Why did I do it?

For fun. As an excercise in unix programming. And there was no such thing. See below, design objectives.

Design objectives

With two elaborate standards at hand, one for C, one for Forth, it should be possible to build one language in terms of the other and thus provide both where at least one is available.

While I leave the writing of an ANSI-C compiler in Forth to those who really believe in Forth's superiority and universality, I concentrated on the manageable part: Providing a Forth-system in ANSI-C that is

CORRECT	this means not just "no bugs" -- it means all the interaction of all parts works as specified by the standard AND by traditions as far as I know about them.
COMPLETE	with powerful computers on the desk and power- ful programming environments there is no need to and no use in strategies like "here kernel, there xyz word set" that might be appropriate on a microcontroller.
PORTABLE	I hate to rewrite code because of environmental dependencies. Most of the system should be a mere interface between two well defined languages. There are of course such dependencies. I tried to resort to a common subset of all UNIX bahavior known to me. Fortunately this in fact allowed porting to DOS and OS/2 thanks to the great emx-package.
USABLE	A sole kernel helps noone. It should be a development environment. You should be able to manage multiple-source-file-projects from within.
SIMPLE	or transparent. At least to my taste.

Maybe you miss the design objective SPEED. It was not my goal to provide the fastest C-based Forth-environment. This would have led to conflicts with more important goals. We all have fast computers, haven't we?

After all pfe isn't slow. With a little tuning using GNU-C's global register variable feature it is pretty fast. On an ix86-based system it seems to run with about two thirds the speed of a direct threaded assembler Forth (eforth) i.e. benchmarks take 50% more time.

Status

Did I achieve the above objectives? Some of them. I'll continue -- slowly -- working on it.

The system is in use for eighteen month now. Several quite obvious and several quite subtle bugs have been fixed. While there surely are some more, they are not as obvious since the system passes several test programs, some of them rather sophisticated.

Once you get the system running, you'll have

• All ANSI Forth word sets with the exception of the assembler hooks `CODE', `END-CODE' etc. Nothing else is missing.
• Several more words provided for compatibility and convenience.
• An interface to a text editor to edit plain text sources and the ability to include them.
• A word star lookalike block file editor to write source in the old style block format.
• A simple source level debugger -- maybe I should better say "single stepper".

Thus you'll be able to edit, compile an run programs in a moderately comfortable way. For the final design of the development environment your suggestions still are welcome!

If you try this system, please keep in mind that it is still under development. Sometimes new --even stupid-- bugs are introduced while enhancing functionality or while fixing old ones. I appreciate every hint to a bug and I fixed every bug I've been told about in the last months. So please don't hesitate to tell me about whatever seems wrong. Please check for the latest version via anonymous ftp from
roxi.rz.fht-mannheim.de:/pub/languages/forth/pfe-?.?.?.tar.gz
(accessible from germany only) or
sunsite.unc.edu:/pub/linux/devel/forth/pfe-?.?.?.tar.gz
Please send suggestions and bug reports via e-mail to
duz@roxi.rz.fht-mannheim.de

UPDATE: The new PFE can be found at guidod@gmx.de
(the new website from Dirk-Zoller can be found at www.dirk-zoller.de )

Usage

For installation refer to the files `INSTALL' and `TUNING'.
(update: the new pfe uses autoconf/automake, i.e. gnu-style)

Once you have it running and see the "ok" prompt after typing return you can interactively type in forth words. If you mistype, you can edit the command line and recall old command lines with the arrow keys. (If you can't then your termcap doesn't work all right and you can resort to wordstar-like control keys.)

To write some more statements try "EDIT-TEXT filename". This will invoke your favorite text-file editor on the given file. If it doesn't, first check the environment variable EDITOR, then check the file "const.h" for the #defined symbol "EDITOR".
(update: the new pfe has the editor not builtin by default)

Having written some code you can load it by "INCLUDE filename".

If you prefer the old style block files, give a file to use as block-file with the -b commandline option. Alternatively you can say `USING filename' or if the file doesn't exist yet: `USING-NEW filename'. Then you can edit a block by `n EDIT-BLOCK'.

If your termcap-mechanism works well, the arrow keys and some other function keys should be active. Quit the editor with ^U and load blocks with "n LOAD". If it doesn't work well you might not even get a picture.

For more commandline options try the option -h.

The interrupt key is remapped to ^U (normally it's ^C) and leads back to the FORTH input loop. Use it to break out of infinite loops. If you find the ^U annoying see src/options.h for how to change it. I needed the ^C for the Wordstar look-and-feel.

To terminate the system, type BYE at the command-prompt or press the keyboard quit key of your system (usually ^\).

Documentation

I started writing a documentation in texinfo format. This will allow you to view it online or print it in good quality. There's also an outdated and hopefully soon superseded man-page that explains some of the command line options. All documentation is highly unfinished.
(update: the new pfe uses docbook format to achieve the same goals)

For more information please try to get the dpANS-document, which is an EXCELLENT REFERENCE to this system! You can (could?) ftp it at ftp.uu.net in the directory /vendor/minerva/x3j14.

Recently the ANSI standard (or the last draft) bacame available as Word document and in HTML format. From a post of Norm Smith to c.l.f:

"I have uploaded a new HTML version of the DPANS94 document to Taygeta.oc.nps.navy.mil. The contents are identical to the original posting. The new version, V2.0, has many more hyperlinks. Specifically, all (well, at least most :-) of the See: references are now linked. Anchors have been added for each section in my working copy. This will faculitate filling in many additional links in the next version of the document."

(update: the best place to view the dpans is the FIG's website at forth.org - see forth.org/dpans/dpans.html htmlpage)

Contributions

are welcome! After the kernel is finished now it makes much sense to share the burden of creating and improving a well rounded programming environment for all tasks a Forth programmer wants to do. PFE is YOUR tool. Get involved in it's design!

There is a mailing list on pfe which you should subscribe to if you want to be up to date with the development of the system. Send e-mail to duz@roxi.rz.fht-mannheim.de if you want to subscribe to that mailing list.
(update: there is a new maintainer and a new mailinglist at pfe.sourceforge.net )

Acknowledgements

I want to express my gratitude to the people who put their efforts in the precise descriptions I found in these documents:

• draft proposed American National Standard -- Programming Languages -- Forth (X3J14 dpANS-6, June 30, 1993)
• FORTH-83 Standard (August 1983)
• fig-FORTH Installation Manual -- Glossary, Model, Editor (Version 1.3, November 1980)

Thanks for providing superb development tools:

• Richard Stallman and the FSF, Linus Thorvalds and all who contribute to Linux and Eberhard Mattes.

Several nice people on the net continuously gave me valuable hints to bugs and possible improvements, were patient enough to try the very first releases, made ports to machines I never saw myself and thus kept me from frustration. If pfe is stable now it is thanks to it's users. Most notably:

Lennert Benschop, Sean Conner, Holger Dietze, Kevin Haddock, Rob Hooft, Giorgio Richelli, Marko Teiste, Guenther Thomsen.

Thank You.