LEVELS OF FORTH

Glen B. Haydon
Mountain View Press
Box 429 Route 2
La Honda, CA 94020


This paper was presented at the 1991 Rochester Forth Conference.

Abstract

Match the level of Forth with the user and the application.

Introduction

Ontogeny recapitulate phylogeny. I guess it was John Dewey who emphasized that you start a student where he is. The needs of novices perhaps evolve in somewhat the same way as the history. If you are already at the end, you have no place to go.

A problem people in the Forth community face is the variety of applications which lend themselves to the language and the diversity of experience of the potential Forth users. One way to address the problems is to identify progressive levels of Forth with which to match the users and the applications. These Levels follow the history.

Level 0.

On several occasions C.H. Moore has listed approximately 63 functions which he considers the essence of Forth. I noted them at one presentation as follows:

+ - * /MOD MIN MAX = AND OR XOR

NEGATE ABS NOT */ DUP DROP SWAP OVER

DECIMAL HEX OCTAL . n .R

CR EMIT KEY : ; CREATE , ALLOT

IF ELSE THEN FOR NEXT I

There are only 45 functions in this list. You can see about which period this was. The list does not include BEGIN, UNTIL, and several other functions belong in Level 0. A figure of about 63 is appropriate.

Three functions are used for output to the display. He included no functions for any kind of storage. Screens are not included much less any means of writing on such a screen.

Something more than Level 0 is necessary to provide anyone but Chuck enough to work with.

Level 1.

The fig-Forth Installation Manual provides a complete set of Forth functions. The implementation with the Installation Manual is for the 6502. Since then, over a dozen implementations of fig-Forth for other processors were written and are still available in hard copy. About forty hours of careful typing will allow one to enter a listing.

At that time we lacked an editor. It was a chicken and egg problem. Though one had screens for compiling code there was no way to write to a screen until an editor was included. A simple single function editor allowed one to place code on a specific line of an already listed screen.

The fig-Forth model served as a wonderful learning tool. There was little in the way of documentation beyond the Installation Manual and the source listings. A few computer teaching programs did appear. The older publications such as the Kitt Peak Primer and Using Forth did not mesh well with the public domain version available.

Execellent applications have been done in fig-Forth.

The fig-Forth implementations took over the disk precluding its use for other programs. The error messages were stored on a disk.

Soon there came the 79 Standard which made several changes. It was then proposed that the Forth Interest Group stop supplying the fig-Forth source listings.

Level 2.

The 79 Standard made about 40 changes to the functions in fig-Forth. These made some things much easier. About the same time Leo Brody came out with the first edition of Starting Forth. This is an excellent book for a beginner. He developed the first part of the book in a very careful and artistic manner. Then he had to get on with the job.

At that time, Mountain View Press came out with a Forth implementation incorporating the 79 Standard changes. It put the error messages in line and added supplemental functions so that all but one of the functions matched those covered in Starting Forth. The single problem was 'tic'.

The assembler for several processors was included. A Forth editor described in Chapter 3 of Starting Forth written by Sam Daniels is included. Note: Starting Forth does not point out that you have to invoke the EDITOR vocabulary for it to work!

The complete source listing of MVP Forth was published for three processors. The appropriate source is included on each distribution disk. The Glossary, ALL ABOUT FORTH 2nd Edition, included the functional definitions, source code for the 8080, examples and comments for each of about 240 functions. The 3rd. Edition of ALL ABOUT FORTH has been expanded to include some 500 functions in common usage.

The distribution Version 03 of MVP Forth has remained unchanged since April 1983. Unfortunately several copies which have appeared on bulletin boards while being designated MVP Forth have been modified. The original Version 3 is now available of the Gene Forth Round Table and the Mountain View Press bulletin board.

The MVP Forth implementation takes over a disk format. But the file on Gene includes several added utilities which allow one to access screens in files and the compiling of text files written with any word processor.

Most of the Forth text books and tutorials begin with a discussion of stacks. After all it is the stacks which makes Forth what it is. But for most beginners stacks are a complication and a distraction. In the FORTH GUIDE, a different introduction is used.

The hardware designer and engineer have other interests besides learning yet another language. A small language which allows him to get on with his work would be sufficient. Levels 1 and 2 are closely bound to the hardware and serves such programmers well.

Level 3.

Several developments led to a new implementation. The 83 Standard was finally adopted. Few people had a computer dedicated to Forth. They needed their computer for other things. There was also a growing library of Forth functions which could be incorporated.

F83, by Henry Laxon and Mike Perry, took the next step. F83 includes nearly 1000 functions. A substantial library of Forth functions was added. Ting's book, Inside F83, provided much documentation.

The most popular implementation of F83 is for the Intel based systems. It makes use of screens but now in files. A dedicated disk is no longer necessary and programs could be developed and run from a hard disk. Many of the functions were optimized assembler code which improved the speed. With the library of functions, we were on the way to an implementation of Forth for programmers. Ting's book, Inside F83, was a big help in printed documentation.

But F83 with its increased vocabulary presented a greater learning hurdle for the engineer and hardware type who had no interest in programming. There was much more to learn than was necessary for him. Of course, if he learned the language, he had more power available to him.

Level 4.

With a programmers demand for a more complete development system Tom Wempe developed his PADS system. At the same time, Tom Zimmer with many collaborators, went through many verisions of what became known as FPC>.

FPC is a full Forth Development System with a built in editor, debugger and many other features making it an outstanding package. However, the vocabulary has grown to nearly 2500 functions - A real learning challenge.

All of the source code is in text files. In fact, there are more source and help files in FPC than there are functions in fig-Forth. But the many features do not detract from speed for the experienced user. All possible speed is milked out of the processor. When added to the TCOM program small target compiled applications can be made.

The program and files are large. One needs a couple of megabytes of disk space available. To do all that it does, the space is needed. For the professional programmer, FPC is an excellent model. Ting has provided us the F-PC Technical Reference Manual and the F-PC Users Manual.

But for the engineer and hardware designer who is not interested in programming languages, the learning curve has increased. It may be more than he wants. However, with such a development system, Tom Zimmer has shown that hardware problems can be easily solved.

Level 5.

For many purposes, 32-bit values are overkill, but they are the coming processors. The segmented architecture of the Intel processors has led to severe limitations. Various schemes of memory management have been developed to work around some of those limitations. A 32-bit Forth system should have 32-bit stacks and address space. Some means should be found to deal efficiently with 16-bit addresses in the segmented architecture.

F32, by Rick VanNorman, solves some of these issues. He has created a full development package including a text editor and assembler in about 900 words. This has been a outgrowth of his years of experience with Forth. He has placed the program in the public domain and is available from several sources including the Mountain View Press bulletin board.

His code is neat and well organized providing reasonable documentation for an experienced Forth programmer. However, there is as yet no printed documentation available. I expect that there will be some before long. This implementation is not intended for a beginning Forth programmer.

Discussion.

These Levels are oriented toward systems based on Intel processors. Most of them operate under a PC-DOS or MS-DOS environment. Other systems also make use of Forth. The Sun work station is built with a Forth monitor, and runs Unix. The Macintosh has available several Forth implementations. A Forth like Yerkes evolved out of NEON and runs on the Macintosh. These implementations are not for the beginner who wants a small language to get started. World wide there are still more implementations of Forth.

This is not to ignore the many professional Forth implementations. I consider these levels of Forth as models. Depending on ones interest and time, he can delve as deeply as he wishes.

For professional programmers, the professionally written implementations are well honed. They provide full development platforms. The big advantage of the professional systems is the many extensions and the level of documentation and support available from the respective companies.

The professional products include an appropriate tutorial. I have read the documentation and tried a number of the professional implementations. They are good. Having an interest in implementation as well, I find the hidden proprietary nature of most of them to be frustrating. For a programmer these works offer the professional touch not found in the public domain implementations.

The problem then boils down to determining the programming level of the user and his applications. Select a tutorial or instructional manual and work through it. The order in each is different. Trying to learn from several at once will only confuse the student. Don't demand that the hardware designer and engineer be master computer scientists as well. Don't make the computer scientist work with minimal amateurish tools.

I think that there is a niche for Forth. The full blown development systems, whatever the language used, is overwhelming for many engineer and hardware designers. The interactive capability of Forth supplies these users a rapid means of progress. A small set of functions which meet his needs is sufficient.

All users are not equal.

Addendum:

Added and Modified 24 March 1995
Since this article was written, in 1994 an ANSI Forth Standard has been adopted. This dialect is the result of many man years of work by experienced Forth programmers. A number of people have attempted to implement this dialect and in 32 bit address space for the PC and clones.

Unfortunately, the ANSI Standard Document is priced at nearly $300.0 (sic) and will be hard for may Forth users to justify. However, a hypertext copy of the final draft before adoption is available.

Also there is an implementation, THIS4TH, which is available in C source code and complied to a number of systems. These are also available.

Another implementation which is near the ANSI Standard Forth for use on PC's in 32 bit address space using DPMI. DPMI is available when shelling out from WINDOWS. This was written by Rick van Norman and is contained in S4.ZIP.

There will be other public domain implementations and these will all be tested by use. Eventually, there will probably appear selected versions and some documentation will be written.

The goal of these Forth Levels and these comments is to provide the beginner with something of a perspective.

---

Glen B. Haydon, M.D.
ghaydon@theforthsource.com