GAP Manual: Preface.

GAP stands for Groups, Algorithms and Programming. The name was chosen to reflect the aim of the system, which is introduced in this manual.

Until well into the eighties the interest of pure mathematicians in computational group theory was stirred by, but in most cases also confined to the information that was produced by group theoretical software for their special research problems - and hampered by the uneasy feeling that one was using black boxes of uncontrollable reliability. However the last years have seen a rapid spread of interest in the understanding, design and even implementation of group theoretical algorithms. These are gradually becoming accepted both as standard tools for a working group theoretician, like certain methods of proof, and as worthwhile objects of study, like connections between notions expressed in theorems.

GAP was started as an attempt to meet this interest. Therefore a primary design goal has been to give its user full access to algorithms and the data structures used by them, thus allowing critical study as well as modification of existing methods. We also intend to relieve the user from unwanted technical chores and to assist him in the programming, thus supporting invention and implementation of new algorithms as well as experimentation with them.

We have tried to achieve these goals by a design which in addition makes GAP easily portable, even to computers such as Atari ST and Amiga, and at the same time facilitates the maintenance of GAP with the limited resources of an academic environment.

While I had felt for some time rather strongly the wish for such a truly open system for computational group theory, the concrete idea of GAP was born when, together with a larger group of students, among whom were Johannes Meier, Werner Nickel, Alice Niemeyer, and Martin Schönert who eventually wrote the first version of GAP, I had my first contact with the Maple system at the EUROCAL meeting in Linz/Austria in 1985. Maple demonstrated to us the feasibility of a strong and efficient computer algebra system built from a small kernel, with an interpreted library of routines written in a problem-adapted language. The discussion of the plan of a system for computational group theory organized in a similar way started in the fall of 1985, programming only in the second half of 1986. A first version of GAP was operational by the end of 1986. The system was first presented at the Oberwolfach meeting on computational group theory in May 1988. Version 2.4 was the first officially to be given away from Aachen starting in December 1988. The strong interest in this version, in spite of its still rather small collection of group theoretical routines, as well as constructive criticism by many colleagues, confirmed our belief in the general design principles of the system. Nevertheless over three years had passed until in April 1992 version 3.1 was released, which was followed in February 1993 by version 3.2, in November 1993 by version 3.3 and is now in June 1994 followed by version 3.4.

A main reason for the long time between versions 2.4 and 3.1 and the fact that there had not been intermediate releases was that we had found it advisable to make a number of changes to basic data structures until with version 3.1 we hoped to have reached a state where we could maintain upward compatibility over further releases, which were planned to follow much more frequently. Both goals have been achieved over the last two years. Of course the time has also been used to extend the scope of the methods implemented in GAP. A rough estimate puts the size of the program library of version 3.4 at about sixteen times the size of that of version 2.4, while for version 3.1 the factor was about eight. Compared to GAP 3.2, which was the last version with major additions, new features of GAP 3.4 include the following:

New data types (and extensions of methods) for algebras, modules and characters

Further methods for working with finite presentations (IMD, a fast size function)

Some "Almost linear", methods and (rational) conjugacy classes for permutation groups

Methods based on "special AG systems", for finite soluble groups

A package for the calculation of Galois groups and field extensions

Extensions of the library of data (transitive permutation groups, crystallographic groups)

An X-window based X-GAP for display of subgroup lattices

Five further share libraries (ANU SQ, MEATAXE, SISYPHOS, VECTORENUMERATOR, SMASH)

Work on the extension of GAP is going on in Aachen as well as in an increasing number of other places. We hope to be able to have the next release of GAP after about 9 months again, that is in the first half of 1995.

The system that you are getting now consists of four parts:

A comparatively small kernel, written in C, which provides the user with:

automatic dynamic storage management, which the user needn't bother about in his programming;

a set of time-critical basic functions, e.g. "arithmetic", operations for integers, finite fields, permutations and words, as well as natural operations for lists and records;

an interpreter for the GAP language, which belongs to the Pascal family, but, while allowing additional types for group theoretical objects, does not require type declarations;

a set of programming tools for testing, debugging, and timing algorithms.

A much larger library of GAP functions that implement group theoretical and other algorithms. Since this is written entirely in the GAP language, in contrast to the situation in older group theoretical software, the GAP language is both the main implementation language and the user language of the system. Therefore the user can as easily as the original programmers investigate and vary algorithms of the library and add new ones to it, first for own use and eventually for the benefit of all GAP users. We hope that moreover the structuring of the library using the concept of domains and the techniques used for their handling that have been introduced into GAP 3.1 by Martin Schönert will be further helpful in this respect.

A library of group theoretical data which already contains various libraries of groups (cf. chapter "Group Libraries"), large libraries of ordinary character tables, including all of the Cambridge Atlas of Finite Groups and modular tables (cf. chapter "Character Table Libraries"), and a library of tables of marks. We hope to extend this collection further with the help of colleagues who have undertaken larger classifications of groups.

The documentation. This is available as a file that can either be used for on-line help or be printed out to form this manual. Some advice for using this manual may be helpful. The first chapter About GAP is really an introduction to the use of the system, starting from scratch and, for the beginning, assuming neither much knowledge about group theory nor much versatility in using a computer. Some of the later sections of chapter 1 assume more, however. For instance section "About Character Tables" definitely assumes familiarity with representation theory of finite groups, while in particular sections "About the Implementation of Domains" to "About Defining New Group Elements" address more advanced users who want to extend the system to meet their special needs. The further chapters of the manual give then a full description of the functions presently available in GAP.

Together with the system we distribute GAP share libraries, which are separate packages which have been written by various groups of people and remain under their responsibility. Some of these packages are written completely in the GAP language, others totally or in parts in C (or even other languages). However the functions in these packages can be called directly from GAP and results are returned to GAP. At present there are 10 such share libraries (cf. chapter "Share Libraries").

The policy for the further development of GAP is to keep the kernel as small as possible, extending the set of basic functions only by very selected ones that have proved to be time-critical and, wherever feasible, of general use. In the interest of the possibility of exchanging functions written in the GAP language the kernel has to be maintained in a single place which in the foreseeable future will be Aachen. On the other hand we hoped from the beginning that the design of GAP would allow the library of GAP functions and the library of data to grow not only by continued work in Aachen but, as does any other part of mathematics, by contributions from many sides, and these hopes have been fulfilled very well.

There are some other points to make on further policy:

When we began work on GAP the typical user that we had in mind was the one wanting to implement his own algorithmic ideas. While we certainly hope that we still serve such users well it has become clear from the experience of the last years that there are even more users of two different species, on the one hand the established theorist, sometimes with little experience in the use of computers, who wants an easily understandable tool, on the other hand the student, often quite familiar with computers, who wants to get assistance in learning the theory by being able to do nontrivial examples. We think that in fact GAP can well be used by both, but we realize that for each a special introduction would be desirable. We apologize that we have not had the time yet to write such, however have learned (through the GAP forum) that in a couple of places work on the development of Laboratory Manuals for the use of GAP alongside with standard Algebra texts is undertaken.

When we began work on GAP, we designed it as a system for doing group theory. It has already turned out that in fact the design of the system is general enough, and some of its functions are also useful, for doing work in other neighbouring areas. For instance Leonard Soicher has used GAP to develop a system GRAPE for working with graphs, which meanwhile is available as a share library. We certainly enjoy seeing this happen, but we want to emphasize that in Aachen our primary interest is the development of a group theory system and that we do not plan to try to extend it beyond our abilities into a general computer algebra system.

Rather we hope to provide tools for linking GAP to other systems that represent years of work and experience in areas such as commutative algebra, or to very efficient special purpose stand-alone programs. A link of this kind exists e.g. to the MOC system for the work with modular characters.

We invite you to further extend GAP. We are willing either to include such extensions into GAP or to make them available through the same channels as GAP in the form of the above mentioned share libraries. Of course, we will do this only if the extension can be distributed free of charge like GAP. The copyright for such share libraries shall remain with you.

Finally to answer an often asked question: The GAP language is in principle designed to be compilable. Work on a compiler is on the way, but this is not yet ready for inclusion with this release.

GAP is given away under the conditions that have always been in use between mathematicians, i.e. in particular completely in source and free of charge. We hope that the possibility offered by modern technology of depositing GAP on a number of computers to be fetched from them by ftp, will assist us in this policy. We want to emphasize, however, two points. GAP is not public domain software; we want to maintain a copyright that in particular forbids commercialization of GAP. Further we ask that use of GAP be quoted in publications like the use of any other mathematical work, and we would be grateful if we could keep track of where GAP is implemented. Therefore we ask you to notify us if you have got GAP, e.g., by sending a short e-mail message to gap@dcs.st-and.ac.uk. The simple reason, on top of our curiosity, is that as anybody else in an academic environment we have from time to time to prove that we are doing meaningful work.

We have established a GAP forum, where interested users can discuss GAP related topics by e-mail. In particular this forum is for questions about GAP, general comments, bug reports, and maybe bug fixes. We will read this forum and answer questions and comments, and distribute bug fixes. Of course others are also invited to answer questions, etc. We will also announce future releases of GAP in this forum.

To subscribe send an e-mail message to miles@dcs.st-and.ac.uk containing the line

   subscribe gap-forum your-name

miles@dcs.st-and.ac.uk also accepts the following requests. help for a short help on how to use miles, unsubscribe gap-forum to unsubscribe, recipients gap-forum to get a list of subscribers, and statistics gap-forum to see how many e-mail messages each subscriber has sent so far.

The reliability of large systems of computer programs is a well known general problem and, although over the past year the record of GAP in this respect has not been too bad, of course GAP is not exempt from this problem. We therefore feel that it is mandatory that we, but also other users, are warned of bugs that have been encountered in GAP or when doubts have arisen. We ask all users of GAP to use the GAP forum for issuing such warnings.

We have also established an e-mail address gap-trouble@dcs.st-and.ac.uk to which technical problems of a more local character such as installation problems can be sent. Together with some experienced GAP users abroad we try to give advice on such problems.

GAP was started as a joint Diplom project of four students whose names have already been mentioned. Since then many more finished Diplom projects have contributed to GAP as well as other members of Lehrstuhl D and colleagues from other institutes. Their individual contributions to the programs and to the manual are documented in the respective files. To all of them as well as to all who have helped proofreading and improving this manual I want to express my thanks for their engagement and enthusiasm as well as to many users of GAP who have helped us by pointing out deficiencies and suggesting improvements. Very special thanks however go to Martin Schönert. Not only does GAP owe many of its basic design features to his profound knowledge of computer languages and the techniques for their implementation, but in many long discussions he has in the name of future users always been the strongest defender of clarity of the design against my impatience and the temptation for "quick and dirty", solutions.

Since 1992 the development of GAP has been financially supported by the Deutsche Forschungsgemeinschaft in the context of the Forschungsschwerpunkt ``Algorithmische Zahlentheorie und Algebra''. This very important help is gratefully acknowledged.

As with the previous versions we send this version out hoping for further feedback of constructive criticism. Of course we ask to be notified about bugs, but moreover we shall appreciate any suggestion for the improvement of the basic system as well as of the algorithms in the library. Most of all, however, we hope that in spite of such criticism you will enjoy working with GAP.

Aachen, June 1., 1994, Joachim Neubüser.