1: \input texinfo @c -*-texinfo-*-
2: @comment The source is gforth.ds, from which gforth.texi is generated
3: @comment %**start of header (This is for running Texinfo on a region.)
4: @setfilename gforth.info
5: @settitle Gforth Manual
6: @dircategory GNU programming tools
7: @direntry
8: * Gforth: (gforth). A fast interpreter for the Forth language.
9: @end direntry
10: @comment @setchapternewpage odd
11: @macro progstyle {}
12: Programming style note:
13: @end macro
14: @comment %**end of header (This is for running Texinfo on a region.)
15:
16: @include version.texi
17:
18: @ifinfo
19: This file documents Gforth @value{VERSION}
20:
21: Copyright @copyright{} 1995-1998 Free Software Foundation, Inc.
22:
23: Permission is granted to make and distribute verbatim copies of
24: this manual provided the copyright notice and this permission notice
25: are preserved on all copies.
26:
27: @ignore
28: Permission is granted to process this file through TeX and print the
29: results, provided the printed document carries a copying permission
30: notice identical to this one except for the removal of this paragraph
31: (this paragraph not being relevant to the printed manual).
32:
33: @end ignore
34: Permission is granted to copy and distribute modified versions of this
35: manual under the conditions for verbatim copying, provided also that the
36: sections entitled "Distribution" and "General Public License" are
37: included exactly as in the original, and provided that the entire
38: resulting derived work is distributed under the terms of a permission
39: notice identical to this one.
40:
41: Permission is granted to copy and distribute translations of this manual
42: into another language, under the above conditions for modified versions,
43: except that the sections entitled "Distribution" and "General Public
44: License" may be included in a translation approved by the author instead
45: of in the original English.
46: @end ifinfo
47:
48: @finalout
49: @titlepage
50: @sp 10
51: @center @titlefont{Gforth Manual}
52: @sp 2
53: @center for version @value{VERSION}
54: @sp 2
55: @center Anton Ertl
56: @center Bernd Paysan
57: @center Jens Wilke
58: @sp 3
59: @center This manual is permanently under construction
60:
61: @comment The following two commands start the copyright page.
62: @page
63: @vskip 0pt plus 1filll
64: Copyright @copyright{} 1995--1998 Free Software Foundation, Inc.
65:
66: @comment !! Published by ... or You can get a copy of this manual ...
67:
68: Permission is granted to make and distribute verbatim copies of
69: this manual provided the copyright notice and this permission notice
70: are preserved on all copies.
71:
72: Permission is granted to copy and distribute modified versions of this
73: manual under the conditions for verbatim copying, provided also that the
74: sections entitled "Distribution" and "General Public License" are
75: included exactly as in the original, and provided that the entire
76: resulting derived work is distributed under the terms of a permission
77: notice identical to this one.
78:
79: Permission is granted to copy and distribute translations of this manual
80: into another language, under the above conditions for modified versions,
81: except that the sections entitled "Distribution" and "General Public
82: License" may be included in a translation approved by the author instead
83: of in the original English.
84: @end titlepage
85:
86:
87: @node Top, License, (dir), (dir)
88: @ifinfo
89: Gforth is a free implementation of ANS Forth available on many
90: personal machines. This manual corresponds to version @value{VERSION}.
91: @end ifinfo
92:
93: @menu
94: * License::
95: * Goals:: About the Gforth Project
96: * Other Books:: Things you might want to read
97: * Invoking Gforth:: Starting Gforth
98: * Words:: Forth words available in Gforth
99: * Tools:: Programming tools
100: * ANS conformance:: Implementation-defined options etc.
101: * Model:: The abstract machine of Gforth
102: * Integrating Gforth:: Forth as scripting language for applications
103: * Emacs and Gforth:: The Gforth Mode
104: * Image Files:: @code{.fi} files contain compiled code
105: * Engine:: The inner interpreter and the primitives
106: * Cross Compiler:: The Cross Compiler
107: * Bugs:: How to report them
108: * Origin:: Authors and ancestors of Gforth
109: * Word Index:: An item for each Forth word
110: * Concept Index:: A menu covering many topics
111:
112: --- The Detailed Node Listing ---
113:
114: Forth Words
115:
116: * Notation::
117: * Arithmetic::
118: * Stack Manipulation::
119: * Memory::
120: * Control Structures::
121: * Locals::
122: * Defining Words::
123: * Structures::
124: * Object-oriented Forth::
125: * Tokens for Words::
126: * Wordlists::
127: * Files::
128: * Including Files::
129: * Blocks::
130: * Other I/O::
131: * Programming Tools::
132: * Assembler and Code Words::
133: * Threading Words::
134:
135: Arithmetic
136:
137: * Single precision::
138: * Bitwise operations::
139: * Mixed precision:: operations with single and double-cell integers
140: * Double precision:: Double-cell integer arithmetic
141: * Floating Point::
142:
143: Stack Manipulation
144:
145: * Data stack::
146: * Floating point stack::
147: * Return stack::
148: * Locals stack::
149: * Stack pointer manipulation::
150:
151: Memory
152:
153: * Memory Access::
154: * Address arithmetic::
155: * Memory Blocks::
156:
157: Control Structures
158:
159: * Selection::
160: * Simple Loops::
161: * Counted Loops::
162: * Arbitrary control structures::
163: * Calls and returns::
164: * Exception Handling::
165:
166: Locals
167:
168: * Gforth locals::
169: * ANS Forth locals::
170:
171: Gforth locals
172:
173: * Where are locals visible by name?::
174: * How long do locals live?::
175: * Programming Style::
176: * Implementation::
177:
178: Defining Words
179:
180: * Simple Defining Words::
181: * Colon Definitions::
182: * User-defined Defining Words::
183: * Supplying names::
184: * Interpretation and Compilation Semantics::
185:
186: Structures
187:
188: * Why explicit structure support?::
189: * Structure Usage::
190: * Structure Naming Convention::
191: * Structure Implementation::
192: * Structure Glossary::
193:
194: Object-oriented Forth
195:
196: * Objects::
197: * OOF::
198: * Mini-OOF::
199:
200: Objects
201:
202: * Properties of the Objects model::
203: * Why object-oriented programming?::
204: * Object-Oriented Terminology::
205: * Basic Objects Usage::
206: * The class Object::
207: * Creating objects::
208: * Object-Oriented Programming Style::
209: * Class Binding::
210: * Method conveniences::
211: * Classes and Scoping::
212: * Object Interfaces::
213: * Objects Implementation::
214: * Comparison with other object models::
215: * Objects Glossary::
216:
217: OOF
218:
219: * Properties of the OOF model::
220: * Basic OOF Usage::
221: * The base class object::
222: * Class Declaration::
223: * Class Implementation::
224:
225: Including Files
226:
227: * Words for Including::
228: * Search Path::
229: * Changing the Search Path::
230: * General Search Paths::
231:
232: Programming Tools
233:
234: * Debugging:: Simple and quick.
235: * Assertions:: Making your programs self-checking.
236: * Singlestep Debugger:: Executing your program word by word.
237:
238: Tools
239:
240: * ANS Report:: Report the words used, sorted by wordset.
241:
242: ANS conformance
243:
244: * The Core Words::
245: * The optional Block word set::
246: * The optional Double Number word set::
247: * The optional Exception word set::
248: * The optional Facility word set::
249: * The optional File-Access word set::
250: * The optional Floating-Point word set::
251: * The optional Locals word set::
252: * The optional Memory-Allocation word set::
253: * The optional Programming-Tools word set::
254: * The optional Search-Order word set::
255:
256: The Core Words
257:
258: * core-idef:: Implementation Defined Options
259: * core-ambcond:: Ambiguous Conditions
260: * core-other:: Other System Documentation
261:
262: The optional Block word set
263:
264: * block-idef:: Implementation Defined Options
265: * block-ambcond:: Ambiguous Conditions
266: * block-other:: Other System Documentation
267:
268: The optional Double Number word set
269:
270: * double-ambcond:: Ambiguous Conditions
271:
272: The optional Exception word set
273:
274: * exception-idef:: Implementation Defined Options
275:
276: The optional Facility word set
277:
278: * facility-idef:: Implementation Defined Options
279: * facility-ambcond:: Ambiguous Conditions
280:
281: The optional File-Access word set
282:
283: * file-idef:: Implementation Defined Options
284: * file-ambcond:: Ambiguous Conditions
285:
286: The optional Floating-Point word set
287:
288: * floating-idef:: Implementation Defined Options
289: * floating-ambcond:: Ambiguous Conditions
290:
291: The optional Locals word set
292:
293: * locals-idef:: Implementation Defined Options
294: * locals-ambcond:: Ambiguous Conditions
295:
296: The optional Memory-Allocation word set
297:
298: * memory-idef:: Implementation Defined Options
299:
300: The optional Programming-Tools word set
301:
302: * programming-idef:: Implementation Defined Options
303: * programming-ambcond:: Ambiguous Conditions
304:
305: The optional Search-Order word set
306:
307: * search-idef:: Implementation Defined Options
308: * search-ambcond:: Ambiguous Conditions
309:
310: Image Files
311:
312: * Image File Background:: Why have image files?
313: * Non-Relocatable Image Files:: don't always work.
314: * Data-Relocatable Image Files:: are better.
315: * Fully Relocatable Image Files:: better yet.
316: * Stack and Dictionary Sizes:: Setting the default sizes for an image.
317: * Running Image Files:: @code{gforth -i @var{file}} or @var{file}.
318: * Modifying the Startup Sequence:: and turnkey applications.
319:
320: Fully Relocatable Image Files
321:
322: * gforthmi:: The normal way
323: * cross.fs:: The hard way
324:
325: Engine
326:
327: * Portability::
328: * Threading::
329: * Primitives::
330: * Performance::
331:
332: Threading
333:
334: * Scheduling::
335: * Direct or Indirect Threaded?::
336: * DOES>::
337:
338: Primitives
339:
340: * Automatic Generation::
341: * TOS Optimization::
342: * Produced code::
343:
344: System Libraries
345:
346: * Binding to System Library::
347:
348: Cross Compiler
349:
350: * Using the Cross Compiler::
351: * How the Cross Compiler Works::
352:
353: @end menu
354:
355: @node License, Goals, Top, Top
356: @unnumbered GNU GENERAL PUBLIC LICENSE
357: @center Version 2, June 1991
358:
359: @display
360: Copyright @copyright{} 1989, 1991 Free Software Foundation, Inc.
361: 675 Mass Ave, Cambridge, MA 02139, USA
362:
363: Everyone is permitted to copy and distribute verbatim copies
364: of this license document, but changing it is not allowed.
365: @end display
366:
367: @unnumberedsec Preamble
368:
369: The licenses for most software are designed to take away your
370: freedom to share and change it. By contrast, the GNU General Public
371: License is intended to guarantee your freedom to share and change free
372: software---to make sure the software is free for all its users. This
373: General Public License applies to most of the Free Software
374: Foundation's software and to any other program whose authors commit to
375: using it. (Some other Free Software Foundation software is covered by
376: the GNU Library General Public License instead.) You can apply it to
377: your programs, too.
378:
379: When we speak of free software, we are referring to freedom, not
380: price. Our General Public Licenses are designed to make sure that you
381: have the freedom to distribute copies of free software (and charge for
382: this service if you wish), that you receive source code or can get it
383: if you want it, that you can change the software or use pieces of it
384: in new free programs; and that you know you can do these things.
385:
386: To protect your rights, we need to make restrictions that forbid
387: anyone to deny you these rights or to ask you to surrender the rights.
388: These restrictions translate to certain responsibilities for you if you
389: distribute copies of the software, or if you modify it.
390:
391: For example, if you distribute copies of such a program, whether
392: gratis or for a fee, you must give the recipients all the rights that
393: you have. You must make sure that they, too, receive or can get the
394: source code. And you must show them these terms so they know their
395: rights.
396:
397: We protect your rights with two steps: (1) copyright the software, and
398: (2) offer you this license which gives you legal permission to copy,
399: distribute and/or modify the software.
400:
401: Also, for each author's protection and ours, we want to make certain
402: that everyone understands that there is no warranty for this free
403: software. If the software is modified by someone else and passed on, we
404: want its recipients to know that what they have is not the original, so
405: that any problems introduced by others will not reflect on the original
406: authors' reputations.
407:
408: Finally, any free program is threatened constantly by software
409: patents. We wish to avoid the danger that redistributors of a free
410: program will individually obtain patent licenses, in effect making the
411: program proprietary. To prevent this, we have made it clear that any
412: patent must be licensed for everyone's free use or not licensed at all.
413:
414: The precise terms and conditions for copying, distribution and
415: modification follow.
416:
417: @iftex
418: @unnumberedsec TERMS AND CONDITIONS FOR COPYING, DISTRIBUTION AND MODIFICATION
419: @end iftex
420: @ifinfo
421: @center TERMS AND CONDITIONS FOR COPYING, DISTRIBUTION AND MODIFICATION
422: @end ifinfo
423:
424: @enumerate 0
425: @item
426: This License applies to any program or other work which contains
427: a notice placed by the copyright holder saying it may be distributed
428: under the terms of this General Public License. The ``Program'', below,
429: refers to any such program or work, and a ``work based on the Program''
430: means either the Program or any derivative work under copyright law:
431: that is to say, a work containing the Program or a portion of it,
432: either verbatim or with modifications and/or translated into another
433: language. (Hereinafter, translation is included without limitation in
434: the term ``modification''.) Each licensee is addressed as ``you''.
435:
436: Activities other than copying, distribution and modification are not
437: covered by this License; they are outside its scope. The act of
438: running the Program is not restricted, and the output from the Program
439: is covered only if its contents constitute a work based on the
440: Program (independent of having been made by running the Program).
441: Whether that is true depends on what the Program does.
442:
443: @item
444: You may copy and distribute verbatim copies of the Program's
445: source code as you receive it, in any medium, provided that you
446: conspicuously and appropriately publish on each copy an appropriate
447: copyright notice and disclaimer of warranty; keep intact all the
448: notices that refer to this License and to the absence of any warranty;
449: and give any other recipients of the Program a copy of this License
450: along with the Program.
451:
452: You may charge a fee for the physical act of transferring a copy, and
453: you may at your option offer warranty protection in exchange for a fee.
454:
455: @item
456: You may modify your copy or copies of the Program or any portion
457: of it, thus forming a work based on the Program, and copy and
458: distribute such modifications or work under the terms of Section 1
459: above, provided that you also meet all of these conditions:
460:
461: @enumerate a
462: @item
463: You must cause the modified files to carry prominent notices
464: stating that you changed the files and the date of any change.
465:
466: @item
467: You must cause any work that you distribute or publish, that in
468: whole or in part contains or is derived from the Program or any
469: part thereof, to be licensed as a whole at no charge to all third
470: parties under the terms of this License.
471:
472: @item
473: If the modified program normally reads commands interactively
474: when run, you must cause it, when started running for such
475: interactive use in the most ordinary way, to print or display an
476: announcement including an appropriate copyright notice and a
477: notice that there is no warranty (or else, saying that you provide
478: a warranty) and that users may redistribute the program under
479: these conditions, and telling the user how to view a copy of this
480: License. (Exception: if the Program itself is interactive but
481: does not normally print such an announcement, your work based on
482: the Program is not required to print an announcement.)
483: @end enumerate
484:
485: These requirements apply to the modified work as a whole. If
486: identifiable sections of that work are not derived from the Program,
487: and can be reasonably considered independent and separate works in
488: themselves, then this License, and its terms, do not apply to those
489: sections when you distribute them as separate works. But when you
490: distribute the same sections as part of a whole which is a work based
491: on the Program, the distribution of the whole must be on the terms of
492: this License, whose permissions for other licensees extend to the
493: entire whole, and thus to each and every part regardless of who wrote it.
494:
495: Thus, it is not the intent of this section to claim rights or contest
496: your rights to work written entirely by you; rather, the intent is to
497: exercise the right to control the distribution of derivative or
498: collective works based on the Program.
499:
500: In addition, mere aggregation of another work not based on the Program
501: with the Program (or with a work based on the Program) on a volume of
502: a storage or distribution medium does not bring the other work under
503: the scope of this License.
504:
505: @item
506: You may copy and distribute the Program (or a work based on it,
507: under Section 2) in object code or executable form under the terms of
508: Sections 1 and 2 above provided that you also do one of the following:
509:
510: @enumerate a
511: @item
512: Accompany it with the complete corresponding machine-readable
513: source code, which must be distributed under the terms of Sections
514: 1 and 2 above on a medium customarily used for software interchange; or,
515:
516: @item
517: Accompany it with a written offer, valid for at least three
518: years, to give any third party, for a charge no more than your
519: cost of physically performing source distribution, a complete
520: machine-readable copy of the corresponding source code, to be
521: distributed under the terms of Sections 1 and 2 above on a medium
522: customarily used for software interchange; or,
523:
524: @item
525: Accompany it with the information you received as to the offer
526: to distribute corresponding source code. (This alternative is
527: allowed only for noncommercial distribution and only if you
528: received the program in object code or executable form with such
529: an offer, in accord with Subsection b above.)
530: @end enumerate
531:
532: The source code for a work means the preferred form of the work for
533: making modifications to it. For an executable work, complete source
534: code means all the source code for all modules it contains, plus any
535: associated interface definition files, plus the scripts used to
536: control compilation and installation of the executable. However, as a
537: special exception, the source code distributed need not include
538: anything that is normally distributed (in either source or binary
539: form) with the major components (compiler, kernel, and so on) of the
540: operating system on which the executable runs, unless that component
541: itself accompanies the executable.
542:
543: If distribution of executable or object code is made by offering
544: access to copy from a designated place, then offering equivalent
545: access to copy the source code from the same place counts as
546: distribution of the source code, even though third parties are not
547: compelled to copy the source along with the object code.
548:
549: @item
550: You may not copy, modify, sublicense, or distribute the Program
551: except as expressly provided under this License. Any attempt
552: otherwise to copy, modify, sublicense or distribute the Program is
553: void, and will automatically terminate your rights under this License.
554: However, parties who have received copies, or rights, from you under
555: this License will not have their licenses terminated so long as such
556: parties remain in full compliance.
557:
558: @item
559: You are not required to accept this License, since you have not
560: signed it. However, nothing else grants you permission to modify or
561: distribute the Program or its derivative works. These actions are
562: prohibited by law if you do not accept this License. Therefore, by
563: modifying or distributing the Program (or any work based on the
564: Program), you indicate your acceptance of this License to do so, and
565: all its terms and conditions for copying, distributing or modifying
566: the Program or works based on it.
567:
568: @item
569: Each time you redistribute the Program (or any work based on the
570: Program), the recipient automatically receives a license from the
571: original licensor to copy, distribute or modify the Program subject to
572: these terms and conditions. You may not impose any further
573: restrictions on the recipients' exercise of the rights granted herein.
574: You are not responsible for enforcing compliance by third parties to
575: this License.
576:
577: @item
578: If, as a consequence of a court judgment or allegation of patent
579: infringement or for any other reason (not limited to patent issues),
580: conditions are imposed on you (whether by court order, agreement or
581: otherwise) that contradict the conditions of this License, they do not
582: excuse you from the conditions of this License. If you cannot
583: distribute so as to satisfy simultaneously your obligations under this
584: License and any other pertinent obligations, then as a consequence you
585: may not distribute the Program at all. For example, if a patent
586: license would not permit royalty-free redistribution of the Program by
587: all those who receive copies directly or indirectly through you, then
588: the only way you could satisfy both it and this License would be to
589: refrain entirely from distribution of the Program.
590:
591: If any portion of this section is held invalid or unenforceable under
592: any particular circumstance, the balance of the section is intended to
593: apply and the section as a whole is intended to apply in other
594: circumstances.
595:
596: It is not the purpose of this section to induce you to infringe any
597: patents or other property right claims or to contest validity of any
598: such claims; this section has the sole purpose of protecting the
599: integrity of the free software distribution system, which is
600: implemented by public license practices. Many people have made
601: generous contributions to the wide range of software distributed
602: through that system in reliance on consistent application of that
603: system; it is up to the author/donor to decide if he or she is willing
604: to distribute software through any other system and a licensee cannot
605: impose that choice.
606:
607: This section is intended to make thoroughly clear what is believed to
608: be a consequence of the rest of this License.
609:
610: @item
611: If the distribution and/or use of the Program is restricted in
612: certain countries either by patents or by copyrighted interfaces, the
613: original copyright holder who places the Program under this License
614: may add an explicit geographical distribution limitation excluding
615: those countries, so that distribution is permitted only in or among
616: countries not thus excluded. In such case, this License incorporates
617: the limitation as if written in the body of this License.
618:
619: @item
620: The Free Software Foundation may publish revised and/or new versions
621: of the General Public License from time to time. Such new versions will
622: be similar in spirit to the present version, but may differ in detail to
623: address new problems or concerns.
624:
625: Each version is given a distinguishing version number. If the Program
626: specifies a version number of this License which applies to it and ``any
627: later version'', you have the option of following the terms and conditions
628: either of that version or of any later version published by the Free
629: Software Foundation. If the Program does not specify a version number of
630: this License, you may choose any version ever published by the Free Software
631: Foundation.
632:
633: @item
634: If you wish to incorporate parts of the Program into other free
635: programs whose distribution conditions are different, write to the author
636: to ask for permission. For software which is copyrighted by the Free
637: Software Foundation, write to the Free Software Foundation; we sometimes
638: make exceptions for this. Our decision will be guided by the two goals
639: of preserving the free status of all derivatives of our free software and
640: of promoting the sharing and reuse of software generally.
641:
642: @iftex
643: @heading NO WARRANTY
644: @end iftex
645: @ifinfo
646: @center NO WARRANTY
647: @end ifinfo
648:
649: @item
650: BECAUSE THE PROGRAM IS LICENSED FREE OF CHARGE, THERE IS NO WARRANTY
651: FOR THE PROGRAM, TO THE EXTENT PERMITTED BY APPLICABLE LAW. EXCEPT WHEN
652: OTHERWISE STATED IN WRITING THE COPYRIGHT HOLDERS AND/OR OTHER PARTIES
653: PROVIDE THE PROGRAM ``AS IS'' WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED
654: OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
655: MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. THE ENTIRE RISK AS
656: TO THE QUALITY AND PERFORMANCE OF THE PROGRAM IS WITH YOU. SHOULD THE
657: PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF ALL NECESSARY SERVICING,
658: REPAIR OR CORRECTION.
659:
660: @item
661: IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING
662: WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MAY MODIFY AND/OR
663: REDISTRIBUTE THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES,
664: INCLUDING ANY GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING
665: OUT OF THE USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED
666: TO LOSS OF DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY
667: YOU OR THIRD PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER
668: PROGRAMS), EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE
669: POSSIBILITY OF SUCH DAMAGES.
670: @end enumerate
671:
672: @iftex
673: @heading END OF TERMS AND CONDITIONS
674: @end iftex
675: @ifinfo
676: @center END OF TERMS AND CONDITIONS
677: @end ifinfo
678:
679: @page
680: @unnumberedsec How to Apply These Terms to Your New Programs
681:
682: If you develop a new program, and you want it to be of the greatest
683: possible use to the public, the best way to achieve this is to make it
684: free software which everyone can redistribute and change under these terms.
685:
686: To do so, attach the following notices to the program. It is safest
687: to attach them to the start of each source file to most effectively
688: convey the exclusion of warranty; and each file should have at least
689: the ``copyright'' line and a pointer to where the full notice is found.
690:
691: @smallexample
692: @var{one line to give the program's name and a brief idea of what it does.}
693: Copyright (C) 19@var{yy} @var{name of author}
694:
695: This program is free software; you can redistribute it and/or modify
696: it under the terms of the GNU General Public License as published by
697: the Free Software Foundation; either version 2 of the License, or
698: (at your option) any later version.
699:
700: This program is distributed in the hope that it will be useful,
701: but WITHOUT ANY WARRANTY; without even the implied warranty of
702: MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
703: GNU General Public License for more details.
704:
705: You should have received a copy of the GNU General Public License
706: along with this program; if not, write to the Free Software
707: Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
708: @end smallexample
709:
710: Also add information on how to contact you by electronic and paper mail.
711:
712: If the program is interactive, make it output a short notice like this
713: when it starts in an interactive mode:
714:
715: @smallexample
716: Gnomovision version 69, Copyright (C) 19@var{yy} @var{name of author}
717: Gnomovision comes with ABSOLUTELY NO WARRANTY; for details
718: type `show w'.
719: This is free software, and you are welcome to redistribute it
720: under certain conditions; type `show c' for details.
721: @end smallexample
722:
723: The hypothetical commands @samp{show w} and @samp{show c} should show
724: the appropriate parts of the General Public License. Of course, the
725: commands you use may be called something other than @samp{show w} and
726: @samp{show c}; they could even be mouse-clicks or menu items---whatever
727: suits your program.
728:
729: You should also get your employer (if you work as a programmer) or your
730: school, if any, to sign a ``copyright disclaimer'' for the program, if
731: necessary. Here is a sample; alter the names:
732:
733: @smallexample
734: Yoyodyne, Inc., hereby disclaims all copyright interest in the program
735: `Gnomovision' (which makes passes at compilers) written by James Hacker.
736:
737: @var{signature of Ty Coon}, 1 April 1989
738: Ty Coon, President of Vice
739: @end smallexample
740:
741: This General Public License does not permit incorporating your program into
742: proprietary programs. If your program is a subroutine library, you may
743: consider it more useful to permit linking proprietary applications with the
744: library. If this is what you want to do, use the GNU Library General
745: Public License instead of this License.
746:
747: @iftex
748: @unnumbered Preface
749: @cindex Preface
750: This manual documents Gforth. The reader is expected to know
751: Forth. This manual is primarily a reference manual. @xref{Other Books}
752: for introductory material.
753: @end iftex
754:
755: @node Goals, Other Books, License, Top
756: @comment node-name, next, previous, up
757: @chapter Goals of Gforth
758: @cindex Goals
759: The goal of the Gforth Project is to develop a standard model for
760: ANS Forth. This can be split into several subgoals:
761:
762: @itemize @bullet
763: @item
764: Gforth should conform to the Forth standard (ANS Forth).
765: @item
766: It should be a model, i.e. it should define all the
767: implementation-dependent things.
768: @item
769: It should become standard, i.e. widely accepted and used. This goal
770: is the most difficult one.
771: @end itemize
772:
773: To achieve these goals Gforth should be
774: @itemize @bullet
775: @item
776: Similar to previous models (fig-Forth, F83)
777: @item
778: Powerful. It should provide for all the things that are considered
779: necessary today and even some that are not yet considered necessary.
780: @item
781: Efficient. It should not get the reputation of being exceptionally
782: slow.
783: @item
784: Free.
785: @item
786: Available on many machines/easy to port.
787: @end itemize
788:
789: Have we achieved these goals? Gforth conforms to the ANS Forth
790: standard. It may be considered a model, but we have not yet documented
791: which parts of the model are stable and which parts we are likely to
792: change. It certainly has not yet become a de facto standard, but it
793: appears to be quite popular. It has some similarities to and some
794: differences from previous models. It has some powerful features, but not
795: yet everything that we envisioned. We certainly have achieved our
796: execution speed goals (@pxref{Performance}). It is free and available
797: on many machines.
798:
799: @node Other Books, Invoking Gforth, Goals, Top
800: @chapter Other books on ANS Forth
801: @cindex books on Forth
802:
803: As the standard is relatively new, there are not many books out yet. It
804: is not recommended to learn Forth by using Gforth and a book that is not
805: written for ANS Forth, as you will not know your mistakes from the
806: deviations of the book. However, books based on the Forth-83 standard
807: should be ok, because ANS Forth is primarily an extension of Forth-83.
808:
809: @cindex standard document for ANS Forth
810: @cindex ANS Forth document
811: There is, of course, the standard, the definite reference if you want to
812: write ANS Forth programs. It is available in printed form from the
813: National Standards Institute Sales Department (Tel.: USA (212) 642-4900;
814: Fax.: USA (212) 302-1286) as document @cite{X3.215-1994} for about $200. You
815: can also get it from Global Engineering Documents (Tel.: USA (800)
816: 854-7179; Fax.: (303) 843-9880) for about $300.
817:
818: @cite{dpANS6}, the last draft of the standard, which was then submitted
819: to ANSI for publication is available electronically and for free in some
820: MS Word format, and it has been converted to HTML
821: (@url{http://www.taygeta.com/forth/dpans.html}; this is my favourite
822: format); this HTML version also includes the answers to Requests for
823: Interpretation (RFIs). Some pointers to these versions can be found
824: through @*@url{http://www.complang.tuwien.ac.at/projects/forth.html}.
825:
826: @cindex introductory book
827: @cindex book, introductory
828: @cindex Woehr, Jack: @cite{Forth: The New Model}
829: @cindex @cite{Forth: The new model} (book)
830: @cite{Forth: The New Model} by Jack Woehr (Prentice-Hall, 1993) is an
831: introductory book based on a draft version of the standard. It does not
832: cover the whole standard. It also contains interesting background
833: information (Jack Woehr was in the ANS Forth Technical Committee). It is
834: not appropriate for complete newbies, but programmers experienced in
835: other languages should find it ok.
836:
837: @cindex Conklin, Edward K., and Elizabeth Rather: @cite{Forth Programmer's Handbook}
838: @cindex Rather, Elizabeth and Edward K. Conklin: @cite{Forth Programmer's Handbook}
839: @cindex @cite{Forth Programmer's Handbook} (book)
840: @cite{Forth Programmer's Handbook} by Edward K. Conklin, Elizabeth
841: D. Rather and the technical staff of Forth, Inc. (Forth, Inc., 1997;
842: ISBN 0-9662156-0-5) contains little introductory material. The majority
843: of the book is similar to @ref{Words}, but the book covers most of the
844: standard words and some non-standard words (whereas this manual is
845: quite incomplete). In addition, the book contains a chapter on
846: programming style. The major drawback of this book is that it usually
847: does not identify what is standard and what is specific to the Forth
848: system described in the book (probably one of Forth, Inc.'s systems).
849: Fortunately, many of the non-standard programming practices described in
850: the book work in Gforth, too. Still, this drawback makes the book
851: hardly more useful than a pre-ANS book.
852:
853: @node Invoking Gforth, Words, Other Books, Top
854: @chapter Invoking Gforth
855: @cindex invoking Gforth
856: @cindex running Gforth
857: @cindex command-line options
858: @cindex options on the command line
859: @cindex flags on the command line
860:
861: You will usually just say @code{gforth}. In many other cases the default
862: Gforth image will be invoked like this:
863: @example
864: gforth [files] [-e forth-code]
865: @end example
866: This interprets the contents of the files and the Forth code in the order they
867: are given.
868:
869: In general, the command line looks like this:
870:
871: @example
872: gforth [initialization options] [image-specific options]
873: @end example
874:
875: The initialization options must come before the rest of the command
876: line. They are:
877:
878: @table @code
879: @cindex -i, command-line option
880: @cindex --image-file, command-line option
881: @item --image-file @var{file}
882: @itemx -i @var{file}
883: Loads the Forth image @var{file} instead of the default
884: @file{gforth.fi} (@pxref{Image Files}).
885:
886: @cindex --path, command-line option
887: @cindex -p, command-line option
888: @item --path @var{path}
889: @itemx -p @var{path}
890: Uses @var{path} for searching the image file and Forth source code files
891: instead of the default in the environment variable @code{GFORTHPATH} or
892: the path specified at installation time (e.g.,
893: @file{/usr/local/share/gforth/0.2.0:.}). A path is given as a list of
894: directories, separated by @samp{:} (on Unix) or @samp{;} (on other OSs).
895:
896: @cindex --dictionary-size, command-line option
897: @cindex -m, command-line option
898: @cindex @var{size} parameters for command-line options
899: @cindex size of the dictionary and the stacks
900: @item --dictionary-size @var{size}
901: @itemx -m @var{size}
902: Allocate @var{size} space for the Forth dictionary space instead of
903: using the default specified in the image (typically 256K). The
904: @var{size} specification consists of an integer and a unit (e.g.,
905: @code{4M}). The unit can be one of @code{b} (bytes), @code{e} (element
906: size, in this case Cells), @code{k} (kilobytes), @code{M} (Megabytes),
907: @code{G} (Gigabytes), and @code{T} (Terabytes). If no unit is specified,
908: @code{e} is used.
909:
910: @cindex --data-stack-size, command-line option
911: @cindex -d, command-line option
912: @item --data-stack-size @var{size}
913: @itemx -d @var{size}
914: Allocate @var{size} space for the data stack instead of using the
915: default specified in the image (typically 16K).
916:
917: @cindex --return-stack-size, command-line option
918: @cindex -r, command-line option
919: @item --return-stack-size @var{size}
920: @itemx -r @var{size}
921: Allocate @var{size} space for the return stack instead of using the
922: default specified in the image (typically 15K).
923:
924: @cindex --fp-stack-size, command-line option
925: @cindex -f, command-line option
926: @item --fp-stack-size @var{size}
927: @itemx -f @var{size}
928: Allocate @var{size} space for the floating point stack instead of
929: using the default specified in the image (typically 15.5K). In this case
930: the unit specifier @code{e} refers to floating point numbers.
931:
932: @cindex --locals-stack-size, command-line option
933: @cindex -l, command-line option
934: @item --locals-stack-size @var{size}
935: @itemx -l @var{size}
936: Allocate @var{size} space for the locals stack instead of using the
937: default specified in the image (typically 14.5K).
938:
939: @cindex -h, command-line option
940: @cindex --help, command-line option
941: @item --help
942: @itemx -h
943: Print a message about the command-line options
944:
945: @cindex -v, command-line option
946: @cindex --version, command-line option
947: @item --version
948: @itemx -v
949: Print version and exit
950:
951: @cindex --debug, command-line option
952: @item --debug
953: Print some information useful for debugging on startup.
954:
955: @cindex --offset-image, command-line option
956: @item --offset-image
957: Start the dictionary at a slightly different position than would be used
958: otherwise (useful for creating data-relocatable images,
959: @pxref{Data-Relocatable Image Files}).
960:
961: @cindex --no-offset-im, command-line option
962: @item --no-offset-im
963: Start the dictionary at the normal position.
964:
965: @cindex --clear-dictionary, command-line option
966: @item --clear-dictionary
967: Initialize all bytes in the dictionary to 0 before loading the image
968: (@pxref{Data-Relocatable Image Files}).
969:
970: @cindex --die-on-signal, command-line-option
971: @item --die-on-signal
972: Normally Gforth handles most signals (e.g., the user interrupt SIGINT,
973: or the segmentation violation SIGSEGV) by translating it into a Forth
974: @code{THROW}. With this option, Gforth exits if it receives such a
975: signal. This option is useful when the engine and/or the image might be
976: severely broken (such that it causes another signal before recovering
977: from the first); this option avoids endless loops in such cases.
978: @end table
979:
980: @cindex loading files at startup
981: @cindex executing code on startup
982: @cindex batch processing with Gforth
983: As explained above, the image-specific command-line arguments for the
984: default image @file{gforth.fi} consist of a sequence of filenames and
985: @code{-e @var{forth-code}} options that are interpreted in the sequence
986: in which they are given. The @code{-e @var{forth-code}} or
987: @code{--evaluate @var{forth-code}} option evaluates the forth
988: code. This option takes only one argument; if you want to evaluate more
989: Forth words, you have to quote them or use several @code{-e}s. To exit
990: after processing the command line (instead of entering interactive mode)
991: append @code{-e bye} to the command line.
992:
993: @cindex versions, invoking other versions of Gforth
994: If you have several versions of Gforth installed, @code{gforth} will
995: invoke the version that was installed last. @code{gforth-@var{version}}
996: invokes a specific version. You may want to use the option
997: @code{--path}, if your environment contains the variable
998: @code{GFORTHPATH}.
999:
1000: Not yet implemented:
1001: On startup the system first executes the system initialization file
1002: (unless the option @code{--no-init-file} is given; note that the system
1003: resulting from using this option may not be ANS Forth conformant). Then
1004: the user initialization file @file{.gforth.fs} is executed, unless the
1005: option @code{--no-rc} is given; this file is first searched in @file{.},
1006: then in @file{~}, then in the normal path (see above).
1007:
1008: @node Words, Tools, Invoking Gforth, Top
1009: @chapter Forth Words
1010: @cindex Words
1011:
1012: @menu
1013: * Notation::
1014: * Arithmetic::
1015: * Stack Manipulation::
1016: * Memory::
1017: * Control Structures::
1018: * Locals::
1019: * Defining Words::
1020: * Structures::
1021: * Object-oriented Forth::
1022: * Tokens for Words::
1023: * Wordlists::
1024: * Files::
1025: * Including Files::
1026: * Blocks::
1027: * Other I/O::
1028: * Programming Tools::
1029: * Assembler and Code Words::
1030: * Threading Words::
1031: @end menu
1032:
1033: @node Notation, Arithmetic, Words, Words
1034: @section Notation
1035: @cindex notation of glossary entries
1036: @cindex format of glossary entries
1037: @cindex glossary notation format
1038: @cindex word glossary entry format
1039:
1040: The Forth words are described in this section in the glossary notation
1041: that has become a de-facto standard for Forth texts, i.e.,
1042:
1043: @format
1044: @var{word} @var{Stack effect} @var{wordset} @var{pronunciation}
1045: @end format
1046: @var{Description}
1047:
1048: @table @var
1049: @item word
1050: @cindex case insensitivity
1051: The name of the word. BTW, Gforth is case insensitive, so you can
1052: type the words in in lower case (However, @pxref{core-idef}).
1053:
1054: @item Stack effect
1055: @cindex stack effect
1056: The stack effect is written in the notation @code{@var{before} --
1057: @var{after}}, where @var{before} and @var{after} describe the top of
1058: stack entries before and after the execution of the word. The rest of
1059: the stack is not touched by the word. The top of stack is rightmost,
1060: i.e., a stack sequence is written as it is typed in. Note that Gforth
1061: uses a separate floating point stack, but a unified stack
1062: notation. Also, return stack effects are not shown in @var{stack
1063: effect}, but in @var{Description}. The name of a stack item describes
1064: the type and/or the function of the item. See below for a discussion of
1065: the types.
1066:
1067: All words have two stack effects: A compile-time stack effect and a
1068: run-time stack effect. The compile-time stack-effect of most words is
1069: @var{ -- }. If the compile-time stack-effect of a word deviates from
1070: this standard behaviour, or the word does other unusual things at
1071: compile time, both stack effects are shown; otherwise only the run-time
1072: stack effect is shown.
1073:
1074: @cindex pronounciation of words
1075: @item pronunciation
1076: How the word is pronounced.
1077:
1078: @cindex wordset
1079: @item wordset
1080: The ANS Forth standard is divided into several wordsets. A standard
1081: system need not support all of them. So, the fewer wordsets your program
1082: uses the more portable it will be in theory. However, we suspect that
1083: most ANS Forth systems on personal machines will feature all
1084: wordsets. Words that are not defined in the ANS standard have
1085: @code{gforth} or @code{gforth-internal} as wordset. @code{gforth}
1086: describes words that will work in future releases of Gforth;
1087: @code{gforth-internal} words are more volatile. Environmental query
1088: strings are also displayed like words; you can recognize them by the
1089: @code{environment} in the wordset field.
1090:
1091: @item Description
1092: A description of the behaviour of the word.
1093: @end table
1094:
1095: @cindex types of stack items
1096: @cindex stack item types
1097: The type of a stack item is specified by the character(s) the name
1098: starts with:
1099:
1100: @table @code
1101: @item f
1102: @cindex @code{f}, stack item type
1103: Boolean flags, i.e. @code{false} or @code{true}.
1104: @item c
1105: @cindex @code{c}, stack item type
1106: Char
1107: @item w
1108: @cindex @code{w}, stack item type
1109: Cell, can contain an integer or an address
1110: @item n
1111: @cindex @code{n}, stack item type
1112: signed integer
1113: @item u
1114: @cindex @code{u}, stack item type
1115: unsigned integer
1116: @item d
1117: @cindex @code{d}, stack item type
1118: double sized signed integer
1119: @item ud
1120: @cindex @code{ud}, stack item type
1121: double sized unsigned integer
1122: @item r
1123: @cindex @code{r}, stack item type
1124: Float (on the FP stack)
1125: @item a_
1126: @cindex @code{a_}, stack item type
1127: Cell-aligned address
1128: @item c_
1129: @cindex @code{c_}, stack item type
1130: Char-aligned address (note that a Char may have two bytes in Windows NT)
1131: @item f_
1132: @cindex @code{f_}, stack item type
1133: Float-aligned address
1134: @item df_
1135: @cindex @code{df_}, stack item type
1136: Address aligned for IEEE double precision float
1137: @item sf_
1138: @cindex @code{sf_}, stack item type
1139: Address aligned for IEEE single precision float
1140: @item xt
1141: @cindex @code{xt}, stack item type
1142: Execution token, same size as Cell
1143: @item wid
1144: @cindex @code{wid}, stack item type
1145: Wordlist ID, same size as Cell
1146: @item f83name
1147: @cindex @code{f83name}, stack item type
1148: Pointer to a name structure
1149: @item "
1150: @cindex @code{"}, stack item type
1151: string in the input stream (not on the stack). The terminating character
1152: is a blank by default. If it is not a blank, it is shown in @code{<>}
1153: quotes.
1154: @end table
1155:
1156: @node Arithmetic, Stack Manipulation, Notation, Words
1157: @section Arithmetic
1158: @cindex arithmetic words
1159:
1160: @cindex division with potentially negative operands
1161: Forth arithmetic is not checked, i.e., you will not hear about integer
1162: overflow on addition or multiplication, you may hear about division by
1163: zero if you are lucky. The operator is written after the operands, but
1164: the operands are still in the original order. I.e., the infix @code{2-1}
1165: corresponds to @code{2 1 -}. Forth offers a variety of division
1166: operators. If you perform division with potentially negative operands,
1167: you do not want to use @code{/} or @code{/mod} with its undefined
1168: behaviour, but rather @code{fm/mod} or @code{sm/mod} (probably the
1169: former, @pxref{Mixed precision}).
1170:
1171: @menu
1172: * Single precision::
1173: * Bitwise operations::
1174: * Mixed precision:: operations with single and double-cell integers
1175: * Double precision:: Double-cell integer arithmetic
1176: * Floating Point::
1177: @end menu
1178:
1179: @node Single precision, Bitwise operations, Arithmetic, Arithmetic
1180: @subsection Single precision
1181: @cindex single precision arithmetic words
1182:
1183: doc-+
1184: doc--
1185: doc-*
1186: doc-/
1187: doc-mod
1188: doc-/mod
1189: doc-negate
1190: doc-abs
1191: doc-min
1192: doc-max
1193:
1194: @node Bitwise operations, Mixed precision, Single precision, Arithmetic
1195: @subsection Bitwise operations
1196: @cindex bitwise operation words
1197:
1198: doc-and
1199: doc-or
1200: doc-xor
1201: doc-invert
1202: doc-2*
1203: doc-2/
1204:
1205: @node Mixed precision, Double precision, Bitwise operations, Arithmetic
1206: @subsection Mixed precision
1207: @cindex mixed precision arithmetic words
1208:
1209: doc-m+
1210: doc-*/
1211: doc-*/mod
1212: doc-m*
1213: doc-um*
1214: doc-m*/
1215: doc-um/mod
1216: doc-fm/mod
1217: doc-sm/rem
1218:
1219: @node Double precision, Floating Point, Mixed precision, Arithmetic
1220: @subsection Double precision
1221: @cindex double precision arithmetic words
1222:
1223: @cindex double-cell numbers, input format
1224: @cindex input format for double-cell numbers
1225: The outer (aka text) interpreter converts numbers containing a dot into
1226: a double precision number. Note that only numbers with the dot as last
1227: character are standard-conforming.
1228:
1229: doc-d+
1230: doc-d-
1231: doc-dnegate
1232: doc-dabs
1233: doc-dmin
1234: doc-dmax
1235:
1236: @node Floating Point, , Double precision, Arithmetic
1237: @subsection Floating Point
1238: @cindex floating point arithmetic words
1239:
1240: @cindex floating-point numbers, input format
1241: @cindex input format for floating-point numbers
1242: The format of floating point numbers recognized by the outer (aka text)
1243: interpreter is: a signed decimal number, possibly containing a decimal
1244: point (@code{.}), followed by @code{E} or @code{e}, optionally followed
1245: by a signed integer (the exponent). E.g., @code{1e} is the same as
1246: @code{+1.0e+0}. Note that a number without @code{e} is not interpreted
1247: as floating-point number, but as double (if the number contains a
1248: @code{.}) or single precision integer. Also, conversions between string
1249: and floating point numbers always use base 10, irrespective of the value
1250: of @code{BASE} (in Gforth; for the standard this is an ambiguous
1251: condition). If @code{BASE} contains a value greater then 14, the
1252: @code{E} may be interpreted as digit and the number will be interpreted
1253: as integer, unless it has a signed exponent (both @code{+} and @code{-}
1254: are allowed as signs).
1255:
1256: @cindex angles in trigonometric operations
1257: @cindex trigonometric operations
1258: Angles in floating point operations are given in radians (a full circle
1259: has 2 pi radians). Note, that Gforth has a separate floating point
1260: stack, but we use the unified notation.
1261:
1262: @cindex floating-point arithmetic, pitfalls
1263: Floating point numbers have a number of unpleasant surprises for the
1264: unwary (e.g., floating point addition is not associative) and even a few
1265: for the wary. You should not use them unless you know what you are doing
1266: or you don't care that the results you get are totally bogus. If you
1267: want to learn about the problems of floating point numbers (and how to
1268: avoid them), you might start with @cite{David Goldberg, What Every
1269: Computer Scientist Should Know About Floating-Point Arithmetic, ACM
1270: Computing Surveys 23(1):5@minus{}48, March 1991}
1271: (@url{http://www.validgh.com/goldberg/paper.ps}).
1272:
1273: doc-f+
1274: doc-f-
1275: doc-f*
1276: doc-f/
1277: doc-fnegate
1278: doc-fabs
1279: doc-fmax
1280: doc-fmin
1281: doc-floor
1282: doc-fround
1283: doc-f**
1284: doc-fsqrt
1285: doc-fexp
1286: doc-fexpm1
1287: doc-fln
1288: doc-flnp1
1289: doc-flog
1290: doc-falog
1291: doc-fsin
1292: doc-fcos
1293: doc-fsincos
1294: doc-ftan
1295: doc-fasin
1296: doc-facos
1297: doc-fatan
1298: doc-fatan2
1299: doc-fsinh
1300: doc-fcosh
1301: doc-ftanh
1302: doc-fasinh
1303: doc-facosh
1304: doc-fatanh
1305:
1306: @node Stack Manipulation, Memory, Arithmetic, Words
1307: @section Stack Manipulation
1308: @cindex stack manipulation words
1309:
1310: @cindex floating-point stack in the standard
1311: Gforth has a data stack (aka parameter stack) for characters, cells,
1312: addresses, and double cells, a floating point stack for floating point
1313: numbers, a return stack for storing the return addresses of colon
1314: definitions and other data, and a locals stack for storing local
1315: variables. Note that while every sane Forth has a separate floating
1316: point stack, this is not strictly required; an ANS Forth system could
1317: theoretically keep floating point numbers on the data stack. As an
1318: additional difficulty, you don't know how many cells a floating point
1319: number takes. It is reportedly possible to write words in a way that
1320: they work also for a unified stack model, but we do not recommend trying
1321: it. Instead, just say that your program has an environmental dependency
1322: on a separate FP stack.
1323:
1324: @cindex return stack and locals
1325: @cindex locals and return stack
1326: Also, a Forth system is allowed to keep the local variables on the
1327: return stack. This is reasonable, as local variables usually eliminate
1328: the need to use the return stack explicitly. So, if you want to produce
1329: a standard complying program and if you are using local variables in a
1330: word, forget about return stack manipulations in that word (see the
1331: standard document for the exact rules).
1332:
1333: @menu
1334: * Data stack::
1335: * Floating point stack::
1336: * Return stack::
1337: * Locals stack::
1338: * Stack pointer manipulation::
1339: @end menu
1340:
1341: @node Data stack, Floating point stack, Stack Manipulation, Stack Manipulation
1342: @subsection Data stack
1343: @cindex data stack manipulation words
1344: @cindex stack manipulations words, data stack
1345:
1346: doc-drop
1347: doc-nip
1348: doc-dup
1349: doc-over
1350: doc-tuck
1351: doc-swap
1352: doc-rot
1353: doc--rot
1354: doc-?dup
1355: doc-pick
1356: doc-roll
1357: doc-2drop
1358: doc-2nip
1359: doc-2dup
1360: doc-2over
1361: doc-2tuck
1362: doc-2swap
1363: doc-2rot
1364:
1365: @node Floating point stack, Return stack, Data stack, Stack Manipulation
1366: @subsection Floating point stack
1367: @cindex floating-point stack manipulation words
1368: @cindex stack manipulation words, floating-point stack
1369:
1370: doc-fdrop
1371: doc-fnip
1372: doc-fdup
1373: doc-fover
1374: doc-ftuck
1375: doc-fswap
1376: doc-frot
1377:
1378: @node Return stack, Locals stack, Floating point stack, Stack Manipulation
1379: @subsection Return stack
1380: @cindex return stack manipulation words
1381: @cindex stack manipulation words, return stack
1382:
1383: doc->r
1384: doc-r>
1385: doc-r@
1386: doc-rdrop
1387: doc-2>r
1388: doc-2r>
1389: doc-2r@
1390: doc-2rdrop
1391:
1392: @node Locals stack, Stack pointer manipulation, Return stack, Stack Manipulation
1393: @subsection Locals stack
1394:
1395: @node Stack pointer manipulation, , Locals stack, Stack Manipulation
1396: @subsection Stack pointer manipulation
1397: @cindex stack pointer manipulation words
1398:
1399: doc-sp@
1400: doc-sp!
1401: doc-fp@
1402: doc-fp!
1403: doc-rp@
1404: doc-rp!
1405: doc-lp@
1406: doc-lp!
1407:
1408: @node Memory, Control Structures, Stack Manipulation, Words
1409: @section Memory
1410: @cindex Memory words
1411:
1412: @menu
1413: * Memory Access::
1414: * Address arithmetic::
1415: * Memory Blocks::
1416: @end menu
1417:
1418: @node Memory Access, Address arithmetic, Memory, Memory
1419: @subsection Memory Access
1420: @cindex memory access words
1421:
1422: doc-@
1423: doc-!
1424: doc-+!
1425: doc-c@
1426: doc-c!
1427: doc-2@
1428: doc-2!
1429: doc-f@
1430: doc-f!
1431: doc-sf@
1432: doc-sf!
1433: doc-df@
1434: doc-df!
1435:
1436: @node Address arithmetic, Memory Blocks, Memory Access, Memory
1437: @subsection Address arithmetic
1438: @cindex address arithmetic words
1439:
1440: ANS Forth does not specify the sizes of the data types. Instead, it
1441: offers a number of words for computing sizes and doing address
1442: arithmetic. Basically, address arithmetic is performed in terms of
1443: address units (aus); on most systems the address unit is one byte. Note
1444: that a character may have more than one au, so @code{chars} is no noop
1445: (on systems where it is a noop, it compiles to nothing).
1446:
1447: @cindex alignment of addresses for types
1448: ANS Forth also defines words for aligning addresses for specific
1449: types. Many computers require that accesses to specific data types
1450: must only occur at specific addresses; e.g., that cells may only be
1451: accessed at addresses divisible by 4. Even if a machine allows unaligned
1452: accesses, it can usually perform aligned accesses faster.
1453:
1454: For the performance-conscious: alignment operations are usually only
1455: necessary during the definition of a data structure, not during the
1456: (more frequent) accesses to it.
1457:
1458: ANS Forth defines no words for character-aligning addresses. This is not
1459: an oversight, but reflects the fact that addresses that are not
1460: char-aligned have no use in the standard and therefore will not be
1461: created.
1462:
1463: @cindex @code{CREATE} and alignment
1464: The standard guarantees that addresses returned by @code{CREATE}d words
1465: are cell-aligned; in addition, Gforth guarantees that these addresses
1466: are aligned for all purposes.
1467:
1468: Note that the standard defines a word @code{char}, which has nothing to
1469: do with address arithmetic.
1470:
1471: doc-chars
1472: doc-char+
1473: doc-cells
1474: doc-cell+
1475: doc-cell
1476: doc-align
1477: doc-aligned
1478: doc-floats
1479: doc-float+
1480: doc-float
1481: doc-falign
1482: doc-faligned
1483: doc-sfloats
1484: doc-sfloat+
1485: doc-sfalign
1486: doc-sfaligned
1487: doc-dfloats
1488: doc-dfloat+
1489: doc-dfalign
1490: doc-dfaligned
1491: doc-maxalign
1492: doc-maxaligned
1493: doc-cfalign
1494: doc-cfaligned
1495: doc-address-unit-bits
1496:
1497: @node Memory Blocks, , Address arithmetic, Memory
1498: @subsection Memory Blocks
1499: @cindex memory block words
1500:
1501: doc-move
1502: doc-erase
1503:
1504: While the previous words work on address units, the rest works on
1505: characters.
1506:
1507: doc-cmove
1508: doc-cmove>
1509: doc-fill
1510: doc-blank
1511:
1512: @node Control Structures, Locals, Memory, Words
1513: @section Control Structures
1514: @cindex control structures
1515:
1516: Control structures in Forth cannot be used in interpret state, only in
1517: compile state@footnote{More precisely, they have no interpretation
1518: semantics (@pxref{Interpretation and Compilation Semantics})}, i.e., in
1519: a colon definition. We do not like this limitation, but have not seen a
1520: satisfying way around it yet, although many schemes have been proposed.
1521:
1522: @menu
1523: * Selection::
1524: * Simple Loops::
1525: * Counted Loops::
1526: * Arbitrary control structures::
1527: * Calls and returns::
1528: * Exception Handling::
1529: @end menu
1530:
1531: @node Selection, Simple Loops, Control Structures, Control Structures
1532: @subsection Selection
1533: @cindex selection control structures
1534: @cindex control structures for selection
1535:
1536: @cindex @code{IF} control structure
1537: @example
1538: @var{flag}
1539: IF
1540: @var{code}
1541: ENDIF
1542: @end example
1543: or
1544: @example
1545: @var{flag}
1546: IF
1547: @var{code1}
1548: ELSE
1549: @var{code2}
1550: ENDIF
1551: @end example
1552:
1553: You can use @code{THEN} instead of @code{ENDIF}. Indeed, @code{THEN} is
1554: standard, and @code{ENDIF} is not, although it is quite popular. We
1555: recommend using @code{ENDIF}, because it is less confusing for people
1556: who also know other languages (and is not prone to reinforcing negative
1557: prejudices against Forth in these people). Adding @code{ENDIF} to a
1558: system that only supplies @code{THEN} is simple:
1559: @example
1560: : endif POSTPONE then ; immediate
1561: @end example
1562:
1563: [According to @cite{Webster's New Encyclopedic Dictionary}, @dfn{then
1564: (adv.)} has the following meanings:
1565: @quotation
1566: ... 2b: following next after in order ... 3d: as a necessary consequence
1567: (if you were there, then you saw them).
1568: @end quotation
1569: Forth's @code{THEN} has the meaning 2b, whereas @code{THEN} in Pascal
1570: and many other programming languages has the meaning 3d.]
1571:
1572: Gforth also provides the words @code{?dup-if} and @code{?dup-0=-if}, so
1573: you can avoid using @code{?dup}. Using these alternatives is also more
1574: efficient than using @code{?dup}. Definitions in plain standard Forth
1575: for @code{ENDIF}, @code{?DUP-IF} and @code{?DUP-0=-IF} are provided in
1576: @file{compat/control.fs}.
1577:
1578: @cindex @code{CASE} control structure
1579: @example
1580: @var{n}
1581: CASE
1582: @var{n1} OF @var{code1} ENDOF
1583: @var{n2} OF @var{code2} ENDOF
1584: @dots{}
1585: ENDCASE
1586: @end example
1587:
1588: Executes the first @var{codei}, where the @var{ni} is equal to
1589: @var{n}. A default case can be added by simply writing the code after
1590: the last @code{ENDOF}. It may use @var{n}, which is on top of the stack,
1591: but must not consume it.
1592:
1593: @node Simple Loops, Counted Loops, Selection, Control Structures
1594: @subsection Simple Loops
1595: @cindex simple loops
1596: @cindex loops without count
1597:
1598: @cindex @code{WHILE} loop
1599: @example
1600: BEGIN
1601: @var{code1}
1602: @var{flag}
1603: WHILE
1604: @var{code2}
1605: REPEAT
1606: @end example
1607:
1608: @var{code1} is executed and @var{flag} is computed. If it is true,
1609: @var{code2} is executed and the loop is restarted; If @var{flag} is
1610: false, execution continues after the @code{REPEAT}.
1611:
1612: @cindex @code{UNTIL} loop
1613: @example
1614: BEGIN
1615: @var{code}
1616: @var{flag}
1617: UNTIL
1618: @end example
1619:
1620: @var{code} is executed. The loop is restarted if @code{flag} is false.
1621:
1622: @cindex endless loop
1623: @cindex loops, endless
1624: @example
1625: BEGIN
1626: @var{code}
1627: AGAIN
1628: @end example
1629:
1630: This is an endless loop.
1631:
1632: @node Counted Loops, Arbitrary control structures, Simple Loops, Control Structures
1633: @subsection Counted Loops
1634: @cindex counted loops
1635: @cindex loops, counted
1636: @cindex @code{DO} loops
1637:
1638: The basic counted loop is:
1639: @example
1640: @var{limit} @var{start}
1641: ?DO
1642: @var{body}
1643: LOOP
1644: @end example
1645:
1646: This performs one iteration for every integer, starting from @var{start}
1647: and up to, but excluding @var{limit}. The counter, aka index, can be
1648: accessed with @code{i}. E.g., the loop
1649: @example
1650: 10 0 ?DO
1651: i .
1652: LOOP
1653: @end example
1654: prints
1655: @example
1656: 0 1 2 3 4 5 6 7 8 9
1657: @end example
1658: The index of the innermost loop can be accessed with @code{i}, the index
1659: of the next loop with @code{j}, and the index of the third loop with
1660: @code{k}.
1661:
1662: doc-i
1663: doc-j
1664: doc-k
1665:
1666: The loop control data are kept on the return stack, so there are some
1667: restrictions on mixing return stack accesses and counted loop
1668: words. E.g., if you put values on the return stack outside the loop, you
1669: cannot read them inside the loop. If you put values on the return stack
1670: within a loop, you have to remove them before the end of the loop and
1671: before accessing the index of the loop.
1672:
1673: There are several variations on the counted loop:
1674:
1675: @code{LEAVE} leaves the innermost counted loop immediately.
1676:
1677: If @var{start} is greater than @var{limit}, a @code{?DO} loop is entered
1678: (and @code{LOOP} iterates until they become equal by wrap-around
1679: arithmetic). This behaviour is usually not what you want. Therefore,
1680: Gforth offers @code{+DO} and @code{U+DO} (as replacements for
1681: @code{?DO}), which do not enter the loop if @var{start} is greater than
1682: @var{limit}; @code{+DO} is for signed loop parameters, @code{U+DO} for
1683: unsigned loop parameters.
1684:
1685: @code{LOOP} can be replaced with @code{@var{n} +LOOP}; this updates the
1686: index by @var{n} instead of by 1. The loop is terminated when the border
1687: between @var{limit-1} and @var{limit} is crossed. E.g.:
1688:
1689: @code{4 0 +DO i . 2 +LOOP} prints @code{0 2}
1690:
1691: @code{4 1 +DO i . 2 +LOOP} prints @code{1 3}
1692:
1693: @cindex negative increment for counted loops
1694: @cindex counted loops with negative increment
1695: The behaviour of @code{@var{n} +LOOP} is peculiar when @var{n} is negative:
1696:
1697: @code{-1 0 ?DO i . -1 +LOOP} prints @code{0 -1}
1698:
1699: @code{ 0 0 ?DO i . -1 +LOOP} prints nothing
1700:
1701: Therefore we recommend avoiding @code{@var{n} +LOOP} with negative
1702: @var{n}. One alternative is @code{@var{u} -LOOP}, which reduces the
1703: index by @var{u} each iteration. The loop is terminated when the border
1704: between @var{limit+1} and @var{limit} is crossed. Gforth also provides
1705: @code{-DO} and @code{U-DO} for down-counting loops. E.g.:
1706:
1707: @code{-2 0 -DO i . 1 -LOOP} prints @code{0 -1}
1708:
1709: @code{-1 0 -DO i . 1 -LOOP} prints @code{0}
1710:
1711: @code{ 0 0 -DO i . 1 -LOOP} prints nothing
1712:
1713: Unfortunately, @code{+DO}, @code{U+DO}, @code{-DO}, @code{U-DO} and
1714: @code{-LOOP} are not in the ANS Forth standard. However, an
1715: implementation for these words that uses only standard words is provided
1716: in @file{compat/loops.fs}.
1717:
1718: @code{?DO} can also be replaced by @code{DO}. @code{DO} always enters
1719: the loop, independent of the loop parameters. Do not use @code{DO}, even
1720: if you know that the loop is entered in any case. Such knowledge tends
1721: to become invalid during maintenance of a program, and then the
1722: @code{DO} will make trouble.
1723:
1724: @code{UNLOOP} is used to prepare for an abnormal loop exit, e.g., via
1725: @code{EXIT}. @code{UNLOOP} removes the loop control parameters from the
1726: return stack so @code{EXIT} can get to its return address.
1727:
1728: @cindex @code{FOR} loops
1729: Another counted loop is
1730: @example
1731: @var{n}
1732: FOR
1733: @var{body}
1734: NEXT
1735: @end example
1736: This is the preferred loop of native code compiler writers who are too
1737: lazy to optimize @code{?DO} loops properly. In Gforth, this loop
1738: iterates @var{n+1} times; @code{i} produces values starting with @var{n}
1739: and ending with 0. Other Forth systems may behave differently, even if
1740: they support @code{FOR} loops. To avoid problems, don't use @code{FOR}
1741: loops.
1742:
1743: @node Arbitrary control structures, Calls and returns, Counted Loops, Control Structures
1744: @subsection Arbitrary control structures
1745: @cindex control structures, user-defined
1746:
1747: @cindex control-flow stack
1748: ANS Forth permits and supports using control structures in a non-nested
1749: way. Information about incomplete control structures is stored on the
1750: control-flow stack. This stack may be implemented on the Forth data
1751: stack, and this is what we have done in Gforth.
1752:
1753: @cindex @code{orig}, control-flow stack item
1754: @cindex @code{dest}, control-flow stack item
1755: An @i{orig} entry represents an unresolved forward branch, a @i{dest}
1756: entry represents a backward branch target. A few words are the basis for
1757: building any control structure possible (except control structures that
1758: need storage, like calls, coroutines, and backtracking).
1759:
1760: doc-if
1761: doc-ahead
1762: doc-then
1763: doc-begin
1764: doc-until
1765: doc-again
1766: doc-cs-pick
1767: doc-cs-roll
1768:
1769: On many systems control-flow stack items take one word, in Gforth they
1770: currently take three (this may change in the future). Therefore it is a
1771: really good idea to manipulate the control flow stack with
1772: @code{cs-pick} and @code{cs-roll}, not with data stack manipulation
1773: words.
1774:
1775: Some standard control structure words are built from these words:
1776:
1777: doc-else
1778: doc-while
1779: doc-repeat
1780:
1781: Gforth adds some more control-structure words:
1782:
1783: doc-endif
1784: doc-?dup-if
1785: doc-?dup-0=-if
1786:
1787: Counted loop words constitute a separate group of words:
1788:
1789: doc-?do
1790: doc-+do
1791: doc-u+do
1792: doc--do
1793: doc-u-do
1794: doc-do
1795: doc-for
1796: doc-loop
1797: doc-+loop
1798: doc--loop
1799: doc-next
1800: doc-leave
1801: doc-?leave
1802: doc-unloop
1803: doc-done
1804:
1805: The standard does not allow using @code{cs-pick} and @code{cs-roll} on
1806: @i{do-sys}. Our system allows it, but it's your job to ensure that for
1807: every @code{?DO} etc. there is exactly one @code{UNLOOP} on any path
1808: through the definition (@code{LOOP} etc. compile an @code{UNLOOP} on the
1809: fall-through path). Also, you have to ensure that all @code{LEAVE}s are
1810: resolved (by using one of the loop-ending words or @code{DONE}).
1811:
1812: Another group of control structure words are
1813:
1814: doc-case
1815: doc-endcase
1816: doc-of
1817: doc-endof
1818:
1819: @i{case-sys} and @i{of-sys} cannot be processed using @code{cs-pick} and
1820: @code{cs-roll}.
1821:
1822: @subsubsection Programming Style
1823:
1824: In order to ensure readability we recommend that you do not create
1825: arbitrary control structures directly, but define new control structure
1826: words for the control structure you want and use these words in your
1827: program.
1828:
1829: E.g., instead of writing
1830:
1831: @example
1832: begin
1833: ...
1834: if [ 1 cs-roll ]
1835: ...
1836: again then
1837: @end example
1838:
1839: we recommend defining control structure words, e.g.,
1840:
1841: @example
1842: : while ( dest -- orig dest )
1843: POSTPONE if
1844: 1 cs-roll ; immediate
1845:
1846: : repeat ( orig dest -- )
1847: POSTPONE again
1848: POSTPONE then ; immediate
1849: @end example
1850:
1851: and then using these to create the control structure:
1852:
1853: @example
1854: begin
1855: ...
1856: while
1857: ...
1858: repeat
1859: @end example
1860:
1861: That's much easier to read, isn't it? Of course, @code{REPEAT} and
1862: @code{WHILE} are predefined, so in this example it would not be
1863: necessary to define them.
1864:
1865: @node Calls and returns, Exception Handling, Arbitrary control structures, Control Structures
1866: @subsection Calls and returns
1867: @cindex calling a definition
1868: @cindex returning from a definition
1869:
1870: @cindex recursive definitions
1871: A definition can be called simply be writing the name of the definition
1872: to be called. Note that normally a definition is invisible during its
1873: definition. If you want to write a directly recursive definition, you
1874: can use @code{recursive} to make the current definition visible.
1875:
1876: doc-recursive
1877:
1878: Another way to perform a recursive call is
1879:
1880: doc-recurse
1881:
1882: @quotation
1883: @progstyle
1884: I prefer using @code{recursive} to @code{recurse}, because calling the
1885: definition by name is more descriptive (if the name is well-chosen) than
1886: the somewhat cryptic @code{recurse}. E.g., in a quicksort
1887: implementation, it is much better to read (and think) ``now sort the
1888: partitions'' than to read ``now do a recursive call''.
1889: @end quotation
1890:
1891: For mutual recursion, use @code{defer}red words, like this:
1892:
1893: @example
1894: defer foo
1895:
1896: : bar ( ... -- ... )
1897: ... foo ... ;
1898:
1899: :noname ( ... -- ... )
1900: ... bar ... ;
1901: IS foo
1902: @end example
1903:
1904: When the end of the definition is reached, it returns. An earlier return
1905: can be forced using
1906:
1907: doc-exit
1908:
1909: Don't forget to clean up the return stack and @code{UNLOOP} any
1910: outstanding @code{?DO}...@code{LOOP}s before @code{EXIT}ing. The
1911: primitive compiled by @code{EXIT} is
1912:
1913: doc-;s
1914:
1915: @node Exception Handling, , Calls and returns, Control Structures
1916: @subsection Exception Handling
1917: @cindex Exceptions
1918:
1919: doc-catch
1920: doc-throw
1921:
1922: @node Locals, Defining Words, Control Structures, Words
1923: @section Locals
1924: @cindex locals
1925:
1926: Local variables can make Forth programming more enjoyable and Forth
1927: programs easier to read. Unfortunately, the locals of ANS Forth are
1928: laden with restrictions. Therefore, we provide not only the ANS Forth
1929: locals wordset, but also our own, more powerful locals wordset (we
1930: implemented the ANS Forth locals wordset through our locals wordset).
1931:
1932: The ideas in this section have also been published in the paper
1933: @cite{Automatic Scoping of Local Variables} by M. Anton Ertl, presented
1934: at EuroForth '94; it is available at
1935: @*@url{http://www.complang.tuwien.ac.at/papers/ertl94l.ps.gz}.
1936:
1937: @menu
1938: * Gforth locals::
1939: * ANS Forth locals::
1940: @end menu
1941:
1942: @node Gforth locals, ANS Forth locals, Locals, Locals
1943: @subsection Gforth locals
1944: @cindex Gforth locals
1945: @cindex locals, Gforth style
1946:
1947: Locals can be defined with
1948:
1949: @example
1950: @{ local1 local2 ... -- comment @}
1951: @end example
1952: or
1953: @example
1954: @{ local1 local2 ... @}
1955: @end example
1956:
1957: E.g.,
1958: @example
1959: : max @{ n1 n2 -- n3 @}
1960: n1 n2 > if
1961: n1
1962: else
1963: n2
1964: endif ;
1965: @end example
1966:
1967: The similarity of locals definitions with stack comments is intended. A
1968: locals definition often replaces the stack comment of a word. The order
1969: of the locals corresponds to the order in a stack comment and everything
1970: after the @code{--} is really a comment.
1971:
1972: This similarity has one disadvantage: It is too easy to confuse locals
1973: declarations with stack comments, causing bugs and making them hard to
1974: find. However, this problem can be avoided by appropriate coding
1975: conventions: Do not use both notations in the same program. If you do,
1976: they should be distinguished using additional means, e.g. by position.
1977:
1978: @cindex types of locals
1979: @cindex locals types
1980: The name of the local may be preceded by a type specifier, e.g.,
1981: @code{F:} for a floating point value:
1982:
1983: @example
1984: : CX* @{ F: Ar F: Ai F: Br F: Bi -- Cr Ci @}
1985: \ complex multiplication
1986: Ar Br f* Ai Bi f* f-
1987: Ar Bi f* Ai Br f* f+ ;
1988: @end example
1989:
1990: @cindex flavours of locals
1991: @cindex locals flavours
1992: @cindex value-flavoured locals
1993: @cindex variable-flavoured locals
1994: Gforth currently supports cells (@code{W:}, @code{W^}), doubles
1995: (@code{D:}, @code{D^}), floats (@code{F:}, @code{F^}) and characters
1996: (@code{C:}, @code{C^}) in two flavours: a value-flavoured local (defined
1997: with @code{W:}, @code{D:} etc.) produces its value and can be changed
1998: with @code{TO}. A variable-flavoured local (defined with @code{W^} etc.)
1999: produces its address (which becomes invalid when the variable's scope is
2000: left). E.g., the standard word @code{emit} can be defined in terms of
2001: @code{type} like this:
2002:
2003: @example
2004: : emit @{ C^ char* -- @}
2005: char* 1 type ;
2006: @end example
2007:
2008: @cindex default type of locals
2009: @cindex locals, default type
2010: A local without type specifier is a @code{W:} local. Both flavours of
2011: locals are initialized with values from the data or FP stack.
2012:
2013: Currently there is no way to define locals with user-defined data
2014: structures, but we are working on it.
2015:
2016: Gforth allows defining locals everywhere in a colon definition. This
2017: poses the following questions:
2018:
2019: @menu
2020: * Where are locals visible by name?::
2021: * How long do locals live?::
2022: * Programming Style::
2023: * Implementation::
2024: @end menu
2025:
2026: @node Where are locals visible by name?, How long do locals live?, Gforth locals, Gforth locals
2027: @subsubsection Where are locals visible by name?
2028: @cindex locals visibility
2029: @cindex visibility of locals
2030: @cindex scope of locals
2031:
2032: Basically, the answer is that locals are visible where you would expect
2033: it in block-structured languages, and sometimes a little longer. If you
2034: want to restrict the scope of a local, enclose its definition in
2035: @code{SCOPE}...@code{ENDSCOPE}.
2036:
2037: doc-scope
2038: doc-endscope
2039:
2040: These words behave like control structure words, so you can use them
2041: with @code{CS-PICK} and @code{CS-ROLL} to restrict the scope in
2042: arbitrary ways.
2043:
2044: If you want a more exact answer to the visibility question, here's the
2045: basic principle: A local is visible in all places that can only be
2046: reached through the definition of the local@footnote{In compiler
2047: construction terminology, all places dominated by the definition of the
2048: local.}. In other words, it is not visible in places that can be reached
2049: without going through the definition of the local. E.g., locals defined
2050: in @code{IF}...@code{ENDIF} are visible until the @code{ENDIF}, locals
2051: defined in @code{BEGIN}...@code{UNTIL} are visible after the
2052: @code{UNTIL} (until, e.g., a subsequent @code{ENDSCOPE}).
2053:
2054: The reasoning behind this solution is: We want to have the locals
2055: visible as long as it is meaningful. The user can always make the
2056: visibility shorter by using explicit scoping. In a place that can
2057: only be reached through the definition of a local, the meaning of a
2058: local name is clear. In other places it is not: How is the local
2059: initialized at the control flow path that does not contain the
2060: definition? Which local is meant, if the same name is defined twice in
2061: two independent control flow paths?
2062:
2063: This should be enough detail for nearly all users, so you can skip the
2064: rest of this section. If you really must know all the gory details and
2065: options, read on.
2066:
2067: In order to implement this rule, the compiler has to know which places
2068: are unreachable. It knows this automatically after @code{AHEAD},
2069: @code{AGAIN}, @code{EXIT} and @code{LEAVE}; in other cases (e.g., after
2070: most @code{THROW}s), you can use the word @code{UNREACHABLE} to tell the
2071: compiler that the control flow never reaches that place. If
2072: @code{UNREACHABLE} is not used where it could, the only consequence is
2073: that the visibility of some locals is more limited than the rule above
2074: says. If @code{UNREACHABLE} is used where it should not (i.e., if you
2075: lie to the compiler), buggy code will be produced.
2076:
2077: doc-unreachable
2078:
2079: Another problem with this rule is that at @code{BEGIN}, the compiler
2080: does not know which locals will be visible on the incoming
2081: back-edge. All problems discussed in the following are due to this
2082: ignorance of the compiler (we discuss the problems using @code{BEGIN}
2083: loops as examples; the discussion also applies to @code{?DO} and other
2084: loops). Perhaps the most insidious example is:
2085: @example
2086: AHEAD
2087: BEGIN
2088: x
2089: [ 1 CS-ROLL ] THEN
2090: @{ x @}
2091: ...
2092: UNTIL
2093: @end example
2094:
2095: This should be legal according to the visibility rule. The use of
2096: @code{x} can only be reached through the definition; but that appears
2097: textually below the use.
2098:
2099: From this example it is clear that the visibility rules cannot be fully
2100: implemented without major headaches. Our implementation treats common
2101: cases as advertised and the exceptions are treated in a safe way: The
2102: compiler makes a reasonable guess about the locals visible after a
2103: @code{BEGIN}; if it is too pessimistic, the
2104: user will get a spurious error about the local not being defined; if the
2105: compiler is too optimistic, it will notice this later and issue a
2106: warning. In the case above the compiler would complain about @code{x}
2107: being undefined at its use. You can see from the obscure examples in
2108: this section that it takes quite unusual control structures to get the
2109: compiler into trouble, and even then it will often do fine.
2110:
2111: If the @code{BEGIN} is reachable from above, the most optimistic guess
2112: is that all locals visible before the @code{BEGIN} will also be
2113: visible after the @code{BEGIN}. This guess is valid for all loops that
2114: are entered only through the @code{BEGIN}, in particular, for normal
2115: @code{BEGIN}...@code{WHILE}...@code{REPEAT} and
2116: @code{BEGIN}...@code{UNTIL} loops and it is implemented in our
2117: compiler. When the branch to the @code{BEGIN} is finally generated by
2118: @code{AGAIN} or @code{UNTIL}, the compiler checks the guess and
2119: warns the user if it was too optimistic:
2120: @example
2121: IF
2122: @{ x @}
2123: BEGIN
2124: \ x ?
2125: [ 1 cs-roll ] THEN
2126: ...
2127: UNTIL
2128: @end example
2129:
2130: Here, @code{x} lives only until the @code{BEGIN}, but the compiler
2131: optimistically assumes that it lives until the @code{THEN}. It notices
2132: this difference when it compiles the @code{UNTIL} and issues a
2133: warning. The user can avoid the warning, and make sure that @code{x}
2134: is not used in the wrong area by using explicit scoping:
2135: @example
2136: IF
2137: SCOPE
2138: @{ x @}
2139: ENDSCOPE
2140: BEGIN
2141: [ 1 cs-roll ] THEN
2142: ...
2143: UNTIL
2144: @end example
2145:
2146: Since the guess is optimistic, there will be no spurious error messages
2147: about undefined locals.
2148:
2149: If the @code{BEGIN} is not reachable from above (e.g., after
2150: @code{AHEAD} or @code{EXIT}), the compiler cannot even make an
2151: optimistic guess, as the locals visible after the @code{BEGIN} may be
2152: defined later. Therefore, the compiler assumes that no locals are
2153: visible after the @code{BEGIN}. However, the user can use
2154: @code{ASSUME-LIVE} to make the compiler assume that the same locals are
2155: visible at the BEGIN as at the point where the top control-flow stack
2156: item was created.
2157:
2158: doc-assume-live
2159:
2160: E.g.,
2161: @example
2162: @{ x @}
2163: AHEAD
2164: ASSUME-LIVE
2165: BEGIN
2166: x
2167: [ 1 CS-ROLL ] THEN
2168: ...
2169: UNTIL
2170: @end example
2171:
2172: Other cases where the locals are defined before the @code{BEGIN} can be
2173: handled by inserting an appropriate @code{CS-ROLL} before the
2174: @code{ASSUME-LIVE} (and changing the control-flow stack manipulation
2175: behind the @code{ASSUME-LIVE}).
2176:
2177: Cases where locals are defined after the @code{BEGIN} (but should be
2178: visible immediately after the @code{BEGIN}) can only be handled by
2179: rearranging the loop. E.g., the ``most insidious'' example above can be
2180: arranged into:
2181: @example
2182: BEGIN
2183: @{ x @}
2184: ... 0=
2185: WHILE
2186: x
2187: REPEAT
2188: @end example
2189:
2190: @node How long do locals live?, Programming Style, Where are locals visible by name?, Gforth locals
2191: @subsubsection How long do locals live?
2192: @cindex locals lifetime
2193: @cindex lifetime of locals
2194:
2195: The right answer for the lifetime question would be: A local lives at
2196: least as long as it can be accessed. For a value-flavoured local this
2197: means: until the end of its visibility. However, a variable-flavoured
2198: local could be accessed through its address far beyond its visibility
2199: scope. Ultimately, this would mean that such locals would have to be
2200: garbage collected. Since this entails un-Forth-like implementation
2201: complexities, I adopted the same cowardly solution as some other
2202: languages (e.g., C): The local lives only as long as it is visible;
2203: afterwards its address is invalid (and programs that access it
2204: afterwards are erroneous).
2205:
2206: @node Programming Style, Implementation, How long do locals live?, Gforth locals
2207: @subsubsection Programming Style
2208: @cindex locals programming style
2209: @cindex programming style, locals
2210:
2211: The freedom to define locals anywhere has the potential to change
2212: programming styles dramatically. In particular, the need to use the
2213: return stack for intermediate storage vanishes. Moreover, all stack
2214: manipulations (except @code{PICK}s and @code{ROLL}s with run-time
2215: determined arguments) can be eliminated: If the stack items are in the
2216: wrong order, just write a locals definition for all of them; then
2217: write the items in the order you want.
2218:
2219: This seems a little far-fetched and eliminating stack manipulations is
2220: unlikely to become a conscious programming objective. Still, the number
2221: of stack manipulations will be reduced dramatically if local variables
2222: are used liberally (e.g., compare @code{max} in @ref{Gforth locals} with
2223: a traditional implementation of @code{max}).
2224:
2225: This shows one potential benefit of locals: making Forth programs more
2226: readable. Of course, this benefit will only be realized if the
2227: programmers continue to honour the principle of factoring instead of
2228: using the added latitude to make the words longer.
2229:
2230: @cindex single-assignment style for locals
2231: Using @code{TO} can and should be avoided. Without @code{TO},
2232: every value-flavoured local has only a single assignment and many
2233: advantages of functional languages apply to Forth. I.e., programs are
2234: easier to analyse, to optimize and to read: It is clear from the
2235: definition what the local stands for, it does not turn into something
2236: different later.
2237:
2238: E.g., a definition using @code{TO} might look like this:
2239: @example
2240: : strcmp @{ addr1 u1 addr2 u2 -- n @}
2241: u1 u2 min 0
2242: ?do
2243: addr1 c@@ addr2 c@@ -
2244: ?dup-if
2245: unloop exit
2246: then
2247: addr1 char+ TO addr1
2248: addr2 char+ TO addr2
2249: loop
2250: u1 u2 - ;
2251: @end example
2252: Here, @code{TO} is used to update @code{addr1} and @code{addr2} at
2253: every loop iteration. @code{strcmp} is a typical example of the
2254: readability problems of using @code{TO}. When you start reading
2255: @code{strcmp}, you think that @code{addr1} refers to the start of the
2256: string. Only near the end of the loop you realize that it is something
2257: else.
2258:
2259: This can be avoided by defining two locals at the start of the loop that
2260: are initialized with the right value for the current iteration.
2261: @example
2262: : strcmp @{ addr1 u1 addr2 u2 -- n @}
2263: addr1 addr2
2264: u1 u2 min 0
2265: ?do @{ s1 s2 @}
2266: s1 c@@ s2 c@@ -
2267: ?dup-if
2268: unloop exit
2269: then
2270: s1 char+ s2 char+
2271: loop
2272: 2drop
2273: u1 u2 - ;
2274: @end example
2275: Here it is clear from the start that @code{s1} has a different value
2276: in every loop iteration.
2277:
2278: @node Implementation, , Programming Style, Gforth locals
2279: @subsubsection Implementation
2280: @cindex locals implementation
2281: @cindex implementation of locals
2282:
2283: @cindex locals stack
2284: Gforth uses an extra locals stack. The most compelling reason for
2285: this is that the return stack is not float-aligned; using an extra stack
2286: also eliminates the problems and restrictions of using the return stack
2287: as locals stack. Like the other stacks, the locals stack grows toward
2288: lower addresses. A few primitives allow an efficient implementation:
2289:
2290: doc-@local#
2291: doc-f@local#
2292: doc-laddr#
2293: doc-lp+!#
2294: doc-lp!
2295: doc->l
2296: doc-f>l
2297:
2298: In addition to these primitives, some specializations of these
2299: primitives for commonly occurring inline arguments are provided for
2300: efficiency reasons, e.g., @code{@@local0} as specialization of
2301: @code{@@local#} for the inline argument 0. The following compiling words
2302: compile the right specialized version, or the general version, as
2303: appropriate:
2304:
2305: doc-compile-@local
2306: doc-compile-f@local
2307: doc-compile-lp+!
2308:
2309: Combinations of conditional branches and @code{lp+!#} like
2310: @code{?branch-lp+!#} (the locals pointer is only changed if the branch
2311: is taken) are provided for efficiency and correctness in loops.
2312:
2313: A special area in the dictionary space is reserved for keeping the
2314: local variable names. @code{@{} switches the dictionary pointer to this
2315: area and @code{@}} switches it back and generates the locals
2316: initializing code. @code{W:} etc.@ are normal defining words. This
2317: special area is cleared at the start of every colon definition.
2318:
2319: @cindex wordlist for defining locals
2320: A special feature of Gforth's dictionary is used to implement the
2321: definition of locals without type specifiers: every wordlist (aka
2322: vocabulary) has its own methods for searching
2323: etc. (@pxref{Wordlists}). For the present purpose we defined a wordlist
2324: with a special search method: When it is searched for a word, it
2325: actually creates that word using @code{W:}. @code{@{} changes the search
2326: order to first search the wordlist containing @code{@}}, @code{W:} etc.,
2327: and then the wordlist for defining locals without type specifiers.
2328:
2329: The lifetime rules support a stack discipline within a colon
2330: definition: The lifetime of a local is either nested with other locals
2331: lifetimes or it does not overlap them.
2332:
2333: At @code{BEGIN}, @code{IF}, and @code{AHEAD} no code for locals stack
2334: pointer manipulation is generated. Between control structure words
2335: locals definitions can push locals onto the locals stack. @code{AGAIN}
2336: is the simplest of the other three control flow words. It has to
2337: restore the locals stack depth of the corresponding @code{BEGIN}
2338: before branching. The code looks like this:
2339: @format
2340: @code{lp+!#} current-locals-size @minus{} dest-locals-size
2341: @code{branch} <begin>
2342: @end format
2343:
2344: @code{UNTIL} is a little more complicated: If it branches back, it
2345: must adjust the stack just like @code{AGAIN}. But if it falls through,
2346: the locals stack must not be changed. The compiler generates the
2347: following code:
2348: @format
2349: @code{?branch-lp+!#} <begin> current-locals-size @minus{} dest-locals-size
2350: @end format
2351: The locals stack pointer is only adjusted if the branch is taken.
2352:
2353: @code{THEN} can produce somewhat inefficient code:
2354: @format
2355: @code{lp+!#} current-locals-size @minus{} orig-locals-size
2356: <orig target>:
2357: @code{lp+!#} orig-locals-size @minus{} new-locals-size
2358: @end format
2359: The second @code{lp+!#} adjusts the locals stack pointer from the
2360: level at the @var{orig} point to the level after the @code{THEN}. The
2361: first @code{lp+!#} adjusts the locals stack pointer from the current
2362: level to the level at the orig point, so the complete effect is an
2363: adjustment from the current level to the right level after the
2364: @code{THEN}.
2365:
2366: @cindex locals information on the control-flow stack
2367: @cindex control-flow stack items, locals information
2368: In a conventional Forth implementation a dest control-flow stack entry
2369: is just the target address and an orig entry is just the address to be
2370: patched. Our locals implementation adds a wordlist to every orig or dest
2371: item. It is the list of locals visible (or assumed visible) at the point
2372: described by the entry. Our implementation also adds a tag to identify
2373: the kind of entry, in particular to differentiate between live and dead
2374: (reachable and unreachable) orig entries.
2375:
2376: A few unusual operations have to be performed on locals wordlists:
2377:
2378: doc-common-list
2379: doc-sub-list?
2380: doc-list-size
2381:
2382: Several features of our locals wordlist implementation make these
2383: operations easy to implement: The locals wordlists are organised as
2384: linked lists; the tails of these lists are shared, if the lists
2385: contain some of the same locals; and the address of a name is greater
2386: than the address of the names behind it in the list.
2387:
2388: Another important implementation detail is the variable
2389: @code{dead-code}. It is used by @code{BEGIN} and @code{THEN} to
2390: determine if they can be reached directly or only through the branch
2391: that they resolve. @code{dead-code} is set by @code{UNREACHABLE},
2392: @code{AHEAD}, @code{EXIT} etc., and cleared at the start of a colon
2393: definition, by @code{BEGIN} and usually by @code{THEN}.
2394:
2395: Counted loops are similar to other loops in most respects, but
2396: @code{LEAVE} requires special attention: It performs basically the same
2397: service as @code{AHEAD}, but it does not create a control-flow stack
2398: entry. Therefore the information has to be stored elsewhere;
2399: traditionally, the information was stored in the target fields of the
2400: branches created by the @code{LEAVE}s, by organizing these fields into a
2401: linked list. Unfortunately, this clever trick does not provide enough
2402: space for storing our extended control flow information. Therefore, we
2403: introduce another stack, the leave stack. It contains the control-flow
2404: stack entries for all unresolved @code{LEAVE}s.
2405:
2406: Local names are kept until the end of the colon definition, even if
2407: they are no longer visible in any control-flow path. In a few cases
2408: this may lead to increased space needs for the locals name area, but
2409: usually less than reclaiming this space would cost in code size.
2410:
2411:
2412: @node ANS Forth locals, , Gforth locals, Locals
2413: @subsection ANS Forth locals
2414: @cindex locals, ANS Forth style
2415:
2416: The ANS Forth locals wordset does not define a syntax for locals, but
2417: words that make it possible to define various syntaxes. One of the
2418: possible syntaxes is a subset of the syntax we used in the Gforth locals
2419: wordset, i.e.:
2420:
2421: @example
2422: @{ local1 local2 ... -- comment @}
2423: @end example
2424: or
2425: @example
2426: @{ local1 local2 ... @}
2427: @end example
2428:
2429: The order of the locals corresponds to the order in a stack comment. The
2430: restrictions are:
2431:
2432: @itemize @bullet
2433: @item
2434: Locals can only be cell-sized values (no type specifiers are allowed).
2435: @item
2436: Locals can be defined only outside control structures.
2437: @item
2438: Locals can interfere with explicit usage of the return stack. For the
2439: exact (and long) rules, see the standard. If you don't use return stack
2440: accessing words in a definition using locals, you will be all right. The
2441: purpose of this rule is to make locals implementation on the return
2442: stack easier.
2443: @item
2444: The whole definition must be in one line.
2445: @end itemize
2446:
2447: Locals defined in this way behave like @code{VALUE}s (@xref{Simple
2448: Defining Words}). I.e., they are initialized from the stack. Using their
2449: name produces their value. Their value can be changed using @code{TO}.
2450:
2451: Since this syntax is supported by Gforth directly, you need not do
2452: anything to use it. If you want to port a program using this syntax to
2453: another ANS Forth system, use @file{compat/anslocal.fs} to implement the
2454: syntax on the other system.
2455:
2456: Note that a syntax shown in the standard, section A.13 looks
2457: similar, but is quite different in having the order of locals
2458: reversed. Beware!
2459:
2460: The ANS Forth locals wordset itself consists of the following word
2461:
2462: doc-(local)
2463:
2464: The ANS Forth locals extension wordset defines a syntax, but it is so
2465: awful that we strongly recommend not to use it. We have implemented this
2466: syntax to make porting to Gforth easy, but do not document it here. The
2467: problem with this syntax is that the locals are defined in an order
2468: reversed with respect to the standard stack comment notation, making
2469: programs harder to read, and easier to misread and miswrite. The only
2470: merit of this syntax is that it is easy to implement using the ANS Forth
2471: locals wordset.
2472:
2473: @node Defining Words, Structures, Locals, Words
2474: @section Defining Words
2475: @cindex defining words
2476:
2477: @menu
2478: * Simple Defining Words::
2479: * Colon Definitions::
2480: * User-defined Defining Words::
2481: * Supplying names::
2482: * Interpretation and Compilation Semantics::
2483: @end menu
2484:
2485: @node Simple Defining Words, Colon Definitions, Defining Words, Defining Words
2486: @subsection Simple Defining Words
2487: @cindex simple defining words
2488: @cindex defining words, simple
2489:
2490: doc-constant
2491: doc-2constant
2492: doc-fconstant
2493: doc-variable
2494: doc-2variable
2495: doc-fvariable
2496: doc-create
2497: doc-user
2498: doc-value
2499: doc-to
2500: doc-defer
2501: doc-is
2502:
2503: @node Colon Definitions, User-defined Defining Words, Simple Defining Words, Defining Words
2504: @subsection Colon Definitions
2505: @cindex colon definitions
2506:
2507: @example
2508: : name ( ... -- ... )
2509: word1 word2 word3 ;
2510: @end example
2511:
2512: creates a word called @code{name}, that, upon execution, executes
2513: @code{word1 word2 word3}. @code{name} is a @dfn{(colon) definition}.
2514:
2515: The explanation above is somewhat superficial. @xref{Interpretation and
2516: Compilation Semantics} for an in-depth discussion of some of the issues
2517: involved.
2518:
2519: doc-:
2520: doc-;
2521:
2522: @node User-defined Defining Words, Supplying names, Colon Definitions, Defining Words
2523: @subsection User-defined Defining Words
2524: @cindex user-defined defining words
2525: @cindex defining words, user-defined
2526:
2527: You can create new defining words simply by wrapping defining-time code
2528: around existing defining words and putting the sequence in a colon
2529: definition.
2530:
2531: @cindex @code{CREATE} ... @code{DOES>}
2532: If you want the words defined with your defining words to behave
2533: differently from words defined with standard defining words, you can
2534: write your defining word like this:
2535:
2536: @example
2537: : def-word ( "name" -- )
2538: Create @var{code1}
2539: DOES> ( ... -- ... )
2540: @var{code2} ;
2541:
2542: def-word name
2543: @end example
2544:
2545: Technically, this fragment defines a defining word @code{def-word}, and
2546: a word @code{name}; when you execute @code{name}, the address of the
2547: body of @code{name} is put on the data stack and @var{code2} is executed
2548: (the address of the body of @code{name} is the address @code{HERE}
2549: returns immediately after the @code{CREATE}).
2550:
2551: In other words, if you make the following definitions:
2552:
2553: @example
2554: : def-word1 ( "name" -- )
2555: Create @var{code1} ;
2556:
2557: : action1 ( ... -- ... )
2558: @var{code2} ;
2559:
2560: def-word name1
2561: @end example
2562:
2563: Using @code{name1 action1} is equivalent to using @code{name}.
2564:
2565: E.g., you can implement @code{Constant} in this way:
2566:
2567: @example
2568: : constant ( w "name" -- )
2569: create ,
2570: DOES> ( -- w )
2571: @@ ;
2572: @end example
2573:
2574: When you create a constant with @code{5 constant five}, first a new word
2575: @code{five} is created, then the value 5 is laid down in the body of
2576: @code{five} with @code{,}. When @code{five} is invoked, the address of
2577: the body is put on the stack, and @code{@@} retrieves the value 5.
2578:
2579: @cindex stack effect of @code{DOES>}-parts
2580: @cindex @code{DOES>}-parts, stack effect
2581: In the example above the stack comment after the @code{DOES>} specifies
2582: the stack effect of the defined words, not the stack effect of the
2583: following code (the following code expects the address of the body on
2584: the top of stack, which is not reflected in the stack comment). This is
2585: the convention that I use and recommend (it clashes a bit with using
2586: locals declarations for stack effect specification, though).
2587:
2588: @subsubsection Applications of @code{CREATE..DOES>}
2589: @cindex @code{CREATE} ... @code{DOES>}, applications
2590:
2591: You may wonder how to use this feature. Here are some usage patterns:
2592:
2593: @cindex factoring similar colon definitions
2594: When you see a sequence of code occurring several times, and you can
2595: identify a meaning, you will factor it out as a colon definition. When
2596: you see similar colon definitions, you can factor them using
2597: @code{CREATE..DOES>}. E.g., an assembler usually defines several words
2598: that look very similar:
2599: @example
2600: : ori, ( reg-target reg-source n -- )
2601: 0 asm-reg-reg-imm ;
2602: : andi, ( reg-target reg-source n -- )
2603: 1 asm-reg-reg-imm ;
2604: @end example
2605:
2606: This could be factored with:
2607: @example
2608: : reg-reg-imm ( op-code -- )
2609: create ,
2610: DOES> ( reg-target reg-source n -- )
2611: @@ asm-reg-reg-imm ;
2612:
2613: 0 reg-reg-imm ori,
2614: 1 reg-reg-imm andi,
2615: @end example
2616:
2617: @cindex currying
2618: Another view of @code{CREATE..DOES>} is to consider it as a crude way to
2619: supply a part of the parameters for a word (known as @dfn{currying} in
2620: the functional language community). E.g., @code{+} needs two
2621: parameters. Creating versions of @code{+} with one parameter fixed can
2622: be done like this:
2623: @example
2624: : curry+ ( n1 -- )
2625: create ,
2626: DOES> ( n2 -- n1+n2 )
2627: @@ + ;
2628:
2629: 3 curry+ 3+
2630: -2 curry+ 2-
2631: @end example
2632:
2633: @subsubsection The gory details of @code{CREATE..DOES>}
2634: @cindex @code{CREATE} ... @code{DOES>}, details
2635:
2636: doc-does>
2637:
2638: @cindex @code{DOES>} in a separate definition
2639: This means that you need not use @code{CREATE} and @code{DOES>} in the
2640: same definition; E.g., you can put the @code{DOES>}-part in a separate
2641: definition. This allows us to, e.g., select among different DOES>-parts:
2642: @example
2643: : does1
2644: DOES> ( ... -- ... )
2645: ... ;
2646:
2647: : does2
2648: DOES> ( ... -- ... )
2649: ... ;
2650:
2651: : def-word ( ... -- ... )
2652: create ...
2653: IF
2654: does1
2655: ELSE
2656: does2
2657: ENDIF ;
2658: @end example
2659:
2660: @cindex @code{DOES>} in interpretation state
2661: In a standard program you can apply a @code{DOES>}-part only if the last
2662: word was defined with @code{CREATE}. In Gforth, the @code{DOES>}-part
2663: will override the behaviour of the last word defined in any case. In a
2664: standard program, you can use @code{DOES>} only in a colon
2665: definition. In Gforth, you can also use it in interpretation state, in a
2666: kind of one-shot mode:
2667: @example
2668: CREATE name ( ... -- ... )
2669: @var{initialization}
2670: DOES>
2671: @var{code} ;
2672: @end example
2673: This is equivalent to the standard
2674: @example
2675: :noname
2676: DOES>
2677: @var{code} ;
2678: CREATE name EXECUTE ( ... -- ... )
2679: @var{initialization}
2680: @end example
2681:
2682: You can get the address of the body of a word with
2683:
2684: doc->body
2685:
2686: @node Supplying names, Interpretation and Compilation Semantics, User-defined Defining Words, Defining Words
2687: @subsection Supplying names for the defined words
2688: @cindex names for defined words
2689: @cindex defining words, name parameter
2690:
2691: @cindex defining words, name given in a string
2692: By default, defining words take the names for the defined words from the
2693: input stream. Sometimes you want to supply the name from a string. You
2694: can do this with
2695:
2696: doc-nextname
2697:
2698: E.g.,
2699:
2700: @example
2701: s" foo" nextname create
2702: @end example
2703: is equivalent to
2704: @example
2705: create foo
2706: @end example
2707:
2708: @cindex defining words without name
2709: Sometimes you want to define a word without a name. You can do this with
2710:
2711: doc-noname
2712:
2713: @cindex execution token of last defined word
2714: To make any use of the newly defined word, you need its execution
2715: token. You can get it with
2716:
2717: doc-lastxt
2718:
2719: E.g., you can initialize a deferred word with an anonymous colon
2720: definition:
2721: @example
2722: Defer deferred
2723: noname : ( ... -- ... )
2724: ... ;
2725: lastxt IS deferred
2726: @end example
2727:
2728: @code{lastxt} also works when the last word was not defined as
2729: @code{noname}.
2730:
2731: The standard has also recognized the need for anonymous words and
2732: provides
2733:
2734: doc-:noname
2735:
2736: This leaves the execution token for the word on the stack after the
2737: closing @code{;}. You can rewrite the last example with @code{:noname}:
2738: @example
2739: Defer deferred
2740: :noname ( ... -- ... )
2741: ... ;
2742: IS deferred
2743: @end example
2744:
2745: @node Interpretation and Compilation Semantics, , Supplying names, Defining Words
2746: @subsection Interpretation and Compilation Semantics
2747: @cindex semantics, interpretation and compilation
2748:
2749: @cindex interpretation semantics
2750: The @dfn{interpretation semantics} of a word are what the text
2751: interpreter does when it encounters the word in interpret state. It also
2752: appears in some other contexts, e.g., the execution token returned by
2753: @code{' @var{word}} identifies the interpretation semantics of
2754: @var{word} (in other words, @code{' @var{word} execute} is equivalent to
2755: interpret-state text interpretation of @code{@var{word}}).
2756:
2757: @cindex compilation semantics
2758: The @dfn{compilation semantics} of a word are what the text interpreter
2759: does when it encounters the word in compile state. It also appears in
2760: other contexts, e.g, @code{POSTPONE @var{word}} compiles@footnote{In
2761: standard terminology, ``appends to the current definition''.} the
2762: compilation semantics of @var{word}.
2763:
2764: @cindex execution semantics
2765: The standard also talks about @dfn{execution semantics}. They are used
2766: only for defining the interpretation and compilation semantics of many
2767: words. By default, the interpretation semantics of a word are to
2768: @code{execute} its execution semantics, and the compilation semantics of
2769: a word are to @code{compile,} its execution semantics.@footnote{In
2770: standard terminology: The default interpretation semantics are its
2771: execution semantics; the default compilation semantics are to append its
2772: execution semantics to the execution semantics of the current
2773: definition.}
2774:
2775: @cindex immediate words
2776: You can change the compilation semantics into @code{execute}ing the
2777: execution semantics with
2778:
2779: doc-immediate
2780:
2781: @cindex compile-only words
2782: You can remove the interpretation semantics of a word with
2783:
2784: doc-compile-only
2785: doc-restrict
2786:
2787: Note that ticking (@code{'}) compile-only words gives an error
2788: (``Interpreting a compile-only word'').
2789:
2790: Gforth also allows you to define words with arbitrary combinations of
2791: interpretation and compilation semantics.
2792:
2793: doc-interpret/compile:
2794:
2795: This feature was introduced for implementing @code{TO} and @code{S"}. I
2796: recommend that you do not define such words, as cute as they may be:
2797: they make it hard to get at both parts of the word in some contexts.
2798: E.g., assume you want to get an execution token for the compilation
2799: part. Instead, define two words, one that embodies the interpretation
2800: part, and one that embodies the compilation part. Once you have done
2801: that, you can define a combined word with @code{interpret/compile:} for
2802: the convenience of your users.
2803:
2804: You also might try to provide an optimizing implementation of the
2805: default compilation semantics with this feature, like this:
2806:
2807: @example
2808: :noname
2809: foo bar ;
2810: :noname
2811: POSTPONE foo POSTPONE bar ;
2812: interpret/compile: foobar
2813: @end example
2814:
2815: as an optimizing version of
2816:
2817: @example
2818: : foobar
2819: foo bar ;
2820: @end example
2821:
2822: Unfortunately, this does not work correctly with @code{[compile]},
2823: because @code{[compile]} assumes that the compilation semantics of all
2824: @code{interpret/compile:} words are non-default. I.e., @code{[compile]
2825: foobar} would compile the compilation semantics for the optimizing
2826: @code{foobar}, whereas it would compile the interpretation semantics for
2827: the non-optimizing @code{foobar}.
2828:
2829: @cindex state-smart words are a bad idea
2830: Some people try to use state-smart words to emulate the feature provided
2831: by @code{interpret/compile:} (words are state-smart if they check
2832: @code{STATE} during execution). E.g., they would try to code
2833: @code{foobar} like this:
2834:
2835: @example
2836: : foobar
2837: STATE @@
2838: IF ( compilation state )
2839: POSTPONE foo POSTPONE bar
2840: ELSE
2841: foo bar
2842: ENDIF ; immediate
2843: @end example
2844:
2845: While this works if @code{foobar} is processed only by the text
2846: interpreter, it does not work in other contexts (like @code{'} or
2847: @code{POSTPONE}). E.g., @code{' foobar} will produce an execution token
2848: for a state-smart word, not for the interpretation semantics of the
2849: original @code{foobar}; when you execute this execution token (directly
2850: with @code{EXECUTE} or indirectly through @code{COMPILE,}) in compile
2851: state, the result will not be what you expected (i.e., it will not
2852: perform @code{foo bar}). State-smart words are a bad idea. Simply don't
2853: write them!
2854:
2855: @cindex defining words with arbitrary semantics combinations
2856: It is also possible to write defining words that define words with
2857: arbitrary combinations of interpretation and compilation semantics. In
2858: general, this looks like:
2859:
2860: @example
2861: : def-word
2862: create-interpret/compile
2863: @var{code1}
2864: interpretation>
2865: @var{code2}
2866: <interpretation
2867: compilation>
2868: @var{code3}
2869: <compilation ;
2870: @end example
2871:
2872: For a @var{word} defined with @code{def-word}, the interpretation
2873: semantics are to push the address of the body of @var{word} and perform
2874: @var{code2}, and the compilation semantics are to push the address of
2875: the body of @var{word} and perform @var{code3}. E.g., @code{constant}
2876: can also be defined like this (except that the defined constants don't
2877: behave correctly when @code{[compile]}d):
2878:
2879: @example
2880: : constant ( n "name" -- )
2881: create-interpret/compile
2882: ,
2883: interpretation> ( -- n )
2884: @@
2885: <interpretation
2886: compilation> ( compilation. -- ; run-time. -- n )
2887: @@ postpone literal
2888: <compilation ;
2889: @end example
2890:
2891: doc-create-interpret/compile
2892: doc-interpretation>
2893: doc-<interpretation
2894: doc-compilation>
2895: doc-<compilation
2896:
2897: Note that words defined with @code{interpret/compile:} and
2898: @code{create-interpret/compile} have an extended header structure that
2899: differs from other words; however, unless you try to access them with
2900: plain address arithmetic, you should not notice this. Words for
2901: accessing the header structure usually know how to deal with this; e.g.,
2902: @code{' word >body} also gives you the body of a word created with
2903: @code{create-interpret/compile}.
2904:
2905: @c ----------------------------------------------------------
2906: @node Structures, Object-oriented Forth, Defining Words, Words
2907: @section Structures
2908: @cindex structures
2909: @cindex records
2910:
2911: This section presents the structure package that comes with Gforth. A
2912: version of the package implemented in plain ANS Forth is available in
2913: @file{compat/struct.fs}. This package was inspired by a posting on
2914: comp.lang.forth in 1989 (unfortunately I don't remember, by whom;
2915: possibly John Hayes). A version of this section has been published in
2916: ???. Marcel Hendrix provided helpful comments.
2917:
2918: @menu
2919: * Why explicit structure support?::
2920: * Structure Usage::
2921: * Structure Naming Convention::
2922: * Structure Implementation::
2923: * Structure Glossary::
2924: @end menu
2925:
2926: @node Why explicit structure support?, Structure Usage, Structures, Structures
2927: @subsection Why explicit structure support?
2928:
2929: @cindex address arithmetic for structures
2930: @cindex structures using address arithmetic
2931: If we want to use a structure containing several fields, we could simply
2932: reserve memory for it, and access the fields using address arithmetic
2933: (@pxref{Address arithmetic}). As an example, consider a structure with
2934: the following fields
2935:
2936: @table @code
2937: @item a
2938: is a float
2939: @item b
2940: is a cell
2941: @item c
2942: is a float
2943: @end table
2944:
2945: Given the (float-aligned) base address of the structure we get the
2946: address of the field
2947:
2948: @table @code
2949: @item a
2950: without doing anything further.
2951: @item b
2952: with @code{float+}
2953: @item c
2954: with @code{float+ cell+ faligned}
2955: @end table
2956:
2957: It is easy to see that this can become quite tiring.
2958:
2959: Moreover, it is not very readable, because seeing a
2960: @code{cell+} tells us neither which kind of structure is
2961: accessed nor what field is accessed; we have to somehow infer the kind
2962: of structure, and then look up in the documentation, which field of
2963: that structure corresponds to that offset.
2964:
2965: Finally, this kind of address arithmetic also causes maintenance
2966: troubles: If you add or delete a field somewhere in the middle of the
2967: structure, you have to find and change all computations for the fields
2968: afterwards.
2969:
2970: So, instead of using @code{cell+} and friends directly, how
2971: about storing the offsets in constants:
2972:
2973: @example
2974: 0 constant a-offset
2975: 0 float+ constant b-offset
2976: 0 float+ cell+ faligned c-offset
2977: @end example
2978:
2979: Now we can get the address of field @code{x} with @code{x-offset
2980: +}. This is much better in all respects. Of course, you still
2981: have to change all later offset definitions if you add a field. You can
2982: fix this by declaring the offsets in the following way:
2983:
2984: @example
2985: 0 constant a-offset
2986: a-offset float+ constant b-offset
2987: b-offset cell+ faligned constant c-offset
2988: @end example
2989:
2990: Since we always use the offsets with @code{+}, using a defining
2991: word @code{cfield} that includes the @code{+} in the
2992: action of the defined word offers itself:
2993:
2994: @example
2995: : cfield ( n "name" -- )
2996: create ,
2997: does> ( name execution: addr1 -- addr2 )
2998: @@ + ;
2999:
3000: 0 cfield a
3001: 0 a float+ cfield b
3002: 0 b cell+ faligned cfield c
3003: @end example
3004:
3005: Instead of @code{x-offset +}, we now simply write @code{x}.
3006:
3007: The structure field words now can be used quite nicely. However,
3008: their definition is still a bit cumbersome: We have to repeat the
3009: name, the information about size and alignment is distributed before
3010: and after the field definitions etc. The structure package presented
3011: here addresses these problems.
3012:
3013: @node Structure Usage, Structure Naming Convention, Why explicit structure support?, Structures
3014: @subsection Structure Usage
3015: @cindex structure usage
3016:
3017: @cindex @code{field} usage
3018: @cindex @code{struct} usage
3019: @cindex @code{end-struct} usage
3020: You can define a structure for a (data-less) linked list with
3021: @example
3022: struct
3023: cell% field list-next
3024: end-struct list%
3025: @end example
3026:
3027: With the address of the list node on the stack, you can compute the
3028: address of the field that contains the address of the next node with
3029: @code{list-next}. E.g., you can determine the length of a list
3030: with:
3031:
3032: @example
3033: : list-length ( list -- n )
3034: \ "list" is a pointer to the first element of a linked list
3035: \ "n" is the length of the list
3036: 0 begin ( list1 n1 )
3037: over
3038: while ( list1 n1 )
3039: 1+ swap list-next @@ swap
3040: repeat
3041: nip ;
3042: @end example
3043:
3044: You can reserve memory for a list node in the dictionary with
3045: @code{list% %allot}, which leaves the address of the list node on the
3046: stack. For the equivalent allocation on the heap you can use @code{list%
3047: %alloc} (or, for an @code{allocate}-like stack effect (i.e., with ior),
3048: use @code{list% %allocate}). You can also get the the size of a list
3049: node with @code{list% %size} and it's alignment with @code{list%
3050: %alignment}.
3051:
3052: Note that in ANS Forth the body of a @code{create}d word is
3053: @code{aligned} but not necessarily @code{faligned};
3054: therefore, if you do a
3055: @example
3056: create @emph{name} foo% %allot
3057: @end example
3058:
3059: then the memory alloted for @code{foo%} is
3060: guaranteed to start at the body of @code{@emph{name}} only if
3061: @code{foo%} contains only character, cell and double fields.
3062:
3063: @cindex strcutures containing structures
3064: You can also include a structure @code{foo%} as field of
3065: another structure, with:
3066: @example
3067: struct
3068: ...
3069: foo% field ...
3070: ...
3071: end-struct ...
3072: @end example
3073:
3074: @cindex structure extension
3075: @cindex extended records
3076: Instead of starting with an empty structure, you can also extend an
3077: existing structure. E.g., a plain linked list without data, as defined
3078: above, is hardly useful; You can extend it to a linked list of integers,
3079: like this:@footnote{This feature is also known as @emph{extended
3080: records}. It is the main innovation in the Oberon language; in other
3081: words, adding this feature to Modula-2 led Wirth to create a new
3082: language, write a new compiler etc. Adding this feature to Forth just
3083: requires a few lines of code.}
3084:
3085: @example
3086: list%
3087: cell% field intlist-int
3088: end-struct intlist%
3089: @end example
3090:
3091: @code{intlist%} is a structure with two fields:
3092: @code{list-next} and @code{intlist-int}.
3093:
3094: @cindex structures containing arrays
3095: You can specify an array type containing @emph{n} elements of
3096: type @code{foo%} like this:
3097:
3098: @example
3099: foo% @emph{n} *
3100: @end example
3101:
3102: You can use this array type in any place where you can use a normal
3103: type, e.g., when defining a @code{field}, or with
3104: @code{%allot}.
3105:
3106: @cindex first field optimization
3107: The first field is at the base address of a structure and the word
3108: for this field (e.g., @code{list-next}) actually does not change
3109: the address on the stack. You may be tempted to leave it away in the
3110: interest of run-time and space efficiency. This is not necessary,
3111: because the structure package optimizes this case and compiling such
3112: words does not generate any code. So, in the interest of readability
3113: and maintainability you should include the word for the field when
3114: accessing the field.
3115:
3116: @node Structure Naming Convention, Structure Implementation, Structure Usage, Structures
3117: @subsection Structure Naming Convention
3118: @cindex structure naming conventions
3119:
3120: The field names that come to (my) mind are often quite generic, and,
3121: if used, would cause frequent name clashes. E.g., many structures
3122: probably contain a @code{counter} field. The structure names
3123: that come to (my) mind are often also the logical choice for the names
3124: of words that create such a structure.
3125:
3126: Therefore, I have adopted the following naming conventions:
3127:
3128: @itemize @bullet
3129: @cindex field naming convention
3130: @item
3131: The names of fields are of the form
3132: @code{@emph{struct}-@emph{field}}, where
3133: @code{@emph{struct}} is the basic name of the structure, and
3134: @code{@emph{field}} is the basic name of the field. You can
3135: think about field words as converting converts the (address of the)
3136: structure into the (address of the) field.
3137:
3138: @cindex structure naming convention
3139: @item
3140: The names of structures are of the form
3141: @code{@emph{struct}%}, where
3142: @code{@emph{struct}} is the basic name of the structure.
3143: @end itemize
3144:
3145: This naming convention does not work that well for fields of extended
3146: structures; e.g., the integer list structure has a field
3147: @code{intlist-int}, but has @code{list-next}, not
3148: @code{intlist-next}.
3149:
3150: @node Structure Implementation, Structure Glossary, Structure Naming Convention, Structures
3151: @subsection Structure Implementation
3152: @cindex structure implementation
3153: @cindex implementation of structures
3154:
3155: The central idea in the implementation is to pass the data about the
3156: structure being built on the stack, not in some global
3157: variable. Everything else falls into place naturally once this design
3158: decision is made.
3159:
3160: The type description on the stack is of the form @emph{align
3161: size}. Keeping the size on the top-of-stack makes dealing with arrays
3162: very simple.
3163:
3164: @code{field} is a defining word that uses @code{create}
3165: and @code{does>}. The body of the field contains the offset
3166: of the field, and the normal @code{does>} action is
3167:
3168: @example
3169: @ +
3170: @end example
3171:
3172: i.e., add the offset to the address, giving the stack effect
3173: @code{addr1 -- addr2} for a field.
3174:
3175: @cindex first field optimization, implementation
3176: This simple structure is slightly complicated by the optimization
3177: for fields with offset 0, which requires a different
3178: @code{does>}-part (because we cannot rely on there being
3179: something on the stack if such a field is invoked during
3180: compilation). Therefore, we put the different @code{does>}-parts
3181: in separate words, and decide which one to invoke based on the
3182: offset. For a zero offset, the field is basically a noop; it is
3183: immediate, and therefore no code is generated when it is compiled.
3184:
3185: @node Structure Glossary, , Structure Implementation, Structures
3186: @subsection Structure Glossary
3187: @cindex structure glossary
3188:
3189: doc-%align
3190: doc-%alignment
3191: doc-%alloc
3192: doc-%allocate
3193: doc-%allot
3194: doc-cell%
3195: doc-char%
3196: doc-dfloat%
3197: doc-double%
3198: doc-end-struct
3199: doc-field
3200: doc-float%
3201: doc-nalign
3202: doc-sfloat%
3203: doc-%size
3204: doc-struct
3205:
3206: @c -------------------------------------------------------------
3207: @node Object-oriented Forth, Tokens for Words, Structures, Words
3208: @section Object-oriented Forth
3209:
3210: Gforth comes with three packets for object-oriented programming,
3211: @file{objects.fs}, @file{oof.fs}, and @file{mini-oof.fs}; none of them
3212: is preloaded, so you have to @code{include} them before use. The most
3213: important differences between these packets (and others) are discussed
3214: in @ref{Comparison with other object models}. All packets are written
3215: in ANS Forth and can be used with any other ANS Forth.
3216:
3217: @menu
3218: * Objects::
3219: * OOF::
3220: * Mini-OOF::
3221: @end menu
3222:
3223: @node Objects, OOF, Object-oriented Forth, Object-oriented Forth
3224: @subsection Objects
3225: @cindex objects
3226: @cindex object-oriented programming
3227:
3228: @cindex @file{objects.fs}
3229: @cindex @file{oof.fs}
3230:
3231: This section describes the @file{objects.fs} packet. This material also has been published in @cite{Yet Another Forth Objects Package} by Anton Ertl and appeared in Forth Dimensions 19(2), pages 37--43 (@url{http://www.complang.tuwien.ac.at/forth/objects/objects.html}).
3232: @c McKewan's and Zsoter's packages
3233:
3234: This section assumes (in some places) that you have read @ref{Structures}.
3235:
3236: @menu
3237: * Properties of the Objects model::
3238: * Why object-oriented programming?::
3239: * Object-Oriented Terminology::
3240: * Basic Objects Usage::
3241: * The class Object::
3242: * Creating objects::
3243: * Object-Oriented Programming Style::
3244: * Class Binding::
3245: * Method conveniences::
3246: * Classes and Scoping::
3247: * Object Interfaces::
3248: * Objects Implementation::
3249: * Comparison with other object models::
3250: * Objects Glossary::
3251: @end menu
3252:
3253: Marcel Hendrix provided helpful comments on this section. Andras Zsoter
3254: and Bernd Paysan helped me with the related works section.
3255:
3256: @node Properties of the Objects model, Why object-oriented programming?, Objects, Objects
3257: @subsubsection Properties of the @file{objects.fs} model
3258: @cindex @file{objects.fs} properties
3259:
3260: @itemize @bullet
3261: @item
3262: It is straightforward to pass objects on the stack. Passing
3263: selectors on the stack is a little less convenient, but possible.
3264:
3265: @item
3266: Objects are just data structures in memory, and are referenced by
3267: their address. You can create words for objects with normal defining
3268: words like @code{constant}. Likewise, there is no difference
3269: between instance variables that contain objects and those
3270: that contain other data.
3271:
3272: @item
3273: Late binding is efficient and easy to use.
3274:
3275: @item
3276: It avoids parsing, and thus avoids problems with state-smartness
3277: and reduced extensibility; for convenience there are a few parsing
3278: words, but they have non-parsing counterparts. There are also a few
3279: defining words that parse. This is hard to avoid, because all standard
3280: defining words parse (except @code{:noname}); however, such
3281: words are not as bad as many other parsing words, because they are not
3282: state-smart.
3283:
3284: @item
3285: It does not try to incorporate everything. It does a few things
3286: and does them well (IMO). In particular, I did not intend to support
3287: information hiding with this model (although it has features that may
3288: help); you can use a separate package for achieving this.
3289:
3290: @item
3291: It is layered; you don't have to learn and use all features to use this
3292: model. Only a few features are necessary (@xref{Basic Objects Usage},
3293: @xref{The class Object}, @xref{Creating objects}.), the others
3294: are optional and independent of each other.
3295:
3296: @item
3297: An implementation in ANS Forth is available.
3298:
3299: @end itemize
3300:
3301: I have used the technique, on which this model is based, for
3302: implementing the parser generator Gray; we have also used this technique
3303: in Gforth for implementing the various flavours of wordlists (hashed or
3304: not, case-sensitive or not, special-purpose wordlists for locals etc.).
3305:
3306: @node Why object-oriented programming?, Object-Oriented Terminology, Properties of the Objects model, Objects
3307: @subsubsection Why object-oriented programming?
3308: @cindex object-oriented programming motivation
3309: @cindex motivation for object-oriented programming
3310:
3311: Often we have to deal with several data structures (@emph{objects}),
3312: that have to be treated similarly in some respects, but differ in
3313: others. Graphical objects are the textbook example: circles,
3314: triangles, dinosaurs, icons, and others, and we may want to add more
3315: during program development. We want to apply some operations to any
3316: graphical object, e.g., @code{draw} for displaying it on the
3317: screen. However, @code{draw} has to do something different for
3318: every kind of object.
3319:
3320: We could implement @code{draw} as a big @code{CASE}
3321: control structure that executes the appropriate code depending on the
3322: kind of object to be drawn. This would be not be very elegant, and,
3323: moreover, we would have to change @code{draw} every time we add
3324: a new kind of graphical object (say, a spaceship).
3325:
3326: What we would rather do is: When defining spaceships, we would tell
3327: the system: "Here's how you @code{draw} a spaceship; you figure
3328: out the rest."
3329:
3330: This is the problem that all systems solve that (rightfully) call
3331: themselves object-oriented, and the object-oriented package I present
3332: here also solves this problem (and not much else).
3333:
3334: @node Object-Oriented Terminology, Basic Objects Usage, Why object-oriented programming?, Objects
3335: @subsubsection Object-Oriented Terminology
3336: @cindex object-oriented terminology
3337: @cindex terminology for object-oriented programming
3338:
3339: This section is mainly for reference, so you don't have to understand
3340: all of it right away. The terminology is mainly Smalltalk-inspired. In
3341: short:
3342:
3343: @table @emph
3344: @cindex class
3345: @item class
3346: a data structure definition with some extras.
3347:
3348: @cindex object
3349: @item object
3350: an instance of the data structure described by the class definition.
3351:
3352: @cindex instance variables
3353: @item instance variables
3354: fields of the data structure.
3355:
3356: @cindex selector
3357: @cindex method selector
3358: @cindex virtual function
3359: @item selector
3360: (or @emph{method selector}) a word (e.g.,
3361: @code{draw}) for performing an operation on a variety of data
3362: structures (classes). A selector describes @emph{what} operation to
3363: perform. In C++ terminology: a (pure) virtual function.
3364:
3365: @cindex method
3366: @item method
3367: the concrete definition that performs the operation
3368: described by the selector for a specific class. A method specifies
3369: @emph{how} the operation is performed for a specific class.
3370:
3371: @cindex selector invocation
3372: @cindex message send
3373: @cindex invoking a selector
3374: @item selector invocation
3375: a call of a selector. One argument of the call (the TOS (top-of-stack))
3376: is used for determining which method is used. In Smalltalk terminology:
3377: a message (consisting of the selector and the other arguments) is sent
3378: to the object.
3379:
3380: @cindex receiving object
3381: @item receiving object
3382: the object used for determining the method executed by a selector
3383: invocation. In our model it is the object that is on the TOS when the
3384: selector is invoked. (@emph{Receiving} comes from Smalltalks
3385: @emph{message} terminology.)
3386:
3387: @cindex child class
3388: @cindex parent class
3389: @cindex inheritance
3390: @item child class
3391: a class that has (@emph{inherits}) all properties (instance variables,
3392: selectors, methods) from a @emph{parent class}. In Smalltalk
3393: terminology: The subclass inherits from the superclass. In C++
3394: terminology: The derived class inherits from the base class.
3395:
3396: @end table
3397:
3398: @c If you wonder about the message sending terminology, it comes from
3399: @c a time when each object had it's own task and objects communicated via
3400: @c message passing; eventually the Smalltalk developers realized that
3401: @c they can do most things through simple (indirect) calls. They kept the
3402: @c terminology.
3403:
3404: @node Basic Objects Usage, The class Object, Object-Oriented Terminology, Objects
3405: @subsubsection Basic Objects Usage
3406: @cindex basic objects usage
3407: @cindex objects, basic usage
3408:
3409: You can define a class for graphical objects like this:
3410:
3411: @cindex @code{class} usage
3412: @cindex @code{end-class} usage
3413: @cindex @code{selector} usage
3414: @example
3415: object class \ "object" is the parent class
3416: selector draw ( x y graphical -- )
3417: end-class graphical
3418: @end example
3419:
3420: This code defines a class @code{graphical} with an
3421: operation @code{draw}. We can perform the operation
3422: @code{draw} on any @code{graphical} object, e.g.:
3423:
3424: @example
3425: 100 100 t-rex draw
3426: @end example
3427:
3428: where @code{t-rex} is a word (say, a constant) that produces a
3429: graphical object.
3430:
3431: @cindex abstract class
3432: How do we create a graphical object? With the present definitions,
3433: we cannot create a useful graphical object. The class
3434: @code{graphical} describes graphical objects in general, but not
3435: any concrete graphical object type (C++ users would call it an
3436: @emph{abstract class}); e.g., there is no method for the selector
3437: @code{draw} in the class @code{graphical}.
3438:
3439: For concrete graphical objects, we define child classes of the
3440: class @code{graphical}, e.g.:
3441:
3442: @cindex @code{overrides} usage
3443: @cindex @code{field} usage in class definition
3444: @example
3445: graphical class \ "graphical" is the parent class
3446: cell% field circle-radius
3447:
3448: :noname ( x y circle -- )
3449: circle-radius @@ draw-circle ;
3450: overrides draw
3451:
3452: :noname ( n-radius circle -- )
3453: circle-radius ! ;
3454: overrides construct
3455:
3456: end-class circle
3457: @end example
3458:
3459: Here we define a class @code{circle} as a child of @code{graphical},
3460: with a field @code{circle-radius} (which behaves just like a field in
3461: @pxref{Structures}); it defines new methods for the selectors
3462: @code{draw} and @code{construct} (@code{construct} is defined in
3463: @code{object}, the parent class of @code{graphical}).
3464:
3465: Now we can create a circle on the heap (i.e.,
3466: @code{allocate}d memory) with
3467:
3468: @cindex @code{heap-new} usage
3469: @example
3470: 50 circle heap-new constant my-circle
3471: @end example
3472:
3473: @code{heap-new} invokes @code{construct}, thus
3474: initializing the field @code{circle-radius} with 50. We can draw
3475: this new circle at (100,100) with
3476:
3477: @example
3478: 100 100 my-circle draw
3479: @end example
3480:
3481: @cindex selector invocation, restrictions
3482: @cindex class definition, restrictions
3483: Note: You can invoke a selector only if the object on the TOS
3484: (the receiving object) belongs to the class where the selector was
3485: defined or one of its descendents; e.g., you can invoke
3486: @code{draw} only for objects belonging to @code{graphical}
3487: or its descendents (e.g., @code{circle}). Immediately before
3488: @code{end-class}, the search order has to be the same as
3489: immediately after @code{class}.
3490:
3491: @node The class Object, Creating objects, Basic Objects Usage, Objects
3492: @subsubsection The class @code{object}
3493: @cindex @code{object} class
3494:
3495: When you define a class, you have to specify a parent class. So how do
3496: you start defining classes? There is one class available from the start:
3497: @code{object}. You can use it as ancestor for all classes. It is the
3498: only class that has no parent. It has two selectors: @code{construct}
3499: and @code{print}.
3500:
3501: @node Creating objects, Object-Oriented Programming Style, The class Object, Objects
3502: @subsubsection Creating objects
3503: @cindex creating objects
3504: @cindex object creation
3505: @cindex object allocation options
3506:
3507: @cindex @code{heap-new} discussion
3508: @cindex @code{dict-new} discussion
3509: @cindex @code{construct} discussion
3510: You can create and initialize an object of a class on the heap with
3511: @code{heap-new} ( ... class -- object ) and in the dictionary
3512: (allocation with @code{allot}) with @code{dict-new} (
3513: ... class -- object ). Both words invoke @code{construct}, which
3514: consumes the stack items indicated by "..." above.
3515:
3516: @cindex @code{init-object} discussion
3517: @cindex @code{class-inst-size} discussion
3518: If you want to allocate memory for an object yourself, you can get its
3519: alignment and size with @code{class-inst-size 2@@} ( class --
3520: align size ). Once you have memory for an object, you can initialize
3521: it with @code{init-object} ( ... class object -- );
3522: @code{construct} does only a part of the necessary work.
3523:
3524: @node Object-Oriented Programming Style, Class Binding, Creating objects, Objects
3525: @subsubsection Object-Oriented Programming Style
3526: @cindex object-oriented programming style
3527:
3528: This section is not exhaustive.
3529:
3530: @cindex stack effects of selectors
3531: @cindex selectors and stack effects
3532: In general, it is a good idea to ensure that all methods for the
3533: same selector have the same stack effect: when you invoke a selector,
3534: you often have no idea which method will be invoked, so, unless all
3535: methods have the same stack effect, you will not know the stack effect
3536: of the selector invocation.
3537:
3538: One exception to this rule is methods for the selector
3539: @code{construct}. We know which method is invoked, because we
3540: specify the class to be constructed at the same place. Actually, I
3541: defined @code{construct} as a selector only to give the users a
3542: convenient way to specify initialization. The way it is used, a
3543: mechanism different from selector invocation would be more natural
3544: (but probably would take more code and more space to explain).
3545:
3546: @node Class Binding, Method conveniences, Object-Oriented Programming Style, Objects
3547: @subsubsection Class Binding
3548: @cindex class binding
3549: @cindex early binding
3550:
3551: @cindex late binding
3552: Normal selector invocations determine the method at run-time depending
3553: on the class of the receiving object (late binding).
3554:
3555: Sometimes we want to invoke a different method. E.g., assume that
3556: you want to use the simple method for @code{print}ing
3557: @code{object}s instead of the possibly long-winded
3558: @code{print} method of the receiver class. You can achieve this
3559: by replacing the invocation of @code{print} with
3560:
3561: @cindex @code{[bind]} usage
3562: @example
3563: [bind] object print
3564: @end example
3565:
3566: in compiled code or
3567:
3568: @cindex @code{bind} usage
3569: @example
3570: bind object print
3571: @end example
3572:
3573: @cindex class binding, alternative to
3574: in interpreted code. Alternatively, you can define the method with a
3575: name (e.g., @code{print-object}), and then invoke it through the
3576: name. Class binding is just a (often more convenient) way to achieve
3577: the same effect; it avoids name clutter and allows you to invoke
3578: methods directly without naming them first.
3579:
3580: @cindex superclass binding
3581: @cindex parent class binding
3582: A frequent use of class binding is this: When we define a method
3583: for a selector, we often want the method to do what the selector does
3584: in the parent class, and a little more. There is a special word for
3585: this purpose: @code{[parent]}; @code{[parent]
3586: @emph{selector}} is equivalent to @code{[bind] @emph{parent
3587: selector}}, where @code{@emph{parent}} is the parent
3588: class of the current class. E.g., a method definition might look like:
3589:
3590: @cindex @code{[parent]} usage
3591: @example
3592: :noname
3593: dup [parent] foo \ do parent's foo on the receiving object
3594: ... \ do some more
3595: ; overrides foo
3596: @end example
3597:
3598: @cindex class binding as optimization
3599: In @cite{Object-oriented programming in ANS Forth} (Forth Dimensions,
3600: March 1997), Andrew McKewan presents class binding as an optimization
3601: technique. I recommend not using it for this purpose unless you are in
3602: an emergency. Late binding is pretty fast with this model anyway, so the
3603: benefit of using class binding is small; the cost of using class binding
3604: where it is not appropriate is reduced maintainability.
3605:
3606: While we are at programming style questions: You should bind
3607: selectors only to ancestor classes of the receiving object. E.g., say,
3608: you know that the receiving object is of class @code{foo} or its
3609: descendents; then you should bind only to @code{foo} and its
3610: ancestors.
3611:
3612: @node Method conveniences, Classes and Scoping, Class Binding, Objects
3613: @subsubsection Method conveniences
3614: @cindex method conveniences
3615:
3616: In a method you usually access the receiving object pretty often. If
3617: you define the method as a plain colon definition (e.g., with
3618: @code{:noname}), you may have to do a lot of stack
3619: gymnastics. To avoid this, you can define the method with @code{m:
3620: ... ;m}. E.g., you could define the method for
3621: @code{draw}ing a @code{circle} with
3622:
3623: @cindex @code{this} usage
3624: @cindex @code{m:} usage
3625: @cindex @code{;m} usage
3626: @example
3627: m: ( x y circle -- )
3628: ( x y ) this circle-radius @@ draw-circle ;m
3629: @end example
3630:
3631: @cindex @code{exit} in @code{m: ... ;m}
3632: @cindex @code{exitm} discussion
3633: @cindex @code{catch} in @code{m: ... ;m}
3634: When this method is executed, the receiver object is removed from the
3635: stack; you can access it with @code{this} (admittedly, in this
3636: example the use of @code{m: ... ;m} offers no advantage). Note
3637: that I specify the stack effect for the whole method (i.e. including
3638: the receiver object), not just for the code between @code{m:}
3639: and @code{;m}. You cannot use @code{exit} in
3640: @code{m:...;m}; instead, use
3641: @code{exitm}.@footnote{Moreover, for any word that calls
3642: @code{catch} and was defined before loading
3643: @code{objects.fs}, you have to redefine it like I redefined
3644: @code{catch}: @code{: catch this >r catch r> to-this ;}}
3645:
3646: @cindex @code{inst-var} usage
3647: You will frequently use sequences of the form @code{this
3648: @emph{field}} (in the example above: @code{this
3649: circle-radius}). If you use the field only in this way, you can
3650: define it with @code{inst-var} and eliminate the
3651: @code{this} before the field name. E.g., the @code{circle}
3652: class above could also be defined with:
3653:
3654: @example
3655: graphical class
3656: cell% inst-var radius
3657:
3658: m: ( x y circle -- )
3659: radius @@ draw-circle ;m
3660: overrides draw
3661:
3662: m: ( n-radius circle -- )
3663: radius ! ;m
3664: overrides construct
3665:
3666: end-class circle
3667: @end example
3668:
3669: @code{radius} can only be used in @code{circle} and its
3670: descendent classes and inside @code{m:...;m}.
3671:
3672: @cindex @code{inst-value} usage
3673: You can also define fields with @code{inst-value}, which is
3674: to @code{inst-var} what @code{value} is to
3675: @code{variable}. You can change the value of such a field with
3676: @code{[to-inst]}. E.g., we could also define the class
3677: @code{circle} like this:
3678:
3679: @example
3680: graphical class
3681: inst-value radius
3682:
3683: m: ( x y circle -- )
3684: radius draw-circle ;m
3685: overrides draw
3686:
3687: m: ( n-radius circle -- )
3688: [to-inst] radius ;m
3689: overrides construct
3690:
3691: end-class circle
3692: @end example
3693:
3694:
3695: @node Classes and Scoping, Object Interfaces, Method conveniences, Objects
3696: @subsubsection Classes and Scoping
3697: @cindex classes and scoping
3698: @cindex scoping and classes
3699:
3700: Inheritance is frequent, unlike structure extension. This exacerbates
3701: the problem with the field name convention (@pxref{Structure Naming
3702: Convention}): One always has to remember in which class the field was
3703: originally defined; changing a part of the class structure would require
3704: changes for renaming in otherwise unaffected code.
3705:
3706: @cindex @code{inst-var} visibility
3707: @cindex @code{inst-value} visibility
3708: To solve this problem, I added a scoping mechanism (which was not in my
3709: original charter): A field defined with @code{inst-var} (or
3710: @code{inst-value}) is visible only in the class where it is defined and in
3711: the descendent classes of this class. Using such fields only makes
3712: sense in @code{m:}-defined methods in these classes anyway.
3713:
3714: This scoping mechanism allows us to use the unadorned field name,
3715: because name clashes with unrelated words become much less likely.
3716:
3717: @cindex @code{protected} discussion
3718: @cindex @code{private} discussion
3719: Once we have this mechanism, we can also use it for controlling the
3720: visibility of other words: All words defined after
3721: @code{protected} are visible only in the current class and its
3722: descendents. @code{public} restores the compilation
3723: (i.e. @code{current}) wordlist that was in effect before. If you
3724: have several @code{protected}s without an intervening
3725: @code{public} or @code{set-current}, @code{public}
3726: will restore the compilation wordlist in effect before the first of
3727: these @code{protected}s.
3728:
3729: @node Object Interfaces, Objects Implementation, Classes and Scoping, Objects
3730: @subsubsection Object Interfaces
3731: @cindex object interfaces
3732: @cindex interfaces for objects
3733:
3734: In this model you can only call selectors defined in the class of the
3735: receiving objects or in one of its ancestors. If you call a selector
3736: with a receiving object that is not in one of these classes, the
3737: result is undefined; if you are lucky, the program crashes
3738: immediately.
3739:
3740: @cindex selectors common to hardly-related classes
3741: Now consider the case when you want to have a selector (or several)
3742: available in two classes: You would have to add the selector to a
3743: common ancestor class, in the worst case to @code{object}. You
3744: may not want to do this, e.g., because someone else is responsible for
3745: this ancestor class.
3746:
3747: The solution for this problem is interfaces. An interface is a
3748: collection of selectors. If a class implements an interface, the
3749: selectors become available to the class and its descendents. A class
3750: can implement an unlimited number of interfaces. For the problem
3751: discussed above, we would define an interface for the selector(s), and
3752: both classes would implement the interface.
3753:
3754: As an example, consider an interface @code{storage} for
3755: writing objects to disk and getting them back, and a class
3756: @code{foo} foo that implements it. The code for this would look
3757: like this:
3758:
3759: @cindex @code{interface} usage
3760: @cindex @code{end-interface} usage
3761: @cindex @code{implementation} usage
3762: @example
3763: interface
3764: selector write ( file object -- )
3765: selector read1 ( file object -- )
3766: end-interface storage
3767:
3768: bar class
3769: storage implementation
3770:
3771: ... overrides write
3772: ... overrides read
3773: ...
3774: end-class foo
3775: @end example
3776:
3777: (I would add a word @code{read} ( file -- object ) that uses
3778: @code{read1} internally, but that's beyond the point illustrated
3779: here.)
3780:
3781: Note that you cannot use @code{protected} in an interface; and
3782: of course you cannot define fields.
3783:
3784: In the Neon model, all selectors are available for all classes;
3785: therefore it does not need interfaces. The price you pay in this model
3786: is slower late binding, and therefore, added complexity to avoid late
3787: binding.
3788:
3789: @node Objects Implementation, Comparison with other object models, Object Interfaces, Objects
3790: @subsubsection @file{objects.fs} Implementation
3791: @cindex @file{objects.fs} implementation
3792:
3793: @cindex @code{object-map} discussion
3794: An object is a piece of memory, like one of the data structures
3795: described with @code{struct...end-struct}. It has a field
3796: @code{object-map} that points to the method map for the object's
3797: class.
3798:
3799: @cindex method map
3800: @cindex virtual function table
3801: The @emph{method map}@footnote{This is Self terminology; in C++
3802: terminology: virtual function table.} is an array that contains the
3803: execution tokens (XTs) of the methods for the object's class. Each
3804: selector contains an offset into the method maps.
3805:
3806: @cindex @code{selector} implementation, class
3807: @code{selector} is a defining word that uses
3808: @code{create} and @code{does>}. The body of the
3809: selector contains the offset; the @code{does>} action for a
3810: class selector is, basically:
3811:
3812: @example
3813: ( object addr ) @@ over object-map @@ + @@ execute
3814: @end example
3815:
3816: Since @code{object-map} is the first field of the object, it
3817: does not generate any code. As you can see, calling a selector has a
3818: small, constant cost.
3819:
3820: @cindex @code{current-interface} discussion
3821: @cindex class implementation and representation
3822: A class is basically a @code{struct} combined with a method
3823: map. During the class definition the alignment and size of the class
3824: are passed on the stack, just as with @code{struct}s, so
3825: @code{field} can also be used for defining class
3826: fields. However, passing more items on the stack would be
3827: inconvenient, so @code{class} builds a data structure in memory,
3828: which is accessed through the variable
3829: @code{current-interface}. After its definition is complete, the
3830: class is represented on the stack by a pointer (e.g., as parameter for
3831: a child class definition).
3832:
3833: At the start, a new class has the alignment and size of its parent,
3834: and a copy of the parent's method map. Defining new fields extends the
3835: size and alignment; likewise, defining new selectors extends the
3836: method map. @code{overrides} just stores a new XT in the method
3837: map at the offset given by the selector.
3838:
3839: @cindex class binding, implementation
3840: Class binding just gets the XT at the offset given by the selector
3841: from the class's method map and @code{compile,}s (in the case of
3842: @code{[bind]}) it.
3843:
3844: @cindex @code{this} implementation
3845: @cindex @code{catch} and @code{this}
3846: @cindex @code{this} and @code{catch}
3847: I implemented @code{this} as a @code{value}. At the
3848: start of an @code{m:...;m} method the old @code{this} is
3849: stored to the return stack and restored at the end; and the object on
3850: the TOS is stored @code{TO this}. This technique has one
3851: disadvantage: If the user does not leave the method via
3852: @code{;m}, but via @code{throw} or @code{exit},
3853: @code{this} is not restored (and @code{exit} may
3854: crash). To deal with the @code{throw} problem, I have redefined
3855: @code{catch} to save and restore @code{this}; the same
3856: should be done with any word that can catch an exception. As for
3857: @code{exit}, I simply forbid it (as a replacement, there is
3858: @code{exitm}).
3859:
3860: @cindex @code{inst-var} implementation
3861: @code{inst-var} is just the same as @code{field}, with
3862: a different @code{does>} action:
3863: @example
3864: @@ this +
3865: @end example
3866: Similar for @code{inst-value}.
3867:
3868: @cindex class scoping implementation
3869: Each class also has a wordlist that contains the words defined with
3870: @code{inst-var} and @code{inst-value}, and its protected
3871: words. It also has a pointer to its parent. @code{class} pushes
3872: the wordlists of the class an all its ancestors on the search order,
3873: and @code{end-class} drops them.
3874:
3875: @cindex interface implementation
3876: An interface is like a class without fields, parent and protected
3877: words; i.e., it just has a method map. If a class implements an
3878: interface, its method map contains a pointer to the method map of the
3879: interface. The positive offsets in the map are reserved for class
3880: methods, therefore interface map pointers have negative
3881: offsets. Interfaces have offsets that are unique throughout the
3882: system, unlike class selectors, whose offsets are only unique for the
3883: classes where the selector is available (invokable).
3884:
3885: This structure means that interface selectors have to perform one
3886: indirection more than class selectors to find their method. Their body
3887: contains the interface map pointer offset in the class method map, and
3888: the method offset in the interface method map. The
3889: @code{does>} action for an interface selector is, basically:
3890:
3891: @example
3892: ( object selector-body )
3893: 2dup selector-interface @@ ( object selector-body object interface-offset )
3894: swap object-map @@ + @@ ( object selector-body map )
3895: swap selector-offset @@ + @@ execute
3896: @end example
3897:
3898: where @code{object-map} and @code{selector-offset} are
3899: first fields and generate no code.
3900:
3901: As a concrete example, consider the following code:
3902:
3903: @example
3904: interface
3905: selector if1sel1
3906: selector if1sel2
3907: end-interface if1
3908:
3909: object class
3910: if1 implementation
3911: selector cl1sel1
3912: cell% inst-var cl1iv1
3913:
3914: ' m1 overrides construct
3915: ' m2 overrides if1sel1
3916: ' m3 overrides if1sel2
3917: ' m4 overrides cl1sel2
3918: end-class cl1
3919:
3920: create obj1 object dict-new drop
3921: create obj2 cl1 dict-new drop
3922: @end example
3923:
3924: The data structure created by this code (including the data structure
3925: for @code{object}) is shown in the <a
3926: href="objects-implementation.eps">figure</a>, assuming a cell size of 4.
3927:
3928: @node Comparison with other object models, Objects Glossary, Objects Implementation, Objects
3929: @subsubsection Comparison with other object models
3930: @cindex comparison of object models
3931: @cindex object models, comparison
3932:
3933: Many object-oriented Forth extensions have been proposed (@cite{A survey
3934: of object-oriented Forths} (SIGPLAN Notices, April 1996) by Bradford
3935: J. Rodriguez and W. F. S. Poehlman lists 17). Here I'll discuss the
3936: relation of @file{objects.fs} to two well-known and two closely-related
3937: (by the use of method maps) models.
3938:
3939: @cindex Neon model
3940: The most popular model currently seems to be the Neon model (see
3941: @cite{Object-oriented programming in ANS Forth} (Forth Dimensions, March
3942: 1997) by Andrew McKewan). The Neon model uses a @code{@emph{selector
3943: object}} syntax, which makes it unnatural to pass objects on the
3944: stack. It also requires that the selector parses the input stream (at
3945: compile time); this leads to reduced extensibility and to bugs that are
3946: hard to find. Finally, it allows using every selector to every object;
3947: this eliminates the need for classes, but makes it harder to create
3948: efficient implementations. A longer version of this critique can be
3949: found in @cite{On Standardizing Object-Oriented Forth Extensions} (Forth
3950: Dimensions, May 1997) by Anton Ertl.
3951:
3952: @cindex Pountain's object-oriented model
3953: Another well-known publication is @cite{Object-Oriented Forth} (Academic
3954: Press, London, 1987) by Dick Pountain. However, it is not really about
3955: object-oriented programming, because it hardly deals with late
3956: binding. Instead, it focuses on features like information hiding and
3957: overloading that are characteristic of modular languages like Ada (83).
3958:
3959: @cindex Zsoter's object-oriented model
3960: In @cite{Does late binding have to be slow?} (Forth Dimensions ??? 1996)
3961: Andras Zsoter describes a model that makes heavy use of an active object
3962: (like @code{this} in @file{objects.fs}): The active object is not only
3963: used for accessing all fields, but also specifies the receiving object
3964: of every selector invocation; you have to change the active object
3965: explicitly with @code{@{ ... @}}, whereas in @file{objects.fs} it
3966: changes more or less implicitly at @code{m: ... ;m}. Such a change at
3967: the method entry point is unnecessary with the Zsoter's model, because
3968: the receiving object is the active object already; OTOH, the explicit
3969: change is absolutely necessary in that model, because otherwise no one
3970: could ever change the active object. An ANS Forth implementation of this
3971: model is available at @url{http://www.forth.org/fig/oopf.html}.
3972:
3973: @cindex @file{oof.fs}, differences to other models
3974: The @file{oof.fs} model combines information hiding and overloading
3975: resolution (by keeping names in various wordlists) with object-oriented
3976: programming. It sets the active object implicitly on method entry, but
3977: also allows explicit changing (with @code{>o...o>} or with
3978: @code{with...endwith}). It uses parsing and state-smart objects and
3979: classes for resolving overloading and for early binding: the object or
3980: class parses the selector and determines the method from this. If the
3981: selector is not parsed by an object or class, it performs a call to the
3982: selector for the active object (late binding), like Zsoter's model.
3983: Fields are always accessed through the active object. The big
3984: disadvantage of this model is the parsing and the state-smartness, which
3985: reduces extensibility and increases the opportunities for subtle bugs;
3986: essentially, you are only safe if you never tick or @code{postpone} an
3987: object or class (Bernd disagrees, but I (Anton) am not convinced).
3988:
3989: @cindex @file{mini-oof.fs}, differences to other models
3990: The Mini-OOF model is quite similar to a very stripped-down version of
3991: the Objects model, but syntactically it is a mixture of the Objects and
3992: the OOF model.
3993:
3994:
3995: @node Objects Glossary, , Comparison with other object models, Objects
3996: @subsubsection @file{objects.fs} Glossary
3997: @cindex @file{objects.fs} Glossary
3998:
3999: doc-bind
4000: doc-<bind>
4001: doc-bind'
4002: doc-[bind]
4003: doc-class
4004: doc-class->map
4005: doc-class-inst-size
4006: doc-class-override!
4007: doc-construct
4008: doc-current'
4009: doc-[current]
4010: doc-current-interface
4011: doc-dict-new
4012: doc-drop-order
4013: doc-end-class
4014: doc-end-class-noname
4015: doc-end-interface
4016: doc-end-interface-noname
4017: doc-exitm
4018: doc-heap-new
4019: doc-implementation
4020: doc-init-object
4021: doc-inst-value
4022: doc-inst-var
4023: doc-interface
4024: doc-;m
4025: doc-m:
4026: doc-method
4027: doc-object
4028: doc-overrides
4029: doc-[parent]
4030: doc-print
4031: doc-protected
4032: doc-public
4033: doc-push-order
4034: doc-selector
4035: doc-this
4036: doc-<to-inst>
4037: doc-[to-inst]
4038: doc-to-this
4039: doc-xt-new
4040:
4041: @c -------------------------------------------------------------
4042: @node OOF, Mini-OOF, Objects, Object-oriented Forth
4043: @subsection OOF
4044: @cindex oof
4045: @cindex object-oriented programming
4046:
4047: @cindex @file{objects.fs}
4048: @cindex @file{oof.fs}
4049:
4050: This section describes the @file{oof.fs} packet. This section uses the
4051: same rationale why using object-oriented programming, and the same
4052: terminology.
4053:
4054: The packet described in this section is used in bigFORTH since 1991, and
4055: used for two large applications: a chromatographic system used to
4056: create new medicaments, and a graphic user interface library (MINOS).
4057:
4058: You can find a description (in German) of @file{oof.fs} in @cite{Object
4059: oriented bigFORTH} by Bernd Paysan, published in @cite{Vierte Dimension}
4060: 10(2), 1994.
4061:
4062: @menu
4063: * Properties of the OOF model::
4064: * Basic OOF Usage::
4065: * The base class object::
4066: * Class Declaration::
4067: * Class Implementation::
4068: @end menu
4069:
4070: @node Properties of the OOF model, Basic OOF Usage, OOF, OOF
4071: @subsubsection Properties of the OOF model
4072: @cindex @file{oof.fs} properties
4073:
4074: @itemize @bullet
4075: @item
4076: This model combines object oriented programming with information
4077: hiding. It helps you writing large application, where scoping is
4078: necessary, because it provides class-oriented scoping.
4079:
4080: @item
4081: Named objects, object pointers, and object arrays can be created,
4082: selector invocation uses the "object selector" syntax. Selector invocation
4083: to objects and/or selectors on the stack is a bit less convenient, but
4084: possible.
4085:
4086: @item
4087: Selector invocation and instance variable usage of the active object is
4088: straight forward, since both make use of the active object.
4089:
4090: @item
4091: Late binding is efficient and easy to use.
4092:
4093: @item
4094: State-smart objects parse selectors. However, extensibility is provided
4095: using a (parsing) selector @code{postpone} and a selector @code{'}.
4096:
4097: @item
4098: An implementation in ANS Forth is available.
4099:
4100: @end itemize
4101:
4102:
4103: @node Basic OOF Usage, The base class object, Properties of the OOF model, OOF
4104: @subsubsection Basic OOF Usage
4105: @cindex @file{oof.fs} usage
4106:
4107: Here, I use the same example as for @code{objects} (@pxref{Basic Objects Usage}).
4108:
4109: You can define a class for graphical objects like this:
4110:
4111: @cindex @code{class} usage
4112: @cindex @code{class;} usage
4113: @cindex @code{method} usage
4114: @example
4115: object class graphical \ "object" is the parent class
4116: method draw ( x y graphical -- )
4117: class;
4118: @end example
4119:
4120: This code defines a class @code{graphical} with an
4121: operation @code{draw}. We can perform the operation
4122: @code{draw} on any @code{graphical} object, e.g.:
4123:
4124: @example
4125: 100 100 t-rex draw
4126: @end example
4127:
4128: where @code{t-rex} is an object or object pointer, created with e.g.
4129: @code{graphical : t-rex}.
4130:
4131: @cindex abstract class
4132: How do we create a graphical object? With the present definitions,
4133: we cannot create a useful graphical object. The class
4134: @code{graphical} describes graphical objects in general, but not
4135: any concrete graphical object type (C++ users would call it an
4136: @emph{abstract class}); e.g., there is no method for the selector
4137: @code{draw} in the class @code{graphical}.
4138:
4139: For concrete graphical objects, we define child classes of the
4140: class @code{graphical}, e.g.:
4141:
4142: @example
4143: graphical class circle \ "graphical" is the parent class
4144: cell var circle-radius
4145: how:
4146: : draw ( x y -- )
4147: circle-radius @@ draw-circle ;
4148:
4149: : init ( n-radius -- (
4150: circle-radius ! ;
4151: class;
4152: @end example
4153:
4154: Here we define a class @code{circle} as a child of @code{graphical},
4155: with a field @code{circle-radius}; it defines new methods for the
4156: selectors @code{draw} and @code{init} (@code{init} is defined in
4157: @code{object}, the parent class of @code{graphical}).
4158:
4159: Now we can create a circle in the dictionary with
4160:
4161: @example
4162: 50 circle : my-circle
4163: @end example
4164:
4165: @code{:} invokes @code{init}, thus initializing the field
4166: @code{circle-radius} with 50. We can draw this new circle at (100,100)
4167: with
4168:
4169: @example
4170: 100 100 my-circle draw
4171: @end example
4172:
4173: @cindex selector invocation, restrictions
4174: @cindex class definition, restrictions
4175: Note: You can invoke a selector only if the receiving object belongs to
4176: the class where the selector was defined or one of its descendents;
4177: e.g., you can invoke @code{draw} only for objects belonging to
4178: @code{graphical} or its descendents (e.g., @code{circle}). The scoping
4179: mechanism will check if you try to invoke a selector that is not
4180: defined in this class hierarchy, so you'll get an error at compilation
4181: time.
4182:
4183:
4184: @node The base class object, Class Declaration, Basic OOF Usage, OOF
4185: @subsubsection The base class @file{object}
4186: @cindex @file{oof.fs} base class
4187:
4188: When you define a class, you have to specify a parent class. So how do
4189: you start defining classes? There is one class available from the start:
4190: @code{object}. You have to use it as ancestor for all classes. It is the
4191: only class that has no parent. Classes are also objects, except that
4192: they don't have instance variables; class manipulation such as
4193: inheritance or changing definitions of a class is handled through
4194: selectors of the class @code{object}.
4195:
4196: @code{object} provides a number of selectors:
4197:
4198: @itemize @bullet
4199: @item
4200: @code{class} for subclassing, @code{definitions} to add definitions
4201: later on, and @code{class?} to get type informations (is the class a
4202: subclass of the class passed on the stack?).
4203: doc---object-class
4204: doc---object-definitions
4205: doc---object-class?
4206:
4207: @item
4208: @code{init} and @code{dispose} as constructor and destroctor of the
4209: object. @code{init} is invocated after the object's memory is allocated,
4210: while @code{dispose} also handles deallocation. Thus if you redefine
4211: @code{dispose}, you have to call the parent's dispose with @code{super
4212: dispose}, too.
4213: doc---object-init
4214: doc---object-dispose
4215:
4216: @item
4217: @code{new}, @code{new[]}, @code{:}, @code{ptr}, @code{asptr}, and
4218: @code{[]} to create named and unnamed objects and object arrays or
4219: object pointers.
4220: doc---object-new
4221: doc---object-new[]
4222: doc---object-:
4223: doc---object-ptr
4224: doc---object-asptr
4225: doc---object-[]
4226:
4227: @item
4228: @code{::} and @code{super} for expicit scoping. You should use expicit
4229: scoping only for super classes or classes with the same set of instance
4230: variables. Explicit scoped selectors use early binding.
4231: doc---object-::
4232: doc---object-super
4233:
4234: @item
4235: @code{self} to get the address of the object
4236: doc---object-self
4237:
4238: @item
4239: @code{bind}, @code{bound}, @code{link}, and @code{is} to assign object
4240: pointers and instance defers.
4241: doc---object-bind
4242: doc---object-bound
4243: doc---object-link
4244: doc---object-is
4245:
4246: @item
4247: @code{'} to obtain selector tokens, @code{send} to invocate selectors
4248: form the stack, and @code{postpone} to generate selector invocation code.
4249: doc---object-'
4250: doc---object-postpone
4251:
4252: @item
4253: @code{with} and @code{endwith} to select the active object from the
4254: stack, and enabling it's scope. Using @code{with} and @code{endwith}
4255: also allows to create code using selector @code{postpone} without being
4256: trapped bye the state-smart objects.
4257: doc---object-with
4258: doc---object-endwith
4259:
4260: @end itemize
4261:
4262: @node Class Declaration, Class Implementation, The base class object, OOF
4263: @subsubsection Class Declaration
4264: @cindex class declaration
4265:
4266: @itemize @bullet
4267: @item
4268: Instance variables
4269: doc---oof-var
4270:
4271: @item
4272: Object pointers
4273: doc---oof-ptr
4274: doc---oof-asptr
4275:
4276: @item
4277: Instance defers
4278: doc---oof-defer
4279:
4280: @item
4281: Method selectors
4282: doc---oof-early
4283: doc---oof-method
4284:
4285: @item
4286: Class wide variables
4287: doc---oof-static
4288:
4289: @item
4290: End declaration
4291: doc---oof-how:
4292: doc---oof-class;
4293:
4294: @end itemize
4295:
4296: @c -------------------------------------------------------------
4297: @node Class Implementation, , Class Declaration, OOF
4298: @subsubsection Class Implementation
4299: @cindex class implementation
4300:
4301: @c -------------------------------------------------------------
4302: @node Mini-OOF, , OOF, Object-oriented Forth
4303: @subsection Mini-OOF
4304: @cindex mini-oof
4305:
4306: Gforth's third object oriented Forth package is a 12-liner. It uses a
4307: bit of a mixture of the @file{object.fs} and the @file{oof.fs} syntax,
4308: and reduces to the bare minimum of features. This is based on a posting
4309: of Bernd Paysan in comp.arch.
4310:
4311: @menu
4312: * Mini-OOF Usage::
4313: * Mini-OOF Example::
4314: @end menu
4315:
4316: @c -------------------------------------------------------------
4317: @node Mini-OOF Usage, Mini-OOF Example, , Mini-OOF
4318: @subsubsection Usage
4319: @cindex mini-oof usage
4320:
4321: Basically, there are seven words, to define a method, a variable, a
4322: class; to end a class, to define a method, to allocate an object, to
4323: resolve binding, and the base class (which allocates one cell for the
4324: object pointer).
4325:
4326: doc-method
4327:
4328: Defines a method
4329:
4330: doc-var
4331:
4332: Defines a variable with size bytes
4333:
4334: doc-class
4335:
4336: Starts the definition of a sub-class
4337:
4338: doc-end-class
4339:
4340: Ends the definition of a class
4341:
4342: doc-defines
4343:
4344: Binds the xt to the method name in the class
4345:
4346: doc-new
4347:
4348: Creates a new incarnation of the class
4349:
4350: doc-::
4351:
4352: Compiles the method name of the class (not immediate!)
4353:
4354: doc-object
4355:
4356: Is the base class of all objects
4357:
4358: @c -------------------------------------------------------------
4359: @node Mini-OOF Example, , Mini-OOF Usage, Mini-OOF
4360: @subsubsection Mini-OOF Example
4361: @cindex mini-oof example
4362:
4363: A short example shows how to use this package.
4364:
4365: @example
4366: object class
4367: method init
4368: method draw
4369: end-class graphical
4370: @end example
4371:
4372: This code defines a class @code{graphical} with an
4373: operation @code{draw}. We can perform the operation
4374: @code{draw} on any @code{graphical} object, e.g.:
4375:
4376: @example
4377: 100 100 t-rex draw
4378: @end example
4379:
4380: where @code{t-rex} is an object or object pointer, created with e.g.
4381: @code{graphical new Constant t-rex}.
4382:
4383: For concrete graphical objects, we define child classes of the
4384: class @code{graphical}, e.g.:
4385:
4386: @example
4387: graphical class
4388: cell var circle-radius
4389: end-class circle \ "graphical" is the parent class
4390:
4391: :noname ( x y -- )
4392: circle-radius @@ draw-circle ; circle defines draw
4393: :noname ( r -- )
4394: circle-radius ! ; circle defines init
4395: @end example
4396:
4397: There is no implicit init method, so we have to define one. The creation
4398: code of the object now has to call init explicitely.
4399:
4400: @example
4401: circle new Constant my-circle
4402: 50 my-circle init
4403: @end example
4404:
4405: It is also possible to add a function to create named objects with
4406: automatic call of @code{init}, given that all objects have @code{init}
4407: on the same place
4408:
4409: @example
4410: : new: ( .. o "name" -- )
4411: new dup Constant init ;
4412: 80 circle new: large-circle
4413: @end example
4414:
4415: We can draw this new circle at (100,100)
4416: with
4417:
4418: @example
4419: 100 100 my-circle draw
4420: @end example
4421:
4422: @c -------------------------------------------------------------
4423: @node Tokens for Words, Wordlists, Object-oriented Forth, Words
4424: @section Tokens for Words
4425: @cindex tokens for words
4426:
4427: This chapter describes the creation and use of tokens that represent
4428: words on the stack (and in data space).
4429:
4430: Named words have interpretation and compilation semantics. Unnamed words
4431: just have execution semantics.
4432:
4433: @cindex execution token
4434: An @dfn{execution token} represents the execution semantics of an
4435: unnamed word. An execution token occupies one cell. As explained in
4436: section @ref{Supplying names}, the execution token of the last words
4437: defined can be produced with
4438:
4439: short-lastxt
4440:
4441: You can perform the semantics represented by an execution token with
4442: doc-execute
4443: You can compile the word with
4444: doc-compile,
4445:
4446: @cindex code field address
4447: @cindex CFA
4448: In Gforth, the abstract data type @emph{execution token} is implemented
4449: as CFA (code field address).
4450:
4451: The interpretation semantics of a named word are also represented by an
4452: execution token. You can get it with
4453:
4454: doc-[']
4455: doc-'
4456:
4457: For literals, you use @code{'} in interpreted code and @code{[']} in
4458: compiled code. Gforth's @code{'} and @code{[']} behave somewhat unusual
4459: by complaining about compile-only words. To get an execution token for a
4460: compiling word @var{X}, use @code{COMP' @var{X} drop} or @code{[COMP']
4461: @var{X} drop}.
4462:
4463: @cindex compilation token
4464: The compilation semantics are represented by a @dfn{compilation token}
4465: consisting of two cells: @var{w xt}. The top cell @var{xt} is an
4466: execution token. The compilation semantics represented by the
4467: compilation token can be performed with @code{execute}, which consumes
4468: the whole compilation token, with an additional stack effect determined
4469: by the represented compilation semantics.
4470:
4471: doc-[comp']
4472: doc-comp'
4473:
4474: You can compile the compilation semantics with @code{postpone,}. I.e.,
4475: @code{COMP' @var{word} POSTPONE,} is equivalent to @code{POSTPONE
4476: @var{word}}.
4477:
4478: doc-postpone,
4479:
4480: At present, the @var{w} part of a compilation token is an execution
4481: token, and the @var{xt} part represents either @code{execute} or
4482: @code{compile,}. However, don't rely on that knowledge, unless necessary;
4483: we may introduce unusual compilation tokens in the future (e.g.,
4484: compilation tokens representing the compilation semantics of literals).
4485:
4486: @cindex name token
4487: @cindex name field address
4488: @cindex NFA
4489: Named words are also represented by the @dfn{name token}. The abstract
4490: data type @emph{name token} is implemented as NFA (name field address).
4491:
4492: doc-find-name
4493: doc-name>int
4494: doc-name?int
4495: doc-name>comp
4496: doc-name>string
4497:
4498: @node Wordlists, Files, Tokens for Words, Words
4499: @section Wordlists
4500:
4501: @node Files, Including Files, Wordlists, Words
4502: @section Files
4503:
4504: @node Including Files, Blocks, Files, Words
4505: @section Including Files
4506: @cindex including files
4507:
4508: @menu
4509: * Words for Including::
4510: * Search Path::
4511: * Changing the Search Path::
4512: * General Search Paths::
4513: @end menu
4514:
4515: @node Words for Including, Search Path, Including Files, Including Files
4516: @subsection Words for Including
4517:
4518: doc-include-file
4519: doc-included
4520: doc-include
4521:
4522: Usually you want to include a file only if it is not included already
4523: (by, say, another source file):
4524:
4525: doc-required
4526: doc-require
4527: doc-needs
4528:
4529: @cindex stack effect of included files
4530: @cindex including files, stack effect
4531: I recommend that you write your source files such that interpreting them
4532: does not change the stack. This allows using these files with
4533: @code{required} and friends without complications. E.g.,
4534:
4535: @example
4536: 1 require foo.fs drop
4537: @end example
4538:
4539: @node Search Path, Changing the Search Path, Words for Including, Including Files
4540: @subsection Search Path
4541: @cindex path for @code{included}
4542: @cindex file search path
4543: @cindex include search path
4544: @cindex search path for files
4545:
4546: If you specify an absolute filename (i.e., a filename starting with
4547: @file{/} or @file{~}, or with @file{:} in the second position (as in
4548: @samp{C:...})) for @code{included} and friends, that file is included
4549: just as you would expect.
4550:
4551: For relative filenames, Gforth uses a search path similar to Forth's
4552: search order (@pxref{Wordlists}). It tries to find the given filename in
4553: the directories present in the path, and includes the first one it
4554: finds.
4555:
4556: If the search path contains the directory @file{.} (as it should), this
4557: refers to the directory that the present file was @code{included}
4558: from. This allows files to include other files relative to their own
4559: position (irrespective of the current working directory or the absolute
4560: position). This feature is essential for libraries consisting of
4561: several files, where a file may include other files from the library.
4562: It corresponds to @code{#include "..."} in C. If the current input
4563: source is not a file, @file{.} refers to the directory of the innermost
4564: file being included, or, if there is no file being included, to the
4565: current working directory.
4566:
4567: Use @file{~+} to refer to the current working directory (as in the
4568: @code{bash}).
4569:
4570: If the filename starts with @file{./}, the search path is not searched
4571: (just as with absolute filenames), and the @file{.} has the same meaning
4572: as described above.
4573:
4574: @node Changing the Search Path, General Search Paths, Search Path, Including Files
4575: @subsection Changing the Search Path
4576: @cindex search path, changes
4577:
4578: The search path is initialized when you start Gforth (@pxref{Invoking
4579: Gforth}). You can display it with
4580:
4581: doc-.fpath
4582:
4583: You can change it later with the following words:
4584:
4585: doc-fpath+
4586: doc-fpath=
4587:
4588: Using fpath and require would look like:
4589:
4590: @example
4591: fpath= /usr/lib/forth/|./
4592:
4593: require timer.fs
4594: @end example
4595:
4596: If you have the need to look for a file in the Forth search path, you could
4597: use this Gforth feature in your application.
4598:
4599: doc-open-fpath-file
4600:
4601:
4602: @node General Search Paths, , Changing the Search Path, Including Files
4603: @subsection General Search Paths
4604: @cindex search paths for user applications
4605:
4606: Your application may need to search files in sevaral directories, like
4607: @code{included} does. For this purpose you can define and use your own
4608: search paths. Create a search path like this:
4609:
4610: @example
4611:
4612: Make a buffer for the path:
4613: create mypath 100 chars , \ maximum length (is checked)
4614: 0 , \ real len
4615: 100 chars allot \ space for path
4616: @end example
4617:
4618: You have the same functions for the forth search path in a generic version
4619: for different paths.
4620:
4621: doc-path+
4622: doc-path=
4623: doc-.path
4624: doc-open-path-file
4625:
4626:
4627: @node Blocks, Other I/O, Including Files, Words
4628: @section Blocks
4629:
4630: @node Other I/O, Programming Tools, Blocks, Words
4631: @section Other I/O
4632:
4633: @node Programming Tools, Assembler and Code Words, Other I/O, Words
4634: @section Programming Tools
4635: @cindex programming tools
4636:
4637: @menu
4638: * Debugging:: Simple and quick.
4639: * Assertions:: Making your programs self-checking.
4640: * Singlestep Debugger:: Executing your program word by word.
4641: @end menu
4642:
4643: @node Debugging, Assertions, Programming Tools, Programming Tools
4644: @subsection Debugging
4645: @cindex debugging
4646:
4647: The simple debugging aids provided in @file{debugs.fs}
4648: are meant to support a different style of debugging than the
4649: tracing/stepping debuggers used in languages with long turn-around
4650: times.
4651:
4652: A much better (faster) way in fast-compiling languages is to add
4653: printing code at well-selected places, let the program run, look at
4654: the output, see where things went wrong, add more printing code, etc.,
4655: until the bug is found.
4656:
4657: The word @code{~~} is easy to insert. It just prints debugging
4658: information (by default the source location and the stack contents). It
4659: is also easy to remove (@kbd{C-x ~} in the Emacs Forth mode to
4660: query-replace them with nothing). The deferred words
4661: @code{printdebugdata} and @code{printdebugline} control the output of
4662: @code{~~}. The default source location output format works well with
4663: Emacs' compilation mode, so you can step through the program at the
4664: source level using @kbd{C-x `} (the advantage over a stepping debugger
4665: is that you can step in any direction and you know where the crash has
4666: happened or where the strange data has occurred).
4667:
4668: Note that the default actions clobber the contents of the pictured
4669: numeric output string, so you should not use @code{~~}, e.g., between
4670: @code{<#} and @code{#>}.
4671:
4672: doc-~~
4673: doc-printdebugdata
4674: doc-printdebugline
4675:
4676: @node Assertions, Singlestep Debugger, Debugging, Programming Tools
4677: @subsection Assertions
4678: @cindex assertions
4679:
4680: It is a good idea to make your programs self-checking, in particular, if
4681: you use an assumption (e.g., that a certain field of a data structure is
4682: never zero) that may become wrong during maintenance. Gforth supports
4683: assertions for this purpose. They are used like this:
4684:
4685: @example
4686: assert( @var{flag} )
4687: @end example
4688:
4689: The code between @code{assert(} and @code{)} should compute a flag, that
4690: should be true if everything is alright and false otherwise. It should
4691: not change anything else on the stack. The overall stack effect of the
4692: assertion is @code{( -- )}. E.g.
4693:
4694: @example
4695: assert( 1 1 + 2 = ) \ what we learn in school
4696: assert( dup 0<> ) \ assert that the top of stack is not zero
4697: assert( false ) \ this code should not be reached
4698: @end example
4699:
4700: The need for assertions is different at different times. During
4701: debugging, we want more checking, in production we sometimes care more
4702: for speed. Therefore, assertions can be turned off, i.e., the assertion
4703: becomes a comment. Depending on the importance of an assertion and the
4704: time it takes to check it, you may want to turn off some assertions and
4705: keep others turned on. Gforth provides several levels of assertions for
4706: this purpose:
4707:
4708: doc-assert0(
4709: doc-assert1(
4710: doc-assert2(
4711: doc-assert3(
4712: doc-assert(
4713: doc-)
4714:
4715: @code{Assert(} is the same as @code{assert1(}. The variable
4716: @code{assert-level} specifies the highest assertions that are turned
4717: on. I.e., at the default @code{assert-level} of one, @code{assert0(} and
4718: @code{assert1(} assertions perform checking, while @code{assert2(} and
4719: @code{assert3(} assertions are treated as comments.
4720:
4721: Note that the @code{assert-level} is evaluated at compile-time, not at
4722: run-time. I.e., you cannot turn assertions on or off at run-time, you
4723: have to set the @code{assert-level} appropriately before compiling a
4724: piece of code. You can compile several pieces of code at several
4725: @code{assert-level}s (e.g., a trusted library at level 1 and newly
4726: written code at level 3).
4727:
4728: doc-assert-level
4729:
4730: If an assertion fails, a message compatible with Emacs' compilation mode
4731: is produced and the execution is aborted (currently with @code{ABORT"}.
4732: If there is interest, we will introduce a special throw code. But if you
4733: intend to @code{catch} a specific condition, using @code{throw} is
4734: probably more appropriate than an assertion).
4735:
4736: @node Singlestep Debugger, , Assertions, Programming Tools
4737: @subsection Singlestep Debugger
4738: @cindex singlestep Debugger
4739: @cindex debugging Singlestep
4740: @cindex @code{dbg}
4741: @cindex @code{BREAK:}
4742: @cindex @code{BREAK"}
4743:
4744: When a new word is created there's often the need to check whether it behaves
4745: correctly or not. You can do this by typing @code{dbg badword}. This might
4746: look like:
4747: @example
4748: : badword 0 DO i . LOOP ; ok
4749: 2 dbg badword
4750: : badword
4751: Scanning code...
4752:
4753: Nesting debugger ready!
4754:
4755: 400D4738 8049BC4 0 -> [ 2 ] 00002 00000
4756: 400D4740 8049F68 DO -> [ 0 ]
4757: 400D4744 804A0C8 i -> [ 1 ] 00000
4758: 400D4748 400C5E60 . -> 0 [ 0 ]
4759: 400D474C 8049D0C LOOP -> [ 0 ]
4760: 400D4744 804A0C8 i -> [ 1 ] 00001
4761: 400D4748 400C5E60 . -> 1 [ 0 ]
4762: 400D474C 8049D0C LOOP -> [ 0 ]
4763: 400D4758 804B384 ; -> ok
4764: @end example
4765:
4766: Each line displayed is one step. You always have to hit return to
4767: execute the next word that is displayed. If you don't want to execute
4768: the next word in a whole, you have to type @kbd{n} for @code{nest}. Here is
4769: an overview what keys are available:
4770:
4771: @table @i
4772:
4773: @item <return>
4774: Next; Execute the next word.
4775:
4776: @item n
4777: Nest; Single step through next word.
4778:
4779: @item u
4780: Unnest; Stop debugging and execute rest of word. If we got to this word
4781: with nest, continue debugging with the calling word.
4782:
4783: @item d
4784: Done; Stop debugging and execute rest.
4785:
4786: @item s
4787: Stopp; Abort immediately.
4788:
4789: @end table
4790:
4791: Debugging large application with this mechanism is very difficult, because
4792: you have to nest very deep into the program before the interesting part
4793: begins. This takes a lot of time.
4794:
4795: To do it more directly put a @code{BREAK:} command into your source code.
4796: When program execution reaches @code{BREAK:} the single step debugger is
4797: invoked and you have all the features described above.
4798:
4799: If you have more than one part to debug it is useful to know where the
4800: program has stopped at the moment. You can do this by the
4801: @code{BREAK" string"} command. This behaves like @code{BREAK:} except that
4802: string is typed out when the ``breakpoint'' is reached.
4803:
4804: @node Assembler and Code Words, Threading Words, Programming Tools, Words
4805: @section Assembler and Code Words
4806: @cindex assembler
4807: @cindex code words
4808:
4809: Gforth provides some words for defining primitives (words written in
4810: machine code), and for defining the the machine-code equivalent of
4811: @code{DOES>}-based defining words. However, the machine-independent
4812: nature of Gforth poses a few problems: First of all, Gforth runs on
4813: several architectures, so it can provide no standard assembler. What's
4814: worse is that the register allocation not only depends on the processor,
4815: but also on the @code{gcc} version and options used.
4816:
4817: The words that Gforth offers encapsulate some system dependences (e.g., the
4818: header structure), so a system-independent assembler may be used in
4819: Gforth. If you do not have an assembler, you can compile machine code
4820: directly with @code{,} and @code{c,}.
4821:
4822: doc-assembler
4823: doc-code
4824: doc-end-code
4825: doc-;code
4826: doc-flush-icache
4827:
4828: If @code{flush-icache} does not work correctly, @code{code} words
4829: etc. will not work (reliably), either.
4830:
4831: These words are rarely used. Therefore they reside in @code{code.fs},
4832: which is usually not loaded (except @code{flush-icache}, which is always
4833: present). You can load them with @code{require code.fs}.
4834:
4835: @cindex registers of the inner interpreter
4836: In the assembly code you will want to refer to the inner interpreter's
4837: registers (e.g., the data stack pointer) and you may want to use other
4838: registers for temporary storage. Unfortunately, the register allocation
4839: is installation-dependent.
4840:
4841: The easiest solution is to use explicit register declarations
4842: (@pxref{Explicit Reg Vars, , Variables in Specified Registers, gcc.info,
4843: GNU C Manual}) for all of the inner interpreter's registers: You have to
4844: compile Gforth with @code{-DFORCE_REG} (configure option
4845: @code{--enable-force-reg}) and the appropriate declarations must be
4846: present in the @code{machine.h} file (see @code{mips.h} for an example;
4847: you can find a full list of all declarable register symbols with
4848: @code{grep register engine.c}). If you give explicit registers to all
4849: variables that are declared at the beginning of @code{engine()}, you
4850: should be able to use the other caller-saved registers for temporary
4851: storage. Alternatively, you can use the @code{gcc} option
4852: @code{-ffixed-REG} (@pxref{Code Gen Options, , Options for Code
4853: Generation Conventions, gcc.info, GNU C Manual}) to reserve a register
4854: (however, this restriction on register allocation may slow Gforth
4855: significantly).
4856:
4857: If this solution is not viable (e.g., because @code{gcc} does not allow
4858: you to explicitly declare all the registers you need), you have to find
4859: out by looking at the code where the inner interpreter's registers
4860: reside and which registers can be used for temporary storage. You can
4861: get an assembly listing of the engine's code with @code{make engine.s}.
4862:
4863: In any case, it is good practice to abstract your assembly code from the
4864: actual register allocation. E.g., if the data stack pointer resides in
4865: register @code{$17}, create an alias for this register called @code{sp},
4866: and use that in your assembly code.
4867:
4868: @cindex code words, portable
4869: Another option for implementing normal and defining words efficiently
4870: is: adding the wanted functionality to the source of Gforth. For normal
4871: words you just have to edit @file{primitives} (@pxref{Automatic
4872: Generation}), defining words (equivalent to @code{;CODE} words, for fast
4873: defined words) may require changes in @file{engine.c}, @file{kernal.fs},
4874: @file{prims2x.fs}, and possibly @file{cross.fs}.
4875:
4876:
4877: @node Threading Words, , Assembler and Code Words, Words
4878: @section Threading Words
4879: @cindex threading words
4880:
4881: @cindex code address
4882: These words provide access to code addresses and other threading stuff
4883: in Gforth (and, possibly, other interpretive Forths). It more or less
4884: abstracts away the differences between direct and indirect threading
4885: (and, for direct threading, the machine dependences). However, at
4886: present this wordset is still incomplete. It is also pretty low-level;
4887: some day it will hopefully be made unnecessary by an internals wordset
4888: that abstracts implementation details away completely.
4889:
4890: doc->code-address
4891: doc->does-code
4892: doc-code-address!
4893: doc-does-code!
4894: doc-does-handler!
4895: doc-/does-handler
4896:
4897: The code addresses produced by various defining words are produced by
4898: the following words:
4899:
4900: doc-docol:
4901: doc-docon:
4902: doc-dovar:
4903: doc-douser:
4904: doc-dodefer:
4905: doc-dofield:
4906:
4907: You can recognize words defined by a @code{CREATE}...@code{DOES>} word
4908: with @code{>DOES-CODE}. If the word was defined in that way, the value
4909: returned is different from 0 and identifies the @code{DOES>} used by the
4910: defining word.
4911:
4912: @c ******************************************************************
4913: @node Tools, ANS conformance, Words, Top
4914: @chapter Tools
4915:
4916: @menu
4917: * ANS Report:: Report the words used, sorted by wordset.
4918: @end menu
4919:
4920: See also @ref{Emacs and Gforth}.
4921:
4922: @node ANS Report, , Tools, Tools
4923: @section @file{ans-report.fs}: Report the words used, sorted by wordset
4924: @cindex @file{ans-report.fs}
4925: @cindex report the words used in your program
4926: @cindex words used in your program
4927:
4928: If you want to label a Forth program as ANS Forth Program, you must
4929: document which wordsets the program uses; for extension wordsets, it is
4930: helpful to list the words the program requires from these wordsets
4931: (because Forth systems are allowed to provide only some words of them).
4932:
4933: The @file{ans-report.fs} tool makes it easy for you to determine which
4934: words from which wordset and which non-ANS words your application
4935: uses. You simply have to include @file{ans-report.fs} before loading the
4936: program you want to check. After loading your program, you can get the
4937: report with @code{print-ans-report}. A typical use is to run this as
4938: batch job like this:
4939: @example
4940: gforth ans-report.fs myprog.fs -e "print-ans-report bye"
4941: @end example
4942:
4943: The output looks like this (for @file{compat/control.fs}):
4944: @example
4945: The program uses the following words
4946: from CORE :
4947: : POSTPONE THEN ; immediate ?dup IF 0=
4948: from BLOCK-EXT :
4949: \
4950: from FILE :
4951: (
4952: @end example
4953:
4954: @subsection Caveats
4955:
4956: Note that @file{ans-report.fs} just checks which words are used, not whether
4957: they are used in an ANS Forth conforming way!
4958:
4959: Some words are defined in several wordsets in the
4960: standard. @file{ans-report.fs} reports them for only one of the
4961: wordsets, and not necessarily the one you expect. It depends on usage
4962: which wordset is the right one to specify. E.g., if you only use the
4963: compilation semantics of @code{S"}, it is a Core word; if you also use
4964: its interpretation semantics, it is a File word.
4965:
4966: @c ******************************************************************
4967: @node ANS conformance, Model, Tools, Top
4968: @chapter ANS conformance
4969: @cindex ANS conformance of Gforth
4970:
4971: To the best of our knowledge, Gforth is an
4972:
4973: ANS Forth System
4974: @itemize @bullet
4975: @item providing the Core Extensions word set
4976: @item providing the Block word set
4977: @item providing the Block Extensions word set
4978: @item providing the Double-Number word set
4979: @item providing the Double-Number Extensions word set
4980: @item providing the Exception word set
4981: @item providing the Exception Extensions word set
4982: @item providing the Facility word set
4983: @item providing @code{MS} and @code{TIME&DATE} from the Facility Extensions word set
4984: @item providing the File Access word set
4985: @item providing the File Access Extensions word set
4986: @item providing the Floating-Point word set
4987: @item providing the Floating-Point Extensions word set
4988: @item providing the Locals word set
4989: @item providing the Locals Extensions word set
4990: @item providing the Memory-Allocation word set
4991: @item providing the Memory-Allocation Extensions word set (that one's easy)
4992: @item providing the Programming-Tools word set
4993: @item providing @code{;CODE}, @code{AHEAD}, @code{ASSEMBLER}, @code{BYE}, @code{CODE}, @code{CS-PICK}, @code{CS-ROLL}, @code{STATE}, @code{[ELSE]}, @code{[IF]}, @code{[THEN]} from the Programming-Tools Extensions word set
4994: @item providing the Search-Order word set
4995: @item providing the Search-Order Extensions word set
4996: @item providing the String word set
4997: @item providing the String Extensions word set (another easy one)
4998: @end itemize
4999:
5000: @cindex system documentation
5001: In addition, ANS Forth systems are required to document certain
5002: implementation choices. This chapter tries to meet these
5003: requirements. In many cases it gives a way to ask the system for the
5004: information instead of providing the information directly, in
5005: particular, if the information depends on the processor, the operating
5006: system or the installation options chosen, or if they are likely to
5007: change during the maintenance of Gforth.
5008:
5009: @comment The framework for the rest has been taken from pfe.
5010:
5011: @menu
5012: * The Core Words::
5013: * The optional Block word set::
5014: * The optional Double Number word set::
5015: * The optional Exception word set::
5016: * The optional Facility word set::
5017: * The optional File-Access word set::
5018: * The optional Floating-Point word set::
5019: * The optional Locals word set::
5020: * The optional Memory-Allocation word set::
5021: * The optional Programming-Tools word set::
5022: * The optional Search-Order word set::
5023: @end menu
5024:
5025:
5026: @c =====================================================================
5027: @node The Core Words, The optional Block word set, ANS conformance, ANS conformance
5028: @comment node-name, next, previous, up
5029: @section The Core Words
5030: @c =====================================================================
5031: @cindex core words, system documentation
5032: @cindex system documentation, core words
5033:
5034: @menu
5035: * core-idef:: Implementation Defined Options
5036: * core-ambcond:: Ambiguous Conditions
5037: * core-other:: Other System Documentation
5038: @end menu
5039:
5040: @c ---------------------------------------------------------------------
5041: @node core-idef, core-ambcond, The Core Words, The Core Words
5042: @subsection Implementation Defined Options
5043: @c ---------------------------------------------------------------------
5044: @cindex core words, implementation-defined options
5045: @cindex implementation-defined options, core words
5046:
5047:
5048: @table @i
5049: @item (Cell) aligned addresses:
5050: @cindex cell-aligned addresses
5051: @cindex aligned addresses
5052: processor-dependent. Gforth's alignment words perform natural alignment
5053: (e.g., an address aligned for a datum of size 8 is divisible by
5054: 8). Unaligned accesses usually result in a @code{-23 THROW}.
5055:
5056: @item @code{EMIT} and non-graphic characters:
5057: @cindex @code{EMIT} and non-graphic characters
5058: @cindex non-graphic characters and @code{EMIT}
5059: The character is output using the C library function (actually, macro)
5060: @code{putc}.
5061:
5062: @item character editing of @code{ACCEPT} and @code{EXPECT}:
5063: @cindex character editing of @code{ACCEPT} and @code{EXPECT}
5064: @cindex editing in @code{ACCEPT} and @code{EXPECT}
5065: @cindex @code{ACCEPT}, editing
5066: @cindex @code{EXPECT}, editing
5067: This is modeled on the GNU readline library (@pxref{Readline
5068: Interaction, , Command Line Editing, readline, The GNU Readline
5069: Library}) with Emacs-like key bindings. @kbd{Tab} deviates a little by
5070: producing a full word completion every time you type it (instead of
5071: producing the common prefix of all completions).
5072:
5073: @item character set:
5074: @cindex character set
5075: The character set of your computer and display device. Gforth is
5076: 8-bit-clean (but some other component in your system may make trouble).
5077:
5078: @item Character-aligned address requirements:
5079: @cindex character-aligned address requirements
5080: installation-dependent. Currently a character is represented by a C
5081: @code{unsigned char}; in the future we might switch to @code{wchar_t}
5082: (Comments on that requested).
5083:
5084: @item character-set extensions and matching of names:
5085: @cindex character-set extensions and matching of names
5086: @cindex case sensitivity for name lookup
5087: @cindex name lookup, case sensitivity
5088: @cindex locale and case sensitivity
5089: Any character except the ASCII NUL charcter can be used in a
5090: name. Matching is case-insensitive (except in @code{TABLE}s). The
5091: matching is performed using the C function @code{strncasecmp}, whose
5092: function is probably influenced by the locale. E.g., the @code{C} locale
5093: does not know about accents and umlauts, so they are matched
5094: case-sensitively in that locale. For portability reasons it is best to
5095: write programs such that they work in the @code{C} locale. Then one can
5096: use libraries written by a Polish programmer (who might use words
5097: containing ISO Latin-2 encoded characters) and by a French programmer
5098: (ISO Latin-1) in the same program (of course, @code{WORDS} will produce
5099: funny results for some of the words (which ones, depends on the font you
5100: are using)). Also, the locale you prefer may not be available in other
5101: operating systems. Hopefully, Unicode will solve these problems one day.
5102:
5103: @item conditions under which control characters match a space delimiter:
5104: @cindex space delimiters
5105: @cindex control characters as delimiters
5106: If @code{WORD} is called with the space character as a delimiter, all
5107: white-space characters (as identified by the C macro @code{isspace()})
5108: are delimiters. @code{PARSE}, on the other hand, treats space like other
5109: delimiters. @code{PARSE-WORD} treats space like @code{WORD}, but behaves
5110: like @code{PARSE} otherwise. @code{(NAME)}, which is used by the outer
5111: interpreter (aka text interpreter) by default, treats all white-space
5112: characters as delimiters.
5113:
5114: @item format of the control flow stack:
5115: @cindex control flow stack, format
5116: The data stack is used as control flow stack. The size of a control flow
5117: stack item in cells is given by the constant @code{cs-item-size}. At the
5118: time of this writing, an item consists of a (pointer to a) locals list
5119: (third), an address in the code (second), and a tag for identifying the
5120: item (TOS). The following tags are used: @code{defstart},
5121: @code{live-orig}, @code{dead-orig}, @code{dest}, @code{do-dest},
5122: @code{scopestart}.
5123:
5124: @item conversion of digits > 35
5125: @cindex digits > 35
5126: The characters @code{[\]^_'} are the digits with the decimal value
5127: 36@minus{}41. There is no way to input many of the larger digits.
5128:
5129: @item display after input terminates in @code{ACCEPT} and @code{EXPECT}:
5130: @cindex @code{EXPECT}, display after end of input
5131: @cindex @code{ACCEPT}, display after end of input
5132: The cursor is moved to the end of the entered string. If the input is
5133: terminated using the @kbd{Return} key, a space is typed.
5134:
5135: @item exception abort sequence of @code{ABORT"}:
5136: @cindex exception abort sequence of @code{ABORT"}
5137: @cindex @code{ABORT"}, exception abort sequence
5138: The error string is stored into the variable @code{"error} and a
5139: @code{-2 throw} is performed.
5140:
5141: @item input line terminator:
5142: @cindex input line terminator
5143: @cindex line terminator on input
5144: @cindex newline charcter on input
5145: For interactive input, @kbd{C-m} (CR) and @kbd{C-j} (LF) terminate
5146: lines. One of these characters is typically produced when you type the
5147: @kbd{Enter} or @kbd{Return} key.
5148:
5149: @item maximum size of a counted string:
5150: @cindex maximum size of a counted string
5151: @cindex counted string, maximum size
5152: @code{s" /counted-string" environment? drop .}. Currently 255 characters
5153: on all ports, but this may change.
5154:
5155: @item maximum size of a parsed string:
5156: @cindex maximum size of a parsed string
5157: @cindex parsed string, maximum size
5158: Given by the constant @code{/line}. Currently 255 characters.
5159:
5160: @item maximum size of a definition name, in characters:
5161: @cindex maximum size of a definition name, in characters
5162: @cindex name, maximum length
5163: 31
5164:
5165: @item maximum string length for @code{ENVIRONMENT?}, in characters:
5166: @cindex maximum string length for @code{ENVIRONMENT?}, in characters
5167: @cindex @code{ENVIRONMENT?} string length, maximum
5168: 31
5169:
5170: @item method of selecting the user input device:
5171: @cindex user input device, method of selecting
5172: The user input device is the standard input. There is currently no way to
5173: change it from within Gforth. However, the input can typically be
5174: redirected in the command line that starts Gforth.
5175:
5176: @item method of selecting the user output device:
5177: @cindex user output device, method of selecting
5178: @code{EMIT} and @code{TYPE} output to the file-id stored in the value
5179: @code{outfile-id} (@code{stdout} by default). Gforth uses unbuffered
5180: output when the user output device is a terminal, otherwise the output
5181: is buffered.
5182:
5183: @item methods of dictionary compilation:
5184: What are we expected to document here?
5185:
5186: @item number of bits in one address unit:
5187: @cindex number of bits in one address unit
5188: @cindex address unit, size in bits
5189: @code{s" address-units-bits" environment? drop .}. 8 in all current
5190: ports.
5191:
5192: @item number representation and arithmetic:
5193: @cindex number representation and arithmetic
5194: Processor-dependent. Binary two's complement on all current ports.
5195:
5196: @item ranges for integer types:
5197: @cindex ranges for integer types
5198: @cindex integer types, ranges
5199: Installation-dependent. Make environmental queries for @code{MAX-N},
5200: @code{MAX-U}, @code{MAX-D} and @code{MAX-UD}. The lower bounds for
5201: unsigned (and positive) types is 0. The lower bound for signed types on
5202: two's complement and one's complement machines machines can be computed
5203: by adding 1 to the upper bound.
5204:
5205: @item read-only data space regions:
5206: @cindex read-only data space regions
5207: @cindex data-space, read-only regions
5208: The whole Forth data space is writable.
5209:
5210: @item size of buffer at @code{WORD}:
5211: @cindex size of buffer at @code{WORD}
5212: @cindex @code{WORD} buffer size
5213: @code{PAD HERE - .}. 104 characters on 32-bit machines. The buffer is
5214: shared with the pictured numeric output string. If overwriting
5215: @code{PAD} is acceptable, it is as large as the remaining dictionary
5216: space, although only as much can be sensibly used as fits in a counted
5217: string.
5218:
5219: @item size of one cell in address units:
5220: @cindex cell size
5221: @code{1 cells .}.
5222:
5223: @item size of one character in address units:
5224: @cindex char size
5225: @code{1 chars .}. 1 on all current ports.
5226:
5227: @item size of the keyboard terminal buffer:
5228: @cindex size of the keyboard terminal buffer
5229: @cindex terminal buffer, size
5230: Varies. You can determine the size at a specific time using @code{lp@@
5231: tib - .}. It is shared with the locals stack and TIBs of files that
5232: include the current file. You can change the amount of space for TIBs
5233: and locals stack at Gforth startup with the command line option
5234: @code{-l}.
5235:
5236: @item size of the pictured numeric output buffer:
5237: @cindex size of the pictured numeric output buffer
5238: @cindex pictured numeric output buffer, size
5239: @code{PAD HERE - .}. 104 characters on 32-bit machines. The buffer is
5240: shared with @code{WORD}.
5241:
5242: @item size of the scratch area returned by @code{PAD}:
5243: @cindex size of the scratch area returned by @code{PAD}
5244: @cindex @code{PAD} size
5245: The remainder of dictionary space. @code{unused pad here - - .}.
5246:
5247: @item system case-sensitivity characteristics:
5248: @cindex case-sensitivity characteristics
5249: Dictionary searches are case insensitive (except in
5250: @code{TABLE}s). However, as explained above under @i{character-set
5251: extensions}, the matching for non-ASCII characters is determined by the
5252: locale you are using. In the default @code{C} locale all non-ASCII
5253: characters are matched case-sensitively.
5254:
5255: @item system prompt:
5256: @cindex system prompt
5257: @cindex prompt
5258: @code{ ok} in interpret state, @code{ compiled} in compile state.
5259:
5260: @item division rounding:
5261: @cindex division rounding
5262: installation dependent. @code{s" floored" environment? drop .}. We leave
5263: the choice to @code{gcc} (what to use for @code{/}) and to you (whether
5264: to use @code{fm/mod}, @code{sm/rem} or simply @code{/}).
5265:
5266: @item values of @code{STATE} when true:
5267: @cindex @code{STATE} values
5268: -1.
5269:
5270: @item values returned after arithmetic overflow:
5271: On two's complement machines, arithmetic is performed modulo
5272: 2**bits-per-cell for single arithmetic and 4**bits-per-cell for double
5273: arithmetic (with appropriate mapping for signed types). Division by zero
5274: typically results in a @code{-55 throw} (Floating-point unidentified
5275: fault), although a @code{-10 throw} (divide by zero) would be more
5276: appropriate.
5277:
5278: @item whether the current definition can be found after @t{DOES>}:
5279: @cindex @t{DOES>}, visibility of current definition
5280: No.
5281:
5282: @end table
5283:
5284: @c ---------------------------------------------------------------------
5285: @node core-ambcond, core-other, core-idef, The Core Words
5286: @subsection Ambiguous conditions
5287: @c ---------------------------------------------------------------------
5288: @cindex core words, ambiguous conditions
5289: @cindex ambiguous conditions, core words
5290:
5291: @table @i
5292:
5293: @item a name is neither a word nor a number:
5294: @cindex name not found
5295: @cindex Undefined word
5296: @code{-13 throw} (Undefined word). Actually, @code{-13 bounce}, which
5297: preserves the data and FP stack, so you don't lose more work than
5298: necessary.
5299:
5300: @item a definition name exceeds the maximum length allowed:
5301: @cindex Word name too long
5302: @code{-19 throw} (Word name too long)
5303:
5304: @item addressing a region not inside the various data spaces of the forth system:
5305: @cindex Invalid memory address
5306: The stacks, code space and name space are accessible. Machine code space is
5307: typically readable. Accessing other addresses gives results dependent on
5308: the operating system. On decent systems: @code{-9 throw} (Invalid memory
5309: address).
5310:
5311: @item argument type incompatible with parameter:
5312: @cindex Argument type mismatch
5313: This is usually not caught. Some words perform checks, e.g., the control
5314: flow words, and issue a @code{ABORT"} or @code{-12 THROW} (Argument type
5315: mismatch).
5316:
5317: @item attempting to obtain the execution token of a word with undefined execution semantics:
5318: @cindex Interpreting a compile-only word, for @code{'} etc.
5319: @cindex execution token of words with undefined execution semantics
5320: @code{-14 throw} (Interpreting a compile-only word). In some cases, you
5321: get an execution token for @code{compile-only-error} (which performs a
5322: @code{-14 throw} when executed).
5323:
5324: @item dividing by zero:
5325: @cindex dividing by zero
5326: @cindex floating point unidentified fault, integer division
5327: @cindex divide by zero
5328: typically results in a @code{-55 throw} (floating point unidentified
5329: fault), although a @code{-10 throw} (divide by zero) would be more
5330: appropriate.
5331:
5332: @item insufficient data stack or return stack space:
5333: @cindex insufficient data stack or return stack space
5334: @cindex stack overflow
5335: @cindex Address alignment exception, stack overflow
5336: @cindex Invalid memory address, stack overflow
5337: Depending on the operating system, the installation, and the invocation
5338: of Gforth, this is either checked by the memory management hardware, or
5339: it is not checked. If it is checked, you typically get a @code{-9 throw}
5340: (Invalid memory address) as soon as the overflow happens. If it is not
5341: check, overflows typically result in mysterious illegal memory accesses,
5342: producing @code{-9 throw} (Invalid memory address) or @code{-23 throw}
5343: (Address alignment exception); they might also destroy the internal data
5344: structure of @code{ALLOCATE} and friends, resulting in various errors in
5345: these words.
5346:
5347: @item insufficient space for loop control parameters:
5348: @cindex insufficient space for loop control parameters
5349: like other return stack overflows.
5350:
5351: @item insufficient space in the dictionary:
5352: @cindex insufficient space in the dictionary
5353: @cindex dictionary overflow
5354: If you try to allot (either directly with @code{allot}, or indirectly
5355: with @code{,}, @code{create} etc.) more memory than available in the
5356: dictionary, you get a @code{-8 throw} (Dictionary overflow). If you try
5357: to access memory beyond the end of the dictionary, the results are
5358: similar to stack overflows.
5359:
5360: @item interpreting a word with undefined interpretation semantics:
5361: @cindex interpreting a word with undefined interpretation semantics
5362: @cindex Interpreting a compile-only word
5363: For some words, we have defined interpretation semantics. For the
5364: others: @code{-14 throw} (Interpreting a compile-only word).
5365:
5366: @item modifying the contents of the input buffer or a string literal:
5367: @cindex modifying the contents of the input buffer or a string literal
5368: These are located in writable memory and can be modified.
5369:
5370: @item overflow of the pictured numeric output string:
5371: @cindex overflow of the pictured numeric output string
5372: @cindex pictured numeric output string, overflow
5373: Not checked. Runs into the dictionary and destroys it (at least,
5374: partially).
5375:
5376: @item parsed string overflow:
5377: @cindex parsed string overflow
5378: @code{PARSE} cannot overflow. @code{WORD} does not check for overflow.
5379:
5380: @item producing a result out of range:
5381: @cindex result out of range
5382: On two's complement machines, arithmetic is performed modulo
5383: 2**bits-per-cell for single arithmetic and 4**bits-per-cell for double
5384: arithmetic (with appropriate mapping for signed types). Division by zero
5385: typically results in a @code{-55 throw} (floatingpoint unidentified
5386: fault), although a @code{-10 throw} (divide by zero) would be more
5387: appropriate. @code{convert} and @code{>number} currently overflow
5388: silently.
5389:
5390: @item reading from an empty data or return stack:
5391: @cindex stack empty
5392: @cindex stack underflow
5393: The data stack is checked by the outer (aka text) interpreter after
5394: every word executed. If it has underflowed, a @code{-4 throw} (Stack
5395: underflow) is performed. Apart from that, stacks may be checked or not,
5396: depending on operating system, installation, and invocation. The
5397: consequences of stack underflows are similar to the consequences of
5398: stack overflows. Note that even if the system uses checking (through the
5399: MMU), your program may have to underflow by a significant number of
5400: stack items to trigger the reaction (the reason for this is that the
5401: MMU, and therefore the checking, works with a page-size granularity).
5402:
5403: @item unexpected end of the input buffer, resulting in an attempt to use a zero-length string as a name:
5404: @cindex unexpected end of the input buffer
5405: @cindex zero-length string as a name
5406: @cindex Attempt to use zero-length string as a name
5407: @code{Create} and its descendants perform a @code{-16 throw} (Attempt to
5408: use zero-length string as a name). Words like @code{'} probably will not
5409: find what they search. Note that it is possible to create zero-length
5410: names with @code{nextname} (should it not?).
5411:
5412: @item @code{>IN} greater than input buffer:
5413: @cindex @code{>IN} greater than input buffer
5414: The next invocation of a parsing word returns a string with length 0.
5415:
5416: @item @code{RECURSE} appears after @code{DOES>}:
5417: @cindex @code{RECURSE} appears after @code{DOES>}
5418: Compiles a recursive call to the defining word, not to the defined word.
5419:
5420: @item argument input source different than current input source for @code{RESTORE-INPUT}:
5421: @cindex argument input source different than current input source for @code{RESTORE-INPUT}
5422: @cindex Argument type mismatch, @code{RESTORE-INPUT}
5423: @cindex @code{RESTORE-INPUT}, Argument type mismatch
5424: @code{-12 THROW}. Note that, once an input file is closed (e.g., because
5425: the end of the file was reached), its source-id may be
5426: reused. Therefore, restoring an input source specification referencing a
5427: closed file may lead to unpredictable results instead of a @code{-12
5428: THROW}.
5429:
5430: In the future, Gforth may be able to restore input source specifications
5431: from other than the current input source.
5432:
5433: @item data space containing definitions gets de-allocated:
5434: @cindex data space containing definitions gets de-allocated
5435: Deallocation with @code{allot} is not checked. This typically results in
5436: memory access faults or execution of illegal instructions.
5437:
5438: @item data space read/write with incorrect alignment:
5439: @cindex data space read/write with incorrect alignment
5440: @cindex alignment faults
5441: @cindex Address alignment exception
5442: Processor-dependent. Typically results in a @code{-23 throw} (Address
5443: alignment exception). Under Linux-Intel on a 486 or later processor with
5444: alignment turned on, incorrect alignment results in a @code{-9 throw}
5445: (Invalid memory address). There are reportedly some processors with
5446: alignment restrictions that do not report violations.
5447:
5448: @item data space pointer not properly aligned, @code{,}, @code{C,}:
5449: @cindex data space pointer not properly aligned, @code{,}, @code{C,}
5450: Like other alignment errors.
5451:
5452: @item less than u+2 stack items (@code{PICK} and @code{ROLL}):
5453: Like other stack underflows.
5454:
5455: @item loop control parameters not available:
5456: @cindex loop control parameters not available
5457: Not checked. The counted loop words simply assume that the top of return
5458: stack items are loop control parameters and behave accordingly.
5459:
5460: @item most recent definition does not have a name (@code{IMMEDIATE}):
5461: @cindex most recent definition does not have a name (@code{IMMEDIATE})
5462: @cindex last word was headerless
5463: @code{abort" last word was headerless"}.
5464:
5465: @item name not defined by @code{VALUE} used by @code{TO}:
5466: @cindex name not defined by @code{VALUE} used by @code{TO}
5467: @cindex @code{TO} on non-@code{VALUE}s
5468: @cindex Invalid name argument, @code{TO}
5469: @code{-32 throw} (Invalid name argument) (unless name is a local or was
5470: defined by @code{CONSTANT}; in the latter case it just changes the constant).
5471:
5472: @item name not found (@code{'}, @code{POSTPONE}, @code{[']}, @code{[COMPILE]}):
5473: @cindex name not found (@code{'}, @code{POSTPONE}, @code{[']}, @code{[COMPILE]})
5474: @cindex Undefined word, @code{'}, @code{POSTPONE}, @code{[']}, @code{[COMPILE]}
5475: @code{-13 throw} (Undefined word)
5476:
5477: @item parameters are not of the same type (@code{DO}, @code{?DO}, @code{WITHIN}):
5478: @cindex parameters are not of the same type (@code{DO}, @code{?DO}, @code{WITHIN})
5479: Gforth behaves as if they were of the same type. I.e., you can predict
5480: the behaviour by interpreting all parameters as, e.g., signed.
5481:
5482: @item @code{POSTPONE} or @code{[COMPILE]} applied to @code{TO}:
5483: @cindex @code{POSTPONE} or @code{[COMPILE]} applied to @code{TO}
5484: Assume @code{: X POSTPONE TO ; IMMEDIATE}. @code{X} performs the
5485: compilation semantics of @code{TO}.
5486:
5487: @item String longer than a counted string returned by @code{WORD}:
5488: @cindex String longer than a counted string returned by @code{WORD}
5489: @cindex @code{WORD}, string overflow
5490: Not checked. The string will be ok, but the count will, of course,
5491: contain only the least significant bits of the length.
5492:
5493: @item u greater than or equal to the number of bits in a cell (@code{LSHIFT}, @code{RSHIFT}):
5494: @cindex @code{LSHIFT}, large shift counts
5495: @cindex @code{RSHIFT}, large shift counts
5496: Processor-dependent. Typical behaviours are returning 0 and using only
5497: the low bits of the shift count.
5498:
5499: @item word not defined via @code{CREATE}:
5500: @cindex @code{>BODY} of non-@code{CREATE}d words
5501: @code{>BODY} produces the PFA of the word no matter how it was defined.
5502:
5503: @cindex @code{DOES>} of non-@code{CREATE}d words
5504: @code{DOES>} changes the execution semantics of the last defined word no
5505: matter how it was defined. E.g., @code{CONSTANT DOES>} is equivalent to
5506: @code{CREATE , DOES>}.
5507:
5508: @item words improperly used outside @code{<#} and @code{#>}:
5509: Not checked. As usual, you can expect memory faults.
5510:
5511: @end table
5512:
5513:
5514: @c ---------------------------------------------------------------------
5515: @node core-other, , core-ambcond, The Core Words
5516: @subsection Other system documentation
5517: @c ---------------------------------------------------------------------
5518: @cindex other system documentation, core words
5519: @cindex core words, other system documentation
5520:
5521: @table @i
5522: @item nonstandard words using @code{PAD}:
5523: @cindex @code{PAD} use by nonstandard words
5524: None.
5525:
5526: @item operator's terminal facilities available:
5527: @cindex operator's terminal facilities available
5528: After processing the command line, Gforth goes into interactive mode,
5529: and you can give commands to Gforth interactively. The actual facilities
5530: available depend on how you invoke Gforth.
5531:
5532: @item program data space available:
5533: @cindex program data space available
5534: @cindex data space available
5535: @code{UNUSED .} gives the remaining dictionary space. The total
5536: dictionary space can be specified with the @code{-m} switch
5537: (@pxref{Invoking Gforth}) when Gforth starts up.
5538:
5539: @item return stack space available:
5540: @cindex return stack space available
5541: You can compute the total return stack space in cells with
5542: @code{s" RETURN-STACK-CELLS" environment? drop .}. You can specify it at
5543: startup time with the @code{-r} switch (@pxref{Invoking Gforth}).
5544:
5545: @item stack space available:
5546: @cindex stack space available
5547: You can compute the total data stack space in cells with
5548: @code{s" STACK-CELLS" environment? drop .}. You can specify it at
5549: startup time with the @code{-d} switch (@pxref{Invoking Gforth}).
5550:
5551: @item system dictionary space required, in address units:
5552: @cindex system dictionary space required, in address units
5553: Type @code{here forthstart - .} after startup. At the time of this
5554: writing, this gives 80080 (bytes) on a 32-bit system.
5555: @end table
5556:
5557:
5558: @c =====================================================================
5559: @node The optional Block word set, The optional Double Number word set, The Core Words, ANS conformance
5560: @section The optional Block word set
5561: @c =====================================================================
5562: @cindex system documentation, block words
5563: @cindex block words, system documentation
5564:
5565: @menu
5566: * block-idef:: Implementation Defined Options
5567: * block-ambcond:: Ambiguous Conditions
5568: * block-other:: Other System Documentation
5569: @end menu
5570:
5571:
5572: @c ---------------------------------------------------------------------
5573: @node block-idef, block-ambcond, The optional Block word set, The optional Block word set
5574: @subsection Implementation Defined Options
5575: @c ---------------------------------------------------------------------
5576: @cindex implementation-defined options, block words
5577: @cindex block words, implementation-defined options
5578:
5579: @table @i
5580: @item the format for display by @code{LIST}:
5581: @cindex @code{LIST} display format
5582: First the screen number is displayed, then 16 lines of 64 characters,
5583: each line preceded by the line number.
5584:
5585: @item the length of a line affected by @code{\}:
5586: @cindex length of a line affected by @code{\}
5587: @cindex @code{\}, line length in blocks
5588: 64 characters.
5589: @end table
5590:
5591:
5592: @c ---------------------------------------------------------------------
5593: @node block-ambcond, block-other, block-idef, The optional Block word set
5594: @subsection Ambiguous conditions
5595: @c ---------------------------------------------------------------------
5596: @cindex block words, ambiguous conditions
5597: @cindex ambiguous conditions, block words
5598:
5599: @table @i
5600: @item correct block read was not possible:
5601: @cindex block read not possible
5602: Typically results in a @code{throw} of some OS-derived value (between
5603: -512 and -2048). If the blocks file was just not long enough, blanks are
5604: supplied for the missing portion.
5605:
5606: @item I/O exception in block transfer:
5607: @cindex I/O exception in block transfer
5608: @cindex block transfer, I/O exception
5609: Typically results in a @code{throw} of some OS-derived value (between
5610: -512 and -2048).
5611:
5612: @item invalid block number:
5613: @cindex invalid block number
5614: @cindex block number invalid
5615: @code{-35 throw} (Invalid block number)
5616:
5617: @item a program directly alters the contents of @code{BLK}:
5618: @cindex @code{BLK}, altering @code{BLK}
5619: The input stream is switched to that other block, at the same
5620: position. If the storing to @code{BLK} happens when interpreting
5621: non-block input, the system will get quite confused when the block ends.
5622:
5623: @item no current block buffer for @code{UPDATE}:
5624: @cindex @code{UPDATE}, no current block buffer
5625: @code{UPDATE} has no effect.
5626:
5627: @end table
5628:
5629: @c ---------------------------------------------------------------------
5630: @node block-other, , block-ambcond, The optional Block word set
5631: @subsection Other system documentation
5632: @c ---------------------------------------------------------------------
5633: @cindex other system documentation, block words
5634: @cindex block words, other system documentation
5635:
5636: @table @i
5637: @item any restrictions a multiprogramming system places on the use of buffer addresses:
5638: No restrictions (yet).
5639:
5640: @item the number of blocks available for source and data:
5641: depends on your disk space.
5642:
5643: @end table
5644:
5645:
5646: @c =====================================================================
5647: @node The optional Double Number word set, The optional Exception word set, The optional Block word set, ANS conformance
5648: @section The optional Double Number word set
5649: @c =====================================================================
5650: @cindex system documentation, double words
5651: @cindex double words, system documentation
5652:
5653: @menu
5654: * double-ambcond:: Ambiguous Conditions
5655: @end menu
5656:
5657:
5658: @c ---------------------------------------------------------------------
5659: @node double-ambcond, , The optional Double Number word set, The optional Double Number word set
5660: @subsection Ambiguous conditions
5661: @c ---------------------------------------------------------------------
5662: @cindex double words, ambiguous conditions
5663: @cindex ambiguous conditions, double words
5664:
5665: @table @i
5666: @item @var{d} outside of range of @var{n} in @code{D>S}:
5667: @cindex @code{D>S}, @var{d} out of range of @var{n}
5668: The least significant cell of @var{d} is produced.
5669:
5670: @end table
5671:
5672:
5673: @c =====================================================================
5674: @node The optional Exception word set, The optional Facility word set, The optional Double Number word set, ANS conformance
5675: @section The optional Exception word set
5676: @c =====================================================================
5677: @cindex system documentation, exception words
5678: @cindex exception words, system documentation
5679:
5680: @menu
5681: * exception-idef:: Implementation Defined Options
5682: @end menu
5683:
5684:
5685: @c ---------------------------------------------------------------------
5686: @node exception-idef, , The optional Exception word set, The optional Exception word set
5687: @subsection Implementation Defined Options
5688: @c ---------------------------------------------------------------------
5689: @cindex implementation-defined options, exception words
5690: @cindex exception words, implementation-defined options
5691:
5692: @table @i
5693: @item @code{THROW}-codes used in the system:
5694: @cindex @code{THROW}-codes used in the system
5695: The codes -256@minus{}-511 are used for reporting signals. The mapping
5696: from OS signal numbers to throw codes is -256@minus{}@var{signal}. The
5697: codes -512@minus{}-2047 are used for OS errors (for file and memory
5698: allocation operations). The mapping from OS error numbers to throw codes
5699: is -512@minus{}@code{errno}. One side effect of this mapping is that
5700: undefined OS errors produce a message with a strange number; e.g.,
5701: @code{-1000 THROW} results in @code{Unknown error 488} on my system.
5702: @end table
5703:
5704: @c =====================================================================
5705: @node The optional Facility word set, The optional File-Access word set, The optional Exception word set, ANS conformance
5706: @section The optional Facility word set
5707: @c =====================================================================
5708: @cindex system documentation, facility words
5709: @cindex facility words, system documentation
5710:
5711: @menu
5712: * facility-idef:: Implementation Defined Options
5713: * facility-ambcond:: Ambiguous Conditions
5714: @end menu
5715:
5716:
5717: @c ---------------------------------------------------------------------
5718: @node facility-idef, facility-ambcond, The optional Facility word set, The optional Facility word set
5719: @subsection Implementation Defined Options
5720: @c ---------------------------------------------------------------------
5721: @cindex implementation-defined options, facility words
5722: @cindex facility words, implementation-defined options
5723:
5724: @table @i
5725: @item encoding of keyboard events (@code{EKEY}):
5726: @cindex keyboard events, encoding in @code{EKEY}
5727: @cindex @code{EKEY}, encoding of keyboard events
5728: Not yet implemented.
5729:
5730: @item duration of a system clock tick:
5731: @cindex duration of a system clock tick
5732: @cindex clock tick duration
5733: System dependent. With respect to @code{MS}, the time is specified in
5734: microseconds. How well the OS and the hardware implement this, is
5735: another question.
5736:
5737: @item repeatability to be expected from the execution of @code{MS}:
5738: @cindex repeatability to be expected from the execution of @code{MS}
5739: @cindex @code{MS}, repeatability to be expected
5740: System dependent. On Unix, a lot depends on load. If the system is
5741: lightly loaded, and the delay is short enough that Gforth does not get
5742: swapped out, the performance should be acceptable. Under MS-DOS and
5743: other single-tasking systems, it should be good.
5744:
5745: @end table
5746:
5747:
5748: @c ---------------------------------------------------------------------
5749: @node facility-ambcond, , facility-idef, The optional Facility word set
5750: @subsection Ambiguous conditions
5751: @c ---------------------------------------------------------------------
5752: @cindex facility words, ambiguous conditions
5753: @cindex ambiguous conditions, facility words
5754:
5755: @table @i
5756: @item @code{AT-XY} can't be performed on user output device:
5757: @cindex @code{AT-XY} can't be performed on user output device
5758: Largely terminal dependent. No range checks are done on the arguments.
5759: No errors are reported. You may see some garbage appearing, you may see
5760: simply nothing happen.
5761:
5762: @end table
5763:
5764:
5765: @c =====================================================================
5766: @node The optional File-Access word set, The optional Floating-Point word set, The optional Facility word set, ANS conformance
5767: @section The optional File-Access word set
5768: @c =====================================================================
5769: @cindex system documentation, file words
5770: @cindex file words, system documentation
5771:
5772: @menu
5773: * file-idef:: Implementation Defined Options
5774: * file-ambcond:: Ambiguous Conditions
5775: @end menu
5776:
5777: @c ---------------------------------------------------------------------
5778: @node file-idef, file-ambcond, The optional File-Access word set, The optional File-Access word set
5779: @subsection Implementation Defined Options
5780: @c ---------------------------------------------------------------------
5781: @cindex implementation-defined options, file words
5782: @cindex file words, implementation-defined options
5783:
5784: @table @i
5785: @item file access methods used:
5786: @cindex file access methods used
5787: @code{R/O}, @code{R/W} and @code{BIN} work as you would
5788: expect. @code{W/O} translates into the C file opening mode @code{w} (or
5789: @code{wb}): The file is cleared, if it exists, and created, if it does
5790: not (with both @code{open-file} and @code{create-file}). Under Unix
5791: @code{create-file} creates a file with 666 permissions modified by your
5792: umask.
5793:
5794: @item file exceptions:
5795: @cindex file exceptions
5796: The file words do not raise exceptions (except, perhaps, memory access
5797: faults when you pass illegal addresses or file-ids).
5798:
5799: @item file line terminator:
5800: @cindex file line terminator
5801: System-dependent. Gforth uses C's newline character as line
5802: terminator. What the actual character code(s) of this are is
5803: system-dependent.
5804:
5805: @item file name format:
5806: @cindex file name format
5807: System dependent. Gforth just uses the file name format of your OS.
5808:
5809: @item information returned by @code{FILE-STATUS}:
5810: @cindex @code{FILE-STATUS}, returned information
5811: @code{FILE-STATUS} returns the most powerful file access mode allowed
5812: for the file: Either @code{R/O}, @code{W/O} or @code{R/W}. If the file
5813: cannot be accessed, @code{R/O BIN} is returned. @code{BIN} is applicable
5814: along with the returned mode.
5815:
5816: @item input file state after an exception when including source:
5817: @cindex exception when including source
5818: All files that are left via the exception are closed.
5819:
5820: @item @var{ior} values and meaning:
5821: @cindex @var{ior} values and meaning
5822: The @var{ior}s returned by the file and memory allocation words are
5823: intended as throw codes. They typically are in the range
5824: -512@minus{}-2047 of OS errors. The mapping from OS error numbers to
5825: @var{ior}s is -512@minus{}@var{errno}.
5826:
5827: @item maximum depth of file input nesting:
5828: @cindex maximum depth of file input nesting
5829: @cindex file input nesting, maximum depth
5830: limited by the amount of return stack, locals/TIB stack, and the number
5831: of open files available. This should not give you troubles.
5832:
5833: @item maximum size of input line:
5834: @cindex maximum size of input line
5835: @cindex input line size, maximum
5836: @code{/line}. Currently 255.
5837:
5838: @item methods of mapping block ranges to files:
5839: @cindex mapping block ranges to files
5840: @cindex files containing blocks
5841: @cindex blocks in files
5842: By default, blocks are accessed in the file @file{blocks.fb} in the
5843: current working directory. The file can be switched with @code{USE}.
5844:
5845: @item number of string buffers provided by @code{S"}:
5846: @cindex @code{S"}, number of string buffers
5847: 1
5848:
5849: @item size of string buffer used by @code{S"}:
5850: @cindex @code{S"}, size of string buffer
5851: @code{/line}. currently 255.
5852:
5853: @end table
5854:
5855: @c ---------------------------------------------------------------------
5856: @node file-ambcond, , file-idef, The optional File-Access word set
5857: @subsection Ambiguous conditions
5858: @c ---------------------------------------------------------------------
5859: @cindex file words, ambiguous conditions
5860: @cindex ambiguous conditions, file words
5861:
5862: @table @i
5863: @item attempting to position a file outside its boundaries:
5864: @cindex @code{REPOSITION-FILE}, outside the file's boundaries
5865: @code{REPOSITION-FILE} is performed as usual: Afterwards,
5866: @code{FILE-POSITION} returns the value given to @code{REPOSITION-FILE}.
5867:
5868: @item attempting to read from file positions not yet written:
5869: @cindex reading from file positions not yet written
5870: End-of-file, i.e., zero characters are read and no error is reported.
5871:
5872: @item @var{file-id} is invalid (@code{INCLUDE-FILE}):
5873: @cindex @code{INCLUDE-FILE}, @var{file-id} is invalid
5874: An appropriate exception may be thrown, but a memory fault or other
5875: problem is more probable.
5876:
5877: @item I/O exception reading or closing @var{file-id} (@code{INCLUDE-FILE}, @code{INCLUDED}):
5878: @cindex @code{INCLUDE-FILE}, I/O exception reading or closing @var{file-id}
5879: @cindex @code{INCLUDED}, I/O exception reading or closing @var{file-id}
5880: The @var{ior} produced by the operation, that discovered the problem, is
5881: thrown.
5882:
5883: @item named file cannot be opened (@code{INCLUDED}):
5884: @cindex @code{INCLUDED}, named file cannot be opened
5885: The @var{ior} produced by @code{open-file} is thrown.
5886:
5887: @item requesting an unmapped block number:
5888: @cindex unmapped block numbers
5889: There are no unmapped legal block numbers. On some operating systems,
5890: writing a block with a large number may overflow the file system and
5891: have an error message as consequence.
5892:
5893: @item using @code{source-id} when @code{blk} is non-zero:
5894: @cindex @code{SOURCE-ID}, behaviour when @code{BLK} is non-zero
5895: @code{source-id} performs its function. Typically it will give the id of
5896: the source which loaded the block. (Better ideas?)
5897:
5898: @end table
5899:
5900:
5901: @c =====================================================================
5902: @node The optional Floating-Point word set, The optional Locals word set, The optional File-Access word set, ANS conformance
5903: @section The optional Floating-Point word set
5904: @c =====================================================================
5905: @cindex system documentation, floating-point words
5906: @cindex floating-point words, system documentation
5907:
5908: @menu
5909: * floating-idef:: Implementation Defined Options
5910: * floating-ambcond:: Ambiguous Conditions
5911: @end menu
5912:
5913:
5914: @c ---------------------------------------------------------------------
5915: @node floating-idef, floating-ambcond, The optional Floating-Point word set, The optional Floating-Point word set
5916: @subsection Implementation Defined Options
5917: @c ---------------------------------------------------------------------
5918: @cindex implementation-defined options, floating-point words
5919: @cindex floating-point words, implementation-defined options
5920:
5921: @table @i
5922: @item format and range of floating point numbers:
5923: @cindex format and range of floating point numbers
5924: @cindex floating point numbers, format and range
5925: System-dependent; the @code{double} type of C.
5926:
5927: @item results of @code{REPRESENT} when @var{float} is out of range:
5928: @cindex @code{REPRESENT}, results when @var{float} is out of range
5929: System dependent; @code{REPRESENT} is implemented using the C library
5930: function @code{ecvt()} and inherits its behaviour in this respect.
5931:
5932: @item rounding or truncation of floating-point numbers:
5933: @cindex rounding of floating-point numbers
5934: @cindex truncation of floating-point numbers
5935: @cindex floating-point numbers, rounding or truncation
5936: System dependent; the rounding behaviour is inherited from the hosting C
5937: compiler. IEEE-FP-based (i.e., most) systems by default round to
5938: nearest, and break ties by rounding to even (i.e., such that the last
5939: bit of the mantissa is 0).
5940:
5941: @item size of floating-point stack:
5942: @cindex floating-point stack size
5943: @code{s" FLOATING-STACK" environment? drop .} gives the total size of
5944: the floating-point stack (in floats). You can specify this on startup
5945: with the command-line option @code{-f} (@pxref{Invoking Gforth}).
5946:
5947: @item width of floating-point stack:
5948: @cindex floating-point stack width
5949: @code{1 floats}.
5950:
5951: @end table
5952:
5953:
5954: @c ---------------------------------------------------------------------
5955: @node floating-ambcond, , floating-idef, The optional Floating-Point word set
5956: @subsection Ambiguous conditions
5957: @c ---------------------------------------------------------------------
5958: @cindex floating-point words, ambiguous conditions
5959: @cindex ambiguous conditions, floating-point words
5960:
5961: @table @i
5962: @item @code{df@@} or @code{df!} used with an address that is not double-float aligned:
5963: @cindex @code{df@@} or @code{df!} used with an address that is not double-float aligned
5964: System-dependent. Typically results in a @code{-23 THROW} like other
5965: alignment violations.
5966:
5967: @item @code{f@@} or @code{f!} used with an address that is not float aligned:
5968: @cindex @code{f@@} used with an address that is not float aligned
5969: @cindex @code{f!} used with an address that is not float aligned
5970: System-dependent. Typically results in a @code{-23 THROW} like other
5971: alignment violations.
5972:
5973: @item floating-point result out of range:
5974: @cindex floating-point result out of range
5975: System-dependent. Can result in a @code{-55 THROW} (Floating-point
5976: unidentified fault), or can produce a special value representing, e.g.,
5977: Infinity.
5978:
5979: @item @code{sf@@} or @code{sf!} used with an address that is not single-float aligned:
5980: @cindex @code{sf@@} or @code{sf!} used with an address that is not single-float aligned
5981: System-dependent. Typically results in an alignment fault like other
5982: alignment violations.
5983:
5984: @item @code{BASE} is not decimal (@code{REPRESENT}, @code{F.}, @code{FE.}, @code{FS.}):
5985: @cindex @code{BASE} is not decimal (@code{REPRESENT}, @code{F.}, @code{FE.}, @code{FS.})
5986: The floating-point number is converted into decimal nonetheless.
5987:
5988: @item Both arguments are equal to zero (@code{FATAN2}):
5989: @cindex @code{FATAN2}, both arguments are equal to zero
5990: System-dependent. @code{FATAN2} is implemented using the C library
5991: function @code{atan2()}.
5992:
5993: @item Using @code{FTAN} on an argument @var{r1} where cos(@var{r1}) is zero:
5994: @cindex @code{FTAN} on an argument @var{r1} where cos(@var{r1}) is zero
5995: System-dependent. Anyway, typically the cos of @var{r1} will not be zero
5996: because of small errors and the tan will be a very large (or very small)
5997: but finite number.
5998:
5999: @item @var{d} cannot be presented precisely as a float in @code{D>F}:
6000: @cindex @code{D>F}, @var{d} cannot be presented precisely as a float
6001: The result is rounded to the nearest float.
6002:
6003: @item dividing by zero:
6004: @cindex dividing by zero, floating-point
6005: @cindex floating-point dividing by zero
6006: @cindex floating-point unidentified fault, FP divide-by-zero
6007: @code{-55 throw} (Floating-point unidentified fault)
6008:
6009: @item exponent too big for conversion (@code{DF!}, @code{DF@@}, @code{SF!}, @code{SF@@}):
6010: @cindex exponent too big for conversion (@code{DF!}, @code{DF@@}, @code{SF!}, @code{SF@@})
6011: System dependent. On IEEE-FP based systems the number is converted into
6012: an infinity.
6013:
6014: @item @var{float}<1 (@code{FACOSH}):
6015: @cindex @code{FACOSH}, @var{float}<1
6016: @cindex floating-point unidentified fault, @code{FACOSH}
6017: @code{-55 throw} (Floating-point unidentified fault)
6018:
6019: @item @var{float}=<-1 (@code{FLNP1}):
6020: @cindex @code{FLNP1}, @var{float}=<-1
6021: @cindex floating-point unidentified fault, @code{FLNP1}
6022: @code{-55 throw} (Floating-point unidentified fault). On IEEE-FP systems
6023: negative infinity is typically produced for @var{float}=-1.
6024:
6025: @item @var{float}=<0 (@code{FLN}, @code{FLOG}):
6026: @cindex @code{FLN}, @var{float}=<0
6027: @cindex @code{FLOG}, @var{float}=<0
6028: @cindex floating-point unidentified fault, @code{FLN} or @code{FLOG}
6029: @code{-55 throw} (Floating-point unidentified fault). On IEEE-FP systems
6030: negative infinity is typically produced for @var{float}=0.
6031:
6032: @item @var{float}<0 (@code{FASINH}, @code{FSQRT}):
6033: @cindex @code{FASINH}, @var{float}<0
6034: @cindex @code{FSQRT}, @var{float}<0
6035: @cindex floating-point unidentified fault, @code{FASINH} or @code{FSQRT}
6036: @code{-55 throw} (Floating-point unidentified fault). @code{fasinh}
6037: produces values for these inputs on my Linux box (Bug in the C library?)
6038:
6039: @item |@var{float}|>1 (@code{FACOS}, @code{FASIN}, @code{FATANH}):
6040: @cindex @code{FACOS}, |@var{float}|>1
6041: @cindex @code{FASIN}, |@var{float}|>1
6042: @cindex @code{FATANH}, |@var{float}|>1
6043: @cindex floating-point unidentified fault, @code{FACOS}, @code{FASIN} or @code{FATANH}
6044: @code{-55 throw} (Floating-point unidentified fault).
6045:
6046: @item integer part of float cannot be represented by @var{d} in @code{F>D}:
6047: @cindex @code{F>D}, integer part of float cannot be represented by @var{d}
6048: @cindex floating-point unidentified fault, @code{F>D}
6049: @code{-55 throw} (Floating-point unidentified fault).
6050:
6051: @item string larger than pictured numeric output area (@code{f.}, @code{fe.}, @code{fs.}):
6052: @cindex string larger than pictured numeric output area (@code{f.}, @code{fe.}, @code{fs.})
6053: This does not happen.
6054: @end table
6055:
6056: @c =====================================================================
6057: @node The optional Locals word set, The optional Memory-Allocation word set, The optional Floating-Point word set, ANS conformance
6058: @section The optional Locals word set
6059: @c =====================================================================
6060: @cindex system documentation, locals words
6061: @cindex locals words, system documentation
6062:
6063: @menu
6064: * locals-idef:: Implementation Defined Options
6065: * locals-ambcond:: Ambiguous Conditions
6066: @end menu
6067:
6068:
6069: @c ---------------------------------------------------------------------
6070: @node locals-idef, locals-ambcond, The optional Locals word set, The optional Locals word set
6071: @subsection Implementation Defined Options
6072: @c ---------------------------------------------------------------------
6073: @cindex implementation-defined options, locals words
6074: @cindex locals words, implementation-defined options
6075:
6076: @table @i
6077: @item maximum number of locals in a definition:
6078: @cindex maximum number of locals in a definition
6079: @cindex locals, maximum number in a definition
6080: @code{s" #locals" environment? drop .}. Currently 15. This is a lower
6081: bound, e.g., on a 32-bit machine there can be 41 locals of up to 8
6082: characters. The number of locals in a definition is bounded by the size
6083: of locals-buffer, which contains the names of the locals.
6084:
6085: @end table
6086:
6087:
6088: @c ---------------------------------------------------------------------
6089: @node locals-ambcond, , locals-idef, The optional Locals word set
6090: @subsection Ambiguous conditions
6091: @c ---------------------------------------------------------------------
6092: @cindex locals words, ambiguous conditions
6093: @cindex ambiguous conditions, locals words
6094:
6095: @table @i
6096: @item executing a named local in interpretation state:
6097: @cindex local in interpretation state
6098: @cindex Interpreting a compile-only word, for a local
6099: Locals have no interpretation semantics. If you try to perform the
6100: interpretation semantics, you will get a @code{-14 throw} somewhere
6101: (Interpreting a compile-only word). If you perform the compilation
6102: semantics, the locals access will be compiled (irrespective of state).
6103:
6104: @item @var{name} not defined by @code{VALUE} or @code{(LOCAL)} (@code{TO}):
6105: @cindex name not defined by @code{VALUE} or @code{(LOCAL)} used by @code{TO}
6106: @cindex @code{TO} on non-@code{VALUE}s and non-locals
6107: @cindex Invalid name argument, @code{TO}
6108: @code{-32 throw} (Invalid name argument)
6109:
6110: @end table
6111:
6112:
6113: @c =====================================================================
6114: @node The optional Memory-Allocation word set, The optional Programming-Tools word set, The optional Locals word set, ANS conformance
6115: @section The optional Memory-Allocation word set
6116: @c =====================================================================
6117: @cindex system documentation, memory-allocation words
6118: @cindex memory-allocation words, system documentation
6119:
6120: @menu
6121: * memory-idef:: Implementation Defined Options
6122: @end menu
6123:
6124:
6125: @c ---------------------------------------------------------------------
6126: @node memory-idef, , The optional Memory-Allocation word set, The optional Memory-Allocation word set
6127: @subsection Implementation Defined Options
6128: @c ---------------------------------------------------------------------
6129: @cindex implementation-defined options, memory-allocation words
6130: @cindex memory-allocation words, implementation-defined options
6131:
6132: @table @i
6133: @item values and meaning of @var{ior}:
6134: @cindex @var{ior} values and meaning
6135: The @var{ior}s returned by the file and memory allocation words are
6136: intended as throw codes. They typically are in the range
6137: -512@minus{}-2047 of OS errors. The mapping from OS error numbers to
6138: @var{ior}s is -512@minus{}@var{errno}.
6139:
6140: @end table
6141:
6142: @c =====================================================================
6143: @node The optional Programming-Tools word set, The optional Search-Order word set, The optional Memory-Allocation word set, ANS conformance
6144: @section The optional Programming-Tools word set
6145: @c =====================================================================
6146: @cindex system documentation, programming-tools words
6147: @cindex programming-tools words, system documentation
6148:
6149: @menu
6150: * programming-idef:: Implementation Defined Options
6151: * programming-ambcond:: Ambiguous Conditions
6152: @end menu
6153:
6154:
6155: @c ---------------------------------------------------------------------
6156: @node programming-idef, programming-ambcond, The optional Programming-Tools word set, The optional Programming-Tools word set
6157: @subsection Implementation Defined Options
6158: @c ---------------------------------------------------------------------
6159: @cindex implementation-defined options, programming-tools words
6160: @cindex programming-tools words, implementation-defined options
6161:
6162: @table @i
6163: @item ending sequence for input following @code{;CODE} and @code{CODE}:
6164: @cindex @code{;CODE} ending sequence
6165: @cindex @code{CODE} ending sequence
6166: @code{END-CODE}
6167:
6168: @item manner of processing input following @code{;CODE} and @code{CODE}:
6169: @cindex @code{;CODE}, processing input
6170: @cindex @code{CODE}, processing input
6171: The @code{ASSEMBLER} vocabulary is pushed on the search order stack, and
6172: the input is processed by the text interpreter, (starting) in interpret
6173: state.
6174:
6175: @item search order capability for @code{EDITOR} and @code{ASSEMBLER}:
6176: @cindex @code{ASSEMBLER}, search order capability
6177: The ANS Forth search order word set.
6178:
6179: @item source and format of display by @code{SEE}:
6180: @cindex @code{SEE}, source and format of output
6181: The source for @code{see} is the intermediate code used by the inner
6182: interpreter. The current @code{see} tries to output Forth source code
6183: as well as possible.
6184:
6185: @end table
6186:
6187: @c ---------------------------------------------------------------------
6188: @node programming-ambcond, , programming-idef, The optional Programming-Tools word set
6189: @subsection Ambiguous conditions
6190: @c ---------------------------------------------------------------------
6191: @cindex programming-tools words, ambiguous conditions
6192: @cindex ambiguous conditions, programming-tools words
6193:
6194: @table @i
6195:
6196: @item deleting the compilation wordlist (@code{FORGET}):
6197: @cindex @code{FORGET}, deleting the compilation wordlist
6198: Not implemented (yet).
6199:
6200: @item fewer than @var{u}+1 items on the control flow stack (@code{CS-PICK}, @code{CS-ROLL}):
6201: @cindex @code{CS-PICK}, fewer than @var{u}+1 items on the control flow stack
6202: @cindex @code{CS-ROLL}, fewer than @var{u}+1 items on the control flow stack
6203: @cindex control-flow stack underflow
6204: This typically results in an @code{abort"} with a descriptive error
6205: message (may change into a @code{-22 throw} (Control structure mismatch)
6206: in the future). You may also get a memory access error. If you are
6207: unlucky, this ambiguous condition is not caught.
6208:
6209: @item @var{name} can't be found (@code{FORGET}):
6210: @cindex @code{FORGET}, @var{name} can't be found
6211: Not implemented (yet).
6212:
6213: @item @var{name} not defined via @code{CREATE}:
6214: @cindex @code{;CODE}, @var{name} not defined via @code{CREATE}
6215: @code{;CODE} behaves like @code{DOES>} in this respect, i.e., it changes
6216: the execution semantics of the last defined word no matter how it was
6217: defined.
6218:
6219: @item @code{POSTPONE} applied to @code{[IF]}:
6220: @cindex @code{POSTPONE} applied to @code{[IF]}
6221: @cindex @code{[IF]} and @code{POSTPONE}
6222: After defining @code{: X POSTPONE [IF] ; IMMEDIATE}. @code{X} is
6223: equivalent to @code{[IF]}.
6224:
6225: @item reaching the end of the input source before matching @code{[ELSE]} or @code{[THEN]}:
6226: @cindex @code{[IF]}, end of the input source before matching @code{[ELSE]} or @code{[THEN]}
6227: Continue in the same state of conditional compilation in the next outer
6228: input source. Currently there is no warning to the user about this.
6229:
6230: @item removing a needed definition (@code{FORGET}):
6231: @cindex @code{FORGET}, removing a needed definition
6232: Not implemented (yet).
6233:
6234: @end table
6235:
6236:
6237: @c =====================================================================
6238: @node The optional Search-Order word set, , The optional Programming-Tools word set, ANS conformance
6239: @section The optional Search-Order word set
6240: @c =====================================================================
6241: @cindex system documentation, search-order words
6242: @cindex search-order words, system documentation
6243:
6244: @menu
6245: * search-idef:: Implementation Defined Options
6246: * search-ambcond:: Ambiguous Conditions
6247: @end menu
6248:
6249:
6250: @c ---------------------------------------------------------------------
6251: @node search-idef, search-ambcond, The optional Search-Order word set, The optional Search-Order word set
6252: @subsection Implementation Defined Options
6253: @c ---------------------------------------------------------------------
6254: @cindex implementation-defined options, search-order words
6255: @cindex search-order words, implementation-defined options
6256:
6257: @table @i
6258: @item maximum number of word lists in search order:
6259: @cindex maximum number of word lists in search order
6260: @cindex search order, maximum depth
6261: @code{s" wordlists" environment? drop .}. Currently 16.
6262:
6263: @item minimum search order:
6264: @cindex minimum search order
6265: @cindex search order, minimum
6266: @code{root root}.
6267:
6268: @end table
6269:
6270: @c ---------------------------------------------------------------------
6271: @node search-ambcond, , search-idef, The optional Search-Order word set
6272: @subsection Ambiguous conditions
6273: @c ---------------------------------------------------------------------
6274: @cindex search-order words, ambiguous conditions
6275: @cindex ambiguous conditions, search-order words
6276:
6277: @table @i
6278: @item changing the compilation wordlist (during compilation):
6279: @cindex changing the compilation wordlist (during compilation)
6280: @cindex compilation wordlist, change before definition ends
6281: The word is entered into the wordlist that was the compilation wordlist
6282: at the start of the definition. Any changes to the name field (e.g.,
6283: @code{immediate}) or the code field (e.g., when executing @code{DOES>})
6284: are applied to the latest defined word (as reported by @code{last} or
6285: @code{lastxt}), if possible, irrespective of the compilation wordlist.
6286:
6287: @item search order empty (@code{previous}):
6288: @cindex @code{previous}, search order empty
6289: @cindex Vocstack empty, @code{previous}
6290: @code{abort" Vocstack empty"}.
6291:
6292: @item too many word lists in search order (@code{also}):
6293: @cindex @code{also}, too many word lists in search order
6294: @cindex Vocstack full, @code{also}
6295: @code{abort" Vocstack full"}.
6296:
6297: @end table
6298:
6299: @c ***************************************************************
6300: @node Model, Integrating Gforth, ANS conformance, Top
6301: @chapter Model
6302:
6303: This chapter has yet to be written. It will contain information, on
6304: which internal structures you can rely.
6305:
6306: @c ***************************************************************
6307: @node Integrating Gforth, Emacs and Gforth, Model, Top
6308: @chapter Integrating Gforth into C programs
6309:
6310: This is not yet implemented.
6311:
6312: Several people like to use Forth as scripting language for applications
6313: that are otherwise written in C, C++, or some other language.
6314:
6315: The Forth system ATLAST provides facilities for embedding it into
6316: applications; unfortunately it has several disadvantages: most
6317: importantly, it is not based on ANS Forth, and it is apparently dead
6318: (i.e., not developed further and not supported). The facilities
6319: provided by Gforth in this area are inspired by ATLASTs facilities, so
6320: making the switch should not be hard.
6321:
6322: We also tried to design the interface such that it can easily be
6323: implemented by other Forth systems, so that we may one day arrive at a
6324: standardized interface. Such a standard interface would allow you to
6325: replace the Forth system without having to rewrite C code.
6326:
6327: You embed the Gforth interpreter by linking with the library
6328: @code{libgforth.a} (give the compiler the option @code{-lgforth}). All
6329: global symbols in this library that belong to the interface, have the
6330: prefix @code{forth_}. (Global symbols that are used internally have the
6331: prefix @code{gforth_}).
6332:
6333: You can include the declarations of Forth types and the functions and
6334: variables of the interface with @code{#include <forth.h>}.
6335:
6336: Types.
6337:
6338: Variables.
6339:
6340: Data and FP Stack pointer. Area sizes.
6341:
6342: functions.
6343:
6344: forth_init(imagefile)
6345: forth_evaluate(string) exceptions?
6346: forth_goto(address) (or forth_execute(xt)?)
6347: forth_continue() (a corountining mechanism)
6348:
6349: Adding primitives.
6350:
6351: No checking.
6352:
6353: Signals?
6354:
6355: Accessing the Stacks
6356:
6357: @node Emacs and Gforth, Image Files, Integrating Gforth, Top
6358: @chapter Emacs and Gforth
6359: @cindex Emacs and Gforth
6360:
6361: @cindex @file{gforth.el}
6362: @cindex @file{forth.el}
6363: @cindex Rydqvist, Goran
6364: @cindex comment editing commands
6365: @cindex @code{\}, editing with Emacs
6366: @cindex debug tracer editing commands
6367: @cindex @code{~~}, removal with Emacs
6368: @cindex Forth mode in Emacs
6369: Gforth comes with @file{gforth.el}, an improved version of
6370: @file{forth.el} by Goran Rydqvist (included in the TILE package). The
6371: improvements are a better (but still not perfect) handling of
6372: indentation. I have also added comment paragraph filling (@kbd{M-q}),
6373: commenting (@kbd{C-x \}) and uncommenting (@kbd{C-u C-x \}) regions and
6374: removing debugging tracers (@kbd{C-x ~}, @pxref{Debugging}). I left the
6375: stuff I do not use alone, even though some of it only makes sense for
6376: TILE. To get a description of these features, enter Forth mode and type
6377: @kbd{C-h m}.
6378:
6379: @cindex source location of error or debugging output in Emacs
6380: @cindex error output, finding the source location in Emacs
6381: @cindex debugging output, finding the source location in Emacs
6382: In addition, Gforth supports Emacs quite well: The source code locations
6383: given in error messages, debugging output (from @code{~~}) and failed
6384: assertion messages are in the right format for Emacs' compilation mode
6385: (@pxref{Compilation, , Running Compilations under Emacs, emacs, Emacs
6386: Manual}) so the source location corresponding to an error or other
6387: message is only a few keystrokes away (@kbd{C-x `} for the next error,
6388: @kbd{C-c C-c} for the error under the cursor).
6389:
6390: @cindex @file{TAGS} file
6391: @cindex @file{etags.fs}
6392: @cindex viewing the source of a word in Emacs
6393: Also, if you @code{include} @file{etags.fs}, a new @file{TAGS} file
6394: (@pxref{Tags, , Tags Tables, emacs, Emacs Manual}) will be produced that
6395: contains the definitions of all words defined afterwards. You can then
6396: find the source for a word using @kbd{M-.}. Note that emacs can use
6397: several tags files at the same time (e.g., one for the Gforth sources
6398: and one for your program, @pxref{Select Tags Table,,Selecting a Tags
6399: Table,emacs, Emacs Manual}). The TAGS file for the preloaded words is
6400: @file{$(datadir)/gforth/$(VERSION)/TAGS} (e.g.,
6401: @file{/usr/local/share/gforth/0.2.0/TAGS}).
6402:
6403: @cindex @file{.emacs}
6404: To get all these benefits, add the following lines to your @file{.emacs}
6405: file:
6406:
6407: @example
6408: (autoload 'forth-mode "gforth.el")
6409: (setq auto-mode-alist (cons '("\\.fs\\'" . forth-mode) auto-mode-alist))
6410: @end example
6411:
6412: @node Image Files, Engine, Emacs and Gforth, Top
6413: @chapter Image Files
6414: @cindex image files
6415: @cindex @code{.fi} files
6416: @cindex precompiled Forth code
6417: @cindex dictionary in persistent form
6418: @cindex persistent form of dictionary
6419:
6420: An image file is a file containing an image of the Forth dictionary,
6421: i.e., compiled Forth code and data residing in the dictionary. By
6422: convention, we use the extension @code{.fi} for image files.
6423:
6424: @menu
6425: * Image Licensing Issues:: Distribution terms for images.
6426: * Image File Background:: Why have image files?
6427: * Non-Relocatable Image Files:: don't always work.
6428: * Data-Relocatable Image Files:: are better.
6429: * Fully Relocatable Image Files:: better yet.
6430: * Stack and Dictionary Sizes:: Setting the default sizes for an image.
6431: * Running Image Files:: @code{gforth -i @var{file}} or @var{file}.
6432: * Modifying the Startup Sequence:: and turnkey applications.
6433: @end menu
6434:
6435: @node Image Licensing Issues, Image File Background, Image Files, Image Files
6436: @section Image Licensing Issues
6437: @cindex license for images
6438: @cindex image license
6439:
6440: An image created with @code{gforthmi} (@pxref{gforthmi}) or
6441: @code{savesystem} (@pxref{Non-Relocatable Image Files}) includes the
6442: original image; i.e., according to copyright law it is a derived work of
6443: the original image.
6444:
6445: Since Gforth is distributed under the GNU GPL, the newly created image
6446: falls under the GNU GPL, too. In particular, this means that if you
6447: distribute the image, you have to make all of the sources for the image
6448: available, including those you wrote. For details see @ref{License, ,
6449: GNU General Public License (Section 3)}.
6450:
6451: If you create an image with @code{cross} (@pxref{cross.fs}), the image
6452: contains only code compiled from the sources you gave it; if none of
6453: these sources is under the GPL, the terms discussed above do not apply
6454: to the image. However, if your image needs an engine (a gforth binary)
6455: that is under the GPL, you should make sure that you distribute both in
6456: a way that is at most a @emph{mere aggregation}, if you don't want the
6457: terms of the GPL to apply to the image.
6458:
6459: @node Image File Background, Non-Relocatable Image Files, Image Licensing Issues, Image Files
6460: @section Image File Background
6461: @cindex image file background
6462:
6463: Our Forth system consists not only of primitives, but also of
6464: definitions written in Forth. Since the Forth compiler itself belongs to
6465: those definitions, it is not possible to start the system with the
6466: primitives and the Forth source alone. Therefore we provide the Forth
6467: code as an image file in nearly executable form. At the start of the
6468: system a C routine loads the image file into memory, optionally
6469: relocates the addresses, then sets up the memory (stacks etc.) according
6470: to information in the image file, and starts executing Forth code.
6471:
6472: The image file variants represent different compromises between the
6473: goals of making it easy to generate image files and making them
6474: portable.
6475:
6476: @cindex relocation at run-time
6477: Win32Forth 3.4 and Mitch Bradleys @code{cforth} use relocation at
6478: run-time. This avoids many of the complications discussed below (image
6479: files are data relocatable without further ado), but costs performance
6480: (one addition per memory access).
6481:
6482: @cindex relocation at load-time
6483: By contrast, our loader performs relocation at image load time. The
6484: loader also has to replace tokens standing for primitive calls with the
6485: appropriate code-field addresses (or code addresses in the case of
6486: direct threading).
6487:
6488: There are three kinds of image files, with different degrees of
6489: relocatability: non-relocatable, data-relocatable, and fully relocatable
6490: image files.
6491:
6492: @cindex image file loader
6493: @cindex relocating loader
6494: @cindex loader for image files
6495: These image file variants have several restrictions in common; they are
6496: caused by the design of the image file loader:
6497:
6498: @itemize @bullet
6499: @item
6500: There is only one segment; in particular, this means, that an image file
6501: cannot represent @code{ALLOCATE}d memory chunks (and pointers to
6502: them). And the contents of the stacks are not represented, either.
6503:
6504: @item
6505: The only kinds of relocation supported are: adding the same offset to
6506: all cells that represent data addresses; and replacing special tokens
6507: with code addresses or with pieces of machine code.
6508:
6509: If any complex computations involving addresses are performed, the
6510: results cannot be represented in the image file. Several applications that
6511: use such computations come to mind:
6512: @itemize @minus
6513: @item
6514: Hashing addresses (or data structures which contain addresses) for table
6515: lookup. If you use Gforth's @code{table}s or @code{wordlist}s for this
6516: purpose, you will have no problem, because the hash tables are
6517: recomputed automatically when the system is started. If you use your own
6518: hash tables, you will have to do something similar.
6519:
6520: @item
6521: There's a cute implementation of doubly-linked lists that uses
6522: @code{XOR}ed addresses. You could represent such lists as singly-linked
6523: in the image file, and restore the doubly-linked representation on
6524: startup.@footnote{In my opinion, though, you should think thrice before
6525: using a doubly-linked list (whatever implementation).}
6526:
6527: @item
6528: The code addresses of run-time routines like @code{docol:} cannot be
6529: represented in the image file (because their tokens would be replaced by
6530: machine code in direct threaded implementations). As a workaround,
6531: compute these addresses at run-time with @code{>code-address} from the
6532: executions tokens of appropriate words (see the definitions of
6533: @code{docol:} and friends in @file{kernel.fs}).
6534:
6535: @item
6536: On many architectures addresses are represented in machine code in some
6537: shifted or mangled form. You cannot put @code{CODE} words that contain
6538: absolute addresses in this form in a relocatable image file. Workarounds
6539: are representing the address in some relative form (e.g., relative to
6540: the CFA, which is present in some register), or loading the address from
6541: a place where it is stored in a non-mangled form.
6542: @end itemize
6543: @end itemize
6544:
6545: @node Non-Relocatable Image Files, Data-Relocatable Image Files, Image File Background, Image Files
6546: @section Non-Relocatable Image Files
6547: @cindex non-relocatable image files
6548: @cindex image files, non-relocatable
6549:
6550: These files are simple memory dumps of the dictionary. They are specific
6551: to the executable (i.e., @file{gforth} file) they were created
6552: with. What's worse, they are specific to the place on which the
6553: dictionary resided when the image was created. Now, there is no
6554: guarantee that the dictionary will reside at the same place the next
6555: time you start Gforth, so there's no guarantee that a non-relocatable
6556: image will work the next time (Gforth will complain instead of crashing,
6557: though).
6558:
6559: You can create a non-relocatable image file with
6560:
6561: doc-savesystem
6562:
6563: @node Data-Relocatable Image Files, Fully Relocatable Image Files, Non-Relocatable Image Files, Image Files
6564: @section Data-Relocatable Image Files
6565: @cindex data-relocatable image files
6566: @cindex image files, data-relocatable
6567:
6568: These files contain relocatable data addresses, but fixed code addresses
6569: (instead of tokens). They are specific to the executable (i.e.,
6570: @file{gforth} file) they were created with. For direct threading on some
6571: architectures (e.g., the i386), data-relocatable images do not work. You
6572: get a data-relocatable image, if you use @file{gforthmi} with a
6573: Gforth binary that is not doubly indirect threaded (@pxref{Fully
6574: Relocatable Image Files}).
6575:
6576: @node Fully Relocatable Image Files, Stack and Dictionary Sizes, Data-Relocatable Image Files, Image Files
6577: @section Fully Relocatable Image Files
6578: @cindex fully relocatable image files
6579: @cindex image files, fully relocatable
6580:
6581: @cindex @file{kern*.fi}, relocatability
6582: @cindex @file{gforth.fi}, relocatability
6583: These image files have relocatable data addresses, and tokens for code
6584: addresses. They can be used with different binaries (e.g., with and
6585: without debugging) on the same machine, and even across machines with
6586: the same data formats (byte order, cell size, floating point
6587: format). However, they are usually specific to the version of Gforth
6588: they were created with. The files @file{gforth.fi} and @file{kernl*.fi}
6589: are fully relocatable.
6590:
6591: There are two ways to create a fully relocatable image file:
6592:
6593: @menu
6594: * gforthmi:: The normal way
6595: * cross.fs:: The hard way
6596: @end menu
6597:
6598: @node gforthmi, cross.fs, Fully Relocatable Image Files, Fully Relocatable Image Files
6599: @subsection @file{gforthmi}
6600: @cindex @file{comp-i.fs}
6601: @cindex @file{gforthmi}
6602:
6603: You will usually use @file{gforthmi}. If you want to create an
6604: image @var{file} that contains everything you would load by invoking
6605: Gforth with @code{gforth @var{options}}, you simply say
6606: @example
6607: gforthmi @var{file} @var{options}
6608: @end example
6609:
6610: E.g., if you want to create an image @file{asm.fi} that has the file
6611: @file{asm.fs} loaded in addition to the usual stuff, you could do it
6612: like this:
6613:
6614: @example
6615: gforthmi asm.fi asm.fs
6616: @end example
6617:
6618: @file{gforthmi} works like this: It produces two non-relocatable
6619: images for different addresses and then compares them. Its output
6620: reflects this: first you see the output (if any) of the two Gforth
6621: invocations that produce the nonrelocatable image files, then you see
6622: the output of the comparing program: It displays the offset used for
6623: data addresses and the offset used for code addresses;
6624: moreover, for each cell that cannot be represented correctly in the
6625: image files, it displays a line like the following one:
6626:
6627: @example
6628: 78DC BFFFFA50 BFFFFA40
6629: @end example
6630:
6631: This means that at offset $78dc from @code{forthstart}, one input image
6632: contains $bffffa50, and the other contains $bffffa40. Since these cells
6633: cannot be represented correctly in the output image, you should examine
6634: these places in the dictionary and verify that these cells are dead
6635: (i.e., not read before they are written).
6636:
6637: @cindex @code{savesystem} during @file{gforthmi}
6638: @cindex @code{bye} during @file{gforthmi}
6639: @cindex doubly indirect threaded code
6640: @cindex environment variable @code{GFORTHD}
6641: @cindex @code{GFORTHD} environment variable
6642: @cindex @code{gforth-ditc}
6643: There are a few wrinkles: After processing the passed @var{options}, the
6644: words @code{savesystem} and @code{bye} must be visible. A special doubly
6645: indirect threaded version of the @file{gforth} executable is used for
6646: creating the nonrelocatable images; you can pass the exact filename of
6647: this executable through the environment variable @code{GFORTHD}
6648: (default: @file{gforth-ditc}); if you pass a version that is not doubly
6649: indirect threaded, you will not get a fully relocatable image, but a
6650: data-relocatable image (because there is no code address offset).
6651:
6652: @node cross.fs, , gforthmi, Fully Relocatable Image Files
6653: @subsection @file{cross.fs}
6654: @cindex @file{cross.fs}
6655: @cindex cross-compiler
6656: @cindex metacompiler
6657:
6658: You can also use @code{cross}, a batch compiler that accepts a Forth-like
6659: programming language. This @code{cross} language has to be documented
6660: yet.
6661:
6662: @cindex target compiler
6663: @code{cross} also allows you to create image files for machines with
6664: different data sizes and data formats than the one used for generating
6665: the image file. You can also use it to create an application image that
6666: does not contain a Forth compiler. These features are bought with
6667: restrictions and inconveniences in programming. E.g., addresses have to
6668: be stored in memory with special words (@code{A!}, @code{A,}, etc.) in
6669: order to make the code relocatable.
6670:
6671:
6672: @node Stack and Dictionary Sizes, Running Image Files, Fully Relocatable Image Files, Image Files
6673: @section Stack and Dictionary Sizes
6674: @cindex image file, stack and dictionary sizes
6675: @cindex dictionary size default
6676: @cindex stack size default
6677:
6678: If you invoke Gforth with a command line flag for the size
6679: (@pxref{Invoking Gforth}), the size you specify is stored in the
6680: dictionary. If you save the dictionary with @code{savesystem} or create
6681: an image with @file{gforthmi}, this size will become the default
6682: for the resulting image file. E.g., the following will create a
6683: fully relocatable version of gforth.fi with a 1MB dictionary:
6684:
6685: @example
6686: gforthmi gforth.fi -m 1M
6687: @end example
6688:
6689: In other words, if you want to set the default size for the dictionary
6690: and the stacks of an image, just invoke @file{gforthmi} with the
6691: appropriate options when creating the image.
6692:
6693: @cindex stack size, cache-friendly
6694: Note: For cache-friendly behaviour (i.e., good performance), you should
6695: make the sizes of the stacks modulo, say, 2K, somewhat different. E.g.,
6696: the default stack sizes are: data: 16k (mod 2k=0); fp: 15.5k (mod
6697: 2k=1.5k); return: 15k(mod 2k=1k); locals: 14.5k (mod 2k=0.5k).
6698:
6699: @node Running Image Files, Modifying the Startup Sequence, Stack and Dictionary Sizes, Image Files
6700: @section Running Image Files
6701: @cindex running image files
6702: @cindex invoking image files
6703: @cindex image file invocation
6704:
6705: @cindex -i, invoke image file
6706: @cindex --image file, invoke image file
6707: You can invoke Gforth with an image file @var{image} instead of the
6708: default @file{gforth.fi} with the @code{-i} flag (@pxref{Invoking Gforth}):
6709: @example
6710: gforth -i @var{image}
6711: @end example
6712:
6713: @cindex executable image file
6714: @cindex image files, executable
6715: If your operating system supports starting scripts with a line of the
6716: form @code{#! ...}, you just have to type the image file name to start
6717: Gforth with this image file (note that the file extension @code{.fi} is
6718: just a convention). I.e., to run Gforth with the image file @var{image},
6719: you can just type @var{image} instead of @code{gforth -i @var{image}}.
6720:
6721: doc-#!
6722:
6723: @node Modifying the Startup Sequence, , Running Image Files, Image Files
6724: @section Modifying the Startup Sequence
6725: @cindex startup sequence for image file
6726: @cindex image file initialization sequence
6727: @cindex initialization sequence of image file
6728:
6729: You can add your own initialization to the startup sequence through the
6730: deferred word
6731:
6732: doc-'cold
6733:
6734: @code{'cold} is invoked just before the image-specific command line
6735: processing (by default, loading files and evaluating (@code{-e}) strings)
6736: starts.
6737:
6738: A sequence for adding your initialization usually looks like this:
6739:
6740: @example
6741: :noname
6742: Defers 'cold \ do other initialization stuff (e.g., rehashing wordlists)
6743: ... \ your stuff
6744: ; IS 'cold
6745: @end example
6746:
6747: @cindex turnkey image files
6748: @cindex image files, turnkey applications
6749: You can make a turnkey image by letting @code{'cold} execute a word
6750: (your turnkey application) that never returns; instead, it exits Gforth
6751: via @code{bye} or @code{throw}.
6752:
6753: @cindex command-line arguments, access
6754: @cindex arguments on the command line, access
6755: You can access the (image-specific) command-line arguments through the
6756: variables @code{argc} and @code{argv}. @code{arg} provides conventient
6757: access to @code{argv}.
6758:
6759: doc-argc
6760: doc-argv
6761: doc-arg
6762:
6763: If @code{'cold} exits normally, Gforth processes the command-line
6764: arguments as files to be loaded and strings to be evaluated. Therefore,
6765: @code{'cold} should remove the arguments it has used in this case.
6766:
6767: @c ******************************************************************
6768: @node Engine, Binding to System Library, Image Files, Top
6769: @chapter Engine
6770: @cindex engine
6771: @cindex virtual machine
6772:
6773: Reading this section is not necessary for programming with Gforth. It
6774: may be helpful for finding your way in the Gforth sources.
6775:
6776: The ideas in this section have also been published in the papers
6777: @cite{ANS fig/GNU/??? Forth} (in German) by Bernd Paysan, presented at
6778: the Forth-Tagung '93 and @cite{A Portable Forth Engine} by M. Anton
6779: Ertl, presented at EuroForth '93; the latter is available at
6780: @*@url{http://www.complang.tuwien.ac.at/papers/ertl93.ps.Z}.
6781:
6782: @menu
6783: * Portability::
6784: * Threading::
6785: * Primitives::
6786: * Performance::
6787: @end menu
6788:
6789: @node Portability, Threading, Engine, Engine
6790: @section Portability
6791: @cindex engine portability
6792:
6793: One of the main goals of the effort is availability across a wide range
6794: of personal machines. fig-Forth, and, to a lesser extent, F83, achieved
6795: this goal by manually coding the engine in assembly language for several
6796: then-popular processors. This approach is very labor-intensive and the
6797: results are short-lived due to progress in computer architecture.
6798:
6799: @cindex C, using C for the engine
6800: Others have avoided this problem by coding in C, e.g., Mitch Bradley
6801: (cforth), Mikael Patel (TILE) and Dirk Zoller (pfe). This approach is
6802: particularly popular for UNIX-based Forths due to the large variety of
6803: architectures of UNIX machines. Unfortunately an implementation in C
6804: does not mix well with the goals of efficiency and with using
6805: traditional techniques: Indirect or direct threading cannot be expressed
6806: in C, and switch threading, the fastest technique available in C, is
6807: significantly slower. Another problem with C is that it is very
6808: cumbersome to express double integer arithmetic.
6809:
6810: @cindex GNU C for the engine
6811: @cindex long long
6812: Fortunately, there is a portable language that does not have these
6813: limitations: GNU C, the version of C processed by the GNU C compiler
6814: (@pxref{C Extensions, , Extensions to the C Language Family, gcc.info,
6815: GNU C Manual}). Its labels as values feature (@pxref{Labels as Values, ,
6816: Labels as Values, gcc.info, GNU C Manual}) makes direct and indirect
6817: threading possible, its @code{long long} type (@pxref{Long Long, ,
6818: Double-Word Integers, gcc.info, GNU C Manual}) corresponds to Forth's
6819: double numbers@footnote{Unfortunately, long longs are not implemented
6820: properly on all machines (e.g., on alpha-osf1, long longs are only 64
6821: bits, the same size as longs (and pointers), but they should be twice as
6822: long according to @pxref{Long Long, , Double-Word Integers, gcc.info, GNU
6823: C Manual}). So, we had to implement doubles in C after all. Still, on
6824: most machines we can use long longs and achieve better performance than
6825: with the emulation package.}. GNU C is available for free on all
6826: important (and many unimportant) UNIX machines, VMS, 80386s running
6827: MS-DOS, the Amiga, and the Atari ST, so a Forth written in GNU C can run
6828: on all these machines.
6829:
6830: Writing in a portable language has the reputation of producing code that
6831: is slower than assembly. For our Forth engine we repeatedly looked at
6832: the code produced by the compiler and eliminated most compiler-induced
6833: inefficiencies by appropriate changes in the source code.
6834:
6835: @cindex explicit register declarations
6836: @cindex --enable-force-reg, configuration flag
6837: @cindex -DFORCE_REG
6838: However, register allocation cannot be portably influenced by the
6839: programmer, leading to some inefficiencies on register-starved
6840: machines. We use explicit register declarations (@pxref{Explicit Reg
6841: Vars, , Variables in Specified Registers, gcc.info, GNU C Manual}) to
6842: improve the speed on some machines. They are turned on by using the
6843: configuration flag @code{--enable-force-reg} (@code{gcc} switch
6844: @code{-DFORCE_REG}). Unfortunately, this feature not only depends on the
6845: machine, but also on the compiler version: On some machines some
6846: compiler versions produce incorrect code when certain explicit register
6847: declarations are used. So by default @code{-DFORCE_REG} is not used.
6848:
6849: @node Threading, Primitives, Portability, Engine
6850: @section Threading
6851: @cindex inner interpreter implementation
6852: @cindex threaded code implementation
6853:
6854: @cindex labels as values
6855: GNU C's labels as values extension (available since @code{gcc-2.0},
6856: @pxref{Labels as Values, , Labels as Values, gcc.info, GNU C Manual})
6857: makes it possible to take the address of @var{label} by writing
6858: @code{&&@var{label}}. This address can then be used in a statement like
6859: @code{goto *@var{address}}. I.e., @code{goto *&&x} is the same as
6860: @code{goto x}.
6861:
6862: @cindex NEXT, indirect threaded
6863: @cindex indirect threaded inner interpreter
6864: @cindex inner interpreter, indirect threaded
6865: With this feature an indirect threaded NEXT looks like:
6866: @example
6867: cfa = *ip++;
6868: ca = *cfa;
6869: goto *ca;
6870: @end example
6871: @cindex instruction pointer
6872: For those unfamiliar with the names: @code{ip} is the Forth instruction
6873: pointer; the @code{cfa} (code-field address) corresponds to ANS Forths
6874: execution token and points to the code field of the next word to be
6875: executed; The @code{ca} (code address) fetched from there points to some
6876: executable code, e.g., a primitive or the colon definition handler
6877: @code{docol}.
6878:
6879: @cindex NEXT, direct threaded
6880: @cindex direct threaded inner interpreter
6881: @cindex inner interpreter, direct threaded
6882: Direct threading is even simpler:
6883: @example
6884: ca = *ip++;
6885: goto *ca;
6886: @end example
6887:
6888: Of course we have packaged the whole thing neatly in macros called
6889: @code{NEXT} and @code{NEXT1} (the part of NEXT after fetching the cfa).
6890:
6891: @menu
6892: * Scheduling::
6893: * Direct or Indirect Threaded?::
6894: * DOES>::
6895: @end menu
6896:
6897: @node Scheduling, Direct or Indirect Threaded?, Threading, Threading
6898: @subsection Scheduling
6899: @cindex inner interpreter optimization
6900:
6901: There is a little complication: Pipelined and superscalar processors,
6902: i.e., RISC and some modern CISC machines can process independent
6903: instructions while waiting for the results of an instruction. The
6904: compiler usually reorders (schedules) the instructions in a way that
6905: achieves good usage of these delay slots. However, on our first tries
6906: the compiler did not do well on scheduling primitives. E.g., for
6907: @code{+} implemented as
6908: @example
6909: n=sp[0]+sp[1];
6910: sp++;
6911: sp[0]=n;
6912: NEXT;
6913: @end example
6914: the NEXT comes strictly after the other code, i.e., there is nearly no
6915: scheduling. After a little thought the problem becomes clear: The
6916: compiler cannot know that sp and ip point to different addresses (and
6917: the version of @code{gcc} we used would not know it even if it was
6918: possible), so it could not move the load of the cfa above the store to
6919: the TOS. Indeed the pointers could be the same, if code on or very near
6920: the top of stack were executed. In the interest of speed we chose to
6921: forbid this probably unused ``feature'' and helped the compiler in
6922: scheduling: NEXT is divided into the loading part (@code{NEXT_P1}) and
6923: the goto part (@code{NEXT_P2}). @code{+} now looks like:
6924: @example
6925: n=sp[0]+sp[1];
6926: sp++;
6927: NEXT_P1;
6928: sp[0]=n;
6929: NEXT_P2;
6930: @end example
6931: This can be scheduled optimally by the compiler.
6932:
6933: This division can be turned off with the switch @code{-DCISC_NEXT}. This
6934: switch is on by default on machines that do not profit from scheduling
6935: (e.g., the 80386), in order to preserve registers.
6936:
6937: @node Direct or Indirect Threaded?, DOES>, Scheduling, Threading
6938: @subsection Direct or Indirect Threaded?
6939: @cindex threading, direct or indirect?
6940:
6941: @cindex -DDIRECT_THREADED
6942: Both! After packaging the nasty details in macro definitions we
6943: realized that we could switch between direct and indirect threading by
6944: simply setting a compilation flag (@code{-DDIRECT_THREADED}) and
6945: defining a few machine-specific macros for the direct-threading case.
6946: On the Forth level we also offer access words that hide the
6947: differences between the threading methods (@pxref{Threading Words}).
6948:
6949: Indirect threading is implemented completely machine-independently.
6950: Direct threading needs routines for creating jumps to the executable
6951: code (e.g. to docol or dodoes). These routines are inherently
6952: machine-dependent, but they do not amount to many source lines. I.e.,
6953: even porting direct threading to a new machine is a small effort.
6954:
6955: @cindex --enable-indirect-threaded, configuration flag
6956: @cindex --enable-direct-threaded, configuration flag
6957: The default threading method is machine-dependent. You can enforce a
6958: specific threading method when building Gforth with the configuration
6959: flag @code{--enable-direct-threaded} or
6960: @code{--enable-indirect-threaded}. Note that direct threading is not
6961: supported on all machines.
6962:
6963: @node DOES>, , Direct or Indirect Threaded?, Threading
6964: @subsection DOES>
6965: @cindex @code{DOES>} implementation
6966:
6967: @cindex dodoes routine
6968: @cindex DOES-code
6969: One of the most complex parts of a Forth engine is @code{dodoes}, i.e.,
6970: the chunk of code executed by every word defined by a
6971: @code{CREATE}...@code{DOES>} pair. The main problem here is: How to find
6972: the Forth code to be executed, i.e. the code after the
6973: @code{DOES>} (the DOES-code)? There are two solutions:
6974:
6975: In fig-Forth the code field points directly to the dodoes and the
6976: DOES-code address is stored in the cell after the code address (i.e. at
6977: @code{@var{cfa} cell+}). It may seem that this solution is illegal in
6978: the Forth-79 and all later standards, because in fig-Forth this address
6979: lies in the body (which is illegal in these standards). However, by
6980: making the code field larger for all words this solution becomes legal
6981: again. We use this approach for the indirect threaded version and for
6982: direct threading on some machines. Leaving a cell unused in most words
6983: is a bit wasteful, but on the machines we are targeting this is hardly a
6984: problem. The other reason for having a code field size of two cells is
6985: to avoid having different image files for direct and indirect threaded
6986: systems (direct threaded systems require two-cell code fields on many
6987: machines).
6988:
6989: @cindex DOES-handler
6990: The other approach is that the code field points or jumps to the cell
6991: after @code{DOES}. In this variant there is a jump to @code{dodoes} at
6992: this address (the DOES-handler). @code{dodoes} can then get the
6993: DOES-code address by computing the code address, i.e., the address of
6994: the jump to dodoes, and add the length of that jump field. A variant of
6995: this is to have a call to @code{dodoes} after the @code{DOES>}; then the
6996: return address (which can be found in the return register on RISCs) is
6997: the DOES-code address. Since the two cells available in the code field
6998: are used up by the jump to the code address in direct threading on many
6999: architectures, we use this approach for direct threading on these
7000: architectures. We did not want to add another cell to the code field.
7001:
7002: @node Primitives, Performance, Threading, Engine
7003: @section Primitives
7004: @cindex primitives, implementation
7005: @cindex virtual machine instructions, implementation
7006:
7007: @menu
7008: * Automatic Generation::
7009: * TOS Optimization::
7010: * Produced code::
7011: @end menu
7012:
7013: @node Automatic Generation, TOS Optimization, Primitives, Primitives
7014: @subsection Automatic Generation
7015: @cindex primitives, automatic generation
7016:
7017: @cindex @file{prims2x.fs}
7018: Since the primitives are implemented in a portable language, there is no
7019: longer any need to minimize the number of primitives. On the contrary,
7020: having many primitives has an advantage: speed. In order to reduce the
7021: number of errors in primitives and to make programming them easier, we
7022: provide a tool, the primitive generator (@file{prims2x.fs}), that
7023: automatically generates most (and sometimes all) of the C code for a
7024: primitive from the stack effect notation. The source for a primitive
7025: has the following form:
7026:
7027: @cindex primitive source format
7028: @format
7029: @var{Forth-name} @var{stack-effect} @var{category} [@var{pronounc.}]
7030: [@code{""}@var{glossary entry}@code{""}]
7031: @var{C code}
7032: [@code{:}
7033: @var{Forth code}]
7034: @end format
7035:
7036: The items in brackets are optional. The category and glossary fields
7037: are there for generating the documentation, the Forth code is there
7038: for manual implementations on machines without GNU C. E.g., the source
7039: for the primitive @code{+} is:
7040: @example
7041: + n1 n2 -- n core plus
7042: n = n1+n2;
7043: @end example
7044:
7045: This looks like a specification, but in fact @code{n = n1+n2} is C
7046: code. Our primitive generation tool extracts a lot of information from
7047: the stack effect notations@footnote{We use a one-stack notation, even
7048: though we have separate data and floating-point stacks; The separate
7049: notation can be generated easily from the unified notation.}: The number
7050: of items popped from and pushed on the stack, their type, and by what
7051: name they are referred to in the C code. It then generates a C code
7052: prelude and postlude for each primitive. The final C code for @code{+}
7053: looks like this:
7054:
7055: @example
7056: I_plus: /* + ( n1 n2 -- n ) */ /* label, stack effect */
7057: /* */ /* documentation */
7058: @{
7059: DEF_CA /* definition of variable ca (indirect threading) */
7060: Cell n1; /* definitions of variables */
7061: Cell n2;
7062: Cell n;
7063: n1 = (Cell) sp[1]; /* input */
7064: n2 = (Cell) TOS;
7065: sp += 1; /* stack adjustment */
7066: NAME("+") /* debugging output (with -DDEBUG) */
7067: @{
7068: n = n1+n2; /* C code taken from the source */
7069: @}
7070: NEXT_P1; /* NEXT part 1 */
7071: TOS = (Cell)n; /* output */
7072: NEXT_P2; /* NEXT part 2 */
7073: @}
7074: @end example
7075:
7076: This looks long and inefficient, but the GNU C compiler optimizes quite
7077: well and produces optimal code for @code{+} on, e.g., the R3000 and the
7078: HP RISC machines: Defining the @code{n}s does not produce any code, and
7079: using them as intermediate storage also adds no cost.
7080:
7081: There are also other optimizations, that are not illustrated by this
7082: example: Assignments between simple variables are usually for free (copy
7083: propagation). If one of the stack items is not used by the primitive
7084: (e.g. in @code{drop}), the compiler eliminates the load from the stack
7085: (dead code elimination). On the other hand, there are some things that
7086: the compiler does not do, therefore they are performed by
7087: @file{prims2x.fs}: The compiler does not optimize code away that stores
7088: a stack item to the place where it just came from (e.g., @code{over}).
7089:
7090: While programming a primitive is usually easy, there are a few cases
7091: where the programmer has to take the actions of the generator into
7092: account, most notably @code{?dup}, but also words that do not (always)
7093: fall through to NEXT.
7094:
7095: @node TOS Optimization, Produced code, Automatic Generation, Primitives
7096: @subsection TOS Optimization
7097: @cindex TOS optimization for primitives
7098: @cindex primitives, keeping the TOS in a register
7099:
7100: An important optimization for stack machine emulators, e.g., Forth
7101: engines, is keeping one or more of the top stack items in
7102: registers. If a word has the stack effect @var{in1}...@var{inx} @code{--}
7103: @var{out1}...@var{outy}, keeping the top @var{n} items in registers
7104: @itemize @bullet
7105: @item
7106: is better than keeping @var{n-1} items, if @var{x>=n} and @var{y>=n},
7107: due to fewer loads from and stores to the stack.
7108: @item is slower than keeping @var{n-1} items, if @var{x<>y} and @var{x<n} and
7109: @var{y<n}, due to additional moves between registers.
7110: @end itemize
7111:
7112: @cindex -DUSE_TOS
7113: @cindex -DUSE_NO_TOS
7114: In particular, keeping one item in a register is never a disadvantage,
7115: if there are enough registers. Keeping two items in registers is a
7116: disadvantage for frequent words like @code{?branch}, constants,
7117: variables, literals and @code{i}. Therefore our generator only produces
7118: code that keeps zero or one items in registers. The generated C code
7119: covers both cases; the selection between these alternatives is made at
7120: C-compile time using the switch @code{-DUSE_TOS}. @code{TOS} in the C
7121: code for @code{+} is just a simple variable name in the one-item case,
7122: otherwise it is a macro that expands into @code{sp[0]}. Note that the
7123: GNU C compiler tries to keep simple variables like @code{TOS} in
7124: registers, and it usually succeeds, if there are enough registers.
7125:
7126: @cindex -DUSE_FTOS
7127: @cindex -DUSE_NO_FTOS
7128: The primitive generator performs the TOS optimization for the
7129: floating-point stack, too (@code{-DUSE_FTOS}). For floating-point
7130: operations the benefit of this optimization is even larger:
7131: floating-point operations take quite long on most processors, but can be
7132: performed in parallel with other operations as long as their results are
7133: not used. If the FP-TOS is kept in a register, this works. If
7134: it is kept on the stack, i.e., in memory, the store into memory has to
7135: wait for the result of the floating-point operation, lengthening the
7136: execution time of the primitive considerably.
7137:
7138: The TOS optimization makes the automatic generation of primitives a
7139: bit more complicated. Just replacing all occurrences of @code{sp[0]} by
7140: @code{TOS} is not sufficient. There are some special cases to
7141: consider:
7142: @itemize @bullet
7143: @item In the case of @code{dup ( w -- w w )} the generator must not
7144: eliminate the store to the original location of the item on the stack,
7145: if the TOS optimization is turned on.
7146: @item Primitives with stack effects of the form @code{--}
7147: @var{out1}...@var{outy} must store the TOS to the stack at the start.
7148: Likewise, primitives with the stack effect @var{in1}...@var{inx} @code{--}
7149: must load the TOS from the stack at the end. But for the null stack
7150: effect @code{--} no stores or loads should be generated.
7151: @end itemize
7152:
7153: @node Produced code, , TOS Optimization, Primitives
7154: @subsection Produced code
7155: @cindex primitives, assembly code listing
7156:
7157: @cindex @file{engine.s}
7158: To see what assembly code is produced for the primitives on your machine
7159: with your compiler and your flag settings, type @code{make engine.s} and
7160: look at the resulting file @file{engine.s}.
7161:
7162: @node Performance, , Primitives, Engine
7163: @section Performance
7164: @cindex performance of some Forth interpreters
7165: @cindex engine performance
7166: @cindex benchmarking Forth systems
7167: @cindex Gforth performance
7168:
7169: On RISCs the Gforth engine is very close to optimal; i.e., it is usually
7170: impossible to write a significantly faster engine.
7171:
7172: On register-starved machines like the 386 architecture processors
7173: improvements are possible, because @code{gcc} does not utilize the
7174: registers as well as a human, even with explicit register declarations;
7175: e.g., Bernd Beuster wrote a Forth system fragment in assembly language
7176: and hand-tuned it for the 486; this system is 1.19 times faster on the
7177: Sieve benchmark on a 486DX2/66 than Gforth compiled with
7178: @code{gcc-2.6.3} with @code{-DFORCE_REG}.
7179:
7180: @cindex Win32Forth performance
7181: @cindex NT Forth performance
7182: @cindex eforth performance
7183: @cindex ThisForth performance
7184: @cindex PFE performance
7185: @cindex TILE performance
7186: However, this potential advantage of assembly language implementations
7187: is not necessarily realized in complete Forth systems: We compared
7188: Gforth (direct threaded, compiled with @code{gcc-2.6.3} and
7189: @code{-DFORCE_REG}) with Win32Forth 1.2093, LMI's NT Forth (Beta, May
7190: 1994) and Eforth (with and without peephole (aka pinhole) optimization
7191: of the threaded code); all these systems were written in assembly
7192: language. We also compared Gforth with three systems written in C:
7193: PFE-0.9.14 (compiled with @code{gcc-2.6.3} with the default
7194: configuration for Linux: @code{-O2 -fomit-frame-pointer -DUSE_REGS
7195: -DUNROLL_NEXT}), ThisForth Beta (compiled with gcc-2.6.3 -O3
7196: -fomit-frame-pointer; ThisForth employs peephole optimization of the
7197: threaded code) and TILE (compiled with @code{make opt}). We benchmarked
7198: Gforth, PFE, ThisForth and TILE on a 486DX2/66 under Linux. Kenneth
7199: O'Heskin kindly provided the results for Win32Forth and NT Forth on a
7200: 486DX2/66 with similar memory performance under Windows NT. Marcel
7201: Hendrix ported Eforth to Linux, then extended it to run the benchmarks,
7202: added the peephole optimizer, ran the benchmarks and reported the
7203: results.
7204:
7205: We used four small benchmarks: the ubiquitous Sieve; bubble-sorting and
7206: matrix multiplication come from the Stanford integer benchmarks and have
7207: been translated into Forth by Martin Fraeman; we used the versions
7208: included in the TILE Forth package, but with bigger data set sizes; and
7209: a recursive Fibonacci number computation for benchmarking calling
7210: performance. The following table shows the time taken for the benchmarks
7211: scaled by the time taken by Gforth (in other words, it shows the speedup
7212: factor that Gforth achieved over the other systems).
7213:
7214: @example
7215: relative Win32- NT eforth This-
7216: time Gforth Forth Forth eforth +opt PFE Forth TILE
7217: sieve 1.00 1.39 1.14 1.39 0.85 1.58 3.18 8.58
7218: bubble 1.00 1.31 1.41 1.48 0.88 1.50 3.88
7219: matmul 1.00 1.47 1.35 1.46 0.74 1.58 4.09
7220: fib 1.00 1.52 1.34 1.22 0.86 1.74 2.99 4.30
7221: @end example
7222:
7223: You may find the good performance of Gforth compared with the systems
7224: written in assembly language quite surprising. One important reason for
7225: the disappointing performance of these systems is probably that they are
7226: not written optimally for the 486 (e.g., they use the @code{lods}
7227: instruction). In addition, Win32Forth uses a comfortable, but costly
7228: method for relocating the Forth image: like @code{cforth}, it computes
7229: the actual addresses at run time, resulting in two address computations
7230: per NEXT (@pxref{Image File Background}).
7231:
7232: Only Eforth with the peephole optimizer performs comparable to
7233: Gforth. The speedups achieved with peephole optimization of threaded
7234: code are quite remarkable. Adding a peephole optimizer to Gforth should
7235: cause similar speedups.
7236:
7237: The speedup of Gforth over PFE, ThisForth and TILE can be easily
7238: explained with the self-imposed restriction of the latter systems to
7239: standard C, which makes efficient threading impossible (however, the
7240: measured implementation of PFE uses a GNU C extension: @pxref{Global Reg
7241: Vars, , Defining Global Register Variables, gcc.info, GNU C Manual}).
7242: Moreover, current C compilers have a hard time optimizing other aspects
7243: of the ThisForth and the TILE source.
7244:
7245: Note that the performance of Gforth on 386 architecture processors
7246: varies widely with the version of @code{gcc} used. E.g., @code{gcc-2.5.8}
7247: failed to allocate any of the virtual machine registers into real
7248: machine registers by itself and would not work correctly with explicit
7249: register declarations, giving a 1.3 times slower engine (on a 486DX2/66
7250: running the Sieve) than the one measured above.
7251:
7252: Note also that there have been several releases of Win32Forth since the
7253: release presented here, so the results presented here may have little
7254: predictive value for the performance of Win32Forth today.
7255:
7256: @cindex @file{Benchres}
7257: In @cite{Translating Forth to Efficient C} by M. Anton Ertl and Martin
7258: Maierhofer (presented at EuroForth '95), an indirect threaded version of
7259: Gforth is compared with Win32Forth, NT Forth, PFE, and ThisForth; that
7260: version of Gforth is 2%@minus{}8% slower on a 486 than the direct
7261: threaded version used here. The paper available at
7262: @*@url{http://www.complang.tuwien.ac.at/papers/ertl&maierhofer95.ps.gz};
7263: it also contains numbers for some native code systems. You can find a
7264: newer version of these measurements at
7265: @url{http://www.complang.tuwien.ac.at/forth/performance.html}. You can
7266: find numbers for Gforth on various machines in @file{Benchres}.
7267:
7268: @node Binding to System Library, Cross Compiler, Engine, Top
7269: @chapter Binding to System Library
7270:
7271: @node Cross Compiler, Bugs, Binding to System Library, Top
7272: @chapter Cross Compiler
7273:
7274: Cross Compiler
7275:
7276: @menu
7277: * Using the Cross Compiler::
7278: * How the Cross Compiler Works::
7279: @end menu
7280:
7281: @node Using the Cross Compiler, , How the Cross Compiler Works, Cross Compiler
7282: @section Using the Cross Compiler
7283:
7284: @node How the Cross Compiler Works, Using the Cross Compiler, , Cross Compiler
7285: @section How the Cross Compiler Works
7286:
7287: @node Bugs, Origin, Cross Compiler, Top
7288: @chapter Bugs
7289: @cindex bug reporting
7290:
7291: Known bugs are described in the file BUGS in the Gforth distribution.
7292:
7293: If you find a bug, please send a bug report to
7294: @email{bug-gforth@@gnu.ai.mit.edu}. A bug report should
7295: describe the Gforth version used (it is announced at the start of an
7296: interactive Gforth session), the machine and operating system (on Unix
7297: systems you can use @code{uname -a} to produce this information), the
7298: installation options (send the @file{config.status} file), and a
7299: complete list of changes you (or your installer) have made to the Gforth
7300: sources (if any); it should contain a program (or a sequence of keyboard
7301: commands) that reproduces the bug and a description of what you think
7302: constitutes the buggy behaviour.
7303:
7304: For a thorough guide on reporting bugs read @ref{Bug Reporting, , How
7305: to Report Bugs, gcc.info, GNU C Manual}.
7306:
7307:
7308: @node Origin, Word Index, Bugs, Top
7309: @chapter Authors and Ancestors of Gforth
7310:
7311: @section Authors and Contributors
7312: @cindex authors of Gforth
7313: @cindex contributors to Gforth
7314:
7315: The Gforth project was started in mid-1992 by Bernd Paysan and Anton
7316: Ertl. The third major author was Jens Wilke. Lennart Benschop (who was
7317: one of Gforth's first users, in mid-1993) and Stuart Ramsden inspired us
7318: with their continuous feedback. Lennart Benshop contributed
7319: @file{glosgen.fs}, while Stuart Ramsden has been working on automatic
7320: support for calling C libraries. Helpful comments also came from Paul
7321: Kleinrubatscher, Christian Pirker, Dirk Zoller, Marcel Hendrix, John
7322: Wavrik, Barrie Stott, Marc de Groot, and Jorge Acerada. Since the
7323: release of Gforth-0.2.1 there were also helpful comments from many
7324: others; thank you all, sorry for not listing you here (but digging
7325: through my mailbox to extract your names is on my to-do list).
7326:
7327: Gforth also owes a lot to the authors of the tools we used (GCC, CVS,
7328: and autoconf, among others), and to the creators of the Internet: Gforth
7329: was developed across the Internet, and its authors have not met
7330: physically yet.
7331:
7332: @section Pedigree
7333: @cindex Pedigree of Gforth
7334:
7335: Gforth descends from BigForth (1993) and fig-Forth. Gforth and PFE (by
7336: Dirk Zoller) will cross-fertilize each other. Of course, a significant
7337: part of the design of Gforth was prescribed by ANS Forth.
7338:
7339: Bernd Paysan wrote BigForth, a descendent from TurboForth, an unreleased
7340: 32 bit native code version of VolksForth for the Atari ST, written
7341: mostly by Dietrich Weineck.
7342:
7343: VolksForth descends from F83. It was written by Klaus Schleisiek, Bernd
7344: Pennemann, Georg Rehfeld and Dietrich Weineck for the C64 (called
7345: UltraForth there) in the mid-80s and ported to the Atari ST in 1986.
7346:
7347: Henry Laxen and Mike Perry wrote F83 as a model implementation of the
7348: Forth-83 standard. !! Pedigree? When?
7349:
7350: A team led by Bill Ragsdale implemented fig-Forth on many processors in
7351: 1979. Robert Selzer and Bill Ragsdale developed the original
7352: implementation of fig-Forth for the 6502 based on microForth.
7353:
7354: The principal architect of microForth was Dean Sanderson. microForth was
7355: FORTH, Inc.'s first off-the-shelf product. It was developed in 1976 for
7356: the 1802, and subsequently implemented on the 8080, the 6800 and the
7357: Z80.
7358:
7359: All earlier Forth systems were custom-made, usually by Charles Moore,
7360: who discovered (as he puts it) Forth during the late 60s. The first full
7361: Forth existed in 1971.
7362:
7363: A part of the information in this section comes from @cite{The Evolution
7364: of Forth} by Elizabeth D. Rather, Donald R. Colburn and Charles
7365: H. Moore, presented at the HOPL-II conference and preprinted in SIGPLAN
7366: Notices 28(3), 1993. You can find more historical and genealogical
7367: information about Forth there.
7368:
7369: @node Word Index, Concept Index, Origin, Top
7370: @unnumbered Word Index
7371:
7372: This index is as incomplete as the manual. Each word is listed with
7373: stack effect and wordset.
7374:
7375: @printindex fn
7376:
7377: @node Concept Index, , Word Index, Top
7378: @unnumbered Concept and Word Index
7379:
7380: This index is as incomplete as the manual. Not all entries listed are
7381: present verbatim in the text. Only the names are listed for the words
7382: here.
7383:
7384: @printindex cp
7385:
7386: @contents
7387: @bye
7388:
FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>
![[BACK]](/cvsweb/icons/back.gif){kind=icon}
Return to gforth.ds CVS log ![[TXT]](/cvsweb/icons/text.gif){kind=icon}
![[DIR]](/cvsweb/icons/dir.gif){kind=icon}
Up to [gforth] / gforth / doc