Annotation of gforth/doc/gforth.ds, revision 1.16
1.1 anton 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
1.12 anton 11: @macro progstyle {}
12: Programming style note:
1.3 anton 13: @end macro
1.1 anton 14: @comment %**end of header (This is for running Texinfo on a region.)
15:
1.10 anton 16: @include version.texi
17:
1.1 anton 18: @ifinfo
1.11 anton 19: This file documents Gforth @value{VERSION}
1.1 anton 20:
1.13 pazsan 21: Copyright @copyright{} 1995-1998 Free Software Foundation, Inc.
1.1 anton 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
1.11 anton 53: @center for version @value{VERSION}
1.1 anton 54: @sp 2
55: @center Anton Ertl
1.6 pazsan 56: @center Bernd Paysan
1.5 anton 57: @center Jens Wilke
1.1 anton 58: @sp 3
1.6 pazsan 59: @center This manual is permanently under construction
1.1 anton 60:
61: @comment The following two commands start the copyright page.
62: @page
63: @vskip 0pt plus 1filll
1.13 pazsan 64: Copyright @copyright{} 1995--1998 Free Software Foundation, Inc.
1.1 anton 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
1.11 anton 90: personal machines. This manual corresponds to version @value{VERSION}.
1.1 anton 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
1.13 pazsan 106: * Cross Compiler:: The Cross Compiler
1.1 anton 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
1.12 anton 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::
1.13 pazsan 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:
1.1 anton 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
1.12 anton 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.
1.1 anton 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
1.16 ! anton 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.
1.1 anton 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:
1.12 anton 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
1.16 ! anton 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}.
1.1 anton 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:
1.16 ! anton 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.
1.12 anton 852:
1.1 anton 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
1.12 anton 866: This interprets the contents of the files and the Forth code in the order they
1.1 anton 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
1.12 anton 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.
1.1 anton 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:
1.5 anton 961: @cindex --no-offset-im, command-line option
962: @item --no-offset-im
963: Start the dictionary at the normal position.
964:
1.1 anton 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}).
1.5 anton 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.
1.1 anton 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::
1.5 anton 1016: * Memory::
1.1 anton 1017: * Control Structures::
1018: * Locals::
1019: * Defining Words::
1.5 anton 1020: * Structures::
1.12 anton 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::
1.1 anton 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
1.12 anton 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{<>}
1.1 anton 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
1.12 anton 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).
1.1 anton 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:
1272: doc-f+
1273: doc-f-
1274: doc-f*
1275: doc-f/
1276: doc-fnegate
1277: doc-fabs
1278: doc-fmax
1279: doc-fmin
1280: doc-floor
1281: doc-fround
1282: doc-f**
1283: doc-fsqrt
1284: doc-fexp
1285: doc-fexpm1
1286: doc-fln
1287: doc-flnp1
1288: doc-flog
1289: doc-falog
1290: doc-fsin
1291: doc-fcos
1292: doc-fsincos
1293: doc-ftan
1294: doc-fasin
1295: doc-facos
1296: doc-fatan
1297: doc-fatan2
1298: doc-fsinh
1299: doc-fcosh
1300: doc-ftanh
1301: doc-fasinh
1302: doc-facosh
1303: doc-fatanh
1304:
1305: @node Stack Manipulation, Memory, Arithmetic, Words
1306: @section Stack Manipulation
1307: @cindex stack manipulation words
1308:
1309: @cindex floating-point stack in the standard
1310: Gforth has a data stack (aka parameter stack) for characters, cells,
1311: addresses, and double cells, a floating point stack for floating point
1312: numbers, a return stack for storing the return addresses of colon
1313: definitions and other data, and a locals stack for storing local
1314: variables. Note that while every sane Forth has a separate floating
1315: point stack, this is not strictly required; an ANS Forth system could
1316: theoretically keep floating point numbers on the data stack. As an
1317: additional difficulty, you don't know how many cells a floating point
1318: number takes. It is reportedly possible to write words in a way that
1319: they work also for a unified stack model, but we do not recommend trying
1320: it. Instead, just say that your program has an environmental dependency
1321: on a separate FP stack.
1322:
1323: @cindex return stack and locals
1324: @cindex locals and return stack
1325: Also, a Forth system is allowed to keep the local variables on the
1326: return stack. This is reasonable, as local variables usually eliminate
1327: the need to use the return stack explicitly. So, if you want to produce
1328: a standard complying program and if you are using local variables in a
1329: word, forget about return stack manipulations in that word (see the
1330: standard document for the exact rules).
1331:
1332: @menu
1333: * Data stack::
1334: * Floating point stack::
1335: * Return stack::
1336: * Locals stack::
1337: * Stack pointer manipulation::
1338: @end menu
1339:
1340: @node Data stack, Floating point stack, Stack Manipulation, Stack Manipulation
1341: @subsection Data stack
1342: @cindex data stack manipulation words
1343: @cindex stack manipulations words, data stack
1344:
1345: doc-drop
1346: doc-nip
1347: doc-dup
1348: doc-over
1349: doc-tuck
1350: doc-swap
1351: doc-rot
1352: doc--rot
1353: doc-?dup
1354: doc-pick
1355: doc-roll
1356: doc-2drop
1357: doc-2nip
1358: doc-2dup
1359: doc-2over
1360: doc-2tuck
1361: doc-2swap
1362: doc-2rot
1363:
1364: @node Floating point stack, Return stack, Data stack, Stack Manipulation
1365: @subsection Floating point stack
1366: @cindex floating-point stack manipulation words
1367: @cindex stack manipulation words, floating-point stack
1368:
1369: doc-fdrop
1370: doc-fnip
1371: doc-fdup
1372: doc-fover
1373: doc-ftuck
1374: doc-fswap
1375: doc-frot
1376:
1377: @node Return stack, Locals stack, Floating point stack, Stack Manipulation
1378: @subsection Return stack
1379: @cindex return stack manipulation words
1380: @cindex stack manipulation words, return stack
1381:
1382: doc->r
1383: doc-r>
1384: doc-r@
1385: doc-rdrop
1386: doc-2>r
1387: doc-2r>
1388: doc-2r@
1389: doc-2rdrop
1390:
1391: @node Locals stack, Stack pointer manipulation, Return stack, Stack Manipulation
1392: @subsection Locals stack
1393:
1394: @node Stack pointer manipulation, , Locals stack, Stack Manipulation
1395: @subsection Stack pointer manipulation
1396: @cindex stack pointer manipulation words
1397:
1398: doc-sp@
1399: doc-sp!
1400: doc-fp@
1401: doc-fp!
1402: doc-rp@
1403: doc-rp!
1404: doc-lp@
1405: doc-lp!
1406:
1407: @node Memory, Control Structures, Stack Manipulation, Words
1408: @section Memory
1409: @cindex Memory words
1410:
1411: @menu
1412: * Memory Access::
1413: * Address arithmetic::
1414: * Memory Blocks::
1415: @end menu
1416:
1417: @node Memory Access, Address arithmetic, Memory, Memory
1418: @subsection Memory Access
1419: @cindex memory access words
1420:
1421: doc-@
1422: doc-!
1423: doc-+!
1424: doc-c@
1425: doc-c!
1426: doc-2@
1427: doc-2!
1428: doc-f@
1429: doc-f!
1430: doc-sf@
1431: doc-sf!
1432: doc-df@
1433: doc-df!
1434:
1435: @node Address arithmetic, Memory Blocks, Memory Access, Memory
1436: @subsection Address arithmetic
1437: @cindex address arithmetic words
1438:
1439: ANS Forth does not specify the sizes of the data types. Instead, it
1440: offers a number of words for computing sizes and doing address
1441: arithmetic. Basically, address arithmetic is performed in terms of
1442: address units (aus); on most systems the address unit is one byte. Note
1443: that a character may have more than one au, so @code{chars} is no noop
1444: (on systems where it is a noop, it compiles to nothing).
1445:
1446: @cindex alignment of addresses for types
1447: ANS Forth also defines words for aligning addresses for specific
1448: types. Many computers require that accesses to specific data types
1449: must only occur at specific addresses; e.g., that cells may only be
1450: accessed at addresses divisible by 4. Even if a machine allows unaligned
1451: accesses, it can usually perform aligned accesses faster.
1452:
1453: For the performance-conscious: alignment operations are usually only
1454: necessary during the definition of a data structure, not during the
1455: (more frequent) accesses to it.
1456:
1457: ANS Forth defines no words for character-aligning addresses. This is not
1458: an oversight, but reflects the fact that addresses that are not
1459: char-aligned have no use in the standard and therefore will not be
1460: created.
1461:
1462: @cindex @code{CREATE} and alignment
1463: The standard guarantees that addresses returned by @code{CREATE}d words
1464: are cell-aligned; in addition, Gforth guarantees that these addresses
1465: are aligned for all purposes.
1466:
1467: Note that the standard defines a word @code{char}, which has nothing to
1468: do with address arithmetic.
1469:
1470: doc-chars
1471: doc-char+
1472: doc-cells
1473: doc-cell+
1474: doc-cell
1475: doc-align
1476: doc-aligned
1477: doc-floats
1478: doc-float+
1479: doc-float
1480: doc-falign
1481: doc-faligned
1482: doc-sfloats
1483: doc-sfloat+
1484: doc-sfalign
1485: doc-sfaligned
1486: doc-dfloats
1487: doc-dfloat+
1488: doc-dfalign
1489: doc-dfaligned
1490: doc-maxalign
1491: doc-maxaligned
1492: doc-cfalign
1493: doc-cfaligned
1494: doc-address-unit-bits
1495:
1496: @node Memory Blocks, , Address arithmetic, Memory
1497: @subsection Memory Blocks
1498: @cindex memory block words
1499:
1500: doc-move
1501: doc-erase
1502:
1503: While the previous words work on address units, the rest works on
1504: characters.
1505:
1506: doc-cmove
1507: doc-cmove>
1508: doc-fill
1509: doc-blank
1510:
1511: @node Control Structures, Locals, Memory, Words
1512: @section Control Structures
1513: @cindex control structures
1514:
1515: Control structures in Forth cannot be used in interpret state, only in
1516: compile state@footnote{More precisely, they have no interpretation
1517: semantics (@pxref{Interpretation and Compilation Semantics})}, i.e., in
1518: a colon definition. We do not like this limitation, but have not seen a
1519: satisfying way around it yet, although many schemes have been proposed.
1520:
1521: @menu
1522: * Selection::
1523: * Simple Loops::
1524: * Counted Loops::
1525: * Arbitrary control structures::
1526: * Calls and returns::
1527: * Exception Handling::
1528: @end menu
1529:
1530: @node Selection, Simple Loops, Control Structures, Control Structures
1531: @subsection Selection
1532: @cindex selection control structures
1533: @cindex control structures for selection
1534:
1535: @cindex @code{IF} control structure
1536: @example
1537: @var{flag}
1538: IF
1539: @var{code}
1540: ENDIF
1541: @end example
1542: or
1543: @example
1544: @var{flag}
1545: IF
1546: @var{code1}
1547: ELSE
1548: @var{code2}
1549: ENDIF
1550: @end example
1551:
1552: You can use @code{THEN} instead of @code{ENDIF}. Indeed, @code{THEN} is
1553: standard, and @code{ENDIF} is not, although it is quite popular. We
1554: recommend using @code{ENDIF}, because it is less confusing for people
1555: who also know other languages (and is not prone to reinforcing negative
1556: prejudices against Forth in these people). Adding @code{ENDIF} to a
1557: system that only supplies @code{THEN} is simple:
1558: @example
1559: : endif POSTPONE then ; immediate
1560: @end example
1561:
1562: [According to @cite{Webster's New Encyclopedic Dictionary}, @dfn{then
1563: (adv.)} has the following meanings:
1564: @quotation
1565: ... 2b: following next after in order ... 3d: as a necessary consequence
1566: (if you were there, then you saw them).
1567: @end quotation
1568: Forth's @code{THEN} has the meaning 2b, whereas @code{THEN} in Pascal
1569: and many other programming languages has the meaning 3d.]
1570:
1571: Gforth also provides the words @code{?dup-if} and @code{?dup-0=-if}, so
1572: you can avoid using @code{?dup}. Using these alternatives is also more
1573: efficient than using @code{?dup}. Definitions in plain standard Forth
1574: for @code{ENDIF}, @code{?DUP-IF} and @code{?DUP-0=-IF} are provided in
1575: @file{compat/control.fs}.
1576:
1577: @cindex @code{CASE} control structure
1578: @example
1579: @var{n}
1580: CASE
1581: @var{n1} OF @var{code1} ENDOF
1582: @var{n2} OF @var{code2} ENDOF
1583: @dots{}
1584: ENDCASE
1585: @end example
1586:
1587: Executes the first @var{codei}, where the @var{ni} is equal to
1588: @var{n}. A default case can be added by simply writing the code after
1589: the last @code{ENDOF}. It may use @var{n}, which is on top of the stack,
1590: but must not consume it.
1591:
1592: @node Simple Loops, Counted Loops, Selection, Control Structures
1593: @subsection Simple Loops
1594: @cindex simple loops
1595: @cindex loops without count
1596:
1597: @cindex @code{WHILE} loop
1598: @example
1599: BEGIN
1600: @var{code1}
1601: @var{flag}
1602: WHILE
1603: @var{code2}
1604: REPEAT
1605: @end example
1606:
1607: @var{code1} is executed and @var{flag} is computed. If it is true,
1608: @var{code2} is executed and the loop is restarted; If @var{flag} is
1609: false, execution continues after the @code{REPEAT}.
1610:
1611: @cindex @code{UNTIL} loop
1612: @example
1613: BEGIN
1614: @var{code}
1615: @var{flag}
1616: UNTIL
1617: @end example
1618:
1619: @var{code} is executed. The loop is restarted if @code{flag} is false.
1620:
1621: @cindex endless loop
1622: @cindex loops, endless
1623: @example
1624: BEGIN
1625: @var{code}
1626: AGAIN
1627: @end example
1628:
1629: This is an endless loop.
1630:
1631: @node Counted Loops, Arbitrary control structures, Simple Loops, Control Structures
1632: @subsection Counted Loops
1633: @cindex counted loops
1634: @cindex loops, counted
1635: @cindex @code{DO} loops
1636:
1637: The basic counted loop is:
1638: @example
1639: @var{limit} @var{start}
1640: ?DO
1641: @var{body}
1642: LOOP
1643: @end example
1644:
1645: This performs one iteration for every integer, starting from @var{start}
1646: and up to, but excluding @var{limit}. The counter, aka index, can be
1647: accessed with @code{i}. E.g., the loop
1648: @example
1649: 10 0 ?DO
1650: i .
1651: LOOP
1652: @end example
1653: prints
1654: @example
1655: 0 1 2 3 4 5 6 7 8 9
1656: @end example
1657: The index of the innermost loop can be accessed with @code{i}, the index
1658: of the next loop with @code{j}, and the index of the third loop with
1659: @code{k}.
1660:
1661: doc-i
1662: doc-j
1663: doc-k
1664:
1665: The loop control data are kept on the return stack, so there are some
1666: restrictions on mixing return stack accesses and counted loop
1667: words. E.g., if you put values on the return stack outside the loop, you
1668: cannot read them inside the loop. If you put values on the return stack
1669: within a loop, you have to remove them before the end of the loop and
1670: before accessing the index of the loop.
1671:
1672: There are several variations on the counted loop:
1673:
1674: @code{LEAVE} leaves the innermost counted loop immediately.
1675:
1676: If @var{start} is greater than @var{limit}, a @code{?DO} loop is entered
1677: (and @code{LOOP} iterates until they become equal by wrap-around
1678: arithmetic). This behaviour is usually not what you want. Therefore,
1679: Gforth offers @code{+DO} and @code{U+DO} (as replacements for
1680: @code{?DO}), which do not enter the loop if @var{start} is greater than
1681: @var{limit}; @code{+DO} is for signed loop parameters, @code{U+DO} for
1682: unsigned loop parameters.
1683:
1684: @code{LOOP} can be replaced with @code{@var{n} +LOOP}; this updates the
1685: index by @var{n} instead of by 1. The loop is terminated when the border
1686: between @var{limit-1} and @var{limit} is crossed. E.g.:
1687:
1688: @code{4 0 +DO i . 2 +LOOP} prints @code{0 2}
1689:
1690: @code{4 1 +DO i . 2 +LOOP} prints @code{1 3}
1691:
1692: @cindex negative increment for counted loops
1693: @cindex counted loops with negative increment
1694: The behaviour of @code{@var{n} +LOOP} is peculiar when @var{n} is negative:
1695:
1696: @code{-1 0 ?DO i . -1 +LOOP} prints @code{0 -1}
1697:
1698: @code{ 0 0 ?DO i . -1 +LOOP} prints nothing
1699:
1700: Therefore we recommend avoiding @code{@var{n} +LOOP} with negative
1701: @var{n}. One alternative is @code{@var{u} -LOOP}, which reduces the
1702: index by @var{u} each iteration. The loop is terminated when the border
1703: between @var{limit+1} and @var{limit} is crossed. Gforth also provides
1704: @code{-DO} and @code{U-DO} for down-counting loops. E.g.:
1705:
1706: @code{-2 0 -DO i . 1 -LOOP} prints @code{0 -1}
1707:
1708: @code{-1 0 -DO i . 1 -LOOP} prints @code{0}
1709:
1710: @code{ 0 0 -DO i . 1 -LOOP} prints nothing
1711:
1712: Unfortunately, @code{+DO}, @code{U+DO}, @code{-DO}, @code{U-DO} and
1713: @code{-LOOP} are not in the ANS Forth standard. However, an
1714: implementation for these words that uses only standard words is provided
1715: in @file{compat/loops.fs}.
1716:
1717: @code{?DO} can also be replaced by @code{DO}. @code{DO} always enters
1718: the loop, independent of the loop parameters. Do not use @code{DO}, even
1719: if you know that the loop is entered in any case. Such knowledge tends
1720: to become invalid during maintenance of a program, and then the
1721: @code{DO} will make trouble.
1722:
1723: @code{UNLOOP} is used to prepare for an abnormal loop exit, e.g., via
1724: @code{EXIT}. @code{UNLOOP} removes the loop control parameters from the
1725: return stack so @code{EXIT} can get to its return address.
1726:
1727: @cindex @code{FOR} loops
1728: Another counted loop is
1729: @example
1730: @var{n}
1731: FOR
1732: @var{body}
1733: NEXT
1734: @end example
1735: This is the preferred loop of native code compiler writers who are too
1736: lazy to optimize @code{?DO} loops properly. In Gforth, this loop
1737: iterates @var{n+1} times; @code{i} produces values starting with @var{n}
1738: and ending with 0. Other Forth systems may behave differently, even if
1739: they support @code{FOR} loops. To avoid problems, don't use @code{FOR}
1740: loops.
1741:
1742: @node Arbitrary control structures, Calls and returns, Counted Loops, Control Structures
1743: @subsection Arbitrary control structures
1744: @cindex control structures, user-defined
1745:
1746: @cindex control-flow stack
1747: ANS Forth permits and supports using control structures in a non-nested
1748: way. Information about incomplete control structures is stored on the
1749: control-flow stack. This stack may be implemented on the Forth data
1750: stack, and this is what we have done in Gforth.
1751:
1752: @cindex @code{orig}, control-flow stack item
1753: @cindex @code{dest}, control-flow stack item
1754: An @i{orig} entry represents an unresolved forward branch, a @i{dest}
1755: entry represents a backward branch target. A few words are the basis for
1756: building any control structure possible (except control structures that
1757: need storage, like calls, coroutines, and backtracking).
1758:
1759: doc-if
1760: doc-ahead
1761: doc-then
1762: doc-begin
1763: doc-until
1764: doc-again
1765: doc-cs-pick
1766: doc-cs-roll
1767:
1768: On many systems control-flow stack items take one word, in Gforth they
1769: currently take three (this may change in the future). Therefore it is a
1770: really good idea to manipulate the control flow stack with
1771: @code{cs-pick} and @code{cs-roll}, not with data stack manipulation
1772: words.
1773:
1774: Some standard control structure words are built from these words:
1775:
1776: doc-else
1777: doc-while
1778: doc-repeat
1779:
1780: Gforth adds some more control-structure words:
1781:
1782: doc-endif
1783: doc-?dup-if
1784: doc-?dup-0=-if
1785:
1786: Counted loop words constitute a separate group of words:
1787:
1788: doc-?do
1789: doc-+do
1790: doc-u+do
1791: doc--do
1792: doc-u-do
1793: doc-do
1794: doc-for
1795: doc-loop
1796: doc-+loop
1797: doc--loop
1798: doc-next
1799: doc-leave
1800: doc-?leave
1801: doc-unloop
1802: doc-done
1803:
1804: The standard does not allow using @code{cs-pick} and @code{cs-roll} on
1805: @i{do-sys}. Our system allows it, but it's your job to ensure that for
1806: every @code{?DO} etc. there is exactly one @code{UNLOOP} on any path
1807: through the definition (@code{LOOP} etc. compile an @code{UNLOOP} on the
1808: fall-through path). Also, you have to ensure that all @code{LEAVE}s are
1809: resolved (by using one of the loop-ending words or @code{DONE}).
1810:
1811: Another group of control structure words are
1812:
1813: doc-case
1814: doc-endcase
1815: doc-of
1816: doc-endof
1817:
1818: @i{case-sys} and @i{of-sys} cannot be processed using @code{cs-pick} and
1819: @code{cs-roll}.
1820:
1821: @subsubsection Programming Style
1822:
1823: In order to ensure readability we recommend that you do not create
1824: arbitrary control structures directly, but define new control structure
1825: words for the control structure you want and use these words in your
1826: program.
1827:
1828: E.g., instead of writing
1829:
1830: @example
1831: begin
1832: ...
1833: if [ 1 cs-roll ]
1834: ...
1835: again then
1836: @end example
1837:
1838: we recommend defining control structure words, e.g.,
1839:
1840: @example
1841: : while ( dest -- orig dest )
1842: POSTPONE if
1843: 1 cs-roll ; immediate
1844:
1845: : repeat ( orig dest -- )
1846: POSTPONE again
1847: POSTPONE then ; immediate
1848: @end example
1849:
1850: and then using these to create the control structure:
1851:
1852: @example
1853: begin
1854: ...
1855: while
1856: ...
1857: repeat
1858: @end example
1859:
1860: That's much easier to read, isn't it? Of course, @code{REPEAT} and
1861: @code{WHILE} are predefined, so in this example it would not be
1862: necessary to define them.
1863:
1864: @node Calls and returns, Exception Handling, Arbitrary control structures, Control Structures
1865: @subsection Calls and returns
1866: @cindex calling a definition
1867: @cindex returning from a definition
1868:
1.3 anton 1869: @cindex recursive definitions
1870: A definition can be called simply be writing the name of the definition
1871: to be called. Note that normally a definition is invisible during its
1872: definition. If you want to write a directly recursive definition, you
1873: can use @code{recursive} to make the current definition visible.
1874:
1875: doc-recursive
1876:
1877: Another way to perform a recursive call is
1878:
1879: doc-recurse
1880:
1.12 anton 1881: @quotation
1882: @progstyle
1883: I prefer using @code{recursive} to @code{recurse}, because calling the
1884: definition by name is more descriptive (if the name is well-chosen) than
1885: the somewhat cryptic @code{recurse}. E.g., in a quicksort
1886: implementation, it is much better to read (and think) ``now sort the
1887: partitions'' than to read ``now do a recursive call''.
1888: @end quotation
1.3 anton 1889:
1890: For mutual recursion, use @code{defer}red words, like this:
1891:
1892: @example
1893: defer foo
1894:
1895: : bar ( ... -- ... )
1896: ... foo ... ;
1897:
1898: :noname ( ... -- ... )
1899: ... bar ... ;
1900: IS foo
1901: @end example
1902:
1903: When the end of the definition is reached, it returns. An earlier return
1904: can be forced using
1.1 anton 1905:
1906: doc-exit
1907:
1908: Don't forget to clean up the return stack and @code{UNLOOP} any
1909: outstanding @code{?DO}...@code{LOOP}s before @code{EXIT}ing. The
1910: primitive compiled by @code{EXIT} is
1911:
1912: doc-;s
1913:
1914: @node Exception Handling, , Calls and returns, Control Structures
1915: @subsection Exception Handling
1916: @cindex Exceptions
1917:
1918: doc-catch
1919: doc-throw
1920:
1921: @node Locals, Defining Words, Control Structures, Words
1922: @section Locals
1923: @cindex locals
1924:
1925: Local variables can make Forth programming more enjoyable and Forth
1926: programs easier to read. Unfortunately, the locals of ANS Forth are
1927: laden with restrictions. Therefore, we provide not only the ANS Forth
1928: locals wordset, but also our own, more powerful locals wordset (we
1929: implemented the ANS Forth locals wordset through our locals wordset).
1930:
1931: The ideas in this section have also been published in the paper
1932: @cite{Automatic Scoping of Local Variables} by M. Anton Ertl, presented
1933: at EuroForth '94; it is available at
1934: @*@url{http://www.complang.tuwien.ac.at/papers/ertl94l.ps.gz}.
1935:
1936: @menu
1937: * Gforth locals::
1938: * ANS Forth locals::
1939: @end menu
1940:
1941: @node Gforth locals, ANS Forth locals, Locals, Locals
1942: @subsection Gforth locals
1943: @cindex Gforth locals
1944: @cindex locals, Gforth style
1945:
1946: Locals can be defined with
1947:
1948: @example
1949: @{ local1 local2 ... -- comment @}
1950: @end example
1951: or
1952: @example
1953: @{ local1 local2 ... @}
1954: @end example
1955:
1956: E.g.,
1957: @example
1958: : max @{ n1 n2 -- n3 @}
1959: n1 n2 > if
1960: n1
1961: else
1962: n2
1963: endif ;
1964: @end example
1965:
1966: The similarity of locals definitions with stack comments is intended. A
1967: locals definition often replaces the stack comment of a word. The order
1968: of the locals corresponds to the order in a stack comment and everything
1969: after the @code{--} is really a comment.
1970:
1971: This similarity has one disadvantage: It is too easy to confuse locals
1972: declarations with stack comments, causing bugs and making them hard to
1973: find. However, this problem can be avoided by appropriate coding
1974: conventions: Do not use both notations in the same program. If you do,
1975: they should be distinguished using additional means, e.g. by position.
1976:
1977: @cindex types of locals
1978: @cindex locals types
1979: The name of the local may be preceded by a type specifier, e.g.,
1980: @code{F:} for a floating point value:
1981:
1982: @example
1983: : CX* @{ F: Ar F: Ai F: Br F: Bi -- Cr Ci @}
1984: \ complex multiplication
1985: Ar Br f* Ai Bi f* f-
1986: Ar Bi f* Ai Br f* f+ ;
1987: @end example
1988:
1989: @cindex flavours of locals
1990: @cindex locals flavours
1991: @cindex value-flavoured locals
1992: @cindex variable-flavoured locals
1993: Gforth currently supports cells (@code{W:}, @code{W^}), doubles
1994: (@code{D:}, @code{D^}), floats (@code{F:}, @code{F^}) and characters
1995: (@code{C:}, @code{C^}) in two flavours: a value-flavoured local (defined
1996: with @code{W:}, @code{D:} etc.) produces its value and can be changed
1997: with @code{TO}. A variable-flavoured local (defined with @code{W^} etc.)
1998: produces its address (which becomes invalid when the variable's scope is
1999: left). E.g., the standard word @code{emit} can be defined in terms of
2000: @code{type} like this:
2001:
2002: @example
2003: : emit @{ C^ char* -- @}
2004: char* 1 type ;
2005: @end example
2006:
2007: @cindex default type of locals
2008: @cindex locals, default type
2009: A local without type specifier is a @code{W:} local. Both flavours of
2010: locals are initialized with values from the data or FP stack.
2011:
2012: Currently there is no way to define locals with user-defined data
2013: structures, but we are working on it.
2014:
2015: Gforth allows defining locals everywhere in a colon definition. This
2016: poses the following questions:
2017:
2018: @menu
2019: * Where are locals visible by name?::
2020: * How long do locals live?::
2021: * Programming Style::
2022: * Implementation::
2023: @end menu
2024:
2025: @node Where are locals visible by name?, How long do locals live?, Gforth locals, Gforth locals
2026: @subsubsection Where are locals visible by name?
2027: @cindex locals visibility
2028: @cindex visibility of locals
2029: @cindex scope of locals
2030:
2031: Basically, the answer is that locals are visible where you would expect
2032: it in block-structured languages, and sometimes a little longer. If you
2033: want to restrict the scope of a local, enclose its definition in
2034: @code{SCOPE}...@code{ENDSCOPE}.
2035:
2036: doc-scope
2037: doc-endscope
2038:
2039: These words behave like control structure words, so you can use them
2040: with @code{CS-PICK} and @code{CS-ROLL} to restrict the scope in
2041: arbitrary ways.
2042:
2043: If you want a more exact answer to the visibility question, here's the
2044: basic principle: A local is visible in all places that can only be
2045: reached through the definition of the local@footnote{In compiler
2046: construction terminology, all places dominated by the definition of the
2047: local.}. In other words, it is not visible in places that can be reached
2048: without going through the definition of the local. E.g., locals defined
2049: in @code{IF}...@code{ENDIF} are visible until the @code{ENDIF}, locals
2050: defined in @code{BEGIN}...@code{UNTIL} are visible after the
2051: @code{UNTIL} (until, e.g., a subsequent @code{ENDSCOPE}).
2052:
2053: The reasoning behind this solution is: We want to have the locals
2054: visible as long as it is meaningful. The user can always make the
2055: visibility shorter by using explicit scoping. In a place that can
2056: only be reached through the definition of a local, the meaning of a
2057: local name is clear. In other places it is not: How is the local
2058: initialized at the control flow path that does not contain the
2059: definition? Which local is meant, if the same name is defined twice in
2060: two independent control flow paths?
2061:
2062: This should be enough detail for nearly all users, so you can skip the
2063: rest of this section. If you really must know all the gory details and
2064: options, read on.
2065:
2066: In order to implement this rule, the compiler has to know which places
2067: are unreachable. It knows this automatically after @code{AHEAD},
2068: @code{AGAIN}, @code{EXIT} and @code{LEAVE}; in other cases (e.g., after
2069: most @code{THROW}s), you can use the word @code{UNREACHABLE} to tell the
2070: compiler that the control flow never reaches that place. If
2071: @code{UNREACHABLE} is not used where it could, the only consequence is
2072: that the visibility of some locals is more limited than the rule above
2073: says. If @code{UNREACHABLE} is used where it should not (i.e., if you
2074: lie to the compiler), buggy code will be produced.
2075:
2076: doc-unreachable
2077:
2078: Another problem with this rule is that at @code{BEGIN}, the compiler
2079: does not know which locals will be visible on the incoming
2080: back-edge. All problems discussed in the following are due to this
2081: ignorance of the compiler (we discuss the problems using @code{BEGIN}
2082: loops as examples; the discussion also applies to @code{?DO} and other
2083: loops). Perhaps the most insidious example is:
2084: @example
2085: AHEAD
2086: BEGIN
2087: x
2088: [ 1 CS-ROLL ] THEN
2089: @{ x @}
2090: ...
2091: UNTIL
2092: @end example
2093:
2094: This should be legal according to the visibility rule. The use of
2095: @code{x} can only be reached through the definition; but that appears
2096: textually below the use.
2097:
2098: From this example it is clear that the visibility rules cannot be fully
2099: implemented without major headaches. Our implementation treats common
2100: cases as advertised and the exceptions are treated in a safe way: The
2101: compiler makes a reasonable guess about the locals visible after a
2102: @code{BEGIN}; if it is too pessimistic, the
2103: user will get a spurious error about the local not being defined; if the
2104: compiler is too optimistic, it will notice this later and issue a
2105: warning. In the case above the compiler would complain about @code{x}
2106: being undefined at its use. You can see from the obscure examples in
2107: this section that it takes quite unusual control structures to get the
2108: compiler into trouble, and even then it will often do fine.
2109:
2110: If the @code{BEGIN} is reachable from above, the most optimistic guess
2111: is that all locals visible before the @code{BEGIN} will also be
2112: visible after the @code{BEGIN}. This guess is valid for all loops that
2113: are entered only through the @code{BEGIN}, in particular, for normal
2114: @code{BEGIN}...@code{WHILE}...@code{REPEAT} and
2115: @code{BEGIN}...@code{UNTIL} loops and it is implemented in our
2116: compiler. When the branch to the @code{BEGIN} is finally generated by
2117: @code{AGAIN} or @code{UNTIL}, the compiler checks the guess and
2118: warns the user if it was too optimistic:
2119: @example
2120: IF
2121: @{ x @}
2122: BEGIN
2123: \ x ?
2124: [ 1 cs-roll ] THEN
2125: ...
2126: UNTIL
2127: @end example
2128:
2129: Here, @code{x} lives only until the @code{BEGIN}, but the compiler
2130: optimistically assumes that it lives until the @code{THEN}. It notices
2131: this difference when it compiles the @code{UNTIL} and issues a
2132: warning. The user can avoid the warning, and make sure that @code{x}
2133: is not used in the wrong area by using explicit scoping:
2134: @example
2135: IF
2136: SCOPE
2137: @{ x @}
2138: ENDSCOPE
2139: BEGIN
2140: [ 1 cs-roll ] THEN
2141: ...
2142: UNTIL
2143: @end example
2144:
2145: Since the guess is optimistic, there will be no spurious error messages
2146: about undefined locals.
2147:
2148: If the @code{BEGIN} is not reachable from above (e.g., after
2149: @code{AHEAD} or @code{EXIT}), the compiler cannot even make an
2150: optimistic guess, as the locals visible after the @code{BEGIN} may be
2151: defined later. Therefore, the compiler assumes that no locals are
2152: visible after the @code{BEGIN}. However, the user can use
2153: @code{ASSUME-LIVE} to make the compiler assume that the same locals are
2154: visible at the BEGIN as at the point where the top control-flow stack
2155: item was created.
2156:
2157: doc-assume-live
2158:
2159: E.g.,
2160: @example
2161: @{ x @}
2162: AHEAD
2163: ASSUME-LIVE
2164: BEGIN
2165: x
2166: [ 1 CS-ROLL ] THEN
2167: ...
2168: UNTIL
2169: @end example
2170:
2171: Other cases where the locals are defined before the @code{BEGIN} can be
2172: handled by inserting an appropriate @code{CS-ROLL} before the
2173: @code{ASSUME-LIVE} (and changing the control-flow stack manipulation
2174: behind the @code{ASSUME-LIVE}).
2175:
2176: Cases where locals are defined after the @code{BEGIN} (but should be
2177: visible immediately after the @code{BEGIN}) can only be handled by
2178: rearranging the loop. E.g., the ``most insidious'' example above can be
2179: arranged into:
2180: @example
2181: BEGIN
2182: @{ x @}
2183: ... 0=
2184: WHILE
2185: x
2186: REPEAT
2187: @end example
2188:
2189: @node How long do locals live?, Programming Style, Where are locals visible by name?, Gforth locals
2190: @subsubsection How long do locals live?
2191: @cindex locals lifetime
2192: @cindex lifetime of locals
2193:
2194: The right answer for the lifetime question would be: A local lives at
2195: least as long as it can be accessed. For a value-flavoured local this
2196: means: until the end of its visibility. However, a variable-flavoured
2197: local could be accessed through its address far beyond its visibility
2198: scope. Ultimately, this would mean that such locals would have to be
2199: garbage collected. Since this entails un-Forth-like implementation
2200: complexities, I adopted the same cowardly solution as some other
2201: languages (e.g., C): The local lives only as long as it is visible;
2202: afterwards its address is invalid (and programs that access it
2203: afterwards are erroneous).
2204:
2205: @node Programming Style, Implementation, How long do locals live?, Gforth locals
2206: @subsubsection Programming Style
2207: @cindex locals programming style
2208: @cindex programming style, locals
2209:
2210: The freedom to define locals anywhere has the potential to change
2211: programming styles dramatically. In particular, the need to use the
2212: return stack for intermediate storage vanishes. Moreover, all stack
2213: manipulations (except @code{PICK}s and @code{ROLL}s with run-time
2214: determined arguments) can be eliminated: If the stack items are in the
2215: wrong order, just write a locals definition for all of them; then
2216: write the items in the order you want.
2217:
2218: This seems a little far-fetched and eliminating stack manipulations is
2219: unlikely to become a conscious programming objective. Still, the number
2220: of stack manipulations will be reduced dramatically if local variables
2221: are used liberally (e.g., compare @code{max} in @ref{Gforth locals} with
2222: a traditional implementation of @code{max}).
2223:
2224: This shows one potential benefit of locals: making Forth programs more
2225: readable. Of course, this benefit will only be realized if the
2226: programmers continue to honour the principle of factoring instead of
2227: using the added latitude to make the words longer.
2228:
2229: @cindex single-assignment style for locals
2230: Using @code{TO} can and should be avoided. Without @code{TO},
2231: every value-flavoured local has only a single assignment and many
2232: advantages of functional languages apply to Forth. I.e., programs are
2233: easier to analyse, to optimize and to read: It is clear from the
2234: definition what the local stands for, it does not turn into something
2235: different later.
2236:
2237: E.g., a definition using @code{TO} might look like this:
2238: @example
2239: : strcmp @{ addr1 u1 addr2 u2 -- n @}
2240: u1 u2 min 0
2241: ?do
2242: addr1 c@@ addr2 c@@ -
2243: ?dup-if
2244: unloop exit
2245: then
2246: addr1 char+ TO addr1
2247: addr2 char+ TO addr2
2248: loop
2249: u1 u2 - ;
2250: @end example
2251: Here, @code{TO} is used to update @code{addr1} and @code{addr2} at
2252: every loop iteration. @code{strcmp} is a typical example of the
2253: readability problems of using @code{TO}. When you start reading
2254: @code{strcmp}, you think that @code{addr1} refers to the start of the
2255: string. Only near the end of the loop you realize that it is something
2256: else.
2257:
2258: This can be avoided by defining two locals at the start of the loop that
2259: are initialized with the right value for the current iteration.
2260: @example
2261: : strcmp @{ addr1 u1 addr2 u2 -- n @}
2262: addr1 addr2
2263: u1 u2 min 0
2264: ?do @{ s1 s2 @}
2265: s1 c@@ s2 c@@ -
2266: ?dup-if
2267: unloop exit
2268: then
2269: s1 char+ s2 char+
2270: loop
2271: 2drop
2272: u1 u2 - ;
2273: @end example
2274: Here it is clear from the start that @code{s1} has a different value
2275: in every loop iteration.
2276:
2277: @node Implementation, , Programming Style, Gforth locals
2278: @subsubsection Implementation
2279: @cindex locals implementation
2280: @cindex implementation of locals
2281:
2282: @cindex locals stack
2283: Gforth uses an extra locals stack. The most compelling reason for
2284: this is that the return stack is not float-aligned; using an extra stack
2285: also eliminates the problems and restrictions of using the return stack
2286: as locals stack. Like the other stacks, the locals stack grows toward
2287: lower addresses. A few primitives allow an efficient implementation:
2288:
2289: doc-@local#
2290: doc-f@local#
2291: doc-laddr#
2292: doc-lp+!#
2293: doc-lp!
2294: doc->l
2295: doc-f>l
2296:
2297: In addition to these primitives, some specializations of these
2298: primitives for commonly occurring inline arguments are provided for
2299: efficiency reasons, e.g., @code{@@local0} as specialization of
2300: @code{@@local#} for the inline argument 0. The following compiling words
2301: compile the right specialized version, or the general version, as
2302: appropriate:
2303:
2304: doc-compile-@local
2305: doc-compile-f@local
2306: doc-compile-lp+!
2307:
2308: Combinations of conditional branches and @code{lp+!#} like
2309: @code{?branch-lp+!#} (the locals pointer is only changed if the branch
2310: is taken) are provided for efficiency and correctness in loops.
2311:
2312: A special area in the dictionary space is reserved for keeping the
2313: local variable names. @code{@{} switches the dictionary pointer to this
2314: area and @code{@}} switches it back and generates the locals
2315: initializing code. @code{W:} etc.@ are normal defining words. This
2316: special area is cleared at the start of every colon definition.
2317:
2318: @cindex wordlist for defining locals
2319: A special feature of Gforth's dictionary is used to implement the
2320: definition of locals without type specifiers: every wordlist (aka
2321: vocabulary) has its own methods for searching
2322: etc. (@pxref{Wordlists}). For the present purpose we defined a wordlist
2323: with a special search method: When it is searched for a word, it
2324: actually creates that word using @code{W:}. @code{@{} changes the search
2325: order to first search the wordlist containing @code{@}}, @code{W:} etc.,
2326: and then the wordlist for defining locals without type specifiers.
2327:
2328: The lifetime rules support a stack discipline within a colon
2329: definition: The lifetime of a local is either nested with other locals
2330: lifetimes or it does not overlap them.
2331:
2332: At @code{BEGIN}, @code{IF}, and @code{AHEAD} no code for locals stack
2333: pointer manipulation is generated. Between control structure words
2334: locals definitions can push locals onto the locals stack. @code{AGAIN}
2335: is the simplest of the other three control flow words. It has to
2336: restore the locals stack depth of the corresponding @code{BEGIN}
2337: before branching. The code looks like this:
2338: @format
2339: @code{lp+!#} current-locals-size @minus{} dest-locals-size
2340: @code{branch} <begin>
2341: @end format
2342:
2343: @code{UNTIL} is a little more complicated: If it branches back, it
2344: must adjust the stack just like @code{AGAIN}. But if it falls through,
2345: the locals stack must not be changed. The compiler generates the
2346: following code:
2347: @format
2348: @code{?branch-lp+!#} <begin> current-locals-size @minus{} dest-locals-size
2349: @end format
2350: The locals stack pointer is only adjusted if the branch is taken.
2351:
2352: @code{THEN} can produce somewhat inefficient code:
2353: @format
2354: @code{lp+!#} current-locals-size @minus{} orig-locals-size
2355: <orig target>:
2356: @code{lp+!#} orig-locals-size @minus{} new-locals-size
2357: @end format
2358: The second @code{lp+!#} adjusts the locals stack pointer from the
2359: level at the @var{orig} point to the level after the @code{THEN}. The
2360: first @code{lp+!#} adjusts the locals stack pointer from the current
2361: level to the level at the orig point, so the complete effect is an
2362: adjustment from the current level to the right level after the
2363: @code{THEN}.
2364:
2365: @cindex locals information on the control-flow stack
2366: @cindex control-flow stack items, locals information
2367: In a conventional Forth implementation a dest control-flow stack entry
2368: is just the target address and an orig entry is just the address to be
2369: patched. Our locals implementation adds a wordlist to every orig or dest
2370: item. It is the list of locals visible (or assumed visible) at the point
2371: described by the entry. Our implementation also adds a tag to identify
2372: the kind of entry, in particular to differentiate between live and dead
2373: (reachable and unreachable) orig entries.
2374:
2375: A few unusual operations have to be performed on locals wordlists:
2376:
2377: doc-common-list
2378: doc-sub-list?
2379: doc-list-size
2380:
2381: Several features of our locals wordlist implementation make these
2382: operations easy to implement: The locals wordlists are organised as
2383: linked lists; the tails of these lists are shared, if the lists
2384: contain some of the same locals; and the address of a name is greater
2385: than the address of the names behind it in the list.
2386:
2387: Another important implementation detail is the variable
2388: @code{dead-code}. It is used by @code{BEGIN} and @code{THEN} to
2389: determine if they can be reached directly or only through the branch
2390: that they resolve. @code{dead-code} is set by @code{UNREACHABLE},
2391: @code{AHEAD}, @code{EXIT} etc., and cleared at the start of a colon
2392: definition, by @code{BEGIN} and usually by @code{THEN}.
2393:
2394: Counted loops are similar to other loops in most respects, but
2395: @code{LEAVE} requires special attention: It performs basically the same
2396: service as @code{AHEAD}, but it does not create a control-flow stack
2397: entry. Therefore the information has to be stored elsewhere;
2398: traditionally, the information was stored in the target fields of the
2399: branches created by the @code{LEAVE}s, by organizing these fields into a
2400: linked list. Unfortunately, this clever trick does not provide enough
2401: space for storing our extended control flow information. Therefore, we
2402: introduce another stack, the leave stack. It contains the control-flow
2403: stack entries for all unresolved @code{LEAVE}s.
2404:
2405: Local names are kept until the end of the colon definition, even if
2406: they are no longer visible in any control-flow path. In a few cases
2407: this may lead to increased space needs for the locals name area, but
2408: usually less than reclaiming this space would cost in code size.
2409:
2410:
2411: @node ANS Forth locals, , Gforth locals, Locals
2412: @subsection ANS Forth locals
2413: @cindex locals, ANS Forth style
2414:
2415: The ANS Forth locals wordset does not define a syntax for locals, but
2416: words that make it possible to define various syntaxes. One of the
2417: possible syntaxes is a subset of the syntax we used in the Gforth locals
2418: wordset, i.e.:
2419:
2420: @example
2421: @{ local1 local2 ... -- comment @}
2422: @end example
2423: or
2424: @example
2425: @{ local1 local2 ... @}
2426: @end example
2427:
2428: The order of the locals corresponds to the order in a stack comment. The
2429: restrictions are:
2430:
2431: @itemize @bullet
2432: @item
2433: Locals can only be cell-sized values (no type specifiers are allowed).
2434: @item
2435: Locals can be defined only outside control structures.
2436: @item
2437: Locals can interfere with explicit usage of the return stack. For the
2438: exact (and long) rules, see the standard. If you don't use return stack
2439: accessing words in a definition using locals, you will be all right. The
2440: purpose of this rule is to make locals implementation on the return
2441: stack easier.
2442: @item
2443: The whole definition must be in one line.
2444: @end itemize
2445:
2446: Locals defined in this way behave like @code{VALUE}s (@xref{Simple
2447: Defining Words}). I.e., they are initialized from the stack. Using their
2448: name produces their value. Their value can be changed using @code{TO}.
2449:
2450: Since this syntax is supported by Gforth directly, you need not do
2451: anything to use it. If you want to port a program using this syntax to
2452: another ANS Forth system, use @file{compat/anslocal.fs} to implement the
2453: syntax on the other system.
2454:
2455: Note that a syntax shown in the standard, section A.13 looks
2456: similar, but is quite different in having the order of locals
2457: reversed. Beware!
2458:
2459: The ANS Forth locals wordset itself consists of the following word
2460:
2461: doc-(local)
2462:
2463: The ANS Forth locals extension wordset defines a syntax, but it is so
2464: awful that we strongly recommend not to use it. We have implemented this
2465: syntax to make porting to Gforth easy, but do not document it here. The
2466: problem with this syntax is that the locals are defined in an order
2467: reversed with respect to the standard stack comment notation, making
2468: programs harder to read, and easier to misread and miswrite. The only
2469: merit of this syntax is that it is easy to implement using the ANS Forth
2470: locals wordset.
2471:
1.5 anton 2472: @node Defining Words, Structures, Locals, Words
1.1 anton 2473: @section Defining Words
2474: @cindex defining words
2475:
2476: @menu
2477: * Simple Defining Words::
2478: * Colon Definitions::
2479: * User-defined Defining Words::
2480: * Supplying names::
2481: * Interpretation and Compilation Semantics::
2482: @end menu
2483:
2484: @node Simple Defining Words, Colon Definitions, Defining Words, Defining Words
2485: @subsection Simple Defining Words
2486: @cindex simple defining words
2487: @cindex defining words, simple
2488:
2489: doc-constant
2490: doc-2constant
2491: doc-fconstant
2492: doc-variable
2493: doc-2variable
2494: doc-fvariable
2495: doc-create
2496: doc-user
2497: doc-value
2498: doc-to
2499: doc-defer
2500: doc-is
2501:
2502: @node Colon Definitions, User-defined Defining Words, Simple Defining Words, Defining Words
2503: @subsection Colon Definitions
2504: @cindex colon definitions
2505:
2506: @example
2507: : name ( ... -- ... )
2508: word1 word2 word3 ;
2509: @end example
2510:
2511: creates a word called @code{name}, that, upon execution, executes
2512: @code{word1 word2 word3}. @code{name} is a @dfn{(colon) definition}.
2513:
2514: The explanation above is somewhat superficial. @xref{Interpretation and
2515: Compilation Semantics} for an in-depth discussion of some of the issues
2516: involved.
2517:
2518: doc-:
2519: doc-;
2520:
2521: @node User-defined Defining Words, Supplying names, Colon Definitions, Defining Words
2522: @subsection User-defined Defining Words
2523: @cindex user-defined defining words
2524: @cindex defining words, user-defined
2525:
2526: You can create new defining words simply by wrapping defining-time code
2527: around existing defining words and putting the sequence in a colon
2528: definition.
2529:
2530: @cindex @code{CREATE} ... @code{DOES>}
2531: If you want the words defined with your defining words to behave
2532: differently from words defined with standard defining words, you can
2533: write your defining word like this:
2534:
2535: @example
2536: : def-word ( "name" -- )
2537: Create @var{code1}
2538: DOES> ( ... -- ... )
2539: @var{code2} ;
2540:
2541: def-word name
2542: @end example
2543:
2544: Technically, this fragment defines a defining word @code{def-word}, and
2545: a word @code{name}; when you execute @code{name}, the address of the
2546: body of @code{name} is put on the data stack and @var{code2} is executed
2547: (the address of the body of @code{name} is the address @code{HERE}
2548: returns immediately after the @code{CREATE}).
2549:
2550: In other words, if you make the following definitions:
2551:
2552: @example
2553: : def-word1 ( "name" -- )
2554: Create @var{code1} ;
2555:
2556: : action1 ( ... -- ... )
2557: @var{code2} ;
2558:
2559: def-word name1
2560: @end example
2561:
2562: Using @code{name1 action1} is equivalent to using @code{name}.
2563:
2564: E.g., you can implement @code{Constant} in this way:
2565:
2566: @example
2567: : constant ( w "name" -- )
2568: create ,
2569: DOES> ( -- w )
2570: @@ ;
2571: @end example
2572:
2573: When you create a constant with @code{5 constant five}, first a new word
2574: @code{five} is created, then the value 5 is laid down in the body of
2575: @code{five} with @code{,}. When @code{five} is invoked, the address of
2576: the body is put on the stack, and @code{@@} retrieves the value 5.
2577:
2578: @cindex stack effect of @code{DOES>}-parts
2579: @cindex @code{DOES>}-parts, stack effect
2580: In the example above the stack comment after the @code{DOES>} specifies
2581: the stack effect of the defined words, not the stack effect of the
2582: following code (the following code expects the address of the body on
2583: the top of stack, which is not reflected in the stack comment). This is
2584: the convention that I use and recommend (it clashes a bit with using
2585: locals declarations for stack effect specification, though).
2586:
2587: @subsubsection Applications of @code{CREATE..DOES>}
2588: @cindex @code{CREATE} ... @code{DOES>}, applications
2589:
2590: You may wonder how to use this feature. Here are some usage patterns:
2591:
2592: @cindex factoring similar colon definitions
2593: When you see a sequence of code occurring several times, and you can
2594: identify a meaning, you will factor it out as a colon definition. When
2595: you see similar colon definitions, you can factor them using
2596: @code{CREATE..DOES>}. E.g., an assembler usually defines several words
2597: that look very similar:
2598: @example
2599: : ori, ( reg-target reg-source n -- )
2600: 0 asm-reg-reg-imm ;
2601: : andi, ( reg-target reg-source n -- )
2602: 1 asm-reg-reg-imm ;
2603: @end example
2604:
2605: This could be factored with:
2606: @example
2607: : reg-reg-imm ( op-code -- )
2608: create ,
2609: DOES> ( reg-target reg-source n -- )
2610: @@ asm-reg-reg-imm ;
2611:
2612: 0 reg-reg-imm ori,
2613: 1 reg-reg-imm andi,
2614: @end example
2615:
2616: @cindex currying
2617: Another view of @code{CREATE..DOES>} is to consider it as a crude way to
2618: supply a part of the parameters for a word (known as @dfn{currying} in
2619: the functional language community). E.g., @code{+} needs two
2620: parameters. Creating versions of @code{+} with one parameter fixed can
2621: be done like this:
2622: @example
2623: : curry+ ( n1 -- )
2624: create ,
2625: DOES> ( n2 -- n1+n2 )
2626: @@ + ;
2627:
2628: 3 curry+ 3+
2629: -2 curry+ 2-
2630: @end example
2631:
2632: @subsubsection The gory details of @code{CREATE..DOES>}
2633: @cindex @code{CREATE} ... @code{DOES>}, details
2634:
2635: doc-does>
2636:
2637: @cindex @code{DOES>} in a separate definition
2638: This means that you need not use @code{CREATE} and @code{DOES>} in the
2639: same definition; E.g., you can put the @code{DOES>}-part in a separate
2640: definition. This allows us to, e.g., select among different DOES>-parts:
2641: @example
2642: : does1
2643: DOES> ( ... -- ... )
2644: ... ;
2645:
2646: : does2
2647: DOES> ( ... -- ... )
2648: ... ;
2649:
2650: : def-word ( ... -- ... )
2651: create ...
2652: IF
2653: does1
2654: ELSE
2655: does2
2656: ENDIF ;
2657: @end example
2658:
2659: @cindex @code{DOES>} in interpretation state
2660: In a standard program you can apply a @code{DOES>}-part only if the last
2661: word was defined with @code{CREATE}. In Gforth, the @code{DOES>}-part
2662: will override the behaviour of the last word defined in any case. In a
2663: standard program, you can use @code{DOES>} only in a colon
2664: definition. In Gforth, you can also use it in interpretation state, in a
2665: kind of one-shot mode:
2666: @example
2667: CREATE name ( ... -- ... )
2668: @var{initialization}
2669: DOES>
2670: @var{code} ;
2671: @end example
2672: This is equivalent to the standard
2673: @example
2674: :noname
2675: DOES>
2676: @var{code} ;
2677: CREATE name EXECUTE ( ... -- ... )
2678: @var{initialization}
2679: @end example
2680:
2681: You can get the address of the body of a word with
2682:
2683: doc->body
2684:
2685: @node Supplying names, Interpretation and Compilation Semantics, User-defined Defining Words, Defining Words
2686: @subsection Supplying names for the defined words
2687: @cindex names for defined words
2688: @cindex defining words, name parameter
2689:
2690: @cindex defining words, name given in a string
2691: By default, defining words take the names for the defined words from the
2692: input stream. Sometimes you want to supply the name from a string. You
2693: can do this with
2694:
2695: doc-nextname
2696:
2697: E.g.,
2698:
2699: @example
2700: s" foo" nextname create
2701: @end example
2702: is equivalent to
2703: @example
2704: create foo
2705: @end example
2706:
2707: @cindex defining words without name
2708: Sometimes you want to define a word without a name. You can do this with
2709:
2710: doc-noname
2711:
2712: @cindex execution token of last defined word
2713: To make any use of the newly defined word, you need its execution
2714: token. You can get it with
2715:
2716: doc-lastxt
2717:
2718: E.g., you can initialize a deferred word with an anonymous colon
2719: definition:
2720: @example
2721: Defer deferred
2722: noname : ( ... -- ... )
2723: ... ;
2724: lastxt IS deferred
2725: @end example
2726:
2727: @code{lastxt} also works when the last word was not defined as
2728: @code{noname}.
2729:
2730: The standard has also recognized the need for anonymous words and
2731: provides
2732:
2733: doc-:noname
2734:
2735: This leaves the execution token for the word on the stack after the
2736: closing @code{;}. You can rewrite the last example with @code{:noname}:
2737: @example
2738: Defer deferred
2739: :noname ( ... -- ... )
2740: ... ;
2741: IS deferred
2742: @end example
2743:
2744: @node Interpretation and Compilation Semantics, , Supplying names, Defining Words
2745: @subsection Interpretation and Compilation Semantics
2746: @cindex semantics, interpretation and compilation
2747:
2748: @cindex interpretation semantics
2749: The @dfn{interpretation semantics} of a word are what the text
2750: interpreter does when it encounters the word in interpret state. It also
2751: appears in some other contexts, e.g., the execution token returned by
2752: @code{' @var{word}} identifies the interpretation semantics of
2753: @var{word} (in other words, @code{' @var{word} execute} is equivalent to
2754: interpret-state text interpretation of @code{@var{word}}).
2755:
2756: @cindex compilation semantics
2757: The @dfn{compilation semantics} of a word are what the text interpreter
2758: does when it encounters the word in compile state. It also appears in
2759: other contexts, e.g, @code{POSTPONE @var{word}} compiles@footnote{In
2760: standard terminology, ``appends to the current definition''.} the
2761: compilation semantics of @var{word}.
2762:
2763: @cindex execution semantics
2764: The standard also talks about @dfn{execution semantics}. They are used
2765: only for defining the interpretation and compilation semantics of many
2766: words. By default, the interpretation semantics of a word are to
2767: @code{execute} its execution semantics, and the compilation semantics of
2768: a word are to @code{compile,} its execution semantics.@footnote{In
2769: standard terminology: The default interpretation semantics are its
2770: execution semantics; the default compilation semantics are to append its
2771: execution semantics to the execution semantics of the current
2772: definition.}
2773:
2774: @cindex immediate words
2775: You can change the compilation semantics into @code{execute}ing the
2776: execution semantics with
2777:
2778: doc-immediate
2779:
2780: @cindex compile-only words
2781: You can remove the interpretation semantics of a word with
2782:
2783: doc-compile-only
2784: doc-restrict
2785:
2786: Note that ticking (@code{'}) compile-only words gives an error
2787: (``Interpreting a compile-only word'').
2788:
2789: Gforth also allows you to define words with arbitrary combinations of
2790: interpretation and compilation semantics.
2791:
2792: doc-interpret/compile:
2793:
2794: This feature was introduced for implementing @code{TO} and @code{S"}. I
2795: recommend that you do not define such words, as cute as they may be:
2796: they make it hard to get at both parts of the word in some contexts.
2797: E.g., assume you want to get an execution token for the compilation
2798: part. Instead, define two words, one that embodies the interpretation
1.15 anton 2799: part, and one that embodies the compilation part. Once you have done
2800: that, you can define a combined word with @code{interpret/compile:} for
2801: the convenience of your users.
1.1 anton 2802:
1.15 anton 2803: You also might try to provide an optimizing implementation of the
2804: default compilation semantics with this feature, like this:
1.1 anton 2805:
2806: @example
2807: :noname
2808: foo bar ;
2809: :noname
2810: POSTPONE foo POSTPONE bar ;
2811: interpret/compile: foobar
2812: @end example
2813:
1.15 anton 2814: as an optimizing version of
2815:
2816: @example
2817: : foobar
2818: foo bar ;
2819: @end example
2820:
2821: Unfortunately, this does not work correctly with @code{[compile]},
2822: because @code{[compile]} assumes that the compilation semantics of all
2823: @code{interpret/compile:} words are non-default. I.e., @code{[compile]
2824: foobar} would compile the compilation semantics for the optimizing
2825: @code{foobar}, whereas it would compile the interpretation semantics for
2826: the non-optimizing @code{foobar}.
1.1 anton 2827:
2828: @cindex state-smart words are a bad idea
2829: Some people try to use state-smart words to emulate the feature provided
2830: by @code{interpret/compile:} (words are state-smart if they check
2831: @code{STATE} during execution). E.g., they would try to code
2832: @code{foobar} like this:
2833:
2834: @example
2835: : foobar
2836: STATE @@
2837: IF ( compilation state )
2838: POSTPONE foo POSTPONE bar
2839: ELSE
2840: foo bar
2841: ENDIF ; immediate
2842: @end example
2843:
2844: While this works if @code{foobar} is processed only by the text
2845: interpreter, it does not work in other contexts (like @code{'} or
2846: @code{POSTPONE}). E.g., @code{' foobar} will produce an execution token
2847: for a state-smart word, not for the interpretation semantics of the
2848: original @code{foobar}; when you execute this execution token (directly
2849: with @code{EXECUTE} or indirectly through @code{COMPILE,}) in compile
2850: state, the result will not be what you expected (i.e., it will not
2851: perform @code{foo bar}). State-smart words are a bad idea. Simply don't
2852: write them!
2853:
2854: @cindex defining words with arbitrary semantics combinations
2855: It is also possible to write defining words that define words with
1.15 anton 2856: arbitrary combinations of interpretation and compilation semantics. In
2857: general, this looks like:
1.1 anton 2858:
2859: @example
2860: : def-word
2861: create-interpret/compile
2862: @var{code1}
2863: interpretation>
2864: @var{code2}
2865: <interpretation
2866: compilation>
2867: @var{code3}
2868: <compilation ;
2869: @end example
2870:
2871: For a @var{word} defined with @code{def-word}, the interpretation
2872: semantics are to push the address of the body of @var{word} and perform
2873: @var{code2}, and the compilation semantics are to push the address of
2874: the body of @var{word} and perform @var{code3}. E.g., @code{constant}
1.15 anton 2875: can also be defined like this (except that the defined constants don't
2876: behave correctly when @code{[compile]}d):
1.1 anton 2877:
2878: @example
2879: : constant ( n "name" -- )
2880: create-interpret/compile
2881: ,
2882: interpretation> ( -- n )
2883: @@
2884: <interpretation
2885: compilation> ( compilation. -- ; run-time. -- n )
2886: @@ postpone literal
2887: <compilation ;
2888: @end example
2889:
2890: doc-create-interpret/compile
2891: doc-interpretation>
2892: doc-<interpretation
2893: doc-compilation>
2894: doc-<compilation
2895:
2896: Note that words defined with @code{interpret/compile:} and
2897: @code{create-interpret/compile} have an extended header structure that
2898: differs from other words; however, unless you try to access them with
2899: plain address arithmetic, you should not notice this. Words for
2900: accessing the header structure usually know how to deal with this; e.g.,
2901: @code{' word >body} also gives you the body of a word created with
2902: @code{create-interpret/compile}.
2903:
1.5 anton 2904: @c ----------------------------------------------------------
1.12 anton 2905: @node Structures, Object-oriented Forth, Defining Words, Words
1.5 anton 2906: @section Structures
2907: @cindex structures
2908: @cindex records
2909:
2910: This section presents the structure package that comes with Gforth. A
2911: version of the package implemented in plain ANS Forth is available in
2912: @file{compat/struct.fs}. This package was inspired by a posting on
2913: comp.lang.forth in 1989 (unfortunately I don't remember, by whom;
2914: possibly John Hayes). A version of this section has been published in
2915: ???. Marcel Hendrix provided helpful comments.
2916:
2917: @menu
2918: * Why explicit structure support?::
2919: * Structure Usage::
2920: * Structure Naming Convention::
2921: * Structure Implementation::
2922: * Structure Glossary::
2923: @end menu
2924:
2925: @node Why explicit structure support?, Structure Usage, Structures, Structures
2926: @subsection Why explicit structure support?
2927:
2928: @cindex address arithmetic for structures
2929: @cindex structures using address arithmetic
2930: If we want to use a structure containing several fields, we could simply
2931: reserve memory for it, and access the fields using address arithmetic
2932: (@pxref{Address arithmetic}). As an example, consider a structure with
2933: the following fields
2934:
2935: @table @code
2936: @item a
2937: is a float
2938: @item b
2939: is a cell
2940: @item c
2941: is a float
2942: @end table
2943:
2944: Given the (float-aligned) base address of the structure we get the
2945: address of the field
2946:
2947: @table @code
2948: @item a
2949: without doing anything further.
2950: @item b
2951: with @code{float+}
2952: @item c
2953: with @code{float+ cell+ faligned}
2954: @end table
2955:
2956: It is easy to see that this can become quite tiring.
2957:
2958: Moreover, it is not very readable, because seeing a
2959: @code{cell+} tells us neither which kind of structure is
2960: accessed nor what field is accessed; we have to somehow infer the kind
2961: of structure, and then look up in the documentation, which field of
2962: that structure corresponds to that offset.
2963:
2964: Finally, this kind of address arithmetic also causes maintenance
2965: troubles: If you add or delete a field somewhere in the middle of the
2966: structure, you have to find and change all computations for the fields
2967: afterwards.
2968:
2969: So, instead of using @code{cell+} and friends directly, how
2970: about storing the offsets in constants:
2971:
2972: @example
2973: 0 constant a-offset
2974: 0 float+ constant b-offset
2975: 0 float+ cell+ faligned c-offset
2976: @end example
2977:
2978: Now we can get the address of field @code{x} with @code{x-offset
2979: +}. This is much better in all respects. Of course, you still
2980: have to change all later offset definitions if you add a field. You can
2981: fix this by declaring the offsets in the following way:
2982:
2983: @example
2984: 0 constant a-offset
2985: a-offset float+ constant b-offset
2986: b-offset cell+ faligned constant c-offset
2987: @end example
2988:
2989: Since we always use the offsets with @code{+}, using a defining
2990: word @code{cfield} that includes the @code{+} in the
2991: action of the defined word offers itself:
2992:
2993: @example
2994: : cfield ( n "name" -- )
2995: create ,
2996: does> ( name execution: addr1 -- addr2 )
2997: @@ + ;
2998:
2999: 0 cfield a
3000: 0 a float+ cfield b
3001: 0 b cell+ faligned cfield c
3002: @end example
3003:
3004: Instead of @code{x-offset +}, we now simply write @code{x}.
3005:
3006: The structure field words now can be used quite nicely. However,
3007: their definition is still a bit cumbersome: We have to repeat the
3008: name, the information about size and alignment is distributed before
3009: and after the field definitions etc. The structure package presented
3010: here addresses these problems.
3011:
3012: @node Structure Usage, Structure Naming Convention, Why explicit structure support?, Structures
3013: @subsection Structure Usage
3014: @cindex structure usage
3015:
3016: @cindex @code{field} usage
3017: @cindex @code{struct} usage
3018: @cindex @code{end-struct} usage
3019: You can define a structure for a (data-less) linked list with
3020: @example
3021: struct
3022: cell% field list-next
3023: end-struct list%
3024: @end example
3025:
3026: With the address of the list node on the stack, you can compute the
3027: address of the field that contains the address of the next node with
3028: @code{list-next}. E.g., you can determine the length of a list
3029: with:
3030:
3031: @example
3032: : list-length ( list -- n )
3033: \ "list" is a pointer to the first element of a linked list
3034: \ "n" is the length of the list
3035: 0 begin ( list1 n1 )
3036: over
3037: while ( list1 n1 )
3038: 1+ swap list-next @@ swap
3039: repeat
3040: nip ;
3041: @end example
3042:
3043: You can reserve memory for a list node in the dictionary with
3044: @code{list% %allot}, which leaves the address of the list node on the
3045: stack. For the equivalent allocation on the heap you can use @code{list%
3046: %alloc} (or, for an @code{allocate}-like stack effect (i.e., with ior),
3047: use @code{list% %allocate}). You can also get the the size of a list
3048: node with @code{list% %size} and it's alignment with @code{list%
3049: %alignment}.
3050:
3051: Note that in ANS Forth the body of a @code{create}d word is
3052: @code{aligned} but not necessarily @code{faligned};
3053: therefore, if you do a
3054: @example
3055: create @emph{name} foo% %allot
3056: @end example
3057:
3058: then the memory alloted for @code{foo%} is
3059: guaranteed to start at the body of @code{@emph{name}} only if
3060: @code{foo%} contains only character, cell and double fields.
3061:
3062: @cindex strcutures containing structures
3063: You can also include a structure @code{foo%} as field of
3064: another structure, with:
3065: @example
3066: struct
3067: ...
3068: foo% field ...
3069: ...
3070: end-struct ...
3071: @end example
3072:
3073: @cindex structure extension
3074: @cindex extended records
3075: Instead of starting with an empty structure, you can also extend an
3076: existing structure. E.g., a plain linked list without data, as defined
3077: above, is hardly useful; You can extend it to a linked list of integers,
3078: like this:@footnote{This feature is also known as @emph{extended
3079: records}. It is the main innovation in the Oberon language; in other
3080: words, adding this feature to Modula-2 led Wirth to create a new
3081: language, write a new compiler etc. Adding this feature to Forth just
3082: requires a few lines of code.}
3083:
3084: @example
3085: list%
3086: cell% field intlist-int
3087: end-struct intlist%
3088: @end example
3089:
3090: @code{intlist%} is a structure with two fields:
3091: @code{list-next} and @code{intlist-int}.
3092:
3093: @cindex structures containing arrays
3094: You can specify an array type containing @emph{n} elements of
3095: type @code{foo%} like this:
3096:
3097: @example
3098: foo% @emph{n} *
3099: @end example
3100:
3101: You can use this array type in any place where you can use a normal
3102: type, e.g., when defining a @code{field}, or with
3103: @code{%allot}.
3104:
3105: @cindex first field optimization
3106: The first field is at the base address of a structure and the word
3107: for this field (e.g., @code{list-next}) actually does not change
3108: the address on the stack. You may be tempted to leave it away in the
3109: interest of run-time and space efficiency. This is not necessary,
3110: because the structure package optimizes this case and compiling such
3111: words does not generate any code. So, in the interest of readability
3112: and maintainability you should include the word for the field when
3113: accessing the field.
3114:
3115: @node Structure Naming Convention, Structure Implementation, Structure Usage, Structures
3116: @subsection Structure Naming Convention
3117: @cindex structure naming conventions
3118:
3119: The field names that come to (my) mind are often quite generic, and,
3120: if used, would cause frequent name clashes. E.g., many structures
3121: probably contain a @code{counter} field. The structure names
3122: that come to (my) mind are often also the logical choice for the names
3123: of words that create such a structure.
3124:
3125: Therefore, I have adopted the following naming conventions:
3126:
3127: @itemize @bullet
3128: @cindex field naming convention
3129: @item
3130: The names of fields are of the form
3131: @code{@emph{struct}-@emph{field}}, where
3132: @code{@emph{struct}} is the basic name of the structure, and
3133: @code{@emph{field}} is the basic name of the field. You can
3134: think about field words as converting converts the (address of the)
3135: structure into the (address of the) field.
3136:
3137: @cindex structure naming convention
3138: @item
3139: The names of structures are of the form
3140: @code{@emph{struct}%}, where
3141: @code{@emph{struct}} is the basic name of the structure.
3142: @end itemize
3143:
3144: This naming convention does not work that well for fields of extended
3145: structures; e.g., the integer list structure has a field
3146: @code{intlist-int}, but has @code{list-next}, not
3147: @code{intlist-next}.
3148:
3149: @node Structure Implementation, Structure Glossary, Structure Naming Convention, Structures
3150: @subsection Structure Implementation
3151: @cindex structure implementation
3152: @cindex implementation of structures
3153:
3154: The central idea in the implementation is to pass the data about the
3155: structure being built on the stack, not in some global
3156: variable. Everything else falls into place naturally once this design
3157: decision is made.
3158:
3159: The type description on the stack is of the form @emph{align
3160: size}. Keeping the size on the top-of-stack makes dealing with arrays
3161: very simple.
3162:
3163: @code{field} is a defining word that uses @code{create}
3164: and @code{does>}. The body of the field contains the offset
3165: of the field, and the normal @code{does>} action is
3166:
3167: @example
3168: @ +
3169: @end example
3170:
3171: i.e., add the offset to the address, giving the stack effect
3172: @code{addr1 -- addr2} for a field.
3173:
3174: @cindex first field optimization, implementation
3175: This simple structure is slightly complicated by the optimization
3176: for fields with offset 0, which requires a different
3177: @code{does>}-part (because we cannot rely on there being
3178: something on the stack if such a field is invoked during
3179: compilation). Therefore, we put the different @code{does>}-parts
3180: in separate words, and decide which one to invoke based on the
3181: offset. For a zero offset, the field is basically a noop; it is
3182: immediate, and therefore no code is generated when it is compiled.
3183:
3184: @node Structure Glossary, , Structure Implementation, Structures
3185: @subsection Structure Glossary
3186: @cindex structure glossary
3187:
3188: doc-%align
3189: doc-%alignment
3190: doc-%alloc
3191: doc-%allocate
3192: doc-%allot
3193: doc-cell%
3194: doc-char%
3195: doc-dfloat%
3196: doc-double%
3197: doc-end-struct
3198: doc-field
3199: doc-float%
3200: doc-nalign
3201: doc-sfloat%
3202: doc-%size
3203: doc-struct
3204:
3205: @c -------------------------------------------------------------
1.12 anton 3206: @node Object-oriented Forth, Tokens for Words, Structures, Words
3207: @section Object-oriented Forth
3208:
3209: Gforth comes with three packets for object-oriented programming,
3210: @file{objects.fs}, @file{oof.fs}, and @file{mini-oof.fs}; none of them
3211: is preloaded, so you have to @code{include} them before use. The most
3212: important differences between these packets (and others) are discussed
3213: in @ref{Comparison with other object models}. All packets are written
3214: in ANS Forth and can be used with any other ANS Forth.
3215:
3216: @menu
3217: * Objects::
3218: * OOF::
3219: * Mini-OOF::
3220: @end menu
3221:
3222: @node Objects, OOF, Object-oriented Forth, Object-oriented Forth
3223: @subsection Objects
1.5 anton 3224: @cindex objects
3225: @cindex object-oriented programming
3226:
3227: @cindex @file{objects.fs}
3228: @cindex @file{oof.fs}
1.12 anton 3229:
3230: 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}).
1.5 anton 3231: @c McKewan's and Zsoter's packages
3232:
3233: This section assumes (in some places) that you have read @ref{Structures}.
3234:
3235: @menu
3236: * Properties of the Objects model::
3237: * Why object-oriented programming?::
3238: * Object-Oriented Terminology::
3239: * Basic Objects Usage::
3240: * The class Object::
3241: * Creating objects::
3242: * Object-Oriented Programming Style::
3243: * Class Binding::
3244: * Method conveniences::
3245: * Classes and Scoping::
3246: * Object Interfaces::
3247: * Objects Implementation::
3248: * Comparison with other object models::
3249: * Objects Glossary::
3250: @end menu
3251:
3252: Marcel Hendrix provided helpful comments on this section. Andras Zsoter
3253: and Bernd Paysan helped me with the related works section.
3254:
3255: @node Properties of the Objects model, Why object-oriented programming?, Objects, Objects
1.12 anton 3256: @subsubsection Properties of the @file{objects.fs} model
1.5 anton 3257: @cindex @file{objects.fs} properties
3258:
3259: @itemize @bullet
3260: @item
3261: It is straightforward to pass objects on the stack. Passing
3262: selectors on the stack is a little less convenient, but possible.
3263:
3264: @item
3265: Objects are just data structures in memory, and are referenced by
3266: their address. You can create words for objects with normal defining
3267: words like @code{constant}. Likewise, there is no difference
3268: between instance variables that contain objects and those
3269: that contain other data.
3270:
3271: @item
3272: Late binding is efficient and easy to use.
3273:
3274: @item
3275: It avoids parsing, and thus avoids problems with state-smartness
3276: and reduced extensibility; for convenience there are a few parsing
3277: words, but they have non-parsing counterparts. There are also a few
3278: defining words that parse. This is hard to avoid, because all standard
3279: defining words parse (except @code{:noname}); however, such
3280: words are not as bad as many other parsing words, because they are not
3281: state-smart.
3282:
3283: @item
3284: It does not try to incorporate everything. It does a few things
3285: and does them well (IMO). In particular, I did not intend to support
3286: information hiding with this model (although it has features that may
3287: help); you can use a separate package for achieving this.
3288:
3289: @item
3290: It is layered; you don't have to learn and use all features to use this
3291: model. Only a few features are necessary (@xref{Basic Objects Usage},
3292: @xref{The class Object}, @xref{Creating objects}.), the others
3293: are optional and independent of each other.
3294:
3295: @item
3296: An implementation in ANS Forth is available.
3297:
3298: @end itemize
3299:
3300: I have used the technique, on which this model is based, for
3301: implementing the parser generator Gray; we have also used this technique
3302: in Gforth for implementing the various flavours of wordlists (hashed or
3303: not, case-sensitive or not, special-purpose wordlists for locals etc.).
3304:
3305: @node Why object-oriented programming?, Object-Oriented Terminology, Properties of the Objects model, Objects
1.12 anton 3306: @subsubsection Why object-oriented programming?
1.5 anton 3307: @cindex object-oriented programming motivation
3308: @cindex motivation for object-oriented programming
3309:
3310: Often we have to deal with several data structures (@emph{objects}),
3311: that have to be treated similarly in some respects, but differ in
3312: others. Graphical objects are the textbook example: circles,
3313: triangles, dinosaurs, icons, and others, and we may want to add more
3314: during program development. We want to apply some operations to any
3315: graphical object, e.g., @code{draw} for displaying it on the
3316: screen. However, @code{draw} has to do something different for
3317: every kind of object.
3318:
3319: We could implement @code{draw} as a big @code{CASE}
3320: control structure that executes the appropriate code depending on the
3321: kind of object to be drawn. This would be not be very elegant, and,
3322: moreover, we would have to change @code{draw} every time we add
3323: a new kind of graphical object (say, a spaceship).
3324:
3325: What we would rather do is: When defining spaceships, we would tell
3326: the system: "Here's how you @code{draw} a spaceship; you figure
3327: out the rest."
3328:
3329: This is the problem that all systems solve that (rightfully) call
3330: themselves object-oriented, and the object-oriented package I present
3331: here also solves this problem (and not much else).
3332:
3333: @node Object-Oriented Terminology, Basic Objects Usage, Why object-oriented programming?, Objects
1.12 anton 3334: @subsubsection Object-Oriented Terminology
1.5 anton 3335: @cindex object-oriented terminology
3336: @cindex terminology for object-oriented programming
3337:
3338: This section is mainly for reference, so you don't have to understand
3339: all of it right away. The terminology is mainly Smalltalk-inspired. In
3340: short:
3341:
3342: @table @emph
3343: @cindex class
3344: @item class
3345: a data structure definition with some extras.
3346:
3347: @cindex object
3348: @item object
3349: an instance of the data structure described by the class definition.
3350:
3351: @cindex instance variables
3352: @item instance variables
3353: fields of the data structure.
3354:
3355: @cindex selector
3356: @cindex method selector
3357: @cindex virtual function
3358: @item selector
3359: (or @emph{method selector}) a word (e.g.,
3360: @code{draw}) for performing an operation on a variety of data
3361: structures (classes). A selector describes @emph{what} operation to
3362: perform. In C++ terminology: a (pure) virtual function.
3363:
3364: @cindex method
3365: @item method
3366: the concrete definition that performs the operation
3367: described by the selector for a specific class. A method specifies
3368: @emph{how} the operation is performed for a specific class.
3369:
3370: @cindex selector invocation
3371: @cindex message send
3372: @cindex invoking a selector
3373: @item selector invocation
3374: a call of a selector. One argument of the call (the TOS (top-of-stack))
3375: is used for determining which method is used. In Smalltalk terminology:
3376: a message (consisting of the selector and the other arguments) is sent
3377: to the object.
3378:
3379: @cindex receiving object
3380: @item receiving object
3381: the object used for determining the method executed by a selector
3382: invocation. In our model it is the object that is on the TOS when the
3383: selector is invoked. (@emph{Receiving} comes from Smalltalks
3384: @emph{message} terminology.)
3385:
3386: @cindex child class
3387: @cindex parent class
3388: @cindex inheritance
3389: @item child class
3390: a class that has (@emph{inherits}) all properties (instance variables,
3391: selectors, methods) from a @emph{parent class}. In Smalltalk
3392: terminology: The subclass inherits from the superclass. In C++
3393: terminology: The derived class inherits from the base class.
3394:
3395: @end table
3396:
3397: @c If you wonder about the message sending terminology, it comes from
3398: @c a time when each object had it's own task and objects communicated via
3399: @c message passing; eventually the Smalltalk developers realized that
3400: @c they can do most things through simple (indirect) calls. They kept the
3401: @c terminology.
3402:
3403: @node Basic Objects Usage, The class Object, Object-Oriented Terminology, Objects
1.12 anton 3404: @subsubsection Basic Objects Usage
1.5 anton 3405: @cindex basic objects usage
3406: @cindex objects, basic usage
3407:
3408: You can define a class for graphical objects like this:
3409:
3410: @cindex @code{class} usage
3411: @cindex @code{end-class} usage
3412: @cindex @code{selector} usage
3413: @example
3414: object class \ "object" is the parent class
3415: selector draw ( x y graphical -- )
3416: end-class graphical
3417: @end example
3418:
3419: This code defines a class @code{graphical} with an
3420: operation @code{draw}. We can perform the operation
3421: @code{draw} on any @code{graphical} object, e.g.:
3422:
3423: @example
3424: 100 100 t-rex draw
3425: @end example
3426:
3427: where @code{t-rex} is a word (say, a constant) that produces a
3428: graphical object.
3429:
3430: @cindex abstract class
3431: How do we create a graphical object? With the present definitions,
3432: we cannot create a useful graphical object. The class
3433: @code{graphical} describes graphical objects in general, but not
3434: any concrete graphical object type (C++ users would call it an
3435: @emph{abstract class}); e.g., there is no method for the selector
3436: @code{draw} in the class @code{graphical}.
3437:
3438: For concrete graphical objects, we define child classes of the
3439: class @code{graphical}, e.g.:
3440:
3441: @cindex @code{overrides} usage
3442: @cindex @code{field} usage in class definition
3443: @example
3444: graphical class \ "graphical" is the parent class
3445: cell% field circle-radius
3446:
3447: :noname ( x y circle -- )
3448: circle-radius @@ draw-circle ;
3449: overrides draw
3450:
3451: :noname ( n-radius circle -- )
3452: circle-radius ! ;
3453: overrides construct
3454:
3455: end-class circle
3456: @end example
3457:
3458: Here we define a class @code{circle} as a child of @code{graphical},
3459: with a field @code{circle-radius} (which behaves just like a field in
3460: @pxref{Structures}); it defines new methods for the selectors
3461: @code{draw} and @code{construct} (@code{construct} is defined in
3462: @code{object}, the parent class of @code{graphical}).
3463:
3464: Now we can create a circle on the heap (i.e.,
3465: @code{allocate}d memory) with
3466:
3467: @cindex @code{heap-new} usage
3468: @example
3469: 50 circle heap-new constant my-circle
3470: @end example
3471:
3472: @code{heap-new} invokes @code{construct}, thus
3473: initializing the field @code{circle-radius} with 50. We can draw
3474: this new circle at (100,100) with
3475:
3476: @example
3477: 100 100 my-circle draw
3478: @end example
3479:
3480: @cindex selector invocation, restrictions
3481: @cindex class definition, restrictions
3482: Note: You can invoke a selector only if the object on the TOS
3483: (the receiving object) belongs to the class where the selector was
3484: defined or one of its descendents; e.g., you can invoke
3485: @code{draw} only for objects belonging to @code{graphical}
3486: or its descendents (e.g., @code{circle}). Immediately before
3487: @code{end-class}, the search order has to be the same as
3488: immediately after @code{class}.
3489:
3490: @node The class Object, Creating objects, Basic Objects Usage, Objects
1.12 anton 3491: @subsubsection The class @code{object}
1.5 anton 3492: @cindex @code{object} class
3493:
3494: When you define a class, you have to specify a parent class. So how do
3495: you start defining classes? There is one class available from the start:
3496: @code{object}. You can use it as ancestor for all classes. It is the
3497: only class that has no parent. It has two selectors: @code{construct}
3498: and @code{print}.
3499:
3500: @node Creating objects, Object-Oriented Programming Style, The class Object, Objects
1.12 anton 3501: @subsubsection Creating objects
1.5 anton 3502: @cindex creating objects
3503: @cindex object creation
3504: @cindex object allocation options
3505:
3506: @cindex @code{heap-new} discussion
3507: @cindex @code{dict-new} discussion
3508: @cindex @code{construct} discussion
3509: You can create and initialize an object of a class on the heap with
3510: @code{heap-new} ( ... class -- object ) and in the dictionary
3511: (allocation with @code{allot}) with @code{dict-new} (
3512: ... class -- object ). Both words invoke @code{construct}, which
3513: consumes the stack items indicated by "..." above.
3514:
3515: @cindex @code{init-object} discussion
3516: @cindex @code{class-inst-size} discussion
3517: If you want to allocate memory for an object yourself, you can get its
3518: alignment and size with @code{class-inst-size 2@@} ( class --
3519: align size ). Once you have memory for an object, you can initialize
3520: it with @code{init-object} ( ... class object -- );
3521: @code{construct} does only a part of the necessary work.
3522:
3523: @node Object-Oriented Programming Style, Class Binding, Creating objects, Objects
1.12 anton 3524: @subsubsection Object-Oriented Programming Style
1.5 anton 3525: @cindex object-oriented programming style
3526:
3527: This section is not exhaustive.
3528:
3529: @cindex stack effects of selectors
3530: @cindex selectors and stack effects
3531: In general, it is a good idea to ensure that all methods for the
3532: same selector have the same stack effect: when you invoke a selector,
3533: you often have no idea which method will be invoked, so, unless all
3534: methods have the same stack effect, you will not know the stack effect
3535: of the selector invocation.
3536:
3537: One exception to this rule is methods for the selector
3538: @code{construct}. We know which method is invoked, because we
3539: specify the class to be constructed at the same place. Actually, I
3540: defined @code{construct} as a selector only to give the users a
3541: convenient way to specify initialization. The way it is used, a
3542: mechanism different from selector invocation would be more natural
3543: (but probably would take more code and more space to explain).
3544:
3545: @node Class Binding, Method conveniences, Object-Oriented Programming Style, Objects
1.12 anton 3546: @subsubsection Class Binding
1.5 anton 3547: @cindex class binding
3548: @cindex early binding
3549:
3550: @cindex late binding
3551: Normal selector invocations determine the method at run-time depending
3552: on the class of the receiving object (late binding).
3553:
3554: Sometimes we want to invoke a different method. E.g., assume that
3555: you want to use the simple method for @code{print}ing
3556: @code{object}s instead of the possibly long-winded
3557: @code{print} method of the receiver class. You can achieve this
3558: by replacing the invocation of @code{print} with
3559:
3560: @cindex @code{[bind]} usage
3561: @example
3562: [bind] object print
3563: @end example
3564:
3565: in compiled code or
3566:
3567: @cindex @code{bind} usage
3568: @example
3569: bind object print
3570: @end example
3571:
3572: @cindex class binding, alternative to
3573: in interpreted code. Alternatively, you can define the method with a
3574: name (e.g., @code{print-object}), and then invoke it through the
3575: name. Class binding is just a (often more convenient) way to achieve
3576: the same effect; it avoids name clutter and allows you to invoke
3577: methods directly without naming them first.
3578:
3579: @cindex superclass binding
3580: @cindex parent class binding
3581: A frequent use of class binding is this: When we define a method
3582: for a selector, we often want the method to do what the selector does
3583: in the parent class, and a little more. There is a special word for
3584: this purpose: @code{[parent]}; @code{[parent]
3585: @emph{selector}} is equivalent to @code{[bind] @emph{parent
3586: selector}}, where @code{@emph{parent}} is the parent
3587: class of the current class. E.g., a method definition might look like:
3588:
3589: @cindex @code{[parent]} usage
3590: @example
3591: :noname
3592: dup [parent] foo \ do parent's foo on the receiving object
3593: ... \ do some more
3594: ; overrides foo
3595: @end example
3596:
3597: @cindex class binding as optimization
3598: In @cite{Object-oriented programming in ANS Forth} (Forth Dimensions,
3599: March 1997), Andrew McKewan presents class binding as an optimization
3600: technique. I recommend not using it for this purpose unless you are in
3601: an emergency. Late binding is pretty fast with this model anyway, so the
3602: benefit of using class binding is small; the cost of using class binding
3603: where it is not appropriate is reduced maintainability.
3604:
3605: While we are at programming style questions: You should bind
3606: selectors only to ancestor classes of the receiving object. E.g., say,
3607: you know that the receiving object is of class @code{foo} or its
3608: descendents; then you should bind only to @code{foo} and its
3609: ancestors.
3610:
3611: @node Method conveniences, Classes and Scoping, Class Binding, Objects
1.12 anton 3612: @subsubsection Method conveniences
1.5 anton 3613: @cindex method conveniences
3614:
3615: In a method you usually access the receiving object pretty often. If
3616: you define the method as a plain colon definition (e.g., with
3617: @code{:noname}), you may have to do a lot of stack
3618: gymnastics. To avoid this, you can define the method with @code{m:
3619: ... ;m}. E.g., you could define the method for
3620: @code{draw}ing a @code{circle} with
3621:
3622: @cindex @code{this} usage
3623: @cindex @code{m:} usage
3624: @cindex @code{;m} usage
3625: @example
3626: m: ( x y circle -- )
3627: ( x y ) this circle-radius @@ draw-circle ;m
3628: @end example
3629:
3630: @cindex @code{exit} in @code{m: ... ;m}
3631: @cindex @code{exitm} discussion
3632: @cindex @code{catch} in @code{m: ... ;m}
3633: When this method is executed, the receiver object is removed from the
3634: stack; you can access it with @code{this} (admittedly, in this
3635: example the use of @code{m: ... ;m} offers no advantage). Note
3636: that I specify the stack effect for the whole method (i.e. including
3637: the receiver object), not just for the code between @code{m:}
3638: and @code{;m}. You cannot use @code{exit} in
3639: @code{m:...;m}; instead, use
3640: @code{exitm}.@footnote{Moreover, for any word that calls
3641: @code{catch} and was defined before loading
3642: @code{objects.fs}, you have to redefine it like I redefined
3643: @code{catch}: @code{: catch this >r catch r> to-this ;}}
3644:
3645: @cindex @code{inst-var} usage
3646: You will frequently use sequences of the form @code{this
3647: @emph{field}} (in the example above: @code{this
3648: circle-radius}). If you use the field only in this way, you can
3649: define it with @code{inst-var} and eliminate the
3650: @code{this} before the field name. E.g., the @code{circle}
3651: class above could also be defined with:
3652:
3653: @example
3654: graphical class
3655: cell% inst-var radius
3656:
3657: m: ( x y circle -- )
3658: radius @@ draw-circle ;m
3659: overrides draw
3660:
3661: m: ( n-radius circle -- )
3662: radius ! ;m
3663: overrides construct
3664:
3665: end-class circle
3666: @end example
3667:
3668: @code{radius} can only be used in @code{circle} and its
3669: descendent classes and inside @code{m:...;m}.
3670:
3671: @cindex @code{inst-value} usage
3672: You can also define fields with @code{inst-value}, which is
3673: to @code{inst-var} what @code{value} is to
3674: @code{variable}. You can change the value of such a field with
3675: @code{[to-inst]}. E.g., we could also define the class
3676: @code{circle} like this:
3677:
3678: @example
3679: graphical class
3680: inst-value radius
3681:
3682: m: ( x y circle -- )
3683: radius draw-circle ;m
3684: overrides draw
3685:
3686: m: ( n-radius circle -- )
3687: [to-inst] radius ;m
3688: overrides construct
3689:
3690: end-class circle
3691: @end example
3692:
3693:
3694: @node Classes and Scoping, Object Interfaces, Method conveniences, Objects
1.12 anton 3695: @subsubsection Classes and Scoping
1.5 anton 3696: @cindex classes and scoping
3697: @cindex scoping and classes
3698:
3699: Inheritance is frequent, unlike structure extension. This exacerbates
3700: the problem with the field name convention (@pxref{Structure Naming
3701: Convention}): One always has to remember in which class the field was
3702: originally defined; changing a part of the class structure would require
3703: changes for renaming in otherwise unaffected code.
3704:
3705: @cindex @code{inst-var} visibility
3706: @cindex @code{inst-value} visibility
3707: To solve this problem, I added a scoping mechanism (which was not in my
3708: original charter): A field defined with @code{inst-var} (or
3709: @code{inst-value}) is visible only in the class where it is defined and in
3710: the descendent classes of this class. Using such fields only makes
3711: sense in @code{m:}-defined methods in these classes anyway.
3712:
3713: This scoping mechanism allows us to use the unadorned field name,
3714: because name clashes with unrelated words become much less likely.
3715:
3716: @cindex @code{protected} discussion
3717: @cindex @code{private} discussion
3718: Once we have this mechanism, we can also use it for controlling the
3719: visibility of other words: All words defined after
3720: @code{protected} are visible only in the current class and its
3721: descendents. @code{public} restores the compilation
3722: (i.e. @code{current}) wordlist that was in effect before. If you
3723: have several @code{protected}s without an intervening
3724: @code{public} or @code{set-current}, @code{public}
3725: will restore the compilation wordlist in effect before the first of
3726: these @code{protected}s.
3727:
3728: @node Object Interfaces, Objects Implementation, Classes and Scoping, Objects
1.12 anton 3729: @subsubsection Object Interfaces
1.5 anton 3730: @cindex object interfaces
3731: @cindex interfaces for objects
3732:
3733: In this model you can only call selectors defined in the class of the
3734: receiving objects or in one of its ancestors. If you call a selector
3735: with a receiving object that is not in one of these classes, the
3736: result is undefined; if you are lucky, the program crashes
3737: immediately.
3738:
3739: @cindex selectors common to hardly-related classes
3740: Now consider the case when you want to have a selector (or several)
3741: available in two classes: You would have to add the selector to a
3742: common ancestor class, in the worst case to @code{object}. You
3743: may not want to do this, e.g., because someone else is responsible for
3744: this ancestor class.
3745:
3746: The solution for this problem is interfaces. An interface is a
3747: collection of selectors. If a class implements an interface, the
3748: selectors become available to the class and its descendents. A class
3749: can implement an unlimited number of interfaces. For the problem
3750: discussed above, we would define an interface for the selector(s), and
3751: both classes would implement the interface.
3752:
3753: As an example, consider an interface @code{storage} for
3754: writing objects to disk and getting them back, and a class
3755: @code{foo} foo that implements it. The code for this would look
3756: like this:
3757:
3758: @cindex @code{interface} usage
3759: @cindex @code{end-interface} usage
3760: @cindex @code{implementation} usage
3761: @example
3762: interface
3763: selector write ( file object -- )
3764: selector read1 ( file object -- )
3765: end-interface storage
3766:
3767: bar class
3768: storage implementation
3769:
3770: ... overrides write
3771: ... overrides read
3772: ...
3773: end-class foo
3774: @end example
3775:
3776: (I would add a word @code{read} ( file -- object ) that uses
3777: @code{read1} internally, but that's beyond the point illustrated
3778: here.)
3779:
3780: Note that you cannot use @code{protected} in an interface; and
3781: of course you cannot define fields.
3782:
3783: In the Neon model, all selectors are available for all classes;
3784: therefore it does not need interfaces. The price you pay in this model
3785: is slower late binding, and therefore, added complexity to avoid late
3786: binding.
3787:
3788: @node Objects Implementation, Comparison with other object models, Object Interfaces, Objects
1.12 anton 3789: @subsubsection @file{objects.fs} Implementation
1.5 anton 3790: @cindex @file{objects.fs} implementation
3791:
3792: @cindex @code{object-map} discussion
3793: An object is a piece of memory, like one of the data structures
3794: described with @code{struct...end-struct}. It has a field
3795: @code{object-map} that points to the method map for the object's
3796: class.
3797:
3798: @cindex method map
3799: @cindex virtual function table
3800: The @emph{method map}@footnote{This is Self terminology; in C++
3801: terminology: virtual function table.} is an array that contains the
3802: execution tokens (XTs) of the methods for the object's class. Each
3803: selector contains an offset into the method maps.
3804:
3805: @cindex @code{selector} implementation, class
3806: @code{selector} is a defining word that uses
3807: @code{create} and @code{does>}. The body of the
3808: selector contains the offset; the @code{does>} action for a
3809: class selector is, basically:
3810:
3811: @example
3812: ( object addr ) @@ over object-map @@ + @@ execute
3813: @end example
3814:
3815: Since @code{object-map} is the first field of the object, it
3816: does not generate any code. As you can see, calling a selector has a
3817: small, constant cost.
3818:
3819: @cindex @code{current-interface} discussion
3820: @cindex class implementation and representation
3821: A class is basically a @code{struct} combined with a method
3822: map. During the class definition the alignment and size of the class
3823: are passed on the stack, just as with @code{struct}s, so
3824: @code{field} can also be used for defining class
3825: fields. However, passing more items on the stack would be
3826: inconvenient, so @code{class} builds a data structure in memory,
3827: which is accessed through the variable
3828: @code{current-interface}. After its definition is complete, the
3829: class is represented on the stack by a pointer (e.g., as parameter for
3830: a child class definition).
3831:
3832: At the start, a new class has the alignment and size of its parent,
3833: and a copy of the parent's method map. Defining new fields extends the
3834: size and alignment; likewise, defining new selectors extends the
3835: method map. @code{overrides} just stores a new XT in the method
3836: map at the offset given by the selector.
3837:
3838: @cindex class binding, implementation
3839: Class binding just gets the XT at the offset given by the selector
3840: from the class's method map and @code{compile,}s (in the case of
3841: @code{[bind]}) it.
3842:
3843: @cindex @code{this} implementation
3844: @cindex @code{catch} and @code{this}
3845: @cindex @code{this} and @code{catch}
3846: I implemented @code{this} as a @code{value}. At the
3847: start of an @code{m:...;m} method the old @code{this} is
3848: stored to the return stack and restored at the end; and the object on
3849: the TOS is stored @code{TO this}. This technique has one
3850: disadvantage: If the user does not leave the method via
3851: @code{;m}, but via @code{throw} or @code{exit},
3852: @code{this} is not restored (and @code{exit} may
3853: crash). To deal with the @code{throw} problem, I have redefined
3854: @code{catch} to save and restore @code{this}; the same
3855: should be done with any word that can catch an exception. As for
3856: @code{exit}, I simply forbid it (as a replacement, there is
3857: @code{exitm}).
3858:
3859: @cindex @code{inst-var} implementation
3860: @code{inst-var} is just the same as @code{field}, with
3861: a different @code{does>} action:
3862: @example
3863: @@ this +
3864: @end example
3865: Similar for @code{inst-value}.
3866:
3867: @cindex class scoping implementation
3868: Each class also has a wordlist that contains the words defined with
3869: @code{inst-var} and @code{inst-value}, and its protected
3870: words. It also has a pointer to its parent. @code{class} pushes
3871: the wordlists of the class an all its ancestors on the search order,
3872: and @code{end-class} drops them.
3873:
3874: @cindex interface implementation
3875: An interface is like a class without fields, parent and protected
3876: words; i.e., it just has a method map. If a class implements an
3877: interface, its method map contains a pointer to the method map of the
3878: interface. The positive offsets in the map are reserved for class
3879: methods, therefore interface map pointers have negative
3880: offsets. Interfaces have offsets that are unique throughout the
3881: system, unlike class selectors, whose offsets are only unique for the
3882: classes where the selector is available (invokable).
3883:
3884: This structure means that interface selectors have to perform one
3885: indirection more than class selectors to find their method. Their body
3886: contains the interface map pointer offset in the class method map, and
3887: the method offset in the interface method map. The
3888: @code{does>} action for an interface selector is, basically:
3889:
3890: @example
3891: ( object selector-body )
3892: 2dup selector-interface @@ ( object selector-body object interface-offset )
3893: swap object-map @@ + @@ ( object selector-body map )
3894: swap selector-offset @@ + @@ execute
3895: @end example
3896:
3897: where @code{object-map} and @code{selector-offset} are
3898: first fields and generate no code.
3899:
3900: As a concrete example, consider the following code:
3901:
3902: @example
3903: interface
3904: selector if1sel1
3905: selector if1sel2
3906: end-interface if1
3907:
3908: object class
3909: if1 implementation
3910: selector cl1sel1
3911: cell% inst-var cl1iv1
3912:
3913: ' m1 overrides construct
3914: ' m2 overrides if1sel1
3915: ' m3 overrides if1sel2
3916: ' m4 overrides cl1sel2
3917: end-class cl1
3918:
3919: create obj1 object dict-new drop
3920: create obj2 cl1 dict-new drop
3921: @end example
3922:
3923: The data structure created by this code (including the data structure
3924: for @code{object}) is shown in the <a
3925: href="objects-implementation.eps">figure</a>, assuming a cell size of 4.
3926:
3927: @node Comparison with other object models, Objects Glossary, Objects Implementation, Objects
1.12 anton 3928: @subsubsection Comparison with other object models
1.5 anton 3929: @cindex comparison of object models
3930: @cindex object models, comparison
3931:
3932: Many object-oriented Forth extensions have been proposed (@cite{A survey
3933: of object-oriented Forths} (SIGPLAN Notices, April 1996) by Bradford
3934: J. Rodriguez and W. F. S. Poehlman lists 17). Here I'll discuss the
3935: relation of @file{objects.fs} to two well-known and two closely-related
3936: (by the use of method maps) models.
3937:
3938: @cindex Neon model
3939: The most popular model currently seems to be the Neon model (see
3940: @cite{Object-oriented programming in ANS Forth} (Forth Dimensions, March
3941: 1997) by Andrew McKewan). The Neon model uses a @code{@emph{selector
3942: object}} syntax, which makes it unnatural to pass objects on the
3943: stack. It also requires that the selector parses the input stream (at
3944: compile time); this leads to reduced extensibility and to bugs that are
3945: hard to find. Finally, it allows using every selector to every object;
3946: this eliminates the need for classes, but makes it harder to create
3947: efficient implementations. A longer version of this critique can be
3948: found in @cite{On Standardizing Object-Oriented Forth Extensions} (Forth
3949: Dimensions, May 1997) by Anton Ertl.
3950:
3951: @cindex Pountain's object-oriented model
3952: Another well-known publication is @cite{Object-Oriented Forth} (Academic
3953: Press, London, 1987) by Dick Pountain. However, it is not really about
3954: object-oriented programming, because it hardly deals with late
3955: binding. Instead, it focuses on features like information hiding and
3956: overloading that are characteristic of modular languages like Ada (83).
3957:
3958: @cindex Zsoter's object-oriented model
3959: In @cite{Does late binding have to be slow?} (Forth Dimensions ??? 1996)
3960: Andras Zsoter describes a model that makes heavy use of an active object
3961: (like @code{this} in @file{objects.fs}): The active object is not only
3962: used for accessing all fields, but also specifies the receiving object
3963: of every selector invocation; you have to change the active object
3964: explicitly with @code{@{ ... @}}, whereas in @file{objects.fs} it
3965: changes more or less implicitly at @code{m: ... ;m}. Such a change at
3966: the method entry point is unnecessary with the Zsoter's model, because
3967: the receiving object is the active object already; OTOH, the explicit
3968: change is absolutely necessary in that model, because otherwise no one
3969: could ever change the active object. An ANS Forth implementation of this
3970: model is available at @url{http://www.forth.org/fig/oopf.html}.
3971:
1.12 anton 3972: @cindex @file{oof.fs}, differences to other models
1.5 anton 3973: The @file{oof.fs} model combines information hiding and overloading
3974: resolution (by keeping names in various wordlists) with object-oriented
3975: programming. It sets the active object implicitly on method entry, but
3976: also allows explicit changing (with @code{>o...o>} or with
3977: @code{with...endwith}). It uses parsing and state-smart objects and
3978: classes for resolving overloading and for early binding: the object or
3979: class parses the selector and determines the method from this. If the
3980: selector is not parsed by an object or class, it performs a call to the
3981: selector for the active object (late binding), like Zsoter's model.
3982: Fields are always accessed through the active object. The big
3983: disadvantage of this model is the parsing and the state-smartness, which
3984: reduces extensibility and increases the opportunities for subtle bugs;
3985: essentially, you are only safe if you never tick or @code{postpone} an
1.12 anton 3986: object or class (Bernd disagrees, but I (Anton) am not convinced).
3987:
3988: @cindex @file{mini-oof.fs}, differences to other models
3989: The Mini-OOF model is quite similar to a very stripped-down version of
3990: the Objects model, but syntactically it is a mixture of the Objects and
3991: the OOF model.
3992:
1.5 anton 3993:
3994: @node Objects Glossary, , Comparison with other object models, Objects
1.12 anton 3995: @subsubsection @file{objects.fs} Glossary
1.5 anton 3996: @cindex @file{objects.fs} Glossary
3997:
3998: doc-bind
3999: doc-<bind>
4000: doc-bind'
4001: doc-[bind]
4002: doc-class
4003: doc-class->map
4004: doc-class-inst-size
4005: doc-class-override!
4006: doc-construct
4007: doc-current'
4008: doc-[current]
4009: doc-current-interface
4010: doc-dict-new
4011: doc-drop-order
4012: doc-end-class
4013: doc-end-class-noname
4014: doc-end-interface
4015: doc-end-interface-noname
4016: doc-exitm
4017: doc-heap-new
4018: doc-implementation
4019: doc-init-object
4020: doc-inst-value
4021: doc-inst-var
4022: doc-interface
4023: doc-;m
4024: doc-m:
4025: doc-method
4026: doc-object
4027: doc-overrides
4028: doc-[parent]
4029: doc-print
4030: doc-protected
4031: doc-public
4032: doc-push-order
4033: doc-selector
4034: doc-this
4035: doc-<to-inst>
4036: doc-[to-inst]
4037: doc-to-this
4038: doc-xt-new
4039:
4040: @c -------------------------------------------------------------
1.12 anton 4041: @node OOF, Mini-OOF, Objects, Object-oriented Forth
4042: @subsection OOF
1.6 pazsan 4043: @cindex oof
4044: @cindex object-oriented programming
4045:
4046: @cindex @file{objects.fs}
4047: @cindex @file{oof.fs}
1.12 anton 4048:
4049: This section describes the @file{oof.fs} packet. This section uses the
4050: same rationale why using object-oriented programming, and the same
1.6 pazsan 4051: terminology.
4052:
4053: The packet described in this section is used in bigFORTH since 1991, and
4054: used for two large applications: a chromatographic system used to
4055: create new medicaments, and a graphic user interface library (MINOS).
4056:
1.12 anton 4057: You can find a description (in German) of @file{oof.fs} in @cite{Object
4058: oriented bigFORTH} by Bernd Paysan, published in @cite{Vierte Dimension}
4059: 10(2), 1994.
4060:
1.6 pazsan 4061: @menu
4062: * Properties of the OOF model::
4063: * Basic OOF Usage::
4064: * The base class object::
1.7 pazsan 4065: * Class Declaration::
4066: * Class Implementation::
1.6 pazsan 4067: @end menu
4068:
1.12 anton 4069: @node Properties of the OOF model, Basic OOF Usage, OOF, OOF
4070: @subsubsection Properties of the OOF model
1.6 pazsan 4071: @cindex @file{oof.fs} properties
4072:
4073: @itemize @bullet
4074: @item
4075: This model combines object oriented programming with information
4076: hiding. It helps you writing large application, where scoping is
4077: necessary, because it provides class-oriented scoping.
4078:
4079: @item
4080: Named objects, object pointers, and object arrays can be created,
4081: selector invocation uses the "object selector" syntax. Selector invocation
4082: to objects and/or selectors on the stack is a bit less convenient, but
4083: possible.
4084:
4085: @item
4086: Selector invocation and instance variable usage of the active object is
4087: straight forward, since both make use of the active object.
4088:
4089: @item
4090: Late binding is efficient and easy to use.
4091:
4092: @item
4093: State-smart objects parse selectors. However, extensibility is provided
4094: using a (parsing) selector @code{postpone} and a selector @code{'}.
4095:
4096: @item
4097: An implementation in ANS Forth is available.
4098:
4099: @end itemize
4100:
4101:
1.12 anton 4102: @node Basic OOF Usage, The base class object, Properties of the OOF model, OOF
4103: @subsubsection Basic OOF Usage
1.6 pazsan 4104: @cindex @file{oof.fs} usage
4105:
4106: Here, I use the same example as for @code{objects} (@pxref{Basic Objects Usage}).
4107:
4108: You can define a class for graphical objects like this:
4109:
4110: @cindex @code{class} usage
4111: @cindex @code{class;} usage
4112: @cindex @code{method} usage
4113: @example
4114: object class graphical \ "object" is the parent class
4115: method draw ( x y graphical -- )
4116: class;
4117: @end example
4118:
4119: This code defines a class @code{graphical} with an
4120: operation @code{draw}. We can perform the operation
4121: @code{draw} on any @code{graphical} object, e.g.:
4122:
4123: @example
4124: 100 100 t-rex draw
4125: @end example
4126:
4127: where @code{t-rex} is an object or object pointer, created with e.g.
1.13 pazsan 4128: @code{graphical : t-rex}.
1.6 pazsan 4129:
4130: @cindex abstract class
4131: How do we create a graphical object? With the present definitions,
4132: we cannot create a useful graphical object. The class
4133: @code{graphical} describes graphical objects in general, but not
4134: any concrete graphical object type (C++ users would call it an
4135: @emph{abstract class}); e.g., there is no method for the selector
4136: @code{draw} in the class @code{graphical}.
4137:
4138: For concrete graphical objects, we define child classes of the
4139: class @code{graphical}, e.g.:
4140:
4141: @example
4142: graphical class circle \ "graphical" is the parent class
4143: cell var circle-radius
4144: how:
4145: : draw ( x y -- )
4146: circle-radius @@ draw-circle ;
4147:
4148: : init ( n-radius -- (
4149: circle-radius ! ;
4150: class;
4151: @end example
4152:
4153: Here we define a class @code{circle} as a child of @code{graphical},
4154: with a field @code{circle-radius}; it defines new methods for the
4155: selectors @code{draw} and @code{init} (@code{init} is defined in
4156: @code{object}, the parent class of @code{graphical}).
4157:
4158: Now we can create a circle in the dictionary with
4159:
4160: @example
4161: 50 circle : my-circle
4162: @end example
4163:
4164: @code{:} invokes @code{init}, thus initializing the field
4165: @code{circle-radius} with 50. We can draw this new circle at (100,100)
4166: with
4167:
4168: @example
4169: 100 100 my-circle draw
4170: @end example
4171:
4172: @cindex selector invocation, restrictions
4173: @cindex class definition, restrictions
4174: Note: You can invoke a selector only if the receiving object belongs to
4175: the class where the selector was defined or one of its descendents;
4176: e.g., you can invoke @code{draw} only for objects belonging to
4177: @code{graphical} or its descendents (e.g., @code{circle}). The scoping
1.7 pazsan 4178: mechanism will check if you try to invoke a selector that is not
1.6 pazsan 4179: defined in this class hierarchy, so you'll get an error at compilation
4180: time.
4181:
4182:
1.12 anton 4183: @node The base class object, Class Declaration, Basic OOF Usage, OOF
4184: @subsubsection The base class @file{object}
1.6 pazsan 4185: @cindex @file{oof.fs} base class
4186:
4187: When you define a class, you have to specify a parent class. So how do
4188: you start defining classes? There is one class available from the start:
4189: @code{object}. You have to use it as ancestor for all classes. It is the
4190: only class that has no parent. Classes are also objects, except that
4191: they don't have instance variables; class manipulation such as
4192: inheritance or changing definitions of a class is handled through
4193: selectors of the class @code{object}.
4194:
4195: @code{object} provides a number of selectors:
4196:
4197: @itemize @bullet
4198: @item
4199: @code{class} for subclassing, @code{definitions} to add definitions
4200: later on, and @code{class?} to get type informations (is the class a
4201: subclass of the class passed on the stack?).
1.7 pazsan 4202: doc---object-class
4203: doc---object-definitions
4204: doc---object-class?
1.6 pazsan 4205:
4206: @item
4207: @code{init} and @code{dispose} as constructor and destroctor of the
4208: object. @code{init} is invocated after the object's memory is allocated,
4209: while @code{dispose} also handles deallocation. Thus if you redefine
4210: @code{dispose}, you have to call the parent's dispose with @code{super
4211: dispose}, too.
1.7 pazsan 4212: doc---object-init
4213: doc---object-dispose
1.6 pazsan 4214:
4215: @item
1.7 pazsan 4216: @code{new}, @code{new[]}, @code{:}, @code{ptr}, @code{asptr}, and
4217: @code{[]} to create named and unnamed objects and object arrays or
4218: object pointers.
4219: doc---object-new
4220: doc---object-new[]
4221: doc---object-:
4222: doc---object-ptr
4223: doc---object-asptr
4224: doc---object-[]
1.6 pazsan 4225:
4226: @item
4227: @code{::} and @code{super} for expicit scoping. You should use expicit
4228: scoping only for super classes or classes with the same set of instance
4229: variables. Explicit scoped selectors use early binding.
1.7 pazsan 4230: doc---object-::
4231: doc---object-super
1.6 pazsan 4232:
4233: @item
4234: @code{self} to get the address of the object
1.7 pazsan 4235: doc---object-self
1.6 pazsan 4236:
4237: @item
4238: @code{bind}, @code{bound}, @code{link}, and @code{is} to assign object
4239: pointers and instance defers.
1.7 pazsan 4240: doc---object-bind
4241: doc---object-bound
4242: doc---object-link
4243: doc---object-is
1.6 pazsan 4244:
4245: @item
4246: @code{'} to obtain selector tokens, @code{send} to invocate selectors
4247: form the stack, and @code{postpone} to generate selector invocation code.
1.7 pazsan 4248: doc---object-'
4249: doc---object-postpone
1.6 pazsan 4250:
4251: @item
4252: @code{with} and @code{endwith} to select the active object from the
4253: stack, and enabling it's scope. Using @code{with} and @code{endwith}
4254: also allows to create code using selector @code{postpone} without being
4255: trapped bye the state-smart objects.
1.7 pazsan 4256: doc---object-with
4257: doc---object-endwith
1.6 pazsan 4258:
4259: @end itemize
4260:
1.12 anton 4261: @node Class Declaration, Class Implementation, The base class object, OOF
4262: @subsubsection Class Declaration
1.7 pazsan 4263: @cindex class declaration
4264:
4265: @itemize @bullet
4266: @item
4267: Instance variables
4268: doc---oof-var
4269:
4270: @item
4271: Object pointers
4272: doc---oof-ptr
4273: doc---oof-asptr
4274:
4275: @item
4276: Instance defers
4277: doc---oof-defer
4278:
4279: @item
4280: Method selectors
4281: doc---oof-early
4282: doc---oof-method
4283:
4284: @item
4285: Class wide variables
4286: doc---oof-static
4287:
4288: @item
4289: End declaration
4290: doc---oof-how:
4291: doc---oof-class;
4292:
4293: @end itemize
4294:
1.13 pazsan 4295: @c -------------------------------------------------------------
1.12 anton 4296: @node Class Implementation, , Class Declaration, OOF
4297: @subsubsection Class Implementation
1.7 pazsan 4298: @cindex class implementation
4299:
1.13 pazsan 4300: @c -------------------------------------------------------------
4301: @node Mini-OOF, , OOF, Object-oriented Forth
1.12 anton 4302: @subsection Mini-OOF
1.8 pazsan 4303: @cindex mini-oof
4304:
4305: Gforth's third object oriented Forth package is a 12-liner. It uses a
4306: bit of a mixture of the @file{object.fs} and the @file{oof.fs} syntax,
1.13 pazsan 4307: and reduces to the bare minimum of features. This is based on a posting
4308: of Bernd Paysan in comp.arch.
4309:
4310: @menu
4311: * Mini-OOF Usage::
4312: * Mini-OOF Example::
4313: @end menu
4314:
4315: @c -------------------------------------------------------------
4316: @node Mini-OOF Usage, Mini-OOF Example, , Mini-OOF
4317: @subsubsection Usage
4318: @cindex mini-oof usage
4319:
4320: Basically, there are seven words, to define a method, a variable, a
4321: class; to end a class, to define a method, to allocate an object, to
4322: resolve binding, and the base class (which allocates one cell for the
4323: object pointer).
4324:
4325: doc-method
4326:
4327: Defines a method
4328:
4329: doc-var
4330:
4331: Defines a variable with size bytes
4332:
4333: doc-class
4334:
4335: Starts the definition of a sub-class
4336:
4337: doc-end-class
4338:
4339: Ends the definition of a class
4340:
4341: doc-defines
4342:
4343: Binds the xt to the method name in the class
4344:
4345: doc-new
4346:
4347: Creates a new incarnation of the class
4348:
4349: doc-::
4350:
4351: Compiles the method name of the class (not immediate!)
4352:
4353: doc-object
4354:
4355: Is the base class of all objects
4356:
4357: @c -------------------------------------------------------------
4358: @node Mini-OOF Example, , Mini-OOF Usage, Mini-OOF
4359: @subsubsection Mini-OOF Example
4360: @cindex mini-oof example
4361:
4362: A short example shows how to use this package.
4363:
4364: @example
4365: object class
4366: method init
4367: method draw
4368: end-class graphical
4369: @end example
4370:
4371: This code defines a class @code{graphical} with an
4372: operation @code{draw}. We can perform the operation
4373: @code{draw} on any @code{graphical} object, e.g.:
4374:
4375: @example
4376: 100 100 t-rex draw
4377: @end example
4378:
4379: where @code{t-rex} is an object or object pointer, created with e.g.
4380: @code{graphical new Constant t-rex}.
4381:
4382: For concrete graphical objects, we define child classes of the
4383: class @code{graphical}, e.g.:
1.8 pazsan 4384:
4385: @example
1.13 pazsan 4386: graphical class
4387: cell var circle-radius
4388: end-class circle \ "graphical" is the parent class
4389:
4390: :noname ( x y -- )
4391: circle-radius @@ draw-circle ; circle defines draw
4392: :noname ( r -- )
4393: circle-radius ! ; circle defines init
4394: @end example
4395:
4396: There is no implicit init method, so we have to define one. The creation
4397: code of the object now has to call init explicitely.
4398:
4399: @example
4400: circle new Constant my-circle
4401: 50 my-circle init
4402: @end example
4403:
4404: It is also possible to add a function to create named objects with
4405: automatic call of @code{init}, given that all objects have @code{init}
4406: on the same place
4407:
4408: @example
4409: : new: ( .. o "name" -- )
4410: new dup Constant init ;
4411: 80 circle new: large-circle
4412: @end example
4413:
4414: We can draw this new circle at (100,100)
4415: with
4416:
4417: @example
4418: 100 100 my-circle draw
1.8 pazsan 4419: @end example
4420:
1.6 pazsan 4421: @c -------------------------------------------------------------
1.12 anton 4422: @node Tokens for Words, Wordlists, Object-oriented Forth, Words
1.1 anton 4423: @section Tokens for Words
4424: @cindex tokens for words
4425:
4426: This chapter describes the creation and use of tokens that represent
4427: words on the stack (and in data space).
4428:
4429: Named words have interpretation and compilation semantics. Unnamed words
4430: just have execution semantics.
4431:
4432: @cindex execution token
4433: An @dfn{execution token} represents the execution semantics of an
4434: unnamed word. An execution token occupies one cell. As explained in
4435: section @ref{Supplying names}, the execution token of the last words
4436: defined can be produced with
4437:
4438: short-lastxt
4439:
4440: You can perform the semantics represented by an execution token with
4441: doc-execute
4442: You can compile the word with
4443: doc-compile,
4444:
4445: @cindex code field address
4446: @cindex CFA
4447: In Gforth, the abstract data type @emph{execution token} is implemented
4448: as CFA (code field address).
4449:
4450: The interpretation semantics of a named word are also represented by an
4451: execution token. You can get it with
4452:
4453: doc-[']
4454: doc-'
4455:
4456: For literals, you use @code{'} in interpreted code and @code{[']} in
4457: compiled code. Gforth's @code{'} and @code{[']} behave somewhat unusual
4458: by complaining about compile-only words. To get an execution token for a
4459: compiling word @var{X}, use @code{COMP' @var{X} drop} or @code{[COMP']
4460: @var{X} drop}.
4461:
4462: @cindex compilation token
4463: The compilation semantics are represented by a @dfn{compilation token}
4464: consisting of two cells: @var{w xt}. The top cell @var{xt} is an
4465: execution token. The compilation semantics represented by the
4466: compilation token can be performed with @code{execute}, which consumes
4467: the whole compilation token, with an additional stack effect determined
4468: by the represented compilation semantics.
4469:
4470: doc-[comp']
4471: doc-comp'
4472:
4473: You can compile the compilation semantics with @code{postpone,}. I.e.,
4474: @code{COMP' @var{word} POSTPONE,} is equivalent to @code{POSTPONE
4475: @var{word}}.
4476:
4477: doc-postpone,
4478:
4479: At present, the @var{w} part of a compilation token is an execution
4480: token, and the @var{xt} part represents either @code{execute} or
4481: @code{compile,}. However, don't rely on that knowledge, unless necessary;
4482: we may introduce unusual compilation tokens in the future (e.g.,
4483: compilation tokens representing the compilation semantics of literals).
4484:
4485: @cindex name token
4486: @cindex name field address
4487: @cindex NFA
4488: Named words are also represented by the @dfn{name token}. The abstract
4489: data type @emph{name token} is implemented as NFA (name field address).
4490:
4491: doc-find-name
4492: doc-name>int
4493: doc-name?int
4494: doc-name>comp
4495: doc-name>string
4496:
4497: @node Wordlists, Files, Tokens for Words, Words
4498: @section Wordlists
4499:
1.12 anton 4500: @node Files, Including Files, Wordlists, Words
1.1 anton 4501: @section Files
4502:
1.12 anton 4503: @node Including Files, Blocks, Files, Words
4504: @section Including Files
4505: @cindex including files
4506:
4507: @menu
4508: * Words for Including::
4509: * Search Path::
4510: * Changing the Search Path::
4511: * General Search Paths::
4512: @end menu
4513:
4514: @node Words for Including, Search Path, Including Files, Including Files
4515: @subsection Words for Including
4516:
4517: doc-include-file
4518: doc-included
4519: doc-include
4520:
4521: Usually you want to include a file only if it is not included already
4522: (by, say, another source file):
4523:
4524: doc-required
4525: doc-require
4526: doc-needs
4527:
4528: @cindex stack effect of included files
4529: @cindex including files, stack effect
4530: I recommend that you write your source files such that interpreting them
4531: does not change the stack. This allows using these files with
4532: @code{required} and friends without complications. E.g.,
4533:
4534: @example
4535: 1 require foo.fs drop
4536: @end example
4537:
4538: @node Search Path, Changing the Search Path, Words for Including, Including Files
4539: @subsection Search Path
4540: @cindex path for @code{included}
4541: @cindex file search path
4542: @cindex include search path
4543: @cindex search path for files
4544:
4545: If you specify an absolute filename (i.e., a filename starting with
4546: @file{/} or @file{~}, or with @file{:} in the second position (as in
4547: @samp{C:...})) for @code{included} and friends, that file is included
4548: just as you would expect.
4549:
4550: For relative filenames, Gforth uses a search path similar to Forth's
4551: search order (@pxref{Wordlists}). It tries to find the given filename in
4552: the directories present in the path, and includes the first one it
4553: finds.
4554:
4555: If the search path contains the directory @file{.} (as it should), this
4556: refers to the directory that the present file was @code{included}
4557: from. This allows files to include other files relative to their own
4558: position (irrespective of the current working directory or the absolute
4559: position). This feature is essential for libraries consisting of
4560: several files, where a file may include other files from the library.
4561: It corresponds to @code{#include "..."} in C. If the current input
4562: source is not a file, @file{.} refers to the directory of the innermost
4563: file being included, or, if there is no file being included, to the
4564: current working directory.
4565:
4566: Use @file{~+} to refer to the current working directory (as in the
4567: @code{bash}).
4568:
4569: If the filename starts with @file{./}, the search path is not searched
4570: (just as with absolute filenames), and the @file{.} has the same meaning
4571: as described above.
4572:
4573: @node Changing the Search Path, General Search Paths, Search Path, Including Files
4574: @subsection Changing the Search Path
4575: @cindex search path, changes
4576:
4577: The search path is initialized when you start Gforth (@pxref{Invoking
4578: Gforth}). You can display it with
4579:
4580: doc-.fpath
4581:
4582: You can change it later with the following words:
4583:
4584: doc-fpath+
4585: doc-fpath=
4586:
4587: Using fpath and require would look like:
4588:
4589: @example
4590: fpath= /usr/lib/forth/|./
4591:
4592: require timer.fs
4593: @end example
4594:
4595: If you have the need to look for a file in the Forth search path, you could
4596: use this Gforth feature in your application.
4597:
4598: doc-open-fpath-file
4599:
4600:
4601: @node General Search Paths, , Changing the Search Path, Including Files
4602: @subsection General Search Paths
4603: @cindex search paths for user applications
4604:
4605: Your application may need to search files in sevaral directories, like
4606: @code{included} does. For this purpose you can define and use your own
4607: search paths. Create a search path like this:
4608:
4609: @example
4610:
4611: Make a buffer for the path:
4612: create mypath 100 chars , \ maximum length (is checked)
4613: 0 , \ real len
4614: 100 chars allot \ space for path
4615: @end example
4616:
4617: You have the same functions for the forth search path in a generic version
4618: for different paths.
4619:
4620: doc-path+
4621: doc-path=
4622: doc-.path
4623: doc-open-path-file
4624:
4625:
4626: @node Blocks, Other I/O, Including Files, Words
1.1 anton 4627: @section Blocks
4628:
4629: @node Other I/O, Programming Tools, Blocks, Words
4630: @section Other I/O
4631:
1.7 pazsan 4632: @node Programming Tools, Assembler and Code Words, Other I/O, Words
1.1 anton 4633: @section Programming Tools
4634: @cindex programming tools
4635:
4636: @menu
4637: * Debugging:: Simple and quick.
4638: * Assertions:: Making your programs self-checking.
1.6 pazsan 4639: * Singlestep Debugger:: Executing your program word by word.
1.1 anton 4640: @end menu
4641:
4642: @node Debugging, Assertions, Programming Tools, Programming Tools
4643: @subsection Debugging
4644: @cindex debugging
4645:
1.2 jwilke 4646: The simple debugging aids provided in @file{debugs.fs}
1.1 anton 4647: are meant to support a different style of debugging than the
4648: tracing/stepping debuggers used in languages with long turn-around
4649: times.
4650:
4651: A much better (faster) way in fast-compiling languages is to add
4652: printing code at well-selected places, let the program run, look at
4653: the output, see where things went wrong, add more printing code, etc.,
4654: until the bug is found.
4655:
4656: The word @code{~~} is easy to insert. It just prints debugging
4657: information (by default the source location and the stack contents). It
4658: is also easy to remove (@kbd{C-x ~} in the Emacs Forth mode to
4659: query-replace them with nothing). The deferred words
4660: @code{printdebugdata} and @code{printdebugline} control the output of
4661: @code{~~}. The default source location output format works well with
4662: Emacs' compilation mode, so you can step through the program at the
4663: source level using @kbd{C-x `} (the advantage over a stepping debugger
4664: is that you can step in any direction and you know where the crash has
4665: happened or where the strange data has occurred).
4666:
4667: Note that the default actions clobber the contents of the pictured
4668: numeric output string, so you should not use @code{~~}, e.g., between
4669: @code{<#} and @code{#>}.
4670:
4671: doc-~~
4672: doc-printdebugdata
4673: doc-printdebugline
4674:
1.2 jwilke 4675: @node Assertions, Singlestep Debugger, Debugging, Programming Tools
1.1 anton 4676: @subsection Assertions
4677: @cindex assertions
4678:
4679: It is a good idea to make your programs self-checking, in particular, if
4680: you use an assumption (e.g., that a certain field of a data structure is
4681: never zero) that may become wrong during maintenance. Gforth supports
4682: assertions for this purpose. They are used like this:
4683:
4684: @example
4685: assert( @var{flag} )
4686: @end example
4687:
4688: The code between @code{assert(} and @code{)} should compute a flag, that
4689: should be true if everything is alright and false otherwise. It should
4690: not change anything else on the stack. The overall stack effect of the
4691: assertion is @code{( -- )}. E.g.
4692:
4693: @example
4694: assert( 1 1 + 2 = ) \ what we learn in school
4695: assert( dup 0<> ) \ assert that the top of stack is not zero
4696: assert( false ) \ this code should not be reached
4697: @end example
4698:
4699: The need for assertions is different at different times. During
4700: debugging, we want more checking, in production we sometimes care more
4701: for speed. Therefore, assertions can be turned off, i.e., the assertion
4702: becomes a comment. Depending on the importance of an assertion and the
4703: time it takes to check it, you may want to turn off some assertions and
4704: keep others turned on. Gforth provides several levels of assertions for
4705: this purpose:
4706:
4707: doc-assert0(
4708: doc-assert1(
4709: doc-assert2(
4710: doc-assert3(
4711: doc-assert(
4712: doc-)
4713:
4714: @code{Assert(} is the same as @code{assert1(}. The variable
4715: @code{assert-level} specifies the highest assertions that are turned
4716: on. I.e., at the default @code{assert-level} of one, @code{assert0(} and
4717: @code{assert1(} assertions perform checking, while @code{assert2(} and
4718: @code{assert3(} assertions are treated as comments.
4719:
4720: Note that the @code{assert-level} is evaluated at compile-time, not at
4721: run-time. I.e., you cannot turn assertions on or off at run-time, you
4722: have to set the @code{assert-level} appropriately before compiling a
4723: piece of code. You can compile several pieces of code at several
4724: @code{assert-level}s (e.g., a trusted library at level 1 and newly
4725: written code at level 3).
4726:
4727: doc-assert-level
4728:
4729: If an assertion fails, a message compatible with Emacs' compilation mode
4730: is produced and the execution is aborted (currently with @code{ABORT"}.
4731: If there is interest, we will introduce a special throw code. But if you
4732: intend to @code{catch} a specific condition, using @code{throw} is
4733: probably more appropriate than an assertion).
4734:
1.2 jwilke 4735: @node Singlestep Debugger, , Assertions, Programming Tools
4736: @subsection Singlestep Debugger
4737: @cindex singlestep Debugger
4738: @cindex debugging Singlestep
4739: @cindex @code{dbg}
4740: @cindex @code{BREAK:}
4741: @cindex @code{BREAK"}
4742:
4743: When a new word is created there's often the need to check whether it behaves
1.5 anton 4744: correctly or not. You can do this by typing @code{dbg badword}. This might
1.2 jwilke 4745: look like:
4746: @example
4747: : badword 0 DO i . LOOP ; ok
4748: 2 dbg badword
4749: : badword
4750: Scanning code...
4751:
4752: Nesting debugger ready!
4753:
4754: 400D4738 8049BC4 0 -> [ 2 ] 00002 00000
4755: 400D4740 8049F68 DO -> [ 0 ]
4756: 400D4744 804A0C8 i -> [ 1 ] 00000
4757: 400D4748 400C5E60 . -> 0 [ 0 ]
4758: 400D474C 8049D0C LOOP -> [ 0 ]
4759: 400D4744 804A0C8 i -> [ 1 ] 00001
4760: 400D4748 400C5E60 . -> 1 [ 0 ]
4761: 400D474C 8049D0C LOOP -> [ 0 ]
4762: 400D4758 804B384 ; -> ok
4763: @end example
4764:
1.5 anton 4765: Each line displayed is one step. You always have to hit return to
4766: execute the next word that is displayed. If you don't want to execute
4767: the next word in a whole, you have to type @kbd{n} for @code{nest}. Here is
4768: an overview what keys are available:
1.2 jwilke 4769:
4770: @table @i
4771:
1.4 anton 4772: @item <return>
1.5 anton 4773: Next; Execute the next word.
1.2 jwilke 4774:
4775: @item n
1.5 anton 4776: Nest; Single step through next word.
1.2 jwilke 4777:
4778: @item u
1.5 anton 4779: Unnest; Stop debugging and execute rest of word. If we got to this word
4780: with nest, continue debugging with the calling word.
1.2 jwilke 4781:
4782: @item d
1.5 anton 4783: Done; Stop debugging and execute rest.
1.2 jwilke 4784:
4785: @item s
1.5 anton 4786: Stopp; Abort immediately.
1.2 jwilke 4787:
4788: @end table
4789:
4790: Debugging large application with this mechanism is very difficult, because
4791: you have to nest very deep into the program before the interesting part
4792: begins. This takes a lot of time.
4793:
4794: To do it more directly put a @code{BREAK:} command into your source code.
4795: When program execution reaches @code{BREAK:} the single step debugger is
4796: invoked and you have all the features described above.
4797:
4798: If you have more than one part to debug it is useful to know where the
4799: program has stopped at the moment. You can do this by the
4800: @code{BREAK" string"} command. This behaves like @code{BREAK:} except that
4801: string is typed out when the ``breakpoint'' is reached.
4802:
1.7 pazsan 4803: @node Assembler and Code Words, Threading Words, Programming Tools, Words
4804: @section Assembler and Code Words
1.1 anton 4805: @cindex assembler
4806: @cindex code words
4807:
4808: Gforth provides some words for defining primitives (words written in
4809: machine code), and for defining the the machine-code equivalent of
4810: @code{DOES>}-based defining words. However, the machine-independent
4811: nature of Gforth poses a few problems: First of all, Gforth runs on
4812: several architectures, so it can provide no standard assembler. What's
4813: worse is that the register allocation not only depends on the processor,
4814: but also on the @code{gcc} version and options used.
4815:
4816: The words that Gforth offers encapsulate some system dependences (e.g., the
4817: header structure), so a system-independent assembler may be used in
4818: Gforth. If you do not have an assembler, you can compile machine code
4819: directly with @code{,} and @code{c,}.
4820:
4821: doc-assembler
4822: doc-code
4823: doc-end-code
4824: doc-;code
4825: doc-flush-icache
4826:
4827: If @code{flush-icache} does not work correctly, @code{code} words
4828: etc. will not work (reliably), either.
4829:
4830: These words are rarely used. Therefore they reside in @code{code.fs},
4831: which is usually not loaded (except @code{flush-icache}, which is always
4832: present). You can load them with @code{require code.fs}.
4833:
4834: @cindex registers of the inner interpreter
4835: In the assembly code you will want to refer to the inner interpreter's
4836: registers (e.g., the data stack pointer) and you may want to use other
4837: registers for temporary storage. Unfortunately, the register allocation
4838: is installation-dependent.
4839:
4840: The easiest solution is to use explicit register declarations
4841: (@pxref{Explicit Reg Vars, , Variables in Specified Registers, gcc.info,
4842: GNU C Manual}) for all of the inner interpreter's registers: You have to
4843: compile Gforth with @code{-DFORCE_REG} (configure option
4844: @code{--enable-force-reg}) and the appropriate declarations must be
4845: present in the @code{machine.h} file (see @code{mips.h} for an example;
4846: you can find a full list of all declarable register symbols with
4847: @code{grep register engine.c}). If you give explicit registers to all
4848: variables that are declared at the beginning of @code{engine()}, you
4849: should be able to use the other caller-saved registers for temporary
4850: storage. Alternatively, you can use the @code{gcc} option
4851: @code{-ffixed-REG} (@pxref{Code Gen Options, , Options for Code
4852: Generation Conventions, gcc.info, GNU C Manual}) to reserve a register
4853: (however, this restriction on register allocation may slow Gforth
4854: significantly).
4855:
4856: If this solution is not viable (e.g., because @code{gcc} does not allow
4857: you to explicitly declare all the registers you need), you have to find
4858: out by looking at the code where the inner interpreter's registers
4859: reside and which registers can be used for temporary storage. You can
4860: get an assembly listing of the engine's code with @code{make engine.s}.
4861:
4862: In any case, it is good practice to abstract your assembly code from the
4863: actual register allocation. E.g., if the data stack pointer resides in
4864: register @code{$17}, create an alias for this register called @code{sp},
4865: and use that in your assembly code.
4866:
4867: @cindex code words, portable
4868: Another option for implementing normal and defining words efficiently
4869: is: adding the wanted functionality to the source of Gforth. For normal
4870: words you just have to edit @file{primitives} (@pxref{Automatic
4871: Generation}), defining words (equivalent to @code{;CODE} words, for fast
4872: defined words) may require changes in @file{engine.c}, @file{kernal.fs},
4873: @file{prims2x.fs}, and possibly @file{cross.fs}.
4874:
4875:
1.12 anton 4876: @node Threading Words, , Assembler and Code Words, Words
1.1 anton 4877: @section Threading Words
4878: @cindex threading words
4879:
4880: @cindex code address
4881: These words provide access to code addresses and other threading stuff
4882: in Gforth (and, possibly, other interpretive Forths). It more or less
4883: abstracts away the differences between direct and indirect threading
4884: (and, for direct threading, the machine dependences). However, at
4885: present this wordset is still incomplete. It is also pretty low-level;
4886: some day it will hopefully be made unnecessary by an internals wordset
4887: that abstracts implementation details away completely.
4888:
4889: doc->code-address
4890: doc->does-code
4891: doc-code-address!
4892: doc-does-code!
4893: doc-does-handler!
4894: doc-/does-handler
4895:
4896: The code addresses produced by various defining words are produced by
4897: the following words:
4898:
4899: doc-docol:
4900: doc-docon:
4901: doc-dovar:
4902: doc-douser:
4903: doc-dodefer:
4904: doc-dofield:
4905:
4906: You can recognize words defined by a @code{CREATE}...@code{DOES>} word
4907: with @code{>DOES-CODE}. If the word was defined in that way, the value
4908: returned is different from 0 and identifies the @code{DOES>} used by the
4909: defining word.
1.2 jwilke 4910:
1.5 anton 4911: @c ******************************************************************
1.1 anton 4912: @node Tools, ANS conformance, Words, Top
4913: @chapter Tools
4914:
4915: @menu
4916: * ANS Report:: Report the words used, sorted by wordset.
4917: @end menu
4918:
4919: See also @ref{Emacs and Gforth}.
4920:
4921: @node ANS Report, , Tools, Tools
4922: @section @file{ans-report.fs}: Report the words used, sorted by wordset
4923: @cindex @file{ans-report.fs}
4924: @cindex report the words used in your program
4925: @cindex words used in your program
4926:
4927: If you want to label a Forth program as ANS Forth Program, you must
4928: document which wordsets the program uses; for extension wordsets, it is
4929: helpful to list the words the program requires from these wordsets
4930: (because Forth systems are allowed to provide only some words of them).
4931:
4932: The @file{ans-report.fs} tool makes it easy for you to determine which
4933: words from which wordset and which non-ANS words your application
4934: uses. You simply have to include @file{ans-report.fs} before loading the
4935: program you want to check. After loading your program, you can get the
4936: report with @code{print-ans-report}. A typical use is to run this as
4937: batch job like this:
4938: @example
4939: gforth ans-report.fs myprog.fs -e "print-ans-report bye"
4940: @end example
4941:
4942: The output looks like this (for @file{compat/control.fs}):
4943: @example
4944: The program uses the following words
4945: from CORE :
4946: : POSTPONE THEN ; immediate ?dup IF 0=
4947: from BLOCK-EXT :
4948: \
4949: from FILE :
4950: (
4951: @end example
4952:
4953: @subsection Caveats
4954:
4955: Note that @file{ans-report.fs} just checks which words are used, not whether
4956: they are used in an ANS Forth conforming way!
4957:
4958: Some words are defined in several wordsets in the
4959: standard. @file{ans-report.fs} reports them for only one of the
4960: wordsets, and not necessarily the one you expect. It depends on usage
4961: which wordset is the right one to specify. E.g., if you only use the
4962: compilation semantics of @code{S"}, it is a Core word; if you also use
4963: its interpretation semantics, it is a File word.
4964:
4965: @c ******************************************************************
4966: @node ANS conformance, Model, Tools, Top
4967: @chapter ANS conformance
4968: @cindex ANS conformance of Gforth
4969:
4970: To the best of our knowledge, Gforth is an
4971:
4972: ANS Forth System
4973: @itemize @bullet
4974: @item providing the Core Extensions word set
4975: @item providing the Block word set
4976: @item providing the Block Extensions word set
4977: @item providing the Double-Number word set
4978: @item providing the Double-Number Extensions word set
4979: @item providing the Exception word set
4980: @item providing the Exception Extensions word set
4981: @item providing the Facility word set
4982: @item providing @code{MS} and @code{TIME&DATE} from the Facility Extensions word set
4983: @item providing the File Access word set
4984: @item providing the File Access Extensions word set
4985: @item providing the Floating-Point word set
4986: @item providing the Floating-Point Extensions word set
4987: @item providing the Locals word set
4988: @item providing the Locals Extensions word set
4989: @item providing the Memory-Allocation word set
4990: @item providing the Memory-Allocation Extensions word set (that one's easy)
4991: @item providing the Programming-Tools word set
4992: @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
4993: @item providing the Search-Order word set
4994: @item providing the Search-Order Extensions word set
4995: @item providing the String word set
4996: @item providing the String Extensions word set (another easy one)
4997: @end itemize
4998:
4999: @cindex system documentation
5000: In addition, ANS Forth systems are required to document certain
5001: implementation choices. This chapter tries to meet these
5002: requirements. In many cases it gives a way to ask the system for the
5003: information instead of providing the information directly, in
5004: particular, if the information depends on the processor, the operating
5005: system or the installation options chosen, or if they are likely to
5006: change during the maintenance of Gforth.
5007:
5008: @comment The framework for the rest has been taken from pfe.
5009:
5010: @menu
5011: * The Core Words::
5012: * The optional Block word set::
5013: * The optional Double Number word set::
5014: * The optional Exception word set::
5015: * The optional Facility word set::
5016: * The optional File-Access word set::
5017: * The optional Floating-Point word set::
5018: * The optional Locals word set::
5019: * The optional Memory-Allocation word set::
5020: * The optional Programming-Tools word set::
5021: * The optional Search-Order word set::
5022: @end menu
5023:
5024:
5025: @c =====================================================================
5026: @node The Core Words, The optional Block word set, ANS conformance, ANS conformance
5027: @comment node-name, next, previous, up
5028: @section The Core Words
5029: @c =====================================================================
5030: @cindex core words, system documentation
5031: @cindex system documentation, core words
5032:
5033: @menu
5034: * core-idef:: Implementation Defined Options
5035: * core-ambcond:: Ambiguous Conditions
5036: * core-other:: Other System Documentation
5037: @end menu
5038:
5039: @c ---------------------------------------------------------------------
5040: @node core-idef, core-ambcond, The Core Words, The Core Words
5041: @subsection Implementation Defined Options
5042: @c ---------------------------------------------------------------------
5043: @cindex core words, implementation-defined options
5044: @cindex implementation-defined options, core words
5045:
5046:
5047: @table @i
5048: @item (Cell) aligned addresses:
5049: @cindex cell-aligned addresses
5050: @cindex aligned addresses
5051: processor-dependent. Gforth's alignment words perform natural alignment
5052: (e.g., an address aligned for a datum of size 8 is divisible by
5053: 8). Unaligned accesses usually result in a @code{-23 THROW}.
5054:
5055: @item @code{EMIT} and non-graphic characters:
5056: @cindex @code{EMIT} and non-graphic characters
5057: @cindex non-graphic characters and @code{EMIT}
5058: The character is output using the C library function (actually, macro)
5059: @code{putc}.
5060:
5061: @item character editing of @code{ACCEPT} and @code{EXPECT}:
5062: @cindex character editing of @code{ACCEPT} and @code{EXPECT}
5063: @cindex editing in @code{ACCEPT} and @code{EXPECT}
5064: @cindex @code{ACCEPT}, editing
5065: @cindex @code{EXPECT}, editing
5066: This is modeled on the GNU readline library (@pxref{Readline
5067: Interaction, , Command Line Editing, readline, The GNU Readline
5068: Library}) with Emacs-like key bindings. @kbd{Tab} deviates a little by
5069: producing a full word completion every time you type it (instead of
5070: producing the common prefix of all completions).
5071:
5072: @item character set:
5073: @cindex character set
5074: The character set of your computer and display device. Gforth is
5075: 8-bit-clean (but some other component in your system may make trouble).
5076:
5077: @item Character-aligned address requirements:
5078: @cindex character-aligned address requirements
5079: installation-dependent. Currently a character is represented by a C
5080: @code{unsigned char}; in the future we might switch to @code{wchar_t}
5081: (Comments on that requested).
5082:
5083: @item character-set extensions and matching of names:
5084: @cindex character-set extensions and matching of names
5085: @cindex case sensitivity for name lookup
5086: @cindex name lookup, case sensitivity
5087: @cindex locale and case sensitivity
5088: Any character except the ASCII NUL charcter can be used in a
5089: name. Matching is case-insensitive (except in @code{TABLE}s). The
5090: matching is performed using the C function @code{strncasecmp}, whose
5091: function is probably influenced by the locale. E.g., the @code{C} locale
5092: does not know about accents and umlauts, so they are matched
5093: case-sensitively in that locale. For portability reasons it is best to
5094: write programs such that they work in the @code{C} locale. Then one can
5095: use libraries written by a Polish programmer (who might use words
5096: containing ISO Latin-2 encoded characters) and by a French programmer
5097: (ISO Latin-1) in the same program (of course, @code{WORDS} will produce
5098: funny results for some of the words (which ones, depends on the font you
5099: are using)). Also, the locale you prefer may not be available in other
5100: operating systems. Hopefully, Unicode will solve these problems one day.
5101:
5102: @item conditions under which control characters match a space delimiter:
5103: @cindex space delimiters
5104: @cindex control characters as delimiters
5105: If @code{WORD} is called with the space character as a delimiter, all
5106: white-space characters (as identified by the C macro @code{isspace()})
5107: are delimiters. @code{PARSE}, on the other hand, treats space like other
5108: delimiters. @code{PARSE-WORD} treats space like @code{WORD}, but behaves
5109: like @code{PARSE} otherwise. @code{(NAME)}, which is used by the outer
5110: interpreter (aka text interpreter) by default, treats all white-space
5111: characters as delimiters.
5112:
5113: @item format of the control flow stack:
5114: @cindex control flow stack, format
5115: The data stack is used as control flow stack. The size of a control flow
5116: stack item in cells is given by the constant @code{cs-item-size}. At the
5117: time of this writing, an item consists of a (pointer to a) locals list
5118: (third), an address in the code (second), and a tag for identifying the
5119: item (TOS). The following tags are used: @code{defstart},
5120: @code{live-orig}, @code{dead-orig}, @code{dest}, @code{do-dest},
5121: @code{scopestart}.
5122:
5123: @item conversion of digits > 35
5124: @cindex digits > 35
5125: The characters @code{[\]^_'} are the digits with the decimal value
5126: 36@minus{}41. There is no way to input many of the larger digits.
5127:
5128: @item display after input terminates in @code{ACCEPT} and @code{EXPECT}:
5129: @cindex @code{EXPECT}, display after end of input
5130: @cindex @code{ACCEPT}, display after end of input
5131: The cursor is moved to the end of the entered string. If the input is
5132: terminated using the @kbd{Return} key, a space is typed.
5133:
5134: @item exception abort sequence of @code{ABORT"}:
5135: @cindex exception abort sequence of @code{ABORT"}
5136: @cindex @code{ABORT"}, exception abort sequence
5137: The error string is stored into the variable @code{"error} and a
5138: @code{-2 throw} is performed.
5139:
5140: @item input line terminator:
5141: @cindex input line terminator
5142: @cindex line terminator on input
5143: @cindex newline charcter on input
5144: For interactive input, @kbd{C-m} (CR) and @kbd{C-j} (LF) terminate
5145: lines. One of these characters is typically produced when you type the
5146: @kbd{Enter} or @kbd{Return} key.
5147:
5148: @item maximum size of a counted string:
5149: @cindex maximum size of a counted string
5150: @cindex counted string, maximum size
5151: @code{s" /counted-string" environment? drop .}. Currently 255 characters
5152: on all ports, but this may change.
5153:
5154: @item maximum size of a parsed string:
5155: @cindex maximum size of a parsed string
5156: @cindex parsed string, maximum size
5157: Given by the constant @code{/line}. Currently 255 characters.
5158:
5159: @item maximum size of a definition name, in characters:
5160: @cindex maximum size of a definition name, in characters
5161: @cindex name, maximum length
5162: 31
5163:
5164: @item maximum string length for @code{ENVIRONMENT?}, in characters:
5165: @cindex maximum string length for @code{ENVIRONMENT?}, in characters
5166: @cindex @code{ENVIRONMENT?} string length, maximum
5167: 31
5168:
5169: @item method of selecting the user input device:
5170: @cindex user input device, method of selecting
5171: The user input device is the standard input. There is currently no way to
5172: change it from within Gforth. However, the input can typically be
5173: redirected in the command line that starts Gforth.
5174:
5175: @item method of selecting the user output device:
5176: @cindex user output device, method of selecting
5177: @code{EMIT} and @code{TYPE} output to the file-id stored in the value
1.10 anton 5178: @code{outfile-id} (@code{stdout} by default). Gforth uses unbuffered
5179: output when the user output device is a terminal, otherwise the output
5180: is buffered.
1.1 anton 5181:
5182: @item methods of dictionary compilation:
5183: What are we expected to document here?
5184:
5185: @item number of bits in one address unit:
5186: @cindex number of bits in one address unit
5187: @cindex address unit, size in bits
5188: @code{s" address-units-bits" environment? drop .}. 8 in all current
5189: ports.
5190:
5191: @item number representation and arithmetic:
5192: @cindex number representation and arithmetic
5193: Processor-dependent. Binary two's complement on all current ports.
5194:
5195: @item ranges for integer types:
5196: @cindex ranges for integer types
5197: @cindex integer types, ranges
5198: Installation-dependent. Make environmental queries for @code{MAX-N},
5199: @code{MAX-U}, @code{MAX-D} and @code{MAX-UD}. The lower bounds for
5200: unsigned (and positive) types is 0. The lower bound for signed types on
5201: two's complement and one's complement machines machines can be computed
5202: by adding 1 to the upper bound.
5203:
5204: @item read-only data space regions:
5205: @cindex read-only data space regions
5206: @cindex data-space, read-only regions
5207: The whole Forth data space is writable.
5208:
5209: @item size of buffer at @code{WORD}:
5210: @cindex size of buffer at @code{WORD}
5211: @cindex @code{WORD} buffer size
5212: @code{PAD HERE - .}. 104 characters on 32-bit machines. The buffer is
5213: shared with the pictured numeric output string. If overwriting
5214: @code{PAD} is acceptable, it is as large as the remaining dictionary
5215: space, although only as much can be sensibly used as fits in a counted
5216: string.
5217:
5218: @item size of one cell in address units:
5219: @cindex cell size
5220: @code{1 cells .}.
5221:
5222: @item size of one character in address units:
5223: @cindex char size
5224: @code{1 chars .}. 1 on all current ports.
5225:
5226: @item size of the keyboard terminal buffer:
5227: @cindex size of the keyboard terminal buffer
5228: @cindex terminal buffer, size
5229: Varies. You can determine the size at a specific time using @code{lp@@
5230: tib - .}. It is shared with the locals stack and TIBs of files that
5231: include the current file. You can change the amount of space for TIBs
5232: and locals stack at Gforth startup with the command line option
5233: @code{-l}.
5234:
5235: @item size of the pictured numeric output buffer:
5236: @cindex size of the pictured numeric output buffer
5237: @cindex pictured numeric output buffer, size
5238: @code{PAD HERE - .}. 104 characters on 32-bit machines. The buffer is
5239: shared with @code{WORD}.
5240:
5241: @item size of the scratch area returned by @code{PAD}:
5242: @cindex size of the scratch area returned by @code{PAD}
5243: @cindex @code{PAD} size
5244: The remainder of dictionary space. @code{unused pad here - - .}.
5245:
5246: @item system case-sensitivity characteristics:
5247: @cindex case-sensitivity characteristics
5248: Dictionary searches are case insensitive (except in
5249: @code{TABLE}s). However, as explained above under @i{character-set
5250: extensions}, the matching for non-ASCII characters is determined by the
5251: locale you are using. In the default @code{C} locale all non-ASCII
5252: characters are matched case-sensitively.
5253:
5254: @item system prompt:
5255: @cindex system prompt
5256: @cindex prompt
5257: @code{ ok} in interpret state, @code{ compiled} in compile state.
5258:
5259: @item division rounding:
5260: @cindex division rounding
5261: installation dependent. @code{s" floored" environment? drop .}. We leave
5262: the choice to @code{gcc} (what to use for @code{/}) and to you (whether
5263: to use @code{fm/mod}, @code{sm/rem} or simply @code{/}).
5264:
5265: @item values of @code{STATE} when true:
5266: @cindex @code{STATE} values
5267: -1.
5268:
5269: @item values returned after arithmetic overflow:
5270: On two's complement machines, arithmetic is performed modulo
5271: 2**bits-per-cell for single arithmetic and 4**bits-per-cell for double
5272: arithmetic (with appropriate mapping for signed types). Division by zero
5273: typically results in a @code{-55 throw} (Floating-point unidentified
5274: fault), although a @code{-10 throw} (divide by zero) would be more
5275: appropriate.
5276:
5277: @item whether the current definition can be found after @t{DOES>}:
5278: @cindex @t{DOES>}, visibility of current definition
5279: No.
5280:
5281: @end table
5282:
5283: @c ---------------------------------------------------------------------
5284: @node core-ambcond, core-other, core-idef, The Core Words
5285: @subsection Ambiguous conditions
5286: @c ---------------------------------------------------------------------
5287: @cindex core words, ambiguous conditions
5288: @cindex ambiguous conditions, core words
5289:
5290: @table @i
5291:
5292: @item a name is neither a word nor a number:
5293: @cindex name not found
5294: @cindex Undefined word
5295: @code{-13 throw} (Undefined word). Actually, @code{-13 bounce}, which
5296: preserves the data and FP stack, so you don't lose more work than
5297: necessary.
5298:
5299: @item a definition name exceeds the maximum length allowed:
5300: @cindex Word name too long
5301: @code{-19 throw} (Word name too long)
5302:
5303: @item addressing a region not inside the various data spaces of the forth system:
5304: @cindex Invalid memory address
5305: The stacks, code space and name space are accessible. Machine code space is
5306: typically readable. Accessing other addresses gives results dependent on
5307: the operating system. On decent systems: @code{-9 throw} (Invalid memory
5308: address).
5309:
5310: @item argument type incompatible with parameter:
5311: @cindex Argument type mismatch
5312: This is usually not caught. Some words perform checks, e.g., the control
5313: flow words, and issue a @code{ABORT"} or @code{-12 THROW} (Argument type
5314: mismatch).
5315:
5316: @item attempting to obtain the execution token of a word with undefined execution semantics:
5317: @cindex Interpreting a compile-only word, for @code{'} etc.
5318: @cindex execution token of words with undefined execution semantics
5319: @code{-14 throw} (Interpreting a compile-only word). In some cases, you
5320: get an execution token for @code{compile-only-error} (which performs a
5321: @code{-14 throw} when executed).
5322:
5323: @item dividing by zero:
5324: @cindex dividing by zero
5325: @cindex floating point unidentified fault, integer division
5326: @cindex divide by zero
5327: typically results in a @code{-55 throw} (floating point unidentified
5328: fault), although a @code{-10 throw} (divide by zero) would be more
5329: appropriate.
5330:
5331: @item insufficient data stack or return stack space:
5332: @cindex insufficient data stack or return stack space
5333: @cindex stack overflow
5334: @cindex Address alignment exception, stack overflow
5335: @cindex Invalid memory address, stack overflow
5336: Depending on the operating system, the installation, and the invocation
5337: of Gforth, this is either checked by the memory management hardware, or
5338: it is not checked. If it is checked, you typically get a @code{-9 throw}
5339: (Invalid memory address) as soon as the overflow happens. If it is not
5340: check, overflows typically result in mysterious illegal memory accesses,
5341: producing @code{-9 throw} (Invalid memory address) or @code{-23 throw}
5342: (Address alignment exception); they might also destroy the internal data
5343: structure of @code{ALLOCATE} and friends, resulting in various errors in
5344: these words.
5345:
5346: @item insufficient space for loop control parameters:
5347: @cindex insufficient space for loop control parameters
5348: like other return stack overflows.
5349:
5350: @item insufficient space in the dictionary:
5351: @cindex insufficient space in the dictionary
5352: @cindex dictionary overflow
1.12 anton 5353: If you try to allot (either directly with @code{allot}, or indirectly
5354: with @code{,}, @code{create} etc.) more memory than available in the
5355: dictionary, you get a @code{-8 throw} (Dictionary overflow). If you try
5356: to access memory beyond the end of the dictionary, the results are
5357: similar to stack overflows.
1.1 anton 5358:
5359: @item interpreting a word with undefined interpretation semantics:
5360: @cindex interpreting a word with undefined interpretation semantics
5361: @cindex Interpreting a compile-only word
5362: For some words, we have defined interpretation semantics. For the
5363: others: @code{-14 throw} (Interpreting a compile-only word).
5364:
5365: @item modifying the contents of the input buffer or a string literal:
5366: @cindex modifying the contents of the input buffer or a string literal
5367: These are located in writable memory and can be modified.
5368:
5369: @item overflow of the pictured numeric output string:
5370: @cindex overflow of the pictured numeric output string
5371: @cindex pictured numeric output string, overflow
5372: Not checked. Runs into the dictionary and destroys it (at least,
5373: partially).
5374:
5375: @item parsed string overflow:
5376: @cindex parsed string overflow
5377: @code{PARSE} cannot overflow. @code{WORD} does not check for overflow.
5378:
5379: @item producing a result out of range:
5380: @cindex result out of range
5381: On two's complement machines, arithmetic is performed modulo
5382: 2**bits-per-cell for single arithmetic and 4**bits-per-cell for double
5383: arithmetic (with appropriate mapping for signed types). Division by zero
5384: typically results in a @code{-55 throw} (floatingpoint unidentified
5385: fault), although a @code{-10 throw} (divide by zero) would be more
5386: appropriate. @code{convert} and @code{>number} currently overflow
5387: silently.
5388:
5389: @item reading from an empty data or return stack:
5390: @cindex stack empty
5391: @cindex stack underflow
5392: The data stack is checked by the outer (aka text) interpreter after
5393: every word executed. If it has underflowed, a @code{-4 throw} (Stack
5394: underflow) is performed. Apart from that, stacks may be checked or not,
5395: depending on operating system, installation, and invocation. The
5396: consequences of stack underflows are similar to the consequences of
5397: stack overflows. Note that even if the system uses checking (through the
5398: MMU), your program may have to underflow by a significant number of
5399: stack items to trigger the reaction (the reason for this is that the
5400: MMU, and therefore the checking, works with a page-size granularity).
5401:
5402: @item unexpected end of the input buffer, resulting in an attempt to use a zero-length string as a name:
5403: @cindex unexpected end of the input buffer
5404: @cindex zero-length string as a name
5405: @cindex Attempt to use zero-length string as a name
5406: @code{Create} and its descendants perform a @code{-16 throw} (Attempt to
5407: use zero-length string as a name). Words like @code{'} probably will not
5408: find what they search. Note that it is possible to create zero-length
5409: names with @code{nextname} (should it not?).
5410:
5411: @item @code{>IN} greater than input buffer:
5412: @cindex @code{>IN} greater than input buffer
5413: The next invocation of a parsing word returns a string with length 0.
5414:
5415: @item @code{RECURSE} appears after @code{DOES>}:
5416: @cindex @code{RECURSE} appears after @code{DOES>}
5417: Compiles a recursive call to the defining word, not to the defined word.
5418:
5419: @item argument input source different than current input source for @code{RESTORE-INPUT}:
5420: @cindex argument input source different than current input source for @code{RESTORE-INPUT}
5421: @cindex Argument type mismatch, @code{RESTORE-INPUT}
5422: @cindex @code{RESTORE-INPUT}, Argument type mismatch
5423: @code{-12 THROW}. Note that, once an input file is closed (e.g., because
5424: the end of the file was reached), its source-id may be
5425: reused. Therefore, restoring an input source specification referencing a
5426: closed file may lead to unpredictable results instead of a @code{-12
5427: THROW}.
5428:
5429: In the future, Gforth may be able to restore input source specifications
5430: from other than the current input source.
5431:
5432: @item data space containing definitions gets de-allocated:
5433: @cindex data space containing definitions gets de-allocated
5434: Deallocation with @code{allot} is not checked. This typically results in
5435: memory access faults or execution of illegal instructions.
5436:
5437: @item data space read/write with incorrect alignment:
5438: @cindex data space read/write with incorrect alignment
5439: @cindex alignment faults
5440: @cindex Address alignment exception
5441: Processor-dependent. Typically results in a @code{-23 throw} (Address
1.12 anton 5442: alignment exception). Under Linux-Intel on a 486 or later processor with
1.1 anton 5443: alignment turned on, incorrect alignment results in a @code{-9 throw}
5444: (Invalid memory address). There are reportedly some processors with
1.12 anton 5445: alignment restrictions that do not report violations.
1.1 anton 5446:
5447: @item data space pointer not properly aligned, @code{,}, @code{C,}:
5448: @cindex data space pointer not properly aligned, @code{,}, @code{C,}
5449: Like other alignment errors.
5450:
5451: @item less than u+2 stack items (@code{PICK} and @code{ROLL}):
5452: Like other stack underflows.
5453:
5454: @item loop control parameters not available:
5455: @cindex loop control parameters not available
5456: Not checked. The counted loop words simply assume that the top of return
5457: stack items are loop control parameters and behave accordingly.
5458:
5459: @item most recent definition does not have a name (@code{IMMEDIATE}):
5460: @cindex most recent definition does not have a name (@code{IMMEDIATE})
5461: @cindex last word was headerless
5462: @code{abort" last word was headerless"}.
5463:
5464: @item name not defined by @code{VALUE} used by @code{TO}:
5465: @cindex name not defined by @code{VALUE} used by @code{TO}
5466: @cindex @code{TO} on non-@code{VALUE}s
5467: @cindex Invalid name argument, @code{TO}
5468: @code{-32 throw} (Invalid name argument) (unless name is a local or was
5469: defined by @code{CONSTANT}; in the latter case it just changes the constant).
5470:
5471: @item name not found (@code{'}, @code{POSTPONE}, @code{[']}, @code{[COMPILE]}):
5472: @cindex name not found (@code{'}, @code{POSTPONE}, @code{[']}, @code{[COMPILE]})
5473: @cindex Undefined word, @code{'}, @code{POSTPONE}, @code{[']}, @code{[COMPILE]}
5474: @code{-13 throw} (Undefined word)
5475:
5476: @item parameters are not of the same type (@code{DO}, @code{?DO}, @code{WITHIN}):
5477: @cindex parameters are not of the same type (@code{DO}, @code{?DO}, @code{WITHIN})
5478: Gforth behaves as if they were of the same type. I.e., you can predict
5479: the behaviour by interpreting all parameters as, e.g., signed.
5480:
5481: @item @code{POSTPONE} or @code{[COMPILE]} applied to @code{TO}:
5482: @cindex @code{POSTPONE} or @code{[COMPILE]} applied to @code{TO}
5483: Assume @code{: X POSTPONE TO ; IMMEDIATE}. @code{X} performs the
5484: compilation semantics of @code{TO}.
5485:
5486: @item String longer than a counted string returned by @code{WORD}:
5487: @cindex String longer than a counted string returned by @code{WORD}
5488: @cindex @code{WORD}, string overflow
5489: Not checked. The string will be ok, but the count will, of course,
5490: contain only the least significant bits of the length.
5491:
5492: @item u greater than or equal to the number of bits in a cell (@code{LSHIFT}, @code{RSHIFT}):
5493: @cindex @code{LSHIFT}, large shift counts
5494: @cindex @code{RSHIFT}, large shift counts
5495: Processor-dependent. Typical behaviours are returning 0 and using only
5496: the low bits of the shift count.
5497:
5498: @item word not defined via @code{CREATE}:
5499: @cindex @code{>BODY} of non-@code{CREATE}d words
5500: @code{>BODY} produces the PFA of the word no matter how it was defined.
5501:
5502: @cindex @code{DOES>} of non-@code{CREATE}d words
5503: @code{DOES>} changes the execution semantics of the last defined word no
5504: matter how it was defined. E.g., @code{CONSTANT DOES>} is equivalent to
5505: @code{CREATE , DOES>}.
5506:
5507: @item words improperly used outside @code{<#} and @code{#>}:
5508: Not checked. As usual, you can expect memory faults.
5509:
5510: @end table
5511:
5512:
5513: @c ---------------------------------------------------------------------
5514: @node core-other, , core-ambcond, The Core Words
5515: @subsection Other system documentation
5516: @c ---------------------------------------------------------------------
5517: @cindex other system documentation, core words
5518: @cindex core words, other system documentation
5519:
5520: @table @i
5521: @item nonstandard words using @code{PAD}:
5522: @cindex @code{PAD} use by nonstandard words
5523: None.
5524:
5525: @item operator's terminal facilities available:
5526: @cindex operator's terminal facilities available
5527: After processing the command line, Gforth goes into interactive mode,
5528: and you can give commands to Gforth interactively. The actual facilities
5529: available depend on how you invoke Gforth.
5530:
5531: @item program data space available:
5532: @cindex program data space available
5533: @cindex data space available
5534: @code{UNUSED .} gives the remaining dictionary space. The total
5535: dictionary space can be specified with the @code{-m} switch
5536: (@pxref{Invoking Gforth}) when Gforth starts up.
5537:
5538: @item return stack space available:
5539: @cindex return stack space available
5540: You can compute the total return stack space in cells with
5541: @code{s" RETURN-STACK-CELLS" environment? drop .}. You can specify it at
5542: startup time with the @code{-r} switch (@pxref{Invoking Gforth}).
5543:
5544: @item stack space available:
5545: @cindex stack space available
5546: You can compute the total data stack space in cells with
5547: @code{s" STACK-CELLS" environment? drop .}. You can specify it at
5548: startup time with the @code{-d} switch (@pxref{Invoking Gforth}).
5549:
5550: @item system dictionary space required, in address units:
5551: @cindex system dictionary space required, in address units
5552: Type @code{here forthstart - .} after startup. At the time of this
5553: writing, this gives 80080 (bytes) on a 32-bit system.
5554: @end table
5555:
5556:
5557: @c =====================================================================
5558: @node The optional Block word set, The optional Double Number word set, The Core Words, ANS conformance
5559: @section The optional Block word set
5560: @c =====================================================================
5561: @cindex system documentation, block words
5562: @cindex block words, system documentation
5563:
5564: @menu
5565: * block-idef:: Implementation Defined Options
5566: * block-ambcond:: Ambiguous Conditions
5567: * block-other:: Other System Documentation
5568: @end menu
5569:
5570:
5571: @c ---------------------------------------------------------------------
5572: @node block-idef, block-ambcond, The optional Block word set, The optional Block word set
5573: @subsection Implementation Defined Options
5574: @c ---------------------------------------------------------------------
5575: @cindex implementation-defined options, block words
5576: @cindex block words, implementation-defined options
5577:
5578: @table @i
5579: @item the format for display by @code{LIST}:
5580: @cindex @code{LIST} display format
5581: First the screen number is displayed, then 16 lines of 64 characters,
5582: each line preceded by the line number.
5583:
5584: @item the length of a line affected by @code{\}:
5585: @cindex length of a line affected by @code{\}
5586: @cindex @code{\}, line length in blocks
5587: 64 characters.
5588: @end table
5589:
5590:
5591: @c ---------------------------------------------------------------------
5592: @node block-ambcond, block-other, block-idef, The optional Block word set
5593: @subsection Ambiguous conditions
5594: @c ---------------------------------------------------------------------
5595: @cindex block words, ambiguous conditions
5596: @cindex ambiguous conditions, block words
5597:
5598: @table @i
5599: @item correct block read was not possible:
5600: @cindex block read not possible
5601: Typically results in a @code{throw} of some OS-derived value (between
5602: -512 and -2048). If the blocks file was just not long enough, blanks are
5603: supplied for the missing portion.
5604:
5605: @item I/O exception in block transfer:
5606: @cindex I/O exception in block transfer
5607: @cindex block transfer, I/O exception
5608: Typically results in a @code{throw} of some OS-derived value (between
5609: -512 and -2048).
5610:
5611: @item invalid block number:
5612: @cindex invalid block number
5613: @cindex block number invalid
5614: @code{-35 throw} (Invalid block number)
5615:
5616: @item a program directly alters the contents of @code{BLK}:
5617: @cindex @code{BLK}, altering @code{BLK}
5618: The input stream is switched to that other block, at the same
5619: position. If the storing to @code{BLK} happens when interpreting
5620: non-block input, the system will get quite confused when the block ends.
5621:
5622: @item no current block buffer for @code{UPDATE}:
5623: @cindex @code{UPDATE}, no current block buffer
5624: @code{UPDATE} has no effect.
5625:
5626: @end table
5627:
5628: @c ---------------------------------------------------------------------
5629: @node block-other, , block-ambcond, The optional Block word set
5630: @subsection Other system documentation
5631: @c ---------------------------------------------------------------------
5632: @cindex other system documentation, block words
5633: @cindex block words, other system documentation
5634:
5635: @table @i
5636: @item any restrictions a multiprogramming system places on the use of buffer addresses:
5637: No restrictions (yet).
5638:
5639: @item the number of blocks available for source and data:
5640: depends on your disk space.
5641:
5642: @end table
5643:
5644:
5645: @c =====================================================================
5646: @node The optional Double Number word set, The optional Exception word set, The optional Block word set, ANS conformance
5647: @section The optional Double Number word set
5648: @c =====================================================================
5649: @cindex system documentation, double words
5650: @cindex double words, system documentation
5651:
5652: @menu
5653: * double-ambcond:: Ambiguous Conditions
5654: @end menu
5655:
5656:
5657: @c ---------------------------------------------------------------------
5658: @node double-ambcond, , The optional Double Number word set, The optional Double Number word set
5659: @subsection Ambiguous conditions
5660: @c ---------------------------------------------------------------------
5661: @cindex double words, ambiguous conditions
5662: @cindex ambiguous conditions, double words
5663:
5664: @table @i
5665: @item @var{d} outside of range of @var{n} in @code{D>S}:
5666: @cindex @code{D>S}, @var{d} out of range of @var{n}
5667: The least significant cell of @var{d} is produced.
5668:
5669: @end table
5670:
5671:
5672: @c =====================================================================
5673: @node The optional Exception word set, The optional Facility word set, The optional Double Number word set, ANS conformance
5674: @section The optional Exception word set
5675: @c =====================================================================
5676: @cindex system documentation, exception words
5677: @cindex exception words, system documentation
5678:
5679: @menu
5680: * exception-idef:: Implementation Defined Options
5681: @end menu
5682:
5683:
5684: @c ---------------------------------------------------------------------
5685: @node exception-idef, , The optional Exception word set, The optional Exception word set
5686: @subsection Implementation Defined Options
5687: @c ---------------------------------------------------------------------
5688: @cindex implementation-defined options, exception words
5689: @cindex exception words, implementation-defined options
5690:
5691: @table @i
5692: @item @code{THROW}-codes used in the system:
5693: @cindex @code{THROW}-codes used in the system
5694: The codes -256@minus{}-511 are used for reporting signals. The mapping
5695: from OS signal numbers to throw codes is -256@minus{}@var{signal}. The
5696: codes -512@minus{}-2047 are used for OS errors (for file and memory
5697: allocation operations). The mapping from OS error numbers to throw codes
5698: is -512@minus{}@code{errno}. One side effect of this mapping is that
5699: undefined OS errors produce a message with a strange number; e.g.,
5700: @code{-1000 THROW} results in @code{Unknown error 488} on my system.
5701: @end table
5702:
5703: @c =====================================================================
5704: @node The optional Facility word set, The optional File-Access word set, The optional Exception word set, ANS conformance
5705: @section The optional Facility word set
5706: @c =====================================================================
5707: @cindex system documentation, facility words
5708: @cindex facility words, system documentation
5709:
5710: @menu
5711: * facility-idef:: Implementation Defined Options
5712: * facility-ambcond:: Ambiguous Conditions
5713: @end menu
5714:
5715:
5716: @c ---------------------------------------------------------------------
5717: @node facility-idef, facility-ambcond, The optional Facility word set, The optional Facility word set
5718: @subsection Implementation Defined Options
5719: @c ---------------------------------------------------------------------
5720: @cindex implementation-defined options, facility words
5721: @cindex facility words, implementation-defined options
5722:
5723: @table @i
5724: @item encoding of keyboard events (@code{EKEY}):
5725: @cindex keyboard events, encoding in @code{EKEY}
5726: @cindex @code{EKEY}, encoding of keyboard events
5727: Not yet implemented.
5728:
5729: @item duration of a system clock tick:
5730: @cindex duration of a system clock tick
5731: @cindex clock tick duration
5732: System dependent. With respect to @code{MS}, the time is specified in
5733: microseconds. How well the OS and the hardware implement this, is
5734: another question.
5735:
5736: @item repeatability to be expected from the execution of @code{MS}:
5737: @cindex repeatability to be expected from the execution of @code{MS}
5738: @cindex @code{MS}, repeatability to be expected
5739: System dependent. On Unix, a lot depends on load. If the system is
5740: lightly loaded, and the delay is short enough that Gforth does not get
5741: swapped out, the performance should be acceptable. Under MS-DOS and
5742: other single-tasking systems, it should be good.
5743:
5744: @end table
5745:
5746:
5747: @c ---------------------------------------------------------------------
5748: @node facility-ambcond, , facility-idef, The optional Facility word set
5749: @subsection Ambiguous conditions
5750: @c ---------------------------------------------------------------------
5751: @cindex facility words, ambiguous conditions
5752: @cindex ambiguous conditions, facility words
5753:
5754: @table @i
5755: @item @code{AT-XY} can't be performed on user output device:
5756: @cindex @code{AT-XY} can't be performed on user output device
5757: Largely terminal dependent. No range checks are done on the arguments.
5758: No errors are reported. You may see some garbage appearing, you may see
5759: simply nothing happen.
5760:
5761: @end table
5762:
5763:
5764: @c =====================================================================
5765: @node The optional File-Access word set, The optional Floating-Point word set, The optional Facility word set, ANS conformance
5766: @section The optional File-Access word set
5767: @c =====================================================================
5768: @cindex system documentation, file words
5769: @cindex file words, system documentation
5770:
5771: @menu
5772: * file-idef:: Implementation Defined Options
5773: * file-ambcond:: Ambiguous Conditions
5774: @end menu
5775:
5776: @c ---------------------------------------------------------------------
5777: @node file-idef, file-ambcond, The optional File-Access word set, The optional File-Access word set
5778: @subsection Implementation Defined Options
5779: @c ---------------------------------------------------------------------
5780: @cindex implementation-defined options, file words
5781: @cindex file words, implementation-defined options
5782:
5783: @table @i
5784: @item file access methods used:
5785: @cindex file access methods used
5786: @code{R/O}, @code{R/W} and @code{BIN} work as you would
5787: expect. @code{W/O} translates into the C file opening mode @code{w} (or
5788: @code{wb}): The file is cleared, if it exists, and created, if it does
5789: not (with both @code{open-file} and @code{create-file}). Under Unix
5790: @code{create-file} creates a file with 666 permissions modified by your
5791: umask.
5792:
5793: @item file exceptions:
5794: @cindex file exceptions
5795: The file words do not raise exceptions (except, perhaps, memory access
5796: faults when you pass illegal addresses or file-ids).
5797:
5798: @item file line terminator:
5799: @cindex file line terminator
5800: System-dependent. Gforth uses C's newline character as line
5801: terminator. What the actual character code(s) of this are is
5802: system-dependent.
5803:
5804: @item file name format:
5805: @cindex file name format
5806: System dependent. Gforth just uses the file name format of your OS.
5807:
5808: @item information returned by @code{FILE-STATUS}:
5809: @cindex @code{FILE-STATUS}, returned information
5810: @code{FILE-STATUS} returns the most powerful file access mode allowed
5811: for the file: Either @code{R/O}, @code{W/O} or @code{R/W}. If the file
5812: cannot be accessed, @code{R/O BIN} is returned. @code{BIN} is applicable
5813: along with the returned mode.
5814:
5815: @item input file state after an exception when including source:
5816: @cindex exception when including source
5817: All files that are left via the exception are closed.
5818:
5819: @item @var{ior} values and meaning:
5820: @cindex @var{ior} values and meaning
5821: The @var{ior}s returned by the file and memory allocation words are
5822: intended as throw codes. They typically are in the range
5823: -512@minus{}-2047 of OS errors. The mapping from OS error numbers to
5824: @var{ior}s is -512@minus{}@var{errno}.
5825:
5826: @item maximum depth of file input nesting:
5827: @cindex maximum depth of file input nesting
5828: @cindex file input nesting, maximum depth
5829: limited by the amount of return stack, locals/TIB stack, and the number
5830: of open files available. This should not give you troubles.
5831:
5832: @item maximum size of input line:
5833: @cindex maximum size of input line
5834: @cindex input line size, maximum
5835: @code{/line}. Currently 255.
5836:
5837: @item methods of mapping block ranges to files:
5838: @cindex mapping block ranges to files
5839: @cindex files containing blocks
5840: @cindex blocks in files
5841: By default, blocks are accessed in the file @file{blocks.fb} in the
5842: current working directory. The file can be switched with @code{USE}.
5843:
5844: @item number of string buffers provided by @code{S"}:
5845: @cindex @code{S"}, number of string buffers
5846: 1
5847:
5848: @item size of string buffer used by @code{S"}:
5849: @cindex @code{S"}, size of string buffer
5850: @code{/line}. currently 255.
5851:
5852: @end table
5853:
5854: @c ---------------------------------------------------------------------
5855: @node file-ambcond, , file-idef, The optional File-Access word set
5856: @subsection Ambiguous conditions
5857: @c ---------------------------------------------------------------------
5858: @cindex file words, ambiguous conditions
5859: @cindex ambiguous conditions, file words
5860:
5861: @table @i
5862: @item attempting to position a file outside its boundaries:
5863: @cindex @code{REPOSITION-FILE}, outside the file's boundaries
5864: @code{REPOSITION-FILE} is performed as usual: Afterwards,
5865: @code{FILE-POSITION} returns the value given to @code{REPOSITION-FILE}.
5866:
5867: @item attempting to read from file positions not yet written:
5868: @cindex reading from file positions not yet written
5869: End-of-file, i.e., zero characters are read and no error is reported.
5870:
5871: @item @var{file-id} is invalid (@code{INCLUDE-FILE}):
5872: @cindex @code{INCLUDE-FILE}, @var{file-id} is invalid
5873: An appropriate exception may be thrown, but a memory fault or other
5874: problem is more probable.
5875:
5876: @item I/O exception reading or closing @var{file-id} (@code{INCLUDE-FILE}, @code{INCLUDED}):
5877: @cindex @code{INCLUDE-FILE}, I/O exception reading or closing @var{file-id}
5878: @cindex @code{INCLUDED}, I/O exception reading or closing @var{file-id}
5879: The @var{ior} produced by the operation, that discovered the problem, is
5880: thrown.
5881:
5882: @item named file cannot be opened (@code{INCLUDED}):
5883: @cindex @code{INCLUDED}, named file cannot be opened
5884: The @var{ior} produced by @code{open-file} is thrown.
5885:
5886: @item requesting an unmapped block number:
5887: @cindex unmapped block numbers
5888: There are no unmapped legal block numbers. On some operating systems,
5889: writing a block with a large number may overflow the file system and
5890: have an error message as consequence.
5891:
5892: @item using @code{source-id} when @code{blk} is non-zero:
5893: @cindex @code{SOURCE-ID}, behaviour when @code{BLK} is non-zero
5894: @code{source-id} performs its function. Typically it will give the id of
5895: the source which loaded the block. (Better ideas?)
5896:
5897: @end table
5898:
5899:
5900: @c =====================================================================
5901: @node The optional Floating-Point word set, The optional Locals word set, The optional File-Access word set, ANS conformance
5902: @section The optional Floating-Point word set
5903: @c =====================================================================
5904: @cindex system documentation, floating-point words
5905: @cindex floating-point words, system documentation
5906:
5907: @menu
5908: * floating-idef:: Implementation Defined Options
5909: * floating-ambcond:: Ambiguous Conditions
5910: @end menu
5911:
5912:
5913: @c ---------------------------------------------------------------------
5914: @node floating-idef, floating-ambcond, The optional Floating-Point word set, The optional Floating-Point word set
5915: @subsection Implementation Defined Options
5916: @c ---------------------------------------------------------------------
5917: @cindex implementation-defined options, floating-point words
5918: @cindex floating-point words, implementation-defined options
5919:
5920: @table @i
5921: @item format and range of floating point numbers:
5922: @cindex format and range of floating point numbers
5923: @cindex floating point numbers, format and range
5924: System-dependent; the @code{double} type of C.
5925:
5926: @item results of @code{REPRESENT} when @var{float} is out of range:
5927: @cindex @code{REPRESENT}, results when @var{float} is out of range
5928: System dependent; @code{REPRESENT} is implemented using the C library
5929: function @code{ecvt()} and inherits its behaviour in this respect.
5930:
5931: @item rounding or truncation of floating-point numbers:
5932: @cindex rounding of floating-point numbers
5933: @cindex truncation of floating-point numbers
5934: @cindex floating-point numbers, rounding or truncation
5935: System dependent; the rounding behaviour is inherited from the hosting C
5936: compiler. IEEE-FP-based (i.e., most) systems by default round to
5937: nearest, and break ties by rounding to even (i.e., such that the last
5938: bit of the mantissa is 0).
5939:
5940: @item size of floating-point stack:
5941: @cindex floating-point stack size
5942: @code{s" FLOATING-STACK" environment? drop .} gives the total size of
5943: the floating-point stack (in floats). You can specify this on startup
5944: with the command-line option @code{-f} (@pxref{Invoking Gforth}).
5945:
5946: @item width of floating-point stack:
5947: @cindex floating-point stack width
5948: @code{1 floats}.
5949:
5950: @end table
5951:
5952:
5953: @c ---------------------------------------------------------------------
5954: @node floating-ambcond, , floating-idef, The optional Floating-Point word set
5955: @subsection Ambiguous conditions
5956: @c ---------------------------------------------------------------------
5957: @cindex floating-point words, ambiguous conditions
5958: @cindex ambiguous conditions, floating-point words
5959:
5960: @table @i
5961: @item @code{df@@} or @code{df!} used with an address that is not double-float aligned:
5962: @cindex @code{df@@} or @code{df!} used with an address that is not double-float aligned
5963: System-dependent. Typically results in a @code{-23 THROW} like other
5964: alignment violations.
5965:
5966: @item @code{f@@} or @code{f!} used with an address that is not float aligned:
5967: @cindex @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: System-dependent. Typically results in a @code{-23 THROW} like other
5970: alignment violations.
5971:
5972: @item floating-point result out of range:
5973: @cindex floating-point result out of range
5974: System-dependent. Can result in a @code{-55 THROW} (Floating-point
5975: unidentified fault), or can produce a special value representing, e.g.,
5976: Infinity.
5977:
5978: @item @code{sf@@} or @code{sf!} used with an address that is not single-float aligned:
5979: @cindex @code{sf@@} or @code{sf!} used with an address that is not single-float aligned
5980: System-dependent. Typically results in an alignment fault like other
5981: alignment violations.
5982:
5983: @item @code{BASE} is not decimal (@code{REPRESENT}, @code{F.}, @code{FE.}, @code{FS.}):
5984: @cindex @code{BASE} is not decimal (@code{REPRESENT}, @code{F.}, @code{FE.}, @code{FS.})
5985: The floating-point number is converted into decimal nonetheless.
5986:
5987: @item Both arguments are equal to zero (@code{FATAN2}):
5988: @cindex @code{FATAN2}, both arguments are equal to zero
5989: System-dependent. @code{FATAN2} is implemented using the C library
5990: function @code{atan2()}.
5991:
5992: @item Using @code{FTAN} on an argument @var{r1} where cos(@var{r1}) is zero:
5993: @cindex @code{FTAN} on an argument @var{r1} where cos(@var{r1}) is zero
5994: System-dependent. Anyway, typically the cos of @var{r1} will not be zero
5995: because of small errors and the tan will be a very large (or very small)
5996: but finite number.
5997:
5998: @item @var{d} cannot be presented precisely as a float in @code{D>F}:
5999: @cindex @code{D>F}, @var{d} cannot be presented precisely as a float
6000: The result is rounded to the nearest float.
6001:
6002: @item dividing by zero:
6003: @cindex dividing by zero, floating-point
6004: @cindex floating-point dividing by zero
6005: @cindex floating-point unidentified fault, FP divide-by-zero
6006: @code{-55 throw} (Floating-point unidentified fault)
6007:
6008: @item exponent too big for conversion (@code{DF!}, @code{DF@@}, @code{SF!}, @code{SF@@}):
6009: @cindex exponent too big for conversion (@code{DF!}, @code{DF@@}, @code{SF!}, @code{SF@@})
6010: System dependent. On IEEE-FP based systems the number is converted into
6011: an infinity.
6012:
6013: @item @var{float}<1 (@code{FACOSH}):
6014: @cindex @code{FACOSH}, @var{float}<1
6015: @cindex floating-point unidentified fault, @code{FACOSH}
6016: @code{-55 throw} (Floating-point unidentified fault)
6017:
6018: @item @var{float}=<-1 (@code{FLNP1}):
6019: @cindex @code{FLNP1}, @var{float}=<-1
6020: @cindex floating-point unidentified fault, @code{FLNP1}
6021: @code{-55 throw} (Floating-point unidentified fault). On IEEE-FP systems
6022: negative infinity is typically produced for @var{float}=-1.
6023:
6024: @item @var{float}=<0 (@code{FLN}, @code{FLOG}):
6025: @cindex @code{FLN}, @var{float}=<0
6026: @cindex @code{FLOG}, @var{float}=<0
6027: @cindex floating-point unidentified fault, @code{FLN} or @code{FLOG}
6028: @code{-55 throw} (Floating-point unidentified fault). On IEEE-FP systems
6029: negative infinity is typically produced for @var{float}=0.
6030:
6031: @item @var{float}<0 (@code{FASINH}, @code{FSQRT}):
6032: @cindex @code{FASINH}, @var{float}<0
6033: @cindex @code{FSQRT}, @var{float}<0
6034: @cindex floating-point unidentified fault, @code{FASINH} or @code{FSQRT}
6035: @code{-55 throw} (Floating-point unidentified fault). @code{fasinh}
6036: produces values for these inputs on my Linux box (Bug in the C library?)
6037:
6038: @item |@var{float}|>1 (@code{FACOS}, @code{FASIN}, @code{FATANH}):
6039: @cindex @code{FACOS}, |@var{float}|>1
6040: @cindex @code{FASIN}, |@var{float}|>1
6041: @cindex @code{FATANH}, |@var{float}|>1
6042: @cindex floating-point unidentified fault, @code{FACOS}, @code{FASIN} or @code{FATANH}
6043: @code{-55 throw} (Floating-point unidentified fault).
6044:
6045: @item integer part of float cannot be represented by @var{d} in @code{F>D}:
6046: @cindex @code{F>D}, integer part of float cannot be represented by @var{d}
6047: @cindex floating-point unidentified fault, @code{F>D}
6048: @code{-55 throw} (Floating-point unidentified fault).
6049:
6050: @item string larger than pictured numeric output area (@code{f.}, @code{fe.}, @code{fs.}):
6051: @cindex string larger than pictured numeric output area (@code{f.}, @code{fe.}, @code{fs.})
6052: This does not happen.
6053: @end table
6054:
6055: @c =====================================================================
6056: @node The optional Locals word set, The optional Memory-Allocation word set, The optional Floating-Point word set, ANS conformance
6057: @section The optional Locals word set
6058: @c =====================================================================
6059: @cindex system documentation, locals words
6060: @cindex locals words, system documentation
6061:
6062: @menu
6063: * locals-idef:: Implementation Defined Options
6064: * locals-ambcond:: Ambiguous Conditions
6065: @end menu
6066:
6067:
6068: @c ---------------------------------------------------------------------
6069: @node locals-idef, locals-ambcond, The optional Locals word set, The optional Locals word set
6070: @subsection Implementation Defined Options
6071: @c ---------------------------------------------------------------------
6072: @cindex implementation-defined options, locals words
6073: @cindex locals words, implementation-defined options
6074:
6075: @table @i
6076: @item maximum number of locals in a definition:
6077: @cindex maximum number of locals in a definition
6078: @cindex locals, maximum number in a definition
6079: @code{s" #locals" environment? drop .}. Currently 15. This is a lower
6080: bound, e.g., on a 32-bit machine there can be 41 locals of up to 8
6081: characters. The number of locals in a definition is bounded by the size
6082: of locals-buffer, which contains the names of the locals.
6083:
6084: @end table
6085:
6086:
6087: @c ---------------------------------------------------------------------
6088: @node locals-ambcond, , locals-idef, The optional Locals word set
6089: @subsection Ambiguous conditions
6090: @c ---------------------------------------------------------------------
6091: @cindex locals words, ambiguous conditions
6092: @cindex ambiguous conditions, locals words
6093:
6094: @table @i
6095: @item executing a named local in interpretation state:
6096: @cindex local in interpretation state
6097: @cindex Interpreting a compile-only word, for a local
6098: Locals have no interpretation semantics. If you try to perform the
6099: interpretation semantics, you will get a @code{-14 throw} somewhere
6100: (Interpreting a compile-only word). If you perform the compilation
6101: semantics, the locals access will be compiled (irrespective of state).
6102:
6103: @item @var{name} not defined by @code{VALUE} or @code{(LOCAL)} (@code{TO}):
6104: @cindex name not defined by @code{VALUE} or @code{(LOCAL)} used by @code{TO}
6105: @cindex @code{TO} on non-@code{VALUE}s and non-locals
6106: @cindex Invalid name argument, @code{TO}
6107: @code{-32 throw} (Invalid name argument)
6108:
6109: @end table
6110:
6111:
6112: @c =====================================================================
6113: @node The optional Memory-Allocation word set, The optional Programming-Tools word set, The optional Locals word set, ANS conformance
6114: @section The optional Memory-Allocation word set
6115: @c =====================================================================
6116: @cindex system documentation, memory-allocation words
6117: @cindex memory-allocation words, system documentation
6118:
6119: @menu
6120: * memory-idef:: Implementation Defined Options
6121: @end menu
6122:
6123:
6124: @c ---------------------------------------------------------------------
6125: @node memory-idef, , The optional Memory-Allocation word set, The optional Memory-Allocation word set
6126: @subsection Implementation Defined Options
6127: @c ---------------------------------------------------------------------
6128: @cindex implementation-defined options, memory-allocation words
6129: @cindex memory-allocation words, implementation-defined options
6130:
6131: @table @i
6132: @item values and meaning of @var{ior}:
6133: @cindex @var{ior} values and meaning
6134: The @var{ior}s returned by the file and memory allocation words are
6135: intended as throw codes. They typically are in the range
6136: -512@minus{}-2047 of OS errors. The mapping from OS error numbers to
6137: @var{ior}s is -512@minus{}@var{errno}.
6138:
6139: @end table
6140:
6141: @c =====================================================================
6142: @node The optional Programming-Tools word set, The optional Search-Order word set, The optional Memory-Allocation word set, ANS conformance
6143: @section The optional Programming-Tools word set
6144: @c =====================================================================
6145: @cindex system documentation, programming-tools words
6146: @cindex programming-tools words, system documentation
6147:
6148: @menu
6149: * programming-idef:: Implementation Defined Options
6150: * programming-ambcond:: Ambiguous Conditions
6151: @end menu
6152:
6153:
6154: @c ---------------------------------------------------------------------
6155: @node programming-idef, programming-ambcond, The optional Programming-Tools word set, The optional Programming-Tools word set
6156: @subsection Implementation Defined Options
6157: @c ---------------------------------------------------------------------
6158: @cindex implementation-defined options, programming-tools words
6159: @cindex programming-tools words, implementation-defined options
6160:
6161: @table @i
6162: @item ending sequence for input following @code{;CODE} and @code{CODE}:
6163: @cindex @code{;CODE} ending sequence
6164: @cindex @code{CODE} ending sequence
6165: @code{END-CODE}
6166:
6167: @item manner of processing input following @code{;CODE} and @code{CODE}:
6168: @cindex @code{;CODE}, processing input
6169: @cindex @code{CODE}, processing input
6170: The @code{ASSEMBLER} vocabulary is pushed on the search order stack, and
6171: the input is processed by the text interpreter, (starting) in interpret
6172: state.
6173:
6174: @item search order capability for @code{EDITOR} and @code{ASSEMBLER}:
6175: @cindex @code{ASSEMBLER}, search order capability
6176: The ANS Forth search order word set.
6177:
6178: @item source and format of display by @code{SEE}:
6179: @cindex @code{SEE}, source and format of output
6180: The source for @code{see} is the intermediate code used by the inner
6181: interpreter. The current @code{see} tries to output Forth source code
6182: as well as possible.
6183:
6184: @end table
6185:
6186: @c ---------------------------------------------------------------------
6187: @node programming-ambcond, , programming-idef, The optional Programming-Tools word set
6188: @subsection Ambiguous conditions
6189: @c ---------------------------------------------------------------------
6190: @cindex programming-tools words, ambiguous conditions
6191: @cindex ambiguous conditions, programming-tools words
6192:
6193: @table @i
6194:
6195: @item deleting the compilation wordlist (@code{FORGET}):
6196: @cindex @code{FORGET}, deleting the compilation wordlist
6197: Not implemented (yet).
6198:
6199: @item fewer than @var{u}+1 items on the control flow stack (@code{CS-PICK}, @code{CS-ROLL}):
6200: @cindex @code{CS-PICK}, fewer than @var{u}+1 items on the control flow stack
6201: @cindex @code{CS-ROLL}, fewer than @var{u}+1 items on the control flow stack
6202: @cindex control-flow stack underflow
6203: This typically results in an @code{abort"} with a descriptive error
6204: message (may change into a @code{-22 throw} (Control structure mismatch)
6205: in the future). You may also get a memory access error. If you are
6206: unlucky, this ambiguous condition is not caught.
6207:
6208: @item @var{name} can't be found (@code{FORGET}):
6209: @cindex @code{FORGET}, @var{name} can't be found
6210: Not implemented (yet).
6211:
6212: @item @var{name} not defined via @code{CREATE}:
6213: @cindex @code{;CODE}, @var{name} not defined via @code{CREATE}
6214: @code{;CODE} behaves like @code{DOES>} in this respect, i.e., it changes
6215: the execution semantics of the last defined word no matter how it was
6216: defined.
6217:
6218: @item @code{POSTPONE} applied to @code{[IF]}:
6219: @cindex @code{POSTPONE} applied to @code{[IF]}
6220: @cindex @code{[IF]} and @code{POSTPONE}
6221: After defining @code{: X POSTPONE [IF] ; IMMEDIATE}. @code{X} is
6222: equivalent to @code{[IF]}.
6223:
6224: @item reaching the end of the input source before matching @code{[ELSE]} or @code{[THEN]}:
6225: @cindex @code{[IF]}, end of the input source before matching @code{[ELSE]} or @code{[THEN]}
6226: Continue in the same state of conditional compilation in the next outer
6227: input source. Currently there is no warning to the user about this.
6228:
6229: @item removing a needed definition (@code{FORGET}):
6230: @cindex @code{FORGET}, removing a needed definition
6231: Not implemented (yet).
6232:
6233: @end table
6234:
6235:
6236: @c =====================================================================
6237: @node The optional Search-Order word set, , The optional Programming-Tools word set, ANS conformance
6238: @section The optional Search-Order word set
6239: @c =====================================================================
6240: @cindex system documentation, search-order words
6241: @cindex search-order words, system documentation
6242:
6243: @menu
6244: * search-idef:: Implementation Defined Options
6245: * search-ambcond:: Ambiguous Conditions
6246: @end menu
6247:
6248:
6249: @c ---------------------------------------------------------------------
6250: @node search-idef, search-ambcond, The optional Search-Order word set, The optional Search-Order word set
6251: @subsection Implementation Defined Options
6252: @c ---------------------------------------------------------------------
6253: @cindex implementation-defined options, search-order words
6254: @cindex search-order words, implementation-defined options
6255:
6256: @table @i
6257: @item maximum number of word lists in search order:
6258: @cindex maximum number of word lists in search order
6259: @cindex search order, maximum depth
6260: @code{s" wordlists" environment? drop .}. Currently 16.
6261:
6262: @item minimum search order:
6263: @cindex minimum search order
6264: @cindex search order, minimum
6265: @code{root root}.
6266:
6267: @end table
6268:
6269: @c ---------------------------------------------------------------------
6270: @node search-ambcond, , search-idef, The optional Search-Order word set
6271: @subsection Ambiguous conditions
6272: @c ---------------------------------------------------------------------
6273: @cindex search-order words, ambiguous conditions
6274: @cindex ambiguous conditions, search-order words
6275:
6276: @table @i
6277: @item changing the compilation wordlist (during compilation):
6278: @cindex changing the compilation wordlist (during compilation)
6279: @cindex compilation wordlist, change before definition ends
6280: The word is entered into the wordlist that was the compilation wordlist
6281: at the start of the definition. Any changes to the name field (e.g.,
6282: @code{immediate}) or the code field (e.g., when executing @code{DOES>})
6283: are applied to the latest defined word (as reported by @code{last} or
6284: @code{lastxt}), if possible, irrespective of the compilation wordlist.
6285:
6286: @item search order empty (@code{previous}):
6287: @cindex @code{previous}, search order empty
6288: @cindex Vocstack empty, @code{previous}
6289: @code{abort" Vocstack empty"}.
6290:
6291: @item too many word lists in search order (@code{also}):
6292: @cindex @code{also}, too many word lists in search order
6293: @cindex Vocstack full, @code{also}
6294: @code{abort" Vocstack full"}.
6295:
6296: @end table
6297:
6298: @c ***************************************************************
6299: @node Model, Integrating Gforth, ANS conformance, Top
6300: @chapter Model
6301:
6302: This chapter has yet to be written. It will contain information, on
6303: which internal structures you can rely.
6304:
6305: @c ***************************************************************
6306: @node Integrating Gforth, Emacs and Gforth, Model, Top
6307: @chapter Integrating Gforth into C programs
6308:
6309: This is not yet implemented.
6310:
6311: Several people like to use Forth as scripting language for applications
6312: that are otherwise written in C, C++, or some other language.
6313:
6314: The Forth system ATLAST provides facilities for embedding it into
6315: applications; unfortunately it has several disadvantages: most
6316: importantly, it is not based on ANS Forth, and it is apparently dead
6317: (i.e., not developed further and not supported). The facilities
6318: provided by Gforth in this area are inspired by ATLASTs facilities, so
6319: making the switch should not be hard.
6320:
6321: We also tried to design the interface such that it can easily be
6322: implemented by other Forth systems, so that we may one day arrive at a
6323: standardized interface. Such a standard interface would allow you to
6324: replace the Forth system without having to rewrite C code.
6325:
6326: You embed the Gforth interpreter by linking with the library
6327: @code{libgforth.a} (give the compiler the option @code{-lgforth}). All
6328: global symbols in this library that belong to the interface, have the
6329: prefix @code{forth_}. (Global symbols that are used internally have the
6330: prefix @code{gforth_}).
6331:
6332: You can include the declarations of Forth types and the functions and
6333: variables of the interface with @code{#include <forth.h>}.
6334:
6335: Types.
6336:
6337: Variables.
6338:
6339: Data and FP Stack pointer. Area sizes.
6340:
6341: functions.
6342:
6343: forth_init(imagefile)
6344: forth_evaluate(string) exceptions?
6345: forth_goto(address) (or forth_execute(xt)?)
6346: forth_continue() (a corountining mechanism)
6347:
6348: Adding primitives.
6349:
6350: No checking.
6351:
6352: Signals?
6353:
6354: Accessing the Stacks
6355:
6356: @node Emacs and Gforth, Image Files, Integrating Gforth, Top
6357: @chapter Emacs and Gforth
6358: @cindex Emacs and Gforth
6359:
6360: @cindex @file{gforth.el}
6361: @cindex @file{forth.el}
6362: @cindex Rydqvist, Goran
6363: @cindex comment editing commands
6364: @cindex @code{\}, editing with Emacs
6365: @cindex debug tracer editing commands
6366: @cindex @code{~~}, removal with Emacs
6367: @cindex Forth mode in Emacs
6368: Gforth comes with @file{gforth.el}, an improved version of
6369: @file{forth.el} by Goran Rydqvist (included in the TILE package). The
6370: improvements are a better (but still not perfect) handling of
6371: indentation. I have also added comment paragraph filling (@kbd{M-q}),
6372: commenting (@kbd{C-x \}) and uncommenting (@kbd{C-u C-x \}) regions and
6373: removing debugging tracers (@kbd{C-x ~}, @pxref{Debugging}). I left the
6374: stuff I do not use alone, even though some of it only makes sense for
6375: TILE. To get a description of these features, enter Forth mode and type
6376: @kbd{C-h m}.
6377:
6378: @cindex source location of error or debugging output in Emacs
6379: @cindex error output, finding the source location in Emacs
6380: @cindex debugging output, finding the source location in Emacs
6381: In addition, Gforth supports Emacs quite well: The source code locations
6382: given in error messages, debugging output (from @code{~~}) and failed
6383: assertion messages are in the right format for Emacs' compilation mode
6384: (@pxref{Compilation, , Running Compilations under Emacs, emacs, Emacs
6385: Manual}) so the source location corresponding to an error or other
6386: message is only a few keystrokes away (@kbd{C-x `} for the next error,
6387: @kbd{C-c C-c} for the error under the cursor).
6388:
6389: @cindex @file{TAGS} file
6390: @cindex @file{etags.fs}
6391: @cindex viewing the source of a word in Emacs
6392: Also, if you @code{include} @file{etags.fs}, a new @file{TAGS} file
6393: (@pxref{Tags, , Tags Tables, emacs, Emacs Manual}) will be produced that
6394: contains the definitions of all words defined afterwards. You can then
6395: find the source for a word using @kbd{M-.}. Note that emacs can use
6396: several tags files at the same time (e.g., one for the Gforth sources
6397: and one for your program, @pxref{Select Tags Table,,Selecting a Tags
6398: Table,emacs, Emacs Manual}). The TAGS file for the preloaded words is
6399: @file{$(datadir)/gforth/$(VERSION)/TAGS} (e.g.,
6400: @file{/usr/local/share/gforth/0.2.0/TAGS}).
6401:
6402: @cindex @file{.emacs}
6403: To get all these benefits, add the following lines to your @file{.emacs}
6404: file:
6405:
6406: @example
6407: (autoload 'forth-mode "gforth.el")
6408: (setq auto-mode-alist (cons '("\\.fs\\'" . forth-mode) auto-mode-alist))
6409: @end example
6410:
6411: @node Image Files, Engine, Emacs and Gforth, Top
6412: @chapter Image Files
6413: @cindex image files
6414: @cindex @code{.fi} files
6415: @cindex precompiled Forth code
6416: @cindex dictionary in persistent form
6417: @cindex persistent form of dictionary
6418:
6419: An image file is a file containing an image of the Forth dictionary,
6420: i.e., compiled Forth code and data residing in the dictionary. By
6421: convention, we use the extension @code{.fi} for image files.
6422:
6423: @menu
6424: * Image File Background:: Why have image files?
6425: * Non-Relocatable Image Files:: don't always work.
6426: * Data-Relocatable Image Files:: are better.
6427: * Fully Relocatable Image Files:: better yet.
6428: * Stack and Dictionary Sizes:: Setting the default sizes for an image.
6429: * Running Image Files:: @code{gforth -i @var{file}} or @var{file}.
6430: * Modifying the Startup Sequence:: and turnkey applications.
6431: @end menu
6432:
6433: @node Image File Background, Non-Relocatable Image Files, Image Files, Image Files
6434: @section Image File Background
6435: @cindex image file background
6436:
6437: Our Forth system consists not only of primitives, but also of
6438: definitions written in Forth. Since the Forth compiler itself belongs to
6439: those definitions, it is not possible to start the system with the
6440: primitives and the Forth source alone. Therefore we provide the Forth
6441: code as an image file in nearly executable form. At the start of the
6442: system a C routine loads the image file into memory, optionally
6443: relocates the addresses, then sets up the memory (stacks etc.) according
6444: to information in the image file, and starts executing Forth code.
6445:
6446: The image file variants represent different compromises between the
6447: goals of making it easy to generate image files and making them
6448: portable.
6449:
6450: @cindex relocation at run-time
6451: Win32Forth 3.4 and Mitch Bradleys @code{cforth} use relocation at
6452: run-time. This avoids many of the complications discussed below (image
6453: files are data relocatable without further ado), but costs performance
6454: (one addition per memory access).
6455:
6456: @cindex relocation at load-time
6457: By contrast, our loader performs relocation at image load time. The
6458: loader also has to replace tokens standing for primitive calls with the
6459: appropriate code-field addresses (or code addresses in the case of
6460: direct threading).
6461:
6462: There are three kinds of image files, with different degrees of
6463: relocatability: non-relocatable, data-relocatable, and fully relocatable
6464: image files.
6465:
6466: @cindex image file loader
6467: @cindex relocating loader
6468: @cindex loader for image files
6469: These image file variants have several restrictions in common; they are
6470: caused by the design of the image file loader:
6471:
6472: @itemize @bullet
6473: @item
6474: There is only one segment; in particular, this means, that an image file
6475: cannot represent @code{ALLOCATE}d memory chunks (and pointers to
6476: them). And the contents of the stacks are not represented, either.
6477:
6478: @item
6479: The only kinds of relocation supported are: adding the same offset to
6480: all cells that represent data addresses; and replacing special tokens
6481: with code addresses or with pieces of machine code.
6482:
6483: If any complex computations involving addresses are performed, the
6484: results cannot be represented in the image file. Several applications that
6485: use such computations come to mind:
6486: @itemize @minus
6487: @item
6488: Hashing addresses (or data structures which contain addresses) for table
6489: lookup. If you use Gforth's @code{table}s or @code{wordlist}s for this
6490: purpose, you will have no problem, because the hash tables are
6491: recomputed automatically when the system is started. If you use your own
6492: hash tables, you will have to do something similar.
6493:
6494: @item
6495: There's a cute implementation of doubly-linked lists that uses
6496: @code{XOR}ed addresses. You could represent such lists as singly-linked
6497: in the image file, and restore the doubly-linked representation on
6498: startup.@footnote{In my opinion, though, you should think thrice before
6499: using a doubly-linked list (whatever implementation).}
6500:
6501: @item
6502: The code addresses of run-time routines like @code{docol:} cannot be
6503: represented in the image file (because their tokens would be replaced by
6504: machine code in direct threaded implementations). As a workaround,
6505: compute these addresses at run-time with @code{>code-address} from the
6506: executions tokens of appropriate words (see the definitions of
6507: @code{docol:} and friends in @file{kernel.fs}).
6508:
6509: @item
6510: On many architectures addresses are represented in machine code in some
6511: shifted or mangled form. You cannot put @code{CODE} words that contain
6512: absolute addresses in this form in a relocatable image file. Workarounds
6513: are representing the address in some relative form (e.g., relative to
6514: the CFA, which is present in some register), or loading the address from
6515: a place where it is stored in a non-mangled form.
6516: @end itemize
6517: @end itemize
6518:
6519: @node Non-Relocatable Image Files, Data-Relocatable Image Files, Image File Background, Image Files
6520: @section Non-Relocatable Image Files
6521: @cindex non-relocatable image files
6522: @cindex image files, non-relocatable
6523:
6524: These files are simple memory dumps of the dictionary. They are specific
6525: to the executable (i.e., @file{gforth} file) they were created
6526: with. What's worse, they are specific to the place on which the
6527: dictionary resided when the image was created. Now, there is no
6528: guarantee that the dictionary will reside at the same place the next
6529: time you start Gforth, so there's no guarantee that a non-relocatable
6530: image will work the next time (Gforth will complain instead of crashing,
6531: though).
6532:
6533: You can create a non-relocatable image file with
6534:
6535: doc-savesystem
6536:
6537: @node Data-Relocatable Image Files, Fully Relocatable Image Files, Non-Relocatable Image Files, Image Files
6538: @section Data-Relocatable Image Files
6539: @cindex data-relocatable image files
6540: @cindex image files, data-relocatable
6541:
6542: These files contain relocatable data addresses, but fixed code addresses
6543: (instead of tokens). They are specific to the executable (i.e.,
6544: @file{gforth} file) they were created with. For direct threading on some
6545: architectures (e.g., the i386), data-relocatable images do not work. You
6546: get a data-relocatable image, if you use @file{gforthmi} with a
6547: Gforth binary that is not doubly indirect threaded (@pxref{Fully
6548: Relocatable Image Files}).
6549:
6550: @node Fully Relocatable Image Files, Stack and Dictionary Sizes, Data-Relocatable Image Files, Image Files
6551: @section Fully Relocatable Image Files
6552: @cindex fully relocatable image files
6553: @cindex image files, fully relocatable
6554:
6555: @cindex @file{kern*.fi}, relocatability
6556: @cindex @file{gforth.fi}, relocatability
6557: These image files have relocatable data addresses, and tokens for code
6558: addresses. They can be used with different binaries (e.g., with and
6559: without debugging) on the same machine, and even across machines with
6560: the same data formats (byte order, cell size, floating point
6561: format). However, they are usually specific to the version of Gforth
6562: they were created with. The files @file{gforth.fi} and @file{kernl*.fi}
6563: are fully relocatable.
6564:
6565: There are two ways to create a fully relocatable image file:
6566:
6567: @menu
6568: * gforthmi:: The normal way
6569: * cross.fs:: The hard way
6570: @end menu
6571:
6572: @node gforthmi, cross.fs, Fully Relocatable Image Files, Fully Relocatable Image Files
6573: @subsection @file{gforthmi}
6574: @cindex @file{comp-i.fs}
6575: @cindex @file{gforthmi}
6576:
6577: You will usually use @file{gforthmi}. If you want to create an
6578: image @var{file} that contains everything you would load by invoking
6579: Gforth with @code{gforth @var{options}}, you simply say
6580: @example
6581: gforthmi @var{file} @var{options}
6582: @end example
6583:
6584: E.g., if you want to create an image @file{asm.fi} that has the file
6585: @file{asm.fs} loaded in addition to the usual stuff, you could do it
6586: like this:
6587:
6588: @example
6589: gforthmi asm.fi asm.fs
6590: @end example
6591:
6592: @file{gforthmi} works like this: It produces two non-relocatable
6593: images for different addresses and then compares them. Its output
6594: reflects this: first you see the output (if any) of the two Gforth
6595: invocations that produce the nonrelocatable image files, then you see
6596: the output of the comparing program: It displays the offset used for
6597: data addresses and the offset used for code addresses;
6598: moreover, for each cell that cannot be represented correctly in the
6599: image files, it displays a line like the following one:
6600:
6601: @example
6602: 78DC BFFFFA50 BFFFFA40
6603: @end example
6604:
6605: This means that at offset $78dc from @code{forthstart}, one input image
6606: contains $bffffa50, and the other contains $bffffa40. Since these cells
6607: cannot be represented correctly in the output image, you should examine
6608: these places in the dictionary and verify that these cells are dead
6609: (i.e., not read before they are written).
6610:
6611: @cindex @code{savesystem} during @file{gforthmi}
6612: @cindex @code{bye} during @file{gforthmi}
6613: @cindex doubly indirect threaded code
6614: @cindex environment variable @code{GFORTHD}
6615: @cindex @code{GFORTHD} environment variable
6616: @cindex @code{gforth-ditc}
6617: There are a few wrinkles: After processing the passed @var{options}, the
6618: words @code{savesystem} and @code{bye} must be visible. A special doubly
6619: indirect threaded version of the @file{gforth} executable is used for
6620: creating the nonrelocatable images; you can pass the exact filename of
6621: this executable through the environment variable @code{GFORTHD}
6622: (default: @file{gforth-ditc}); if you pass a version that is not doubly
6623: indirect threaded, you will not get a fully relocatable image, but a
6624: data-relocatable image (because there is no code address offset).
6625:
6626: @node cross.fs, , gforthmi, Fully Relocatable Image Files
6627: @subsection @file{cross.fs}
6628: @cindex @file{cross.fs}
6629: @cindex cross-compiler
6630: @cindex metacompiler
6631:
6632: You can also use @code{cross}, a batch compiler that accepts a Forth-like
6633: programming language. This @code{cross} language has to be documented
6634: yet.
6635:
6636: @cindex target compiler
6637: @code{cross} also allows you to create image files for machines with
6638: different data sizes and data formats than the one used for generating
6639: the image file. You can also use it to create an application image that
6640: does not contain a Forth compiler. These features are bought with
6641: restrictions and inconveniences in programming. E.g., addresses have to
6642: be stored in memory with special words (@code{A!}, @code{A,}, etc.) in
6643: order to make the code relocatable.
6644:
6645:
6646: @node Stack and Dictionary Sizes, Running Image Files, Fully Relocatable Image Files, Image Files
6647: @section Stack and Dictionary Sizes
6648: @cindex image file, stack and dictionary sizes
6649: @cindex dictionary size default
6650: @cindex stack size default
6651:
6652: If you invoke Gforth with a command line flag for the size
6653: (@pxref{Invoking Gforth}), the size you specify is stored in the
6654: dictionary. If you save the dictionary with @code{savesystem} or create
6655: an image with @file{gforthmi}, this size will become the default
6656: for the resulting image file. E.g., the following will create a
6657: fully relocatable version of gforth.fi with a 1MB dictionary:
6658:
6659: @example
6660: gforthmi gforth.fi -m 1M
6661: @end example
6662:
6663: In other words, if you want to set the default size for the dictionary
6664: and the stacks of an image, just invoke @file{gforthmi} with the
6665: appropriate options when creating the image.
6666:
6667: @cindex stack size, cache-friendly
6668: Note: For cache-friendly behaviour (i.e., good performance), you should
6669: make the sizes of the stacks modulo, say, 2K, somewhat different. E.g.,
6670: the default stack sizes are: data: 16k (mod 2k=0); fp: 15.5k (mod
6671: 2k=1.5k); return: 15k(mod 2k=1k); locals: 14.5k (mod 2k=0.5k).
6672:
6673: @node Running Image Files, Modifying the Startup Sequence, Stack and Dictionary Sizes, Image Files
6674: @section Running Image Files
6675: @cindex running image files
6676: @cindex invoking image files
6677: @cindex image file invocation
6678:
6679: @cindex -i, invoke image file
6680: @cindex --image file, invoke image file
6681: You can invoke Gforth with an image file @var{image} instead of the
6682: default @file{gforth.fi} with the @code{-i} flag (@pxref{Invoking Gforth}):
6683: @example
6684: gforth -i @var{image}
6685: @end example
6686:
6687: @cindex executable image file
6688: @cindex image files, executable
6689: If your operating system supports starting scripts with a line of the
6690: form @code{#! ...}, you just have to type the image file name to start
6691: Gforth with this image file (note that the file extension @code{.fi} is
6692: just a convention). I.e., to run Gforth with the image file @var{image},
6693: you can just type @var{image} instead of @code{gforth -i @var{image}}.
6694:
6695: doc-#!
6696:
6697: @node Modifying the Startup Sequence, , Running Image Files, Image Files
6698: @section Modifying the Startup Sequence
6699: @cindex startup sequence for image file
6700: @cindex image file initialization sequence
6701: @cindex initialization sequence of image file
6702:
6703: You can add your own initialization to the startup sequence through the
6704: deferred word
6705:
6706: doc-'cold
6707:
6708: @code{'cold} is invoked just before the image-specific command line
6709: processing (by default, loading files and evaluating (@code{-e}) strings)
6710: starts.
6711:
6712: A sequence for adding your initialization usually looks like this:
6713:
6714: @example
6715: :noname
6716: Defers 'cold \ do other initialization stuff (e.g., rehashing wordlists)
6717: ... \ your stuff
6718: ; IS 'cold
6719: @end example
6720:
6721: @cindex turnkey image files
6722: @cindex image files, turnkey applications
6723: You can make a turnkey image by letting @code{'cold} execute a word
6724: (your turnkey application) that never returns; instead, it exits Gforth
6725: via @code{bye} or @code{throw}.
6726:
6727: @cindex command-line arguments, access
6728: @cindex arguments on the command line, access
6729: You can access the (image-specific) command-line arguments through the
6730: variables @code{argc} and @code{argv}. @code{arg} provides conventient
6731: access to @code{argv}.
6732:
6733: doc-argc
6734: doc-argv
6735: doc-arg
6736:
6737: If @code{'cold} exits normally, Gforth processes the command-line
6738: arguments as files to be loaded and strings to be evaluated. Therefore,
6739: @code{'cold} should remove the arguments it has used in this case.
6740:
6741: @c ******************************************************************
1.13 pazsan 6742: @node Engine, Binding to System Library, Image Files, Top
1.1 anton 6743: @chapter Engine
6744: @cindex engine
6745: @cindex virtual machine
6746:
6747: Reading this section is not necessary for programming with Gforth. It
6748: may be helpful for finding your way in the Gforth sources.
6749:
6750: The ideas in this section have also been published in the papers
6751: @cite{ANS fig/GNU/??? Forth} (in German) by Bernd Paysan, presented at
6752: the Forth-Tagung '93 and @cite{A Portable Forth Engine} by M. Anton
6753: Ertl, presented at EuroForth '93; the latter is available at
6754: @*@url{http://www.complang.tuwien.ac.at/papers/ertl93.ps.Z}.
6755:
6756: @menu
6757: * Portability::
6758: * Threading::
6759: * Primitives::
6760: * Performance::
6761: @end menu
6762:
6763: @node Portability, Threading, Engine, Engine
6764: @section Portability
6765: @cindex engine portability
6766:
6767: One of the main goals of the effort is availability across a wide range
6768: of personal machines. fig-Forth, and, to a lesser extent, F83, achieved
6769: this goal by manually coding the engine in assembly language for several
6770: then-popular processors. This approach is very labor-intensive and the
6771: results are short-lived due to progress in computer architecture.
6772:
6773: @cindex C, using C for the engine
6774: Others have avoided this problem by coding in C, e.g., Mitch Bradley
6775: (cforth), Mikael Patel (TILE) and Dirk Zoller (pfe). This approach is
6776: particularly popular for UNIX-based Forths due to the large variety of
6777: architectures of UNIX machines. Unfortunately an implementation in C
6778: does not mix well with the goals of efficiency and with using
6779: traditional techniques: Indirect or direct threading cannot be expressed
6780: in C, and switch threading, the fastest technique available in C, is
6781: significantly slower. Another problem with C is that it is very
6782: cumbersome to express double integer arithmetic.
6783:
6784: @cindex GNU C for the engine
6785: @cindex long long
6786: Fortunately, there is a portable language that does not have these
6787: limitations: GNU C, the version of C processed by the GNU C compiler
6788: (@pxref{C Extensions, , Extensions to the C Language Family, gcc.info,
6789: GNU C Manual}). Its labels as values feature (@pxref{Labels as Values, ,
6790: Labels as Values, gcc.info, GNU C Manual}) makes direct and indirect
6791: threading possible, its @code{long long} type (@pxref{Long Long, ,
6792: Double-Word Integers, gcc.info, GNU C Manual}) corresponds to Forth's
6793: double numbers@footnote{Unfortunately, long longs are not implemented
6794: properly on all machines (e.g., on alpha-osf1, long longs are only 64
6795: bits, the same size as longs (and pointers), but they should be twice as
1.4 anton 6796: long according to @pxref{Long Long, , Double-Word Integers, gcc.info, GNU
1.1 anton 6797: C Manual}). So, we had to implement doubles in C after all. Still, on
6798: most machines we can use long longs and achieve better performance than
6799: with the emulation package.}. GNU C is available for free on all
6800: important (and many unimportant) UNIX machines, VMS, 80386s running
6801: MS-DOS, the Amiga, and the Atari ST, so a Forth written in GNU C can run
6802: on all these machines.
6803:
6804: Writing in a portable language has the reputation of producing code that
6805: is slower than assembly. For our Forth engine we repeatedly looked at
6806: the code produced by the compiler and eliminated most compiler-induced
6807: inefficiencies by appropriate changes in the source code.
6808:
6809: @cindex explicit register declarations
6810: @cindex --enable-force-reg, configuration flag
6811: @cindex -DFORCE_REG
6812: However, register allocation cannot be portably influenced by the
6813: programmer, leading to some inefficiencies on register-starved
6814: machines. We use explicit register declarations (@pxref{Explicit Reg
6815: Vars, , Variables in Specified Registers, gcc.info, GNU C Manual}) to
6816: improve the speed on some machines. They are turned on by using the
6817: configuration flag @code{--enable-force-reg} (@code{gcc} switch
6818: @code{-DFORCE_REG}). Unfortunately, this feature not only depends on the
6819: machine, but also on the compiler version: On some machines some
6820: compiler versions produce incorrect code when certain explicit register
6821: declarations are used. So by default @code{-DFORCE_REG} is not used.
6822:
6823: @node Threading, Primitives, Portability, Engine
6824: @section Threading
6825: @cindex inner interpreter implementation
6826: @cindex threaded code implementation
6827:
6828: @cindex labels as values
6829: GNU C's labels as values extension (available since @code{gcc-2.0},
6830: @pxref{Labels as Values, , Labels as Values, gcc.info, GNU C Manual})
6831: makes it possible to take the address of @var{label} by writing
6832: @code{&&@var{label}}. This address can then be used in a statement like
6833: @code{goto *@var{address}}. I.e., @code{goto *&&x} is the same as
6834: @code{goto x}.
6835:
6836: @cindex NEXT, indirect threaded
6837: @cindex indirect threaded inner interpreter
6838: @cindex inner interpreter, indirect threaded
6839: With this feature an indirect threaded NEXT looks like:
6840: @example
6841: cfa = *ip++;
6842: ca = *cfa;
6843: goto *ca;
6844: @end example
6845: @cindex instruction pointer
6846: For those unfamiliar with the names: @code{ip} is the Forth instruction
6847: pointer; the @code{cfa} (code-field address) corresponds to ANS Forths
6848: execution token and points to the code field of the next word to be
6849: executed; The @code{ca} (code address) fetched from there points to some
6850: executable code, e.g., a primitive or the colon definition handler
6851: @code{docol}.
6852:
6853: @cindex NEXT, direct threaded
6854: @cindex direct threaded inner interpreter
6855: @cindex inner interpreter, direct threaded
6856: Direct threading is even simpler:
6857: @example
6858: ca = *ip++;
6859: goto *ca;
6860: @end example
6861:
6862: Of course we have packaged the whole thing neatly in macros called
6863: @code{NEXT} and @code{NEXT1} (the part of NEXT after fetching the cfa).
6864:
6865: @menu
6866: * Scheduling::
6867: * Direct or Indirect Threaded?::
6868: * DOES>::
6869: @end menu
6870:
6871: @node Scheduling, Direct or Indirect Threaded?, Threading, Threading
6872: @subsection Scheduling
6873: @cindex inner interpreter optimization
6874:
6875: There is a little complication: Pipelined and superscalar processors,
6876: i.e., RISC and some modern CISC machines can process independent
6877: instructions while waiting for the results of an instruction. The
6878: compiler usually reorders (schedules) the instructions in a way that
6879: achieves good usage of these delay slots. However, on our first tries
6880: the compiler did not do well on scheduling primitives. E.g., for
6881: @code{+} implemented as
6882: @example
6883: n=sp[0]+sp[1];
6884: sp++;
6885: sp[0]=n;
6886: NEXT;
6887: @end example
6888: the NEXT comes strictly after the other code, i.e., there is nearly no
6889: scheduling. After a little thought the problem becomes clear: The
6890: compiler cannot know that sp and ip point to different addresses (and
6891: the version of @code{gcc} we used would not know it even if it was
6892: possible), so it could not move the load of the cfa above the store to
6893: the TOS. Indeed the pointers could be the same, if code on or very near
6894: the top of stack were executed. In the interest of speed we chose to
6895: forbid this probably unused ``feature'' and helped the compiler in
6896: scheduling: NEXT is divided into the loading part (@code{NEXT_P1}) and
6897: the goto part (@code{NEXT_P2}). @code{+} now looks like:
6898: @example
6899: n=sp[0]+sp[1];
6900: sp++;
6901: NEXT_P1;
6902: sp[0]=n;
6903: NEXT_P2;
6904: @end example
6905: This can be scheduled optimally by the compiler.
6906:
6907: This division can be turned off with the switch @code{-DCISC_NEXT}. This
6908: switch is on by default on machines that do not profit from scheduling
6909: (e.g., the 80386), in order to preserve registers.
6910:
6911: @node Direct or Indirect Threaded?, DOES>, Scheduling, Threading
6912: @subsection Direct or Indirect Threaded?
6913: @cindex threading, direct or indirect?
6914:
6915: @cindex -DDIRECT_THREADED
6916: Both! After packaging the nasty details in macro definitions we
6917: realized that we could switch between direct and indirect threading by
6918: simply setting a compilation flag (@code{-DDIRECT_THREADED}) and
6919: defining a few machine-specific macros for the direct-threading case.
6920: On the Forth level we also offer access words that hide the
6921: differences between the threading methods (@pxref{Threading Words}).
6922:
6923: Indirect threading is implemented completely machine-independently.
6924: Direct threading needs routines for creating jumps to the executable
6925: code (e.g. to docol or dodoes). These routines are inherently
6926: machine-dependent, but they do not amount to many source lines. I.e.,
6927: even porting direct threading to a new machine is a small effort.
6928:
6929: @cindex --enable-indirect-threaded, configuration flag
6930: @cindex --enable-direct-threaded, configuration flag
6931: The default threading method is machine-dependent. You can enforce a
6932: specific threading method when building Gforth with the configuration
6933: flag @code{--enable-direct-threaded} or
6934: @code{--enable-indirect-threaded}. Note that direct threading is not
6935: supported on all machines.
6936:
6937: @node DOES>, , Direct or Indirect Threaded?, Threading
6938: @subsection DOES>
6939: @cindex @code{DOES>} implementation
6940:
6941: @cindex dodoes routine
6942: @cindex DOES-code
6943: One of the most complex parts of a Forth engine is @code{dodoes}, i.e.,
6944: the chunk of code executed by every word defined by a
6945: @code{CREATE}...@code{DOES>} pair. The main problem here is: How to find
6946: the Forth code to be executed, i.e. the code after the
6947: @code{DOES>} (the DOES-code)? There are two solutions:
6948:
6949: In fig-Forth the code field points directly to the dodoes and the
6950: DOES-code address is stored in the cell after the code address (i.e. at
6951: @code{@var{cfa} cell+}). It may seem that this solution is illegal in
6952: the Forth-79 and all later standards, because in fig-Forth this address
6953: lies in the body (which is illegal in these standards). However, by
6954: making the code field larger for all words this solution becomes legal
6955: again. We use this approach for the indirect threaded version and for
6956: direct threading on some machines. Leaving a cell unused in most words
6957: is a bit wasteful, but on the machines we are targeting this is hardly a
6958: problem. The other reason for having a code field size of two cells is
6959: to avoid having different image files for direct and indirect threaded
6960: systems (direct threaded systems require two-cell code fields on many
6961: machines).
6962:
6963: @cindex DOES-handler
6964: The other approach is that the code field points or jumps to the cell
6965: after @code{DOES}. In this variant there is a jump to @code{dodoes} at
6966: this address (the DOES-handler). @code{dodoes} can then get the
6967: DOES-code address by computing the code address, i.e., the address of
6968: the jump to dodoes, and add the length of that jump field. A variant of
6969: this is to have a call to @code{dodoes} after the @code{DOES>}; then the
6970: return address (which can be found in the return register on RISCs) is
6971: the DOES-code address. Since the two cells available in the code field
6972: are used up by the jump to the code address in direct threading on many
6973: architectures, we use this approach for direct threading on these
6974: architectures. We did not want to add another cell to the code field.
6975:
6976: @node Primitives, Performance, Threading, Engine
6977: @section Primitives
6978: @cindex primitives, implementation
6979: @cindex virtual machine instructions, implementation
6980:
6981: @menu
6982: * Automatic Generation::
6983: * TOS Optimization::
6984: * Produced code::
6985: @end menu
6986:
6987: @node Automatic Generation, TOS Optimization, Primitives, Primitives
6988: @subsection Automatic Generation
6989: @cindex primitives, automatic generation
6990:
6991: @cindex @file{prims2x.fs}
6992: Since the primitives are implemented in a portable language, there is no
6993: longer any need to minimize the number of primitives. On the contrary,
6994: having many primitives has an advantage: speed. In order to reduce the
6995: number of errors in primitives and to make programming them easier, we
6996: provide a tool, the primitive generator (@file{prims2x.fs}), that
6997: automatically generates most (and sometimes all) of the C code for a
6998: primitive from the stack effect notation. The source for a primitive
6999: has the following form:
7000:
7001: @cindex primitive source format
7002: @format
7003: @var{Forth-name} @var{stack-effect} @var{category} [@var{pronounc.}]
7004: [@code{""}@var{glossary entry}@code{""}]
7005: @var{C code}
7006: [@code{:}
7007: @var{Forth code}]
7008: @end format
7009:
7010: The items in brackets are optional. The category and glossary fields
7011: are there for generating the documentation, the Forth code is there
7012: for manual implementations on machines without GNU C. E.g., the source
7013: for the primitive @code{+} is:
7014: @example
7015: + n1 n2 -- n core plus
7016: n = n1+n2;
7017: @end example
7018:
7019: This looks like a specification, but in fact @code{n = n1+n2} is C
7020: code. Our primitive generation tool extracts a lot of information from
7021: the stack effect notations@footnote{We use a one-stack notation, even
7022: though we have separate data and floating-point stacks; The separate
7023: notation can be generated easily from the unified notation.}: The number
7024: of items popped from and pushed on the stack, their type, and by what
7025: name they are referred to in the C code. It then generates a C code
7026: prelude and postlude for each primitive. The final C code for @code{+}
7027: looks like this:
7028:
7029: @example
7030: I_plus: /* + ( n1 n2 -- n ) */ /* label, stack effect */
7031: /* */ /* documentation */
7032: @{
7033: DEF_CA /* definition of variable ca (indirect threading) */
7034: Cell n1; /* definitions of variables */
7035: Cell n2;
7036: Cell n;
7037: n1 = (Cell) sp[1]; /* input */
7038: n2 = (Cell) TOS;
7039: sp += 1; /* stack adjustment */
7040: NAME("+") /* debugging output (with -DDEBUG) */
7041: @{
7042: n = n1+n2; /* C code taken from the source */
7043: @}
7044: NEXT_P1; /* NEXT part 1 */
7045: TOS = (Cell)n; /* output */
7046: NEXT_P2; /* NEXT part 2 */
7047: @}
7048: @end example
7049:
7050: This looks long and inefficient, but the GNU C compiler optimizes quite
7051: well and produces optimal code for @code{+} on, e.g., the R3000 and the
7052: HP RISC machines: Defining the @code{n}s does not produce any code, and
7053: using them as intermediate storage also adds no cost.
7054:
7055: There are also other optimizations, that are not illustrated by this
7056: example: Assignments between simple variables are usually for free (copy
7057: propagation). If one of the stack items is not used by the primitive
7058: (e.g. in @code{drop}), the compiler eliminates the load from the stack
7059: (dead code elimination). On the other hand, there are some things that
7060: the compiler does not do, therefore they are performed by
7061: @file{prims2x.fs}: The compiler does not optimize code away that stores
7062: a stack item to the place where it just came from (e.g., @code{over}).
7063:
7064: While programming a primitive is usually easy, there are a few cases
7065: where the programmer has to take the actions of the generator into
7066: account, most notably @code{?dup}, but also words that do not (always)
7067: fall through to NEXT.
7068:
7069: @node TOS Optimization, Produced code, Automatic Generation, Primitives
7070: @subsection TOS Optimization
7071: @cindex TOS optimization for primitives
7072: @cindex primitives, keeping the TOS in a register
7073:
7074: An important optimization for stack machine emulators, e.g., Forth
7075: engines, is keeping one or more of the top stack items in
7076: registers. If a word has the stack effect @var{in1}...@var{inx} @code{--}
7077: @var{out1}...@var{outy}, keeping the top @var{n} items in registers
7078: @itemize @bullet
7079: @item
7080: is better than keeping @var{n-1} items, if @var{x>=n} and @var{y>=n},
7081: due to fewer loads from and stores to the stack.
7082: @item is slower than keeping @var{n-1} items, if @var{x<>y} and @var{x<n} and
7083: @var{y<n}, due to additional moves between registers.
7084: @end itemize
7085:
7086: @cindex -DUSE_TOS
7087: @cindex -DUSE_NO_TOS
7088: In particular, keeping one item in a register is never a disadvantage,
7089: if there are enough registers. Keeping two items in registers is a
7090: disadvantage for frequent words like @code{?branch}, constants,
7091: variables, literals and @code{i}. Therefore our generator only produces
7092: code that keeps zero or one items in registers. The generated C code
7093: covers both cases; the selection between these alternatives is made at
7094: C-compile time using the switch @code{-DUSE_TOS}. @code{TOS} in the C
7095: code for @code{+} is just a simple variable name in the one-item case,
7096: otherwise it is a macro that expands into @code{sp[0]}. Note that the
7097: GNU C compiler tries to keep simple variables like @code{TOS} in
7098: registers, and it usually succeeds, if there are enough registers.
7099:
7100: @cindex -DUSE_FTOS
7101: @cindex -DUSE_NO_FTOS
7102: The primitive generator performs the TOS optimization for the
7103: floating-point stack, too (@code{-DUSE_FTOS}). For floating-point
7104: operations the benefit of this optimization is even larger:
7105: floating-point operations take quite long on most processors, but can be
7106: performed in parallel with other operations as long as their results are
7107: not used. If the FP-TOS is kept in a register, this works. If
7108: it is kept on the stack, i.e., in memory, the store into memory has to
7109: wait for the result of the floating-point operation, lengthening the
7110: execution time of the primitive considerably.
7111:
7112: The TOS optimization makes the automatic generation of primitives a
7113: bit more complicated. Just replacing all occurrences of @code{sp[0]} by
7114: @code{TOS} is not sufficient. There are some special cases to
7115: consider:
7116: @itemize @bullet
7117: @item In the case of @code{dup ( w -- w w )} the generator must not
7118: eliminate the store to the original location of the item on the stack,
7119: if the TOS optimization is turned on.
7120: @item Primitives with stack effects of the form @code{--}
7121: @var{out1}...@var{outy} must store the TOS to the stack at the start.
7122: Likewise, primitives with the stack effect @var{in1}...@var{inx} @code{--}
7123: must load the TOS from the stack at the end. But for the null stack
7124: effect @code{--} no stores or loads should be generated.
7125: @end itemize
7126:
7127: @node Produced code, , TOS Optimization, Primitives
7128: @subsection Produced code
7129: @cindex primitives, assembly code listing
7130:
7131: @cindex @file{engine.s}
7132: To see what assembly code is produced for the primitives on your machine
7133: with your compiler and your flag settings, type @code{make engine.s} and
7134: look at the resulting file @file{engine.s}.
7135:
7136: @node Performance, , Primitives, Engine
7137: @section Performance
7138: @cindex performance of some Forth interpreters
7139: @cindex engine performance
7140: @cindex benchmarking Forth systems
7141: @cindex Gforth performance
7142:
7143: On RISCs the Gforth engine is very close to optimal; i.e., it is usually
7144: impossible to write a significantly faster engine.
7145:
7146: On register-starved machines like the 386 architecture processors
7147: improvements are possible, because @code{gcc} does not utilize the
7148: registers as well as a human, even with explicit register declarations;
7149: e.g., Bernd Beuster wrote a Forth system fragment in assembly language
7150: and hand-tuned it for the 486; this system is 1.19 times faster on the
7151: Sieve benchmark on a 486DX2/66 than Gforth compiled with
7152: @code{gcc-2.6.3} with @code{-DFORCE_REG}.
7153:
7154: @cindex Win32Forth performance
7155: @cindex NT Forth performance
7156: @cindex eforth performance
7157: @cindex ThisForth performance
7158: @cindex PFE performance
7159: @cindex TILE performance
7160: However, this potential advantage of assembly language implementations
7161: is not necessarily realized in complete Forth systems: We compared
7162: Gforth (direct threaded, compiled with @code{gcc-2.6.3} and
7163: @code{-DFORCE_REG}) with Win32Forth 1.2093, LMI's NT Forth (Beta, May
7164: 1994) and Eforth (with and without peephole (aka pinhole) optimization
7165: of the threaded code); all these systems were written in assembly
7166: language. We also compared Gforth with three systems written in C:
7167: PFE-0.9.14 (compiled with @code{gcc-2.6.3} with the default
7168: configuration for Linux: @code{-O2 -fomit-frame-pointer -DUSE_REGS
7169: -DUNROLL_NEXT}), ThisForth Beta (compiled with gcc-2.6.3 -O3
7170: -fomit-frame-pointer; ThisForth employs peephole optimization of the
7171: threaded code) and TILE (compiled with @code{make opt}). We benchmarked
7172: Gforth, PFE, ThisForth and TILE on a 486DX2/66 under Linux. Kenneth
7173: O'Heskin kindly provided the results for Win32Forth and NT Forth on a
7174: 486DX2/66 with similar memory performance under Windows NT. Marcel
7175: Hendrix ported Eforth to Linux, then extended it to run the benchmarks,
7176: added the peephole optimizer, ran the benchmarks and reported the
7177: results.
7178:
7179: We used four small benchmarks: the ubiquitous Sieve; bubble-sorting and
7180: matrix multiplication come from the Stanford integer benchmarks and have
7181: been translated into Forth by Martin Fraeman; we used the versions
7182: included in the TILE Forth package, but with bigger data set sizes; and
7183: a recursive Fibonacci number computation for benchmarking calling
7184: performance. The following table shows the time taken for the benchmarks
7185: scaled by the time taken by Gforth (in other words, it shows the speedup
7186: factor that Gforth achieved over the other systems).
7187:
7188: @example
7189: relative Win32- NT eforth This-
7190: time Gforth Forth Forth eforth +opt PFE Forth TILE
7191: sieve 1.00 1.39 1.14 1.39 0.85 1.58 3.18 8.58
7192: bubble 1.00 1.31 1.41 1.48 0.88 1.50 3.88
7193: matmul 1.00 1.47 1.35 1.46 0.74 1.58 4.09
7194: fib 1.00 1.52 1.34 1.22 0.86 1.74 2.99 4.30
7195: @end example
7196:
7197: You may find the good performance of Gforth compared with the systems
7198: written in assembly language quite surprising. One important reason for
7199: the disappointing performance of these systems is probably that they are
7200: not written optimally for the 486 (e.g., they use the @code{lods}
7201: instruction). In addition, Win32Forth uses a comfortable, but costly
7202: method for relocating the Forth image: like @code{cforth}, it computes
7203: the actual addresses at run time, resulting in two address computations
7204: per NEXT (@pxref{Image File Background}).
7205:
7206: Only Eforth with the peephole optimizer performs comparable to
7207: Gforth. The speedups achieved with peephole optimization of threaded
7208: code are quite remarkable. Adding a peephole optimizer to Gforth should
7209: cause similar speedups.
7210:
7211: The speedup of Gforth over PFE, ThisForth and TILE can be easily
7212: explained with the self-imposed restriction of the latter systems to
7213: standard C, which makes efficient threading impossible (however, the
1.4 anton 7214: measured implementation of PFE uses a GNU C extension: @pxref{Global Reg
1.1 anton 7215: Vars, , Defining Global Register Variables, gcc.info, GNU C Manual}).
7216: Moreover, current C compilers have a hard time optimizing other aspects
7217: of the ThisForth and the TILE source.
7218:
7219: Note that the performance of Gforth on 386 architecture processors
7220: varies widely with the version of @code{gcc} used. E.g., @code{gcc-2.5.8}
7221: failed to allocate any of the virtual machine registers into real
7222: machine registers by itself and would not work correctly with explicit
7223: register declarations, giving a 1.3 times slower engine (on a 486DX2/66
7224: running the Sieve) than the one measured above.
7225:
7226: Note also that there have been several releases of Win32Forth since the
7227: release presented here, so the results presented here may have little
7228: predictive value for the performance of Win32Forth today.
7229:
7230: @cindex @file{Benchres}
7231: In @cite{Translating Forth to Efficient C} by M. Anton Ertl and Martin
7232: Maierhofer (presented at EuroForth '95), an indirect threaded version of
7233: Gforth is compared with Win32Forth, NT Forth, PFE, and ThisForth; that
7234: version of Gforth is 2%@minus{}8% slower on a 486 than the direct
7235: threaded version used here. The paper available at
7236: @*@url{http://www.complang.tuwien.ac.at/papers/ertl&maierhofer95.ps.gz};
7237: it also contains numbers for some native code systems. You can find a
7238: newer version of these measurements at
7239: @url{http://www.complang.tuwien.ac.at/forth/performance.html}. You can
7240: find numbers for Gforth on various machines in @file{Benchres}.
7241:
1.13 pazsan 7242: @node Binding to System Library, Cross Compiler, Engine, Top
1.14 pazsan 7243: @chapter Binding to System Library
1.13 pazsan 7244:
7245: @node Cross Compiler, Bugs, Binding to System Library, Top
1.14 pazsan 7246: @chapter Cross Compiler
1.13 pazsan 7247:
7248: Cross Compiler
7249:
7250: @menu
7251: * Using the Cross Compiler::
7252: * How the Cross Compiler Works::
7253: @end menu
7254:
7255: @node Using the Cross Compiler, , How the Cross Compiler Works, Cross Compiler
1.14 pazsan 7256: @section Using the Cross Compiler
1.13 pazsan 7257:
7258: @node How the Cross Compiler Works, Using the Cross Compiler, , Cross Compiler
1.14 pazsan 7259: @section How the Cross Compiler Works
1.13 pazsan 7260:
7261: @node Bugs, Origin, Cross Compiler, Top
1.1 anton 7262: @chapter Bugs
7263: @cindex bug reporting
7264:
7265: Known bugs are described in the file BUGS in the Gforth distribution.
7266:
7267: If you find a bug, please send a bug report to
7268: @email{bug-gforth@@gnu.ai.mit.edu}. A bug report should
7269: describe the Gforth version used (it is announced at the start of an
7270: interactive Gforth session), the machine and operating system (on Unix
7271: systems you can use @code{uname -a} to produce this information), the
7272: installation options (send the @file{config.status} file), and a
7273: complete list of changes you (or your installer) have made to the Gforth
7274: sources (if any); it should contain a program (or a sequence of keyboard
7275: commands) that reproduces the bug and a description of what you think
7276: constitutes the buggy behaviour.
7277:
7278: For a thorough guide on reporting bugs read @ref{Bug Reporting, , How
7279: to Report Bugs, gcc.info, GNU C Manual}.
7280:
7281:
7282: @node Origin, Word Index, Bugs, Top
7283: @chapter Authors and Ancestors of Gforth
7284:
7285: @section Authors and Contributors
7286: @cindex authors of Gforth
7287: @cindex contributors to Gforth
7288:
7289: The Gforth project was started in mid-1992 by Bernd Paysan and Anton
7290: Ertl. The third major author was Jens Wilke. Lennart Benschop (who was
7291: one of Gforth's first users, in mid-1993) and Stuart Ramsden inspired us
7292: with their continuous feedback. Lennart Benshop contributed
7293: @file{glosgen.fs}, while Stuart Ramsden has been working on automatic
7294: support for calling C libraries. Helpful comments also came from Paul
7295: Kleinrubatscher, Christian Pirker, Dirk Zoller, Marcel Hendrix, John
1.12 anton 7296: Wavrik, Barrie Stott, Marc de Groot, and Jorge Acerada. Since the
7297: release of Gforth-0.2.1 there were also helpful comments from many
7298: others; thank you all, sorry for not listing you here (but digging
7299: through my mailbox to extract your names is on my to-do list).
1.1 anton 7300:
7301: Gforth also owes a lot to the authors of the tools we used (GCC, CVS,
7302: and autoconf, among others), and to the creators of the Internet: Gforth
7303: was developed across the Internet, and its authors have not met
7304: physically yet.
7305:
7306: @section Pedigree
7307: @cindex Pedigree of Gforth
7308:
7309: Gforth descends from BigForth (1993) and fig-Forth. Gforth and PFE (by
7310: Dirk Zoller) will cross-fertilize each other. Of course, a significant
7311: part of the design of Gforth was prescribed by ANS Forth.
7312:
7313: Bernd Paysan wrote BigForth, a descendent from TurboForth, an unreleased
7314: 32 bit native code version of VolksForth for the Atari ST, written
7315: mostly by Dietrich Weineck.
7316:
7317: VolksForth descends from F83. It was written by Klaus Schleisiek, Bernd
7318: Pennemann, Georg Rehfeld and Dietrich Weineck for the C64 (called
7319: UltraForth there) in the mid-80s and ported to the Atari ST in 1986.
7320:
7321: Henry Laxen and Mike Perry wrote F83 as a model implementation of the
7322: Forth-83 standard. !! Pedigree? When?
7323:
7324: A team led by Bill Ragsdale implemented fig-Forth on many processors in
7325: 1979. Robert Selzer and Bill Ragsdale developed the original
7326: implementation of fig-Forth for the 6502 based on microForth.
7327:
7328: The principal architect of microForth was Dean Sanderson. microForth was
7329: FORTH, Inc.'s first off-the-shelf product. It was developed in 1976 for
7330: the 1802, and subsequently implemented on the 8080, the 6800 and the
7331: Z80.
7332:
7333: All earlier Forth systems were custom-made, usually by Charles Moore,
7334: who discovered (as he puts it) Forth during the late 60s. The first full
7335: Forth existed in 1971.
7336:
7337: A part of the information in this section comes from @cite{The Evolution
7338: of Forth} by Elizabeth D. Rather, Donald R. Colburn and Charles
7339: H. Moore, presented at the HOPL-II conference and preprinted in SIGPLAN
7340: Notices 28(3), 1993. You can find more historical and genealogical
7341: information about Forth there.
7342:
7343: @node Word Index, Concept Index, Origin, Top
7344: @unnumbered Word Index
7345:
7346: This index is as incomplete as the manual. Each word is listed with
7347: stack effect and wordset.
7348:
7349: @printindex fn
7350:
7351: @node Concept Index, , Word Index, Top
7352: @unnumbered Concept and Word Index
7353:
7354: This index is as incomplete as the manual. Not all entries listed are
7355: present verbatim in the text. Only the names are listed for the words
7356: here.
7357:
7358: @printindex cp
7359:
7360: @contents
7361: @bye
7362:
FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>