File Modes for OS/2, DOS >= 4.0 and UNIX
OS/2 and DOS versions greater than or equal to 4.0, have a special fm_returnerr
mode:
The fm_returnerr specify that "media" errors occurring after the file has been
opened should return an error to the program, rather than be reported through a
pop-up window. Media errors are those indicating a malfunction of the device,
e.g. if writing to a floppy and the drive door is opened - this generates the well-
known 'Drive not ready' error. The standard predicates described in the previous
section do not specify fm_returnerr, so media errors will generate a pop-up
through OS/2's critical error handler.
UNIX, OS/2 and DOS >= 4.0 also have a write- through mode:
The fm_writethru specifies that no write buffering should take place. In this case,
every single byte written to the file cause both the physical disk image and the
directory entry to be updated, giving a secure file. However, disk writes
performed with write-through are excessively slow compared to ordinary writes.

openfile/5
With the general-purpose openfile predicate, files may be created or opened in
nonstandard ways. openfile looks like this:
openfile(SymbolicName,OSName,OpenMode,Attributes,Creation)
/* (i,i,i,i,i) */
The SymbolicName and OSName are the same as for the previously described
standard predicates. The rest of the arguments are as follows (please refer to the
section on File Attributes a few pages back):
OpenMode is the access and sharing modes for the file. It is formed by
adding together one of the fm_access_XX values, one of the
fm_sh_XXXXXX and optionally fm_returnerr and fm_writethru. If no
access mode is specified, it will be set to 'read only'. If no sharing mode is
specified, it will be set to 'deny write'.
Attributes are the attributes for the physical disk file. Valid attributes on
DOS and OS/2 are fa_rdonly, fa_hidden, fa_system, fa_arch and
fa_normal. If nothing (0) is specified, the attributes will be set to
fa_normal. The system and the hidden attributes both have the same effect,
namely to hide the file when a 'dir' command is executed. Note that DOS
and OS/2 automatically set the archive attribute when an updated file is
closed. For UNIX, the attributes correspond to the file's permissions.
Creation specifies how the presence or absence of an existing file with the
same name is to be handled. It is formed by adding at most one from the
cr_ex_XX group and at most one from the cr_noex_XX group. Pay attention
to Creation defaults - if nothing (0) is specified. Note that this is the
equivalent of specifying cr_ex_fail and cr_noex_fail, i.e. fail if it exists and
fail if it doesn't exist. But remember that the actual default Creation action
will be set according to the access mode as follows:
fm_access_ro -> cr_ex_open + cr_noex_fail
fm_access_wo -> cr_ex_replace + cr_noex_create
fm_access_rw -> cr_ex_open + cr_noex_create

A sensible Creation default for read-write access is a bit tricky: If read-

write is specified because the file is opened for 'modification', an existing
file of the same name should be opened, not replaced. This is therefore the
default. However, if read-write is specified because one wants bidirectional
access to a new file, an existing file of the same name should be deleted.
This is possible with a call to openfile as follows:
:
FMode = fm_access_rw + fm_sh_denywr + fm_returnerr,
FCrea = cr_ex_replace + cr_noex_create,
openfile(dbfile,"salient.dba",FMode,fa_normal,FCrea),
:

Chapter 214, Writing, Reading, and Files 293

File and Path Names

A set of standard predicates ease file name handling and enable searching for files
on a disk.

filenamepath/3
filenamepath is used to compose and decompose a fully qualified name around
its path and file name. It takes three arguments, as follows:
filenamepath(QualName,Path,Name) /* (i,o,o) (o,i,i)*/

filenamepath converts between QualName on one side, and Path and Name on
the other. The programs ch11e12.pro and ch11e13.pro contain examples for DOS
and UNIX respectively; both examples do essentially the same thing:
/* Program ch234e12.pro */

GOAL
QualName="c:\\vip\\bin\\prolog.err",
FileNamePath(QualName,Path,Name),
write("\nQualName=",QualName),
write("\nPath=",Path),
write("\nName=",Name),
FileNamePath(NewName,Path,Name),
write("\nConverted back: ",NewName),nl.

/* Program ch235e13.pro */

GOAL
QualName="/usr/bin/prolog.err",
FileNamePath(QualName,Path,Name),
write("\nQualName=",QualName),
write("\nPath=",Path),
write("\nName=",Name),
FileNamePath(NewName,Path,Name),
write("\nConverted back: ",NewName),nl.

This will set Path to C:\VIP\BIN and name to PROLOG.ERR; finally, NewName
will be set to C:\VIP\BIN\PROLOG.ERR. Note that under DOS, all Visual
Prolog file name handling converts the name to upper case. This is because there
has in the past been confusion with respect to upper and lower case versions of
some foreign characters.
Please, using the (o,i,i) flow pattern of this predicate, take into account some
special cases described in the filenamepath topic in VDE help.

filenameext/3
filenameext is used to compose and decompose a (fully qualified) file name
around its extension, defined by a dot. It takes three arguments, as follows:
filenameext(Name,BaseName,Ext) /* (i,o,o) (o,i,i)*/

Here is the DOS example:

/* Program ch236e14.pro */

GOAL
Name="c:\\vip\\bin\\win\\16\\vip.exe",
FileNameExt(Name,BaseName,Ext),
write("\nName=",Name),
write("\nBaseName=",BaseName),
write("\nExt=",Ext),
FileNameExt(NewName,BaseName,Ext),
write("\nConverted back: ",NewName),
% Override the old extension
FileNameExt(NewName1,"VIP.EXE",".HLP"),
write("\nNewName1=",NewName1),nl.

This will set BaseName to C:\VIP\BIN\WIN\16\VIP and Ext to .EXE; then

NewName is set to C:\VIP\BIN\WIN\16\VIP.EXE and finally NewName1
demonstrates a direct extension override - it isn't necessary to explicitly remove
the old extension first. Note that the dot is considered a part of the extension and
that - as with filenamepath - in the DOS version, everything is converted to upper
case.

Directory Searching

Visual Prolog includes directory search predicates, enabling file name matching
with wildcards. In addition, the predicates return all relevant information about
the directory entries found.
Directory searching is very file system dependent and you should therefore guard
yourself against future changes by isolating these predicates when they're used.
Don't spread them all over your application, and don't rely on their arguments and
functionality remaining unchanged. In fact, don't even rely on them being the
same for different versions of OS/2: the installable file system concept means that
Chapter 214, Writing, Reading, and Files 295
future versions of OS/2 may behave very differently, although great effort will be
spent in attempts to provide portability. Below, all references to OS/2 refer to the
FAT file system.
Basically, to find matching files the directory has to be opened; this is done with
the diropen predicate, specifying the file name mask and the attributes to look for.
Then, by calling the dirmatch predicate, the matching files are found one by one.
Finally, the directory is closed by a call to the dirclose predicate.
Generally, the predicates behave identically irrespective of platform: a file name -
optionally containing wildcards - is used to specify the names to match, and a set
of search attributes refine the match (for a list of attributes, see the section on File
Attributes earlier in this chapter). However, unlike the DOS and OS/2 directory
search mechanisms, the search attributes don't increase the search beyond 'normal'
files. Visual Prolog considers all attributes as strictly informational, and they may
all be used for file selection. When using the directory search predicates, you
should therefore specify the attributes you are interested in: if you for instance
want everything with the archive bit set, specify fa_arch; if you want everything
with the system bit set, specify fa_system; if you want 'normal' files, specify
fa_normal, etc. You should be aware, though, that the attributes specified are
inclusive of each other: if several attributes are combined, the directory search
will find everything matching at least one of the attributes, but the entry found
won't necessarily match all the attributes. In other words, using set terminology, it
is the union of the matching files, not the intersection, which is returned. Exactly
what is found may be determined by bitwise testing of the returned attribute.
UNIX users should be aware that only one kind of directory entry (such as a
normal file, a pipe, a directory, etc.) may be searched for at a time. No
permissions of the entries are considered, and none should be specified.

diropen/3
diropen is used to gain access to a directory. It takes the following format:
diropen(Wild,Attrib,Block) /* (i,i,o) */

where Wild is a file name, optionally containing wildcards, Attrib are the required
search attributes, and Block is an information block used by subsequent calls to
dirmatch. To the compiler this block looks like a string, but it contains more
information than meets the eye. Therefore, it cannot be asserted in a database and
then retracted for use at a later stage - as long as the directory is open, it must be
held in a local variable (or an argument, which is the same thing). diropen will
fail if there are no files matching the specification; however, if the file name
includes a directory which doesn't exist, diropen will exit with an error.
Several diropens may be active simultaneously; in other words, they may be
nested and used recursively.

dirmatch/10
dirmatch will, after diropen has returned an information block, return the name
and other information for each matching file, one at each call. It looks as follows:
dirmatch(Block,Name,Attr,Hour,Min,TwoSec,Year,Month,Day,Size)
/* (i,o,o,o,o,o,o,o,o,o) */

The Block is the information block returned by diropen, Name is the matching
name, and Attr are the attributes for the entry found. The rest of the arguments
should be self-explanatory - they're all unsigned integers, apart from Size, which
is an unsigned long. Note that DOS and OS/2 use only 5 bits to encode the
seconds part of the time stamp, giving at most 32 different values - hence the
TwoSec name.
Upon each call, dirmatch returns the next matching directory entry. When there
are no more matches, dirmatch fails; if this happens, dirmatch will automatically
close the directory.
You should be aware that if searching for subdirectories with a name
specification of e.g. "*.*", dirmatch will always return the entries "." and ".." if
these are returned by the operating system. Therefore, dirmatch is likely to find
directories in all directories except perhaps the root.

dirclose/1
dirclose will close a previously opened directory. It takes the form:
dirclose(Block) /* (i) */

where Block is the information block returned by diropen. Note that if dirmatch
is repeatedly called until it fails (because there are no more matching files),
dirclose should not be called, as dirmatch will have closed the directory.

Example
The following demonstrates the use of the directory matching predicates, to make
an existdir predicate to complement the existfile standard predicate described
previously.

Chapter 214, Writing, Reading, and Files 297

 /* Program ch237e16.pro */

include "iodecl.con"

PREDICATES
existdir(string)
exd1(string)
exd2(string,string)

CLAUSES
existdir(Wild):-
diropen(Wild,fa_subdir,Block),
exd1(Block),
dirclose(Block).

exd1(Block):-
dirmatch(Block,Name,_,_,_,_,_,_,_,_),
exd2(Block,Name).

exd2(_,Name):-
not(frontchar(Name,'.',_)),!.
exd2(Block,_):-
exd1(Block).

Given for instance the goal existdir("c:\\*.*") in DOS, it will - unless you have
a rather extraordinary disk organization - say 'yes'. However, it will only find
subdirectories in existing paths - if you ask for e.g. existdir(c:\\jnk\\*.*")
without having a directory called 'JNK' in the root of drive c, it will exit with an
error. You should also be aware that in DOS the root itself can't be matched: there
is no directory called '\', and existdir("c:\\") will fail. This is an operating
system defined restriction of DOS, and is not relevant in UNIX where '/' does
exist.
Note, by the way, how the current and parent directory entries ("." and "..") are
filtered out in the example.

dirfiles/11
Having presented the hard way of finding files, here's the easy way. dirfiles is a
non- deterministic standard predicate which, upon backtracking, returns all
matching files one by one. It looks as follows:
dirfiles(Wild,Attrib,Fnam,RetAttr,Hour,Min,Sec,
Year,Month,Day,Size) /* (i,i,o,o,o,o,o,o,o,o,o) */

The use of dirfiles obviates the need to open and close the directory as this is
handled automatically, but there is a condition attached: in order to use it
correctly, it must be backtracked into until it fails. It is the final failure of the
predicate which automatically closes the directory. You should be aware that
neither the compiler nor the code supporting a running program has any way of
detecting if this won't happen - it is entirely the programmers responsibility.
Having said that, no serious harm will come from leaving a couple of directories
open, but eventually the system will run out of handles.
As with diropen, calls to dirfiles may be nested and used recursively.

DOS Example
Below is a sample program which will traverse all directories on drive C,
searching for entries having the 'system' or 'hidden' attribute set. The OS will
typically have a couple of hidden files in the root directory. However, if there are
hidden files elsewhere on the disk, be suspicious! They're probably harmless
copy- protection or configuration files for commercial software you have
installed, but why hide any files?
/* Program ch238e17.pro */

/* Copyright (c) 1986, '95 by Prolog Development Center */

CONSTANTS
fa_hidden = $02 /* Hidden file */
fa_system = $04 /* System file */
fa_subdir = $10 /* Subdirectory */

fa_hidsys = $06 /* hidden + system */

PREDICATES
findhidden(string,string)
wrattr(integer)

CLAUSES
wrattr(A):-
bitand(A,fa_hidden,AA),
AA<>0,write('H'),fail.
wrattr(A):-bitand(A,fa_system,AA),
AA<>0,write('S'),fail.
wrattr(A):-
bitand(A,fa_subdir,AA),
AA<>0,write('D'),fail.
wrattr(_).

Chapter 214, Writing, Reading, and Files 299

 findhidden(CurrPath,Wild):-
write(CurrPath,":\n"),
filenamepath(FileSpec,CurrPath,Wild),
dirfiles(FileSpec,fa_hidsys,FileName,RetAttr,_,_,_,_,_,_,_),
wrattr(RetAttr),
write('\t',FileName,'\n'),
fail.
findhidden(CurrPath,Wild):-
filenamepath(DirSpec,CurrPath,"*.*"),
dirfiles(DirSpec,fa_subdir,Name,_,_,_,_,_,_,_,_),
not(frontchar(Name,'.',_)),
filenamepath(DirName,CurrPath,Name),
findhidden(DirName,Wild),
fail.
findhidden(_,_).

GOAL
findhidden("C:\\","*.*").

This example also demonstrates decoding the returned attribute (in the wrattr
predicate), by means of bitwise testing.

Manipulating File Attributes

A standard predicate enables getting and setting the (informational) attributes of

files. Although documentation for DOS, OS/2 and MS Windows frequently talks
about a "directory attribute", a file cannot be changed to a directory just by
clearing this attribute.

fileattrib/2
Depending on dataflow fileattrib will get or set the attributes for a file. In UNIX
this corresponds to the file mode, meaning permissions, sticky bits, etc; see
chmod(S).
fileattrib(Name,Attrib) /* (i,o) (i,i) */

The Name must specify an existing file, otherwise fileattrib exits with an error.
Note that the definition of getting or setting attributes is entirely operating system
defined; in particular, you cannot set the file attributes for a directory. The
attributes for the file appear in Attrib as an unsigned short. This may then be
decomposed and/or changed using bitwise manipulation. For instance, the
following will clear the system attribute for the DOS file "JNK":
 CONSTANTS
fa_system = $04 /* System file */
fa_notsys = $FFFB /* ~system file. */

GOAL
fileattrib("jnk",FA),
bitand(FA,fa_notsys,Plain),
fileattrib("jnk",Plain).

The constant fa_notsys is the bitwise negation of fa_system. If you don't know
how to find the negation, use the bitxor (see chapter 16) standard predicate:
CONSTANTS
fa_system = $04 /* System file */

GOAL
bitxor(fa_system,$FFFF,NotSys),
fileattrib("jnk",FA),
bitand(FA,NotSys,Plain),
fileattrib("jnk",Plain).

Handling terms in text files

The readterm predicate makes it possible to access facts in a file. readterm can
read any object written by the write predicate and takes the form
readterm(<name>,TermParam).

where <name> is the name of a domain. The following code excerpt shows how
readterm might be used.
DOMAINS
name,addr = string
one_data_record = p(name, addr)
file = file_of_data_records

PREDICATES
person(name, addr)
moredata(file)

Chapter 214, Writing, Reading, and Files 301

CLAUSES
person(Name,Addr) :-
openread(file_of_data_records, "dd.dat"),
readdevice(file_of_data_records),
moredata(file_of_data_records),
readterm(one_data_record, p(Name, Addr)).

moredata(_).
moredata(File) :-
not(eof(File)),
moredata(File).

If the file DD.DAT contains facts belonging to the one_data_record domain, such
as
p("Peter","28th Street")
p("Curt","Wall Street")

the following is an example of a dialog that retrieves information from that file:
Goal person("Peter",Address).

Address="28th Street"
1 Solution

Goal person("Peter","Not an address").

no

Manipulating Facts Like Terms

Facts that describe database predicates can also be manipulated as though they
were terms. This is made possible by the dbasedom domain (which is
automatically declared by the Visual Prolog system) or any user-defined database
domain name.
Visual Prolog generates one alternative for each predicate in the database. It
describes each database predicate by a functor and by the domains of the
arguments in that predicate. For example, given this database declaration:
DATABASE
person(name, telno)
city(cno, cname)

Visual Prolog generates the corresponding dbasedom domain:

DOMAINS
dbasedom = person(name, telno) ; city(cno, cname)
A named database similarly generates a domain corresponding to the database
name, as in the following example:
DATABASE - test
person(name, telno)
city(cno, cname)

generates the domain

DOMAINS
test = person(name, telno) ; city(cno, cname)

These domains can be used like any other predefined domain.

Example
The following example shows how you could construct a predicate my_consult,
similar to the standard predicate consult.
/* User-defined Predicate my_consult using readterm */

DOMAINS
file = dbase

DATABASE - dba1
/* ... Declare database predicates to be read from file */

PREDICATES
my_consult(string)
repfile(file)

CLAUSES
my_consult(FileName) :-
openread(dbase, FileName),
readdevice(dbase),
repfile(dbase),
readterm(dba1, Term),
assertz(Term),
fail.
my_consult(_) :- eof(dbase).

repfile(_).
repfile(F):-not(eof(F)),repfile(F).

If, for example, the database program section contains the declaration
p(string, string)

Chapter 214, Writing, Reading, and Files 303

and a file called DD.DAT exists (with contents as described on page 329), you
could obtain the following dialog:
Goal my_consult("dd.dat").
yes

Goal p(X,Y).
X=Peter, Y=28th Street
X=Curt", Y=Wall Street
2 Solutions

Summary

These are the important points covered in this chapter:

1. Visual Prolog provides three standard predicates for basic writing:
a. write (for plain output)
b. writef (for output formatted according to format specifiers)
c. nl (for generating new lines)
2. Visual Prolog basic standard predicates for reading are
a. readln (for reading whole lines of characters)
b. readint, readreal, and readchar (for reading integers, reals, and characters,
respectively)
c. readterm (for reading compound objects)
d. file_str (for reading a whole text file into a string)
3. Additionally, binary blocks may be transferred using
a. readblock (reads a binary block from a file)
b. writeblock (writes a binary block to a file)
c. file_bin transfers between whole files and binary blocks
4. Visual Prolog uses a current_read_device (normally the keyboard) for reading
input, and a current_write_device (normally the screen) for sending output.
You can specify other devices, and can reassign the current input and output
devices at run time. (This reassigning is known as redirecting I/O.)
5. Visual Prolog's basic file-access predicates are:
a. openread (open a file for reading)
 b. openwrite (open a file for writing)
c. openappend (open a file for appending)
d. openmodify (open a file for modification)
e. openfile (general purpose file opener)
f. filemode (set a file to text mode or binary mode)
g. closefile (close a file)
h. readdevice (reassign the current_read_device or get its name)
i. writedevice (reassign the current_write_device or get its name)
6. To access files you use the FILE domain, which has five predefined
alternatives:
a. keyboard (for reading from the keyboard)
b. screen (for writing to the screen)
c. stdin (for reading from standard input)
d. stdout (for writing to standard output)
e. stderr (for writing to standard error)
7. To work within and manipulate files, you use the following standard
predicates:
a. filepos (controls the position where reading or writing takes place)
b. eof (checks whether the file position during a read operation is at the end of
the file)
c. flush (forces the contents of the internal buffer to be written to a file)
d. existfile (verifies that a file exists)
e. searchfile (locates a file among several directories)
f. deletefile (deletes a file)
g. renamefile (renames a file)
h. copyfile (copies a file)
i. fileattrib (gets or sets file attributes)
j. disk (changes the current disk and directory/subdirectory)
8. To search directories, the following standard predicates are available:
a. diropen (opens a directory for searching)
Chapter 214, Writing, Reading, and Files 305
 b. dirmatch (finds matching files in a directory)
c. dirclose (closes a directory)
d. dirfiles (non-deterministically matches files in a directory)
9. To manipulate file names, the following are available:
a. filenamepath (joins or splits qualified file names)
b. filenameext (joins or splits file names and extensions)
10.The standard predicate readterm allows your program to access facts in a file
at run time. readterm can read any object written by write, plus facts
describing database predicates.
CHAPTER 239

String-Handling in Visual Prolog

240Visual PrologVisual PrologVisual PrologVisual Prolog provides several
standard predicates for powerful and efficient string manipulations. In this
chapter, we've divided these into two groups: the family of basic string-handling
predicates, and a set of predicates used for converting strings to other types and
vice versa. Strings may also be compared to each other, but this is covered in
chapter 241.

String Processing

A few formalities apply to strings and string processing, in that the backslash acts
as an escape character, allowing you to put non-keyboardable characters into
strings. Please see the description on page 95.

Basic String-Handling Predicates

The predicates described in this section are the backbone of string-handling in
Visual Prolog; as such, they serve several purposes:
dividing a string into component strings or tokens
building a string from specified strings or tokens
verifying that a string is composed of specified strings or tokens
returning a string, token, or list of these from a given string
verifying or returning the length of a string
creating a blank string of specified length
verifying that a string is a valid Visual Prolog name
formatting a variable number of arguments into a string variable

frontchar/3
frontchar operates as if it were defined by the equation

Chapter 242, String-Handling in Visual Prolog 307

 String1 = the concatenation of Char and String2

It takes this format:

frontchar(String1,Char,String2)
/* (i,o,o) (i,i,o) (i,o,i) (i,i,i) (o,i,i) */

frontchar takes three arguments; the first is a string, the second is a char (the first
character of the first string), and the third is the rest of the first string.
frontchar can be used to split a string up into a series of characters, or to create a
string from a series of characters, and to test the characters within a string. If the
argument String1 is bound to a zero-length string, the predicate fails.

Example
In Program 1, frontchar is used to define a predicate that changes a string to a list
of characters. Try the goal
string_chlist("ABC", Z)

This goal will return Z bound to ['A','B','C'].

/* Program ch243e01.pro */

DOMAINS
charlist = char*

PREDICATES
string_chlist(string, charlist)

CLAUSES
string_chlist("", []):-!.
string_chlist(S, [H|T]):-
frontchar(S,H,S1),
string_chlist(S1,T).

fronttoken/3
fronttoken performs three related functions, depending on the type of flow
pattern you use when calling it.
fronttoken(String1, Token, Rest)
/* (i,o,o) (i,i,o) (i,o,i) (i,i,i) (o,i,i) */

In the (i,o,o) flow variant, fronttoken finds the first token of String1, binds it to
Token, and binds the remainder of String1 to Rest. The (i,i,o), (i,o,i), and
(i,i,i) flow variants are tests; if the bound arguments are actually bound to the
corresponding parts of String1 (the first token, everything after the first token, or
both, respectively), fronttoken succeeds; otherwise, it fails.
The last flow variant (o,i,i) constructs a string by concatenating Token and Rest,
then binds String1 to the result.
A sequence of characters is grouped as one token when it constitutes one of the
following:
a name according to normal Visual Prolog syntax
a number (a preceding sign is returned as a separate token)
a non-space character
fronttoken is perfectly suited for decomposing a string into lexical tokens.

Example
Program 2 illustrates how you can use fronttoken to divide a sentence into a list
of names. If 2 is given the goal:
string_namelist("bill fred tom dick harry", X).

X will be bound to:

[bill, fred, tom, dick, harry]

/* Program ch244e02.pro */

DOMAINS
namelist = name*
name = symbol

PREDICATES
string_namelist(string, namelist)

CLAUSES
string_namelist(S,[H|T]):-
fronttoken(S,H,S1),!,
string_namelist(S1,T).
string_namelist(_,[]).

frontstr/4
frontstr splits String1 into two parts. It takes this format:

Chapter 242, String-Handling in Visual Prolog 309

 frontstr(NumberOfChars, String1, StartStr, EndStr)
/* (i,i,o,o) */

StartStr contains the first NumberOfChars characters in String1, and EndStr

contains the rest. When frontstr is called, the first two parameters must be bound,
and the last two must be free.

concat/3
concat states that String3 is the string obtained by concatenating String1 and
String2. It takes this format:
concat(String1, String2, String3)
/* (i,i,o), (i,o,i), (o,i,i), (i,i,i) */

At least two of the parameters must be bound before you invoke concat, which
means that concat always gives only one solution (in other words, it's
deterministic). For example, the call
concat("croco", "dile", In_a_while)

binds In_a_while to crocodile. In the same vein, if See_ya_later is bound, the call
concat("alli", "gator", See_ya_later)

succeeds only if See_ya_later is bound to alligator.

str_len/2
str_len can perform three tasks: It either returns or verifies the length of a string,
or it returns a string of blank spaces of a given length. It takes this format:
str_len(StringArg, Length) /* (i,o), (i,i), (o,i) */

str_len binds Length to the length of StringArg or tests whether StringArg has the
given Length. The Length is an unsigned integer. In the third flow version, str_len
returns a string of spaces with a given length; this can be used to allocate buffers,
etc. allocating buffers with str_len, but makebinary is preferable especially for
binary data.

isname/1
isname verifies that its argument is a valid name in accordance with Visual
Prolog's syntax; it takes this format:
isname(String) /* (i) */
A name is a letter of the alphabet or an underscore character, followed by any
number of letters, digits, and underscore characters. Preceding and succeeding
spaces are ignored.

format/*
format performs the same formatting as writef (see page 272), but format delivers
the result in a string variable.
format(OutputString,FormatString,Arg1,Arg2,Arg3,...,ArgN)
/* (o,i,i,i,...,i) */

subchar/3
subchar returns the character at a given position in a string; it takes the form:
subchar(String,Position,Char) /* (i,i,o) */

The first character has position 1. For example,

subchar("ABC",2,Char)

will bind Char to B. If the position specifies a character beyond the end of the
string, subchar exits with an error.

substring/4
substring returns a part of another string; it takes the form:
substring(Str_in,Pos,Len,Str_out) /* (i,i,i,o) */

Str_out will be bound to a copy of the string starting with the Pos’th character,
Len characters long, in Str_in. For example
substring("GOLORP",2,3,SubStr)]

binds SubStr to OLO. If Pos and Len specify a string partly or wholly outside of
Str_in, substring exits with an error. However, it is not an error to ask for 0 bytes
at the extreme end of the string:
substring("ABC",4,0,SubStr)]

will bind SubStr to an empty string (""), while

substring("ABC",4,1,SubStr)/* WRONG */]

is an error. By the way, so is

Chapter 242, String-Handling in Visual Prolog 311

 substring("ABC",5,-1,SubStr)/* WRONG */]

searchchar/3
searchchar returns the position of the first occurrence of a specified character in a
string; it takes the form:
searchchar(String,Char,Position) /* (i,i,o) */]

For example,
searchchar("ABEKAT",'A',Pos)]

will bind Pos to 1. If the character isn't found, searchchar will fail. Note that
searchchar is not re-satisfiable (i.e. if there are more occurrences of the specified
character in the string, backtracking won't find them), but you can easily make
your own:
/* Program ch245e03.pro */

PREDICATES
nondeterm nd_searchchar(string,char,integer)
nondeterm nd_searchchar1(string,char,integer,integer)
nondeterm nd_sc(string,char,integer,integer,integer)
run

CLAUSES
nd_searchchar(Str,Ch,Pos):-
nd_searchchar1(Str,Ch,Pos,0).

nd_searchchar1(Str,Ch,Pos,Old):-
searchchar(Str,Ch,Pos1),
nd_sc(Str,Ch,Pos,Pos1,Old).

nd_sc(_,_,Pos,Pos1,Old):- Pos = Pos1+Old.

nd_sc(Str,Ch,Pos,Pos1,Old):-
frontstr(Pos1,Str,_,Rest),
Old1 = Old + Pos1,
nd_searchchar1(Rest,Ch,Pos,Old1).

GOAL
nd_searchchar("abbalblablabbala",'a',P),
write(P,'\n'),
fail.

This implements a non-deterministic predicate (nd_searchchar) which is plug-

compatible with searchchar; if you don't mind typing the extra argument (Old) to
nd_searchchar1 yourself, you can of course discard a level of calls.
searchstring/3
searchstring returns the position of the first occurrence of a string in another
string; it takes the form:
searchstring(SourceStr,SearchStr,Pos) /* (i,i,o) */]

For example,
searchstring("ABEKAT","BE",Pos)]

will bind Pos to 2. If the search string isn't found in, or is longer than, the source
string, searchstring will fail. As with searchchar, searchstring isn't re-
satisfiable, but you can easily make your own. As a matter of fact, all that's
necessary is to take 3 and do a global substitution with 'string' replacing 'char',
and change the 'a' in the goal to a suitable search string, e.g. "ab":
GOAL
nd_searchstring("abbalblablabbala","ab",P),
write(P,'\n'),
fail.

Type Conversion

In this section, we summarize the standard predicates available for type

conversion. The predicates are char_int, str_char, str_int, str_real, upper_lower,
and finally term_str which converts between terms of any kind and strings.

char_int/2
char_int converts a character into an integer or an integer into a character; it takes
this format:
char_int(Char, Integer) /* (i,o), (o,i), (i,i) */

With both its arguments bound, char_int tests that the arguments correspond.
With one argument bound and the other free, char_int performs the conversion
and binds the output argument to the converted form of the input one.
Note: This predicate is really not needed in newer versions of Visual Prolog
because there is automatic conversion between characters and integers. We've left
char_int in to be compatible with older versions.

Chapter 242, String-Handling in Visual Prolog 313

str_char/2
str_char converts a string containing one and only one character into a character,
or converts a single character into a string of one character; it takes this format:
str_char(String, Char) /* (i,o), (o,i), (i,i) */]

In the (i,i) flow variant, str_char succeeds if String is bound to the single-
character string equivalent of Char. If the length of the string is not 1, str_char
fails.

str_int/2
str_int converts a string containing an integer into an integer, or converts an
integer into its textual representation; it takes this format:
str_int(String, Integer) /* (i,o), (o,i), (i,i) */]

In the (i,i) flow variant, str_int succeeds if Integer is bound to the integer
equivalent of the integer represented by String.

str_real/2
str_real converts a string containing a real number into a real number, or converts
a real number into a string; it takes this format:
str_real(String, Real) /* (i,o), (o,i), (i,i) */]

In the (i,i) flow variant, str_real succeeds if Real is bound to the real equivalent
of the real number represented by String.

upper_lower/2
upper_lower converts an upper-case (or mixed) string or character to all lower-
case, or a lower-case (or mixed) string or character to all upper-case; it takes this
format:
upper_lower(Upper, Lower) /* (i,o), (o,i), (i,i) */]

With both its arguments bound, upper_lower succeeds if Upper and Lower are
bound to strings that--except for the case of the letters--are identical; for instance,
the goal:
Str1=samPLEstrING,
Str2=sAMpleSTRing,
upper_lower(Str1, Str2)}
succeeds. Otherwise, it fails.
term_str/3
term_str is a general-purpose conversion predicate and will convert between
terms of a specified domain and their string representations. It looks like this:
term_str(Domain,Term,String) /* (i,i,o),(i,_,i) */]

where Domain specifies which domain the term belongs to. term_str could
replace the various str_* predicates above, for instance, str_real could be
implemented as str_real(S,R):- term_str(real,R,S). However, term_str is a
somewhat heavier mechanism.
The Domain need not be one of the standard domains, it can be any user-defined
domain:
/* Program ch246e04.pro */

DOMAINS
intlist = integer*

GOAL
write("Input list (example [66,73,76,83]): "),
readln(L),nl,
str_len(L,Len),
write("The stringlength of ",L),
write(" is ",Len,'\n').

Examples
This example defines the predicate scanner, which transforms a string into a list
of tokens. Tokens are classified by associating a functor with each token. This
example uses the predicates isname, str_int, and str_len to determine the nature
of the tokens returned by fronttoken.
/* Program ch247e05.pro */

/* Copyright (c) 1986, '95 by Prolog Development Center */

DOMAINS
tok = numb(integer); name(string); char(char)
toklist = tok*

PREDICATES
nondeterm scanner(string, toklist)
nondeterm maketok(string, tok)

Chapter 242, String-Handling in Visual Prolog 315

CLAUSES
scanner("",[]).
scanner(Str,[Tok|Rest]):-
fronttoken(Str, Sym, Str1), maketok(Sym, Tok), scanner(Str1, Rest).

maketok(S,name(S)):-isname(S).
maketok(S,numb(N)):-str_int(S,N).
maketok(S,char(C)):-str_char(S, C).

GOAL
write("Enter some text:"),nl,
readln(Text),nl,
scanner(Text,T_List),
write(T_List).

Conversions between the domain types symbol and string, and between char,
integer, and real, are handled automatically when using standard predicates and
during evaluation of arithmetic expressions. Reals will be rounded during
automatic conversions. Visual Prolog performs this automatic conversion as
necessary when a predicate is called, as in the following example:
PREDICATES
p(integer)

CLAUSES
p(X):- write("The integer value is ",X,'\n').

With this example, the following goals have the same effect:
X=97.234, p(X).
X=97, p(X).
X='a', p(X).

The following very simple English parser is a practical example of string parsing.
This example directly parses strings; if the parser were to be extended, the string
should be tokenized using a scanner similar to the one used in Program 4.
Whether you're parsing tokens or strings, the algorithm in this program is a good
example of how to start.
If you are interested in English-language parsing, we recommend that you take a
look at the Sentence Analyzer and Geobase programs in the VPI\PROGRAMS
subdirectory.
/* Program ch248e06.pro */

DOMAINS
sentence = s(noun_phrase,verb_phrase)
 noun_phrase = noun(noun) ; noun_phrase(detrm,noun)
noun = string

verb_phrase = verb(verb) ; verb_phrase(verb,noun_phrase)

verb = string

detrm = string

PREDICATES
nondeterm s_sentence(string,sentence)
nondeterm s_noun_phrase(string,string,noun_phrase)
nondeterm s_verb_phrase(string,verb_phrase)
d(string)
n(string)
v(string)

CLAUSES
s_sentence(Str,s(N_Phrase,V_Phrase)):-
s_noun_phrase(Str, Rest, N_Phrase),
s_verb_phrase(Rest, V_Phrase).

s_noun_phrase(Str,Rest,noun_phrase(Detr,Noun)):-
fronttoken(Str,Detr,Rest1),
d(Detr),
fronttoken(Rest1,Noun,Rest),
n(Noun).

s_noun_phrase(Str,Rest,noun(Noun)):-
fronttoken(STR,Noun,Rest),
(Noun).

s_verb_phrase(Str, verb_phrase(Verb,N_Phrase)):-
fronttoken(Str,Verb,Rest1),
v(Verb),
s_noun_phrase(Rest1,"",N_Phrase).

s_verb_phrase(Str,verb(Verb)):-
fronttoken(STR,Verb,""),
v(Verb).

/* determiner */

d("the").
d("a").

/* nouns */

n("bill").
n("dog").
n("cat").

Chapter 242, String-Handling in Visual Prolog 317

/* verbs */

v("is").

Load and run this program, and enter the following goal:
Goal s_sentence("bill is a cat", Result).

The program will return:

Result = s(noun("bill"),verb_phrase("is", noun_phrase("a","cat")))
1 Solution

Summary

These are the important points covered in this chapter:

1. Visual Prolog's string-handling predicates are divided into two groups: basic
string manipulation and string type conversions.
2. The predicates for basic string manipulation are summarized here:
a. frontchar, fronttoken, and concat for dividing a string into components,
building a string from specified components, and testing if a string is
composed of specified components; these components can be characters,
tokens, or strings
b. subchar and substring for returning a single character from, or a part of,
another string
c. searchchar and searchstring for finding the first occurrence of a character,
or a string, in a string
d. str_len for verifying or returning the length of a string, or creating a blank
string of specified length
e. frontstr for splitting a string into two separate strings
f. isname for verifying that a string is a valid Visual Prolog name
g. format for formatting a variable number of arguments into a string variable
Several of the basic string manipulation predicates have different flow
variants. The variants with only input parameters perform tests that succeed
when the string in question is made up of the specified components (or is of
the specified length).
3. The predicates for type conversion are listed here:
 a. char_int for converting from a character to an integer, or vice versa
b. str_char for converting a single character into a string of one character, or
vice versa
c. str_int for converting from an integer to its textual representation, or vice
versa
d. str_real for converting from a real number to a string, or vice versa
e. upper_lower for converting a string to all upper-case or all lower-case
characters, or testing case-insensitive string equality
f. term_str for conversion between arbitrary domains and strings
The type conversion predicates each have three flow variants; the (i,o) and
(o,i) variants perform the appropriate conversions, and the (i,i) variants are
tests that succeed only if the two arguments are bound to the converted
representations of one another

Chapter 242, String-Handling in Visual Prolog 319

CHAPTER 249

The External Database System

250Visual PrologIn this chapter, we cover Visual Prolog's external database
system. An external database is composed of an external collection of chained
terms; these chains give you direct access to data that is not a part of your Prolog
program. The external database can be stored in any one of three locations: in a
file, in memory, or in EMS-type expanded memory under DOS. The external
database supports B+ trees, which provide fast data retrieval and the ability to sort
quickly, and it supports multi-user access by a mechanism for serializing the file
accesses inside transactions.

External Databases in Visual Prolog

Visual Prolog's internal fact’s database, which uses asserta, assertz, retract, and
retractall, is very simple to use and suitable for many applications. However, the
RAM requirements of a database can easily exceed the capacity of your
computer; the external database system has been designed partly with this
problem in mind. For example, you might want to implement one or more of the
following:
a stock control system with an large number of records
an expert system with many relations but only a few records with complicated
structures
a filing system in which you store large text files in the database
your own database product--which maybe has nothing to do with a relational
database system--in which data is linked together in other, nonrelational ways
a system including several of these possibilities

Visual Prolog's external database system supports these different types of

applications, while meeting the requirement that some database systems must not
lose data during update operations--even in the event of power failure.
Visual Prolog's external database predicates provide the following facilities:
 efficient handling of very large amounts of data on disk
the ability to place the database in a file, in memory, or in EMS-type expanded
memory cards under DOS
multi-user access
greater data-handling flexibility than provided by the sequential nature of
Visual Prolog's automatic backtracking mechanism
the ability to save and load external databases in binary form

An Overview: What's in an External Database?

A Visual Prolog external database consists of two components: the data items--
actually Prolog terms--stored in chains, and corresponding B+ trees, which you
can use to access the data items very quickly.
The external database stores data items in chains (rather than individually) so that
related items stay together. For example, one chain might contain part numbers to
a stock list, while another might contain customer names. Simple database
operations, such as adding new items or replacing and deleting old items, do not
require B+ trees. These come into play when you want to sort data items or search
the database for a given item; they are covered in detail later in this chapter.

Naming Convention
The names of all the standard predicates concerned with database management
follow a certain convention.
The first part of the name (db_, chain_, term_, and so on) is a reminder of
what you must specify as input.
The second part of the name (flush, btrees, delete, and so on) is a reminder of
what action occurs or what is returned or affected.
For example, db_delete deletes a whole database, chain_delete deletes a whole
chain, and term_delete deletes a single term.

Chapter 251, The External Database System 321

 Figure 252.1: Structure of a Visual Prolog External Database253Visual Prolog

External Database Selectors

It is possible to have several external databases simultaneously in memory, on
disk, and in an EMS-type memory expansion card under DOS. With this
flexibility, you can place external databases where they give the best speed and
space compromise.
In order to distinguish between several open databases, you use a selector in
every call to an external database standard predicate. You must declare these
selectors in a domain called db_selector. This works like the file domain in the
file system. For example, the following domains, declarations, external databases
domain declaration declares customers and parts to be external database
selectors:
DOMAINS
db_selector = customers; parts

Chains
An external database is a collection of Prolog terms. Some examples of terms are
integers, reals, strings, symbol values, and compound objects; for instance, 32, -
194, 3.1417, "Wally", wages, and book(dickens, "Wally goes to the zoo").

Inside an external database, the terms are stored in chains. A chain can contain
any number of terms, and an external database can contain any number of chains.
Each chain is selected by a name, which is simply a string.
The following figure illustrates the structure of a chain called MY_CHAIN.

Figure 254.2: Structure of a Chain255

Database relations and database tables are modeled by chains of terms. For
example, suppose you have a customer, supplier, and parts database, and you
want to put all the data into a single database with three relations: one for
customers, one for suppliers, and one for parts. You do this by putting the
customers in one chain called customers, the suppliers in another chain called
suppliers, and the parts in a chain called parts.
To insert a term in an external database, you must insert the term into a named
chain. On the other hand, you can retrieve terms without explicitly naming the
containing chain. In both cases, you must specify the domain to which the term
belongs. In practice, it is best if all terms in the chain belong to the same domain,
but there is actually no restriction on how terms are mixed in a database. It's up to
you to ensure that a term you retrieve belongs to the same domain as it did when
you inserted it.

Chapter 251, The External Database System 323

The following is a simple example of setting up two chained databases, dba1 and
dba2. In this example, all the customer data is in dba1 and all the parts data in
dba2. For now, just look over this example; we need to introduce a lot more
information before we can explain what's happening here.
/* Program ch256e01.pro */

DOMAINS
db_selector = dba1 ; dba2
customers = customer(customer_name, address)
parts = part(part_name, ID, customer_name)
customer_name, part_name = symbol
ID = integer
address = string

PREDICATES
access

CLAUSES
access:-
chain_terms(dba1,chain1,customers,customer(Name, ADDR),_),
chain_terms(dba2,chain2,parts,part(Part, Id, Name),_),
write("send ",Part," part num ",Id," to ",Addr), nl,
fail.
access.

GOAL
% create the databases dba1 and dba2
db_create(dba1, "dd1", in_memory),
db_create(dba2, "dd1.bin", in_file),

% insert customer facts into chain1 in dba1

chain_insertz(dba1, chain1, customers,
customer("Joe Fraser","123 West Side"), _),
chain_insertz(dba1, chain1, customers,
customer("John Smith","31 East Side"), _),
chain_insertz(dba1, chain1, customers,
customer("Diver Dan","1 Water Way"), _),
chain_insertz(dba1, chain1, customers,
customer("Dave Devine","123 Heaven Street"), _),
 % insert parts facts into chain2 in dba2
chain_insertz(dba2, chain2, parts, part("wrench", 231,
"John Smith"), _),
chain_insertz(dba2, chain2, parts, part("knife", 331,
"Diver Dan"), _),
access,
db_close(dba1), db_close(dba2),
db_delete("dd1", in_memory),
db_delete("dd1.bin", in_file).

This program first creates the databases dba1 (in memory) and dba2 (in a disk
file). It then inserts facts into two chains: chain1 and chain2. After inserting the
facts, it looks in these chains for a customer and the part ordered by that
customer; finding these, it returns the address to which the shipper should ship the
part. Finally, it closes and deletes the two databases.

External Database Domains

The external database uses six standard domains, summarized here:

Domain What It's Used For

db_selector Domain for declaring database selectors
bt_selector Domain for declaring B+ tree selectors
place Location of the database: in RAM, in a file, or in an
extended memory system (EMS card under DOS)
accessmode Decides how the file will be used.
denymode Determines how other users can open the file.
ref Reference to the location of a term in a chain

Database Reference Numbers

Every time you insert a new term into an external database, Visual Prolog assigns
it a database reference number. You can use the term's database reference
number to retrieve, remove, or replace that term, or to get the next or previous
term in the chain. You can also insert a database reference number in a B+ tree
(as described later in this chapter), and then use the B+ tree to sort some terms or
to carry out a fast search for a term.
Database reference numbers are independent of the database location and any
possible packing operations. Once a reference number has been associated with a

Chapter 251, The External Database System 325

term, you can use that number to access that term--no matter which database
management operations are subsequently carried out--until the term is deleted.

The ref Domain

Database reference numbers are special because you can insert them in facts
sections and write them out with write or writef, but you can't type them in from
the keyboard. You must declare the arguments to predicates handling database
reference numbers as belonging to the standard domain ref.
When you delete a term with term_delete, the system will reuse that term's
reference number when it inserts the next term into the external database. This
happens automatically; however, if reference numbers have been stored in the
facts section or in a B+ tree for some reason, it is your responsibility to ensure
that a given reference is associated with the correct term.
To assist you in this, there is an error-checking option, enabled with the
db_reuserefs standard predicate:

db_reuserefs/2
db_reuserefs has the following form:
db_reuserefs(DBase,ReUse) /* (i,i)*/

where DBase is a db_selector and ReUse is an unsigned integer. This should be

set to 0 to enable checking for use of released terms, or 1 do disable this. The
overhead of having the check enabled is very small (4 bytes per term, virtually no
CPU overhead), but those 4 bytes will never be released. If you constantly create
and release terms, your database will therefore grow at a steady rate.
db_reuserefs's primary purpose is to assist you in tracking down bugs during
development of programs.

Manipulating Whole External Databases

When you create a new external database, or open an existing one, you can place
it in a file, in memory, or in EMS-type expanded memory under DOS, depending
on the value of the Place argument in your call to db_create or db_open. After
you've finished working with the external database, you close it with a call to
db_close.
When you place an external database in main or expanded memory, closing the
database with db_close does not delete the database from memory. You must do
this explicitly with a call to db_delete, to free the memory the database occupies.
If you close such an external database but don't delete it, you can later reopen it
with the db_open predicate.
Since the external database system relies on the DOS buffer system, it will be
very slow if no buffers have been allocated. To allocate 40 buffers (which isn't an
excessive number), include the following line in your CONFIG.SYS file (a part
of the DOS environment):
buffers = 40

In this section, we discuss the predicates db_create, db_open, db_copy,

db_loadems, db_saveems, db_close, db_delete, db_openinvalid, db_flush,
db_garbagecollect, db_btrees, db_chains, and db_statistics.

db_create/3
db_create creates a new database.
db_create(Dbase, Name, Place) /* (i,i,i) */

If the database is placed in a disk file, the name of the file will be Name; if it's
placed in memory or EMS under DOS, you'll need Name if you close the
database and want to open it later. Dbase and Name correspond to the internal and
external names for files.
Where you place an external database is determined by the Place argument. Place
can take one of the following values:

in_file The external database is placed in a disk file, and there

will be only a minimum of main memory overhead.
in_memory The external database is placed in the main memory--
usually this will be done to achieve maximum
performance.
in_ems The database is placed in EMS-type expanded
memory, if a suitable card is installed in the computer.
in_ems is only relevant for DOS. On other platforms it
has the same effect as in_memory

These values, in_file, in_memory, and in_ems, are elements of the pre-declared
domain place, which corresponds to the following declaration:
DOMAINS
place = in_file; in_memory; in_ems

For example, here are two different calls to db_create:

Chapter 251, The External Database System 327
 db_create(db_sel1,"MYFILE.DBA",in_file)
/* Creates disk file MYFILE.DBA */
db_create(db_sel2,"SymName2",in_memory)
/* Creates memory database SymName2 */

db_open/3
db_open opens a previously created database, identified by Name and Place.
db_open(Dbase, Name, Place) /* (i,i,i) */

If Place is in_memory or in_ems, Name is the database's symbolic file name; if

Place is in_file, Name is the actual DOS-style file name.

db_copy/3
Irrespective of where you initially place an external database, you can later move
it to another location with the db_copy predicate.
db_copy(Dbase, Name, Place) /* (i,i,i) */

For example, in this call to db_copy

db_copy(my_base, "new_EMSbase", in_ems)

Visual Prolog copies the database identified by the database selector my_base into
the new database file new_EMSbase, which is placed in EMS under DOS.
When you copy a database, the original still exists; you will have two copies until
you explicitly delete the original.
Once you've moved a database, all processing can continue as if nothing
happened, since all reference numbers to the external database terms will still be
valid. In this way, if you're maintaining an external database in main memory,
and free storage is running short, you can copy the database to a file and continue
execution with the database in the file. An index set up to the external database in
internal memory is still valid, even after you've copied the database to a file.
db_copy has several uses; you can use it to do the following:
Load a database from disk to memory and later save it again in binary form,
instead of using save and consult with text files.
Copy a medium-sized database from disk to memory for faster access.
Pack a database containing too much free space; when the database is copied
to another file, all free space will be eliminated.
db_loadems/2 and db_saveems/2
db_copy performs a full-scale record-by-record copy of the database in question.
This has the advantage that the resulting database will be compacted and without
unused space, but for large databases the process can be time consuming.
For DOS only, db_loadems and db_saveems will transfer complete images of
databases between disk and EMS:
db_loadems(FileName,EmsName /* (i,i) */
db_saveems(EmsName,FileName) /* (i,i) */

The only restriction on their use is that there can be no more than one database in
EMS.

db_openinvalid/3
db_openinvalid allows you to open a database that's been flagged as invalid.
db_openinvalid(Dbase, Name, Place) /* (i,i,i) */

If the power to the computer fails while a database is being updated, all the data
in the database may be lost because part of some buffer has not been written to
disk. A flag in the database indicates if it's in an invalid state after an update.
A database is recorded as being invalid after a call to any of the predicates that
change the content in the database. These include chain_inserta, chain_insertz,
chain_insertafter, term_replace, term_delete, chain_delete, bt_create,
key_insert, and key_delete. The database is recorded as being valid once again
when it is closed with db_close, or when db_flush is called to flush out the
buffers.
By using db_openinvalid, it is sometimes possible to continue execution when a
database is marked as invalid. This might make it possible to recover some data if
all your backups have disappeared. However, all attempts to use an invalid
database after opening it with db_openinvalid might yield unexpected results.

db_flush/1
db_flush flushes the buffers and writes their contents to the appropriate
destination in your database.
db_flush(Dbase) /* (i) */

When a database is updated it will be marked as invalid, and it remains flagged as

invalid until it is either flushed with db_flush, or closed.

Chapter 251, The External Database System 329

The level of security you employ for a given database will, of course, depend on
how important its data is. The most basic level of data security is to keep backups
on disk. At the intermediate level, you could call db_flush after each important
database update. However, flushing the buffers is a relatively slow operation; if
it's done too often, your database system will grind to a halt. Finally, if the
contents of an external database are especially valuable, you could record all
changes in a special log file or maintain two identical databases--perhaps on
different disks.

db_close/1
A call to db_close closes an open database.
db_close(Dbase) /* (i) */

If the database Dbase is placed in a disk file, the file will be closed. The database
won't be deleted, even if it is placed in memory or in an EMS-type memory
expansion card, and you can reopen it later through a call to db_open. You can
use db_delete to remove a closed database from memory.

db_delete/1
When the database is situated in memory or in an EMS-type memory expansion
card, db_delete releases all the occupied space.
db_delete(Name, Place) /* (i,i) */

When the database is situated in a file, db_delete erases the file. db_delete gives
an error if the database Name does not exist in the given Place.

db_garbagecollect/1
db_garbagecollect scans through the free lists in the database garbage collect and
tries to merge some of the free space together into larger pieces.
db_garbagecollect(Dbase) /* (i) */

This scanning and merging is done automatically when the database is placed in
memory or in an EMS card.
Under normal circumstances, there should be no need to call this predicate.
However, if there seems to be too much free space in the database that is not
being reused when new terms are inserted, db_garbagecollect can regain some
extra space.
db_btrees/2
During backtracking, db_btrees successively binds BtreeName to the name of
each B+ tree in the Dbase database.
nondeterm db_btrees(Dbase, BtreeName) /* (i,o) */

The names are returned in sorted order. B+ trees are described later in this
chapter.

db_chains/2
During backtracking, db_chains successively binds ChainName to the name of
each chain in the Dbase database.
nondeterm db_chains(Dbase, ChainName) /* (i,o) */

The names are returned in sorted order.

db_statistics/5
db_statistics returns statistical information for the database Dbase.
db_statistics(Dbase, NoOfTerms, MemSize, DbaSize, FreeSize)
/* (i,o,o,o,o) */

The arguments to db_statistics represent the following:

NoOfTerms is bound to the total number of terms in the database.

MemSize is bound to the size--in bytes--of the internal tables stored

in memory for the database.
DbaSize is bound to the total number of bytes that the terms and
descriptors in the Dbase database occupy. If Dbase is
stored in a disk file, and DbaSize gets a value much
smaller than the size of that file, the file can be
compressed by using db_copy.
FreeSize becomes bound to a value representing the free memory
space; the value depends on where the database Dbase is
currently placed.
When Dbase is placed in memory, FreeSize is bound
to the number of unused bytes between the top of the
global stack and the top of the heap. (Note: There
might be some additional free bytes that are not
included in this count.)

Chapter 251, The External Database System 331

 If Dbase is placed in EMS-type expanded memory,
FreeSize is bound to the number of unoccupied bytes
in that expansion memory.
When Dbase is placed in a file, FreeSize is bound to
the number of unused bytes on the disk containing the
file.

Manipulating Chains
To insert terms into an external database chain, you use the predicates
chain_inserta, chain_insertz, or chain_insertafter. You can successively bind
the terms in a chain, and their reference numbers, to the arguments of
chain_terms, while chain_delete allows you to delete a whole chain of terms
from the external database.
Four standard predicates return database reference numbers. These are
chain_first, chain_last, chain_next, and chain_prev.

chain_inserta/5 and chain_insertz/5

The predicates chain_inserta and chain_insertz correspond to asserta and
assertz, respectively. These take the following form:
chain_inserta(Dbase, Chain, Domain, Term, Ref) /* (i,i,i,i,o) */
chain_insertz(Dbase, Chain, Domain, Term, Ref) /* (i,i,i,i,o) */

chain_inserta inserts the term Term at the beginning of the chain Chain, while
chain_insertz inserts Term at the chain's end. Dbase is the db_selector of the
database, Domain is the domain of Term, and Ref is the database reference
number corresponding to Term. For example, if my_dba is declared to be in the
domain db_selector, like this:
DOMAINS
db_selector = my_dba; ....

then in this call to chain_inserta

chain_inserta(my_dba, customer, person, p(john,
"1 The Avenue", 32), NewRef)

customer is the name of the chain, and all customers are stored in one chain. It
would be perfectly all right to store the suppliers as terms from the domain person
but in a different chain, perhaps called supplier. person is the name of the domain
to which p(john, "1 The Avenue", 32) belongs, as shown in this domain
declaration:
DOMAINS
person = p(name, address, age)

If Chain doesn't already exist, these predicates will automatically create it.

chain_insertafter/5
chain_insertafter inserts a term after a specified term, returning the inserted
term's new reference number. It takes this format:
chain_insertafter(Dbase, Domain, Ref, Term, NewRef)
/* (i,i,i,i,o) */

chain_insertafter inserts the term Term after the chain element specified by Ref,
while NewRef is bound to the database reference number corresponding to Term
after it's been inserted.

chain_terms/5
During backtracking, chain_terms successively binds Term and Ref to each term
and its associated database reference number in the specified Chain. chain_terms
takes the form:
chain_terms(Dbase, Chain, Domain, Term, Ref) /* (i,i,i,o,o) */

chain_delete/2
chain_delete deletes a specified chain from a given external database; this
predicate takes the form:
chain_delete(Dbase, Chain) /* (i,i) */

chain_first/3 and chain_last/3

chain_first and chain_last return the database reference number for the first and
last terms in a given chain, respectively.
chain_first(Dbase, Chain, FirstRef) /* (i,i,o) */
chain_last(Dbase, Chain, LastRef) /* (i,i,o) */

Chapter 251, The External Database System 333

chain_next/3 and chain_prev/3
chain_next returns the reference number of the term following the given one,
while chain_prev returns the reference number of the term preceding the given
one.
chain_next(Dbase, Ref, NextRef) /* (i,i,o) */
chain_prev(Dbase, Ref, PrevRef) /* (i,i,o) */

Manipulating Terms
Three standard predicates for external database management are all concerned
with terms; these are term_replace, term_delete, and ref_term. Whenever you
call any of the term-handling external database standard predicates, you must give
the domain of the term as one of the arguments. Because of this, it's usually a
good idea to declare all terms in a given database as alternatives in one domain,
as in this declaration:
DOMAINS
terms_for_my_stock_control_database =
customer(Customer, Name, ZipCode, Address);
supplier(SupplierNo, Name, Address);
parts(PartNo, Description, Price, SupplierNo)

Note that there are no restrictions on mixing types (domains) in an external

database. One chain can contain text strings, another integers, a third some kind
of compound structures, and so on. However, external database data items are not
stored with type descriptors; for example, integers don't necessarily occupy just
two bytes. It's your responsibility to retrieve a term into the same domain as that
from which it was inserted. A run-time error will usually result if you attempt to
mix domains.

term_replace/4
term_replace replaces an old term (referenced by Ref, a database reference
number) with a new term, Term.
term_replace(Dbase, Domain, Ref, Term) /* (i,i,i,i) */

term_delete/3
term_delete erases the term stored under Ref, a given database reference number.
term_delete(Dbase, Chain, Ref) /* (i,i,i) */
The storage occupied by the term will be released, and there must be no further
references to Ref.

ref_term/4
ref_term binds Term to the term stored under a given reference number, Ref.
ref_term(Dbase, Domain, Ref, Term) /* (i,i,i,o) */

A Complete Program Example

The following example program 2, uses nearly all the external database predicates
introduced so far. Working first in memory, this program goes through the
following sequence of operations:
1. Writes 100 terms in a database.
2. Reads them back.
3. Replaces every second term.
4. Doubles the number of terms.
5. Erases every second term.
6. Examines every term with ref_term.
7. Calculates the size of the database.
This program then copies the database to a disk file and carries out the same
sequence of activities twice with the database held on disk. Finally, it calculates--
in hundredths of a second--the total time taken to carry out these activities. Note,
however, that for illustration the program generates large amounts of output,
which slows it down considerably. The true speed is only revealed if you remove
the output statements. The program 3 is for UNIX, as time-calculation is done
differently in UNIX, and terminal output is significantly slower in UNIX than in
DOS.
Run the program to see what happens, and then try to alter the number of terms
and study your system's performance. The DOS program appears below.
/* Program ch257e02.pro */

/* Copyright (c) 1986, '95 by Prolog Development Center */

DOMAINS
my_dom = f(string)
db_selector = my_dba

Chapter 251, The External Database System 335

PREDICATES
write_dba(integer)
read_dba
rd(Ref)
count_dba(integer)
count(Ref, integer, integer)
replace_dba
replace(Ref)
double_dba
double(Ref)
half_dba
half(Ref)
mixture

CLAUSES
write_dba(0):-!.
write_dba(N):-
chain_inserta(my_dba,my_chain,my_dom,f("Prolog system"),_),
chain_insertz(my_dba, my_chain, my_dom, f("Prolog Compiler"), _),
N1=N-1,
write_dba(N1).

read_dba:-
db_chains(my_dba, Chain),
chain_terms(my_dba, Chain, my_dom, Term, Ref),nl,
write("Ref=", Ref),
write(", Term=", Term),
fail.
read_dba:-
db_chains(my_dba, Chain),
chain_first(my_dba, Chain, Ref),
rd(Ref),
fail.
read_dba.

rd(Ref):-
ref_term(my_dba, my_dom, Ref, Term), nl,
write(Term),
fail.
rd(Ref):-
chain_next(my_dba,Ref,Next),!,rd(Next).
rd(_).

replace_dba:-
chain_first(my_dba, my_chain, Ref),
replace(Ref).
 replace(Ref):-
term_replace(my_dba, my_dom, Ref, f("Prolog Toolbox")),
chain_next(my_dba, Ref, NN),
chain_next(my_dba, NN, Next),!,
replace(Next).
replace(_).

half_dba:-
chain_last(my_dba, my_chain, Ref),
half(Ref).

half(Ref):-
chain_prev(my_dba, Ref, PP),
chain_prev(my_dba, PP, Prev), !,
term_delete(my_dba, my_chain, Ref),
half(Prev).
half(_).

double_dba:-
chain_first(my_dba, my_chain, Ref),
double(Ref).

double(Ref):-
chain_next(my_dba, Ref, Next),!,
chain_insertafter(my_dba, my_chain, my_dom, Ref,f("Programmers
Guide"), _),
double(Next).
double(_).

count_dba(N):-
chain_first(my_dba, my_chain, Ref),
count(Ref, 1, N).

count(Ref, N, N2):-
chain_next(my_dba, Ref, Next),!,
N1=N+1,
count(Next, N1, N2).
count(_, N, N).

Chapter 251, The External Database System 337

 mixture :-nl,
write("Replace every second term:"),
replace_dba,nl,
write("Double the number of terms:"),
double_dba,nl,
write("Erase every second term:"),
half_dba,nl,
write("Use ref_term for all terms:"),
read_dba,
count_dba(N),nl,
write("There are now ", N, " terms in the database"),
db_statistics(my_dba, NoOfTerms, MemSize, DbaSize, FreSize),nl,
writef("NoOfTerms=%, MemSize=%, DbaSize=%, FreeSize=%", NoOfTerms,
MemSize,DbaSize,FreSize).

GOAL
nl,nl,nl,
write("\tTEST OF DATABASE SYSTEM\n\t***********************\n\n"),
time(H1, M1, S1, D1),
db_create(my_dba, "dd.dat", in_memory),nl,nl,
write("Write some terms in the database:"),
write_dba(50),
read_dba,
mixture,nl,nl,

write("Copy to file"),
db_copy(my_dba, "dd.dat", in_file),
db_close(my_dba), db_delete("dd.dat", in_memory),
db_open(my_dba, "dd.dat", in_file),
mixture,
db_close(my_dba),nl,nl,nl,

write("Open the database on file"),

db_open(my_dba, "dd.dat", in_file),
mixture,
db_close(my_dba),

time(H2, M2, S2, D2),

Time = (D2-D1)+100.0*((S2-S1)+60.0*((M2-M1)+ 60.0*(H2-H1))),nl,nl,
write("Time = ", Time, "/100 Sec" ), nl.
B+ Trees

A B+ tree is a data structure you can use to implement a very sorting, large
amounts of data efficient method for sorting large amounts of data; B+ trees
enable a correspondingly efficient searching algorithm. You can think of a B+
tree as providing an index to a database, which is why B+ trees are sometimes
referred to as indices.
In Visual Prolog, a B+ tree resides in an external database. Each entry in a B+
tree is a pair of values: a key string key string and an associated database
reference number. When building your database, you first insert a record in the
database and establish a key for that record. The Visual Prolog btree predicates
may then be used to insert this key and the database reference number
corresponding to this record into a B+ tree.
When searching a database for a record, all you have to do is to obtain a key for
that record, and the B+ tree will give you the corresponding reference number.
Using this reference number, you can retrieve the record from the database. As a
B+ tree evolves, its entries are kept in key order. This means that you can easily
obtain a sorted listing of the records.
A B+ tree is analogous to a binary tree, with the exception that in a B+ tree, more
than one key string is stored at each node. B+ trees are also balanced; this means
that the search paths to each key in the leaves of the tree have the same length.
Because of this feature, a search for a given key among more than a million keys
can be guaranteed, even in the worst case, to require accessing the disk only a few
times--depending on how many keys are stored at each node.
Although B+ trees are placed in an external database, they don't need to point to
terms in the same database. It is possible to have a database containing a number
of chains, and another database with a B+ tree pointing to terms in those chains.

Pages, Order, and Keylength

In a B+ tree, keys are grouped together in pages; each page has the same size, and
all pages can contain the same number of keys, which means that all the stored
keys for that B+ tree must be the same size. The size of the keys is determined by
the KeyLen argument, which you must supply when creating a B+ tree. If you
attempt to insert strings longer than KeyLen into a B+ tree, Visual Prolog will
truncate them. In general, you should choose the smallest possible value for
KeyLen in order to save space and maximize speed.
When you create a B+ tree, you must also give an argument called its Order. This
argument determines how many keys should be stored in each tree node; usually,
Chapter 251, The External Database System 339
you must determine the best choice by trial and error. A good first try for Order is
4, which stores between 4 and 8 keys at each node. You must choose the value of
Order by experimentation because the B+ tree's search speed depends on the
values KeyLen and Order, the number of keys in the B+ tree, and your computer's
hardware configuration.

Duplicate Keys
When setting up a B+ tree, you must allow for all repeat occurrences of your key.
For example, if you're setting up a B+ tree for a database of customers in which
the key is the customer's last name, you need to allow for all those customers
called Smith. For this reason, it is possible to have duplicate keys in a B+ tree.
When you delete a term in the database, you must delete the corresponding entry
in a B+ tree with duplicate keys by giving both the key and the database reference
number.

Multiple Scans
In order multiple, scans of B+ trees to have more than one internal pointer to the
same B+ tree, you can open the tree more than once. Note, however, that if you
update one copy of a B+ tree, for which you have other copies currently open, the
pointers for the other copies will be repositioned to the top of the tree.

The B+ Tree Standard Predicates

Visual Prolog provides several predicates for handling B+ trees; these predicates
work in a manner that parallels the corresponding db_... predicates.

bt_create/5 and bt_create/6

You create new B+ trees by calling the bt_create predicate.
bt_create(Dbase, BtreeName, Btree_Sel, KeyLen, Order)
/* (i,i,o,i,i) */
bt_create(Dbase, BtreeName, Btree_Sel, KeyLen, Order, Duplicates)
/* (i,i,o,i,i,i) */

The BtreeName argument specifies the name for the new tree. You later use this
name as an argument for bt_open. The arguments KeyLen and Order for the B+
Tree are given when the tree is created and can't be changed afterwards. If you are
calling bt_create/5 or bt_create/6 with the Duplicates argument set to 1,
duplicates will be allowed in the B+Tree. If you call bt_create/6 with the
Duplicates argument set to 0 you will not be allowed to insert duplicates in the
B+Tree.

bt_open/3
bt_open opens an already created B+ tree in a database, which is identified by the
name given in bt_create.
bt_open(Dbase, BtreeName, Btree_Sel) /* (i,i,o) */

When you open or create a B+ tree, the call returns a selector (Btree_Sel) for that
B+ tree. A B+ tree selector belongs to the predefined domain bt_selector and
refers to the B+ tree whenever the system carries out search or positioning
operations. The relationship between a B+ tree's name and its selector is exactly
the same as the relationship between an actual file name and the corresponding
symbolic file name.
You can open a given B+ tree more than once in order to handle several
simultaneous scans. Each time a B+ tree is opened, a descriptor is allocated, and
each descriptor maintains its own internal B+ tree pointer.

bt_close/2 and bt_delete/2

You can close an open B+ tree with a call to bt_close or delete an entire B+ tree
with bt_delete.
bt_close(Dbase, Btree_Sel) /* (i,i) */
bt_delete(Dbase, BtreeName) /* (i,i) */

Calling bt_close releases the internal buffers allocated for the open B+ tree with
BtreeName.

bt_copyselector
bt_copyselector gives you a new pointer for an already open B+ tree selector (a
new scan).
bt_copyselector(Dbase,OldBtree_sel,NewBtree_sel) /* (i,i,o) */

The new selector will point to the same place in the B+ tree as the old selector.
After the creation the two B+ tree selectors can freely be repositioned without
affecting each other.

bt_statistics/8
bt_statistics returns statistical information for the B+ tree identified by Btree_Sel.
Chapter 251, The External Database System 341
 bt_statistics(Dbase,Btree_Sel,NumKeys,NumPages, /* (i,i,o,o, */
Depth,KeyLen,Order,PgSize) /* o,o,o,o) */

The arguments to bt_statistics represent the following:

Dbase is the db_selector identifying the database.

Btree_Sel is the bt_selector identifying the B+ tree.
NumKeys is bound to the total number of keys in the B+ tree
Btree_Sel.
NumPages is bound to the total number of pages in the B+ tree.
Depth is bound to the depth of the B+ tree.
KeyLen is bound to the key length.
Order is bound to the order of the B+ tree.
PgSize is bound to the page size (in bytes).

key_insert/4 and key_delete/4

You use the standard predicates key_insert and key_delete to update the B+ tree.
key_insert(Dbase, Btree_Sel, Key, Ref /* (i,i,i,i) */
key_delete(Dbase, Btree_Sel, Key, Ref) /* (i,i,i,i) */

By giving both Key and Ref to key_delete, you can delete a specific entry in a B+
tree with duplicate keys.

key_first/3, key_last/3, and key_search/4

Each B+ tree maintains an internal pointer to its nodes. key_first and key_last
allow you to position the pointer at the first or last key in a B+ tree, respectively.
key_search positions the pointer on a given key.
key_first(Dbase, Btree_Sel, Ref) /* (i,i,o) */
key_last(Dbase, Btree_Sel, Ref) /* (i,i,o) */
key_search(Dbase, Btree_Sel, Key, Ref) /* (i,i,i,o)(i,i,i,i) */

If the key is found, key_search will succeed; if it's not found, key_search will
fail, but the internal B+ tree pointer will be positioned at the key immediately
after where Key would have been located. You can then use key_current to return
the key and database reference number for this key. If you want to position on an
exact position in a B+ tree with duplicates you can also provide the Ref as an
input argument.
key_next/3 and key_prev/3
You can use the predicates key_next and key_prev to move the B+ tree's pointer
forward or backward in the sorted tree.
key_next(Dbase, Btree_Sel, NextRef) /* (i,i,o) */
key_prev(Dbase, Btree_Sel, PrevRef) /* (i,i,o) */

If the B+ tree is at one of the ends, trying to move the pointer further will cause a
fail, but the B+ tree pointer will act as if it were placed one position outside the
tree.

key_current/4
key_current returns the key and database reference number for the current pointer
in the B+ tree.
key_current(Dbase, Btree_Sel, Key, Ref) /* (i,i,o,o) */

key_current fails after a call to the predicates bt_open, bt_create, key_insert, or

key_delete, or when the pointer is positioned before the first key (using key_prev)
or after the last (with key_next).

Example: Accessing a Database via B+ Trees

The following example program handles several text files in a single database file
at once. You can select and edit the texts as though they were in different files. A
corresponding B+ tree is set up for fast access to the texts and to produce a sorted
list of the file names.
/* Program ch258e04.pro */

DOMAINS
db_selector = dba

PREDICATES
% List all keys in an index
list_keys(db_selector,bt_selector)

Chapter 251, The External Database System 343

CLAUSES
list_keys(dba,Bt_selector):-
key_current(dba,Bt_selector,Key,_),
write(Key,' '),
fail.
list_keys(dba,Bt_selector):-
key_next(dba,Bt_selector,_),!,
list_keys(dba,Bt_selector).
ist_keys(_,_).

PREDICATES
open_dbase(bt_selector)
main(db_selector,bt_selector)
ed(db_selector,bt_selector,string)
ed1(db_selector,bt_selector,string)

CLAUSES
% Loop until escape is pressed
main(dba,Bt_select):-
write("File Name: "),
readln(Name),
ed(dba,Bt_select,Name),!,
main(dba,Bt_select).
main(_,_).

% The ed predicates ensure that the edition will never fail.

ed(dba,Bt_select,Name):-
ed1(dba,Bt_select,Name),!.
ed(_,_,_).

%* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
% There are three choices:
%% a) The name is an empty string - list all the names
% b) The name already exists - modify the contents of the file
% c) The name is a new name - create a new file
%* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * */
 ed1(dba,Bt_select,""):-!,
key_first(dba,Bt_select,_),
list_keys(dba,Bt_select),
nl.
ed1(dba,Bt_select,Name):-
key_search(dba,Bt_select,Name,Ref),!,
ref_term(dba,string,Ref,Str),
edit(Str,Str1,"Edit old",NAME,"",0,"PROLOG.HLP",RET),
clearwindow,
Str><Str1, RET=0,
term_replace(dba, string, Ref, Str1).
ed1(dba,Bt_select,Name):-
edit("",STR1,"Create New",NAME,"",0,"PROLOG.HLP",RET),
clearwindow,
""><Str1, RET=0,
chain_insertz(dba,file_chain,string,Str1,Ref),
key_insert(dba,Bt_select,Name,Ref).

open_dbase(INDEX):-
existfile("dd1.dat"),!,
db_open(dba,"dd1.dat",in_file),
bt_open(dba,"ndx",INDEX).

open_dbase(INDEX):-
db_create(dba,"dd1.dat",in_file),
bt_create(dba,"ndx",INDEX,20,4).

GOAL
open_dbase(INDEX),
main(dba,INDEX),
bt_close(dba,INDEX),
db_close(dba).

External Database Programming

In this section, we provide seven examples that illustrate some general principles
and methods for working with Visual Prolog's external database system. This is a
summary of what the following sections cover:
"Scanning through a Database" shows you the way to perform a sequential scan
through a chain or a B+ tree in an external database.
"Displaying the Contents of a Database" defines a predicate you can use to
display the current state of an external database.

Chapter 251, The External Database System 345

"Making a Database That Won't Break Down" illustrates how to protect your
database from unexpected system power failure and other potential catastrophes.
"Updating the Database" provides an example that makes it easy to change, add
to, and protect your database.
"Using Internal B+ Tree Pointers" supplies you with some predicates for
positioning a pointer within an open B+ tree.
"Changing the Structure of a Database" offers an alternative to the old copy-
while-changing method of changing the structure of a database.

Scanning through a Database

When you are using the database system, it is important to keep Visual Prolog's
storage mechanisms storage mechanisms in mind. Every time Visual Prolog
retrieves a term from an external database with the ref_term predicate, it places
that term on the global stack. The system won't release the space occupied by that
term until the program fails and backtracks to a point before the call to ref_term.
This means that, to do a sequential scan through a chain in an external database,
you should always use a structure like the following:
/* Structure for sequentially scanning through a chain */

scan(db_selector, Chain, ....) :-

chain_first(db_selector, Chain, Ref),
scanloop(db_selector, Ref).

scanloop(db_selector, Ref) :-
ref_term(db_selector, mydom, Ref, Term),
/* ... do your processing ... */
fail.

scanloop(db_selector, _) :-
chain_next(db_selector, Ref, NextRef),
scanloop(db_selector, NextRef).

Similarly, for a sequential scan through an index, you should use a structure like
this:
/* Structure for sequentially scanning through an index */

scan(db_selector, Bt_selector) :-
key_first(db_selector, Bt_selector, FirstRef),
scanloop(db_selector, Bt_selector, FirstRef).
 scanloop(db_selector, Bt_selector, Ref) :-
ref_term(db_selector, mydom, Ref, Term),
/* ... do your processing ... */
fail.

scanloop(db_selector, Bt_selector, _) :-
key_next(db_selector, Bt_selector, NextRef),
scanloop(db_selector, Bt_selector, NextRef).

You can also carry out a sequential scan through a chain in the database by using
chain_terms, like this:
/* Another way to sequentially scan through a chain */

scan(db_selector, Chain) :-
chain_terms(db_selector, Chain, mydom, Term, Ref),
/* ... do your processing ... */
fail.
scan(_, _).

To scan through a B+ tree, you could have also defined and used the predicate
bt_keys. During backtracking, this predicate returns (for a given B+ tree and
database) each key in the tree and its associated database reference number.
/* This fragment goes with Program 5 */

PREDICATES
bt_keys(db_selector, bt_selector, string, ref)
bt_keysloop(db_selector, bt_selector, string, ref)

CLAUSES
bt_keys(Db_selector, Bt_selector, Key, Ref):-
key_first(Db_selector, Bt_selector, _),
bt_keysloop(Db_selector, Bt_selector, Key, Ref).

bt_keysloop(Db_selector, Bt_selector, Key, Ref):-

key_current(Db_selector, Bt_selector, Key, Ref).

bt_keysloop(Db_selector, Bt_selector, Key, Ref):-

key_next(Db_selector, Bt_selector, _),
bt_keysloop(Db_selector, Bt_selector, Key, Ref).

Displaying the Contents of a Database

You can use the predicate listdba, defined in the following program fragment, to
display the current state of an external database. listdba has one argument: the
selector of a database assumed to be open. All terms in the database must belong
Chapter 251, The External Database System 347
to the same domain. In the example, the domain is called mydom; when you use
this predicate, you must replace mydom with the actual name of the appropriate
domain in your program.
/* Program ch259e05.pro */

CONSTANTS
filename = "\\vip\\vpi\\programs\\register\\exe\\register.bin"

DOMAINS
db_selector = mydba
mydom = city(zipcode, cityname);
person(firstname, lastname, street, zipcode, code)
zipcode, cityname, firstname, lastname, street, code = string

PREDICATES
listdba(db_selector)
nondeterm bt_keys(db_selector,bt_selector,string,ref)
nondeterm bt_keysloop(db_selector,bt_selector,string,ref)

CLAUSES
listdba(Db_selector):-nl,
write("********************************************"),nl,
write(" DATABASE LISTING"),nl,
write("********************************************"),
db_statistics(Db_selector,NoOfTerms,MemSize,DbaSize,FreeSize),nl,nl,
write("Total number of records in the database: ",NoOfTerms),nl,
write("Number of bytes used in main memory: ",MemSize),nl,
write("Number of bytes used by the database: ",DbaSize),nl,
write("Number of bytes free on disk: ",FreeSize),nl,
fail.

listdba(Db_selector):-
db_chains(Db_selector,Chain),nl,nl,nl,nl,
write("******* Chain LISTING *************"),nl,nl,
write("Name=",Chain),nl,nl,
write("CONTENT OF: ",Chain),nl,
write("------------------------------\n"),
chain_terms(Db_selector, Chain, mydom,Term, Ref),
write("\n", Ref, ": ",Term),
fail.
 listdba(Db_selector):-
db_btrees(Db_selector,Btree), /* Returns each B+ tree name */
bt_open(Db_selector,Btree,Bt_selector),
bt_statistics(Db_selector,Bt_selector,NoOfKeys,
NoOfPages,Dept,KeyLen,Order,PageSize),nl,nl,nl,
write("******** INDEX LISTING **************"),nl,nl,
write("Name= ", Btree),nl,
write("NoOfKeys= ", NoOfKeys),nl,
write("NoOfPages=", NoOfPages),nl,
write("Dept= ", Dept),nl,
write("Order= ", Order),nl,
write("KeyLen= ", KeyLen),nl,
write("PageSize= ", PageSize), nl,
write("CONTENT OF: ", Btree),nl,
write("-----------------------------\n"),
bt_keys(Db_selector,Bt_selector,Key,Ref),
write("\n",Key, " - ",Ref),
fail.
listdba(_).

bt_keys(Db_selector,Bt_selector,Key, Ref):-
key_first(Db_selector,Bt_selector,_),
bt_keysloop(Db_selector,Bt_selector,Key,Ref).

bt_keysloop(Db_selector,Bt_selector,Key,Ref):-
key_current(Db_selector,Bt_selector,Key,Ref).

bt_keysloop(Db_selector,Bt_selector,Key,Ref):-
key_next(Db_selector,Bt_selector,_),
bt_keysloop(Db_selector,Bt_selector,Key,Ref).

GOAL
db_open(mydba,filename,in_file),
listdba(mydba).

Implementing a Database That Won't Break Down

If you enter a lot of new information into a database, it is important to ensure that
this information won't be lost if the system goes down. In this section, we
illustrate one way of doing this -- by logging all changes in another file.
Making a change involves first updating the database, and then flushing it. If this
operation succeeds, the system then records the change in the log file and flushes
the log file itself. This means that only one file is unsafe at any given time. If the
database file becomes invalid (because the system went down before the file was
flushed, for example), you should be able to reconstruct it by merging the log file
Chapter 251, The External Database System 349
with a backup of the database file. If the log file becomes invalid, you should
create a new log file and make a backup of the database file.
If you record the date and time in the log file, together with the old values from a
modification involving replacement or deletion, you should be able to reconstruct
the database to its state at a given time.
/* This program fragment goes with Program 6 */

DOMAINS
logdom = insert(relation,dbdom,ref);
replace(relation,dbdom,ref,dbdom);
erase(relation,ref,dbdom)

PREDICATES
logdbchange(logdom)

CLAUSES
logdbchange(Logterm):-
chain_insertz(logdba,logchain,logdom,Logterm,_),
db_flush(logdba).

Updating the Database

As a general principle, you shouldn't spread database updating throughout the
program but should keep it in some user-defined predicates. This makes it easier
to change the database and/or to add new B+ trees. When you confine updating
this way, it's also easier to make a robust database system because your program
involves only a small piece of code in which the database is unsafe.
The following example handles updating two different relations, whose objects
are all strings:
person(firstname, lastname, street, zipcode, code)

city(zipcode, cityname)

It handles the updating with the following indexes (keys) on the person and city
relations:
Person's Name............Last Name plus First Name
Person's Address.........Street Name
City Number..............Zip Code

In this example, we assume that the B+ trees are already open, and that their
bt_selectors have been asserted in the database predicate indices.
Before this program initiates the updating, it eliminates the possibility of a
BREAK with the break predicate. After updating is finished, the program flushes
the database with db_flush. Although db_flush makes the updating a slow
process (thanks to DOS), the file will be safe after this call.
To make the system as secure as possible, the program logs changes in a special
file through a call to logdbchange.
/* Program ch260e06.pro */

/* Logging database operations */

DOMAINS
logdom = insert(relation,dbdom,ref);
replace(relation,dbdom,ref,dbdom);
erase(relation,ref,dbdom)

PREDICATES
logdbchange(logdom)

CLAUSES
logdbchange(Logterm):-
chain_insertz(logdba,logchain,logdom,Logterm,_),
db_flush(logdba).

DOMAINS
dbdom = city(zipcode, cityname);
person(firstname, lastname, street, zipcode, code)
zipcode, cityname, firstname, lastname = string
street, code = string
indexName = person_name; person_adr; city_no
relation = city; person
db_selector = dba; logdba

DATABASE
% This takes and index name (a key) that is a person's name or address
%or a city number; it also takes a B+ tree selector
indices(IndexName, bt_selector)

PREDICATES
%and a first name (10 characters)
% This predicate creates an index name from a last name (20 characters)
xname(FirstName,LastName,string)

CLAUSES
xname(F,L,S):-
str_len(L,LEN),LEN>20,!,
frontstr(20,L,L1,_),
format(S,"%-20%",L1,F).

Chapter 251, The External Database System 351

 xname(F,L,S):-
format(S,"%-20%",L,F).

PREDICATES
ba_insert(relation, dbdom)
dba_replace(relation, dbdom, Ref)
dba_erase(relation, Ref)

CLAUSES
dba_insert(person,Term):-!,
break(OldBreak),
break(off),
indices(person_name,I1),
indices(person_adr,I2),!,
Term = person(Fname,Lname,Adr,_,_),
xname(Fname,Lname,Xname),
chain_insertz(dba,person,dbdom,Term,Ref),
key_insert(dba,I1,Xname,Ref),
key_insert(dba,I2,Adr,Ref),
db_flush(dba),
logdbchange(insert(person,Term,Ref)),
break(OldBreak).

dba_insert(city,Term):-
break(OldBreak),
break(off),
indices(city_no,I),!,
Term = city(ZipCode,_),
chain_insertz(dba,city,dbdom,Term,Ref),
key_insert(dba,I,ZipCode,Ref),
db_flush(dba),
logdbchange(insert(city,Term,Ref)),
break(OldBreak).
 dba_replace(person,NewTerm,Ref):-!,
break(OldBreak),
break(off),
indices(person_name,I1),
indices(person_adr,I2),!,
ref_term(dba,dbdom,Ref,OldTerm),
OldTerm=person(OldFname,OldLname,OldAdr,_,_),
xname(OldFname,OldLname,OldXname),
key_delete(dba,I1,OldXname,Ref),
key_delete(dba,I2,Oldadr,Ref),
NewTerm=person(NewFname,NewLname,NewAdr,_,_),
xname(NewFname,NewLname,NewXname),
term_replace(dba,dbdom,Ref,NewTerm),
key_insert(dba,I1,NewXname,Ref),
key_insert(dba,I2,NewAdr,Ref),
db_flush(dba),
logdbchange(replace(person,NewTerm,Ref,OldTerm)),
break(OldBreak).

dba_replace(city,NewTerm,Ref):-!,
break(OldBreak),
break(off),
indices(city_no,I),!,
ref_term(dba,dbdom,Ref,OldTerm),
OldTerm=city(OldZipCode,_),
key_delete(dba,I,OldZipCode,Ref),
NewTerm=city(ZipCode,_),
term_replace(dba,dbdom,Ref,NewTerm),
key_insert(dba,I,ZipCode,Ref),
db_flush(dba),
logdbchange(replace(city,NewTerm,Ref,OldTerm)),
break(OldBreak).

Chapter 251, The External Database System 353

 dba_erase(person,Ref):-!,
break(OldBreak),
break(off),
indices(person_name,I1),
indices(person_adr,I2),!,
ref_term(dba, dbdom, Ref, OldTerm),
OldTerm=person(OldFname,OldLname,OldAdr,_,_),
xname(OldFname,OldLname,OldXname),
key_delete(dba,I1,OldXname,Ref),
key_delete(dba,I2,OldAdr,Ref),
term_delete(dba,person,Ref),
db_flush(dba),
logdbchange(erase(person, Ref, OldTerm)),
break(OldBreak).

dba_erase(city,Ref):-
break(OldBreak),
break(off),
indices(city_no,I),!,
ref_term(dba,dbdom,Ref,OldTerm),
OldTerm=city(OldZipCode,_),
key_delete(dba,I,OldZipCode,Ref),
term_delete(dba,city,Ref),
db_flush(dba),
logdbchange(erase(city,Ref,OldTerm)),
break(OldBreak).

Using Internal B+ Tree Pointers

Each open B+ tree has an associated pointer to its nodes. When you open or
update the B+ tree, this pointer is positioned before the start of the tree. When you
call key_next with the pointer at the last key in the tree, the pointer will be
positioned after the end of the tree. Whenever the pointer moves outside the tree,
key_current fails. If this arrangement is not appropriate for a particular
application, you can model other predicates.
You can use mykey_next, mykey_prev , and mykey_search, defined in this
example, to ensure that the B+ tree pointer is always positioned inside the B+ tree
(provided there are any keys in the tree).
PREDICATES
mykey_next(db_selector, bt_selector, ref)
mykey_prev(db_selector, bt_selector, ref)
mykey_search(db_selector, bt_selector, string, ref)
 CLAUSES
mykey_prev(Dba, Bt_selector, Ref) :-
key_prev(Dba, Bt_selector, Ref), !.
mykey_prev(Dba, Bt_selector, Ref) :-
key_next(Dba, Bt_selector, Ref), fail.

mykey_next(Dba, Bt_selector, Ref) :-

key_next(Dba, Bt_selector, Ref), !.
mykey_next(Dba, Bt_selector, Ref) :-
key_prev(Dba, Bt_selector, Ref), fail.

mykey_search(Dba, Bt_selector, Key, Ref) :-

key_search(Dba, Bt_selector, Key, Ref), !.
mykey_search(Dba, Bt_selector, _, Ref) :-
key_current(Dba, Bt_selector, _, Ref), !.
mykey_search(Dba, Bt_selector, _, Ref) :-
key_last(Dba, Bt_selector, Ref).

You can use the samekey_next and samekey_prev predicates, defined in the next
example, to move the index pointer to the next identical key in a B+ tree that has
duplicate keys.
PREDICATES
samekey_next(db_selector, bt_selector, ref)
try_next(db_selector, bt_selector, ref, string)
samekey_prev(db_selector, bt_selector, ref)
try_prev(db_selector, bt_selector, ref, string)

CLAUSES
Samekey_next(Dba, Bt_selector, Ref) :-
key_current(Dba, Bt_selector, OldKey, _),
try_next(Dba, Bt_selector, Ref, OldKey).
try_next(Dba, Bt_selector, Ref, OldKey) :-
key_next(Dba, Bt_selector, Ref),
key_current(Dba, Bt_selector, NewKey, _),
NewKey = OldKey, !.

try_next(Dba, Bt_selector, _, _) :-
key_prev(Dba, Bt_selector, _),
fail.

samekey_prev(Dba, Bt_selector, Ref) :-

key_current(Dba, Bt_selector, OldKey, _),
try_prev(Dba, Bt_selector, Ref, OldKey).

Chapter 251, The External Database System 355

 try_prev(Dba, Bt_selector, Ref, OldKey) :-
key_prev(Dba, Bt_selector, Ref),
key_current(Dba, Bt_selector, NewKey, _),
NewKey = OldKey, !.

try_prev(Dba, Bt_selector, _, _) :-
key_next(Dba, Bt_selector, _),
fail.

Changing the Structure of a Database

One way to change the structure of a database is to write a small program that
copies the old database to a new one while making external databases, changing
structure of the changes. Another way, which we'll describe here, is to first dump
the database into a text file, make any necessary modifications to the database
with a text editor, and then read the modified database back into a new file.
You can use the predicate dumpDba, defined in the next program fragment, to
dump the contents of an external database into a text file if the database satisfies
the following conditions:
Every chain in the database models a relation.
All terms in the database belong to the same domain.
This method does not dump the B+ trees into the text file; we assume, given the
first condition, that B+ trees can be generated from the relations. In this example,
all terms belong to the generic domain mydom; when you implement this method,
replace mydom with the actual name and a proper declaration.
This code writes the contents of the database to a text file opened by outfile. Each
line of the text file contains a term and the name of the containing chain. The term
and the chain names are combined into the domain chainterm.
/* Program ch261e07.pro */

CONSTANTS
filename = "\\vip\\vpi\\programs\\register\\exe\\register.bin"

DOMAINS
Db_selector = myDba
chainterm = chain(string, mydom)
file = outfile
mydom = city(zipcode, cityname);
person(firstname, lastname, street, zipcode, code)
zipcode, cityname, firstname, lastname = string
street, code = string
PREDICATES
wr(chainterm)
dumpDba(string,string)

CLAUSES
wr(X):-
write(X),nl.

dumpDba(Db_selector,OutFile):-
db_open(myDba,Db_selector,in_file),
openwrite(outfile,OutFile),
writedevice(outfile),
db_chains(myDba,Chain),
chain_terms(myDba,Chain,mydom,Term,_),
wr(chain(Chain,Term)),
fail.

dumpDba(_,_):-
closefile(outfile),
db_close(myDba).

GOAL
dumpDba(filename,"register.txt").

Now, using your customized version of this code, you can generate the text file
by calling dumpDba, and you can reload the database by using readterm with the
chainterm domain. The predicate dba_insert, which we defined in "Updating the
Database" (page 350), takes care of the updating.
DOMAINS
chainterm = chain(string, dbdom)

PREDICATES
nondeterm repfile(file)
copyDba
loadDba(string)

CLAUSES
repfile(_).
repfile(File) :- not(eof(File)), repfile(File).

Chapter 251, The External Database System 357

 loadDba(OutFile) :-
openread(Prn_file, OutFile),
readdevice(Prn_file),
repfile(Prn_file),
readterm(Chainterm, chain(Chain, Term)),
write(Term), nl,
Dba_insert(Chain, Term),
fail.

loadDba(_) :-
closefile(Prn_file).

copyDba :-
createDba,
db_open(Dba, "register.bin", in_file),
open_indices,
loadDba("register.txt"),
db_close(Dba).

Filesharing and the External Database

Visual Prolog supports file-sharing the external database. This means that a file
can be opened by several users or processes simultaneously, which will be useful
if you are using the external database in a LAN-application or with one of the
multitasking platforms. UNIX developers should take note that Visual Prolog
uses advisory filelocking.
Visual Prolog provides the following file-sharing facilities:
opening an existing database with two different access modes and three
different sharing modes for optimal speed.
grouping database accesses in transactions to ensure consistency
predicates that make it possible to check whether other users have updated the
database.

Filesharing Domains
The two special domains which are used for file-sharing have the alternatives:

Domain Functors
accessmode = read; readwrite
denymode = denynone; denywrite; denyall
Opening the Database in Sharemode
In order to access the external database in share mode, you must open an already
existing database file with the four arity version of db_open, specifying
AccessMode and DenyMode.
If AccessMode is read the file will be opened as readonly, and any attempts to
update the file will result in an run-time error, if it is readwrite the file is opened
for both reading and writing. AccessMode is also used with the predicate
db_begintransaction.
If DenyMode is denynone all other users will be able to both update and read the
file, if it is denywrite, other users will not be able to open the file in accessmode =
readwrite, but you will be able to update the file providing it was opened in
accessmode = readwrite. If db_open is called with denymode = denyall no other
users will be able to access the file at all.
The first user that opens the file determines DenyMode for all subsequent
attempts to open the file, and a run-time error will occur if reopened in an
incompatible mode. The following table summarizes the results of opening and
subsequently attempting to reopen the same file for all combinations of
DenyMode and AccessMode:
2ND, 3RD, ..... REOPEN
Denyall Denywrite Denynone

R : AccessMode = read

RW: AccessMode = readwrite

Y : Open by 2ND, 3RD .. allowed

N : Open by 2ND, 3RD .. not allowed

Transactions and Filesharing

If a database file is opened in share mode, all database predicates that access the
database file in any way, must be grouped inside "transactions" this is done by
surrounding the calls to the predicates with db_begintransaction and
db_endtransaction.
Dependent on the combination of the chosen AccessMode and DenyMode the
shared file may be locked for the duration of the transaction. Again dependent on
the severity of the lock, other users may not be able to either read or update the
file, while your transaction takes place. This is of course necessary to avoid
conflicts between reading and writing, but if file-sharing is to have any meaning,
no excessive locking ought to take place. This can be avoided by keeping the
Chapter 251, The External Database System 359
transactions small (as short as possible) and only include those predicates that
access the database inside the transaction.
The concept of transactions in relation to file-sharing is very important. Two
often conflicting requirements, namely database consistency and a minimum of
file locking, must be fulfilled at the same time.
db_begintransaction ensures that database consistency is maintained and that an
appropriate locking of the file is effectuated. Several readers can access the file at
the same time, but only one process at the time is allowed to update the database.
The predicate db_setretry can be called to set for how long db_begintransaction
will wait to gain access to the file before returning with a run-time error. Calling
db_begintransaction with accessmode set to readwrite with a file opened with
accessmode set to read will also result in a run-time error. If db_begintransaction
is called, db_endtransaction must be called before a new call to
db_begintransaction for the same database, otherwise a run-time error will occur.
The following table summarizes the actions taken by db_begintransaction with
different combinations of AccessMode and DenyMode:
AccessMode
read readwrite
DenyMode- denynone WLock\Reload RWLock\Reload

Actions :

WLock : No write. Read allowed.

RWLOCK : No read or write allowed.
Reload : Reloading of file descriptors.

Since reloading and locking takes time, AccessMode and DenyMode should be
selected with care. If no users are going to update the database, set AccessMode to
read and DenyMode to denywrite for a minimal overhead.

Filesharing Predicates
In this section we discuss the file sharing predicates db_open,
db_begintransaction, db_endtransaction, db_updated, bt_updated, and
db_setretry.

db_open/4
This four arity version of db_open opens an existing database on file in share
mode.
db_open(Dbase, Name, AccessMode, DenyMode) /* (i,i,i,i) */
After creating an external database (in_file) with db_create it can be opened in
share mode, where Dbase is a db_selector, Name is the DOS-style file name,
AccessMode is read or readwrite, and DenyMode is denynone, denywrite, or
denyall.

db_begintransaction/2
db_begintransaction(Dbase, AccessMode) /* (i,i) */

This predicate marks the beginning of a transaction, and must be called prior to
any form of access to a database opened in share mode, even if opened with
denyall. In addition to the db_selector for the database, db_begintransaction
must be called with AccessMode bound to either read or readwrite.

db_endtransaction/1
db_endtransaction(Dbase) /* (i) */

db_endtransaction marks the end of a transaction and carries out the appropriate
unlocking of the database. A call of db_endtransaction without a prior call to
db_begintransaction for the db_selector Dbase will result in an run-time error.

db_updated/1
db_updated(Dbase) /* (i) */

If other users have updated the database, a call of db_begintransaction will

ensure that database consistency is maintained. Changes can be detected with the
predicate db_updated, which succeeds if called inside a transaction where
changes made by other users since your last call of db_begintransaction. If no
changes have been made, db_updated will fail. If called outside a transaction a
run-time error will occur.

bt_updated/2
bt_updated(Dbase,Btree_Sel) /* (i,i) */

Similar to db_updated/1, but only succeeds if the named B+ tree has been
updated.

db_setretry/3
db_setretry(Dbase,SleepPeriod,RetryCount) /* (i,i,i) */

Chapter 251, The External Database System 361

If access to a file is denied, because another process has locked the file, you can
have your process wait for a period of time and then try again. The predicate
db_setretry changes the default settings of SleepPeriod, which is the interval in
centiseconds between retries, and RetryCount, which is the maximum number of
times access will be attempted. The default settings are 100 for RetryCount and
10 for SleepPeriod.

Programming with Filesharing

Great care must be taken when using the file sharing predicates. Although they,
when used properly, ensure low-level consistency in a shared database, it is the
application programmers responsibility to provide the demanded high level
consistency for a given application. The term "transaction" is used here for a
group of file accesses, but it should be kept in mind that no back out facilities are
provided, and that program interruption caused by either software or hardware
failure, may cause inconsistencies in the database file.
When several processes share a database, special attention must also be paid to
the domains involved. It's crucial that they are identical and use identical
alignment.
To avoid unnecessary locking of the database file the transactions should be kept
fairly small, in order to ensure that the file will be locked for as short a time as
possible. At the same time it is important that predicates used to locate and access
an item in the database are grouped inside the same transaction:
.....
db_begintransaction(dba,readwrite),
key_current(dba,firstindex,Key,Ref),
ref_term(dba,string,Ref,Term),
db_endtransaction(dba),
write(Term),
.....

In this example the predicates key_current and ref_term should not be placed
inside different transactions, as the term stored under Ref may be deleted by
another user between transactions.
If a B+ tree is updated by another user and the file buffers are reloaded, the B+
tree will be repositioned before the first element of the tree. By calling the
predicate bt_updated you can detect when to reposition your B+ tree. It is still
possible to list the entire index and at the same time keep the transactions small,
by temporarily storing the current key in the internal database, as shown in the
following program fragment. It works under the assumption that no duplicate
keys exist.
DOMAINS
db_selector = dba

DATABASE
determ currentkey(string)

PREDICATES
list_keys(bt_selector)
list_index(bt_selector)
check_update(bt_selector,string)

CLAUSES
check_update(Index,Key):-
not(bt_updated(dba,Index)),!,
key_next(dba,Index,_).
check_update(Index,Key):-
key__search(dba,Index,Key,_),!. % Will fail if current was deleted
check_update(_,_). %by another user

list_keys(Index):-
currentkey(Key),
write(Key),nl,
db_begintransaction(dba,read),
check_update(Index,Key),
key_current(dba,Index,NextKey,_),
db_endtransaction(dba),!,
retract(currentkey(_)),
assert(currentkey(NextKey)),
list_keys(Index).
list_keys(_):-
db_endtransaction(dba).

list_index(Index):-
db_begintransaction(dba,read),
key_first(dba,Index,_),
key_current(dba,Index,Key,_),
db_endtransaction(dba),
retractall(currentkey(_)),
assert(currentkey(Key)),
list_keys(Index).
list_index(_).

key_search is used to reposition the B+ tree at the key that was listed previously.
The my_search predicate insures that the B+ tree will be correctly positioned
even if currentkey was deleted by another user.

Chapter 251, The External Database System 363

The example above also illustrates another important point. A db_endtransaction
must be used after each, and before the next, call of db_begintransaction. In the
predicate list_keys above, the listing stops when key_next fails, indicating that all
the keys have been listed. As db_begintransaction had to be called prior to
accessing the database, db_endtransaction has to be called as well after accessing
is completed. The second list_keys-clause ensures that db_endtransaction will be
called when key_next fails.

Implementing highlevel locking

The examples shown so far have illustrated some of the problems involved in file
sharing, and how they can be avoided.
You are allowed to do all the same operations on a shared database file as if you
were the only user with access to the file. Grouping the accesses to the file inside
db_begintransaction and db_endtransaction ensures that the Visual Prolog
system has consistency in its descriptor tables. But on a higher level you must
yourself ensure that the various logical constraints you have on your application
are conserved over a network with multiple users.
We call this high level locking or application level locking. By using the
primitives db_begintransaction and db_endtransaction you have many ways of
implementing a high level locking facility.
A common example of where high level locking is needed is in a database system
where a user wants to change a record. When he has decided that he wants to
change a record he should perform some kind of action so the application will
place a lock on that record until the user has finished the changes to the record so
the new record can be written back to disk, and the record unlocked.
Some suggestions for implementing an application-level lock of this type are:
Have a special field in that record to tell whether it is locked.
Have a special B+Tree or a chain where you store all references to all the
records that are locked by users.
Associated with a REF store a list of references to all records that are locked.

You might need to implement a kind of supervisor mechanism so a special user

can unlock locked records.
This was just an example, you might want to implement locking on a higher level
like tables or groups of tables, - or knowledge groups etc.
Note: If you want to delete a B+ tree in a database file opened in share mode, it is
up to you to ensure by high level locking that no other users have opened this B+
Tree. In the Visual Prolog system there is no check for a B+Tree selector being
no longer valid because the B+Tree has been deleted by another user.

A Complete Filesharing Example

In the following large example it will be shown how file sharing can be done
more easily by implementing your own locking system. If you manage your own
locks, needless file locking can be avoided, and other users won't have to wait for
access to the file because it is locked.
The example is the file sharing version of the previous 4 example. The program
lets several users create, edit, view and delete texts from a single shared file.
When creating and editing a text, it will be locked until editing is complete. Other
users cannot delete or edit a text while it is locked, but they will be able to view
the text. Run the program and experiment with different settings for db_open and
db_setretry.
/* Program ch262e08.pro */

DATABASE - indexes
determ lockindex(bt_selector)
determ index(bt_selector)
determ mark(real)

DOMAINS
my_dom = f(string)
db_selector = dba

PREDICATES
nondeterm repeat
wr_err(integer)

% List texts and their status

list
list_texts(bt_selector,bt_selector)
show_textname(string,bt_selector)

CLAUSES
show_textname(Key,LockIndex):-
key_search(dba,LockIndex,Key,_),!,
write("\n*",Key).
show_textname(Key,_):-
write("\n ",Key).

Chapter 251, The External Database System 365

 list_texts(Index,LockIndex) :-
key_current(dba,Index,Key,_),
show_textname(Key,LockIndex),
key_next(dba,Index,_),!,
list_texts(Index,LockIndex).
list_texts(_,_).

list:-nl,
write("***************** TEXTS (*=Locked) *******************"),nl,
index(Index),
lockindex(LockIndex),
key_first(dba,Index,_),!,
list_texts(Index, LockIndex),nl,
write("******************************************************"),nl.
list.

repeat.
repeat:-repeat.

wr_err(E):-
errormsg("PROLOG.ERR",E,Errormsg,_),
write(Errormsg),
readchar(_).

PREDICATES
%Logical locking of files
lock(string,bt_selector,bt_selector)

CLAUSES
lock(Name,Index,LockIndex):-
not(key_search(dba,LockIndex,Name,_)),!,
key_search(dba,Index,Name,Ref),
key_insert(dba, LockIndex, Name, Ref).
lock(Name,_,_):-
db_endtransaction(dba),
write(Name," is being updated by another user.\n Access denied"),
fail.

PREDICATES
ed(db_selector, bt_selector, bt_selector, string)
ed1(db_selector, bt_selector, bt_selector, string)

CLAUSES
% The ed predicates ensure that the edition will never fail.
ed(dba,Index,LockIndex,Name):-
ed1(dba,Index,LockIndex,Name),!.
ed(_,_,_,_).
 /* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
* There are two choices: *
* *
* 1) The name already exists - modify the contents of the *
* file *
* 2) The name is a new name - create a new file *
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * */
ed1(dba, Index,LockIndex, Name) :-
db_begintransaction(dba,readwrite),
key_search(dba, Index, Name, Ref),!,
ref_term(dba, string, Ref, Str),
lock(Name,Index,LockIndex),
list,
db_endtransaction(dba),nl,
write("******************************************************"),nl,
write("* EDIT ",Name," *"),nl,
write("******************************************************"),nl,
write(Str),nl,
write("< Press 'r' to replace this string ; else any key >"),nl,
readchar(X),X='r',nl,
write("Enter string and press <ENTER>"),nl,
readln(Str1),nl,
db_begintransaction(dba,readwrite),
term_replace(dba, string, Ref, Str1),
key_delete(dba, LockIndex, Name, Ref), %unlock
list,
db_endtransaction(dba).

%New file
ed1(dba, Index,LockIndex, Name):-
chain_insertz(dba, file_chain, string, "", Ref),
key_insert(dba, Index, Name, Ref),
list,
db_endtransaction(dba),
ed1(dba,Index,LockIndex, Name).

PREDICATES
main(db_selector, bt_selector, bt_selector)
interpret(char, bt_selector, bt_selector)
check_update_view
update_view
get_command(char)

Chapter 251, The External Database System 367

CLAUSES
% Loop until 'Q' is pressed
main(dba,Index,LockIndex) :-
check_update_view,
get_command(Command),
trap(interpret(Command,Index,LockIndex),E,wr_err(E)),!,
main(dba,Index,LockIndex).
main(_,_,_).

check_update_view:-
mark(T),timeout(T),!,
db_begintransaction(dba,read),
update_view,
db_endtransaction(dba),
marktime(100,Mark),
retractall(mark(_)),
assert(mark(Mark)).
check_update_view.

update_view:-nl,
write("******* COMMANDS E:Edit V:View D:DeleteQ:Quit *******"),nl,
write("COMMAND>"),
db_updated(dba),!,
list.
update_view.

get_command(Command):-
readchar(C),!,
upper_lower(Command,C),
write(Command),nl.
get_command(' ').
%interpret commandlineinput
interpret(' ',_,_):-!.
interpret('Q',_,_):-!,fail.
interpret('E',Index,LockIndex):-!,
write("\nFile Name: "),
readln(Name),nl,
ed(dba,Index,LockIndex,Name).
interpret('V',Index,_):-
write("\nFile Name: "),
readln(Name),nl,
db_begintransaction(dba,read),
key_search(dba,Index,Name,Ref),!,
ref_term(dba,string,Ref,Str),
db_endtransaction(dba),
write("******************************************************"),nl,
write("* VIEW ",Name," "),nl,
write("******************************************************"),nl,
write(Str),nl.

interpret('V',_,_):-!,
db_endtransaction(dba).
interpret('D',Index,_):-
write("\nDelete file: "),
readln(Name),nl,
db_begintransaction(dba,readwrite),
key_search(dba,Index,Name,Ref),!,
% not(key_search(dba,LockIndex,Name,_)),!,
key_delete(dba,Index,Name,Ref),
term_delete(dba,file_chain,Ref),
list,
db_endtransaction(dba).
interpret('D',_,_):-!,
db_endtransaction(dba).
nterpret(_,_,_):-beep.

PREDICATES
open_dbase(bt_selector,bt_selector)

CLAUSES
open_dbase(INDEX,LOCKINDEX):-
existfile("share.dba"),!,
db_open(dba, "share.dba",readwrite,denynone),
db_begintransaction(dba,readwrite),
bt_open(dba, "locks", LOCKINDEX),
bt_open(dba, "ndx", INDEX),
db_endtransaction(dba).

Chapter 251, The External Database System 369

 open_dbase(INDEX,LOCKINDEX):-
db_create(dba,"share.dba" , in_file),
bt_create(dba, "locks",TEMPLOCKINDEX,20, 4),
bt_create(dba, "ndx",TEMPINDEX , 20, 4),
bt_close(dba, TEMPINDEX),
bt_close(dba, TEMPLOCKINDEX),
db_close(dba),
open_dbase(INDEX,LOCKINDEX).

GOAL
open_dbase(INDEX,LOCKINDEX),
assert(index(INDEX)),
assert(lockindex(LOCKINDEX)),
marktime(10,Mark),
assert(mark(Mark)),
db_setretry(dba,5,20),
db_begintransaction(dba,read),
list,nl,
db_endtransaction(dba),
main(dba, INDEX,LOCKINDEX),
db_begintransaction(dba,read),
bt_close(dba, INDEX),
bt_close(dba, LOCKINDEX),
db_endtransaction(dba),
db_close(dba).

Implementation Aspects of Visual Prolog Filesharing

Filesharing in Visual Prolog is efficient and fast, because only the necessary parts
of the database file descriptors are loaded after an update by another user. As was
shown earlier in this chapter it is only under certain circumstances that any
reloading of file buffers and locking of files has to be done at all, and the complex
internal management of the database file ensures that after an update a minimum
of disk activity is needed.
The database has a serial number, which is a six byte integer, that is incremented
and written to disk each time an update occurs. The db_begintransaction
predicate compares the local copy of the serial number with the one on the disk,
and if they differ, the descriptors are reloaded. Locking is done in an array with
room for 256 readers. When a reader wishes to access the file, an unlocked space
is located in this lock array, and locked for the duration of the transaction. This
allows several readers to access the file simultaneously. If db_begintransaction is
called with AccessMode = readwrite, it will wait until all present readers have
unlocked their space, and then lock the entire array, allowing no other users to
access the file.

Miscellaneous

Finally, we have provided a couple of small predicates that are handy in special
circumstances. The predicate availableems will in DOS return the amount of
available ems. This can be used before a call to db_open or db_create in order to
see if there is enough space for placing the database in_ems.
availableems(Size) /* (real)-(o) */

Another predicate str_ref can be used to convert a database reference number to a

string so it can be inserted in a B+ Tree.
str_ref(Str,Ref) /* (string,ref)-(i,o)(o,i)(i,i) */

Summary

Visual Prolog's external database system adds power, speed, and efficiency to
your database applications. These are the major points covered in this chapter:
1. External database terms are stored in chains, which you can access directly
with database reference numbers; these reference numbers belong to the
predefined ref domain.
2. Individual databases are identified by a database selector, which belongs to the
standard domain db_selector.
3. You can store your external database in three locations, depending on which
alternative you use for the predefined place domain:
a. in_file places the database in a disk file
b. in_memory places it in memory
c. and in_ems places it in EMS-type expanded memory (same effect as
in_memory on non-DOS platforms
4. If you want to sort the terms in your database, you'll use B+ trees. Like
databases, individual B+ trees are identified by a B+ tree selector of the
standard domain bt_selector.

Chapter 251, The External Database System 371

5. Each entry in a B+ tree node consists of a key string (also known as an index),
which identifies a record, and the database reference number associated with
that record.
6. tree keys are grouped in pages, and the number of keys stored at a node is
specified by the tree's order.
7. Filesharing is done by grouping the predicates that access the database into
transactions.
CHAPTER 263

System-Level Programming
264Visual Prolog provides several predicates that allow you to access your PC's
operating system and - to the extent that operating system allows - the hardware
directly. We summarize those predicates in this chapter, first taking a look at the
ones giving access to the OS, then those that perform bit-level logical and shifting
operations. After that, we discuss a set of predicates that provide low-level
support for manipulating the DOS BIOS, memory, and other hardware elements.
We end this chapter with a couple of examples that demonstrate how to use some
of these predicates within a Visual Prolog application.

Access to the operating system

With a handful of predicates, you can access the operating system while running
the Visual Prolog integrated environment, as well as build the ability to access the
run-time computer's operating system right into your Visual Prolog applications.
You can execute any external program with a call to system, call the date and
time facilities with date and time, investigate the environment table with
envsymbol, and read the command-line arguments with comline. Furthermore,
you can establish the start-up directory and exe-filename of the program by
calling syspath, and the marktime, the timeout and the sleep predicates provide
time-tunneling capacity. Then there's the inevitable sound and beep predicates,
and finally osversion returning the operating system version, diskspace returning
the amount of free disk space, and three versions of storage used to determine
memory used.
This section describes each of these predicates in detail and provides some
practical examples that demonstrate how to use them.

system/1
Visual Prolog programs provide access to the OS through the system predicate,
which takes the following form:
system("command") /* (i) */

Chapter 265, System-Level Programming 373

If the argument is an empty string (""), a new command interpreter will be run in
interactive mode.

Examples
1. To copy the file B:ORIGINAL.FIL to a file A:NEWCOPY.FIL from within
the Visual Prolog system, you could give the goal
system("").

then copy the file using the usual command,

copy b:original.fil newcopy.fil

You could then return to Visual Prolog by typing

exit

after which you are back in your program again.

2. To rename the file (without going out to the OS), you could give the command
system("ren newcopy.fil newcopy.txt").

system/3
This extended version of the system predicate provides two extra features: one for
returning the OS error level, and one for resetting the run-time system's video
mode. The latter has no effect in OS/2 or Windows. In UNIX, this argument is
used to indicate that the process has no interaction with the terminal, and hence
that there's no need to clear and reset it. This is a somewhat unfortunate dual use
of the same argument, but it fulfills the typical needs of users.
system/3 takes this format:
system(CommandString, ResetVideo, ErrorLevel) /* (i,i,o) */

The error level is returned in ErrorLevel. This is the program return code known
by the OS at the time control returns to the program issuing the system call. In
DOS and OS/2 this is only available for .COM and .EXE files.
In textmode DOS, ResetVideo controls whether your program should reset the
video hardware to the state it was in before system/3 was called. ResetVideo = 1
resets the video mode; ResetVideo = 0 does not. When ResetVideo = 0, your
program will run in the new video mode you set, even if that's a mode not
specifically supported by Visual Prolog. (For information about setting the run-
time system's video mode, refer to the reference manual for the video hardware.)
In other words, if your external program MYSETMD sets the video hardware to a
mode not specifically supported by Visual Prolog, and you place the following
calls to system in your Visual Prolog program (running from the development
environment), you can actually make your program run in that unsupported mode:
system("mysetmd", 0, ErrorLevel),

Note: The external program must be compatible with the hardware at least at the
BIOS level (updating the BIOS variables rows and columns on-screen).

envsymbol/2
The envsymbol predicate searches for environment symbols in the application's
environment table; the SET (OS) commands set these symbols. envsymbol takes
this format:
envsymbol(EnvSymb, Value) /* (i,o) */

For example, the command

SET SYSDIR=C:\FOOL

sets the symbol SYSDIR to the string C:\FOOL, and the goal
/*...*/
envsymbol("SYSDIR", SysDir),
/*...*/

searches the environment for the symbol SYSDIR, binding SetValue to C:\FOOL.
envsymbol will fail if the symbol does not exist.

date and time

Visual Prolog has three more handy OS-related standard predicates: two forms of
date and time. The date/3 and time/3 predicates can be used in two ways,
depending on whether their parameters are free or bound on entry.
With input flow, time and date will set the internal system clock to the time
specified (in UNIX you need root privileges to do this). If all variables are free,
the system will bind them to the internal clock's current values.
time(Hours, Minutes, Seconds, Hundredths)
/* (i,i,i,i), (o,o,o,o) */

Note that the UNIX version of time doesn't return anything useful in the
Hundredths argument.

Chapter 265, System-Level Programming 375

date/3 also relies on the internal system clock and operates similarly to time; it
takes the following form:
date(Year, Month, Day) /* (i,i,i), (o,o,o) */

date/4 only has an output flow version. The fourth argument is the weekday
number, but what numbering scheme is used is operating system dependent.
However, it's fairly common that 0 is Sunday, 1 is Monday, etc.
date(Year, Month, Day, WeekDay) /* (o,o,o,o) */

Example
Program 2 uses time to display the time elapsed during a listing of the default
directory.
/* Program ch266e02.pro */

GOAL
time(H1,M1,S1,_),nl,
write("Start time is: ",H1,":",M1,":",S1),nl,
/* This is the activity that is being timed */
system("dir /s/b c:\\*.*"),
time(H2,M2,S2,_),
Time = S2-S1 + 60*(M2-M1 + 60*(H2-H1)),
write("Elapsed time: ",Time," seconds"),nl,
time(H3,M3,S3,_),
write("The time now is: ",H3,":",M3,":",S3).

comline/1
comline reads the command-line parameters used when invoking a program; this
is its format:
comline(CommandLine) /* (o) */

where CommandLine is a string.

syspath/2
syspath returns the start-up directory and name of the program calling it. syspath
looks as follows:
syspath(HomeDir,ExeName) /* (o,o) */
The main use for syspath is to provide programs the possibility of loading files
from their home directory, as well as constructing helpful command-line error
messages: <progname>: Usage: [-foul] <blah> <blah> <blah>.
On UNIX, the start-up directory is not directly available to a program. In order to
use syspath on UNIX, an initialization predicate, initsyspath, must be called. In
particular, this must be called before the program changes its working directory,
if this becomes necessary. If initsyspath isn't called, syspath will exit with an
error code of 1138.

Timing Services
Visual Prolog provides two different timing services: execution suspension, and
elapsed-time testing. Some special notes apply to UNIX, see the description of
difftime below.

sleep/1
sleep suspends program execution for a specified length of time. It looks like this
sleep(CSecs) /* (i) */

where Csecs is the time (in centiseconds, i.e. 1/100ths) to sleep. The exact length
of time the program will wait may vary, depending on CPU / OS activity, and you
shouldn't expect greater accuracy than 20-50 milliseconds.
In UNIX, sleep uses the nap(S) system call for delays and fractions of delays less
than 1 second. This call uses the kernel's callout table, and it may be necessary to
increase the size of this (kernel parameter NCALL) to prevent overflows if more
than 10-20 processes simultaneously use sleep with fractional delays or nap(S).

marktime/2
marktime returns a time-stamp which may later be tested for expiration using the
timeout predicate. marktime has the following format:
marktime(CSecs,Ticket) /* (i,o) */

where CSecs is the required length of time Ticket should last. The Ticket is an
implementation-defined structure holding the timing information, currently
masquerading as a real number.

timeout/1
timeout tests a time-ticket returned by marktime for expiration. If it has expired,
timeout succeeds, otherwise it fails. timeout looks like this:
Chapter 265, System-Level Programming 377
 timeout(Ticket) /* (i) */

As with sleep, don't expect too fine a granularity.

difftime
On UNIX, the standard predicate time doesn't provide a resolution in 100ths, so
any timing calculations will be rather rough. However, the UNIX version of
Visual Prolog has a standard predicate difftime:
difftime(real,real,real) /* (i,i,o) */

which returns the difference between the 1st and the 2nd timemark, in hundredths
of seconds as a floating-point number. The first timemark should be the younger,
and the second the older, i.e. the usage is
marktime(0,M1), lengthy_process, marktime(0,M2),
difftime(M2,M1,Diff).

In order for marktime and difftime to work, they must know how many clock-
ticks the machine has per second. For UNIX executables, they establish this by
calling the sysconf library function (see sysconf(S)), which is a very safe
mechanism. However, for XENIX executables they have to call the library
function gethz (see gethz(S)), which in it's current implementation simply
examines a shell variable called HZ. Thus it is critical that this variable has the
correct value, which, unless it's a whole new world when you read this, is 60. If
gethz fails (e.g. because HZ doesn't exist), marktime will exit with error 1136.
The same applies to difftime if either marktime has never been called, or if
marktime exited due to failure in gethz.
The granularity of sleep and the marktime and timeout predicates is system-
defined, currently being 1/60th of a second. Note that timemarks do not survive
reboots. Under UNIX they're the number of machine clock-ticks since "an
arbitrary point in the past" which in practice means system start-up. With 60
ticks/second, this also means that the tick count wraps around zero after approx.
2.26 years.

Example
Program 4 below demonstrates marktime and timeout.
/* Program ch267e04.pro */

PREDICATES
ttimeout(real)
CLAUSES
ttimeout(TM):-timeout(TM),!.
ttimeout(TM):-
write("No timeout, sleep 0.5 secs"),nl,
sleep(50),
ttimeout(TM).

GOAL
marktime(400,TM), % 4 secs
ttimeout(TM),
write("\nBINGO!\n").

sound/2
sound generates a sound in the PC's speaker:
sound(Duration,Frequency) /* (i,i) */

where Duration is the time in 1/100ths of a second.

On UNIX, sound works only on the ANSI console; whether you're running on
this is established by examining the TERM shell variable. On other terminals,
sound is equivalent to beep.

beep/0
beep /* (no arguments) */

In the DOS-related versions of Visual Prolog, beep is equivalent to

sound(50,1000).
On UNIX, beep writes the bell character to the file used for terminal output. If the
program is in terminal mode, all buffering will be bypassed.

osversion/1
osversion returns the current operating system version and looks like this:
osversion(VerString) /* (o) */

The format for VerString is operating system defined. For DOS and OS/2, it
consists of the major and minor version numbers, separated by a dot (full stop),
e.g. "3.30". Note that the major version number currently returned by OS/2 is 10,
rather than 1. In UNIX, the string contains the information returned by uname(S).

Chapter 265, System-Level Programming 379

diskspace/2
diskspace returns as an unsigned long the available disk space, using the
following format:
diskspace(Where,Space) /* (i,o) */

The space is reported in bytes.

In the DOS-related versions of Visual Prolog, Where should be a character
specifying the drive letter. In the UNIX version, it should be the name of a file
residing on the file system you want to query (see statfs(S)). You may use simply
"/" for the root file system, or an empty or null string in which case information is
retrieved for the file system holding the current working directory. The space
reported will be the smaller of the actual space and the ulimit for the process (see
ulimit(S)).

storage/3
The standard predicate storage returns information about the three run-time
memory areas used by the system (stack, heap, and trail, respectively) as unsigned
longs:
storage(StackSize,HeapSize,TrailSize) /* (o,o,o) */

The values are all in bytes.

In all versions of Visual Prolog, TrailSize contains the amount of memory used
by the trail.
In the DOS-related versions, StackSize indicates how much stack space is left. In
UNIX, StackSize is the exact opposite, namely how much stack that's been used
so far.
Finally, the HeapSize shows how much memory is available to the process.
In UNIX this is the difference between the current break value and the maximum
possible break value (see ulimit(S) and brk(S)), which again is set by the kernel
parameter MAXUMEM. It does not include memory held in freelists in the heap.
In DOS, the HeapSize is the unallocated physical memory between the top of the
GStack and the bottom of the heap. It does not include memory held in freelists in
the heap. The storage predicate returns the size that you can be sure of having
when you're loading a file or going out to the operating system.
In OS/2 the storage will be limited by the virtual memory which is limited by the
size of the disk where the swap file resides. Since there is no easy way to find
where the swap file resides, the storage predicate will for the heap just return the
size of the largest block OS/2 can allocate without compacting and/or swapping
memory. You can't really use that for anything, but it gives you an indication of
memory fragmentation. If you feel certain that you know where the swap file
resides, you can use the diskspace standard predicate to check the free space
available.

storage/0
The 0-arity version of storage is primarily intended for debugging purposes. It
prints in the current window an overview of the amount of memory in use by the
different parts of Visual Prolog's memory management, as well as the number of
backtrack points.

Bit-Level Operations

Visual Prolog provides six predicates for bit-level operations; bitor, bitand,
bitnot, bitxor, bitleft, and bitright. These predicates have one flow variant each,
operate on unsigned integers, and must be used in prefix notation.
bitnot/2
bitnot performs a bit-wise logical NOT.
bitnot(X, Z) /* (i,o) */

With X bound to some integral value, Z will be bound to the bit-wise negation of
X.

Operator X Z
bitnot 1 0
0 1

bitand/3
bitand performs a bit-wise AND.
bitand(X, Y, Z) /* (i,i,o) */

With X and Y bound to some integral values, Z will be bound to the result of bit-
wise ANDing the corresponding bits of X and Y.

Chapter 265, System-Level Programming 381

 Operator X Y Z
bitand 1 1 1
1 0 0
0 1 0
0 0 0

bitor/3
bitor performs a bit-wise OR.
bitor(X, Y, Z) /* (i,i,o) */

With X and Y bound to some integral values, Z will be bound to the result of bit-
wise ORing the corresponding bits of X and Y.

Operator X Y Z
bitor 1 1 1
1 0 1
0 1 1
0 0 0

bitxor/3
bitxor performs a bit-wise XOR.
bitxor(X, Y, Z) /* (i,i,o) */

With X and Y bound to some integral values, Z will be bound to the result of bit-
wise XORing the corresponding bits of X and Y.

Operator X Y Z
bitxor 1 1 0
1 0 1
0 1 1
0 0 0

bitleft/3
bitleft performs a bit-wise left shift.
 bitleft(X, N, Y) /* (i,i,o) */

With X and N are bound to some integral values, Y is bound to the result of
shifting the bit-wise representation of X N places to the left. The new bits will be
zero-filled.

bitright/3
bitright performs a bit-wise right shift.
bitright(X, N, Y) /* (i,i,o) */

With X and N are bound to some integral values, Y is bound to the result of
shifting the bit-wise representation of X N places to the right. The new bits will be
zero-filled.

Exercise
Write a Visual Prolog program to test the theory that
myxor(A, B, Result) :-
bitnot(B, NotB), bitand(A, NotB, AandNotB),
bitnot(A, NotA), bitand(NotA, B, NotAandB),
bitor(AandNotB, NotAandB, Result).
behaves like
bitxor(A, B, Result)

Access to the Hardware: Low-Level Support

The DOS ROM-BIOS (Read Only Memory-Basic Input/Output System) provides

an interface between programs and the operating system to perform various
functions, including disk, file, printer, and screen I/O. For specific information on
the ROM-BIOS, refer to the DOS Technical Reference Manual.
Visual Prolog provides six built-in predicates that give low-level access to the
operating system, I/O ports, and hardware. These predicates are bios (2 versions),
ptr_dword, memword, membyte, and port_byte.
This section describes each of these predicates in detail and provides some
practical examples that demonstrate how to use them.

Chapter 265, System-Level Programming 383

bios/3 and bios/4
bios gives access to the PC's low-level BIOS (Basic I/O System) routines. For
information about these routines, refer to your DOS Reference Manual. Note that
the bios predicates only relate to DOS. Under UNIX, it's possible to access
routines in shared libraries using the nlistnlist library call (see nlist(S)). However,
the process is rather involved and won't be described here. See NLIST.PRO in the
PROGRAMS directory for an example.
Information passes to and from the BIOS functions through the predefined
compound object reg(...). The bios predicate takes the following forms:
bios(InterruptNo, RegistersIn, RegistersOut) /* (i,i,o) */
bios(InterruptNo, RegistersIn, RegistersOut, OutFlags)
/* (i,i,o,o) */

where RegistersIn and RegistersOut are data structures defined as follows:

/* RegistersIn */
reg(AXi, BXi, CXi, DXi, SIi, DIi, DSi, ESi)
/* (i,i,i,i,i,i,i,i) */

/* RegistersOut */
reg(AXo, BXo, CXo, DXo, SIo, DIo, DSo, ESo)
/* (o,o,o,o,o,o,o,o) */

The bios predicates use the arguments

AXi, BXi, CXi, DXi, SIi, DIi, DSi, and ESi to represent the PC's hardware register
values passed to the BIOS.
AXo, ... , ESo for those register values returned by the BIOS.
The flow pattern for bios/3 is (i,i,o); for bios/4 it's (i,i,o,o). When you make a call
to the BIOS, each argument of RegistersIn must be instantiated (bound to a
value), and each argument of RegistersOut must be free (not bound to a value).
The domain for the RegistersIn and RegistersOut compound objects (reg(AX,
BX, ...)) is the reg domain, a predefined data structure created by Visual Prolog
specifically for the bios predicate. Internally, Visual Prolog defines this data
structure as
DOMAINS
reg = reg(integer, integer, integer, ..., integer)

The optional OutFlag argument in the bios/4 predicate is packed coding for the
8086 flag register (see Figure 268.1). OutFlag allows you to read the contents of
the status flags after return from the interrupt. The flags are packed in an integer
value as shown here:
 Figure 269.2: Packing the 8086 Flag Register in an Integer2703

_________________________________________________________________
| | | | | | | | | | | | | | | | |
| U | U | U | U | O | D | I | T | S | Z | U | A | U | P | U | C |
|___|___|___|___|___|___|___|___|___|___|___|___|___|___|___|___|
| | | | | | | | | | | | | | | | |
|15 |14 |13 |12 |11 |10 | 9 | 8 | 7 | 6 | 5 | 4 | 3 | 2 | 1 | 0 |
|___|___|___|___|___|___|___|___|___|___|___|___|___|___|___|___|

Code Flag Flag's Purpose

U Undefined; Not used
indeterminate value
O Overflow flag Indicates arithmetic overflow
D Direction flag Controls left/right direction in repeated
operations
I Interrupt enable flag Enables interrupts (when set)
T Trap flag Generates a trap at end of each
instruction (for trace)
S Sign flag Indicates negative result or comparison
if set
Z Zero flag Indicates zero result or equal
comparison
A Auxiliary flag Need adjustment in BCD (Binary
Coded Decimal) operations
P Parity flag Indicates even number of bits set
C Carry flag Indicates arithmetic carry out bit

ptr_dword
ptr_dword returns the internal address of StringVar, or creates the string ("the
char pointer") StringVar based on the supplied address.
ptr_dword(StringVar, Seg, Off) /* (o,i,i), (i,o,o) */

Chapter 265, System-Level Programming 385

When StringVar is bound, ptr_dword returns the internal segment and offset for
the string. When Seg and Off are bound, ptr_dword binds StringVar to the string
stored at that location. On 32-bit platforms the segment is ignored. ptr_dword has
to a considerable extent been superseded by the cast function.
A string in Visual Prolog is a series of ASCII values terminated by a zero value.
You can use the low-level routines in this chapter on abnormal strings (those that
contain several zero bytes). However, you can't write abnormal strings out or
assert them in the database.

membyte, memword, memdword

Visual Prolog provides three predicates for looking at (peeking) and modifying
(poking) specific elements in memory. membyte, memword and memdword
access byte, word and dword sized locations respectively. All predicates have two
formats:
membyte(Segment, Offset, Byte) /* (i,i,i), (i,i,o) */
memword(Segment, Offset, Word) /* (i,i,i), (i,i,o) */
memdword(Segment, Offset, DWord) /* (i,i,i), (i,i,o) */

and
membyte(StringVar, Byte) /* (i,i), (i,o) */
memword(StringVar, Word) /* (i,i), (i,o) */
memdword(StringVar, DWord) /* (i,i), (i,o) */

The Segment is an ushort, the Offset is an unsigned, and Byte, Word and DWord
are byte, word and dword respectively. Many of the bios calls require pointers to
be passed as Segment:Offset pairs. membyte and memword also require pointers
in this format. In realmode DOS, Memory locations are calculated as ((Segment )
16) + Offset).
The mem* predicates have to a large extent been superseded by the get*entry and
set*entry predicates for the binary datatype.

port_byte/2
The port_byte predicate allows you to read or write a byte to a specific I/O port.
The DOS format for port_byte is
port_byte(PortAddress, Byte) /* (i,i), (i,o) */

where PortAddress and Byte are defined as unsigneds. If you don't know what to
use port_byte for, don't worry and don't think about using it. It's intended for
access to (custom) hardware using ports for I/O.
Summary

These are the major points covered in this chapter:

1. Visual Prolog includes several predicates that
a. give access to the OS
b. perform bit-level logical and shifting operations
c. provide low-level support for manipulating the BIOS, memory, and other
hardware elements
2. These are the predicates giving access to the OS:
a. system (execute external program)
b. time (get or set the system clock)
c. date (get or set the internal calendar)
d. envsymbol (look up an environment variable)
e. comline (read the command-line arguments)
f. syspath (return start-up directory and name of .EXE file)
g. osversion (returns operating system version number)
h. diskspace (returns disk space available)
3. These are the predicates that perform bit-level operations:
a. bitor (bit-wise OR)
b. bitand (bit-wise AND)
c. bitnot (bit-wise NOT)
d. bitxor (bit-wise XOR)
e. bitleft (bit-wise LEFT SHIFT)
f. bitright (bit-wise RIGHT SHIFT)
4. These are the predicates that provide low-level support for various hardware
elements:
a. bios (accesses the PC's low-level BIOS routines)
b. ptr_dword (returns the internal address of its argument or places the
argument at a specified memory location)
c. membyte (peeks or pokes a one-byte value)
Chapter 265, System-Level Programming 387
d. memword (peeks or pokes a two-byte value)
e. memdword (peeks or pokes a four-byte value)
f. port_byte (reads or writes a byte to a specific I/O port)
CHAPTER 271

Example Prolog Programs

272In this final tutorial chapter, we present some small example programs
intended to stimulate your own ideas and to further illustrate the topics covered in
the earlier tutorial chapters. Nearly all of the examples offer plenty of room for
expansion; your own ideas can grow into full-blown programs using one of these
programs as a starting point.

Building a Small Expert System

In this first example, we show you how to construct a small expert system expert
system, sample that figures out which of seven animals (if any) the system's user
has in mind. The expert system will figure out the animal by asking questions
then making deductions from the replies given. This example demonstrates back-
tracking--using facts--and how to use not effectively.
A typical user dialogue with this expert system might be:
has it hair?
yes
does it eat meat?
yes
has it a fawn color?
yes
has it dark spots?
yes

Your animal may be a cheetah!

Visual Prolog's ability to check facts and rules will provide your program with the
reasoning capabilities germane to an expert system. The first step is to provide the
knowledge with which the system can reason; this is known as the inference
engine and is shown in ch273e01.pro.

Chapter 274, Example Prolog Programs 389

 /* Program ch275e01.pro */

DATABASE
xpositive(symbol,symbol)
xnegative(symbol,symbol)

PREDICATES
nondeterm animal_is(symbol)
nondeterm it_is(symbol)
ask(symbol,symbol,symbol)
remember(symbol,symbol,symbol)
positive(symbol,symbol)
negative(symbol,symbol)
clear_facts
run

CLAUSES
animal_is(cheetah):-
it_is(mammal),
it_is(carnivore),
positive(has,tawny_color),
positive(has,dark_spots).

animal_is(tiger):-
it_is(mammal),
it_is(carnivore),
positive(has, tawny_color),
positive(has, black_stripes).

animal_is(giraffe):-
it_is(ungulate),
positive(has,long_neck),
positive(has,long_legs),
positive(has, dark_spots).

animal_is(zebra):-
it_is(ungulate),
positive(has,black_stripes).

animal_is(ostrich):-
it_is(bird),
negative(does,fly),
positive(has,long_neck),
positive(has,long_legs),
positive(has, black_and_white_color).
 animal_is(penguin):-
it_is(bird),
negative(does,fly),
positive(does,swim),
positive(has,black_and_white_color).

animal_is(albatross):-
it_is(bird),positive(does,fly_well).

it_is(mammal):-
positive(has,hair).
it_is(mammal):-
positive(does,give_milk).

it_is(bird):-
positive(has,feathers).
it_is(bird):-
positive(does,fly),
positive(does,lay_eggs).

it_is(carnivore):-
positive(does,eat_meat).

it_is(carnivore):-
positive(has,pointed_teeth),
positive(has, claws),
positive(has,forward_eyes).

it_is(ungulate):-
it_is(mammal),
positive(has,hooves).

it_is(ungulate):-
it_is(mammal),
positive(does,chew_cud).

positive(X,Y):-
xpositive(X,Y),!.
positive(X,Y):-
not(xnegative(X,Y)),
ask(X,Y,yes).

negative(X,Y):-
xnegative(X,Y),!.
negative(X,Y):-
not(xpositive(X,Y)),
ask(X,Y,no).

Chapter 274, Example Prolog Programs 391

 ask(X,Y,yes):-
!,
write(X," it ",Y,'\n'),
readln(Reply),nl,
frontchar(Reply,'y',_),
remember(X,Y,yes).
ask(X,Y,no):-
!,
write(X," it ",Y,'\n'),
readln(Reply),nl,
frontchar(Reply,'n',_),
remember(X,Y,no).

remember(X,Y,yes):-
assertz(xpositive(X,Y)).
remember(X,Y,no):-
assertz(xnegative(X,Y)).

clear_facts:-
write("\n\nPlease press the space bar to exit\n"),
retractall(_,dbasedom),readchar(_).

run:-
animal_is(X),!,
write("\nYour animal may be a (an) ",X),
nl,nl,clear_facts.
run :-
write("\nUnable to determine what"),
write("your animal is.\n\n"),
clear_facts.

GOAL
run.

Each animal is described by a number of attributes that it has (or has not). Those
questions that the user is to reply to are the positive(X,Y) and negative(X,Y)
ones. The system, therefore, might ask something like this:
Does it have hair?

Having received a reply to such a question, you want to be able to add the answer
to the database, so the system will be able to use the previously gathered
information when reasoning.*
For simplicity, this example program will only consider positive and negative
replies, so it uses a database containing two predicates:
 DATABASE
xpositive(symbol, symbol)
xnegative(symbol, symbol)

The fact that the animal doesn't have hair is represented by

xnegative(has,hair).

The rules of positive and negative then check to see if the answer is already
known, before asking the user. askable
positive(X,Y) :-
xpositive(X,Y), !.
positive(X,Y) :-
not(xnegative(X,Y)),
ask(X,Y,yes).

negative(X,Y) :-
xnegative(X,Y), !.
negative(X,Y) :-
not(xpositive(X,Y)),
ask(X,Y,no).

Notice that the second rule for both positive and negative ensures that a
contradiction won't arise before asking the user.
The ask predicate asks the questions and organizes the remembered replies. If a
reply begins with the letter y, the system assumes the answer is Yes; if it begins
with n, the answer is No.
/* Asking Questions and Remembering Answers */

ask(X, Y, yes) :- !, write(X, " it ", Y, '\n'),

readln(Reply),
frontchar(Reply, 'y', _),
remember(X, Y, yes).

ask(X, Y, no) :- !, write(X, " it ", Y, '\n'),

readln(Reply),
frontchar(Reply, 'n', _),
remember(X, Y, no).

remember(X, Y, yes) :- assertz(xpositive(X, Y)).

remember(X, Y, no) :- assertz(xnegative(X, Y)).
/* Clearing Out Old Facts */

clear_facts :- write("\n\nPlease press the space bar to exit\n"),

retractall(_,dbasedom), readchar(_).
Chapter 274, Example Prolog Programs 393
For practice, type in the preceding inference engine and knowledge clauses. Add
appropriate declarations to make a complete program, and then try out the result.
The completed animal expert system is provided as ch276e01.pro.
An example expert systems shell (GENI.PRO) is also provided with Visual
Prolog in the PROGRAMS directory; this shell is based on the same techniques
introduced in this example, with the added feature that it allows you to
dynamically change the rules.

Prototyping: A Simple Routing Problem

Suppose you want to construct a computer system to help decide the best route
between two U.S. cities. You could first use Visual Prolog to build a miniature
version of the system (see 2), since it will then become easier to investigate and
explore different ways of solving the problems involved. You will use the final
system to investigate questions such as:
Is there a direct road from one particular town to another?
Which towns are situated less than ten miles from a particular town?
The following program is a classic example of using backtracking and recursion
to solve route planning.
/* Program ch277e02.pro */

DOMAINS
town = symbol
distance = integer

PREDICATES
nondeterm road(town,town,distance)
nondeterm route(town,town,distance)

CLAUSES
road(tampa,houston,200).
road(gordon,tampa,300).
road(houston,gordon,100).
road(houston,kansas_city,120).
road(gordon,kansas_city,130).
 route(Town1,Town2,Distance):-
road(Town1,Town2,Distance).
route(Town1,Town2,Distance):-
road(Town1,X,Dist1),
route(X,Town2,Dist2),
Distance=Dist1+Dist2,
!.

Figure 278.1 shows a simplified map for the prototype.

Figure 279.2: Prototype Map2803

Each clause for the road predicate is a fact that describes a road of a certain
length (in miles) that goes from one town to another.
The route clauses indicate that it is possible to make a route from one town to
another over several stretches of road. Following the route, the driver travels a
distance given by the third parameter, distance.
The route predicate is defined recursively; a route can simply consist of one
single stretch of road, as in the first clause. In this case, the total distance is
merely the length of the road.
You can also construct a route from Town1 to Town2 by driving first from Town1
to X, then following some other route from X to Town2. The total distance is the
sum of the distance from Town1 to X and the distance from X to Town2, as shown
in the second clause for route.
Try the program with the goal:
route(tampa, kansas_city, X).

Chapter 274, Example Prolog Programs 395

Can the program handle all possible combinations of starting point and
destination? If not, can you modify the program to avoid any omissions?
The next example will give you ideas about how to get this routing program to
make a list of towns visited enroute. Making such a list prevents Visual Prolog
from choosing a route that involves visiting the same town twice, thereby
avoiding going around in circles, and ensures that the program doesn't go into an
infinite loop. When you've solved problems of this type, you can enlarge the
program by adding more cities and roads.

Adventures in a Dangerous Cave

You're an adventurer, and you've heard that there is a vast gold treasure hidden
inside a cave. Many people before you have tried to find it, but to no avail. The
cave is a labyrinth of galleries connecting different rooms in which there are
dangerous beings, like monsters and robbers. In your favor is the fact that the
treasure is all in one room. Which route should you follow to get to the treasure
and escape unhurt with it? Consider the following map of the cave:

Figure 281.4: The Labyrinth of Galleries282

You can construct a Visual Prolog representation of the map to help you find a
safe route. Each gallery is described by a fact. The predicates go and route give
rules. Give the program the goal
go(entry, exit).
The answer will consist of a list of the rooms you should visit to capture the
treasure and return safely with it.
An important design feature of this program is that the rooms already visited are
collected in a catalog. This happens thanks to the route predicate, which is
defined recursively. If you're standing in the exit room, the third parameter in the
route predicate will be a list of the rooms you've already visited. If the
gold_treasure room is a member of this list, you'll have achieved your aim.
Otherwise, the list of rooms visited is enlarged with Nextroom, provided
Nextroom is not one of the dangerous rooms and has not been visited before.
/* Program ch283e03.pro */

DOMAINS
room = symbol
roomlist = room*

PREDICATES
nondeterm gallery(room,room)
% There is a gallery between two rooms
% Necessary because it does not matter
% which direction you go along a gallery
nondeterm neighborroom(room,room)
avoid(roomlist)
nondeterm go(room,room)
nondeterm route(room,room,roomlist)
% This is the route to be followed.
% roomlist consists of a list of rooms already visited.
nondeterm member(room,roomlist)

CLAUSES
gallery(entry,monsters). gallery(entry,fountain).
gallery(fountain,hell). gallery(fountain,food).
gallery(exit,gold_treasure). gallery(fountain,mermaid).
gallery(robbers,gold_treasure). gallery(fountain,robbers).
gallery(food,gold_treasure). gallery(mermaid,exit).
gallery(monsters,gold_treasure). gallery(gold_treasure,exit).

neighborroom(X,Y):-gallery(X,Y).
neighborroom(X,Y):-gallery(Y,X).

avoid([monsters,robbers]).

go(Here,There):-route(Here,There,[Here]).
go(_,_).

Chapter 274, Example Prolog Programs 397

 route(Room,Room,VisitedRooms):-
member(gold_treasure,VisitedRooms),
write(VisitedRooms),nl.
route(Room,Way_out,VisitedRooms):-
neighborroom(Room,Nextroom),
avoid(DangerousRooms),
not(member(NextRoom,DangerousRooms)),
not(member(NextRoom,VisitedRooms)),
route(NextRoom,Way_out,[NextRoom|VisitedRooms]).

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

After verifying that the program does find a solution to the goal
go(entry, exit).

you might want to try adding some more galleries, for example,
gallery(mermaid, gold_treasure).

Or you can add some additional nasty things to avoid.

Even though--once you've made these additions--there is more than one possible
solution to the problem, your program will only come up with one solution. To
obtain all the solutions, you must make Visual Prolog backtrack as soon as it has
found one solution. You can do this by adding the fail predicate to the first rule
for route:
route(Room, Room, VisitedRooms) :-
member(gold_treasure, VisitedRooms),
write(VisitedRooms), nl, fail.

To get a neater output, you could use a list-writing predicate, write_a_list, to

write the list of names without the containing square brackets ([ and ]) or the
separating commas. However, the rooms you've visited are collected in the
VisitedRooms list in reverse order (exit first and entry last). Therefore, you need
to reverse the list or make the list-writing predicate write the list in reverse.

Hardware Simulation

Every logical circuit can be described with a Visual Prolog predicate, where the
predicate indicates the relationship between the signals on the input and output
terminals of the circuit. The fundamental circuits are described by giving a table
of corresponding truth values (see the and_, or_, and not_ predicates in Program
4).
Fundamental circuits can be described by indicating the relationships between the
internal connections, as well as the terminals. To see how this works, construct an
exclusive OR circuit from AND, OR, and NOT circuits, and then check its
operation with a Visual Prolog program. The circuit is shown in Figure 284.5.

Figure 285.6: Fundamental XOR Circuit2867

In Program 4, this network is described by the xor predicate.

/* Program ch287e04.pro */

DOMAINS
d = integer

PREDICATES
nondeterm not_(D,D)
and_(D,D,D)
or_(D,D,D)
nondeterm xor(D,D,D)

CLAUSES
not_(1,0). not_(0,1).
and_(0,0,0). and_(0,1,0).
and_(1,0,0). and_(1,1,1).
or_(0,0,0). or_(0,1,1).
or_(1,0,1). or_(1,1,1).

xor(Input1,Input2,Output):-
not_(Input1,N1),
not_(Input2,N2),
and_(Input1,N2,N3),
and_(Input2,N1,N4),
or_(N3,N4,Output).
Chapter 274, Example Prolog Programs 399
Given the following goal in interactive mode:
xor(Input1, Input2, Output).

this program yields the following result:

Input1=1, Input2=1, Output=0
Input1=1, Input2=0, Output=1
Input1=0, Input2=1, Output=1
Input1=0, Input2=0, Output=0
4 Solutions

Interpreting this result as a truth table, you can see that the circuit does indeed
perform as expected.

Towers of Hanoi

The solution to the Towers of Hanoi puzzle is a classic example of recursion. The
ancient puzzle of the Towers Of Hanoi consists of a number of wooden disks
mounted on three poles, which are in turn attached to a baseboard. The disks each
have different diameters and a hole in the middle large enough for the poles to
pass through. In the beginning, all the disks are on the left pole as shown in
Figure 288.8.

Figure 289.9: The Towers of Hanoi29010

The object of the puzzle is to move all the disks over to the right pole, one at a
time, so that they end up in the original order on that pole. You can use the
middle pole as a temporary resting place for disks, but at no time is a larger disk
to be on top of a smaller one. It's easy to solve the Towers of Hanoi with two or
three disks, but the process becomes more difficult with four or more disks.
A simple strategy for solving the puzzle is as follows:
You can move a single disk directly.
 You can move N disks in three general steps:
Move N-1 disks to the middle pole.
Move the last (Nth) disk directly over to the right pole.
Move the N-1 disks from the middle pole to the right pole.
The Visual Prolog program to solve the Towers Of Hanoi puzzle uses three
predicates:
hanoi, with one parameter that indicates the total number of disks you are
working with.
move, which describes the moving of N disks from one pole to another--using
the remaining pole as a temporary resting place for disks.
inform, which displays what has happened to a particular disk.
/* Program ch291e05.pro */

DOMAINS
loc =right;middle;left

PREDICATES
hanoi(integer)
move(integer,loc,loc,loc)
inform(loc,loc)

CLAUSES
hanoi(N):-
move(N,left,middle,right).

move(1,A,_,C):-
inform(A,C),!.

move(N,A,B,C):-
N1=N-1, move(N1,A,C,B),
inform(A,C),move(N1,B,A,C).

inform(Loc1, Loc2):-nl,
write("Move a disk from ", Loc1, " to ", Loc2).

To solve the Towers of Hanoi with three disks, give the goal hanoi(3). The output
is:

Chapter 274, Example Prolog Programs 401

 Move a disk from left to right
Move a disk from left to middle
Move a disk from right to middle
Move a disk from left to right
Move a disk from middle to left
Move a disk from middle to right
Move a disk from left to right

Dividing Words into Syllables

Using a very simple algorithm that involves looking at the sequence of vowels
and consonants a word contains, a computer program can decide how to divide
words into syllables. For instance, consider the two sequences:
1) vowel consonant vowel
In this case, the word is divided after the first vowel. For example, this rule
can be applied to the following words:
ruler > ru-ler
prolog > pro-log
2) vowel consonant consonant vowel
In this case, the word is divided between the two consonants. For example,
number > num-ber
panter > pan-ter
console > con-sole
These two rules work well for most words but fail with words like handbook and
hungry, which conform to neither pattern. To divide such words, your program
would have to use a library containing all words.
Write a Visual Prolog program to divide a word into syllables. The program will
first ask for a word to be typed in, and then attempt to split it into syllables using
the two rules just given. As we've mentioned, this will not always produce correct
results.
First, the program should split the word up into a list of characters. You therefore
need the following domain declarations:
DOMAINS
letter = symbol
word= letter*
You must have a predicate that determines whether the letter is a vowel or a
consonant. However, the two rules given can also work with the vocals (the usual
vowels--a, e, i, o, and u--plus the letter y). The letter y sounds like (and is
considered to be) a vowel in many words, for example, hyphen, pity, myrrh,
syzygy, and martyr. To account for the vocals, you have the clauses
vocal(a). vocal(e). vocal(i).
vocal(o). vocal(u). vocal(y).

for the predicate vocal. A consonant is defined as a letter that is not a vocal:
consonant(L) :- not(vocal(L)).

You also need two more predicates. First, you need the append predicate.
append(word, word, word)

Second, you need a predicate to convert a string to a list of the characters in that
string:
string_word(string, word)

This predicate will use the standard predicate frontstr (described in chapter 292),
as well as the standard predicates free and bound (where free(X) succeeds if X is
a free variable at the time of calling, and bound(Y) succeeds if Y is bound), to
control which clause to activate, dependent on the flow-pattern.
Now you're ready to attack the main problem: defining the predicate divide that
separates a word into syllables. divide has four parameters and is defined
recursively. The first and second parameters contain, respectively, the Start and
the Remainder of a given word during the recursion. The last two arguments
return, respectively, the first and the last part of the word after the word has been
divided into syllables.
As a example, the first rule for divide is:
divide(Start, [T1, T2, T3|Rest], D, [T2, T3|Rest]) :-
vocal(T1), consonant(T2), vocal(T3),
append(Start, [T1], D).

where Start is a list of the first group of characters in the word to be divided. The
next three characters in the word are represented by T1, T2, and T3, while Rest
represents the remaining characters in the word. In list D, the characters T2 and
T3, and the list Rest represent the complete sequence of letters in the word. The
word is divided into syllables at the end of those letters contained in D.
This rule can be satisfied by the call:

Chapter 274, Example Prolog Programs 403

 divide([p, r], [o, l, o, g], P1, P2)

To see how, insert the appropriate letters into the clause:

divide([p, r], [o, l, o|[g]], [p, r, o], [l, o | [g]]) :-
vocal(o), consonant(l), vocal(o),
append([p, r], [o], [p, r, o]).

The append predicate concatenates the first vocal to the start of the word. P1
becomes bound to [p, r, o], and P2 is bound to [l, o, g].
The second rule for divide is shown in the complete program, 6.
/* Program ch293e06.pro */

DOMAINS
letter = char
word_ = letter*

PREDICATES
nondeterm divide(word_,word_,word_,word_)
vocal(letter)
consonant(letter)
nondeterm string_word(string,word_)
append(word_,word_,word_)
nondeterm repeat

CLAUSES
divide(Start,[T1,T2,T3|Rest],D1,[T2,T3|Rest]):-
vocal(T1),consonant(T2),vocal(T3),
append(Start,[T1],D1).
divide(Start,[T1,T2,T3,T4|Rest],D1,[T3,T4|Rest]):-
vocal(T1),consonant(T2),consonant(T3),vocal(T4),
append(Start,[T1,T2],D1).
divide(Start,[T1|Rest],D1,D2):-
append(Start,[T1],S),
divide(S,Rest,D1,D2).

vocal('a'). vocal('e'). vocal('i').

vocal('o'). vocal('u'). vocal('y').

consonant(B):-
not(vocal(B)),B <= 'z','a' <= B.

string_word("",[]):-!.
string_word(Str,[H|T]):-
bound(Str),frontchar(Str,H,S),string_word(S,T).
string_word(Str,[H|T]):-
free(Str),bound(H),string_word(S,T),frontchar(Str,H,S).
 append([],L,L):-!.
append([X|L1],L2,[X|L3]):-
append(L1,L2,L3).

repeat.
repeat:-repeat.

GOAL
repeat,
write("Write a multi-syllable word: "),
readln(S),nl,
string_word(S,Word),
divide([],Word,Part1,Part2),
string_word(Syllable1,Part1),
string_word(Syllable2,Part2),
write("Division: ",Syllable1,"-",Syllable2),nl,
fail.

The N Queens Problem

In the N Queens problem, the object is to place N queens on a chessboard in such

a way that no two queens can take each other. Accordingly, no two queens can be
placed on the same row, column, or diagonal.
To solve the problem, you'll number the rows and columns of the chessboard
from 1 to N. To number the diagonals, you divide them into two types, so that a
diagonal is uniquely specified by a type and a number calculated from its row and
column numbers:
Diagonal = N + Column - Row (Type 1)
Diagonal = Row + Column - 1 (Type 2)

When you view the chessboard with row 1 at the top and column 1 on the left
side, Type 1 diagonals resemble the backslash (\) character in shape, and Type 2
diagonals resemble the shape of slash (/). Figure 294.11 shows the numbering of
Type 2 diagonals on a 4x4 board.

Chapter 274, Example Prolog Programs 405

 Figure 295.12: The N Queens Chessboard29613

To solve the N Queens Problem with a Visual Prolog program, you must record
which rows, columns, and diagonals are unoccupied, and also make a note of
where the queens are placed.
A queen's position is described with a row number and a column number as in the
domain declaration:
queen = q(integer, integer)

This declaration represents the position of one queen. To describe more positions,
you can use a list:
queens = queen*

Likewise, you need several numerical lists indicating the rows, columns, and
diagonals not occupied by a queen. These lists are described by:
freelist = integer*

You will treat the chessboard as a single object with the following domain
declaration:
board = board(queens, freelist, freelist, freelist, freelist)

The four freelists represent the free rows, columns, and diagonals of Type 1 and
Type 2, respectively.
To see how this is going to work, let board represent a 4 )4 chessboard in two
situations: (1) without queens, and (2) with one queen at the top left corner.
1. board without queens
board([], [1,2,3,4], [1,2,3,4], [1,2,3,4,5,6,7], [1,2,3,4,5,6,7])

2. board with one queen

board([q(1,1)], [2,3,4], [2,3,4], [1,2,3,5,6,7], [2,3,4,5,6,7])

You can now solve the problem by describing the relationship between an empty
board and a board with N queens. You define the predicate
placeN(integer, board, board)

with the two clauses following. Queens are placed one at a time until every row
and column is occupied. You can see this in the first clause, where the two lists of
freerows and freecols are empty:
placeN(_, board(D, [], [], X, Y), board(D, [], [], X, Y)) :- !.

placeN(N, Board1, Result) :-

place_a_queen(N, Board1, Board2),
placeN(N, Board2, Result).

In the second clause, the predicate place_a_queen gives the connection between
Board1 and Board2. (Board2 has one more queen than Board1). Use this
predicate declaration:
place_a_queen(integer, board, board)

The core of the N Queens Problem lies in the description of how to add extra
queens until they have all been successfully placed, starting with an empty board.
To solve this problem, add the new queen to the list of those already placed:
[q(R, C)|Queens]

Among the remaining free rows, Rows, you need to find a row R where you can
place the next queen. At the same time, you must remove R from the list of free
rows, resulting in a new list of free rows, NewR. This is formulated as:
findandremove(R, Rows, NewR)

Correspondingly, you must find and remove a vacant column C. From R and C,
you can calculate the numbers of the occupied diagonals. Then you can determine
if D1 and D2 are among the vacant diagonals.
This is the place_a_queen clause:

Chapter 274, Example Prolog Programs 407

 place_a_queen(N, board(Queens, Rows, Columns, Diag1, Diag2),
board([q(R, C)|Queens], NewR, NewS, NewD1, NewD2)) :-
findandremove(R, Rows, NewR),
findandremove(C, Columns, NewC),
D1=N+S-R, findandremove(D1, Diag1, NewD1),
D2=R+S-1, findandremove(D2, Diag2, NewD2).

Program 7 is the complete program. It contains a number of smaller additions to

define nqueens, so you only need to give a goal like:
nqueens(5)

to obtain a possible solution (in this case, for placing five queens on a 5 )5 board).
/* Program ch297e07.pro */

DOMAINS
queen = q(integer, integer)
queens = queen*
freelist = integer*
board = board(queens, freelist, freelist, freelist, freelist)

PREDICATES
nondeterm placeN(integer, board, board)
nondeterm place_a_queen(integer, board, board)
nondeterm nqueens(integer)
nondeterm makelist(integer, freelist)
nondeterm findandremove(integer, freelist, freelist)
nextrow(integer, freelist, freelist)

CLAUSES
nqueens(N):-
makelist(N,L),Diagonal=N*2-1,makelist(Diagonal,LL),
placeN(N,board([],L,L,LL,LL),Final), write(Final).

placeN(_,board(D,[],[],D1,D2),board(D,[],[],D1,D2)):-!.
placeN(N,Board1,Result):-
place_a_queen(N,Board1,Board2),
placeN(N,Board2,Result).

place_a_queen(N,board(Queens,Rows,Columns,Diag1,Diag2),
board([q(R,C)|Queens],NewR,NewC,NewD1,NewD2)):-
nextrow(R,Rows,NewR),
findandremove(C,Columns,NewC),
D1=N+C-R,findandremove(D1,Diag1,NewD1),
D2=R+C-1,findandremove(D2,Diag2,NewD2).
 findandremove(X,[X|Rest],Rest).
findandremove(X,[Y|Rest],[Y|Tail]):-
findandremove(X,Rest,Tail).

makelist(1,[1]).
makelist(N,[N|Rest]) :-
N1=N-1,makelist(N1,Rest).
nextrow(Row,[Row|Rest],Rest).

Chapter 274, Example Prolog Programs 409

PA RT 4
Programmer’s Guide
CHAPTER 298

Elements of the Language

299In this chapter, we summarize the elements of the Visual Prolog compiler and
language. We discuss some fundamental elements of the language: names,
program sections, compiler directives, and memory management.
After this we give an introduction to handling modules in Visual Prolog, and how
a program can be split up into several modules, which you can then compile
separately and link together.
We've written this chapter for programmers who've already worked some with
Visual Prolog. To get the most benefit out of this chapter, you should be familiar
with the material in the first chapters of the Visual Prolog Language.

Names

In Prolog, names are used to denote symbolic constants, domains, predicates, and
variables. A name consists of a letter, or an underscore character, followed by any
combination of zero or more letters, digits, and underscores. Two important
restrictions are imposed on names:
Names of symbolic constants must start with a lower-case letter.
Names of variables must start with an upper-case letter or an underscore.
Except for these restrictions, you can use upper-case and lower-case letters in
your programs as you please. For instance, you could make a name more readable
by using mixed upper-case and lower-case, as in the variable
MyLongestVariableNameSoFar

or by using underscores, as in
pair_who_might_make_a_happy_couple(henry_viii, ann_boleyn)

The Visual Prolog compiler does not make a distinction between upper and lower
case letters, except for the first letter. This means that the two variables:
SourceCode

Chapter 300, Elements of the Language 411

and
SOURCECODE

are the same.

Keywords
The following are reserved words; you must not employ them as user-defined
names:
and domains goal include
clauses elsedef if or
constants enddef ifdef predicates
database global ifndef

Specially-Handled Predicates
The following list of predicates are handled specially by the compiler.
assert chain_terms free retractall
asserta consult msgrecv save
assertz db_btrees msgsend term_bin
bound db_chains not term_replace
chain_inserta fail readterm term_str
chain_insertafter findall ref_term trap
chain_insertz format retract write
writef

Program Sections

A Visual Prolog program consists of several program sections. Each program

section is identified by a keyword, as shown in this table.

Table 301.1: Contents of Program Sections302

Section Contents
compiler options Options are given at the top of a program.
constants section Zero or more constants.
domains section Zero or more domain declarations.
facts section Zero or more database predicates.
 Class section Zero or more class declarations
Implement section Zero or more implementations of class-
predicates
predicates section Zero or more predicate declarations.
goal section Zero or one goal.
clauses section Zero or more clauses.

To generate an executable stand-alone application, your program must contain a

goal. Usually, a program requires at least a predicates and a clauses section. For
most programs, a domains section is needed to declare lists, compound structures
and your own names for the basic domains.
For modular programming, you can prefix the keywords domains, predicates
and database with the word global, indicating that the subsequent declarations
affect several program modules globally. (Modular programming is discussed on
page 437).
A program can contain several domains, predicates, database or clauses
sections, provided you observe the following restrictions:
Constants, domains and predicates should be defined before you use them.
However, within the domains section you can refer to domains that are
declared at a later point.
Only one goal must be met during compilation. However, the goal can appear
anywhere.
All clauses that describe the same predicate must occur in sequence (one after
the other).
All global declarations must come before any local declarations.
The database sections can be named, but a given name can only appear once.
Because the default name is dbasedom, there can only be one unnamed
database section.

The Domains Section

A domains section contains domain declarations. Five generic formats are used:

Chapter 300, Elements of the Language 413

 name = d /*standard domain*/
mylist = elementDom* /*list domain*/
my_CompDom = f1(d11,d12,...,d1n); /*compound object domain*/
f2(d21,d22,...,d2n);
...
predefdom = name1;name2;...;nameN
/*eg db_selector and file domains*/
pclass = determspec args flow langspec
/* predicate class declaration */

Standard Domains
name = d

This declaration specifies a domain, name, which consists of elements from a

standard domain type d; the domain type d must be char, real, ref, string,
symbol, or one of the integral domains.
This declaration is used for objects that are syntactically alike but semantically
different. For instance, NoOfApples and HeightInFeet could both be represented
as integers, and consequently be mistaken for one another. You can avoid this by
declaring two different domains of integer type, like this:
apples, height = integer

Declaring different domains in this way allows Visual Prolog to perform domain
checks to ensure, for example, that apples and height are never inadvertently
mixed. However both domains can interchangeably be mixed with integers, and
you can use the equal sign to convert between NoOfApples and HeightInFeet.

List Domains
mylist = elementDom*

This is a convenient notation for declaring a list domain. mylist is a domain

consisting of lists of elements, from the domain elementDom. The domain
elementDom can be either a user-defined domain, or one of the standard types of
domain. You read the asteriskasterisk as "list". For example, this domain
declaration:
numberlist = integer*

declares a domain for lists of integers, such as [1, -5, 2, -6].

Compound Object Domains
myCompDom=f1(d11, .., d1N); f2(d21, d22 ..); ...

To declare a domain that consists of compound objects, you state a functor and
the domains for all the subcomponents.
For example, you could declare a domain of owners made up of elements like
this:
owns(john, book(wuthering_heights, bronte))

with this declaration:

owners = owns(symbol, book)
book = book(symbol,symbol)]

where owns is the functor of the compound object, and symbol and book are
domains of the subcomponents.
The right side of this type of domain declaration can define several alternatives,
separated by a semicolon (;). Each alternative must contain a unique functor and a
description of the domains for the actual subcomponents of the functor. For
example, the following domain declaration could be used to say, "For some
predicates a key is either up, down, left, right or a character value."
key = up; down; left; right; char(char)

There is a possibility to include a comment after the domain, for instance

person= p(string name, integer age).

File Domain
file = name1;name2;...;nameN

A file domain must be defined when you need to refer to files (other than the
predefined ones) by symbolic names. A program can have only one domain of
this type, which must be called file. Symbolic file names are then given as
alternatives for the file domain. For example, this declaration:
file = sales ; salaries

introduces the two symbolic file names sales and salaries.

Chapter 300, Elements of the Language 415

The following alternatives are predefined in the file domain:
keyboard stdin
screen stdout
stderr

Specially Handled Predefined Domains

There are several predefined domains; some are handled specially, like the file
domain and the db_selector domain. Here's a summary of these special
predefined domains:

Table 303.2: Specially Handled Predefined Domains304

dbasedom generated domain for terms in the global database

bt_selector returned binary tree selector
db_selector user-defined external database selectors
place in_memory; in_ems; in_file
accesmode read; readwrite
denymode denynone; denywrite; denyall
ref domain for database reference numbers
file symbolic file names
reg reg(AX,BX,CX,DX,SI,DI,DS,ES) used with bios/4
bgi_ilist list of integers used in the BGI predicates.

Shortening Domain Declarations

As shown in the standard domain declaration name=d, the left side of a domain
declaration (except for a file domain) can consist of a list of names, like this:
mydom1, mydom2, ... , mydomN = ...

This feature allows you to declare several domains at the same time.
firstname, lastname, address = string
Declaring Reference Domains
A reference domain is one that can carry unbound variables as input arguments.
To declare a reference domain, precede the right side of the domain declaration
with the keyword reference. When you declare a compound domain as a
reference domain, all its subdomains are automatically declared as reference
domains.
DOMAINS
reflist = reference refint*
refint = reference integer
term = reference int(refint); symb(refsymb)
refsymb = reference symbol

Declaring Predicate Domains

A predicate domain declares a group or class of predicates. In a subsequent
predicate declaration you may then declare one or more predicates as belonging
to such a group, and these may then be specified as arguments to other predicates.
Those other predicates will hence be able to do a variable call.
The declaration for a predicate domain is of the form:
pdom = { determ | nondeterm } [ domain ] arglist
[ - flowpattern ] [ language ]

(curly braces indicate "choose one", square brackets indicate optional items)
where
domain is the return domain, if you're declaring a function
arglist is of the form
( [ domain [ , domain ]* ] )

flowpattern is of the form

( flow )
where flow is
{ i | o | functor flowpattern | listflow }
where listflow is
'[' flow [ , flow ]* [ '|' { i | o | listflow } ] ']'

language is of the form

language { prolog | c | pascal | asm | stdcall | syscall }

The language specification tells the compiler which calling convention to use,
and is only required when declaring domains for routines written in other
languages (see the chapter on foreign language interface). The calling convention
Chapter 300, Elements of the Language 417
defaults to pascal if omitted, but this should not be relied upon if a particular
convention is desired.
The flowpattern specifies how each argument is to be used. It should be the letter
i for an argument with input flow, the letter o for one with output flow, a functor
and flowpattern for a compound term (e.g. (i,o,myfunc(i,i),o) ), or a listflow
(e.g. [i,myfunc(i,o),o] or [o,o|i] ).
You can have no more than one flowpattern declaration for a predicate pointer
domain, and it must be given unless the argument list is empty or all arguments
have input flow.
Hence, the declaration for a group of deterministic predicates taking an integer as
argument and returning an integer, would be:
DOMAINS
list_process = determ integer (integer) - (i)

This group, or class, is now known as list_process.

The Predicates Section

In Visual Prolog, the sections introduced by the keyword predicates contain
predicate declarations. You declare a predicate by its name and the domains of its
arguments, like this:)
PREDICATES
predname(domain1, domain2,...,domainN)

In this example, predname represents the new predicate name and domain1, ...,
domainN stand for user-defined domains or pre-defined domains. Multiple
declarations for one predicate are also allowed. As an example, you could declare
that the predicate member works both on numbers and names by giving the
following declarations:
PREDICATES
member(name, namelist)
member(number, numberlist)

In this example, the arguments name, namelist, number, and numberlist are user-
defined domains.
You can declare a predicate with several different arities.
hanoi % chooses 10 slices as default
hanoi(integer) % moves N slices
If you give more than one declaration for the same name, these declarations must
come right after each other.
You can declare predicates as being deterministic by preceding the predicate
declaration with determ, or you can declare a predicate as being non-deterministic
by preceding the declaration by nondeterm. If you declare a predicate to be
deterministic, the compiler will issue a warning if it finds any non-deterministic
clauses for the predicate. This functions exactly as if you had used the general
compiler directive check_determ. On the other hand, when you declare a
predicate as non-deterministic, the compiler will not complain when you add
check_determ for checking the other predicates.

nondeterm repeat /*repeat is non-deterministic by design*/

determ menuact(Integer,String) /*menuact is deterministic*/

Note that predicates also can be preceded with the following keywords:
Multi: The keyword multi defines non-deterministic predicates that can backtrack
and generate multiple solutions. Predicates declared with the keyword multi
always succeed (never fail) and, therefore, always have at least one solution.
Failure: A predicate declared with the keyword failure should always fail.
Therefore, such a predicate does not produce a solution. In Visual Prolog failure
predicates always enforce a program to backtrack to the nearest backtracking
point.
Erroneous: A predicate declared with the keyword erroneous should never fail
and should not produce solution. Typical used for errorhandling purposes.

Predicate Classes
If you have declared a predicate domain in the domain section, you may declare
one or more predicates as belonging to that domain. The syntax for this is.
PREDICATES
pred1: p_domain
pred2: p_domain
...

where pred1, pred2 etc. are the predicate names and p_domain is the predicate
domain declared in the domain section.

Functions
By prefixing a predicate declaration with a domain name, you declare a function.
The return value is taken from the last argument in the final clause executed, and

Chapter 300, Elements of the Language 419

this argument must not be present in the predicate declaration. A function
returning the cube of its argument would hence be declared as:
PREDICATES
integer cube(integer)

And the clause for this function would be:

CLAUSES
cube(In,Out):- Out = In*In*In.

A function can return any domain.

The Facts/Database Section

A facts database section declares predicates just as the predicates section does.
However, the clauses for database predicates can only consist of plain facts, they
cannot have an associated body. These facts can be inserted at run time by assert,
asserta, assertz, or consult, and you can remove them again with retract or
retractall. You can have a number of database sections in your program; some of
them can be global and some local. You should name your program's database
sections, and each name must be unique within the module. If you don't give a
name for a database section, the compiler will give it the default name dbasedom.
Only one unnamed database is possible. You can precede a database predicate
with determ if you know that there will be only one fact for that predicate. This
enables the compiler to produce better code, and you will not get non-
deterministic warnings for calling such a predicate. This is useful for flags,
counters, and other things that are essentially global variables.
When a database section is declared, the compiler will internally declare a
corresponding domain with the same name as the name of the database section;
this allows predicates to handle facts as terms.
The form of a facts section is:
[global] DATABASE[ - <databasename> ]
[determ|single|nocopy] dbpred1(....)
dbpred2(.....)

An example is:
FACTS - tables
part(name,cost)
salesperson(name,sex)

PREDICATES
write_table_element(tables)
 CLAUSES
write_table_element(part(Name,Cost)):-
writef("\nPart's Name= % Cost = %",Name,Cost).
write_table_element(salesperson(Name,Sex)):-
writef("\nSalesperson's Name= % Sex = %",Name,Sex).

The Clauses Section

A clause is either a fact or a rule corresponding to one of the declared predicates.
In general, a clause consists of either 1) a fact or 2) a clause head followed first
by a colon and hyphen (:-), then by a list of predicate calls separated by commas
or semicolons. Both facts and rules must be terminated by a period (.).
The fact:
same_league(ucla, usc).

consists of a predicate name (same_league), and a bracketed list of arguments

(ucla, usc).

Simple Constants
Simple constants belong to one of the following standard domains:

char A character (an 8-bit ASCII character enclosed

between a pair of single quotation marks) belongs
to the char domain.
An ASCII character is indicated by the escape
character (\) followed by the ASCII code for that
character. \n, \t, \r produce a newline , a tab and a
carriage return character, respectively. A backslash
(\) followed by any other character produces that
character ('\\' produces \ and '\'' produces ').
integral numbers positive and negative numbers can be represented in
the Visual Prologs integral number domains shown
in the following table.
real A real number belongs to the real domain and is a
number in the range -)1e-307 to -)1e+308. real
numbers.
Real numbers are written with a sign, a mantissa, a
decimal point, a fractional part, an e, a sign, and an

Chapter 300, Elements of the Language 421

 exponent, all without included spaces. For example,
the real value -12345.6789 * 1014 can be written as
-1.23456789e+18.

The sign, fractional, and exponent parts are optional

(though if you omit the fractional part, you must
leave out the decimal point, too). Visual Prolog
automatically converts integers to real numbers
when necessary.
string A string (any sequence of characters between a pair
of double quotation marks) belongs to the string
domain. Strings can contain characters produced by
an escape sequence (as mentioned under char);
strings can be up to 64 K in length.

symbol A symbolic constant (a name starting with a lower-

case letter) belongs to the symbol domain type.
Strings are accepted as symbols too, but symbols
are kept in an internal table for quicker matching.
The symbol table takes up some storage space, as
well as the time required to make an entry in the
table. However, if the same symbols are frequently
compared, it's well worth the investment.
binary A binary constant belongs to the binary domain. It
is written as a comma-separated list of integral
values, each less than or equal to 255, enclosed in
square brackets prefixed with a dollar sign:
$[1,0xff,'a'].
predicate pointer A predicate pointer is the name of a predicate
previously declared as belonging to a predicate
pointer domain. It is written simply as the name of
the predicate, with no argument list or brackets.

Table 305.3: Integral Standard Domains306

Domain Description and implementation

 short A small, signed, quantity.

All platforms 16 bits,2s comp 32768 .. 32767

ushort A small, unsigned, quantity.

All platforms 16 bits 0 .. 65535

long A large signed quantity

All platforms 32 bits,2s comp -2147483648 ..

2147483647

ulong A large, unsigned quantity

All platforms 32 bits 0 .. 4294967295

integer A signed quantity, having the natural size for the

machine/platform architecture in question.

16bit platforms 16 bits,2s comp -32768 .. 32767

32bit platforms 32 bits,2s comp -2147483648 ..

2147483647

unsigned An unsigned quantity, having the natural size for the

machine/platform architecture in question.

16bit platforms 16 bits 0 .. 65535

32bit platforms 32 bits 0 .. 4294967295

byte
All platforms ³ 8 bits 0 .. 255

word
All platforms 16 bits 0 .. 65535

dword
All platforms 32 bits 0 .. 4294967295

An integral value may be preceded by 0x or 0o, indicating hexadecimal and octal

syntax respectively.

Terms
A term is, strictly speaking, any Prolog entity. In practice we tend to mean those
(variable) entities holding data or non-compiled information, or compound terms
(consisting of a functor and optional arguments).

Chapter 300, Elements of the Language 423

Variables
Variables are names starting with an upper-case letter or underscore or, to
represent the anonymous variable, a single underscore character underscore
character). The anonymous variable is used when the value of that variable is not
of interest. A variable is said to be free when it is not yet associated with a term,
and bound or instantiated when it is unified with a term.
The Visual Prolog has a option so it can give a warning when it detects that a
variable has been used only once in a clause. This warning will not be given if the
variable starts with an underscore. Note that when a variable name starts with an
underscore like _Win, it is still a normal variable, that unlike the anonymous
variable can be used to pass values from one call to another.

Compound Objects
A compound object is a single object that consists of a collection of other objects
(called subcomponents) and a describing name (the functor). The subcomponents
are enclosed in parentheses and separated by commas. The functor is written just
before the left parenthesis. For example, the following compound term consists of
the functor author and three subcomponents:
author(emily, bronte, 1818)

A compound object belongs to a user-defined domain. The domain declaration

corresponding to the author compound object might look like this:
DOMAINS
author_dom = author(firstname, lastname, year_of_birth)
firstname, lastname = symbol
year_of_birth = integer

Functorless Compound Objects

By prefixing a compound object declaration with the directive struct, you declare
a functorless object.
DOMAINS
author_dom = struct author(firstname, lastname, year_of_birth)

The internal representation of such an object has no functor and there can be no
alternatives in a functorless domain. Functorless terms can be used just like other
terms in your source code, but their primary aim is to be directly compatible with
C structs.
Lists--A Special Kind of Compound Object
Lists are a common data structure in Prolog and is actually a form of compound
object. Syntactically, it is written as a sequence of comma-separated arguments,
enclosed in square brackets. A list of integers would appear as follows:
[1, 2, 3, 9, -3, 2]

Such a list belongs to a user-defined domain, such as:

DOMAINS
ilist = integer*

If the elements in a list are of mixed types (for example, a list containing both
characters and integers), you must state this in a corresponding domain
declaration. For example, the following declarations
DOMAINS
element = c(char) ; i(integer)
list = element*

would allow lists like this one:

[i(12), i(34), i(-567), c('x'), c('y'), c('z'), i(987)]

Memory Alignment
By prefixing a compound or list declaration with an alignment specification, you
can override the default alignment. The syntax is:
DOMAINS
dom = align { byte | word | dword } domdecl

where domdecl is a normal domain declaration:

DOMAINS
element = align byte c(char) ; i(integer)
list = align dword element*

This would make the internal representation for elements byte-aligned and list
dword-aligned.
If you want to override the default alignment for a functorless domain, the struct
directive must precede the align directive.
DOMAINS
bbdom = struct align byte blm(char,integer)

Chapter 300, Elements of the Language 425

The primary aim of overriding alignment is to make compound objects
compatible with external code using a different alignment than the default for
your platform. If several program share an external database or communicate over
pipes, the domains involved must use the same alignment.

The Constants Section

You can define and use constants in your Visual Prolog programs. A constant
declaration section is indicated by the keyword constants, followed by the
declarations themselves, using the following syntax:
<Id> = <definition>

Each <definition> is terminated by a newline character, so there can be only one

constant declaration per line. Constants declared in this way can then be referred
to later in the program.
Consider the following program fragment:
CONSTANTS
blue = 1
green = 2
red = 4
grayfill = [0xaa, 0x55, 0xaa, 0x55, 0xaa, 0x55, 0xaa, 0x55 ]
language = english
project_module = true

Before compiling your program, Visual Prolog will replace each constant with the
actual string to which it corresponds. For instance:
...
menu_colors(red,green,blue),
my_fill_pattern(grayfill),
text_convert(prolog, language),
status(project_module),
...

will be handled by the compiler in exactly the same way as:

...
menu_colors(4, 2, 1),
my_fill_pattern([0xaa, 0x55, 0xaa, 0x55, 0xaa, 0x55, 0xaa, 0x55]),
text_convert(prolog, english),
status(true),
...
There are a few restrictions on the use of symbolic constants.
The definition of a constant can't refer to itself. For example:
list = [1, 2|list]. /* Is not allowed */

will generate the error message Recursion in constant definition. The

system does not distinguish between upper-case and lower-case in a constant
declaration. Consequently, when a constant identifier is used in the clauses
section of a program, the first letter must be lower-case to avoid ambiguity
with variables. So, for example, the following is a valid construction:
CONSTANTS
Two = 2

GOAL
A=two, write(A).

There can be several constants sections in a program, but each constant must
be declared before it is used.
Constant identifiers are global for the rest of the file and can only be declared
once. Multiple declarations of the same identifier will result in an error
message. You can use constants to redefine names of domains and predicates,
except the specially-handled predicates. Refer to "Specially-Handled
Predicates" earlier in this chapter.

Predefined Constants
Depending on the target platform selected for compilation, one or more constants
will be predefined:

Table 307.4: Predefined Constants308

Constant Target selections causing it to be defined

Chapter 300, Elements of the Language 427

 os_dos DOS, Phar Lap and Windows
os_os2 OS/2 or PM
os_nt Windows 95 or Windows NT 32bit mode
os_unix XENIX, UNIX and Motif
ws_win MS Windows
ws_pm Presentation Manager
ws_motif Motif
dosx286 Phar Lap286
platform_16bit 16-bit platforms
platform_32bit 32-bit platforms

Selecting DOS as your target will cause os_dos to be defined, and selecting MS
Windows will cause both os_dos and ws_win to be defined.
These predefined constants enable you to control platform-dependent conditional
compilation.

Conditional Compilation
You use conditional compilation when you need to generate different versions of
the same program; for example, one version that uses graphics and another that
only uses text mode. The syntax for conditional compilation directives is:
[ifdef | ifndef] <constantID>
...
elsedef
...
enddef

<constantID> represents a constant identifier declared in a constants section. The

value of the constant is irrelevant; only its presence matters. The ifdef directive
succeeds if the constant is defined, while the ifndef directive succeeds if the
constant is not defined. The elsedef part is optional. The following program
shows a typical use of the conditional compilation directives.
CONSTANTS
restricted = 1

ifdef restricted /* if restricted is defined, use this */

savebase(_):-
write("\nBase cannot be saved in demo version"),
readchar(_).

elsedef /* otherwise, use this */

 savebase(Name):-
write("\nSaving ",Name),
save(Name).

enddef

Including Files in Your Program

You use include to include the contents of another file in your program during
compilation. The syntax is:
include "OSFileName"

The OSFileName can include a path name, but you must remember that the
backslash character used to give subdirectories in the DOS-related versions of
Visual Prolog is an escape character in Visual Prolog. Because of this, you must
always give two backslash characters when you use the backslash in a path inside
the source text.
include "\\vip\\include\\error.con"

Under Options | Project | Directories you can give one or more paths separated
by semicolons (colons under UNIX) to indicate where the Prolog system should
look for the include files (Here, of course, only a single backslash is required). If
you don't give an absolute path in your OSFileName, the compiler will in turn try
to concatenate each of the paths given in the include directory to your filename in
order to locate the file.
You can only use include files on section boundaries in a program, so include can
appear only where one of the keywords domains, predicates, goal, database, or
clauses is permitted. An include file itself can contain further include directives.
However, include files must not be used recursively in such a way that the same
file is included more than once during compilation.
Include files can contain any sections, provided the restrictions on program
structure are observed (see page 413).

Compiler Directives

A number of compiler features are controlled through compiler directives. You

can introduce one or more of the following directives at the beginning of the
program text:
Chapter 300, Elements of the Language 429
 bgidriver config nobreak shorttrace
bgifont diagnostics nowarnings trace
check_determ errorlevel printermenu
code heap project

Many of the compiler directives can be set both in the Visual Prolog development
environment (from the menus), through command-line options and in the source
code. If a compiler directive exists in the source code, its setting will override
values set elsewhere.
Note, that most of the compiler directives are now obsolete for VPI Programs.

bgidriver
When you want to link a particular BGI graphics driver directly into your
executable BGI program, use the bgidriver compiler directive followed by the
public name for the graphics driver file. This directive is only relevant for plain
DOS.
bgidriver "_CGA_driver_far"

bgifont
When you want to link BGI stroked character fonts directly into your executable
BGI program, use the bgifont compiler directive followed by the public name for
the font file. This directive is only relevant for plain DOS.
bgifont "_gothic_font_far"

check_determ Options|Project|Compiler Options|Nondeterm Check

When you specify check_determ, the Visual Prolog system will give a warning
for each program clause that results in a non-deterministic predicate. There are
two kinds of non-deterministic clauses:
1. If a clause does not contain a cut, and there are one or more clauses that can
match with the same input arguments for that flow pattern.
2. If a clause calls a non-deterministic predicate, and that predicate call is not
followed by a cut.
You can use check_determ to guide the setting of cuts. Visual Prolog itself
performs extensive tests to decide whether a predicate is deterministic or non-
deterministic, so you don't need to fill your programs with cuts merely to save
stack space (as is necessary in many other Prolog implementations).
If some predicates are non-deterministic, and you wish to squelch the warning
messages, you can precede a predicate declaration by determ or nondeterm; these
state that the predicate is to be deterministic or non-deterministic, respectively.
For example,
PREDICATES
nondeterm repeat
determ readname(string)

By default, all global predicates are treated as being deterministic. nondeterm

allows you to declare a global predicate as non-deterministic. You can also use
determ on database predicates intended to have no more than one clause. In this
case, the compiler generates simpler code when accessing the predicate.

code Options|Project|Compiler Options|Default for Local Calls

The code directive specifies the size of the internal code array. The default is
4000 paragraphs (16-byte units) for the 16-bit versions of Visual Prolog,
otherwise 10000 paragraphs. For larger programs you might need to specify a
larger size.
code = Number_of_paragraphs

where Number_of_paragraphs represents the number of memory paragraphs (16

bytes each) required in the code array. For example, the directive:
code = 1024

sets the size of the code array to 16 Kbytes.

The code directive has no influence on the size of an executable file, it simply
controls how much memory the compiler should allocate for the compilation.
When the code size exceeds the value 4095, the compiler will switch over to
generating FAR calls inside that module. For this reason, you should only use a
code size above 4095 if it is really needed. For 32-bit code generation, the size of
the code array is practically irrelevant. All code is NEAR, and the operating
system will only allocate physical memory when the allocated virtual memory is
actually used.

config
This option is only relevant for old DOS textmode windowing applications !
To let a stand-alone application read a configuration file that defines default
window attributes, keyboard setup, etc., place the directive:

Chapter 300, Elements of the Language 431

 config "<ConfigFileName>.cfg"

in your program. The application will read <ConfigFileName>.cfg and set the
configurations the same way Visual Prolog does with its configuration file.
Various options are passed from the environment to the executable file; these will
either be built in during compilation or read from the configuration file during
start-up. Some of the values built into the execuatble file can be overwritten by
corresponding settings in the configuration file. Here's an overview of these
parameters:

Passed during Read only from Overwritten from

compilation <Config>.cfg <Config>.cfg
Stack Size x x
Heap Size x
Snow Check x x
Keyboard x
Layout
Help Lines x
Xedit Setup x

diagnostics Options|Project|Compiler Options|Diagnostics output

When you specify diagnostics, the compiler will display an analysis of your
program containing the following information:
the names of the predicates used
whether a predicate is local, global or defined externally
whether a predicate is deterministic or non-deterministic
the size of the code for each predicate
the domain types of the parameters
the flow patterns for each predicate
The diagnostics will also produce a listing of which domains are treated as
reference domains, and for which domains the compiler generates internal
unification predicates. These predicates are generated when you unify two terms
in certain non-simple ways. As an example if you are writing L1=L2 where both
L1 and L2 are bound to a list, the compiler needs to test all the elements of the list
for equality.
Here's an example of a diagnostics display:
DIAGNOSTICS FOR MODULE: /usr/pdev/test/flow1.pro

Predicate Name Type Determ Size Domains -- flowpattern

---------------- ------ ------ ----- ----------------------------
_PROLOG_Goal local yes 216 --
_p1_0 global yes 112 integerlist -- [i,o,i|o]
_p1_1 global yes 128 integerlist -- o
---------------- ------ ------ ----- ----------------------------
Total size 460

Size of symbol table= 324 bytes

Size of PROCONST segment=1705 bytes

Under Options | Global | Environment, it is possible to log the diagnostics

output into a file.

errorlevel Options | Project | Compiler Options | Errorlevel

The compiler directive errorlevel enables you to control how detailed the error
reporting should be. The syntax is:
errorlevel = d

where d is one of 0, 1, or 2, representing the following levels:

d Level of Error Reporting

0 Generates the most efficient code. No cursor information will be
placed in the code and only the error number will be displayed in an
error occurs.

1 This is the default. When an error occurs, its origin (module name
and include file, if applicable) will be displayed. The place where
the error was detected within the relevant source file will also be
displayed, expressed in terms of the number of bytes from the
beginning of the file.

2 At this level, certain errors not reported at level 1, including stack

overflow heap overflow, trail overflow, etc., are also reported.

Chapter 300, Elements of the Language 433

In a project, it is the error-level option in each module that controls that module's
detail of saving the cursor information. If, however, the error-level option in the
main module is higher than that of the sub-modules, Visual Prolog might generate
misleading error information. For example, if an error occurs in a module
compiled with error level 0, which is included in a main module compiled with
error level 1 or 2, the system will be unable to show the correct location of the
error. Instead, it will indicate the position of some previously executed code.

heap Options | Project | Compiler Options | Heap Size

Note: this is relevant only if you are going to implement a DOS TSR program.
The heap directive specifies how much memory your .EXE file should allocate
when it is started from DOS. If you don't use the heap directive, or if you set it to
the value 0, the program will allocate all available memory. This is normally the
right way to do it, but if you want to implement a RAM-resident Visual Prolog
program, your program should only allocate the necessary memory. The format
is:
heap = Number_of_paragraphs

nobreak Options | Project | Compiler Options | Break Check

Note: this is only relevant for DOS textmode programs.

In the absence of the nobreak compiler directive, the Visual Prolog system will
generate code to check the keyboard before each predicate call, to ensure that the
Ctrl+Break key combination has not been pressed. This slows down program
execution slightly and takes up a little extra program space.
The nobreak directive prevents this automatic generation of code. When nobreak
is in operation, the only way to escape an endless loop is to reboot the computer
(DOS, Phar Lap) or kill the process in some other way. nobreak should only be
used after a program has been thoroughly tested.

nowarnings Options|Project|Compiler Options|Single variable

The nowarnings directive suppresses the warnings given when a variable occurs
only once in a clause.
Note: This directive is only included for completeness. If a variable occurs only
once, either it is a mistake or it should be replaced by the anonymous variable - or
a variable starting with an underscore.
printermenu Options|Project|Compiler Options|Print menu in DOS .EXE

Note: this is only relevant for DOS textmode programs.

When this compiler directive appears in the program, Visual Prolog will place
code in the executable file for handling the Alt-P key. This means that the user
will be able to send screen output to the printer or capture it to a log file.

project Options|Project|Compiler Options|.SYM File Name

The project compiler directive is used in modular programming modular
programming. All Visual Prolog modules involved in a project need to share an
internal symbol table. If the project is called MYPROJ, the symbol table will be
placed in a file called MYPROJ.SYM. The project directive must appear on the
first line of a module to specify which project that module belongs to. For
example, the following line of code states that a module belongs to the MYPROJ
project:
project "myproj"

The project name is not allowed to have a path or an extension.

If the name of the .SYM file is given in the VDE or as option to the commandline
compiler, the project directive will be ignored.
See page 437 for complete details about modular programming.

Chapter 300, Elements of the Language 435

Visual Prolog Memory Management

Visual Prolog uses the following memory area’s:

Stack the stack is used for transferring arguments and return addresses for
predicate calls. The stack also holds the information for
backtrackpoints.
Heap the heap holds all objects that are more or less permanent, such as
database facts, window buffers, file buffers etc.
GStack the global stack, normally called gstack, is the place where lists,
compound structures and strings are placed. The Gstack is only released
during backtracking.
Trail The trail is only used when the program uses reference variables. It
holds information about which reference variables must be unbound
during backtracking. The trail is allocated in the heap.

Releasing Spare Memory Resources

During program execution, the memory is used for several different purposes;
depending upon the purpose, spare memory resources can be released in separate
ways.
To minimize stack use, avoid unnecessary non-determinism; use the
check_determ directive to guide the setting of cuts. Also, take advantage of
tail-recursion elimination by writing your predicates so they are tail-recursive.
The global stack is used for building strings and structures. In order to save
global stack space, write your program so that the outer loop is a repeat...fail
loop.
The trail will seldom be a problem in Visual Prolog. In all versions of Visual
Prolog, the trail is dynamically allocated (in the heap), and will be increased in
size when necessary. However, in the 16-bit versions the trail is limited to 64K
and the first thing to do if the system complains about trail overflow is to
avoid using reference domains. If you want to use reference domains, you
should decrease the number of backtrack points by using some cuts (use
check_determ). The repeat...fail combination will also release the trail. As a
last resort, rearrange your predicate calls so that you create less reference
variables.
 The heap is used when facts are inserted in a database and to store window
buffers, file buffers, graphic drivers, database tables, etc. These areas are
automatically released when facts are retracted, windows are closed, and so
on.

Modular Programming

A Visual Prolog program can be broken up into modules. You can write, edit, and
compile the modules separately, and then link them together to create a single
executable program. If you need to change the program, you only need to edit and
recompile individual modules, not the entire program--a feature you will
appreciate when you write large programs. Also, modular programming allows
you to take advantage of the fact that, by default, all predicate and domain names
are local. This means different modules can use the same name in different ways.
Visual Prolog uses two concepts to manage modular programming: projects and
global declarations. Among other things, these features make it possible to keep a
record of which modules make up a program (this record is called a project), and
to perform type-checking across module boundaries. In this section, we'll define
the two concepts; then, using a simple example, we'll show you how some
modules can be combined into a single, stand-alone program.

Global Declarations
By default, all names used in a module are local. Visual Prolog programs
communicate across module boundaries using the predicates defined in the global
predicates and global database sections. The domains used in these global
sections must be defined as global domains, or else they must be pre-defined
domains.
ALL THE MODULES IN A PROJECT NEED TO HAVE EXACTLY THE
SAME GLOBAL DATABASE AND GLOBAL DOMAINS
DECLARATIONS.
If you mix this up, all sorts of strange things will probably happen, such as a hung
computer under DOS or MS Windows, or a protection violation on other
platforms.
The easiest way to ensure this is correct is by writing all global declarations in a
single file, which you can then include in every relevant module with an include
directive. For example, if all your global declarations are in a file called
global.inc, you can include that file in every relevant module by adding the
directive:

Chapter 300, Elements of the Language 437

 include "global.inc"

to the top of each module.

All global declarations must appear before any local declarations local
declarations.

Global Domains
You make a domain global by writing it in a global domains section. In all other
respects, global domains are the same as ordinary (local) domains.
Note: If any global domain definition is changed, all modules in that project must
be recompiled.

Global Database
You make a database section global to a project by preceding the database
keyword with the keyword global. You can only give initializing facts for global
databases in the main module, which is the one containing the goal section. The
goal section must appear before the global database clauses in the main module.
Note: If any global database definition is changed, all modules in that project
must be recompiled.

Global Predicates
Global predicate declarations differ from ordinary (local) predicate declarations
because they must contain a description of the flow pattern(s) by which each
given predicate can be called. If such one is not specified, all arguments will be
input. The syntax for a global predicate declarations is:
[ { determ | nondeterm | single | nocopy} ] [ domain ] name arglist
[ - flowpattern [ [,] flowpattern ]* ] [ language ] [ namespec ]

(curly braces indicate "choose one", square brackets indicate optional items)
where
Note, that predicate types can also appear before the flowpattern.

domain is the return domain, if you're declaring a function

name is the name of the predicate
arglist is of the form
( [ domain [ , domain ]* ] )
 flowpattern is of the form
( flow [ , flow ]* )
where flow is
{ i | o | functor flowpattern | listflow }
where listflow is
'[' flow [ , flow ]* [ '|' { i | o | listflow } ] ']'

language is of the form

language { c | asm | pascal | prolog | stdcall | syscall }

namespec is of the form

as "extname"

The namespec may be used to specify the public object-code name, overriding the
default naming used by Visual Prolog. The main use of this is when you're
linking in modules written in other languages.
The language directs the calling convention used when calling the predicate. This
defaults to "prolog".
In the following global predicate declaration, name and home are of type string,
and age is of type integer; the arguments to first_pred can either be all bound (i,
i, i) or all free (o, o, o):
first_pred(name,home,age) - (i,i,i) (o,o,o)

Here is the declaration for a predicate with either compound flow of an integer
list, or plain output flow:
p1(integerlist) - ([i,o,i|o]),(o)

Finally, this declaration specifies compound flow for an object declared as

func(string,integer) coming from a domain called mydom:
pred(mydom) - (func(i,o)) (func(o,i))

Note: If any global predicate definition is changed, only the modules that refer to
this predicate need to be recompiled. However, it is rather critical that this
recompilation is done; if you change the flow pattern of a predicate the calls using
it will need different code.
It doesn't matter in which module the clauses for global predicates appear, but--as
with local predicates--all clauses must appear together.

Chapter 300, Elements of the Language 439

CHAPTER 309

Interfacing with Other Languages

310Although Visual Prolog is an excellent tool for many purposes, there are still
reasons to use other languages. For example, it's easier to perform numeric
integration in C, and interrupt-handling and low-level stuff is perhaps better done
in Assembly language. Moreover, if you've developed a large program in another
language that already solves some aspect of the problem, this work should not be
wasted. For these reasons, Visual Prolog allows you to interface your programs
with other languages, as long as those languages produce standard object files and
follows the conventions outlined in this chapter.
In this chapter you will find a number of examples of interfacing C and Visual
Prolog. Their source files are in the DOC\EXAMPLES directory or the
FOREIGN directory of your distribution. In order to run them, you need to have
the appropriate development system and/or C compiler and libraries installed.
The process to compile and link the examples varies considerably between
the different operating systems and the different C compilers. In the foreign
subdirectory of your distribution you will find thorough instructions and
examples for the different platforms and compilers. Read these instructions
carefully.
When using the Visual Prolog Development Environment, you don't, strictly
speaking, need to know how to compile C programs, how to run the linker, or
which libraries to specify and how to do it. This is handled automatically.
However, you should have a fairly thorough understanding about C, and be able
to write, compile and link multi-module C programs yourself.

Using DLL’s

A dynamic-link library (DLL) is a binary file that acts as a shared library of

predicates that can be used simultaneously by multiple applications.
Visual Prolog can generate dll’s and link in dll’s staticly or load dll’s dynamicly.
For more information about Visual Prolog and dll’s please see the examples VPI\
EXAMPLES\DLL and VPI\TOOLEXAMP\BUILD.
Calling Other Languages from Visual Prolog

In this section, we cover what you need to know to call C, Pascal and assembler
routines from Visual Prolog.
Before calling routines and functions written in other languages, you need to
declare them as external predicates in Visual Prolog. You also need to understand
the correct calling conventions and parameter-pushing sequences, and you need to
know how to name the different flow variants of your external predicates.

Declaring External Predicates

To inform the Visual Prolog system that a given global predicate is implemented
in another language, you need to append a language specification to the global
predicates declaration, as briefly mentioned in chapter 311:
GLOBAL PREDICATES
add(integer,integer,integer) - (i,i,o),(i,i,i) language c
scanner(string,token) - (i,o) language pascal
triple(integer,real) - (i,o) language asm

In Visual Prolog, you explicitly list the interfaced language; this simplifies the
problems inherent in calling conventions, such as activation record format,
naming convention and returning conventions.

Calling Conventions and Parameter Passing

The 80x86 processor family gives programmers a choice between NEAR and
FAR subroutine calls, when running 16-bit programs. Visual Prolog requires all
global routines to be FAR. The same applies to pointers to data objects. Many 16-
bit compilers for the 80x86 family require you to choose between 16-bit and 32-
bit pointers, where the 16-bit pointers refer to a default segment. In order to
access all of memory, Visual Prolog always uses 32-bit pointers.
For 32-bit programs, "NEAR" means 32 bits and the above considerations are
irrelevant.

Input parameters
For input parameters, the value is pushed directly, and the size of the parameter
depends on its type.

Index 441
Output parameters
An output parameter is pushed as a 32-bit pointer to where a values must be
assigned.

Return Values
Visual Prolog follows the most widely adopted register convention for function
values on the 80x86 CPU family. This should not be of any concern in most
cases, but is included here for completeness.

Table 312.1: Registers for Return Values313

Operand Size Program Type

16 bit 32 bit

byte (8 bits) AL

word (16 bits) AX

dword (32 bits) DX:AX ³ EAX

Pointers are 32 bits in size and are handled as dwords. The Program Type is
determined by the operating system,
Floating point values are exceedingly troublesome to handle. They may be
returned in registers, on the (emulated) coprocessor stack, and the pascal calling
convention will frequently return them through pointers. Currently pascal
functions cannot return floating point values. See the notes in the FOREIGN
subdirectory of your distribution for any special considerations for your platform.
In any case, floating point values can always be returned in arguments. However,
take special note that Visual Prolog's real corresponds to a C double (8 bytes).
You should also be aware that currently external C functions cannot return C
structs (but they may of course return pointers to structs).

Multiple declarations
In Visual Prolog, a predicate can have several type variants, arities, and flow
variants, and a separate procedure is needed for each type and flow variant. When
you implement predicates, having several versions, in C, each C function must
have a name corresponding to the name generated by Visual Prolog. The naming
convention used by Visual Prolog is straightforward; the predicate name is used
as the root, and the suffix _X is appended to signify the variant number, where X
is an integer starting at 0. If there is only one variant, no suffix is appended.
Consider the following program:
GLOBAL PREDICATES
add(integer,integer,integer) - (i,i,o),(i,o,i),(o,i,i),(i,i,i)
language c
square(integer,integer) - (i,o)

GOAL
add(2,3,X), write("2 + 3 = ",X), nl,
add(2,Y,5), write("5 - 2 = ",Y), nl,
add(Z,3,5), write("5 - 3 = ",Z), nl,
add(2,3,5), write("2 + 3 is 5"), nl,
square(5,Sq), write("5 squared is ",Sq).

A module linked with this program should contain the following C functions:
add_0 for the first flow pattern (i,i,o)
add_1 for the (i,o,i) flow pattern
add_2 for (o,i,i)
add_3 for (i,i,i)
square
As an example, the following C module implements square as well as all flow
patterns for add:
add_0(int x, int y, int *z) /* (i,i,o) flow pattern */
{ *z = x + y; }

add_1(int x, int *y, int z) /* (i,o,i) flow pattern */

{ *y = z - x; }

add_2(int *x, int y, int z) /* (o,i,i) flow pattern */

{ *x = z - y; }

add_3(int x, int y, int z) /* (i,i,i) flow pattern */

{ if ( (x + y) != z ) RUN_Fail(); }

square(int i,int *i_sq)

{ *i_sq = i*i; }

Index 443
Parameter pushing order
When interfacing to a routine written in C, the parameters are pushed onto the
stack in reverse order and, after return, the stack pointer is automatically adjusted
by Visual Prolog.
When calling languages other than C, the parameters are pushed in the normal
order, and the called function is responsible for removing the parameters from the
stack.

Leading underscored
On the 16bit platforms, C compilers will prefix the name of public C functions
with an underscore. Therefore, global predicates declared as language C will also
have their name prefixed with an underscore if the target platform is one of these.

NT naming convention
Under 32-bit Windows the OMF (Object Module Format - used, for example, by
Borland C++) and COFF (Common Object File Format - used, for example, by
Visual C++) object file formats are used:
 In OMF format the predicate name is placed as is. This means that a predicate
p will be named p.
 In COFF format the predicate name is prefixed with an underscore and the
number of bytes pushed on the stack will together with a '@' sign be suffixed
to the predicate name. This means that a predicate p, which has two integer
arguments, will be named _p@8.
When choosing the calling convention stdcall under 32-bit Windows, these
conventions will be used. So it is recommended to use stdcall convention calling
32-bit Windows API functions.

Converting the name to Uppercase (Pascal)

PASCAl uses the convention, that the name is converted to uppercase. So if
language PASCAl is used, the name will during .OBJ module generation be
converted to uppercase.

Adjustment of stackpointer
There are two possibilities of adjusting the SP register . This can be done either
by the called function or the calling function. Traditionally PASCAL does this in
the called function, while C does it in the calling function.
 Table 314.2: Calling conventions315

Convert Add Push Adjust SP NT

name to leading args after naming
upper case under Reversed return convention
Score
pascal X
c X X X
stdcall X X X
syscall X X

The AS "external_name" Declaration

As an alternative to the automatic naming convention, you can use the as
keyword in the global declaration, like this:
GLOBAL PREDICATES
scanner(string,token) - (i,o) language c as "_myscan"

The result of this is that Visual Prolog will refer to the name _myscan in the
object file instead of _scanner. You would still refer to the name scanner in your
Visual Prolog source.
You can only use the as option if there is a single flow variant for the predicate.

Domain Implementation

Most types normally used in C form a subset of Visual Prolog domains, and
hence have direct equivalents. Below we discuss both simple and complex
domain equivalents in C.

Index 445
Simple Domains
The implementation of Visual Prolog's simple domains are outlined in the
following table:

Table 316.3: Visual Prolog Simple Domains317

Domain Implementation
16-bit OS 32-bit OS

char, byte 1 byte (see note) 1 byte (see note)

(u)short, word 2 bytes 2 bytes

(u)long, dword 4 bytes 4 bytes

unsigned, integer 2 bytes 4 bytes

real 8 bytes (IEEE format) 8 bytes (IEEE format)

ref 4 bytes 4 bytes

Note: The char and byte domains occupy a machine word when pushed on the
stack (2 bytes for 16-bit programs, 4 bytes for 32-bit programs).

Complex Domains
All non-simple domains are implemented as pointers to things.
The string and symbol domains are pointers to null-terminated character arrays,
with symbols being hashed and stored in the symbol table in the heap.
The binary domain is a pointer to a block of memory, prefixed by a dword (32bit
platforms) or word (16bit platforms) indicating the net size of the block.
Size bytes

^
|
Pointer

Special consideration must be given to allocation of memory for complex

domains. This will be detailed in a later section.
Ordinary Compound Objects and Structures
User-defined compound objects are pointers to records (structs and unions in C).
The general format of these is a byte representing the functor (domain
alternative), followed by the individual components of the term. These will vary,
depending on which alternative we're dealing with. In any case, components
belonging to simple domains are stored directly in the term record itself, while
complex components are themselves stored as pointers. For example, in this code
fragment:
DOMAINS
mydom = i(integer); c(char); s(string)

the functor number will be 1 for the first alternative, i(integer), 2 for the second,
c(char), and 3 for the third.
A suitable C typedef for mydom would be:
typedef struct {
unsigned char func;
union {
int i;
char c;
char *s;
} u;
} MYDOM;

Here func will have the value 1, 2 or 3, depending on which domain alternative
we're dealing with. This then indicates which of the union's components it's
appropriate to access.

Functorless Terms (structs)

By prefixing a compound domain declaration with the compiler directive struct,
the terms belonging to that domain will not carry functors with them:
DOMAINS
rec = struct record(d1,d2,...)

Apart from the struct directive, terms belonging to a functorless domain are used
and written exactly like other terms in your program, except that there can be no
alternatives in a functorless domain.
Functorless terms allow you to duplicate C structs when interfacing to C routines
and libraries using predefined structs. Apart from that, they'll save you a bit of
memory if you don't need alternatives in the domain.

Index 447
Lists
Lists are implemented exactly like ordinary compound domains, with a field at
the end of the record pointing to the next. This is known as linked lists in C
terminology. From a Prolog perspective lists are merely a notational convenience.
Given for instance a declaration for a list of strings:
DOMAINS
strlist = string*

the C structures relevant for this are identical to those for:

DOMAINS
strlist = elem(string,strlist); endoflist()

The records for the elem alternative will contain:

1. a functor
2. a pointer to a string
3. a pointer to the next element in the list
and the record for the endoflist alternative will only contain a functor.
This collection of fields is reflected in the C data structure for strlist:
struct node {
unsigned char functor; /* The type */
char *value; /* A string pointer */
struct node *next; /* A pointer to struct node */
} strlist;

The functor field indicates the type of list record. The value is 1 if it's a list
element, and 2 if it's the end of the list.

Memory Considerations

While all memory considerations are handled automatically when you write pure
Prolog code, you need to take special care when interfacing to foreign languages.
In this section we'll describe several of these aspects.

Memory Alignment
C compilers for 32-bit platforms will usually align data on dword boundaries,
while those for 16-bit platforms will usually align on byte boundaries. The reason
for aligning on word or dword boundaries is speed. On a 32-bit platform, dword
alignment will give up to 10-12 percent faster execution than byte alignment.
For simple variables alignment isn't important, but in order for Prolog terms and
C structures to be compatible, the data contained in the records must be
identically aligned. To this end, Visual Prolog gives you the option of selecting a
different alignment than what's the default for your platform. The default is
dword if you're using a 32-bit version of Visual Prolog, otherwise byte.
The alignment scheme may be selected with the help of the Options | Project |
Compiler Options menu item, or through the -A command line option.
Additionally, the align compiler directive may be used to override alignment on
selected domains, like this:
DOMAINS
dom = align { byte | word | dword } func(d1,d2,...) [;
func1(...); ... ]

The align directive must appear before any alternatives in the domain, and all
alternatives will have the alignment specified. It's not possible to specify different
alignment for individual alternatives.
For functorless terms, the align directive should appear after the struct directive.
Note that when several processes share a database or communicate over pipes, it's
crucial that the domains involved use identical alignment.

Example
Byte alignment is easy: each element is simply put right after the previous one.
Given the declaration dom = struct my_struct(char,short,char,long) (recall that
the struct directive declares the term to be functorless), the term
my_struct('P',29285,'B',1702063209) is stored in memory like this:

Byte number:

| 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 |

_______________________________
| | | | |
|'P'| 29285 |'B'| 1702063209 |
|___|_______|___|_______________|

Word and dword alignment is a bit trickier. Here, items are stored in memory so
that accessing them won't cross a word or dword boundary. That means that the
individual elements of terms may be followed by a number of unused bytes,
Index 449
depending on the size of the following element. With dword alignment, the term
above would be stored like this:
Byte number:

| 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10| 11|

_______________________________________________
| | | | | | | | |
|'P'|PAD| 29285 |'B'|PAD|PAD|PAD| 1702063209 |
|___|___|_______|___|___|___|___|_______________|

The PADs indicate unused bytes, allowing the values following them to be stored
on suitable boundaries.
Notice that it's sufficient for the value 29285 to be aligned on a word boundary,
because it's a short (16 bits); accessing it on a word boundary won't cross any
undesirable boundaries.

Memory Allocation
When you create and return compound objects to Visual Prolog, memory for the
objects must normally be allocated on the Global Stack. This memory will
automatically be released if you fail back to a point previous to its allocation.
GStack memory is allocated using:
void *MEM_AllocGStack(unsigned size);

You would typically use C's sizeof function to determine how much memory to
allocate. Given for instance the mydom domain discussed previously, the Prolog
declarations for a C routine returning a term belonging to that domain in an
argument would be:
/* Program mydom_p.pro */

project "mydom"

global domains
mydom = i(integer); c(char); s(string)

global predicates
determ make_mydom(mydom) - (o) language C

goal
make_mydom(MD), write(MD), nl.
And the C code for mydom and make_mydom could be:
/* Program mydom_c.c */

typedef struct {
unsigned char func;
union {
int i;
char c;
char *s;
} u;
} MYDOM;

void *MEM_AllocGStack(unsigned);
char *MEM_SaveStringGStack(char *);

void make_mydom(register MYDOM **md)

{
*md = MEM_AllocGStack(sizeof(MYDOM));
(*md)->func = 3;
(*md)->u.s = MEM_SaveStringGStack("wombat");
}

Notice that, as terms are handled through pointers in Prolog, the argument to
make_mydom is a pointer to a term pointer. This example also makes use of
another GStack-related function, MEM_SaveStringGStack, which allocates
GStack space for the string (based on its length), then copies the string into the
allocated space, returning a pointer to it. There's a few other handy functions in
Visual Prolog's library:
char *MEM_SaveStringHeap(char *String);
/* Copies String to heap */

unsigned STR_StrLen(char *String);

/* Returns length (excluding terminating null byte) of String */

void MEM_MovMem(void *Source,void *Dest,unsigned Len);

/* Moves Len bytes from Source to Dest; these may overlap */

Pre-allocation of Memory
Many C library functions require you to specify a pointer to a structure, which the
C routine then fills in. In this case the compound flow pattern for global
predicates should be used to specify what's happening:

Index 451
 GLOBAL DOMAINS
off_t, time_t = long
dev_t = short
stat = struct stat(dev_t,ushort,ushort,short,ushort,ushort,
dev_t,off_t,time_t,time_t,time_t)

GLOBAL PREDICATES
determ integer stat(string,stat) -
(i,stat(o,o,o,o,o,o,o,o,o,o,o)) language C

When you call stat

..., 0 = stat("/unix",Stat), !, write(Stat).

Visual Prolog will allocate memory for the stat structure before the call.

The sizeof function

Visual Prolog has a sizeof function that duplicates C's sizeof function, returning
the size of the specified domain or variable.
For a compound domain with alternatives, sizeof will return the size of the largest
alternative. Given a second argument of a functor from one of the domain
alternatives, sizeof will return the size of that particular alternative.
Given a variable, sizeof will return the size of the corresponding domain
(alternative), except that for string variables or constants, sizeof will return the
number of bytes in the string including the terminating zero byte.
The program align.pro illustrates alignment selection and the use of the sizeof
function:
/* Program align.pro */

DOMAINS
dom = struct f(char,integer)
dom1 = align word f(integer,integer,long); g(string)
refint = reference integer

predicates
refint(refint)

clauses
refint(_).

goal % Find the size of a functorless domain

A = sizeof(dom),
write("\nSize=",A),
% when there are alternatives, the largest is returned
B = sizeof(dom1),
write("\nSize=",B),

% Find size of a single alternative

C = sizeof(dom1,g),
write("\nSize=",C),

% Find size of a term pointed to by a variable

X = f(1,1,1), % This is from dom1
D = sizeof(X),
write("\nSize=",D),

% Find size of a string pointed to by a variable

Y = "hello there",
E = sizeof(Y),
write("\nSize=",E),

% Find size of a reference variable

refint(Z),
F = sizeof(Z),
write("\nSize=",F).

Load and run this program. Try changing the domains and their alignment, and
watch the results.

malloc and free

When writing functions in other languages, you often need to allocate dynamic
memory. You've already seen MEM_AllocGStack, but this allocates memory on
Prolog's Global Stack, which is released automatically. Permanent allocations
should be done in the heap, and because Visual Prolog already has suitable
memory allocation routines, it's generally preferable to use these. In fact, in DOS
it's mandatory to use them, since a foreign memory allocation package would be
allocating memory from the same physical memory as Visual Prolog. On other
platforms, you can use C's malloc and free, but this would duplicate an amount of
code and data, and the two packages would both be holding released memory in
separate pools. Moreover, Visual Prolog's heap allocation system has a
performance far superior to that supplied with most C compilers.
Therefore, on all platforms except UNIX, public routines for malloc and free,
consisting of bindings to Visual Prolog's heap allocation routines, are provided in
the initialization assembler and object files. These files are found in the
subdirectories for the different platforms and compilers in the FOREIGN
directory of your distribution. Note that when linking, it's essential that the

Index 453
appropriate initialization file appears before the (C) library containing malloc and
free.

Examples

List Handling
In this section we give a more useful example that shows how to convert a list to
an array and back to a list again.
The C routine ListToArray takes a list of integers, converts this to an array
placed on the Global Stack, and returns the number of elements. The conversion
is done in three steps:
1. The list is traversed in order to count the number of elements.
2. The array with the needed number of elements is allocated.
3. The list is traversed again while the elements are transferred to the array.
The C routine ArrayToList takes an integer array and the size of the array as
arguments, then converts these to a list of integers. This routine only makes one
pass, building the list as it indexes through the array.
All of this is used in the C-coded predicate inclist. When given a list of integers,
inclist first converts the input list to an array, increments the elements of the array
by 1, then converts the array back to a list of integers.
/* Program lstar_p.pro */

project "lstar"

global domains
ilist = integer*

global predicates
inclist(ilist,ilist) - (i,o) language c

goal
inclist([1,2,3,4,5,6,7],L), write(L).

Here is the C program defining the two C procedures ListToArray and

ArrayToList, and the external Visual Prolog predicate inclist.
 /* Program lstar_c.c */

#define listfno 1
#define nilfno 2
typedef unsigned char BYTE;

void *MEM_AllocGStack(unsigned);

typedef struct ilist {

BYTE Functor;
int Value;
struct ilist *Next;
} INTLIST;

int ListToArray(INTLIST *List,int **ResultArray)

{
INTLIST *SaveList = List;
int *Array, len;
register int *ArrP;
register int i;

/* Count the number of elements in the list */

i = 0;
while ( List->Functor == listfno ) {
i++;
List = List->Next;
}
len = i;

Array = MEM_AllocGStack(i*sizeof(int));
ArrP = Array;

/* Transfer the elements from the list to the array */

List = SaveList;
while ( i != 0 ) {
*ArrP++ = List->Value;
List = List->Next;
i--;
}

*ResultArray = Array;
return(len);
}

Index 455
 void ArrayToList(register int *ArrP,register int n,
register INTLIST **ListPP)
{
while ( n != 0 ) {
*ListPP = MEM_AllocGStack(sizeof(INTLIST));
(*ListPP)->Functor = listfno;
(*ListPP)->Value = *ArrP++;
ListPP = &(*ListPP)->Next;
n--;
}
*ListPP = MEM_AllocGStack(sizeof((*ListPP)->Functor));
/* End of list */
(*ListPP)->Functor = nilfno;
}

void inclist(INTLIST *InList,INTLIST **OutList)

{
register int *ArrP, i, len;
int *Array;

len = ListToArray(InList,&Array);
ArrP = Array;
for ( i = 0; i < len; i++)
++*ArrP++;
ArrayToList(Array,len,OutList);
}

This program belongs to the kind where memory alignment can be critical. If you
intend to compile to several platforms, you're well advised to keep an eye on this.
As a first step, check that the sizes of the structures shared by C and Prolog are
the same; the padding applied when aligning on non-byte boundaries will make
things a bit bigger. The sizeof function comes in handy here. You can write a
small C function:
unsigned c_ilsize(void)
{
return(sizeof(INTLIST));
}

returning the size of the INTLIST structure. This can then be used by a Prolog
predicate to verify that the sizes of INTLIST and ilist are identical:
GLOBAL PREDICATES
unsigned c_ilsize() language C

PREDICATES
scheck
CLAUSES
scheck:- ILSize = sizeof(ilist), ILSize = c_ilsize(), !.
scheck:- write("ilist element sizes differ\n"), exit(1).

Calling Prolog from Foreign Languages

If you supply Prolog clauses for global predicates declared as being of foreign
language, those predicates may be called from foreign languages. They will have
parameter access and entry and exit code, including register preservation, as for
the language specified.

Hello
This small project is hello-world, with a twist.
/* Program hello_p.pro */

global predicates
char prowin_msg(string) - (i) language c
hello_c - language c

clauses
prowin_msg(S,C) :-
write(S," (press any key)"), readchar(C).

goal
prowin_msg("Hello from PDC Prolog"),
hello_c.

The global predicate prowin_msg is now accessible from C and can be called just
like any other C function:
/* Program hello_c.c */

char prowin_msg(char *);

void hello_c()
{
while ( prowin_msg("Hello from C (press 'C')") != 'C' )
;
}

As is evident, values may be returned to foreign languages.

Standard Predicates
Most of Visual Prolog's standard predicates can be called from C, but their public
names and exact functionality are subject to change without notice. It's therefore
Index 457
strongly recommended that you write a small set of interface routines if you want
to call Visual Prolog standard predicates from C. The following illustrates
bindings to a number of Visual Prolog's DOS Textmode I/O predicates:
/* Program spred_p.pro */

project "spred"

global predicates
myfail language c as "_fail"
mymakewindow(integer,integer,integer,string,integer,integer,
integer,integer)
- (i,i,i,i,i,i,i,i) language c as "_makewindow"
myshiftwindow(integer) - (i) language c as "_shiftwindow"
myremovewindow language c as "_removewindow"
write_integer(integer) - (i) language c as "_write_integer"
write_real(real) - (i) language c as "_write_real"
write_string(string) - (i) language c as "_write_string"
myreadchar(char) - (o) language c as "_readchar"
myreadline(string) - (o) language c as "_readline"

extprog language c

clauses
myfail:- fail.

mymakewindow(Wno, Wattr, Fattr, Text, Srow, Scol, Rows, Cols):-

makewindow(Wno, Wattr, Fattr, Text, Srow, Scol, Rows, Cols).

myshiftwindow(WNO):- shiftwindow(WNO).

myremovewindow:- removewindow.

write_integer(I):- write(I).

write_real(R):- write(R).

write_string(S):- write(S).

myreadchar(CH):- readchar(CH).

myreadline(S):- readln(S).

goal
extprog.

These may be accessed freely by C, as illustrated by extprog:

 /* Program spred_c.c */

void extprog(void)
{
char dummychar;
char *Name;

makewindow(1,7,7,"Hello there",5,5,15,60);
write_string("\n\nIsn't it easy");
readchar(&dummychar);
write_string("\nEnter your name: ");
readline(&Name);
write_string("\nYour name is: ");
write_string(Name);
readchar(&dummychar);
removewindow();
}

Calling an Assembler Routine from Visual Prolog

You can also call assembler routines from Visual Prolog. The activation record is
the same as for pascal (that is, parameters are pushed left to right), and the called
routine should pop the stack itself. If you have a C compiler supporting inline
assembler, things will be considerably easier than if you have to do everything
yourself.
In any case there seems to be little point in using assembler since C handles most
things, but a small example is included here for completeness. For obvious
reasons, the code differs between 16 and 32 bit platforms.
Suppose you want to write a routine returning a 32-bit sum of the characters in a
string, and also verifies that all characters are within a certain range, say A-Z.
The Prolog code for this could be:
/* Program csum_p.pro */

project "csum"

global predicates
integer sum_verify(char,char,string,ulong) - (i,i,i,o) language asm

predicates
uc_check(string)

Index 459
clauses
uc_check(S):-
0 = sum_verify('A','Z',S,Sum), !,
write('"',S,"\" OK, sum = ",Sum,'\n').
uc_check(S):- write('"',S,"\" fails\n").

goal
uc_check("UNIX"),
uc_check("Windows").

where we have adopted the convention that a return value of 0 means the string
was OK.
Here is the suitable 16-bit assembler code:
/* Program csum_a16.asm */

;/* Copyright (c) 1986, '92 by Prolog Development Center */

; 16-bit version

CSUM_A16_TEXT SEGMENT WORD PUBLIC 'CODE'

CSUM_A16_TEXT ENDS
_DATA SEGMENT WORD PUBLIC 'DATA'
_DATA ENDS
CONST SEGMENT WORD PUBLIC 'CONST'
CONST ENDS
_BSS SEGMENT WORD PUBLIC 'BSS'
_BSS ENDS
DGROUP GROUP CONST, _BSS, _DATA
ASSUME CS: CSUM_A16_TEXT, DS: DGROUP, SS: DGROUP

CSUM_A16_TEXT SEGMENT
ASSUME CS: CSUM_A16_TEXT

PUBLIC sum_verify
sum_verify PROC FAR
push bp
ov bp,sp

lolim equ 16
hilim equ 14
string equ 10
sum equ 6
 xor dx,dx
xor bx,bx ; Do sum in dx:bx
les di,[bp+string] ; Pointer to string
mov cl,byte ptr [bp+lolim] ; Low limit in cl
mov ch,byte ptr [bp+hilim] ; High limit in ch
xor ax,ax

ALIGN 2
loopy:
add bx,ax ; Add sum
adc dx,0
mov al,byte ptr es:[di]
inc di
cmp al,cl
jb end_check
cmp al,ch
jbe loopy

end_check:
or al,al
jnz go_home
les di,[bp+sum]
mov es:[di],bx
mov es:[di+2],dx
inc ax; ax: 0 -> 1

go_home:
dec ax ; ax: 1 -> 0, or 0 -> -1
mov sp,bp
pop bp
ret 12
sum_verify ENDP

CSUM_A16_TEXT ENDS
END

When writing assembler code, take special care that the sizes of things on the
stack follow the machine's natural word-size. This is 2 bytes on 16-bit machines
and 4 bytes on 32-bit machines. A good first attempt is to compile a dummy C
routine, with the correct parameters and local variables, to assembler, and then
use the entry, exit, and variable access code generated by the C compiler.
It isn't necessary to preserve any of the usual registers when foreign language
routines are called from Prolog, but if you're calling from C or assembler it's
assumed that you preserve si and di (esi and edi on 32-bit platforms). On 32-bit
platforms, ebx must also be preserved.

Index 461
Index

—8— approximate, 196

arc tangent, 192
8086, 384 arctan, 192
80x86 processor family, 442 arguments, 11, 22
arity and, 50
—A— compound data objects, 98
flow pattern, 216
abs, 191 input, 216
absolute, 191 known, 216
absolute values, 191 multiple-type, 110
Abstract, 212 output, 216
access reference domain, 220
error message files, 246 typing in predicate declarations,
external database via B+ trees, 47
343 arguments:, 60
hardware, 383 arithmetic, 186
memory, 386 expressions, 186
OS, 373 comparing, 195
access modes, 291 integer and real, 191
accessmode, 358 operations, 186
accessmode, 325 order of evaluation, 187
actions arities, 418
post- and pre-, 116 arity, 50
adding facts at run time, 174 arrays
addition, 186 code, 432
adventure game, 396 internal, 432
alignment ASCII, 96
memory, 426, 449 assembler
databases and pipes, 362, 450 routines
alternate solutions, 61 calling, 460
alternative, 126 assert, 174
alternatives to file domains, 415 asserta, 174
anonymous variables, 25, 425 assertz, 174
append, 160, 223, 403 assignment statements, 195
appending atoms, 97
lists, 159 attributes, 291
automatic type conversion, 51, 97, creating, 239
222, 316 size of, 240
unifying, 241
binding
—B— flowpatternstopredicatecalls, 216
B+ trees, 321, 339 bios, 384
bt_copyselector, 341 bitand, 381
closing, 341 bitleft, 382
creating, 340 bit-level operations, 381
deleting, 341 bitnot, 381
duplicate keys, 340 bitor, 382
internal pointer, 342 bitright, 383
internal pointers, 354 bitxor, 301, 382
key length, 339 bound variables, 425
multiple scans, 340 break, 251
order B+ trees Break Check menu item, 435
pages, 339 breakpressed, 252
statistics, 341 bt_close, 341
updating, 342 bt_copyselector, 341
backslash, 421 B+ trees, 341
backtrack point bt_create, 340
removing, 259 bt_delete, 341
backtracking, 61, 114 bt_open, 341
basic principles, 68 bt_selector, 221, 325, 416
point, 62 bt_statistics, 341
preventing with the cut, 78 bt_updated, 361
basic byte, 44, 424
concepts of Prolog, 10
program sections, 35
string-handling procedures, 307
—C—
beep, 379 C
bgi_ilist, 416 interface, 441
bgidriver compiler directive, 431 lists, 455
bgifont compiler directive, 431 passing lists to, 449
binary passing structures to, 448
domain, 238 routines
search trees, 139 calling, 444
trees calling conventions, 442
reference domains and, 224 calls
binary terms deterministic, 77
accessing, 240 non-deterministic, 77
comparing, 241 carriage return, 421
case conversion, 314 code compiler directive, 432
cast, 260 coercion
chain_delete, 333 type], 51
chain_first, 334 comline, 376
chain_inserta, 332 command line, 376
chain_insertafter, 333 comments, 29
chain_insertz, 332 comparing, 186
chain_last, 334 arithmetic expressions, 195
chain_next, 334 characters, 199
chain_prev, 334 strings, 199
chain_terms, 333 symbols, 199
chains compilation
deleting, 333 conditional, 429
inserting terms in, 332 compiler, 126
manipulating, 332 compiler directives, 54, 431
names of terms in, 333 bgidriver, 431
of terms, 323 bgifont, 431
char, 45, 96, 421 check_determ, 260, 265, 419
char_int, 313 code, 432
character fonts config, 433
linking, 431 determ, 173
characters, 96, 199 diagnostics, 433
comparing, 199 errorlevel, 434
converting to integers, 313 include, 54, 438
converting to strings, 314 nonbreak, 435
characters:, 96, 199 nowarnings, 436
check_determ, 126 struct, 448
check_determ compiler directive, composebinary, 240
260, 265, 419, 431, 437 compound
Class, 202, 203, 204, 208 data objects, 98
classes, 200 lists, 164
clauses, 18 mixed-domain declarations, 110
head, 421 objects, 425
Horn, 10 declaring domains, 105
nondeterministic, 431 unification, 99
non-deterministic, 265 compound flow pattern, 217
section, 36, 421 compound:, 148
clauses:, 162 compund object, 425
closefile, 282 concat, 310
closing concatenation
B+ trees, 341 strings, 308, 310
external database, 330 conditional compilation, 429
files, 281 conditional:, 51
config compiler directive, 433 —D—
CONFIG.SYS, 327
configuration file, 433 data
constants objects
declaring, 53 compound, 98
predefined, 428 security, 330
section, 53, 427 structures
Constructors, 210 recursive], 132
consult, 175, 249 types
consulterror, 249 trees, 133
controlling the flow analysis, 219 database
conversion global, 439
case, 314 internal, 171
character to integer, 313 declaring, 171
integer to character, 313 updating, 173
integer to string, 314 using, 172
numbers, 194 predicates
real to string, 314 restrictions, 172
single character to string, 314 reference numbers, 325, 334
string to character, 314 section, 52, 171, 420
string to integer, 314 database:, 52
string to real, 314 databases
type, 51 test of system (program), 335
conversion of types, 260 date, 102, 375
converting domains, 222 db_begintransaction, 359, 361, 370
copyfile, 290 db_btrees, 331
copying external databases, 328 db_chains, 331
cos, 192 db_close, 330
cosine, 192 db_copy, 328
counters, 120 db_create, 327
counting list elements, 152 db_delete, 330
cpunters, 128 db_endtransaction, 359, 361
criticalerror, 257 db_flush, 329
cursor predicate, 216 db_garbagecollect, 330
cutbacktrack, 259 db_loadems, 329
cuts, 78, 126 db_open, 328, 360
as goto, 90 db_openinvalid, 329
determinism and, 81 db_reuserefs, 326
dynamic, 259 db_saveems, 329
setting, 432 db_selector, 221, 325, 416
static, 259 db_setretry, 362
using, 78 db_statistics, 331
db_updated, 361

Index 465
dbasedom, 172, 416 difference lists, 165
declarations difftime, 378
accessmode, 358 dirclose, 297
as external name, 446 dirfiles, 298
B+ tree selectors, 325 dirmatch, 297
compound mixed-domain, 110 diropen, 296
constants, 427 disabling breaks, 251
database selectors, 325 discriminant, 196
denymode, 358 diskspace, 380
different domains, 414 displaying external database
domain, 39 contents, 347
domains as reference, 220 div, 191
domains of compound objects, dividing
105 words into syllables, 402
facts section, 171 division, 186
functions, 419 domains
global, 438 binary, 238
lists, 148 compound mixed, 110
local, 439 compound object, 415
predicate domains, 417 converting reference, 222
predicates, 36 db_selector, 323
predicates as deterministic, 419 dbasedom, 172
reference domains, 417 declarations, 39
typing arguments, 47 declaring, 414
declarative language, 10 declaring as reference, 220
default error routines, 257 external databases, 325
deletefile, 289 file, 415
deleting file, 284
B+ trees, 341 global, 439
chains, 333 internal, 420
external database, 330 list, 414
terms, 335 predefined, 416
denymode, 325, 358 predicate, 417
depth-first search, 136 ref, 326
Derived, 213 reference, 220, 417
Destroying, 204 reg, 384
Destructors, 210 section, 413
determ, 432 shortening declarations, 416
determinism, 77 specially handled, 416
cut and, 81 standard, 414
vs. non-determinism, 265 user-defined, 418, 425
deterministic predicates, 419 DOS
diagnostics compiler directive, 433 critical error, 257
double quotation marks, 422 constant identifier declarations,
drivers 428
linking, 431 consult, 249
dumpDba, 356 control in EXE files, 257
dumping external databases to text readterm, 250
file, 356 reporting
duplettes, 340 at run time, 247
duplicate keys run-time, 247, 434
in B+ trees, 340 term reader, 249
dword, 44, 424 trapping, 244
dynamic cutting, 259 escape character, 421
dynamic memory allocation, 454 example Prolog programs, 389
EXE files
error control in, 257
—E— existdir
elements of the language, 411 example, 297
elsedef, 429 existfile, 288
enabling breaks, 251 exit, 245
Encapsulation, 200 exp, 193
enddef, 429 expressions, 186
environment symbols, 375 order of evaluation, 187
envsymbol, 375 external, 320
eof, 286 name
equal declaration, 446
predicate, 195 predicates
sign declaring, 442
unifying compound objects, 99 external databases
equality, 195 accessing via B+ trees, 343
equality:, 196 accessmode, 358
erroneous, 230 B+ tree names, 331
Erroneous, 419 chain names, 331
error closing, 330
memory overflow, 224 copying, 328
error reporting, 247 creating, 327
errorcodes, 244 deleting, 330
reserved, 244 deleting chains, 333
errorlevel, 247 denymode, 358
errorlevel compiler directive, 434 displaying contents of, 347
errormsg, 246 domains, 325
errors dumping to text file, 356
constant definition recursion, 428 filesharing, 359, 370
file-sharing, 358
flushing, 329

Index 467
 index to, 339 file attributes, 291
inserting terms in, 332 file_bin, 279
invalid, 329 file_str, 276
location of, 327 fileattrib, 300
locking of, 360, 364, 370 fileerror, 258
log file, 349 filemode, 282
merging free space, 330 filenameext, 295
moving, 328 filenamepath, 294
non-breakdown, 349 filepos, 285
opening, 328 files
programming, 345 attributes, 290
RAM requirements, 320 closing, 281
reopening, 359 domain, 415
scanning through, 346 domains, 416
selectors, 322 dumped (external databases), 356
sharemode, 359 error message, 246
statistics, 331 external databases
structure of, 321 file-sharing, 358
system, 320 log (external databases), 349
transactions, 359 object, 441
updating, 350 opening, 281
external program, 373 symbolic file names, 422
filesharing, 370
predicates, 360
—F— transaction, 359
factorials:, 120 findall, 162
facts, 11, 420, 421 finding
adding at run time, 174 all solutions at once, 162
loading from files at run time, 175 flag, 384
removing, 176, 178 floating-point numbers, 422
saving at run time, 181 flow pattern, 31, 161, 216
section, 52 compound, 217
facts databases non-existent, 219
using, 172 flush, 288
facts sections flushing an external database, 329
databases fonts
updating, 173 linking, 431
facts:, 60 Formal, 215
fail, 76, 117 format, 311
failure, 230 formatted output
Failure, 419 examples, 273
FAR subroutines, 442 to string variable, 311
formatting arguments into a string, head of clause, 421
307 heap
free, 59, 454 allocation from C, 454
free variables, 425 heap compiler directive, 435
frontchar, 307 Heap menu item, 435
frontstr, 309 heapsize, 380
fronttoken, 308 hexadecimal numbers, 186
functions Horn clauses, 10
declaring, 419 hypertext, 146
return values, 227
sizeof, 453
functorless terms, 448 —I—
functors, 98, 415 I/O
ports, 386
redirecting, 284
—G— Identity, 201
games IEEE standard format, 97
adventures in a cave, 396 if/then, 51
Towers of Hanoi, 400 ifdef, 429
getbacktrack, 259 ifndef, 429
getbinarysize, 240 IMPLEMENT, 202
getentry in B+ trees
binary access, 240 names, 331
global in chains
databases, 439 names, 331
declarations, 438 in_file, 327
domains, 439 in_memory, 327
predicates, 439 include compiler directive, 54, 438
stack, 437 include file
global sections, 54 error.con, 244
goal trees, 67 including files in your program, 430
goals, 26 index
goals:, 43, 59 to external databases, 339
goto:, 90 inference engine, 10, 389
graphics infix
linking drivers, 431 predicate, 99
linking fonts, 431 infix:, 195
Inheritance, 201
input
—H— argument, 31
hardware simulation, 398 arguments, 216
head, 60 parameters, 443

Index 469
 redirecting, 284 keywords, 412
input:, 161
inserting terms in chains, 332
instantiating reference variables, 221 —L—
integer, 44, 424 large, 438
integers last-call, 122
arithmetic, 191 lasterror, 248
converting to characters, 313 length
converting to strings, 314 of a string, 310
division, 191 length:, 152
random, 189 less than, 195
interchangeability of unknowns, 155 linked lists, 449
interfacing with other languages, listdba, 347
441 lists, 111, 147
intermediate, 120 appending, 159
internal as compound objects, 426
databases, 171 compound, 164
using, 172 counting elements, 152
facts-section, 171 declaring, 148
goals, 413 defined, 147
pointer difference, 165
B+ trees, 342 domains, 414
string address, 385 handling, 455
system time clock, 375 length, 152
internal:, 52, 436 linked, 449
invalid external databases, 329 membership, 157
invertible, 162 mixed types, 426
isname, 310 passing to C, 449
IXREF statistics processing, 149
B+ trees, 341 recursion and, 147
using, 150
lists:, 147, 150
—K— ln, 193
key_current, 343 loading
key_delete, 342 facts from a file at run time, 175
key_first, 342 log, 193
key_insert, 342 log file
key_last, 342 external databases, 349
key_next, 343 logarithm, 193
key_prev, 343 logic program
key_search, 342 defined, 10
keyboard, 416 logical
 AND, 381 merging free space, 330
circuit, 398 mod, 191
inference, 10 modes
NOT, 381 access, 291
OR, 382 sharing, 291
XOR, 382 modular arithmetic, 191
long, 43, 424 modular programming, 438
loop variables, 128 modules, 438
loops moving external databases, 328
backtracking, 118 Multi, 419
lowercase multiple
in names, 411 arity, 50
low-level support, 383 solutions, 77
multiple-type arguments, 110
multiplication, 186
—M— multitasking, 358
macro definition, 427 mykey_next, 354
makebinary, 239 mykey_prev, 354
malloc, 454 mykey_search, 354
manipulation
chains, 332
external databases, 326
—N—
terms (external database, 334 N Queens problem, 405
marktime, 377 names, 411
matching, 30, 58 external database predicates, 321
mathematical, 188 predicates, 37
member, 158, 223, 418 redefining, 428
membyte, 386 restrictions in, 411
memdword, 386 terms in chains, 333
memory valid, 310
access, 386 naming conventions, 443
alignment, 426, 449 extended database predicates, 321
databases and pipes, 362, 450 natural
allocation, 451 logarithm, 193
dynamic, 454 NEAR subroutines, 442
freeing, 437 newline, 421
management, 437 nl, 267
overflow error, 224 nobreak compiler directive, 435
regaining (external databases), nondeterm, 432
330 non-determinism, 77
memword, 386 non-determinism vs. determinism,
menus:, 54 265

Index 471
non-deterministic clause warning, echoing to file or printer, 436
431 formatted to string variable, 311
nondeterministic predicates, 419 parameters, 443
not, 82, 389 redirecting, 284
nowarnings compiler directive, 436 output:, 161
numbers, 97 overflow
converting, 194 memory, 224
hexadecimal, 186
octal, 186
—P—
parameter-pushing, 442
—O— parameters
object files, 441 input, 443
objects, 11, 200 output, 443
compound, 425 parser, 316
octal numbers, 186 parsing, 168
openappend, 282 by different lists, 165
openfile, 292 pathname
opening in include files, 430
B+ trees, 341 pattern matcher, 10
external databases, 328 peeking, 386
files, 281 place, 221, 325, 416
invalid external database, 329 pointers
openmodify, 282 B+ trees (internal), 354
openread, 144, 281 stack, 445
openwrite, 144, 281 poking, 386
operands, 186 port_byte, 386
operations, 186 post-actions, 116
bit-level, 381 pre-actions, 116
operators, 186 predefined
precedence of, 187 domains, 416
relational, 195 file names, 416
order predefined constants, 428
B+ trees, 339 predicate logic, 10
of evaluation, 187 predicates, 22
OS arity, 50, 418
accessing from applications, 373 as arguments, 232
osversion, 379 C functions, 444
output declarations
argument, 31 typing arguments in, 47
arguments, 216 declaring as deterministic, 419
diagnostic, 433 equal, 195
 external, 442 error level in, 435
flow variants, 443 Prolog
global, 439 example programs, 389
implementing in other languages, fundamentals, 10
442 objects, 11
infix, 99 predicate logic syntax, 10
multiple declarations, 418 procedural perspective, 87
names, 37 relations, 11
number of arguments, 50 Protected, 213
section, 36, 418 ptr_dword, 385
specially handled, 412
predicates:, 36
preventing backtracking, 78 —Q—
Printer Menu in EXE File menu quadratic, 196
item, 436 queries, 13, 26
printermenu compiler directive, 436 questions, 13
procedural perspective, 87 quotation marks, 421
procedure parameters, 232
procedures, 12
program planning routes, 394 —R—
program sections, 412 random, 189
program structure random numbers
restrictions on, 413 generating, 189
programming initializing, 190
efficiency, 261 randominit, 190
external databases, 345 readblock, 279
modular, 438 readchar, 276
style, 261 readdevice, 144, 283
system-level, 373 reading
programs from I/O port, 386
different versions of same, 429 user-edited files, 249
large, 438 readint, 275
sections, 35 readln, 275
clauses, 36 readreal, 276
constants, 53 readterm, 249, 276, 301
domains, 39 readtermerror, 250
facts, 52 real, 45, 97
global, 54 arithmetic, 191
predicates, 36 converting to string, 314
stand-alone, 413 random, 189
programs:, 10, 43 recursion, 120, 147
project compiler directive, 436 from a procedural viewpoint, 160
projects, 438

Index 473
 lists and, 147 restrictions to using database
repitition and, 114 predicates, 172
recursive retract, 176
data structures, 132 retractall, 178
procedures, 120 RetryCount, 362
recursive:, 148 return values, 419
red cuts, 78 registers for, 443
ref, 325, 416 return values from functions, 227
ref_term, 335 round, 193
reference rounding, 191, 193
domains, 220, 417 route planning
binary trees and, 224 example, 394
sorting with, 225 rules, 11, 12, 421
trail array and, 221 as procedures, 87
numbers, 325 syntax, 50
variable, 220 using like case statements, 88
variables, 221 rules:, 60
Reference, 211 run-time
reg, 221, 416 error reporting, 434
reg domain, 384 run-time errors, 244, 247
register:, 384
registers
preserving, 458, 462 —S—
relational, 172 samekey_next, 355
relational operators, 195 samekey_prev, 355
relations, 11, 22 save, 181, 183, 249
removing saving
backtrack points, 259 facts at run time, 181
facts at run time, 176 scanner, 315
several facts at once, 178 scanning, 168
renamefile, 290 B+ trees, 340
repeat, 118 external databases, 346
repeat...fail, 438 scope
repetition constant identifiers, 428
recursion and, 114 predicates, 433
repetitive processes, 114 screen, 416
replacing terms, 334 search
reporting errors at run time, 247 database for record, 339
reserved words, 412 searchchar, 312
restrictions searchfile, 289
names, 411 searchstring, 313
program structure, 413 selectors
symbolic constants, 428
 external databases, 322 storage (predicate), 380
sentence structure, 109 str_char, 314
separators:, 150 str_int, 314
setentry str_len, 310
binary access, 240 str_real, 314
setting cuts, 432 string, 46
sharing modes, 291 string-handling, 307
short, 43, 424 strings, 97, 199, 422
signal, 252 blank spaces, 310
signals building, 307
enabling and disabling, 251 comparing, 199
signed:, 44 comparison, 199
simple constants, 421 concatenation, 308, 310
sin, 192 converting to other domains, 314
sine, 192 converting to term, 315
single, 230 creating blank, 307
single solutions, 77 dividing, 307
sizeof function, 453 internal address, 385
sleep, 377 length, 310
SleepPeriod, 362 manipulations, 307
solutions parsing, 316
controlling the search, 75 returning, 307
finding all at once, 162 verifying, 307
multiple, 77 length, 307
single, 77 struct compiler directive, 448
solutions:, 126 structure
sorting data, 425
tree-based, 141 external databases, 321, 356
with reference domains, 225 programs, 413
sound, 379 structures
sqrt, 193 passing to C, 448
square roots, 191, 193 style
stack programming, 261
pointer, 445 subchar, 311
standard subcomponents, 425
domains, 414, 421 of functors, 415
object files, 441 subgoals:, 61, 82
Static, 208 substring, 311
statistics subtraction, 186
external databases, 331 symbol, 46, 422
stderr, 416 symbolic
stdin, 416 constants, 53
stdout, 416 symbolic constants, 411, 422

Index 475
 restrictions on, 428 files
symbolic file names, 422 external databases dump, 356
symbols, 97, 199 This, 211
comparing, 199 time, 375
syntax timeout, 377
predicate logic, 10 totals, 120
rules, 50 trail
syspath, 376 array
system, 373 reference domains and, 221
system-level programming, 373 transcendental functions, 191
trap, 245
traversing trees, 135
—T— tree-based sorting, 141
tab, 421 trees
tail recursion, 154 as data type, 133
optimization, 122 binary search, 139
tan, 192 creating, 137
tangent, 188, 192 goal, 67
telephone directory, 48 traversing, 135
term trigonometry, 191
converting to string, 315 trunc, 194
location in chain, 325 truncation, 191, 194
term reader type
handling errors from, 249 coercion, 51
term_bin, 243 conversion, 51, 313
term_delete, 335 automatic, 316
term_replace, 334 type conversion, 260
term_str, 315 type implementation, 447
termination, 43 type variants, 418
terms, 425 type-checking, 177
alignment, 426, 449 typing arguments in predicate
binary conversion of, 242 declarations, 47
chains of, 323
deleting, 335
functorless, 448
—U—
manipulating, 334 ulong, 43, 424
reference number, 335 unbound variables, 417
replacing, 334 underscore symbol, 26
tests unification, 58, 425
external database system external, of compound objects, 99
335 unification:, 131
text unsigned, 44, 424
unsigned:, 44 variables:, 59
updating verifying the length of a string, 307
B+ trees, 342 version
external databases, 350 OS, 379
facts section, 173 Virtual, 206
upper_lower, 314 Visual Prolog
uppercase arithmetic, 186
in names, 411 external database, 320
user-defined handling strings, 307
domains, 425 strings in, 307
error routines, 257 Visual Prolog
ushort, 43, 424 internal facts section, 171
program sections, 35
—V—
value, 186
—W—
Variable Used Once menu item, 436 word, 44, 424
variables, 13, 16, 22, 425 write, 267
anonymous, 25, 425 write_a_list, 398
bound, 425 writeblock, 279
constants and, 428 writedevice, 144, 283
efficient programming with, 261 writef, 272
free, 425 writing
reference, 220 to I/O port, 386
unbound, 417

Index 477