Annotation of gforth/doc/gforth.ds, revision 1.22
1.1 anton 1: \input texinfo @c -*-texinfo-*-
2: @comment The source is gforth.ds, from which gforth.texi is generated
1.21 crook 3: @comment TODO: nac29jan99 - a list of things to add in the next edit:
4: @comment 1. x-ref all ambiguous or implementation-defined features
5: @comment 2. refer to all environment strings
6: @comment 3. gloss and info in blocks section
7: @comment 4. move file and blocks to common sub-section?
8: @comment 5. command-line editing, command completion etc.
9: @comment 6. document more of the words in require.fs
10: @comment 7. document the include files process (Describe the list,
11: @comment including its scope)
12: @comment 8. Describe the use of Auser Avariable etc.
13: @comment 9. cross-compiler
14: @comment 10.words in miscellaneous section need a home.
15: @comment 11.Move structures and oof into their own chapters.
16: @comment 12.search for TODO for other minor works
1.1 anton 17: @comment %**start of header (This is for running Texinfo on a region.)
18: @setfilename gforth.info
19: @settitle Gforth Manual
20: @dircategory GNU programming tools
21: @direntry
22: * Gforth: (gforth). A fast interpreter for the Forth language.
23: @end direntry
24: @comment @setchapternewpage odd
1.12 anton 25: @macro progstyle {}
26: Programming style note:
1.3 anton 27: @end macro
1.1 anton 28: @comment %**end of header (This is for running Texinfo on a region.)
29:
1.10 anton 30: @include version.texi
31:
1.1 anton 32: @ifinfo
1.11 anton 33: This file documents Gforth @value{VERSION}
1.1 anton 34:
1.13 pazsan 35: Copyright @copyright{} 1995-1998 Free Software Foundation, Inc.
1.1 anton 36:
37: Permission is granted to make and distribute verbatim copies of
38: this manual provided the copyright notice and this permission notice
39: are preserved on all copies.
40:
41: @ignore
42: Permission is granted to process this file through TeX and print the
43: results, provided the printed document carries a copying permission
44: notice identical to this one except for the removal of this paragraph
45: (this paragraph not being relevant to the printed manual).
46:
47: @end ignore
48: Permission is granted to copy and distribute modified versions of this
49: manual under the conditions for verbatim copying, provided also that the
50: sections entitled "Distribution" and "General Public License" are
51: included exactly as in the original, and provided that the entire
52: resulting derived work is distributed under the terms of a permission
53: notice identical to this one.
54:
55: Permission is granted to copy and distribute translations of this manual
56: into another language, under the above conditions for modified versions,
57: except that the sections entitled "Distribution" and "General Public
58: License" may be included in a translation approved by the author instead
59: of in the original English.
60: @end ifinfo
61:
62: @finalout
63: @titlepage
64: @sp 10
65: @center @titlefont{Gforth Manual}
66: @sp 2
1.11 anton 67: @center for version @value{VERSION}
1.1 anton 68: @sp 2
69: @center Anton Ertl
1.6 pazsan 70: @center Bernd Paysan
1.5 anton 71: @center Jens Wilke
1.1 anton 72: @sp 3
1.21 crook 73: @center This manual is permanently under construction and was last updated on 18-Jan-1999
1.1 anton 74:
75: @comment The following two commands start the copyright page.
76: @page
77: @vskip 0pt plus 1filll
1.13 pazsan 78: Copyright @copyright{} 1995--1998 Free Software Foundation, Inc.
1.1 anton 79:
80: @comment !! Published by ... or You can get a copy of this manual ...
81:
82: Permission is granted to make and distribute verbatim copies of
83: this manual provided the copyright notice and this permission notice
84: are preserved on all copies.
85:
86: Permission is granted to copy and distribute modified versions of this
87: manual under the conditions for verbatim copying, provided also that the
88: sections entitled "Distribution" and "General Public License" are
89: included exactly as in the original, and provided that the entire
90: resulting derived work is distributed under the terms of a permission
91: notice identical to this one.
92:
93: Permission is granted to copy and distribute translations of this manual
94: into another language, under the above conditions for modified versions,
95: except that the sections entitled "Distribution" and "General Public
96: License" may be included in a translation approved by the author instead
97: of in the original English.
98: @end titlepage
99:
100:
101: @node Top, License, (dir), (dir)
102: @ifinfo
103: Gforth is a free implementation of ANS Forth available on many
1.11 anton 104: personal machines. This manual corresponds to version @value{VERSION}.
1.1 anton 105: @end ifinfo
106:
107: @menu
1.21 crook 108: * License:: The GPL
109: * Introduction:: An introduction to ANS Forth
1.1 anton 110: * Goals:: About the Gforth Project
1.21 crook 111: * Invoking Gforth:: Starting (and exiting) Gforth
1.1 anton 112: * Words:: Forth words available in Gforth
113: * Tools:: Programming tools
114: * ANS conformance:: Implementation-defined options etc.
115: * Model:: The abstract machine of Gforth
116: * Integrating Gforth:: Forth as scripting language for applications
117: * Emacs and Gforth:: The Gforth Mode
118: * Image Files:: @code{.fi} files contain compiled code
119: * Engine:: The inner interpreter and the primitives
1.13 pazsan 120: * Cross Compiler:: The Cross Compiler
1.1 anton 121: * Bugs:: How to report them
122: * Origin:: Authors and ancestors of Gforth
1.21 crook 123: * Forth-related information:: Books and places to look on the WWW
1.1 anton 124: * Word Index:: An item for each Forth word
125: * Concept Index:: A menu covering many topics
1.12 anton 126:
127: --- The Detailed Node Listing ---
128:
1.21 crook 129: Goals
130:
131: * Gforth Extensions Sinful?::
132:
1.12 anton 133: Forth Words
134:
135: * Notation::
1.21 crook 136: * Comments::
137: * Boolean Flags::
1.12 anton 138: * Arithmetic::
139: * Stack Manipulation::
140: * Memory::
141: * Control Structures::
142: * Locals::
143: * Defining Words::
1.21 crook 144: * The Text Interpreter::
1.12 anton 145: * Structures::
146: * Object-oriented Forth::
147: * Tokens for Words::
1.21 crook 148: * Word Lists::
149: * Environmental Queries::
1.12 anton 150: * Files::
151: * Including Files::
152: * Blocks::
153: * Other I/O::
154: * Programming Tools::
155: * Assembler and Code Words::
156: * Threading Words::
1.21 crook 157: * Passing Commands to the OS::
158: * Miscellaneous Words::
1.12 anton 159:
160: Arithmetic
161:
162: * Single precision::
163: * Bitwise operations::
1.21 crook 164: * Double precision:: Double-cell integer arithmetic
165: * Numeric comparison::
1.12 anton 166: * Mixed precision:: operations with single and double-cell integers
167: * Floating Point::
168:
169: Stack Manipulation
170:
171: * Data stack::
172: * Floating point stack::
173: * Return stack::
174: * Locals stack::
175: * Stack pointer manipulation::
176:
177: Memory
178:
179: * Memory Access::
180: * Address arithmetic::
181: * Memory Blocks::
182:
183: Control Structures
184:
185: * Selection::
186: * Simple Loops::
187: * Counted Loops::
188: * Arbitrary control structures::
189: * Calls and returns::
190: * Exception Handling::
191:
192: Locals
193:
194: * Gforth locals::
195: * ANS Forth locals::
196:
197: Gforth locals
198:
199: * Where are locals visible by name?::
200: * How long do locals live?::
201: * Programming Style::
202: * Implementation::
203:
204: Defining Words
205:
206: * Simple Defining Words::
207: * Colon Definitions::
208: * User-defined Defining Words::
209: * Supplying names::
210: * Interpretation and Compilation Semantics::
211:
1.21 crook 212: The Text Interpreter
213:
214: * Number Conversion::
215: * Interpret/Compile states::
216: * Literals::
217: * Interpreter Directives::
218:
1.12 anton 219: Structures
220:
221: * Why explicit structure support?::
222: * Structure Usage::
223: * Structure Naming Convention::
224: * Structure Implementation::
225: * Structure Glossary::
226:
227: Object-oriented Forth
228:
229: * Objects::
230: * OOF::
231: * Mini-OOF::
232:
233: Objects
234:
235: * Properties of the Objects model::
236: * Why object-oriented programming?::
237: * Object-Oriented Terminology::
238: * Basic Objects Usage::
239: * The class Object::
240: * Creating objects::
241: * Object-Oriented Programming Style::
242: * Class Binding::
243: * Method conveniences::
244: * Classes and Scoping::
245: * Object Interfaces::
246: * Objects Implementation::
247: * Comparison with other object models::
248: * Objects Glossary::
249:
250: OOF
251:
252: * Properties of the OOF model::
253: * Basic OOF Usage::
254: * The base class object::
255: * Class Declaration::
256: * Class Implementation::
257:
1.21 crook 258: Word Lists
259:
260: * Why use word lists?::
261: * Word list examples::
262:
1.12 anton 263: Including Files
264:
265: * Words for Including::
266: * Search Path::
1.21 crook 267: * Forth Search Paths::
1.12 anton 268: * General Search Paths::
269:
1.21 crook 270: Other I/O
271:
272: * Simple numeric output::
273: * Formatted numeric output::
274: * String Formats::
275: * Displaying characters and strings::
276: * Input::
277:
1.12 anton 278: Programming Tools
279:
280: * Debugging:: Simple and quick.
281: * Assertions:: Making your programs self-checking.
282: * Singlestep Debugger:: Executing your program word by word.
283:
284: Tools
285:
286: * ANS Report:: Report the words used, sorted by wordset.
287:
288: ANS conformance
289:
290: * The Core Words::
291: * The optional Block word set::
292: * The optional Double Number word set::
293: * The optional Exception word set::
294: * The optional Facility word set::
295: * The optional File-Access word set::
296: * The optional Floating-Point word set::
297: * The optional Locals word set::
298: * The optional Memory-Allocation word set::
299: * The optional Programming-Tools word set::
300: * The optional Search-Order word set::
301:
302: The Core Words
303:
304: * core-idef:: Implementation Defined Options
305: * core-ambcond:: Ambiguous Conditions
306: * core-other:: Other System Documentation
307:
308: The optional Block word set
309:
310: * block-idef:: Implementation Defined Options
311: * block-ambcond:: Ambiguous Conditions
312: * block-other:: Other System Documentation
313:
314: The optional Double Number word set
315:
316: * double-ambcond:: Ambiguous Conditions
317:
318: The optional Exception word set
319:
320: * exception-idef:: Implementation Defined Options
321:
322: The optional Facility word set
323:
324: * facility-idef:: Implementation Defined Options
325: * facility-ambcond:: Ambiguous Conditions
326:
327: The optional File-Access word set
328:
329: * file-idef:: Implementation Defined Options
330: * file-ambcond:: Ambiguous Conditions
331:
332: The optional Floating-Point word set
333:
334: * floating-idef:: Implementation Defined Options
335: * floating-ambcond:: Ambiguous Conditions
336:
337: The optional Locals word set
338:
339: * locals-idef:: Implementation Defined Options
340: * locals-ambcond:: Ambiguous Conditions
341:
342: The optional Memory-Allocation word set
343:
344: * memory-idef:: Implementation Defined Options
345:
346: The optional Programming-Tools word set
347:
348: * programming-idef:: Implementation Defined Options
349: * programming-ambcond:: Ambiguous Conditions
350:
351: The optional Search-Order word set
352:
353: * search-idef:: Implementation Defined Options
354: * search-ambcond:: Ambiguous Conditions
355:
356: Image Files
357:
358: * Image File Background:: Why have image files?
359: * Non-Relocatable Image Files:: don't always work.
360: * Data-Relocatable Image Files:: are better.
361: * Fully Relocatable Image Files:: better yet.
362: * Stack and Dictionary Sizes:: Setting the default sizes for an image.
363: * Running Image Files:: @code{gforth -i @var{file}} or @var{file}.
364: * Modifying the Startup Sequence:: and turnkey applications.
365:
366: Fully Relocatable Image Files
367:
1.21 crook 368: * gforthmi:: The normal way
1.12 anton 369: * cross.fs:: The hard way
370:
371: Engine
372:
373: * Portability::
374: * Threading::
375: * Primitives::
376: * Performance::
377:
378: Threading
379:
380: * Scheduling::
381: * Direct or Indirect Threaded?::
382: * DOES>::
383:
384: Primitives
385:
386: * Automatic Generation::
387: * TOS Optimization::
388: * Produced code::
1.13 pazsan 389:
390: System Libraries
391:
392: * Binding to System Library::
393:
394: Cross Compiler
395:
396: * Using the Cross Compiler::
397: * How the Cross Compiler Works::
398:
1.21 crook 399: Forth-related information
400:
401: * Internet resources::
402: * Books::
403: * The Forth Interest Group::
404: * Conferences::
405:
406:
407:
1.1 anton 408: @end menu
409:
1.21 crook 410: @node License, Introduction, Top, Top
1.1 anton 411: @unnumbered GNU GENERAL PUBLIC LICENSE
412: @center Version 2, June 1991
413:
414: @display
415: Copyright @copyright{} 1989, 1991 Free Software Foundation, Inc.
416: 675 Mass Ave, Cambridge, MA 02139, USA
417:
418: Everyone is permitted to copy and distribute verbatim copies
419: of this license document, but changing it is not allowed.
420: @end display
421:
422: @unnumberedsec Preamble
423:
424: The licenses for most software are designed to take away your
425: freedom to share and change it. By contrast, the GNU General Public
426: License is intended to guarantee your freedom to share and change free
427: software---to make sure the software is free for all its users. This
428: General Public License applies to most of the Free Software
429: Foundation's software and to any other program whose authors commit to
430: using it. (Some other Free Software Foundation software is covered by
431: the GNU Library General Public License instead.) You can apply it to
432: your programs, too.
433:
434: When we speak of free software, we are referring to freedom, not
435: price. Our General Public Licenses are designed to make sure that you
436: have the freedom to distribute copies of free software (and charge for
437: this service if you wish), that you receive source code or can get it
438: if you want it, that you can change the software or use pieces of it
439: in new free programs; and that you know you can do these things.
440:
441: To protect your rights, we need to make restrictions that forbid
442: anyone to deny you these rights or to ask you to surrender the rights.
443: These restrictions translate to certain responsibilities for you if you
444: distribute copies of the software, or if you modify it.
445:
446: For example, if you distribute copies of such a program, whether
447: gratis or for a fee, you must give the recipients all the rights that
448: you have. You must make sure that they, too, receive or can get the
449: source code. And you must show them these terms so they know their
450: rights.
451:
452: We protect your rights with two steps: (1) copyright the software, and
453: (2) offer you this license which gives you legal permission to copy,
454: distribute and/or modify the software.
455:
456: Also, for each author's protection and ours, we want to make certain
457: that everyone understands that there is no warranty for this free
458: software. If the software is modified by someone else and passed on, we
459: want its recipients to know that what they have is not the original, so
460: that any problems introduced by others will not reflect on the original
461: authors' reputations.
462:
463: Finally, any free program is threatened constantly by software
464: patents. We wish to avoid the danger that redistributors of a free
465: program will individually obtain patent licenses, in effect making the
466: program proprietary. To prevent this, we have made it clear that any
467: patent must be licensed for everyone's free use or not licensed at all.
468:
469: The precise terms and conditions for copying, distribution and
470: modification follow.
471:
472: @iftex
473: @unnumberedsec TERMS AND CONDITIONS FOR COPYING, DISTRIBUTION AND MODIFICATION
474: @end iftex
475: @ifinfo
476: @center TERMS AND CONDITIONS FOR COPYING, DISTRIBUTION AND MODIFICATION
477: @end ifinfo
478:
479: @enumerate 0
480: @item
481: This License applies to any program or other work which contains
482: a notice placed by the copyright holder saying it may be distributed
483: under the terms of this General Public License. The ``Program'', below,
484: refers to any such program or work, and a ``work based on the Program''
485: means either the Program or any derivative work under copyright law:
486: that is to say, a work containing the Program or a portion of it,
487: either verbatim or with modifications and/or translated into another
488: language. (Hereinafter, translation is included without limitation in
489: the term ``modification''.) Each licensee is addressed as ``you''.
490:
491: Activities other than copying, distribution and modification are not
492: covered by this License; they are outside its scope. The act of
493: running the Program is not restricted, and the output from the Program
494: is covered only if its contents constitute a work based on the
495: Program (independent of having been made by running the Program).
496: Whether that is true depends on what the Program does.
497:
498: @item
499: You may copy and distribute verbatim copies of the Program's
500: source code as you receive it, in any medium, provided that you
501: conspicuously and appropriately publish on each copy an appropriate
502: copyright notice and disclaimer of warranty; keep intact all the
503: notices that refer to this License and to the absence of any warranty;
504: and give any other recipients of the Program a copy of this License
505: along with the Program.
506:
507: You may charge a fee for the physical act of transferring a copy, and
508: you may at your option offer warranty protection in exchange for a fee.
509:
510: @item
511: You may modify your copy or copies of the Program or any portion
512: of it, thus forming a work based on the Program, and copy and
513: distribute such modifications or work under the terms of Section 1
514: above, provided that you also meet all of these conditions:
515:
516: @enumerate a
517: @item
518: You must cause the modified files to carry prominent notices
519: stating that you changed the files and the date of any change.
520:
521: @item
522: You must cause any work that you distribute or publish, that in
523: whole or in part contains or is derived from the Program or any
524: part thereof, to be licensed as a whole at no charge to all third
525: parties under the terms of this License.
526:
527: @item
528: If the modified program normally reads commands interactively
529: when run, you must cause it, when started running for such
530: interactive use in the most ordinary way, to print or display an
531: announcement including an appropriate copyright notice and a
532: notice that there is no warranty (or else, saying that you provide
533: a warranty) and that users may redistribute the program under
534: these conditions, and telling the user how to view a copy of this
535: License. (Exception: if the Program itself is interactive but
536: does not normally print such an announcement, your work based on
537: the Program is not required to print an announcement.)
538: @end enumerate
539:
540: These requirements apply to the modified work as a whole. If
541: identifiable sections of that work are not derived from the Program,
542: and can be reasonably considered independent and separate works in
543: themselves, then this License, and its terms, do not apply to those
544: sections when you distribute them as separate works. But when you
545: distribute the same sections as part of a whole which is a work based
546: on the Program, the distribution of the whole must be on the terms of
547: this License, whose permissions for other licensees extend to the
548: entire whole, and thus to each and every part regardless of who wrote it.
549:
550: Thus, it is not the intent of this section to claim rights or contest
551: your rights to work written entirely by you; rather, the intent is to
552: exercise the right to control the distribution of derivative or
553: collective works based on the Program.
554:
555: In addition, mere aggregation of another work not based on the Program
556: with the Program (or with a work based on the Program) on a volume of
557: a storage or distribution medium does not bring the other work under
558: the scope of this License.
559:
560: @item
561: You may copy and distribute the Program (or a work based on it,
562: under Section 2) in object code or executable form under the terms of
563: Sections 1 and 2 above provided that you also do one of the following:
564:
565: @enumerate a
566: @item
567: Accompany it with the complete corresponding machine-readable
568: source code, which must be distributed under the terms of Sections
569: 1 and 2 above on a medium customarily used for software interchange; or,
570:
571: @item
572: Accompany it with a written offer, valid for at least three
573: years, to give any third party, for a charge no more than your
574: cost of physically performing source distribution, a complete
575: machine-readable copy of the corresponding source code, to be
576: distributed under the terms of Sections 1 and 2 above on a medium
577: customarily used for software interchange; or,
578:
579: @item
580: Accompany it with the information you received as to the offer
581: to distribute corresponding source code. (This alternative is
582: allowed only for noncommercial distribution and only if you
583: received the program in object code or executable form with such
584: an offer, in accord with Subsection b above.)
585: @end enumerate
586:
587: The source code for a work means the preferred form of the work for
588: making modifications to it. For an executable work, complete source
589: code means all the source code for all modules it contains, plus any
590: associated interface definition files, plus the scripts used to
591: control compilation and installation of the executable. However, as a
592: special exception, the source code distributed need not include
593: anything that is normally distributed (in either source or binary
594: form) with the major components (compiler, kernel, and so on) of the
595: operating system on which the executable runs, unless that component
596: itself accompanies the executable.
597:
598: If distribution of executable or object code is made by offering
599: access to copy from a designated place, then offering equivalent
600: access to copy the source code from the same place counts as
601: distribution of the source code, even though third parties are not
602: compelled to copy the source along with the object code.
603:
604: @item
605: You may not copy, modify, sublicense, or distribute the Program
606: except as expressly provided under this License. Any attempt
607: otherwise to copy, modify, sublicense or distribute the Program is
608: void, and will automatically terminate your rights under this License.
609: However, parties who have received copies, or rights, from you under
610: this License will not have their licenses terminated so long as such
611: parties remain in full compliance.
612:
613: @item
614: You are not required to accept this License, since you have not
615: signed it. However, nothing else grants you permission to modify or
616: distribute the Program or its derivative works. These actions are
617: prohibited by law if you do not accept this License. Therefore, by
618: modifying or distributing the Program (or any work based on the
619: Program), you indicate your acceptance of this License to do so, and
620: all its terms and conditions for copying, distributing or modifying
621: the Program or works based on it.
622:
623: @item
624: Each time you redistribute the Program (or any work based on the
625: Program), the recipient automatically receives a license from the
626: original licensor to copy, distribute or modify the Program subject to
627: these terms and conditions. You may not impose any further
628: restrictions on the recipients' exercise of the rights granted herein.
629: You are not responsible for enforcing compliance by third parties to
630: this License.
631:
632: @item
633: If, as a consequence of a court judgment or allegation of patent
634: infringement or for any other reason (not limited to patent issues),
635: conditions are imposed on you (whether by court order, agreement or
636: otherwise) that contradict the conditions of this License, they do not
637: excuse you from the conditions of this License. If you cannot
638: distribute so as to satisfy simultaneously your obligations under this
639: License and any other pertinent obligations, then as a consequence you
640: may not distribute the Program at all. For example, if a patent
641: license would not permit royalty-free redistribution of the Program by
642: all those who receive copies directly or indirectly through you, then
643: the only way you could satisfy both it and this License would be to
644: refrain entirely from distribution of the Program.
645:
646: If any portion of this section is held invalid or unenforceable under
647: any particular circumstance, the balance of the section is intended to
648: apply and the section as a whole is intended to apply in other
649: circumstances.
650:
651: It is not the purpose of this section to induce you to infringe any
652: patents or other property right claims or to contest validity of any
653: such claims; this section has the sole purpose of protecting the
654: integrity of the free software distribution system, which is
655: implemented by public license practices. Many people have made
656: generous contributions to the wide range of software distributed
657: through that system in reliance on consistent application of that
658: system; it is up to the author/donor to decide if he or she is willing
659: to distribute software through any other system and a licensee cannot
660: impose that choice.
661:
662: This section is intended to make thoroughly clear what is believed to
663: be a consequence of the rest of this License.
664:
665: @item
666: If the distribution and/or use of the Program is restricted in
667: certain countries either by patents or by copyrighted interfaces, the
668: original copyright holder who places the Program under this License
669: may add an explicit geographical distribution limitation excluding
670: those countries, so that distribution is permitted only in or among
671: countries not thus excluded. In such case, this License incorporates
672: the limitation as if written in the body of this License.
673:
674: @item
675: The Free Software Foundation may publish revised and/or new versions
676: of the General Public License from time to time. Such new versions will
677: be similar in spirit to the present version, but may differ in detail to
678: address new problems or concerns.
679:
680: Each version is given a distinguishing version number. If the Program
681: specifies a version number of this License which applies to it and ``any
682: later version'', you have the option of following the terms and conditions
683: either of that version or of any later version published by the Free
684: Software Foundation. If the Program does not specify a version number of
685: this License, you may choose any version ever published by the Free Software
686: Foundation.
687:
688: @item
689: If you wish to incorporate parts of the Program into other free
690: programs whose distribution conditions are different, write to the author
691: to ask for permission. For software which is copyrighted by the Free
692: Software Foundation, write to the Free Software Foundation; we sometimes
693: make exceptions for this. Our decision will be guided by the two goals
694: of preserving the free status of all derivatives of our free software and
695: of promoting the sharing and reuse of software generally.
696:
697: @iftex
698: @heading NO WARRANTY
699: @end iftex
700: @ifinfo
701: @center NO WARRANTY
702: @end ifinfo
703:
704: @item
705: BECAUSE THE PROGRAM IS LICENSED FREE OF CHARGE, THERE IS NO WARRANTY
706: FOR THE PROGRAM, TO THE EXTENT PERMITTED BY APPLICABLE LAW. EXCEPT WHEN
707: OTHERWISE STATED IN WRITING THE COPYRIGHT HOLDERS AND/OR OTHER PARTIES
708: PROVIDE THE PROGRAM ``AS IS'' WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED
709: OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
710: MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. THE ENTIRE RISK AS
711: TO THE QUALITY AND PERFORMANCE OF THE PROGRAM IS WITH YOU. SHOULD THE
712: PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF ALL NECESSARY SERVICING,
713: REPAIR OR CORRECTION.
714:
715: @item
716: IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING
717: WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MAY MODIFY AND/OR
718: REDISTRIBUTE THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES,
719: INCLUDING ANY GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING
720: OUT OF THE USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED
721: TO LOSS OF DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY
722: YOU OR THIRD PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER
723: PROGRAMS), EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE
724: POSSIBILITY OF SUCH DAMAGES.
725: @end enumerate
726:
727: @iftex
728: @heading END OF TERMS AND CONDITIONS
729: @end iftex
730: @ifinfo
731: @center END OF TERMS AND CONDITIONS
732: @end ifinfo
733:
734: @page
735: @unnumberedsec How to Apply These Terms to Your New Programs
736:
737: If you develop a new program, and you want it to be of the greatest
738: possible use to the public, the best way to achieve this is to make it
739: free software which everyone can redistribute and change under these terms.
740:
741: To do so, attach the following notices to the program. It is safest
742: to attach them to the start of each source file to most effectively
743: convey the exclusion of warranty; and each file should have at least
744: the ``copyright'' line and a pointer to where the full notice is found.
745:
746: @smallexample
747: @var{one line to give the program's name and a brief idea of what it does.}
748: Copyright (C) 19@var{yy} @var{name of author}
749:
750: This program is free software; you can redistribute it and/or modify
751: it under the terms of the GNU General Public License as published by
752: the Free Software Foundation; either version 2 of the License, or
753: (at your option) any later version.
754:
755: This program is distributed in the hope that it will be useful,
756: but WITHOUT ANY WARRANTY; without even the implied warranty of
757: MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
758: GNU General Public License for more details.
759:
760: You should have received a copy of the GNU General Public License
761: along with this program; if not, write to the Free Software
762: Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
763: @end smallexample
764:
765: Also add information on how to contact you by electronic and paper mail.
766:
767: If the program is interactive, make it output a short notice like this
768: when it starts in an interactive mode:
769:
770: @smallexample
771: Gnomovision version 69, Copyright (C) 19@var{yy} @var{name of author}
772: Gnomovision comes with ABSOLUTELY NO WARRANTY; for details
773: type `show w'.
774: This is free software, and you are welcome to redistribute it
775: under certain conditions; type `show c' for details.
776: @end smallexample
777:
778: The hypothetical commands @samp{show w} and @samp{show c} should show
779: the appropriate parts of the General Public License. Of course, the
780: commands you use may be called something other than @samp{show w} and
781: @samp{show c}; they could even be mouse-clicks or menu items---whatever
782: suits your program.
783:
784: You should also get your employer (if you work as a programmer) or your
785: school, if any, to sign a ``copyright disclaimer'' for the program, if
786: necessary. Here is a sample; alter the names:
787:
788: @smallexample
789: Yoyodyne, Inc., hereby disclaims all copyright interest in the program
790: `Gnomovision' (which makes passes at compilers) written by James Hacker.
791:
792: @var{signature of Ty Coon}, 1 April 1989
793: Ty Coon, President of Vice
794: @end smallexample
795:
796: This General Public License does not permit incorporating your program into
797: proprietary programs. If your program is a subroutine library, you may
798: consider it more useful to permit linking proprietary applications with the
799: library. If this is what you want to do, use the GNU Library General
800: Public License instead of this License.
801:
802: @iftex
803: @unnumbered Preface
804: @cindex Preface
1.21 crook 805: This manual documents Gforth. Some introductory material is provided for
806: readers who are unfamiliar with Forth or who are migrating to Gforth
807: from other Forth compilers. However, this manual is primarily a
808: reference manual.
1.1 anton 809: @end iftex
810:
1.21 crook 811: @c ----------------------------------------------------------
812: @node Introduction, Goals, License, Top
813: @comment node-name, next, previous, up
814: @chapter An Introduction to ANS Forth
815: @cindex Forth - an introduction
816:
817: The primary purpose of this manual is to document Gforth. However, since
818: Forth is not a widely-known language and there is a lack of up-to-date
819: teaching material, it seems worthwhile to provide some introductory
820: material. @xref{Forth-related information} for other sources of Forth-related
821: information.
822:
823: The examples in this section should work on any ANS Standard Forth, the
824: output shown was produced using Gforth. In each example, I have tried to
825: reproduce the exact output that Gforth produces. If you try out the
826: examples (and you should), what you should type is shown @kbd{like this}
827: and Gforth's response is shown @code{like this}. The single exception is
828: that, where the example shows @kbd{<return>} it means that you should
829: press the "carriage return" key. Unfortunatley, some output formats for
830: this manual cannot show the difference between @kbd{this} and
831: @code{this} which will make trying out the examples harder (but not
832: impossible).
833:
834: Forth is an unusual language. It provides an interactive development
835: environment which includes both an interpreter and compiler. Forth
836: programming style encourages you to break a problem down into many
837: @cindex factoring
838: small fragments (@var{factoring}), and then to develop and test each
839: fragment interactively. Forth advocates assert that breaking the
840: edit-compile-test cycle used by conventional programming languages can
841: lead to great productivity improvements.
842:
843: @menu
844: * Introducing the Text Interpreter::
845: * Stacks and Postfix notation::
846: * Your first definition::
847: * How does that work?::
848: * Forth is written in Forth::
849: * Classifying Forth words::
850: * Review - elements of a Forth system::
851: * Exercises::
852: @end menu
853: @comment TODO add these sections to the top xref lists
854:
855: @comment ----------------------------------------------
856: @node Introducing the Text Interpreter, Stacks and Postfix notation, Introduction, Introduction
857: @section Introducing the Text Interpreter
858: @cindex text interpreter
859: @cindex outer interpreter
860:
861: When you invoke the Forth image, you will see a startup banner printed
862: and nothing else (if you have Gforth installed on your system, try
863: invoking it now, by typing @kbd{gforth<return>}). Forth is now running
864: its command line interpreter, which is called the @var{Text Interpreter}
865: (also known as the @var{Outer Interpreter}). (@pxref{The Text
866: Interpreter} describes it in more detail, but we will learn more about
867: its behaviour as we go through this chapter).
868:
869: Although it may not be obvious, Forth is actually waiting for your
870: input. Type a number and press the <return> key:
871:
872: @example
873: @kbd{45<return>} ok
874: @end example
875:
876: Rather than give you a prompt to invite you to input something, the text
877: interpreter prints a status message @var{after} it has processed a line
878: of input. The status message in this case (" ok" followed by
879: carriage-return) indicates that the text interpreter was able to process
880: all of your input successfully. Now type something illegal:
881:
882: @example
883: @kbd{qwer341<return>}
884: ^^^^^^^
885: Error: Undefined word
886: @end example
887:
888: When the text interpreter detects an error, it discards any remaining
889: text on a line, resets certain internal state and prints an error
890: message.
891:
892: The text interpreter works on input one line at a time. Starting at
893: the beginning of the line, it breaks the line into groups of characters
894: separated by spaces. For each group of characters in turn, it makes two
895: attempts to do something:
896:
897: @itemize @bullet
898: @item
899: It tries to treat it as a command. It does this by searching a @var{name
900: dictionary}. If the group of characters matches an entry in the name
901: dictionary, the name dictionary provides the text interpreter with
902: information that allows the text interpreter perform some actions. In
903: Forth jargon, we say that the group
904: @cindex word
905: @cindex definition
906: @cindex execution token
907: @cindex xt
908: of characters names a @var{word}, that the dictionary search returns an
909: @var{execution token (xt)} corresponding to the @var{definition} of the
910: word, and that the text interpreter executes the xt. Often, the terms
911: @var{word} and @var{definition} are used interchangeably.
912: @item
913: If the text interpreter fails to find a match in the name dictionary, it
914: tries to treat the group of characters as a number in the current number
915: base (when you start up Forth, the current number base is base 10). If
916: the group of characters legitimately represents a number, the text
917: interpreter pushes the number onto a stack (we'll learn more about that
918: in the next section).
919: @end itemize
920:
921: If the text interpreter is unable to do either of these things with any
922: group of characters, it discards the rest of the line and print an error
923: message. If the text interpreter reaches the end of the line without
924: error, it prints the status message " ok" followed by carriage-return.
925:
926: This is the simplest command we can give to the text interpreter:
927:
928: @example
929: @kbd{<return>} ok
930: @end example
931:
932: The text interpreter did everything we asked it to do (nothing) without
933: an error, so it said that everything is "ok". Try a slightly longer
934: command:
935:
936: @example
937: @kbd{12 dup fred dup<return>}
938: ^^^^
939: Error: Undefined word
940: @end example
941:
942: When you pres the <return> key, the text interpreter starts to work its
943: way along the line.
944:
945: @itemize @bullet
946: @item
947: When it gets to the space after the @code{2}, it takes the group of
948: characters @code{12} and looks them up in the name
949: dictionary@footnote{We can't tell if it found them or not, but assume
950: for now that it did not}. There is no match for this group of characters
951: in the name dictionary, so it tries to treat them as a number. It is
952: able to do this successfully, so it puts the number, 12, "on the stack"
953: (whatever that means).
954: @item
955: The text interpreter resumes scanning the line and gets the next group
956: of characters, @code{dup}. It looks them up in the name dictionary and
957: (you'll have to take my word for this) finds them, and executes the word
958: @code{dup} (whatever that means).
959: @item
960: Once again, the text interpreter resumes scanning the line and gets the
961: group of characters @code{fred}. It looks them up in the name
962: dictionary, but can't find them. It tries to treat them as a number, but
963: they don't represent any legal number.
964: @end itemize
965:
966: At this point, the text interpreter gives up and prints an error
967: message. The error message shows exactly how far the text interpreter
968: got in processing the line. In particular, it shows that the text
969: interpreter made no attempt to do anything with the final character
970: group, @code{dup}, even though we have good reason to believe that the
971: text interpreter would have had no problems with looking that word up
972: and executing it a second time.
973:
974:
975: @comment ----------------------------------------------
976: @node Stacks and Postfix notation, Your first definition, Introducing the Text Interpreter, Introduction
977: @section Stacks, postfix notation and parameter passing
978: @cindex text interpreter
979: @cindex outer interpreter
980:
981: In procedural programming languages (like C and Pascal), the
982: building-block of programs is the function or procedure. These
983: functions or procedures are called with explicit parameters. For
984: example, in C we might write:
985:
986: @example
987: total = total + new_volume(length,height,depth);
988: @end example
989:
990: where total, length, height, depth are all variables and new_volume is
991: a function-call to another piece of code.
992:
993: In Forth, the equivalent to the function or procedure is the
994: @var{definition} and parameters are implicitly passed between
995: definitions using a shared stack that is visible to the
996: programmer. Although Forth does support variables, the existence of the
997: stack means that they are used far less often than in most other
998: programming languages. When the text interpreter encounters a number, it
999: will place (@var{push}) it on the stack. There are several stacks (the
1000: actual number is implementation-dependent ..) and the particular stack
1001: used for any operation is implied unambiguously by the operation being
1002: performed. The stack used for all integer operations is called the @var{data
1003: stack} and, since this is the stack used most commonly, references to
1004: "the data stack" are often abbreviated to "the stack".
1005:
1006: The stacks have a last-in, first-out (LIFO) organisation. If you type:
1007:
1008: @example
1009: @kbd{1 2 3<return>} ok
1010: @end example
1011:
1012: Then you (well, the text interpreter, really) have placed three numbers
1013: on the (data) stack. An analogy for the behaviour of the stack is to
1014: take a pack of playing cards and deal out the ace (1), 2 and 3 into a
1015: pile on the table. The 3 was the last card onto the pile ("last-in") and
1016: if you take a card off the pile then, unless you're prepared to fiddle a
1017: bit, the card that you take off will be the 3 ("first-out"). The number
1018: that will be first-out of the stack is called the "top of stack", which
1019: is often abbreviated to @var{TOS}.
1020:
1021: To see how parameters are passed in Forth, we will consider the
1022: behaviour of the definition @code{+} (pronounced "plus"). You will not be
1023: surprised to learn that this definition performs addition. More
1024: precisely, it adds two number together and produces a result. Where does
1025: it get the two numbers from? It takes the first two numbers off the
1026: stack. Where does it place the result? On the stack. You can act-out the
1027: behaviour of @code{+} with your playing cards like this:
1028:
1029: @itemize @bullet
1030: @item
1031: Pick up two cards from the stack
1032: @item
1033: Stare at them intently and ask yourself "what *is* the sum of these two
1034: numbers"
1035: @item
1036: Decide that the answer is 5
1037: @item
1038: Shuffle the two cards back into the pack and find a 5
1039: @item
1040: Put a 5 on the remaining ace that's on the table.
1041: @end itemize
1042:
1043: If you don't have a pack of cards handy but you do have Forth running,
1044: you can use the definition .s to show the current state of the stack,
1045: without affecting the stack. Type:
1046:
1047: @example
1048: @kbd{clearstack 1 2 3<return>} ok
1049: @kbd{.s<return> <3> 1 2 3 } ok
1050: @end example
1051:
1052: The text interpreter looks up the word @code{clearstack} and executes
1053: it; it tidies up the stack and removes any entries that may have been
1054: left on it by earlier examples. The text interpreter pushes each of the
1055: three numbers in turn onto the stack. Finally, the text interpreter
1056: looks up the word @code{.s} and executes it. The effect of executing
1057: @code{.s} is to print the "<3>" (the total number of items on the stack)
1058: followed by a list of all the items and the item on the far right-hand
1059: side is the TOS.
1060:
1061: You can now type:
1062:
1063: + .s<return> <2> 1 5 ok
1064:
1065: which is correct; there are now 2 items on the stack and the result of
1066: the addition is 5.
1067:
1068: If you're playing with cards, try doing a second addition; pick up the
1069: two cards, work out that their sum is 6, shuffle them into the pack,
1070: look for a 6 and place that on the table. You now have just one item
1071: on the stack. What happens if you try to do a third addition? Pick up
1072: the first card, pick up the second card - ah. There is no second
1073: card. This is called a "stack underflow" and consitutes an error. If
1074: you try to do the same thing with Forth it will report an error
1075: (probably a Stack Underflow or an Invalid Memory Address error).
1076:
1077: The opposite situation to a stack underflow is a stack overflow, which
1078: simply accepts that there is a finite amount of storage space reserved
1079: for the stack. To stretch the playing card analogy, if you had enough
1080: packs of cards and you piled the cards up on the table, you would
1081: eventually be unable to add another card; you'd hit the
1082: ceiling. Gforth allows you to set the maximum size of the stacks. In
1083: general, the only time that you will get a stack overflow is because a
1084: definition has a bug in it and is generating data on the stack
1085: uncontrollably.
1086:
1087: There's one final use for the playing card analogy. If you model your
1088: stack using a pack of playing cards, the maximum number of items on
1089: your stack will be 52 (I assume you didn't use the Joker). The maximum
1090: *value* of any item on the stack is 13 (the King). In fact, the only
1091: possible numbers are positive integer numbers 1 through 13; you can't
1092: have (for example) 0 or 27 or 3.52 or -2. If you change the way you
1093: think about some of the cards, you can accommodate different
1094: numbers. For example, you could think of the Jack as representing 0,
1095: the Queen as representing -1 and the King as representing -2. Your
1096: *range* remains unchanged (you can still only represent a total of 13
1097: numbers) but the numbers that you can represent are -2 through 10.
1098:
1099: In that analogy, the limit was the amount of information that a single
1100: stack entry could hold, and Forth has a similar limit. In Forth, the
1101: size of a stack entry is called a "cell". The actual size of a cell is
1102: implementation dependent and affects the maximum value that a stack
1103: entry can hold. A Standard Forth provides a cell size of at least
1104: 16-bits, and most desktop systems use a cell size of 32-bits.
1105:
1106: Forth does not do any type checking for you, so you are free to
1107: manipulate and combine stack items in any way you wish. A convenient
1108: ways of treating stack items is as 2's complement signed integers, and
1109: that is what Standard words like "+" do. Therefore you can type:
1110:
1111: -5 12 + .s<return> <1> 7 ok
1112:
1113: If you use numbers and definitions like "+" in order to turn Forth
1114: into a great big pocket calculator, you will realise that it's rather
1115: different from a normal calculator. Rather than typing 2 + 3 = you had
1116: to type 2 3 + (ignore the fact that you had to use .s to see the
1117: result). The terminology used to describe this difference is to say
1118: that your calculator uses "Infix Notation" (parameters and operators
1119: are mixed) whilst Forth uses "Postfix Notation" (parameters and
1120: operators are separate), also called "Reverse Polish Notation".
1121:
1122: Whilst postfix notation might look confusing to begin with, it has
1123: several important advantages:
1124:
1125: - it is unambiguous
1126: - it is more concise
1127: - it fits naturally with a stack-based system
1128:
1129: To examine these claims in more detail, consider these sums:
1130:
1131: 6 + 5 * 4 =
1132: 4 * 5 + 6 =
1133:
1134: If you're just learning maths or your maths is very rusty, you will
1135: probably come up with the answer 44 for the first and 26 for the
1136: second. If you are a bit of a whizz at maths you will remember the
1137: *convention* that multiplication takes precendence over addition, and
1138: you'd come up with the answer 26 both times. To explain the answer 26
1139: to someone who got the answer 44, you'd probably rewrite the first sum
1140: like this:
1141:
1142: 6 + (5 * 4) =
1143:
1144: If what you really wanted was to perform the addition before the
1145: multiplication, you would have to use parentheses to force it.
1146:
1147: If you did the first two sums on a pocket calculator you would probably
1148: get the right answers, unless you were very cautious and entered them using
1149: these keystroke sequences:
1150:
1151: 6 + 5 = * 4 =
1152: 4 * 5 = + 6 =
1153:
1154: Postfix notation is unambiguous because the order that the operators
1155: are applied is always explicit; that also means that parentheses are
1156: never required. The operators are *active* (the act of quoting the
1157: operator makes the operation occur) which removes the need for "=".
1158:
1159: The sum 6 + 5 * 4 can be written (in postfix notation) in two
1160: equivalent ways:
1161:
1162: 6 5 4 * + or:
1163: 5 4 * 6 +
1164:
1165: TODO point out that the order of number is never changed.
1166:
1167: TODO -- another way of thinking of this is to think of all Forth
1168: definitions as being ACTIVE. They execute as they are encountered by the
1169: text interpreter. With this mental model, it's easy to see that the only
1170: way of implementing an active scheme is to use postfix notation.
1171:
1172:
1173:
1174:
1175: .. up until now we've just been giving lists of commands that once
1176: exeduted are gone forwever (well, not really-- try pressing the up-arrow
1177: key.. you can recall, edit and re-enter )
1178:
1179:
1180: @comment ----------------------------------------------
1181: @node Your first definition, How does that work?, Stacks and Postfix notation, Introduction
1182: @section Your first Forth definition
1183: @cindex first definition
1184:
1185:
1186: The easiest way to create a new definition is to use a "colon
1187: definition". In order to provide a few examples (and give you some
1188: homework) I'm going to introduce a very small set of words but only
1189: describe what they do very informally, by example.
1190:
1191: + add the top two numbers on the stack and place the result on the
1192: stack
1193: . print the top stack item
1194: ." print text until a " delimiter is found
1195: CR print a carriage-return
1196: : start a new definition
1197: ; end a definition
1198: DUP blah
1199: DROP blah
1200:
1201: example 1:
1202: : greet ." Hello and welcome" ;<return> ok
1203: greet<return> Hello and welcome ok
1204: greet greet<return> Hello and welcomeHello and welcome ok
1205:
1206: When you try out this example, be careful to copy the spaces
1207: accurately; there needs to be a space between each group of characters
1208: that will be processed by the text interpreter.
1209:
1210:
1211: example 2:
1212: : add-two 2 + . ;<return> ok
1213: 5 add-two<return> 7 ok
1214:
1215:
1216: - numbers and definitions
1217: - redefining things .. what uses the old defn and what uses the new one
1218: - boundary between system definitions and your definitions
1219: - standards.. a double-edged sword
1220: - philosophy
1221:
1222: - your first set of definitions
1223:
1224:
1225:
1226: @comment ----------------------------------------------
1227: @node How does that work?, Forth is written in Forth, Your first definition, Introduction
1228: @section How does that work?
1229: @cindex parsing words
1230:
1231:
1232: todo parsing words .. trick the text interpreter
1233:
1234: .. switching from intepret to compile and back again
1235:
1236: .. what the text interpreter does.
1237:
1238: Now that we have looked at the behaviour of the text interpreter in
1239: greater detail, we can list all of the things that it knows how to do:
1240:
1241: @itemize @bullet
1242: @item
1243: It knows how to @var{compile} a number
1244: @item
1245: It knows how to @var{compile} a word into a new definition
1246: @item
1247: It knows how to @var{interpret} a number
1248: @item
1249: It knows how to @var{interpret} a word
1250: @end itemize
1251:
1252: The way in which the text interpreter interprets and compiles numbers is
1253: fixed; the effect of interpreting a number is to put that number on the
1254: stack, and the effect of compiling a number into a definition is to
1255: perform some trick whereby the number appears on the stack when the
1256: definition is executed.
1257:
1258: The way in which the text interpreter interprets and compiles words is
1259: not fixed; it is defined at the same time as the word is defined, and
1260: can be overridden in subtle ways later. When the text interpreter
1261: searches the name dictionary for a defintion, it not only retrieves the
1262: xt for the word, it also retrieves information about the way in which
1263: the words can behave.
1264:
1265:
1266: @comment TODO -- fix this up and decide whether I really want it here.
1267: @itemize @bullet
1268: @item
1269: Interpretation
1270: Compilation
1271: Description
1272:
1273: @item
1274: execute
1275: the xt is compiled
1276: Normal non-immediate definition. Created by default (eg using @code{:})
1277:
1278: @item
1279: execute
1280: execute
1281: Normal immediate definition. Created using @code{immediate} after definition.
1282:
1283: @item
1284: illegal (generate error)
1285: the xt is compiled
1286: Compile-only definition. Created using @code{compile-only} after definition.
1287:
1288: @item
1289: illegal (generate error)
1290: execute
1291: Immediate compile-only definition created using @code{immediate} @code{compile-only} after definition.
1292:
1293: @item
1294: execute
1295: illegal
1296: Interpret-only definition. No standard way to generate this.
1297:
1298: @end itemize
1299:
1300:
1301:
1302: @comment ----------------------------------------------
1303: @node Forth is written in Forth, Classifying Forth words, How does that work?, Introduction
1304: @section Forth is written in Forth
1305: @cindex structure of Forth programs
1306:
1307:
1308:
1309: Blah
1310:
1311: When you start up the Forth compiler, a large number of definitions
1312: already exist. To develop a new application, use bottom-up programming
1313: techniques to create new definitions that are defined in terms of
1314: existing definitions. As you create each definition you can test it
1315: interactively. Ultimately, you end up with an environment <blah blah>
1316:
1317: @comment TODO - other defining words
1318: @comment other parsing words
1319: @comment Your first loop
1320: @comment syntax and semantics
1321: @comment DOES>
1322: @comment taste of other elements of Forth
1323:
1324: @comment ----------------------------------------------
1325: @node Classifying Forth words, Review - elements of a Forth system, Forth is written in Forth, Introduction
1326: @section Classifying Forth words
1327: @cindex classifying Forth words
1328:
1329: It can be helpful to classify Forth words into a number of groups. We
1330: can classify any word in several orthogonal ways:
1331:
1332: @itemize @bullet
1333: @item
1334: Based upon the way in which it is implemented
1335: @item
1336: Based upon whether it affects the input stream
1337: @item
1338: Based upon its behaviour at different times
1339: @end itemize
1340:
1341: If we classify a word based upon the way in which it is implemented, we
1342: divide words into two groups:
1343:
1344: @itemize @bullet
1345: @item
1346: Those that are implemented in Forth (often called @var{high-level
1347: definitions}).
1348: @item
1349: Those that are not (often called @var{low-level definitions},
1350: @var{code definitions} or @var{primitives}).
1351: @end itemize
1352:
1353: When you are programming in Forth it should never make any difference to you (or
1354: even be apparent to you) whether any particular word is implemented as a
1355: high-level definition or a low-level definition. If you use the word
1356: disassembler, @code{see} you can easily find both types of words (try
1357: @kbd{see +} and @kbd{see :}).
1358:
1359: If we classify a word based upon the way in which it affects the input
1360: stream we also divide words into two groups:
1361:
1362: @itemize @bullet
1363: @item
1364: Those that do not affect the input stream (the vast majority of Forth
1365: definitions fall into this category).
1366: @item
1367: Those that do affect the input stream (these are called @var{parsing words}).
1368: @end itemize
1369:
1370: Here are some examples of ANS Standard parsing words; you can use the
1371: word index at the back of this manual to find out more about them:
1372:
1373: @code{:} @ @code{CONSTANT} @ @code{[CHAR]} @ @code{CHAR} @ @code{\}
1374:
1375: The most complex way of classifying Forth words is based upon their
1376: behaviour at different times. We have already seen how the text
1377: interpreter knows how to treat words differently depending upon whether
1378: it is interpreting or compiling,
1379:
1380: -- classifying words
1381: Three orthogonal ways:
1382: -- by function
1383: -- classifying words by the way in which they are defined
1384: -- classifying words by their behaviour
1385:
1386:
1387:
1388:
1389: .. interactive stuff
1390: 5 3 + . <return> 8 ok
1391:
1392: could have been split over several lines
1393:
1394: 5 . . <return>
1395:
1396:
1397: .. talk about syntax and semantics
1398:
1399:
1400: -- command-line recall and editing
1401:
1402:
1403: Recode this example to show that, when you define a word, the old
1404: definition becomes unavailable to any *subsequent* definitions.
1405:
1406: @example
1407: : greet ." Hello" ;
1408: : announce ." I just want to say " greet ;
1409: : greet ." Bog off" ;
1410: : another-announce ." I just want to say " greet ;
1411: @end example
1412:
1413: After these four words have been defined, invoking the three distinct words will have this result:
1414:
1415: @example
1416: greet Welcome
1417: announce I just want to say Hello
1418: another-announce I just want to say Bog off
1419: @end example
1420:
1421: The original definition of @code{greet} is no longer available.
1422:
1423: However, if you created two word lists and put alternative definitions of
1424: greet in each of them, you could control which was used by changing the search order, like this:
1425:
1426: @example
1427: <create two word lists>
1428: ALSO POLITE-WORDS DEFINITIONS
1429: : greet ." Hello" ;
1430: ALSO RUDE-WORDS DEFINITIONS
1431: : greet ." Bonjour" ;
1432:
1433: FORTH DEFINITIONS
1434: ALSO POLITE-WORDS
1435: : announce ." I just want to say " greet ;
1436: PREVIOUS
1437: ALSO RUDE-WORDS
1438: : another-announce ." I just want to say " greet ;
1439: PREVIOUS
1440: @end example
1441:
1442:
1443:
1444:
1445:
1446:
1447: - cells and chars
1448:
1449: - the text interpreter in "Compilation" state.
1450:
1451: -- elements of a forth system
1452: - text interpreter (outer interpreter)
1453: - compiler
1454: - inner interpreter
1455: - dictionaries and wordlists
1456: - stacks
1457:
1458: -- disparate spaces .. may be better to describe that elsewhere.
1459:
1460: -- show how to use the rest of the manual and how to use the ANS Forth Standard
1461:
1462: @comment ----------------------------------------------
1463: @node Review - elements of a Forth system, Exercises, Classifying Forth words, Introduction
1464: @section Review - elements of a Forth system
1465: @cindex elements of a Forth system
1466:
1467:
1468:
1469:
1470: @comment ----------------------------------------------
1471: @node Exercises, ,Review - elements of a Forth system, Introduction
1472: @section Exercises
1473: @cindex elements of a Forth system
1474:
1475: Ideally, provide a set of programming excercises linked into the stuff
1476: done already and into other sections of the manual. Provide solutions to
1477: all the exercises in a .fs file in the distribution. Get some
1478: inspiration from Starting Forth and Kelly&Spies.
1479:
1480:
1481: @c ----------------------------------------------------------
1482: @node Goals, Invoking Gforth, Introduction, Top
1.1 anton 1483: @comment node-name, next, previous, up
1484: @chapter Goals of Gforth
1485: @cindex Goals
1486: The goal of the Gforth Project is to develop a standard model for
1487: ANS Forth. This can be split into several subgoals:
1488:
1489: @itemize @bullet
1490: @item
1.21 crook 1491: Gforth should conform to the ANS Forth Standard.
1.1 anton 1492: @item
1493: It should be a model, i.e. it should define all the
1494: implementation-dependent things.
1495: @item
1496: It should become standard, i.e. widely accepted and used. This goal
1497: is the most difficult one.
1498: @end itemize
1499:
1500: To achieve these goals Gforth should be
1501: @itemize @bullet
1502: @item
1503: Similar to previous models (fig-Forth, F83)
1504: @item
1505: Powerful. It should provide for all the things that are considered
1506: necessary today and even some that are not yet considered necessary.
1507: @item
1508: Efficient. It should not get the reputation of being exceptionally
1509: slow.
1510: @item
1511: Free.
1512: @item
1513: Available on many machines/easy to port.
1514: @end itemize
1515:
1516: Have we achieved these goals? Gforth conforms to the ANS Forth
1517: standard. It may be considered a model, but we have not yet documented
1518: which parts of the model are stable and which parts we are likely to
1.12 anton 1519: change. It certainly has not yet become a de facto standard, but it
1520: appears to be quite popular. It has some similarities to and some
1521: differences from previous models. It has some powerful features, but not
1522: yet everything that we envisioned. We certainly have achieved our
1523: execution speed goals (@pxref{Performance}). It is free and available
1524: on many machines.
1.1 anton 1525:
1.21 crook 1526: @menu
1527: * Gforth Extensions Sinful?::
1528: @end menu
1529:
1530: @node Gforth Extensions Sinful?, , Goals, Goals
1531: @comment node-name, next, previous, up
1532: @section Is it a Sin to use Gforth Extensions?
1533: @cindex Gforth extensions
1534:
1535: If you've been paying attention, you will have realised that there is an
1536: ANS Standard for Forth. As you read through the rest of this manual, you
1537: will see documentation for @var{Standard} words, and documentation for
1538: some appealing Gforth @var{extensions}. You might ask yourself the
1539: question: @var{"Given that there is a standard, would I be committing a
1540: sin to use (non-Standard) Gforth extensions?"}
1.1 anton 1541:
1.21 crook 1542: The answer to that question is somewhat pragmatic and somewhat
1543: philosophical. Consider these points:
1.1 anton 1544:
1.21 crook 1545: @itemize @bullet
1546: @item
1547: A number of the Gforth extensions can be implemented in ANS Standard
1548: Forth using files provided in the @file{compat/} directory. These are
1549: mentioned in the text in passing.
1550: @item
1551: Forth has a rich historical precedent for programmers taking advantage
1552: of implementation-dependent features of their tools (for example,
1553: relying on a knowledge of the dictionary structure). Sometimes these
1554: techniques are necessary to extract every last bit of performance from
1555: the hardware, sometimes they are just a programming shorthand.
1556: @item
1557: The best way to break the rules is to know what the rules are. To learn
1558: the rules, there is no substitute for studying the text of the Standard
1559: itself. In particular, Appendix A of the Standard (@var{Rationale})
1560: provides a valuable insight into the thought processes of the technical
1561: committee.
1562: @item
1563: The best reason to break a rule is because you have to; because it's
1564: more productive to do that, because it makes your code run fast enough
1565: or because you can see no Standard way to achieve what you want to
1566: achieve.
1567: @end itemize
1.1 anton 1568:
1.21 crook 1569: The tool @file{ans-report.fs} (@pxref{ANS Report}) makes it easy to
1570: analyse your program and determine what non-Standard definitions it
1571: relies upon.
1.1 anton 1572:
1573:
1.12 anton 1574:
1.21 crook 1575: @c ----------------------------------------------------------
1576: @node Invoking Gforth, Words, Goals, Top
1.1 anton 1577: @chapter Invoking Gforth
1.21 crook 1578: @cindex Gforth - invoking
1.1 anton 1579: @cindex invoking Gforth
1580: @cindex running Gforth
1581: @cindex command-line options
1582: @cindex options on the command line
1583: @cindex flags on the command line
1584:
1585: You will usually just say @code{gforth}. In many other cases the default
1586: Gforth image will be invoked like this:
1587: @example
1588: gforth [files] [-e forth-code]
1589: @end example
1.12 anton 1590: This interprets the contents of the files and the Forth code in the order they
1.1 anton 1591: are given.
1592:
1593: In general, the command line looks like this:
1594:
1595: @example
1596: gforth [initialization options] [image-specific options]
1597: @end example
1598:
1599: The initialization options must come before the rest of the command
1600: line. They are:
1601:
1602: @table @code
1603: @cindex -i, command-line option
1604: @cindex --image-file, command-line option
1605: @item --image-file @var{file}
1606: @itemx -i @var{file}
1607: Loads the Forth image @var{file} instead of the default
1608: @file{gforth.fi} (@pxref{Image Files}).
1609:
1610: @cindex --path, command-line option
1611: @cindex -p, command-line option
1612: @item --path @var{path}
1613: @itemx -p @var{path}
1614: Uses @var{path} for searching the image file and Forth source code files
1615: instead of the default in the environment variable @code{GFORTHPATH} or
1616: the path specified at installation time (e.g.,
1617: @file{/usr/local/share/gforth/0.2.0:.}). A path is given as a list of
1618: directories, separated by @samp{:} (on Unix) or @samp{;} (on other OSs).
1619:
1620: @cindex --dictionary-size, command-line option
1621: @cindex -m, command-line option
1622: @cindex @var{size} parameters for command-line options
1623: @cindex size of the dictionary and the stacks
1624: @item --dictionary-size @var{size}
1625: @itemx -m @var{size}
1626: Allocate @var{size} space for the Forth dictionary space instead of
1627: using the default specified in the image (typically 256K). The
1.21 crook 1628: @var{size} specification for this and subsequent options consists of
1629: an integer and a unit (e.g.,
1.1 anton 1630: @code{4M}). The unit can be one of @code{b} (bytes), @code{e} (element
1.12 anton 1631: size, in this case Cells), @code{k} (kilobytes), @code{M} (Megabytes),
1632: @code{G} (Gigabytes), and @code{T} (Terabytes). If no unit is specified,
1633: @code{e} is used.
1.1 anton 1634:
1635: @cindex --data-stack-size, command-line option
1636: @cindex -d, command-line option
1637: @item --data-stack-size @var{size}
1638: @itemx -d @var{size}
1639: Allocate @var{size} space for the data stack instead of using the
1640: default specified in the image (typically 16K).
1641:
1642: @cindex --return-stack-size, command-line option
1643: @cindex -r, command-line option
1644: @item --return-stack-size @var{size}
1645: @itemx -r @var{size}
1646: Allocate @var{size} space for the return stack instead of using the
1647: default specified in the image (typically 15K).
1648:
1649: @cindex --fp-stack-size, command-line option
1650: @cindex -f, command-line option
1651: @item --fp-stack-size @var{size}
1652: @itemx -f @var{size}
1653: Allocate @var{size} space for the floating point stack instead of
1654: using the default specified in the image (typically 15.5K). In this case
1655: the unit specifier @code{e} refers to floating point numbers.
1656:
1657: @cindex --locals-stack-size, command-line option
1658: @cindex -l, command-line option
1659: @item --locals-stack-size @var{size}
1660: @itemx -l @var{size}
1661: Allocate @var{size} space for the locals stack instead of using the
1662: default specified in the image (typically 14.5K).
1663:
1664: @cindex -h, command-line option
1665: @cindex --help, command-line option
1666: @item --help
1667: @itemx -h
1668: Print a message about the command-line options
1669:
1670: @cindex -v, command-line option
1671: @cindex --version, command-line option
1672: @item --version
1673: @itemx -v
1674: Print version and exit
1675:
1676: @cindex --debug, command-line option
1677: @item --debug
1678: Print some information useful for debugging on startup.
1679:
1680: @cindex --offset-image, command-line option
1681: @item --offset-image
1682: Start the dictionary at a slightly different position than would be used
1683: otherwise (useful for creating data-relocatable images,
1684: @pxref{Data-Relocatable Image Files}).
1685:
1.5 anton 1686: @cindex --no-offset-im, command-line option
1687: @item --no-offset-im
1688: Start the dictionary at the normal position.
1689:
1.1 anton 1690: @cindex --clear-dictionary, command-line option
1691: @item --clear-dictionary
1692: Initialize all bytes in the dictionary to 0 before loading the image
1693: (@pxref{Data-Relocatable Image Files}).
1.5 anton 1694:
1695: @cindex --die-on-signal, command-line-option
1696: @item --die-on-signal
1697: Normally Gforth handles most signals (e.g., the user interrupt SIGINT,
1698: or the segmentation violation SIGSEGV) by translating it into a Forth
1699: @code{THROW}. With this option, Gforth exits if it receives such a
1700: signal. This option is useful when the engine and/or the image might be
1701: severely broken (such that it causes another signal before recovering
1702: from the first); this option avoids endless loops in such cases.
1.1 anton 1703: @end table
1704:
1705: @cindex loading files at startup
1706: @cindex executing code on startup
1707: @cindex batch processing with Gforth
1708: As explained above, the image-specific command-line arguments for the
1709: default image @file{gforth.fi} consist of a sequence of filenames and
1710: @code{-e @var{forth-code}} options that are interpreted in the sequence
1711: in which they are given. The @code{-e @var{forth-code}} or
1.21 crook 1712: @code{--evaluate @var{forth-code}} option evaluates the Forth
1.1 anton 1713: code. This option takes only one argument; if you want to evaluate more
1714: Forth words, you have to quote them or use several @code{-e}s. To exit
1715: after processing the command line (instead of entering interactive mode)
1716: append @code{-e bye} to the command line.
1717:
1718: @cindex versions, invoking other versions of Gforth
1719: If you have several versions of Gforth installed, @code{gforth} will
1720: invoke the version that was installed last. @code{gforth-@var{version}}
1721: invokes a specific version. You may want to use the option
1722: @code{--path}, if your environment contains the variable
1723: @code{GFORTHPATH}.
1724:
1725: Not yet implemented:
1726: On startup the system first executes the system initialization file
1727: (unless the option @code{--no-init-file} is given; note that the system
1728: resulting from using this option may not be ANS Forth conformant). Then
1729: the user initialization file @file{.gforth.fs} is executed, unless the
1730: option @code{--no-rc} is given; this file is first searched in @file{.},
1731: then in @file{~}, then in the normal path (see above).
1732:
1.21 crook 1733:
1734: @cindex Gforth - leaving
1735: @cindex leaving Gforth
1736:
1737: You can leave Gforth by typing @code{bye} or (if you invoked Gforth with
1738: the @code{--die-on-signal} option) Ctrl-C. When you leave Gforth, all of
1739: your definitions and data are discarded. @xref{Image Files} for ways
1740: of saving the state of the system before leaving Gforth.
1741:
1742: doc-bye
1743:
1744:
1.1 anton 1745: @node Words, Tools, Invoking Gforth, Top
1746: @chapter Forth Words
1747: @cindex Words
1748:
1749: @menu
1750: * Notation::
1.21 crook 1751: * Comments::
1752: * Boolean Flags::
1.1 anton 1753: * Arithmetic::
1754: * Stack Manipulation::
1.5 anton 1755: * Memory::
1.1 anton 1756: * Control Structures::
1757: * Locals::
1758: * Defining Words::
1.21 crook 1759: * The Text Interpreter::
1.5 anton 1760: * Structures::
1.12 anton 1761: * Object-oriented Forth::
1762: * Tokens for Words::
1.21 crook 1763: * Word Lists::
1764: * Environmental Queries::
1.12 anton 1765: * Files::
1766: * Including Files::
1767: * Blocks::
1768: * Other I/O::
1769: * Programming Tools::
1770: * Assembler and Code Words::
1771: * Threading Words::
1.21 crook 1772: * Passing Commands to the OS::
1773: * Miscellaneous Words::
1.1 anton 1774: @end menu
1775:
1.21 crook 1776: @node Notation, Comments, Words, Words
1.1 anton 1777: @section Notation
1778: @cindex notation of glossary entries
1779: @cindex format of glossary entries
1780: @cindex glossary notation format
1781: @cindex word glossary entry format
1782:
1783: The Forth words are described in this section in the glossary notation
1784: that has become a de-facto standard for Forth texts, i.e.,
1785:
1786: @format
1787: @var{word} @var{Stack effect} @var{wordset} @var{pronunciation}
1788: @end format
1789: @var{Description}
1790:
1791: @table @var
1792: @item word
1793: @cindex case insensitivity
1794: The name of the word. BTW, Gforth is case insensitive, so you can
1795: type the words in in lower case (However, @pxref{core-idef}).
1796:
1797: @item Stack effect
1798: @cindex stack effect
1799: The stack effect is written in the notation @code{@var{before} --
1800: @var{after}}, where @var{before} and @var{after} describe the top of
1801: stack entries before and after the execution of the word. The rest of
1802: the stack is not touched by the word. The top of stack is rightmost,
1803: i.e., a stack sequence is written as it is typed in. Note that Gforth
1804: uses a separate floating point stack, but a unified stack
1805: notation. Also, return stack effects are not shown in @var{stack
1806: effect}, but in @var{Description}. The name of a stack item describes
1807: the type and/or the function of the item. See below for a discussion of
1808: the types.
1809:
1810: All words have two stack effects: A compile-time stack effect and a
1811: run-time stack effect. The compile-time stack-effect of most words is
1812: @var{ -- }. If the compile-time stack-effect of a word deviates from
1813: this standard behaviour, or the word does other unusual things at
1814: compile time, both stack effects are shown; otherwise only the run-time
1815: stack effect is shown.
1816:
1817: @cindex pronounciation of words
1818: @item pronunciation
1819: How the word is pronounced.
1820:
1821: @cindex wordset
1822: @item wordset
1.21 crook 1823: The ANS Forth standard is divided into several word sets. A standard
1824: system need not support all of them. Therefore, in theory, the fewer
1825: word sets your program uses the more portable it will be. However, we
1826: suspect that most ANS Forth systems on personal machines will feature
1827: all word sets. Words that are not defined in the ANS standard have
1828: @code{gforth} or @code{gforth-internal} as word set. @code{gforth}
1.1 anton 1829: describes words that will work in future releases of Gforth;
1830: @code{gforth-internal} words are more volatile. Environmental query
1831: strings are also displayed like words; you can recognize them by the
1.21 crook 1832: @code{environment} in the word set field.
1.1 anton 1833:
1834: @item Description
1835: A description of the behaviour of the word.
1836: @end table
1837:
1838: @cindex types of stack items
1839: @cindex stack item types
1840: The type of a stack item is specified by the character(s) the name
1841: starts with:
1842:
1843: @table @code
1844: @item f
1845: @cindex @code{f}, stack item type
1846: Boolean flags, i.e. @code{false} or @code{true}.
1847: @item c
1848: @cindex @code{c}, stack item type
1849: Char
1850: @item w
1851: @cindex @code{w}, stack item type
1852: Cell, can contain an integer or an address
1853: @item n
1854: @cindex @code{n}, stack item type
1855: signed integer
1856: @item u
1857: @cindex @code{u}, stack item type
1858: unsigned integer
1859: @item d
1860: @cindex @code{d}, stack item type
1861: double sized signed integer
1862: @item ud
1863: @cindex @code{ud}, stack item type
1864: double sized unsigned integer
1865: @item r
1866: @cindex @code{r}, stack item type
1867: Float (on the FP stack)
1.21 crook 1868: @item a-
1.1 anton 1869: @cindex @code{a_}, stack item type
1870: Cell-aligned address
1.21 crook 1871: @item c-
1.1 anton 1872: @cindex @code{c_}, stack item type
1873: Char-aligned address (note that a Char may have two bytes in Windows NT)
1.21 crook 1874: @item f-
1.1 anton 1875: @cindex @code{f_}, stack item type
1876: Float-aligned address
1.21 crook 1877: @item df-
1.1 anton 1878: @cindex @code{df_}, stack item type
1879: Address aligned for IEEE double precision float
1.21 crook 1880: @item sf-
1.1 anton 1881: @cindex @code{sf_}, stack item type
1882: Address aligned for IEEE single precision float
1883: @item xt
1884: @cindex @code{xt}, stack item type
1885: Execution token, same size as Cell
1886: @item wid
1887: @cindex @code{wid}, stack item type
1.21 crook 1888: Word list ID, same size as Cell
1.1 anton 1889: @item f83name
1890: @cindex @code{f83name}, stack item type
1891: Pointer to a name structure
1892: @item "
1893: @cindex @code{"}, stack item type
1.12 anton 1894: string in the input stream (not on the stack). The terminating character
1895: is a blank by default. If it is not a blank, it is shown in @code{<>}
1.1 anton 1896: quotes.
1897: @end table
1898:
1.21 crook 1899: @node Comments, Boolean Flags, Notation, Words
1900: @section Comments
1901: @cindex Comments
1902:
1903: Forth supports two styles of comment; the traditional "in-line" comment,
1904: @code{(} and its modern cousin, the "comment to end of line"; @code{\}.
1905:
1906: doc-\
1907: doc-(
1908:
1909:
1910: @node Boolean Flags, Arithmetic, Comments, Words
1911: @section Boolean Flags
1912: @cindex Boolean Flags
1913:
1914: A Boolean flag is cell-sized. A cell with all bits clear represents the
1915: flag @code{false} and a flag with all bits set represents the flag
1916: @code{true}. Words that check a flag (for example, @var{IF}) will treat
1917: a cell that has @var{any} bit set as @code{true}.
1918:
1919: doc-true
1920: doc-false
1921:
1922:
1923: @node Arithmetic, Stack Manipulation, Boolean Flags, Words
1.1 anton 1924: @section Arithmetic
1925: @cindex arithmetic words
1926:
1927: @cindex division with potentially negative operands
1928: Forth arithmetic is not checked, i.e., you will not hear about integer
1929: overflow on addition or multiplication, you may hear about division by
1930: zero if you are lucky. The operator is written after the operands, but
1931: the operands are still in the original order. I.e., the infix @code{2-1}
1932: corresponds to @code{2 1 -}. Forth offers a variety of division
1933: operators. If you perform division with potentially negative operands,
1934: you do not want to use @code{/} or @code{/mod} with its undefined
1935: behaviour, but rather @code{fm/mod} or @code{sm/mod} (probably the
1936: former, @pxref{Mixed precision}).
1937:
1938: @menu
1939: * Single precision::
1940: * Bitwise operations::
1.21 crook 1941: * Double precision:: Double-cell integer arithmetic
1942: * Numeric comparison::
1.1 anton 1943: * Mixed precision:: operations with single and double-cell integers
1944: * Floating Point::
1945: @end menu
1946:
1947: @node Single precision, Bitwise operations, Arithmetic, Arithmetic
1948: @subsection Single precision
1949: @cindex single precision arithmetic words
1950:
1.21 crook 1951: By default, numbers in Forth are single-precision integers that are 1
1952: CELL in size. They can be signed or unsigned, depending upon how you
1953: treat them. @xref{Number Conversion} for the rules used by the text
1954: interpreter for recognising single-precision integers.
1955:
1.1 anton 1956: doc-+
1.21 crook 1957: doc-1+
1.1 anton 1958: doc--
1.21 crook 1959: doc-1-
1.1 anton 1960: doc-*
1961: doc-/
1962: doc-mod
1963: doc-/mod
1964: doc-negate
1965: doc-abs
1966: doc-min
1967: doc-max
1.21 crook 1968: doc-d>s
1.1 anton 1969:
1.21 crook 1970: @node Bitwise operations, Double precision, Single precision, Arithmetic
1.1 anton 1971: @subsection Bitwise operations
1972: @cindex bitwise operation words
1973:
1974: doc-and
1975: doc-or
1976: doc-xor
1977: doc-invert
1.21 crook 1978: doc-lshift
1979: doc-rshift
1.1 anton 1980: doc-2*
1.21 crook 1981: doc-d2*
1.1 anton 1982: doc-2/
1.21 crook 1983: doc-d2/
1984:
1985: @node Double precision, Numeric comparison, Bitwise operations, Arithmetic
1986: @subsection Double precision
1987: @cindex double precision arithmetic words
1988:
1989: @xref{Number Conversion} for the rules used by the text interpreter for
1990: recognising double-precision integers.
1991:
1992: A double precision number is represented by a cell pair, with the most
1993: significant digit at the TOS. It is trivial to convert an unsigned single
1994: to an (unsigned) double; simply push a @code{0} onto the TOS. Since numbers
1995: are represented by Gforth using 2's complement arithmetic, converting
1996: a signed single to a (signed) double requires sign-extension across the
1997: most significant digit. This can be achieved using @code{s>d}. The moral
1998: of the story is that you cannot convert a number without knowing what that
1999: number represents.
2000:
2001: doc-s>d
2002: doc-d+
2003: doc-d-
2004: doc-dnegate
2005: doc-dabs
2006: doc-dmin
2007: doc-dmax
2008:
2009: @node Numeric comparison, Mixed precision, Double precision, Arithmetic
2010: @subsection Numeric comparison
2011: @cindex numeric comparison words
2012:
2013: doc-0<
2014: doc-0<>
2015: doc-0=
2016: doc-<
2017: doc-<>
2018: doc-=
2019: doc->
2020: doc-d0<
2021: doc-d0=
2022: doc-d<
2023: doc-d=
2024: doc-u<
2025: doc-du<
2026: doc-u>
2027: doc-within
1.1 anton 2028:
1.21 crook 2029: @node Mixed precision, Floating Point, Numeric comparison, Arithmetic
1.1 anton 2030: @subsection Mixed precision
2031: @cindex mixed precision arithmetic words
2032:
2033: doc-m+
2034: doc-*/
2035: doc-*/mod
2036: doc-m*
2037: doc-um*
2038: doc-m*/
2039: doc-um/mod
2040: doc-fm/mod
2041: doc-sm/rem
2042:
1.21 crook 2043: @node Floating Point, , Mixed precision, Arithmetic
1.1 anton 2044: @subsection Floating Point
2045: @cindex floating point arithmetic words
2046:
1.21 crook 2047: @xref{Number Conversion} for the rules used by the text interpreter for
2048: recognising floating-point numbers.
1.1 anton 2049:
2050: @cindex angles in trigonometric operations
2051: @cindex trigonometric operations
2052: Angles in floating point operations are given in radians (a full circle
2053: has 2 pi radians). Note, that Gforth has a separate floating point
2054: stack, but we use the unified notation.
2055:
2056: @cindex floating-point arithmetic, pitfalls
2057: Floating point numbers have a number of unpleasant surprises for the
2058: unwary (e.g., floating point addition is not associative) and even a few
2059: for the wary. You should not use them unless you know what you are doing
2060: or you don't care that the results you get are totally bogus. If you
2061: want to learn about the problems of floating point numbers (and how to
2062: avoid them), you might start with @cite{David Goldberg, What Every
2063: Computer Scientist Should Know About Floating-Point Arithmetic, ACM
1.17 anton 2064: Computing Surveys 23(1):5@minus{}48, March 1991}
2065: (@url{http://www.validgh.com/goldberg/paper.ps}).
1.1 anton 2066:
1.21 crook 2067: doc-d>f
2068: doc-f>d
1.1 anton 2069: doc-f+
2070: doc-f-
2071: doc-f*
2072: doc-f/
2073: doc-fnegate
2074: doc-fabs
2075: doc-fmax
2076: doc-fmin
2077: doc-floor
2078: doc-fround
2079: doc-f**
2080: doc-fsqrt
2081: doc-fexp
2082: doc-fexpm1
2083: doc-fln
2084: doc-flnp1
2085: doc-flog
2086: doc-falog
2087: doc-fsin
2088: doc-fcos
2089: doc-fsincos
2090: doc-ftan
2091: doc-fasin
2092: doc-facos
2093: doc-fatan
2094: doc-fatan2
2095: doc-fsinh
2096: doc-fcosh
2097: doc-ftanh
2098: doc-fasinh
2099: doc-facosh
2100: doc-fatanh
1.21 crook 2101: doc-pi
2102: doc-f0<
2103: doc-f0=
2104: doc-f<
2105: doc-f<=
2106: doc-f<>
2107: doc-f=
2108: doc-f>
2109: doc-f>=
2110: doc-f2*
2111: doc-f2/
2112: doc-1/f
2113: doc-f~
2114: doc-precision
2115: doc-set-precision
1.1 anton 2116:
2117: @node Stack Manipulation, Memory, Arithmetic, Words
2118: @section Stack Manipulation
2119: @cindex stack manipulation words
2120:
2121: @cindex floating-point stack in the standard
1.21 crook 2122: Gforth maintains a number of separate stacks:
2123:
2124: @itemize @bullet
2125: @item
2126: A data stack (aka parameter stack) -- for characters, cells,
2127: addresses, and double cells.
2128:
2129: @item
2130: A floating point stack -- for floating point numbers.
2131:
2132: @item
2133: A return stack -- for storing the return addresses of colon
2134: definitions and other data.
2135:
2136: @item
2137: A locals stack for storing local variables.
2138: @end itemize
2139:
2140: Whilst every sane Forth has a separate floating-point stack, it is not
2141: strictly required; an ANS Forth system could theoretically keep
2142: floating-point numbers on the data stack. As an additional difficulty,
2143: you don't know how many cells a floating-point number takes. It is
2144: reportedly possible to write words in a way that they work also for a
2145: unified stack model, but we do not recommend trying it. Instead, just
2146: say that your program has an environmental dependency on a separate
2147: floating-point stack.
2148:
2149: doc-floating-stack
1.1 anton 2150:
2151: @cindex return stack and locals
2152: @cindex locals and return stack
1.21 crook 2153: A Forth system is allowed to keep local variables on the
1.1 anton 2154: return stack. This is reasonable, as local variables usually eliminate
2155: the need to use the return stack explicitly. So, if you want to produce
1.21 crook 2156: a standard compliant program and you are using local variables in a
2157: word, forget about return stack manipulations in that word (refer to the
1.1 anton 2158: standard document for the exact rules).
2159:
2160: @menu
2161: * Data stack::
2162: * Floating point stack::
2163: * Return stack::
2164: * Locals stack::
2165: * Stack pointer manipulation::
2166: @end menu
2167:
2168: @node Data stack, Floating point stack, Stack Manipulation, Stack Manipulation
2169: @subsection Data stack
2170: @cindex data stack manipulation words
2171: @cindex stack manipulations words, data stack
2172:
2173: doc-drop
2174: doc-nip
2175: doc-dup
2176: doc-over
2177: doc-tuck
2178: doc-swap
1.21 crook 2179: doc-pick
1.1 anton 2180: doc-rot
2181: doc--rot
2182: doc-?dup
2183: doc-roll
2184: doc-2drop
2185: doc-2nip
2186: doc-2dup
2187: doc-2over
2188: doc-2tuck
2189: doc-2swap
2190: doc-2rot
2191:
2192: @node Floating point stack, Return stack, Data stack, Stack Manipulation
2193: @subsection Floating point stack
2194: @cindex floating-point stack manipulation words
2195: @cindex stack manipulation words, floating-point stack
2196:
2197: doc-fdrop
2198: doc-fnip
2199: doc-fdup
2200: doc-fover
2201: doc-ftuck
2202: doc-fswap
1.21 crook 2203: doc-fpick
1.1 anton 2204: doc-frot
2205:
2206: @node Return stack, Locals stack, Floating point stack, Stack Manipulation
2207: @subsection Return stack
2208: @cindex return stack manipulation words
2209: @cindex stack manipulation words, return stack
2210:
2211: doc->r
2212: doc-r>
2213: doc-r@
2214: doc-rdrop
2215: doc-2>r
2216: doc-2r>
2217: doc-2r@
2218: doc-2rdrop
2219:
2220: @node Locals stack, Stack pointer manipulation, Return stack, Stack Manipulation
2221: @subsection Locals stack
2222:
1.21 crook 2223:
1.1 anton 2224: @node Stack pointer manipulation, , Locals stack, Stack Manipulation
2225: @subsection Stack pointer manipulation
2226: @cindex stack pointer manipulation words
2227:
1.21 crook 2228: doc-sp0
2229: doc-s0
1.1 anton 2230: doc-sp@
2231: doc-sp!
1.21 crook 2232: doc-fp0
1.1 anton 2233: doc-fp@
2234: doc-fp!
1.21 crook 2235: doc-rp0
2236: doc-r0
1.1 anton 2237: doc-rp@
2238: doc-rp!
1.21 crook 2239: doc-lp0
2240: doc-l0
1.1 anton 2241: doc-lp@
2242: doc-lp!
2243:
2244: @node Memory, Control Structures, Stack Manipulation, Words
2245: @section Memory
2246: @cindex Memory words
2247:
2248: @menu
2249: * Memory Access::
2250: * Address arithmetic::
2251: * Memory Blocks::
2252: @end menu
2253:
2254: @node Memory Access, Address arithmetic, Memory, Memory
2255: @subsection Memory Access
2256: @cindex memory access words
2257:
2258: doc-@
2259: doc-!
2260: doc-+!
2261: doc-c@
2262: doc-c!
2263: doc-2@
2264: doc-2!
2265: doc-f@
2266: doc-f!
2267: doc-sf@
2268: doc-sf!
2269: doc-df@
2270: doc-df!
2271:
2272: @node Address arithmetic, Memory Blocks, Memory Access, Memory
2273: @subsection Address arithmetic
2274: @cindex address arithmetic words
2275:
2276: ANS Forth does not specify the sizes of the data types. Instead, it
2277: offers a number of words for computing sizes and doing address
2278: arithmetic. Basically, address arithmetic is performed in terms of
2279: address units (aus); on most systems the address unit is one byte. Note
2280: that a character may have more than one au, so @code{chars} is no noop
2281: (on systems where it is a noop, it compiles to nothing).
2282:
2283: @cindex alignment of addresses for types
2284: ANS Forth also defines words for aligning addresses for specific
2285: types. Many computers require that accesses to specific data types
2286: must only occur at specific addresses; e.g., that cells may only be
2287: accessed at addresses divisible by 4. Even if a machine allows unaligned
2288: accesses, it can usually perform aligned accesses faster.
2289:
2290: For the performance-conscious: alignment operations are usually only
2291: necessary during the definition of a data structure, not during the
2292: (more frequent) accesses to it.
2293:
2294: ANS Forth defines no words for character-aligning addresses. This is not
2295: an oversight, but reflects the fact that addresses that are not
2296: char-aligned have no use in the standard and therefore will not be
2297: created.
2298:
2299: @cindex @code{CREATE} and alignment
2300: The standard guarantees that addresses returned by @code{CREATE}d words
2301: are cell-aligned; in addition, Gforth guarantees that these addresses
2302: are aligned for all purposes.
2303:
2304: Note that the standard defines a word @code{char}, which has nothing to
2305: do with address arithmetic.
2306:
2307: doc-chars
2308: doc-char+
2309: doc-cells
2310: doc-cell+
2311: doc-cell
2312: doc-align
2313: doc-aligned
2314: doc-floats
2315: doc-float+
2316: doc-float
2317: doc-falign
2318: doc-faligned
2319: doc-sfloats
2320: doc-sfloat+
2321: doc-sfalign
2322: doc-sfaligned
2323: doc-dfloats
2324: doc-dfloat+
2325: doc-dfalign
2326: doc-dfaligned
2327: doc-maxalign
2328: doc-maxaligned
2329: doc-cfalign
2330: doc-cfaligned
2331: doc-address-unit-bits
2332:
2333: @node Memory Blocks, , Address arithmetic, Memory
2334: @subsection Memory Blocks
2335: @cindex memory block words
2336:
1.21 crook 2337: Some of these words work on address units (increments of @code{CELL}),
2338: and expect a @code{CELL}-aligned address. Others work on character units
2339: (increments of @code{CHAR}), and expect a @code{CHAR}-aligned
2340: address. Choose the correct operation depending upon your data type. If
2341: you are moving a block of memory (for example, a region reserved by
2342: @code{allot}) it is safe to use @code{move}, and it should be faster
2343: than using @code{cmove}. If you are moving (for example) a string
2344: compiled using @code{S"}, it is not portable to use @code{move}; the
2345: alignment of the string in memory could change, and the relationship
2346: between @code{CELL} and @code{CHAR} could change.
2347:
2348: When copying characters between overlapping memory regions, choose
2349: carefully between @code{cmove} and @code{cmove>}.
2350:
2351: You can only use any of these words @var{portably} to access data space.
2352:
2353: @comment - think the naming of the arguments is wrong for move
1.1 anton 2354: doc-move
2355: doc-erase
2356:
1.21 crook 2357: @comment - think the naming of the arguments is wrong for cmove
1.1 anton 2358: doc-cmove
1.21 crook 2359: @comment - think the naming of the arguments is wrong for cmove>
1.1 anton 2360: doc-cmove>
2361: doc-fill
1.21 crook 2362: @comment - think the naming of the arguments is wrong for blank
1.1 anton 2363: doc-blank
1.21 crook 2364: doc-compare
2365: doc-search
1.1 anton 2366:
2367: @node Control Structures, Locals, Memory, Words
2368: @section Control Structures
2369: @cindex control structures
2370:
2371: Control structures in Forth cannot be used in interpret state, only in
2372: compile state@footnote{More precisely, they have no interpretation
2373: semantics (@pxref{Interpretation and Compilation Semantics})}, i.e., in
2374: a colon definition. We do not like this limitation, but have not seen a
2375: satisfying way around it yet, although many schemes have been proposed.
2376:
2377: @menu
2378: * Selection::
2379: * Simple Loops::
2380: * Counted Loops::
2381: * Arbitrary control structures::
2382: * Calls and returns::
2383: * Exception Handling::
2384: @end menu
2385:
2386: @node Selection, Simple Loops, Control Structures, Control Structures
2387: @subsection Selection
2388: @cindex selection control structures
2389: @cindex control structures for selection
2390:
2391: @cindex @code{IF} control structure
2392: @example
2393: @var{flag}
2394: IF
2395: @var{code}
2396: ENDIF
2397: @end example
1.21 crook 2398: @noindent
1.1 anton 2399: or
2400: @example
2401: @var{flag}
2402: IF
2403: @var{code1}
2404: ELSE
2405: @var{code2}
2406: ENDIF
2407: @end example
2408:
2409: You can use @code{THEN} instead of @code{ENDIF}. Indeed, @code{THEN} is
2410: standard, and @code{ENDIF} is not, although it is quite popular. We
2411: recommend using @code{ENDIF}, because it is less confusing for people
2412: who also know other languages (and is not prone to reinforcing negative
2413: prejudices against Forth in these people). Adding @code{ENDIF} to a
2414: system that only supplies @code{THEN} is simple:
2415: @example
1.21 crook 2416: : ENDIF POSTPONE THEN ; immediate
1.1 anton 2417: @end example
2418:
2419: [According to @cite{Webster's New Encyclopedic Dictionary}, @dfn{then
2420: (adv.)} has the following meanings:
2421: @quotation
2422: ... 2b: following next after in order ... 3d: as a necessary consequence
2423: (if you were there, then you saw them).
2424: @end quotation
2425: Forth's @code{THEN} has the meaning 2b, whereas @code{THEN} in Pascal
2426: and many other programming languages has the meaning 3d.]
2427:
1.21 crook 2428: Gforth also provides the words @code{?DUP-IF} and @code{?DUP-0=-IF}, so
1.1 anton 2429: you can avoid using @code{?dup}. Using these alternatives is also more
1.21 crook 2430: efficient than using @code{?dup}. Definitions in ANS Standard Forth
1.1 anton 2431: for @code{ENDIF}, @code{?DUP-IF} and @code{?DUP-0=-IF} are provided in
2432: @file{compat/control.fs}.
2433:
2434: @cindex @code{CASE} control structure
2435: @example
2436: @var{n}
2437: CASE
2438: @var{n1} OF @var{code1} ENDOF
2439: @var{n2} OF @var{code2} ENDOF
2440: @dots{}
2441: ENDCASE
2442: @end example
2443:
2444: Executes the first @var{codei}, where the @var{ni} is equal to
2445: @var{n}. A default case can be added by simply writing the code after
2446: the last @code{ENDOF}. It may use @var{n}, which is on top of the stack,
2447: but must not consume it.
2448:
2449: @node Simple Loops, Counted Loops, Selection, Control Structures
2450: @subsection Simple Loops
2451: @cindex simple loops
2452: @cindex loops without count
2453:
2454: @cindex @code{WHILE} loop
2455: @example
2456: BEGIN
2457: @var{code1}
2458: @var{flag}
2459: WHILE
2460: @var{code2}
2461: REPEAT
2462: @end example
2463:
2464: @var{code1} is executed and @var{flag} is computed. If it is true,
2465: @var{code2} is executed and the loop is restarted; If @var{flag} is
2466: false, execution continues after the @code{REPEAT}.
2467:
2468: @cindex @code{UNTIL} loop
2469: @example
2470: BEGIN
2471: @var{code}
2472: @var{flag}
2473: UNTIL
2474: @end example
2475:
2476: @var{code} is executed. The loop is restarted if @code{flag} is false.
2477:
2478: @cindex endless loop
2479: @cindex loops, endless
2480: @example
2481: BEGIN
2482: @var{code}
2483: AGAIN
2484: @end example
2485:
2486: This is an endless loop.
2487:
2488: @node Counted Loops, Arbitrary control structures, Simple Loops, Control Structures
2489: @subsection Counted Loops
2490: @cindex counted loops
2491: @cindex loops, counted
2492: @cindex @code{DO} loops
2493:
2494: The basic counted loop is:
2495: @example
2496: @var{limit} @var{start}
2497: ?DO
2498: @var{body}
2499: LOOP
2500: @end example
2501:
2502: This performs one iteration for every integer, starting from @var{start}
1.21 crook 2503: and up to, but excluding @var{limit}. The counter, or @var{index}, can be
2504: accessed with @code{i}. For example, the loop:
1.1 anton 2505: @example
2506: 10 0 ?DO
2507: i .
2508: LOOP
2509: @end example
1.21 crook 2510: @noindent
2511: prints @code{0 1 2 3 4 5 6 7 8 9}
2512:
1.1 anton 2513: The index of the innermost loop can be accessed with @code{i}, the index
2514: of the next loop with @code{j}, and the index of the third loop with
2515: @code{k}.
2516:
2517: doc-i
2518: doc-j
2519: doc-k
2520:
2521: The loop control data are kept on the return stack, so there are some
1.21 crook 2522: restrictions on mixing return stack accesses and counted loop words. In
2523: particuler, if you put values on the return stack outside the loop, you
2524: cannot read them inside the loop@footnote{well, not in a way that is
2525: portable.}. If you put values on the return stack within a loop, you
2526: have to remove them before the end of the loop and before accessing the
2527: index of the loop.
1.1 anton 2528:
2529: There are several variations on the counted loop:
2530:
1.21 crook 2531: @itemize @bullet
2532: @item
2533: @code{LEAVE} leaves the innermost counted loop immediately; execution
2534: continues after the associated @code{LOOP} or @code{NEXT}. For example:
2535:
2536: @example
2537: 10 0 ?DO i DUP . 3 = IF LEAVE THEN LOOP
2538: @end example
2539: prints @code{0 1 2 3}
2540:
1.1 anton 2541:
1.21 crook 2542: @item
2543: @code{UNLOOP} prepares for an abnormal loop exit, e.g., via
2544: @code{EXIT}. @code{UNLOOP} removes the loop control parameters from the
2545: return stack so @code{EXIT} can get to its return address. For example:
2546:
2547: @example
2548: : demo 10 0 ?DO i DUP . 3 = IF UNLOOP EXIT THEN LOOP ." Done" ;
2549: @end example
2550: prints @code{0 1 2 3}
2551:
2552:
2553: @item
1.1 anton 2554: If @var{start} is greater than @var{limit}, a @code{?DO} loop is entered
2555: (and @code{LOOP} iterates until they become equal by wrap-around
2556: arithmetic). This behaviour is usually not what you want. Therefore,
2557: Gforth offers @code{+DO} and @code{U+DO} (as replacements for
2558: @code{?DO}), which do not enter the loop if @var{start} is greater than
2559: @var{limit}; @code{+DO} is for signed loop parameters, @code{U+DO} for
2560: unsigned loop parameters.
2561:
1.21 crook 2562: @item
2563: @code{?DO} can be replaced by @code{DO}. @code{DO} always enters
2564: the loop, independent of the loop parameters. Do not use @code{DO}, even
2565: if you know that the loop is entered in any case. Such knowledge tends
2566: to become invalid during maintenance of a program, and then the
2567: @code{DO} will make trouble.
2568:
2569: @item
1.1 anton 2570: @code{LOOP} can be replaced with @code{@var{n} +LOOP}; this updates the
2571: index by @var{n} instead of by 1. The loop is terminated when the border
2572: between @var{limit-1} and @var{limit} is crossed. E.g.:
2573:
1.21 crook 2574: @example
2575: 4 0 +DO i . 2 +LOOP
2576: @end example
2577: @noindent
2578: prints @code{0 2}
2579:
2580: @example
2581: 4 1 +DO i . 2 +LOOP
2582: @end example
2583: @noindent
2584: prints @code{1 3}
1.1 anton 2585:
2586:
2587: @cindex negative increment for counted loops
2588: @cindex counted loops with negative increment
2589: The behaviour of @code{@var{n} +LOOP} is peculiar when @var{n} is negative:
2590:
1.21 crook 2591: @example
2592: -1 0 ?DO i . -1 +LOOP
2593: @end example
2594: @noindent
2595: prints @code{0 -1}
1.1 anton 2596:
1.21 crook 2597: @example
2598: 0 0 ?DO i . -1 +LOOP
2599: @end example
2600: prints nothing.
1.1 anton 2601:
2602: Therefore we recommend avoiding @code{@var{n} +LOOP} with negative
2603: @var{n}. One alternative is @code{@var{u} -LOOP}, which reduces the
2604: index by @var{u} each iteration. The loop is terminated when the border
2605: between @var{limit+1} and @var{limit} is crossed. Gforth also provides
2606: @code{-DO} and @code{U-DO} for down-counting loops. E.g.:
2607:
1.21 crook 2608: @example
2609: -2 0 -DO i . 1 -LOOP
2610: @end example
2611: @noindent
2612: prints @code{0 -1}
1.1 anton 2613:
1.21 crook 2614: @example
2615: -1 0 -DO i . 1 -LOOP
2616: @end example
2617: @noindent
2618: prints @code{0}
2619:
2620: @example
2621: 0 0 -DO i . 1 -LOOP
2622: @end example
2623: @noindent
2624: prints nothing.
1.1 anton 2625:
1.21 crook 2626: @end itemize
1.1 anton 2627:
2628: Unfortunately, @code{+DO}, @code{U+DO}, @code{-DO}, @code{U-DO} and
2629: @code{-LOOP} are not in the ANS Forth standard. However, an
2630: implementation for these words that uses only standard words is provided
2631: in @file{compat/loops.fs}.
2632:
2633:
2634:
2635: @cindex @code{FOR} loops
2636: Another counted loop is
2637: @example
2638: @var{n}
2639: FOR
2640: @var{body}
2641: NEXT
2642: @end example
2643: This is the preferred loop of native code compiler writers who are too
2644: lazy to optimize @code{?DO} loops properly. In Gforth, this loop
2645: iterates @var{n+1} times; @code{i} produces values starting with @var{n}
2646: and ending with 0. Other Forth systems may behave differently, even if
2647: they support @code{FOR} loops. To avoid problems, don't use @code{FOR}
2648: loops.
2649:
2650: @node Arbitrary control structures, Calls and returns, Counted Loops, Control Structures
2651: @subsection Arbitrary control structures
2652: @cindex control structures, user-defined
2653:
2654: @cindex control-flow stack
2655: ANS Forth permits and supports using control structures in a non-nested
2656: way. Information about incomplete control structures is stored on the
2657: control-flow stack. This stack may be implemented on the Forth data
2658: stack, and this is what we have done in Gforth.
2659:
2660: @cindex @code{orig}, control-flow stack item
2661: @cindex @code{dest}, control-flow stack item
2662: An @i{orig} entry represents an unresolved forward branch, a @i{dest}
2663: entry represents a backward branch target. A few words are the basis for
2664: building any control structure possible (except control structures that
2665: need storage, like calls, coroutines, and backtracking).
2666:
2667: doc-if
2668: doc-ahead
2669: doc-then
2670: doc-begin
2671: doc-until
2672: doc-again
2673: doc-cs-pick
2674: doc-cs-roll
2675:
1.21 crook 2676: The Standard words @code{CS-PICK} and @code{CS-ROLL} allow you to
2677: manipulate the control-flow stack in a portable way. Without them, you
2678: would need to know how many stack items are occupied by a control-flow
2679: entry (many systems use one cell. In Gforth they currently take three,
2680: but this may change in the future).
2681:
1.1 anton 2682:
2683: Some standard control structure words are built from these words:
2684:
2685: doc-else
2686: doc-while
2687: doc-repeat
2688:
2689: Gforth adds some more control-structure words:
2690:
2691: doc-endif
2692: doc-?dup-if
2693: doc-?dup-0=-if
2694:
2695: Counted loop words constitute a separate group of words:
2696:
2697: doc-?do
2698: doc-+do
2699: doc-u+do
2700: doc--do
2701: doc-u-do
2702: doc-do
2703: doc-for
2704: doc-loop
2705: doc-+loop
2706: doc--loop
2707: doc-next
2708: doc-leave
2709: doc-?leave
2710: doc-unloop
2711: doc-done
2712:
1.21 crook 2713: The standard does not allow using @code{CS-PICK} and @code{CS-ROLL} on
2714: @i{do-sys}. Gforth allows it, but it's your job to ensure that for
1.1 anton 2715: every @code{?DO} etc. there is exactly one @code{UNLOOP} on any path
2716: through the definition (@code{LOOP} etc. compile an @code{UNLOOP} on the
2717: fall-through path). Also, you have to ensure that all @code{LEAVE}s are
2718: resolved (by using one of the loop-ending words or @code{DONE}).
2719:
2720: Another group of control structure words are
2721:
2722: doc-case
2723: doc-endcase
2724: doc-of
2725: doc-endof
2726:
1.21 crook 2727: @i{case-sys} and @i{of-sys} cannot be processed using @code{CS-PICK} and
2728: @code{CS-ROLL}.
1.1 anton 2729:
2730: @subsubsection Programming Style
2731:
2732: In order to ensure readability we recommend that you do not create
2733: arbitrary control structures directly, but define new control structure
2734: words for the control structure you want and use these words in your
2735: program.
2736:
1.21 crook 2737: E.g., instead of writing:
1.1 anton 2738:
2739: @example
2740: begin
2741: ...
2742: if [ 1 cs-roll ]
2743: ...
2744: again then
2745: @end example
2746:
1.21 crook 2747: @noindent
1.1 anton 2748: we recommend defining control structure words, e.g.,
2749:
2750: @example
2751: : while ( dest -- orig dest )
2752: POSTPONE if
2753: 1 cs-roll ; immediate
2754:
2755: : repeat ( orig dest -- )
2756: POSTPONE again
2757: POSTPONE then ; immediate
2758: @end example
2759:
1.21 crook 2760: @noindent
1.1 anton 2761: and then using these to create the control structure:
2762:
2763: @example
2764: begin
2765: ...
2766: while
2767: ...
2768: repeat
2769: @end example
2770:
2771: That's much easier to read, isn't it? Of course, @code{REPEAT} and
2772: @code{WHILE} are predefined, so in this example it would not be
2773: necessary to define them.
2774:
2775: @node Calls and returns, Exception Handling, Arbitrary control structures, Control Structures
2776: @subsection Calls and returns
2777: @cindex calling a definition
2778: @cindex returning from a definition
2779:
1.3 anton 2780: @cindex recursive definitions
2781: A definition can be called simply be writing the name of the definition
2782: to be called. Note that normally a definition is invisible during its
2783: definition. If you want to write a directly recursive definition, you
2784: can use @code{recursive} to make the current definition visible.
2785:
2786: doc-recursive
2787:
2788: Another way to perform a recursive call is
2789:
2790: doc-recurse
2791:
1.21 crook 2792: @comment TODO add example of the two recursion methods
1.12 anton 2793: @quotation
2794: @progstyle
2795: I prefer using @code{recursive} to @code{recurse}, because calling the
2796: definition by name is more descriptive (if the name is well-chosen) than
2797: the somewhat cryptic @code{recurse}. E.g., in a quicksort
2798: implementation, it is much better to read (and think) ``now sort the
2799: partitions'' than to read ``now do a recursive call''.
2800: @end quotation
1.3 anton 2801:
1.21 crook 2802: @comment TODO maybe move deferred words to Defining Words section and x-ref
2803: @comment from here.. that is where these two are glossed.
2804:
1.3 anton 2805: For mutual recursion, use @code{defer}red words, like this:
2806:
2807: @example
2808: defer foo
2809:
2810: : bar ( ... -- ... )
2811: ... foo ... ;
2812:
2813: :noname ( ... -- ... )
2814: ... bar ... ;
2815: IS foo
2816: @end example
2817:
2818: When the end of the definition is reached, it returns. An earlier return
2819: can be forced using
1.1 anton 2820:
2821: doc-exit
2822:
2823: Don't forget to clean up the return stack and @code{UNLOOP} any
1.21 crook 2824: outstanding @code{?DO}...@code{LOOP}s before @code{EXIT}ing.
1.1 anton 2825:
2826: doc-;s
2827:
2828: @node Exception Handling, , Calls and returns, Control Structures
2829: @subsection Exception Handling
2830: @cindex Exceptions
2831:
1.21 crook 2832: @comment TODO examples and blurb
1.1 anton 2833: doc-catch
2834: doc-throw
1.21 crook 2835: @comment TODO -- think this will alllcate you a new THROW code?
2836: @comment for reserving new exception numbers. Note the existence of compat/exception.fs
2837: doc---exception-exception
2838: doc-quit
2839: doc-abort
2840: doc-abort"
2841:
1.1 anton 2842:
2843: @node Locals, Defining Words, Control Structures, Words
2844: @section Locals
2845: @cindex locals
2846:
2847: Local variables can make Forth programming more enjoyable and Forth
2848: programs easier to read. Unfortunately, the locals of ANS Forth are
2849: laden with restrictions. Therefore, we provide not only the ANS Forth
2850: locals wordset, but also our own, more powerful locals wordset (we
2851: implemented the ANS Forth locals wordset through our locals wordset).
2852:
2853: The ideas in this section have also been published in the paper
2854: @cite{Automatic Scoping of Local Variables} by M. Anton Ertl, presented
2855: at EuroForth '94; it is available at
2856: @*@url{http://www.complang.tuwien.ac.at/papers/ertl94l.ps.gz}.
2857:
2858: @menu
2859: * Gforth locals::
2860: * ANS Forth locals::
2861: @end menu
2862:
2863: @node Gforth locals, ANS Forth locals, Locals, Locals
2864: @subsection Gforth locals
2865: @cindex Gforth locals
2866: @cindex locals, Gforth style
2867:
2868: Locals can be defined with
2869:
2870: @example
2871: @{ local1 local2 ... -- comment @}
2872: @end example
2873: or
2874: @example
2875: @{ local1 local2 ... @}
2876: @end example
2877:
2878: E.g.,
2879: @example
2880: : max @{ n1 n2 -- n3 @}
2881: n1 n2 > if
2882: n1
2883: else
2884: n2
2885: endif ;
2886: @end example
2887:
2888: The similarity of locals definitions with stack comments is intended. A
2889: locals definition often replaces the stack comment of a word. The order
2890: of the locals corresponds to the order in a stack comment and everything
2891: after the @code{--} is really a comment.
2892:
2893: This similarity has one disadvantage: It is too easy to confuse locals
2894: declarations with stack comments, causing bugs and making them hard to
2895: find. However, this problem can be avoided by appropriate coding
2896: conventions: Do not use both notations in the same program. If you do,
2897: they should be distinguished using additional means, e.g. by position.
2898:
2899: @cindex types of locals
2900: @cindex locals types
2901: The name of the local may be preceded by a type specifier, e.g.,
2902: @code{F:} for a floating point value:
2903:
2904: @example
2905: : CX* @{ F: Ar F: Ai F: Br F: Bi -- Cr Ci @}
2906: \ complex multiplication
2907: Ar Br f* Ai Bi f* f-
2908: Ar Bi f* Ai Br f* f+ ;
2909: @end example
2910:
2911: @cindex flavours of locals
2912: @cindex locals flavours
2913: @cindex value-flavoured locals
2914: @cindex variable-flavoured locals
2915: Gforth currently supports cells (@code{W:}, @code{W^}), doubles
2916: (@code{D:}, @code{D^}), floats (@code{F:}, @code{F^}) and characters
2917: (@code{C:}, @code{C^}) in two flavours: a value-flavoured local (defined
2918: with @code{W:}, @code{D:} etc.) produces its value and can be changed
2919: with @code{TO}. A variable-flavoured local (defined with @code{W^} etc.)
2920: produces its address (which becomes invalid when the variable's scope is
2921: left). E.g., the standard word @code{emit} can be defined in terms of
2922: @code{type} like this:
2923:
2924: @example
2925: : emit @{ C^ char* -- @}
2926: char* 1 type ;
2927: @end example
2928:
2929: @cindex default type of locals
2930: @cindex locals, default type
2931: A local without type specifier is a @code{W:} local. Both flavours of
2932: locals are initialized with values from the data or FP stack.
2933:
2934: Currently there is no way to define locals with user-defined data
2935: structures, but we are working on it.
2936:
2937: Gforth allows defining locals everywhere in a colon definition. This
2938: poses the following questions:
2939:
2940: @menu
2941: * Where are locals visible by name?::
2942: * How long do locals live?::
2943: * Programming Style::
2944: * Implementation::
2945: @end menu
2946:
2947: @node Where are locals visible by name?, How long do locals live?, Gforth locals, Gforth locals
2948: @subsubsection Where are locals visible by name?
2949: @cindex locals visibility
2950: @cindex visibility of locals
2951: @cindex scope of locals
2952:
2953: Basically, the answer is that locals are visible where you would expect
2954: it in block-structured languages, and sometimes a little longer. If you
2955: want to restrict the scope of a local, enclose its definition in
2956: @code{SCOPE}...@code{ENDSCOPE}.
2957:
2958: doc-scope
2959: doc-endscope
2960:
2961: These words behave like control structure words, so you can use them
2962: with @code{CS-PICK} and @code{CS-ROLL} to restrict the scope in
2963: arbitrary ways.
2964:
2965: If you want a more exact answer to the visibility question, here's the
2966: basic principle: A local is visible in all places that can only be
2967: reached through the definition of the local@footnote{In compiler
2968: construction terminology, all places dominated by the definition of the
2969: local.}. In other words, it is not visible in places that can be reached
2970: without going through the definition of the local. E.g., locals defined
2971: in @code{IF}...@code{ENDIF} are visible until the @code{ENDIF}, locals
2972: defined in @code{BEGIN}...@code{UNTIL} are visible after the
2973: @code{UNTIL} (until, e.g., a subsequent @code{ENDSCOPE}).
2974:
2975: The reasoning behind this solution is: We want to have the locals
2976: visible as long as it is meaningful. The user can always make the
2977: visibility shorter by using explicit scoping. In a place that can
2978: only be reached through the definition of a local, the meaning of a
2979: local name is clear. In other places it is not: How is the local
2980: initialized at the control flow path that does not contain the
2981: definition? Which local is meant, if the same name is defined twice in
2982: two independent control flow paths?
2983:
2984: This should be enough detail for nearly all users, so you can skip the
2985: rest of this section. If you really must know all the gory details and
2986: options, read on.
2987:
2988: In order to implement this rule, the compiler has to know which places
2989: are unreachable. It knows this automatically after @code{AHEAD},
2990: @code{AGAIN}, @code{EXIT} and @code{LEAVE}; in other cases (e.g., after
2991: most @code{THROW}s), you can use the word @code{UNREACHABLE} to tell the
2992: compiler that the control flow never reaches that place. If
2993: @code{UNREACHABLE} is not used where it could, the only consequence is
2994: that the visibility of some locals is more limited than the rule above
2995: says. If @code{UNREACHABLE} is used where it should not (i.e., if you
2996: lie to the compiler), buggy code will be produced.
2997:
2998: doc-unreachable
2999:
3000: Another problem with this rule is that at @code{BEGIN}, the compiler
3001: does not know which locals will be visible on the incoming
3002: back-edge. All problems discussed in the following are due to this
3003: ignorance of the compiler (we discuss the problems using @code{BEGIN}
3004: loops as examples; the discussion also applies to @code{?DO} and other
3005: loops). Perhaps the most insidious example is:
3006: @example
3007: AHEAD
3008: BEGIN
3009: x
3010: [ 1 CS-ROLL ] THEN
3011: @{ x @}
3012: ...
3013: UNTIL
3014: @end example
3015:
3016: This should be legal according to the visibility rule. The use of
3017: @code{x} can only be reached through the definition; but that appears
3018: textually below the use.
3019:
3020: From this example it is clear that the visibility rules cannot be fully
3021: implemented without major headaches. Our implementation treats common
3022: cases as advertised and the exceptions are treated in a safe way: The
3023: compiler makes a reasonable guess about the locals visible after a
3024: @code{BEGIN}; if it is too pessimistic, the
3025: user will get a spurious error about the local not being defined; if the
3026: compiler is too optimistic, it will notice this later and issue a
3027: warning. In the case above the compiler would complain about @code{x}
3028: being undefined at its use. You can see from the obscure examples in
3029: this section that it takes quite unusual control structures to get the
3030: compiler into trouble, and even then it will often do fine.
3031:
3032: If the @code{BEGIN} is reachable from above, the most optimistic guess
3033: is that all locals visible before the @code{BEGIN} will also be
3034: visible after the @code{BEGIN}. This guess is valid for all loops that
3035: are entered only through the @code{BEGIN}, in particular, for normal
3036: @code{BEGIN}...@code{WHILE}...@code{REPEAT} and
3037: @code{BEGIN}...@code{UNTIL} loops and it is implemented in our
3038: compiler. When the branch to the @code{BEGIN} is finally generated by
3039: @code{AGAIN} or @code{UNTIL}, the compiler checks the guess and
3040: warns the user if it was too optimistic:
3041: @example
3042: IF
3043: @{ x @}
3044: BEGIN
3045: \ x ?
3046: [ 1 cs-roll ] THEN
3047: ...
3048: UNTIL
3049: @end example
3050:
3051: Here, @code{x} lives only until the @code{BEGIN}, but the compiler
3052: optimistically assumes that it lives until the @code{THEN}. It notices
3053: this difference when it compiles the @code{UNTIL} and issues a
3054: warning. The user can avoid the warning, and make sure that @code{x}
3055: is not used in the wrong area by using explicit scoping:
3056: @example
3057: IF
3058: SCOPE
3059: @{ x @}
3060: ENDSCOPE
3061: BEGIN
3062: [ 1 cs-roll ] THEN
3063: ...
3064: UNTIL
3065: @end example
3066:
3067: Since the guess is optimistic, there will be no spurious error messages
3068: about undefined locals.
3069:
3070: If the @code{BEGIN} is not reachable from above (e.g., after
3071: @code{AHEAD} or @code{EXIT}), the compiler cannot even make an
3072: optimistic guess, as the locals visible after the @code{BEGIN} may be
3073: defined later. Therefore, the compiler assumes that no locals are
3074: visible after the @code{BEGIN}. However, the user can use
3075: @code{ASSUME-LIVE} to make the compiler assume that the same locals are
3076: visible at the BEGIN as at the point where the top control-flow stack
3077: item was created.
3078:
3079: doc-assume-live
3080:
3081: E.g.,
3082: @example
3083: @{ x @}
3084: AHEAD
3085: ASSUME-LIVE
3086: BEGIN
3087: x
3088: [ 1 CS-ROLL ] THEN
3089: ...
3090: UNTIL
3091: @end example
3092:
3093: Other cases where the locals are defined before the @code{BEGIN} can be
3094: handled by inserting an appropriate @code{CS-ROLL} before the
3095: @code{ASSUME-LIVE} (and changing the control-flow stack manipulation
3096: behind the @code{ASSUME-LIVE}).
3097:
3098: Cases where locals are defined after the @code{BEGIN} (but should be
3099: visible immediately after the @code{BEGIN}) can only be handled by
3100: rearranging the loop. E.g., the ``most insidious'' example above can be
3101: arranged into:
3102: @example
3103: BEGIN
3104: @{ x @}
3105: ... 0=
3106: WHILE
3107: x
3108: REPEAT
3109: @end example
3110:
3111: @node How long do locals live?, Programming Style, Where are locals visible by name?, Gforth locals
3112: @subsubsection How long do locals live?
3113: @cindex locals lifetime
3114: @cindex lifetime of locals
3115:
3116: The right answer for the lifetime question would be: A local lives at
3117: least as long as it can be accessed. For a value-flavoured local this
3118: means: until the end of its visibility. However, a variable-flavoured
3119: local could be accessed through its address far beyond its visibility
3120: scope. Ultimately, this would mean that such locals would have to be
3121: garbage collected. Since this entails un-Forth-like implementation
3122: complexities, I adopted the same cowardly solution as some other
3123: languages (e.g., C): The local lives only as long as it is visible;
3124: afterwards its address is invalid (and programs that access it
3125: afterwards are erroneous).
3126:
3127: @node Programming Style, Implementation, How long do locals live?, Gforth locals
3128: @subsubsection Programming Style
3129: @cindex locals programming style
3130: @cindex programming style, locals
3131:
3132: The freedom to define locals anywhere has the potential to change
3133: programming styles dramatically. In particular, the need to use the
3134: return stack for intermediate storage vanishes. Moreover, all stack
3135: manipulations (except @code{PICK}s and @code{ROLL}s with run-time
3136: determined arguments) can be eliminated: If the stack items are in the
3137: wrong order, just write a locals definition for all of them; then
3138: write the items in the order you want.
3139:
3140: This seems a little far-fetched and eliminating stack manipulations is
3141: unlikely to become a conscious programming objective. Still, the number
3142: of stack manipulations will be reduced dramatically if local variables
3143: are used liberally (e.g., compare @code{max} in @ref{Gforth locals} with
3144: a traditional implementation of @code{max}).
3145:
3146: This shows one potential benefit of locals: making Forth programs more
3147: readable. Of course, this benefit will only be realized if the
3148: programmers continue to honour the principle of factoring instead of
3149: using the added latitude to make the words longer.
3150:
3151: @cindex single-assignment style for locals
3152: Using @code{TO} can and should be avoided. Without @code{TO},
3153: every value-flavoured local has only a single assignment and many
3154: advantages of functional languages apply to Forth. I.e., programs are
3155: easier to analyse, to optimize and to read: It is clear from the
3156: definition what the local stands for, it does not turn into something
3157: different later.
3158:
3159: E.g., a definition using @code{TO} might look like this:
3160: @example
3161: : strcmp @{ addr1 u1 addr2 u2 -- n @}
3162: u1 u2 min 0
3163: ?do
3164: addr1 c@@ addr2 c@@ -
3165: ?dup-if
3166: unloop exit
3167: then
3168: addr1 char+ TO addr1
3169: addr2 char+ TO addr2
3170: loop
3171: u1 u2 - ;
3172: @end example
3173: Here, @code{TO} is used to update @code{addr1} and @code{addr2} at
3174: every loop iteration. @code{strcmp} is a typical example of the
3175: readability problems of using @code{TO}. When you start reading
3176: @code{strcmp}, you think that @code{addr1} refers to the start of the
3177: string. Only near the end of the loop you realize that it is something
3178: else.
3179:
3180: This can be avoided by defining two locals at the start of the loop that
3181: are initialized with the right value for the current iteration.
3182: @example
3183: : strcmp @{ addr1 u1 addr2 u2 -- n @}
3184: addr1 addr2
3185: u1 u2 min 0
3186: ?do @{ s1 s2 @}
3187: s1 c@@ s2 c@@ -
3188: ?dup-if
3189: unloop exit
3190: then
3191: s1 char+ s2 char+
3192: loop
3193: 2drop
3194: u1 u2 - ;
3195: @end example
3196: Here it is clear from the start that @code{s1} has a different value
3197: in every loop iteration.
3198:
3199: @node Implementation, , Programming Style, Gforth locals
3200: @subsubsection Implementation
3201: @cindex locals implementation
3202: @cindex implementation of locals
3203:
3204: @cindex locals stack
3205: Gforth uses an extra locals stack. The most compelling reason for
3206: this is that the return stack is not float-aligned; using an extra stack
3207: also eliminates the problems and restrictions of using the return stack
3208: as locals stack. Like the other stacks, the locals stack grows toward
3209: lower addresses. A few primitives allow an efficient implementation:
3210:
3211: doc-@local#
3212: doc-f@local#
3213: doc-laddr#
3214: doc-lp+!#
3215: doc-lp!
3216: doc->l
3217: doc-f>l
3218:
3219: In addition to these primitives, some specializations of these
3220: primitives for commonly occurring inline arguments are provided for
3221: efficiency reasons, e.g., @code{@@local0} as specialization of
3222: @code{@@local#} for the inline argument 0. The following compiling words
3223: compile the right specialized version, or the general version, as
3224: appropriate:
3225:
3226: doc-compile-@local
3227: doc-compile-f@local
3228: doc-compile-lp+!
3229:
3230: Combinations of conditional branches and @code{lp+!#} like
3231: @code{?branch-lp+!#} (the locals pointer is only changed if the branch
3232: is taken) are provided for efficiency and correctness in loops.
3233:
3234: A special area in the dictionary space is reserved for keeping the
3235: local variable names. @code{@{} switches the dictionary pointer to this
3236: area and @code{@}} switches it back and generates the locals
3237: initializing code. @code{W:} etc.@ are normal defining words. This
3238: special area is cleared at the start of every colon definition.
3239:
1.21 crook 3240: @cindex word list for defining locals
1.1 anton 3241: A special feature of Gforth's dictionary is used to implement the
1.21 crook 3242: definition of locals without type specifiers: every word list (aka
1.1 anton 3243: vocabulary) has its own methods for searching
1.21 crook 3244: etc. (@pxref{Word Lists}). For the present purpose we defined a word list
1.1 anton 3245: with a special search method: When it is searched for a word, it
3246: actually creates that word using @code{W:}. @code{@{} changes the search
1.21 crook 3247: order to first search the word list containing @code{@}}, @code{W:} etc.,
3248: and then the word list for defining locals without type specifiers.
1.1 anton 3249:
3250: The lifetime rules support a stack discipline within a colon
3251: definition: The lifetime of a local is either nested with other locals
3252: lifetimes or it does not overlap them.
3253:
3254: At @code{BEGIN}, @code{IF}, and @code{AHEAD} no code for locals stack
3255: pointer manipulation is generated. Between control structure words
3256: locals definitions can push locals onto the locals stack. @code{AGAIN}
3257: is the simplest of the other three control flow words. It has to
3258: restore the locals stack depth of the corresponding @code{BEGIN}
3259: before branching. The code looks like this:
3260: @format
3261: @code{lp+!#} current-locals-size @minus{} dest-locals-size
3262: @code{branch} <begin>
3263: @end format
3264:
3265: @code{UNTIL} is a little more complicated: If it branches back, it
3266: must adjust the stack just like @code{AGAIN}. But if it falls through,
3267: the locals stack must not be changed. The compiler generates the
3268: following code:
3269: @format
3270: @code{?branch-lp+!#} <begin> current-locals-size @minus{} dest-locals-size
3271: @end format
3272: The locals stack pointer is only adjusted if the branch is taken.
3273:
3274: @code{THEN} can produce somewhat inefficient code:
3275: @format
3276: @code{lp+!#} current-locals-size @minus{} orig-locals-size
3277: <orig target>:
3278: @code{lp+!#} orig-locals-size @minus{} new-locals-size
3279: @end format
3280: The second @code{lp+!#} adjusts the locals stack pointer from the
3281: level at the @var{orig} point to the level after the @code{THEN}. The
3282: first @code{lp+!#} adjusts the locals stack pointer from the current
3283: level to the level at the orig point, so the complete effect is an
3284: adjustment from the current level to the right level after the
3285: @code{THEN}.
3286:
3287: @cindex locals information on the control-flow stack
3288: @cindex control-flow stack items, locals information
3289: In a conventional Forth implementation a dest control-flow stack entry
3290: is just the target address and an orig entry is just the address to be
1.21 crook 3291: patched. Our locals implementation adds a word list to every orig or dest
1.1 anton 3292: item. It is the list of locals visible (or assumed visible) at the point
3293: described by the entry. Our implementation also adds a tag to identify
3294: the kind of entry, in particular to differentiate between live and dead
3295: (reachable and unreachable) orig entries.
3296:
1.21 crook 3297: A few unusual operations have to be performed on locals word lists:
1.1 anton 3298:
3299: doc-common-list
3300: doc-sub-list?
3301: doc-list-size
3302:
1.21 crook 3303: Several features of our locals word list implementation make these
3304: operations easy to implement: The locals word lists are organised as
1.1 anton 3305: linked lists; the tails of these lists are shared, if the lists
3306: contain some of the same locals; and the address of a name is greater
3307: than the address of the names behind it in the list.
3308:
3309: Another important implementation detail is the variable
3310: @code{dead-code}. It is used by @code{BEGIN} and @code{THEN} to
3311: determine if they can be reached directly or only through the branch
3312: that they resolve. @code{dead-code} is set by @code{UNREACHABLE},
3313: @code{AHEAD}, @code{EXIT} etc., and cleared at the start of a colon
3314: definition, by @code{BEGIN} and usually by @code{THEN}.
3315:
3316: Counted loops are similar to other loops in most respects, but
3317: @code{LEAVE} requires special attention: It performs basically the same
3318: service as @code{AHEAD}, but it does not create a control-flow stack
3319: entry. Therefore the information has to be stored elsewhere;
3320: traditionally, the information was stored in the target fields of the
3321: branches created by the @code{LEAVE}s, by organizing these fields into a
3322: linked list. Unfortunately, this clever trick does not provide enough
3323: space for storing our extended control flow information. Therefore, we
3324: introduce another stack, the leave stack. It contains the control-flow
3325: stack entries for all unresolved @code{LEAVE}s.
3326:
3327: Local names are kept until the end of the colon definition, even if
3328: they are no longer visible in any control-flow path. In a few cases
3329: this may lead to increased space needs for the locals name area, but
3330: usually less than reclaiming this space would cost in code size.
3331:
3332:
3333: @node ANS Forth locals, , Gforth locals, Locals
3334: @subsection ANS Forth locals
3335: @cindex locals, ANS Forth style
3336:
3337: The ANS Forth locals wordset does not define a syntax for locals, but
3338: words that make it possible to define various syntaxes. One of the
3339: possible syntaxes is a subset of the syntax we used in the Gforth locals
3340: wordset, i.e.:
3341:
3342: @example
3343: @{ local1 local2 ... -- comment @}
3344: @end example
3345: or
3346: @example
3347: @{ local1 local2 ... @}
3348: @end example
3349:
3350: The order of the locals corresponds to the order in a stack comment. The
3351: restrictions are:
3352:
3353: @itemize @bullet
3354: @item
3355: Locals can only be cell-sized values (no type specifiers are allowed).
3356: @item
3357: Locals can be defined only outside control structures.
3358: @item
3359: Locals can interfere with explicit usage of the return stack. For the
3360: exact (and long) rules, see the standard. If you don't use return stack
3361: accessing words in a definition using locals, you will be all right. The
3362: purpose of this rule is to make locals implementation on the return
3363: stack easier.
3364: @item
3365: The whole definition must be in one line.
3366: @end itemize
3367:
3368: Locals defined in this way behave like @code{VALUE}s (@xref{Simple
3369: Defining Words}). I.e., they are initialized from the stack. Using their
3370: name produces their value. Their value can be changed using @code{TO}.
3371:
3372: Since this syntax is supported by Gforth directly, you need not do
3373: anything to use it. If you want to port a program using this syntax to
3374: another ANS Forth system, use @file{compat/anslocal.fs} to implement the
3375: syntax on the other system.
3376:
3377: Note that a syntax shown in the standard, section A.13 looks
3378: similar, but is quite different in having the order of locals
3379: reversed. Beware!
3380:
3381: The ANS Forth locals wordset itself consists of the following word
3382:
3383: doc-(local)
3384:
3385: The ANS Forth locals extension wordset defines a syntax, but it is so
3386: awful that we strongly recommend not to use it. We have implemented this
3387: syntax to make porting to Gforth easy, but do not document it here. The
3388: problem with this syntax is that the locals are defined in an order
3389: reversed with respect to the standard stack comment notation, making
3390: programs harder to read, and easier to misread and miswrite. The only
3391: merit of this syntax is that it is easy to implement using the ANS Forth
3392: locals wordset.
3393:
1.21 crook 3394: @node Defining Words, The Text Interpreter, Locals, Words
1.1 anton 3395: @section Defining Words
3396: @cindex defining words
3397:
3398: @menu
3399: * Simple Defining Words::
3400: * Colon Definitions::
3401: * User-defined Defining Words::
3402: * Supplying names::
3403: * Interpretation and Compilation Semantics::
3404: @end menu
3405:
3406: @node Simple Defining Words, Colon Definitions, Defining Words, Defining Words
3407: @subsection Simple Defining Words
3408: @cindex simple defining words
3409: @cindex defining words, simple
3410:
3411: doc-constant
3412: doc-2constant
3413: doc-fconstant
3414: doc-variable
3415: doc-2variable
3416: doc-fvariable
3417: doc-create
3418: doc-user
3419: doc-value
3420: doc-to
3421: doc-defer
3422: doc-is
3423:
1.21 crook 3424: Definitions in ANS Standard Forth for @code{defer}, @code{<is>} and
3425: @code{[is]} are provided in @file{compat/defer.fs}. TODO - what do
3426: the two is words do?
3427:
1.1 anton 3428: @node Colon Definitions, User-defined Defining Words, Simple Defining Words, Defining Words
3429: @subsection Colon Definitions
3430: @cindex colon definitions
3431:
3432: @example
3433: : name ( ... -- ... )
3434: word1 word2 word3 ;
3435: @end example
3436:
3437: creates a word called @code{name}, that, upon execution, executes
3438: @code{word1 word2 word3}. @code{name} is a @dfn{(colon) definition}.
3439:
3440: The explanation above is somewhat superficial. @xref{Interpretation and
3441: Compilation Semantics} for an in-depth discussion of some of the issues
3442: involved.
3443:
3444: doc-:
3445: doc-;
3446:
3447: @node User-defined Defining Words, Supplying names, Colon Definitions, Defining Words
3448: @subsection User-defined Defining Words
3449: @cindex user-defined defining words
3450: @cindex defining words, user-defined
3451:
3452: You can create new defining words simply by wrapping defining-time code
3453: around existing defining words and putting the sequence in a colon
3454: definition.
3455:
1.21 crook 3456: @comment TODO example
3457:
1.1 anton 3458: @cindex @code{CREATE} ... @code{DOES>}
3459: If you want the words defined with your defining words to behave
3460: differently from words defined with standard defining words, you can
3461: write your defining word like this:
3462:
3463: @example
3464: : def-word ( "name" -- )
3465: Create @var{code1}
3466: DOES> ( ... -- ... )
3467: @var{code2} ;
3468:
3469: def-word name
3470: @end example
3471:
3472: Technically, this fragment defines a defining word @code{def-word}, and
3473: a word @code{name}; when you execute @code{name}, the address of the
3474: body of @code{name} is put on the data stack and @var{code2} is executed
3475: (the address of the body of @code{name} is the address @code{HERE}
1.21 crook 3476: returns immediately after the @code{CREATE}). The word @code{name} is
3477: sometimes called a @var{child} of @code{def-word}.
1.1 anton 3478:
3479: In other words, if you make the following definitions:
3480:
3481: @example
3482: : def-word1 ( "name" -- )
3483: Create @var{code1} ;
3484:
3485: : action1 ( ... -- ... )
3486: @var{code2} ;
3487:
3488: def-word name1
3489: @end example
3490:
3491: Using @code{name1 action1} is equivalent to using @code{name}.
3492:
3493: E.g., you can implement @code{Constant} in this way:
3494:
3495: @example
3496: : constant ( w "name" -- )
3497: create ,
3498: DOES> ( -- w )
3499: @@ ;
3500: @end example
3501:
1.21 crook 3502: @comment that is the classic example.. maybe it should be earlier. There
3503: @comment is a beautiful description of how this works and what it does in
3504: @comment the Forthwrite 100th edition.
3505:
1.1 anton 3506: When you create a constant with @code{5 constant five}, first a new word
3507: @code{five} is created, then the value 5 is laid down in the body of
3508: @code{five} with @code{,}. When @code{five} is invoked, the address of
3509: the body is put on the stack, and @code{@@} retrieves the value 5.
3510:
3511: @cindex stack effect of @code{DOES>}-parts
3512: @cindex @code{DOES>}-parts, stack effect
3513: In the example above the stack comment after the @code{DOES>} specifies
3514: the stack effect of the defined words, not the stack effect of the
3515: following code (the following code expects the address of the body on
3516: the top of stack, which is not reflected in the stack comment). This is
3517: the convention that I use and recommend (it clashes a bit with using
3518: locals declarations for stack effect specification, though).
3519:
3520: @subsubsection Applications of @code{CREATE..DOES>}
3521: @cindex @code{CREATE} ... @code{DOES>}, applications
3522:
3523: You may wonder how to use this feature. Here are some usage patterns:
3524:
3525: @cindex factoring similar colon definitions
3526: When you see a sequence of code occurring several times, and you can
3527: identify a meaning, you will factor it out as a colon definition. When
3528: you see similar colon definitions, you can factor them using
3529: @code{CREATE..DOES>}. E.g., an assembler usually defines several words
3530: that look very similar:
3531: @example
3532: : ori, ( reg-target reg-source n -- )
3533: 0 asm-reg-reg-imm ;
3534: : andi, ( reg-target reg-source n -- )
3535: 1 asm-reg-reg-imm ;
3536: @end example
3537:
1.21 crook 3538: @noindent
1.1 anton 3539: This could be factored with:
3540: @example
3541: : reg-reg-imm ( op-code -- )
1.21 crook 3542: CREATE ,
1.1 anton 3543: DOES> ( reg-target reg-source n -- )
3544: @@ asm-reg-reg-imm ;
3545:
3546: 0 reg-reg-imm ori,
3547: 1 reg-reg-imm andi,
3548: @end example
3549:
3550: @cindex currying
3551: Another view of @code{CREATE..DOES>} is to consider it as a crude way to
3552: supply a part of the parameters for a word (known as @dfn{currying} in
3553: the functional language community). E.g., @code{+} needs two
3554: parameters. Creating versions of @code{+} with one parameter fixed can
3555: be done like this:
3556: @example
3557: : curry+ ( n1 -- )
1.21 crook 3558: CREATE ,
1.1 anton 3559: DOES> ( n2 -- n1+n2 )
3560: @@ + ;
3561:
3562: 3 curry+ 3+
3563: -2 curry+ 2-
3564: @end example
3565:
3566: @subsubsection The gory details of @code{CREATE..DOES>}
3567: @cindex @code{CREATE} ... @code{DOES>}, details
3568:
3569: doc-does>
3570:
3571: @cindex @code{DOES>} in a separate definition
3572: This means that you need not use @code{CREATE} and @code{DOES>} in the
1.21 crook 3573: same definition; you can put the @code{DOES>}-part in a separate
1.1 anton 3574: definition. This allows us to, e.g., select among different DOES>-parts:
3575: @example
3576: : does1
3577: DOES> ( ... -- ... )
3578: ... ;
3579:
3580: : does2
3581: DOES> ( ... -- ... )
3582: ... ;
3583:
3584: : def-word ( ... -- ... )
3585: create ...
3586: IF
3587: does1
3588: ELSE
3589: does2
3590: ENDIF ;
3591: @end example
3592:
1.21 crook 3593: In this example, the selection of whether to use @code{does1} or
3594: @code{does2} is made at compile-time; at the time that the child word is
3595: @code{Create}d.
3596:
1.1 anton 3597: @cindex @code{DOES>} in interpretation state
3598: In a standard program you can apply a @code{DOES>}-part only if the last
3599: word was defined with @code{CREATE}. In Gforth, the @code{DOES>}-part
3600: will override the behaviour of the last word defined in any case. In a
3601: standard program, you can use @code{DOES>} only in a colon
3602: definition. In Gforth, you can also use it in interpretation state, in a
3603: kind of one-shot mode:
3604: @example
3605: CREATE name ( ... -- ... )
3606: @var{initialization}
3607: DOES>
3608: @var{code} ;
3609: @end example
3610: This is equivalent to the standard
3611: @example
3612: :noname
3613: DOES>
3614: @var{code} ;
3615: CREATE name EXECUTE ( ... -- ... )
3616: @var{initialization}
3617: @end example
3618:
3619: You can get the address of the body of a word with
3620:
3621: doc->body
3622:
3623: @node Supplying names, Interpretation and Compilation Semantics, User-defined Defining Words, Defining Words
3624: @subsection Supplying names for the defined words
3625: @cindex names for defined words
3626: @cindex defining words, name parameter
3627:
3628: @cindex defining words, name given in a string
3629: By default, defining words take the names for the defined words from the
3630: input stream. Sometimes you want to supply the name from a string. You
1.21 crook 3631: can do this with:
1.1 anton 3632:
3633: doc-nextname
3634:
1.21 crook 3635: For example:
1.1 anton 3636:
3637: @example
3638: s" foo" nextname create
3639: @end example
1.21 crook 3640: @noindent
3641: is equivalent to:
1.1 anton 3642: @example
3643: create foo
3644: @end example
3645:
3646: @cindex defining words without name
1.21 crook 3647: Sometimes you want to define an @var{anonymous word}; a word without a
3648: name. You can do this with:
3649:
3650: doc-:noname
3651:
3652: This leaves the execution token for the word on the stack after the
3653: closing @code{;}. Here's an example in which a deferred word is
3654: initialised with an @code{xt} from an anonymous colon definition:
3655: @example
3656: Defer deferred
3657: :noname ( ... -- ... )
3658: ... ;
3659: IS deferred
3660: @end example
3661:
3662: Gforth provides an alternative way of doing this, using two separate
3663: words:
1.1 anton 3664:
3665: doc-noname
3666: @cindex execution token of last defined word
1.21 crook 3667: doc-lastxt
1.1 anton 3668:
1.21 crook 3669: The previous example can be rewritten using @code{noname} and
3670: @code{lastxt}:
1.1 anton 3671:
3672: @example
3673: Defer deferred
3674: noname : ( ... -- ... )
3675: ... ;
3676: lastxt IS deferred
3677: @end example
3678:
3679: @code{lastxt} also works when the last word was not defined as
3680: @code{noname}.
3681:
3682:
3683: @node Interpretation and Compilation Semantics, , Supplying names, Defining Words
3684: @subsection Interpretation and Compilation Semantics
3685: @cindex semantics, interpretation and compilation
3686:
3687: @cindex interpretation semantics
3688: The @dfn{interpretation semantics} of a word are what the text
3689: interpreter does when it encounters the word in interpret state. It also
3690: appears in some other contexts, e.g., the execution token returned by
3691: @code{' @var{word}} identifies the interpretation semantics of
3692: @var{word} (in other words, @code{' @var{word} execute} is equivalent to
3693: interpret-state text interpretation of @code{@var{word}}).
3694:
3695: @cindex compilation semantics
3696: The @dfn{compilation semantics} of a word are what the text interpreter
3697: does when it encounters the word in compile state. It also appears in
3698: other contexts, e.g, @code{POSTPONE @var{word}} compiles@footnote{In
3699: standard terminology, ``appends to the current definition''.} the
3700: compilation semantics of @var{word}.
3701:
3702: @cindex execution semantics
3703: The standard also talks about @dfn{execution semantics}. They are used
3704: only for defining the interpretation and compilation semantics of many
3705: words. By default, the interpretation semantics of a word are to
3706: @code{execute} its execution semantics, and the compilation semantics of
3707: a word are to @code{compile,} its execution semantics.@footnote{In
3708: standard terminology: The default interpretation semantics are its
3709: execution semantics; the default compilation semantics are to append its
3710: execution semantics to the execution semantics of the current
3711: definition.}
3712:
1.21 crook 3713: @comment TODO expand, make it co-operate with new sections on text interpreter.
3714:
1.1 anton 3715: @cindex immediate words
3716: You can change the compilation semantics into @code{execute}ing the
3717: execution semantics with
3718:
3719: doc-immediate
3720:
3721: @cindex compile-only words
3722: You can remove the interpretation semantics of a word with
3723:
3724: doc-compile-only
3725: doc-restrict
3726:
3727: Note that ticking (@code{'}) compile-only words gives an error
3728: (``Interpreting a compile-only word'').
3729:
3730: Gforth also allows you to define words with arbitrary combinations of
3731: interpretation and compilation semantics.
3732:
3733: doc-interpret/compile:
3734:
3735: This feature was introduced for implementing @code{TO} and @code{S"}. I
3736: recommend that you do not define such words, as cute as they may be:
3737: they make it hard to get at both parts of the word in some contexts.
3738: E.g., assume you want to get an execution token for the compilation
3739: part. Instead, define two words, one that embodies the interpretation
1.15 anton 3740: part, and one that embodies the compilation part. Once you have done
3741: that, you can define a combined word with @code{interpret/compile:} for
3742: the convenience of your users.
1.1 anton 3743:
1.15 anton 3744: You also might try to provide an optimizing implementation of the
3745: default compilation semantics with this feature, like this:
1.1 anton 3746:
3747: @example
3748: :noname
3749: foo bar ;
3750: :noname
3751: POSTPONE foo POSTPONE bar ;
3752: interpret/compile: foobar
3753: @end example
3754:
1.21 crook 3755: @noindent
3756: as an optimizing version of:
1.15 anton 3757:
3758: @example
3759: : foobar
3760: foo bar ;
3761: @end example
3762:
3763: Unfortunately, this does not work correctly with @code{[compile]},
3764: because @code{[compile]} assumes that the compilation semantics of all
3765: @code{interpret/compile:} words are non-default. I.e., @code{[compile]
3766: foobar} would compile the compilation semantics for the optimizing
3767: @code{foobar}, whereas it would compile the interpretation semantics for
3768: the non-optimizing @code{foobar}.
1.1 anton 3769:
3770: @cindex state-smart words are a bad idea
3771: Some people try to use state-smart words to emulate the feature provided
3772: by @code{interpret/compile:} (words are state-smart if they check
3773: @code{STATE} during execution). E.g., they would try to code
3774: @code{foobar} like this:
3775:
3776: @example
3777: : foobar
3778: STATE @@
3779: IF ( compilation state )
3780: POSTPONE foo POSTPONE bar
3781: ELSE
3782: foo bar
3783: ENDIF ; immediate
3784: @end example
3785:
3786: While this works if @code{foobar} is processed only by the text
3787: interpreter, it does not work in other contexts (like @code{'} or
3788: @code{POSTPONE}). E.g., @code{' foobar} will produce an execution token
3789: for a state-smart word, not for the interpretation semantics of the
3790: original @code{foobar}; when you execute this execution token (directly
3791: with @code{EXECUTE} or indirectly through @code{COMPILE,}) in compile
3792: state, the result will not be what you expected (i.e., it will not
3793: perform @code{foo bar}). State-smart words are a bad idea. Simply don't
1.21 crook 3794: write them@footnote{For a more detailed discussion of this topic, see
3795: @cite{@code{State}-smartness -- Why it is Evil and How to Exorcise it} by Anton
3796: Ertl; presented at EuroForth '98 and available from
3797: @url{http://www.complang.tuwien.ac.at/papers/}}!
1.1 anton 3798:
3799: @cindex defining words with arbitrary semantics combinations
3800: It is also possible to write defining words that define words with
1.15 anton 3801: arbitrary combinations of interpretation and compilation semantics. In
3802: general, this looks like:
1.1 anton 3803:
3804: @example
3805: : def-word
3806: create-interpret/compile
3807: @var{code1}
3808: interpretation>
3809: @var{code2}
3810: <interpretation
3811: compilation>
3812: @var{code3}
1.21 crook 3813: <compilation ;
3814: @end example
3815:
3816: For a @var{word} defined with @code{def-word}, the interpretation
3817: semantics are to push the address of the body of @var{word} and perform
3818: @var{code2}, and the compilation semantics are to push the address of
3819: the body of @var{word} and perform @var{code3}. E.g., @code{constant}
3820: can also be defined like this (except that the defined constants don't
3821: behave correctly when @code{[compile]}d):
3822:
3823: @example
3824: : constant ( n "name" -- )
3825: create-interpret/compile
3826: ,
3827: interpretation> ( -- n )
3828: @@
3829: <interpretation
3830: compilation> ( compilation. -- ; run-time. -- n )
3831: @@ postpone literal
3832: <compilation ;
3833: @end example
3834:
3835: doc-create-interpret/compile
3836: doc-interpretation>
3837: doc-<interpretation
3838: doc-compilation>
3839: doc-<compilation
3840:
3841: Note that words defined with @code{interpret/compile:} and
3842: @code{create-interpret/compile} have an extended header structure that
3843: differs from other words; however, unless you try to access them with
3844: plain address arithmetic, you should not notice this. Words for
3845: accessing the header structure usually know how to deal with this; e.g.,
3846: @code{' word >body} also gives you the body of a word created with
3847: @code{create-interpret/compile}.
3848:
3849: @c ----------------------------------------------------------
3850: @node The Text Interpreter, Structures, Defining Words, Words
3851: @section The Text Interpreter
3852: @cindex interpreter - outer
3853: @cindex text interpreter
3854: @cindex outer interpreter
3855:
3856: Blah blah.
3857:
3858: doc->in
3859:
3860:
3861: @menu
3862: * Number Conversion::
3863: * Interpret/Compile states::
3864: * Literals::
3865: * Interpreter Directives::
3866: @end menu
3867:
3868:
3869:
3870: invoking it now, by typing @kbd{gforth<return>}). Forth is now running
3871: its command line interpreter, which is called the "Text Interpreter"
3872: (also known as the "Outer Interpreter"). The behaviour of the text
3873: interpreter depends upon whether the system is in "Interpret" or
3874: "Compile" state. At startup, the system is always in "Interpret" state.
3875:
3876:
3877: Behaviour of the text interpreter in "Interpret" state
3878: ------------------------------------------------------
3879:
3880: Although it may not be obvious, Forth is actually prompting you for
3881: input. Type a number and press the <return> key:
3882:
3883: 45<return> ok
3884:
3885: Rather than give you a prompt to invite you to input something, the
3886: text interpreter prints a status message *after* it has processed a
3887: line of input. The status message in this case (" ok" followed by
3888: carriage-return) indicates that the text interpreter was able to
3889: process all of your input successfully. Now type something illegal:
3890:
3891: qwer341<return>
3892: ^^^^^^^
3893: Error: Undefined word
3894:
3895: When the text interpreter detects an error, it discards any remaining
3896: text on a line, resets certain internal state (including returning to
3897: "Interpret" state) and prints an error message.
3898:
3899: The text interpreter works on input one line at a time. Starting at
3900: the beginning of the line, it skips leading spaces (called
3901: "delimiters") then parses a string (a sequence of non-space
3902: characters) until it either reaches a space character or it
3903: reaches the end of the line. Having parsed a string, it then makes two
3904: attempts to do something with it:
3905:
3906: * It looks the string up in a dictionary of definitions. If the string
3907: is found in the dictionary, the string names a "definition" (also
3908: known as a "word") and the dictionary search will return an
3909: "Execution token" (xt) for the definition and some flags that show
3910: when the definition can be used legally. If the definition can be
3911: legally executed in "Interpret" mode then the text interpreter will
3912: use the xt to execute it, otherwise it will issue an error
3913: message. The dictionary is described in more detail in <blah>.
3914:
3915: * If the string is not found in the dictionary, the text interpreter
3916: attempts to treat it as a number in the current radix (base 10 after
3917: initial startup). If the string represents a legal number in the
3918: current radix, the number is pushed onto the appropriate parameter
3919: stack. Stacks are discussed in more detail in <blah>. Number
3920: conversion is described in more detail in <section about +, -
3921: numbers and different number formats>.
3922:
3923: If both of these attempts fail, the remainer of the input line is
3924: discarded and the text interpreter isses an error message. If one of
3925: these attempts succeeds, the text interpreter repeats the parsing
3926: process until the end of the line has been reached. At this point,
3927: it prints the status message " ok" and waits for more input.
3928:
3929: There are two important things to note about the behaviour of the text
3930: interpreter:
3931:
3932: * it processes each input string to completion before parsing
3933: additional characters from the input line.
3934:
3935: * it keeps track of its position in the input line using a variable
3936: (called >IN, pronounced "to-in"). The value of >IN can be modified
3937: by the execution of definitions in the input line. This means that
3938: definitions can "trick" the text interpreter either into skipping
3939: sections of the input line or into parsing a section of the
3940: input line more than once.
3941:
3942:
3943: Stacks, postfix notation and parameter passing
3944: ----------------------------------------------
3945:
3946: In procedural programming languages (like C and Pascal), the
3947: building-block of programs is the function or procedure. These
3948: functions or procedures are called with explicit parameters. For
3949: example, in C we might write:
3950:
3951: total = total + new_volume(length,height,depth);
3952:
3953: where total, length, height, depth are all variables and new_volume is
3954: a function-call to another piece of code.
3955:
3956: In Forth, the equivalent to the function or procedure is the
3957: "definition" and parameters are implicitly passed between definitions
3958: using a shared stack that is visible to the programmer. Although Forth
3959: does support variables, the existence of the stack means that they are
3960: used far less often than in most other programming languages. When the
3961: text interpreter encounters a number, it will place it on the
3962: stack. There are several stacks (the actual number is
3963: implementation-dependent ..) and the particular stack used for any
3964: operation is implied unambiguously by the operation being
3965: performed. The stack used for all integer operations is called the
3966: "data stack", and since this is the stack used most commonly,
3967: references to "the data stack" are often abbreviated to "the stack".
3968:
3969: The stacks have a LIFO (last-in, first-out) organisation. If you type:
3970:
3971: 1 2 3<return> ok
3972:
3973: then you have placed three numbers on the (data) stack. An analogy for
3974: the behaviour of the stack is to take a pack of playing cards and deal
3975: out the ace (1), 2 and 3 into a pile on the table. The 3 was the last
3976: card onto the pile ("last-in") and if you take a card off the pile
3977: then, unless you're prepared to fiddle a bit, the card that you take
3978: off will be the 3 ("first-out"). The number that will be first-out of
3979: the stack is called the "top of stack", which is often abbreviated to
3980: TOS.
3981:
3982: To see how parameters are passed in Forth, we will consider the
3983: behaviour of the definition "+" (pronounced "plus"). You will not be
3984: surprised to learn that this definition performs addition. More
3985: precisely, it adds two number together and produces a result. Where
3986: does it get the two numbers from? It takes the first two numbers off
3987: the stack. Where does it place the result? On the stack. To continue
3988: with the playing-cards analogy, you can perform the behaviour of "+"
3989: like this:
3990:
3991: - pick up two cards from the stack
3992: - stare at them intently and ask yourself "what *is* the sum of these
3993: two numbers"
3994: - decide that the answer is 5
3995: - shuffle the two cards back into the pack and find a 5
3996: - put a 5 on the remaining ace that's on the table.
3997:
3998: If you don't have a pack of cards handy but you do have Forth running,
3999: you can use the definition .s to show the current state of the stack,
4000: without affecting the stack. If you already typed "1 2 3" then you
4001: should see:
4002:
4003: .s<return> <3> 1 2 3 ok
4004:
4005: The "<3>" is the total number of items on the stack, and the item on
4006: the far right-hand side is the TOS. You can now type:
4007:
4008: + .s<return> <2> 1 5 ok
4009:
4010: which is correct; there are now 2 items on the stack and the result of
4011: the addition is 5.
4012:
4013: If you're playing with cards, try doing a second addition; pick up the
4014: two cards, work out that their sum is 6, shuffle them into the pack,
4015: look for a 6 and place that on the table. You now have just one item
4016: on the stack. What happens if you try to do a third addition? Pick up
4017: the first card, pick up the second card - ah. There is no second
4018: card. This is called a "stack underflow" and consitutes an error. If
4019: you try to do the same thing with Forth it will report an error
4020: (probably a Stack Underflow or an Invalid Memory Address error).
4021:
4022: The opposite situation to a stack underflow is a stack overflow, which
4023: simply accepts that there is a finite amount of storage space reserved
4024: for the stack. To stretch the playing card analogy, if you had enough
4025: packs of cards and you piled the cards up on the table, you would
4026: eventually be unable to add another card; you'd hit the
4027: ceiling. Gforth allows you to set the maximum size of the stacks. In
4028: general, the only time that you will get a stack overflow is because a
4029: definition has a bug in it and is generating data on the stack
4030: uncontrollably.
4031:
4032: There's one final use for the playing card analogy. If you model your
4033: stack using a pack of playing cards, the maximum number of items on
4034: your stack will be 52 (I assume you didn't use the Joker). The maximum
4035: *value* of any item on the stack is 13 (the King). In fact, the only
4036: possible numbers are positive integer numbers 1 through 13; you can't
4037: have (for example) 0 or 27 or 3.52 or -2. If you change the way you
4038: think about some of the cards, you can accommodate different
4039: numbers. For example, you could think of the Jack as representing 0,
4040: the Queen as representing -1 and the King as representing -2. Your
4041: *range* remains unchanged (you can still only represent a total of 13
4042: numbers) but the numbers that you can represent are -2 through 10.
4043:
4044: In that analogy, the limit was the amount of information that a single
4045: stack entry could hold, and Forth has a similar limit. In Forth, the
4046: size of a stack entry is called a "cell". The actual size of a cell is
4047: implementation dependent and affects the maximum value that a stack
4048: entry can hold. A Standard Forth provides a cell size of at least
4049: 16-bits, and most desktop systems use a cell size of 32-bits.
4050:
4051: Forth does not do any type checking for you, so you are free to
4052: manipulate and combine stack items in any way you wish. A convenient
4053: ways of treating stack items is as 2's complement signed integers, and
4054: that is what Standard words like "+" do. Therefore you can type:
4055:
4056: -5 12 + .s<return> <1> 7 ok
4057:
4058: If you use numbers and definitions like "+" in order to turn Forth
4059: into a great big pocket calculator, you will realise that it's rather
4060: different from a normal calculator. Rather than typing 2 + 3 = you had
4061: to type 2 3 + (ignore the fact that you had to use .s to see the
4062: result). The terminology used to describe this difference is to say
4063: that your calculator uses "Infix Notation" (parameters and operators
4064: are mixed) whilst Forth uses "Postfix Notation" (parameters and
4065: operators are separate), also called "Reverse Polish Notation".
4066:
4067: Whilst postfix notation might look confusing to begin with, it has
4068: several important advantages:
4069:
4070: - it is unambiguous
4071: - it is more concise
4072: - it fits naturally with a stack-based system
4073:
4074: To examine these claims in more detail, consider these sums:
4075:
4076: 6 + 5 * 4 =
4077: 4 * 5 + 6 =
4078:
4079: If you're just learning maths or your maths is very rusty, you will
4080: probably come up with the answer 44 for the first and 26 for the
4081: second. If you are a bit of a whizz at maths you will remember the
4082: *convention* that multiplication takes precendence over addition, and
4083: you'd come up with the answer 26 both times. To explain the answer 26
4084: to someone who got the answer 44, you'd probably rewrite the first sum
4085: like this:
4086:
4087: 6 + (5 * 4) =
4088:
4089: If what you really wanted was to perform the addition before the
4090: multiplication, you would have to use parentheses to force it.
4091:
4092: If you did the first two sums on a pocket calculator you would probably
4093: get the right answers, unless you were very cautious and entered them using
4094: these keystroke sequences:
4095:
4096: 6 + 5 = * 4 =
4097: 4 * 5 = + 6 =
4098:
4099: Postfix notation is unambiguous because the order that the operators
4100: are applied is always explicit; that also means that parentheses are
4101: never required. The operators are *active* (the act of quoting the
4102: operator makes the operation occur) which removes the need for "=".
4103:
4104: The sum 6 + 5 * 4 can be written (in postfix notation) in two
4105: equivalent ways:
4106:
4107: 6 5 4 * + or:
4108: 5 4 * 6 +
4109:
4110: TODO point out that the order of number is never changed.
4111:
4112: The Structure Of Programs In Forth
4113: ----------------------------------
4114:
4115: When you start up the Forth compiler, a large number of definitions
4116: already exist. To develop a new application, use bottom-up programming
4117: techniques to create new definitions that are defined in terms of
4118: existing definitions. As you create each definition you can test it
4119: interactively. Ultimately, you end up with an environment <blah blah>
4120:
4121: Creating new definitions
4122: ------------------------
4123:
4124: The easiest way to create a new definition is to use a "colon
4125: definition". In order to provide a few examples (and give you some
4126: homework) I'm going to introduce a very small set of words but only
4127: describe what they do very informally, by example.
4128:
4129: + add the top two numbers on the stack and place the result on the
4130: stack
4131: . print the top stack item
4132: ." print text until a " delimiter is found
4133: CR print a carriage-return
4134: : start a new definition
4135: ; end a definition
4136: DUP blah
4137: DROP blah
4138:
4139: example 1:
4140: : greet ." Hello and welcome" ;<return> ok
4141: greet<return> Hello and welcome ok
4142: greet greet<return> Hello and welcomeHello and welcome ok
4143:
4144: When you try out this example, be careful to copy the spaces
4145: accurately; there needs to be a space between each group of characters
4146: that will be processed by the text interpreter.
4147:
4148:
4149: example 2:
4150: : add-two 2 + . ;<return> ok
4151: 5 add-two<return> 7 ok
4152:
4153:
4154: - numbers and definitions
4155: - redefining things .. what uses the old defn and what uses the new one
4156: - boundary between system definitions and your definitions
4157: - standards.. a double-edged sword
4158: - philosophy
4159:
4160: - your first set of definitions
4161:
4162:
4163:
4164: .. interactive stuff
4165: 5 3 + . <return> 8 ok
4166:
4167: could have been split over several lines
4168:
4169: 5 . . <return>
4170:
4171: - cells and chars
4172:
4173: - the text interpreter in "Compilation" state.
4174:
4175: -- elements of a forth system
4176: - text interpreter (outer interpreter)
4177: - compiler
4178: - inner interpreter
4179: - dictionaries and wordlists
4180: - stacks
4181:
4182: -- disparate spaces .. may be better to describe that elsewhere.
4183:
4184:
4185:
4186: @node Number Conversion, Interpret/Compile states, The Text Interpreter, The Text Interpreter
4187: @subsection Number Conversion
4188: @cindex Number conversion
4189: @cindex double-cell numbers, input format
4190: @cindex input format for double-cell numbers
4191: @cindex single-cell numbers, input format
4192: @cindex input format for single-cell numbers
4193: @cindex floating-point numbers, input format
4194: @cindex input format for floating-point numbers
4195:
4196: If the text interpreter fails to find a particular string in the name
4197: dictionary, it attempts to convert it to a number using a set of rules.
4198:
4199: Let <digit> represent any character that is a legal digit in the current
4200: number base (for example, 0-9 when the number base is decimal or 0-9, A-F
4201: when the number base is hexadecimal).
4202:
4203: Let <decimal digit> represent any character in the range 0-9.
4204:
4205: @comment TODO need to extend the next defn to support fp format
4206: Let @{+ | -@} represent the optional presence of either a @code{+} or
4207: @code{-} character.
4208:
4209: Let * represent any number of instances of the previous character
4210: (including none).
4211:
4212: Let any other character represent itself.
4213:
4214: Now, the conversion rules are:
4215:
4216: @itemize @bullet
4217: @item
4218: A string of the form <digit><digit>* is treated as a single-precision
4219: (CELL-sized) positive integer. Examples are 0 123 6784532 32343212343456 42
4220: @item
4221: A string of the form -<digit><digit>* is treated as a single-precision
4222: (CELL-sized) negative integer, and is represented using 2's-complement
4223: arithmetic. Examples are -45 -5681 -0
4224: @item
4225: A string of the form <digit><digit>*.<digit>* is treated as a double-precision
4226: (double-CELL-sized) positive integer. Examples are 3465. 3.465 34.65
4227: (and note that these all represent the same number).
4228: @item
4229: A string of the form -<digit><digit>*.<digit>* is treated as a
4230: double-precision (double-CELL-sized) negative integer, and is
4231: represented using 2's-complement arithmetic. Examples are -3465. -3.465
4232: -34.65 (and note that these all represent the same number).
4233: @item
4234: A string of the form @{+ | -@}<decimal digit>@{.@}<decimal digit>*@{e | E@}@{+
4235: | -@}<decimal digit><decimal digit>* is treated as floating-point
4236: number. Examples are 1e0 1.e 1.e0 +1e+0 (which all represent the same
4237: number) +12.E-4
4238: @end itemize
4239:
4240: By default, the number base used for integer number conversion is given
4241: by the contents of a variable named @code{BASE}. Base 10 (decimal) is
4242: always used for floating-point number conversion.
4243:
4244: doc-base
4245: doc-hex
4246: doc-decimal
4247:
4248: @cindex '-prefix for character strings
4249: @cindex &-prefix for decimal numbers
4250: @cindex %-prefix for binary numbers
4251: @cindex $-prefix for hexadecimal numbers
4252: Gforth allows you to override the value of @code{BASE} by using a prefix
4253: before the first digit of an (integer) number. Four prefixes are
4254: supported:
4255:
4256: @itemize @bullet
4257: @item
4258: @code{&} -- decimal number
4259: @item
4260: @code{%} -- binary number
4261: @item
4262: @code{$} -- hexadecimal number
4263: @item
4264: @code{'} -- base 256 number
4265: @end itemize
4266:
4267: Here are some examples, with the equivalent decimal number shown after
4268: in braces:
4269:
4270: -$41 (-65) %1001101 (205) %1001.0001 (145 - a double-precision number)
4271: 'AB (16706; ascii A is 65, ascii B is 66, number is 65*256 + 66)
4272: 'ab (24930; ascii a is 97, ascii B is 98, number is 97*256 + 98)
4273: &905 (905) $abc (2478) $ABC (2478)
4274:
4275: @cindex Number conversion - traps for the unwary
4276: Number conversion has a number of traps for the unwary:
4277:
4278: @itemize @bullet
4279: @item
4280: You cannot determine the current number base using the code sequence
4281: @code{BASE @@ .} -- the number base is always 10 in the current number
4282: base. Instead, use something like @code{BASE @@ DECIMAL DUP . BASE !}
4283: @item
4284: If the number base is set to a value greater than 14 (for example,
4285: hexadecimal), the number 123E4 is ambiguous; the conversion rules allow
4286: it to be intepreted as either a single-precision integer or a
4287: floating-point number (Gforth treats it as an integer). The ambiguity
4288: can be resolved by explicitly stating the sign of the mantissa and/or
4289: exponent: 123E+4 or +123E4 -- if the number base is decimal, no
4290: ambiguity arises; either representation will be treated as a
4291: floating-point number.
4292: @item
4293: There is a word @code{bin} but it does @var{not} set the number base!
4294: It is used to specify file types.
4295: @item
4296: ANS Forth Standard requires the @code{.} of a double-precision number to
4297: be the final character in the string. Allowing the @code{.} to be
4298: anywhere after the first digit is a Gforth extension.
4299: @item
4300: The number conversion process does not check for overflow.
4301: @item
4302: In Gforth, number conversion to floating-point numbers always use base
4303: 10, irrespective of the value of @code{BASE}. For the ANS Forth
4304: Standard, conversion to floating-point numbers whilst the value of
4305: @code{BASE} is not 10 is an ambiguous condition.
4306: @end itemize
4307:
4308:
4309: @node Interpret/Compile states, Literals, Number Conversion, The Text Interpreter
4310: @subsection Interpret/Compile states
4311: @cindex Interpret/Compile states
4312:
4313: Blah
4314:
4315: doc-state
4316: doc-[
4317: doc-]
1.1 anton 4318:
4319:
4320:
1.21 crook 4321: @node Literals, Interpreter Directives, Interpret/Compile states, The Text Interpreter
4322: @subsection Literals
4323: @cindex Literals
4324:
4325: Blah blah
4326:
4327: doc-literal
4328: doc-2literal
4329: doc-fliteral
4330:
4331: @node Interpreter Directives, ,Literals, The Text Interpreter
4332: @subsection Interpreter Directives
4333: @cindex Interpreter Directives
4334:
4335: These words are usually used outside of definitions; for example, to
4336: control which parts of a source file are processed by the text
4337: interpreter. There are only a few ANS Forth Standard words, but Gforth
4338: supplements these with a rich set of immediate control structure words
4339: to compensate for the fact that the non-immediate versions can only be
4340: used in compile state (@pxref{Control Structures}).
4341:
4342: doc-[IF]
4343: doc-[ELSE]
4344: doc-[THEN]
4345: doc-[ENDIF]
4346:
4347: doc-[IFDEF]
4348: doc-[IFUNDEF]
4349:
4350: doc-[?DO]
4351: doc-[DO]
4352: doc-[FOR]
4353: doc-[LOOP]
4354: doc-[+LOOP]
4355: doc-[NEXT]
4356:
4357: doc-[BEGIN]
4358: doc-[UNTIL]
4359: doc-[AGAIN]
4360: doc-[WHILE]
4361: doc-[REPEAT]
1.1 anton 4362:
4363:
1.5 anton 4364: @c ----------------------------------------------------------
1.21 crook 4365: @node Structures, Object-oriented Forth, The Text Interpreter, Words
1.5 anton 4366: @section Structures
4367: @cindex structures
4368: @cindex records
4369:
4370: This section presents the structure package that comes with Gforth. A
1.21 crook 4371: version of the package implemented in ANS Standard Forth is available in
1.5 anton 4372: @file{compat/struct.fs}. This package was inspired by a posting on
4373: comp.lang.forth in 1989 (unfortunately I don't remember, by whom;
4374: possibly John Hayes). A version of this section has been published in
4375: ???. Marcel Hendrix provided helpful comments.
4376:
4377: @menu
4378: * Why explicit structure support?::
4379: * Structure Usage::
4380: * Structure Naming Convention::
4381: * Structure Implementation::
4382: * Structure Glossary::
4383: @end menu
4384:
4385: @node Why explicit structure support?, Structure Usage, Structures, Structures
4386: @subsection Why explicit structure support?
4387:
4388: @cindex address arithmetic for structures
4389: @cindex structures using address arithmetic
4390: If we want to use a structure containing several fields, we could simply
4391: reserve memory for it, and access the fields using address arithmetic
4392: (@pxref{Address arithmetic}). As an example, consider a structure with
4393: the following fields
4394:
4395: @table @code
4396: @item a
4397: is a float
4398: @item b
4399: is a cell
4400: @item c
4401: is a float
4402: @end table
4403:
4404: Given the (float-aligned) base address of the structure we get the
4405: address of the field
4406:
4407: @table @code
4408: @item a
4409: without doing anything further.
4410: @item b
4411: with @code{float+}
4412: @item c
4413: with @code{float+ cell+ faligned}
4414: @end table
4415:
4416: It is easy to see that this can become quite tiring.
4417:
4418: Moreover, it is not very readable, because seeing a
4419: @code{cell+} tells us neither which kind of structure is
4420: accessed nor what field is accessed; we have to somehow infer the kind
4421: of structure, and then look up in the documentation, which field of
4422: that structure corresponds to that offset.
4423:
4424: Finally, this kind of address arithmetic also causes maintenance
4425: troubles: If you add or delete a field somewhere in the middle of the
4426: structure, you have to find and change all computations for the fields
4427: afterwards.
4428:
4429: So, instead of using @code{cell+} and friends directly, how
4430: about storing the offsets in constants:
4431:
4432: @example
4433: 0 constant a-offset
4434: 0 float+ constant b-offset
4435: 0 float+ cell+ faligned c-offset
4436: @end example
4437:
4438: Now we can get the address of field @code{x} with @code{x-offset
4439: +}. This is much better in all respects. Of course, you still
4440: have to change all later offset definitions if you add a field. You can
4441: fix this by declaring the offsets in the following way:
4442:
4443: @example
4444: 0 constant a-offset
4445: a-offset float+ constant b-offset
4446: b-offset cell+ faligned constant c-offset
4447: @end example
4448:
4449: Since we always use the offsets with @code{+}, using a defining
4450: word @code{cfield} that includes the @code{+} in the
4451: action of the defined word offers itself:
4452:
4453: @example
4454: : cfield ( n "name" -- )
4455: create ,
4456: does> ( name execution: addr1 -- addr2 )
4457: @@ + ;
4458:
4459: 0 cfield a
4460: 0 a float+ cfield b
4461: 0 b cell+ faligned cfield c
4462: @end example
4463:
4464: Instead of @code{x-offset +}, we now simply write @code{x}.
4465:
4466: The structure field words now can be used quite nicely. However,
4467: their definition is still a bit cumbersome: We have to repeat the
4468: name, the information about size and alignment is distributed before
4469: and after the field definitions etc. The structure package presented
4470: here addresses these problems.
4471:
4472: @node Structure Usage, Structure Naming Convention, Why explicit structure support?, Structures
4473: @subsection Structure Usage
4474: @cindex structure usage
4475:
4476: @cindex @code{field} usage
4477: @cindex @code{struct} usage
4478: @cindex @code{end-struct} usage
4479: You can define a structure for a (data-less) linked list with
4480: @example
4481: struct
4482: cell% field list-next
4483: end-struct list%
4484: @end example
4485:
4486: With the address of the list node on the stack, you can compute the
4487: address of the field that contains the address of the next node with
4488: @code{list-next}. E.g., you can determine the length of a list
4489: with:
4490:
4491: @example
4492: : list-length ( list -- n )
4493: \ "list" is a pointer to the first element of a linked list
4494: \ "n" is the length of the list
4495: 0 begin ( list1 n1 )
4496: over
4497: while ( list1 n1 )
4498: 1+ swap list-next @@ swap
4499: repeat
4500: nip ;
4501: @end example
4502:
4503: You can reserve memory for a list node in the dictionary with
4504: @code{list% %allot}, which leaves the address of the list node on the
4505: stack. For the equivalent allocation on the heap you can use @code{list%
4506: %alloc} (or, for an @code{allocate}-like stack effect (i.e., with ior),
4507: use @code{list% %allocate}). You can also get the the size of a list
4508: node with @code{list% %size} and it's alignment with @code{list%
4509: %alignment}.
4510:
4511: Note that in ANS Forth the body of a @code{create}d word is
4512: @code{aligned} but not necessarily @code{faligned};
4513: therefore, if you do a
4514: @example
4515: create @emph{name} foo% %allot
4516: @end example
4517:
4518: then the memory alloted for @code{foo%} is
4519: guaranteed to start at the body of @code{@emph{name}} only if
4520: @code{foo%} contains only character, cell and double fields.
4521:
4522: @cindex strcutures containing structures
4523: You can also include a structure @code{foo%} as field of
4524: another structure, with:
4525: @example
4526: struct
4527: ...
4528: foo% field ...
4529: ...
4530: end-struct ...
4531: @end example
4532:
4533: @cindex structure extension
4534: @cindex extended records
4535: Instead of starting with an empty structure, you can also extend an
4536: existing structure. E.g., a plain linked list without data, as defined
4537: above, is hardly useful; You can extend it to a linked list of integers,
4538: like this:@footnote{This feature is also known as @emph{extended
4539: records}. It is the main innovation in the Oberon language; in other
4540: words, adding this feature to Modula-2 led Wirth to create a new
4541: language, write a new compiler etc. Adding this feature to Forth just
4542: requires a few lines of code.}
4543:
4544: @example
4545: list%
4546: cell% field intlist-int
4547: end-struct intlist%
4548: @end example
4549:
4550: @code{intlist%} is a structure with two fields:
4551: @code{list-next} and @code{intlist-int}.
4552:
4553: @cindex structures containing arrays
4554: You can specify an array type containing @emph{n} elements of
4555: type @code{foo%} like this:
4556:
4557: @example
4558: foo% @emph{n} *
4559: @end example
4560:
4561: You can use this array type in any place where you can use a normal
4562: type, e.g., when defining a @code{field}, or with
4563: @code{%allot}.
4564:
4565: @cindex first field optimization
4566: The first field is at the base address of a structure and the word
4567: for this field (e.g., @code{list-next}) actually does not change
4568: the address on the stack. You may be tempted to leave it away in the
4569: interest of run-time and space efficiency. This is not necessary,
4570: because the structure package optimizes this case and compiling such
4571: words does not generate any code. So, in the interest of readability
4572: and maintainability you should include the word for the field when
4573: accessing the field.
4574:
4575: @node Structure Naming Convention, Structure Implementation, Structure Usage, Structures
4576: @subsection Structure Naming Convention
4577: @cindex structure naming conventions
4578:
4579: The field names that come to (my) mind are often quite generic, and,
4580: if used, would cause frequent name clashes. E.g., many structures
4581: probably contain a @code{counter} field. The structure names
4582: that come to (my) mind are often also the logical choice for the names
4583: of words that create such a structure.
4584:
4585: Therefore, I have adopted the following naming conventions:
4586:
4587: @itemize @bullet
4588: @cindex field naming convention
4589: @item
4590: The names of fields are of the form
4591: @code{@emph{struct}-@emph{field}}, where
4592: @code{@emph{struct}} is the basic name of the structure, and
4593: @code{@emph{field}} is the basic name of the field. You can
4594: think about field words as converting converts the (address of the)
4595: structure into the (address of the) field.
4596:
4597: @cindex structure naming convention
4598: @item
4599: The names of structures are of the form
4600: @code{@emph{struct}%}, where
4601: @code{@emph{struct}} is the basic name of the structure.
4602: @end itemize
4603:
4604: This naming convention does not work that well for fields of extended
4605: structures; e.g., the integer list structure has a field
4606: @code{intlist-int}, but has @code{list-next}, not
4607: @code{intlist-next}.
4608:
4609: @node Structure Implementation, Structure Glossary, Structure Naming Convention, Structures
4610: @subsection Structure Implementation
4611: @cindex structure implementation
4612: @cindex implementation of structures
4613:
4614: The central idea in the implementation is to pass the data about the
4615: structure being built on the stack, not in some global
4616: variable. Everything else falls into place naturally once this design
4617: decision is made.
4618:
4619: The type description on the stack is of the form @emph{align
4620: size}. Keeping the size on the top-of-stack makes dealing with arrays
4621: very simple.
4622:
1.21 crook 4623: @code{field} is a defining word that uses @code{Create}
4624: and @code{DOES>}. The body of the field contains the offset
4625: of the field, and the normal @code{DOES>} action is:
1.5 anton 4626:
4627: @example
4628: @ +
4629: @end example
4630:
1.21 crook 4631: @noindent
1.5 anton 4632: i.e., add the offset to the address, giving the stack effect
4633: @code{addr1 -- addr2} for a field.
4634:
4635: @cindex first field optimization, implementation
4636: This simple structure is slightly complicated by the optimization
4637: for fields with offset 0, which requires a different
1.21 crook 4638: @code{DOES>}-part (because we cannot rely on there being
1.5 anton 4639: something on the stack if such a field is invoked during
1.21 crook 4640: compilation). Therefore, we put the different @code{DOES>}-parts
1.5 anton 4641: in separate words, and decide which one to invoke based on the
4642: offset. For a zero offset, the field is basically a noop; it is
4643: immediate, and therefore no code is generated when it is compiled.
4644:
4645: @node Structure Glossary, , Structure Implementation, Structures
4646: @subsection Structure Glossary
4647: @cindex structure glossary
4648:
4649: doc-%align
4650: doc-%alignment
4651: doc-%alloc
4652: doc-%allocate
4653: doc-%allot
4654: doc-cell%
4655: doc-char%
4656: doc-dfloat%
4657: doc-double%
4658: doc-end-struct
4659: doc-field
4660: doc-float%
4661: doc-nalign
4662: doc-sfloat%
4663: doc-%size
4664: doc-struct
4665:
4666: @c -------------------------------------------------------------
1.12 anton 4667: @node Object-oriented Forth, Tokens for Words, Structures, Words
4668: @section Object-oriented Forth
4669:
4670: Gforth comes with three packets for object-oriented programming,
4671: @file{objects.fs}, @file{oof.fs}, and @file{mini-oof.fs}; none of them
4672: is preloaded, so you have to @code{include} them before use. The most
4673: important differences between these packets (and others) are discussed
4674: in @ref{Comparison with other object models}. All packets are written
4675: in ANS Forth and can be used with any other ANS Forth.
4676:
4677: @menu
4678: * Objects::
4679: * OOF::
4680: * Mini-OOF::
4681: @end menu
4682:
4683: @node Objects, OOF, Object-oriented Forth, Object-oriented Forth
4684: @subsection Objects
1.5 anton 4685: @cindex objects
4686: @cindex object-oriented programming
4687:
4688: @cindex @file{objects.fs}
4689: @cindex @file{oof.fs}
1.12 anton 4690:
4691: 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 4692: @c McKewan's and Zsoter's packages
4693:
4694: This section assumes (in some places) that you have read @ref{Structures}.
4695:
4696: @menu
4697: * Properties of the Objects model::
4698: * Why object-oriented programming?::
4699: * Object-Oriented Terminology::
4700: * Basic Objects Usage::
4701: * The class Object::
4702: * Creating objects::
4703: * Object-Oriented Programming Style::
4704: * Class Binding::
4705: * Method conveniences::
4706: * Classes and Scoping::
4707: * Object Interfaces::
4708: * Objects Implementation::
4709: * Comparison with other object models::
4710: * Objects Glossary::
4711: @end menu
4712:
4713: Marcel Hendrix provided helpful comments on this section. Andras Zsoter
4714: and Bernd Paysan helped me with the related works section.
4715:
4716: @node Properties of the Objects model, Why object-oriented programming?, Objects, Objects
1.12 anton 4717: @subsubsection Properties of the @file{objects.fs} model
1.5 anton 4718: @cindex @file{objects.fs} properties
4719:
4720: @itemize @bullet
4721: @item
4722: It is straightforward to pass objects on the stack. Passing
4723: selectors on the stack is a little less convenient, but possible.
4724:
4725: @item
4726: Objects are just data structures in memory, and are referenced by
4727: their address. You can create words for objects with normal defining
4728: words like @code{constant}. Likewise, there is no difference
4729: between instance variables that contain objects and those
4730: that contain other data.
4731:
4732: @item
4733: Late binding is efficient and easy to use.
4734:
4735: @item
4736: It avoids parsing, and thus avoids problems with state-smartness
4737: and reduced extensibility; for convenience there are a few parsing
4738: words, but they have non-parsing counterparts. There are also a few
4739: defining words that parse. This is hard to avoid, because all standard
4740: defining words parse (except @code{:noname}); however, such
4741: words are not as bad as many other parsing words, because they are not
4742: state-smart.
4743:
4744: @item
4745: It does not try to incorporate everything. It does a few things
4746: and does them well (IMO). In particular, I did not intend to support
4747: information hiding with this model (although it has features that may
4748: help); you can use a separate package for achieving this.
4749:
4750: @item
4751: It is layered; you don't have to learn and use all features to use this
4752: model. Only a few features are necessary (@xref{Basic Objects Usage},
4753: @xref{The class Object}, @xref{Creating objects}.), the others
4754: are optional and independent of each other.
4755:
4756: @item
4757: An implementation in ANS Forth is available.
4758:
4759: @end itemize
4760:
4761: I have used the technique, on which this model is based, for
4762: implementing the parser generator Gray; we have also used this technique
1.21 crook 4763: in Gforth for implementing the various flavours of word lists (hashed or
4764: not, case-sensitive or not, special-purpose word lists for locals etc.).
1.5 anton 4765:
4766: @node Why object-oriented programming?, Object-Oriented Terminology, Properties of the Objects model, Objects
1.12 anton 4767: @subsubsection Why object-oriented programming?
1.5 anton 4768: @cindex object-oriented programming motivation
4769: @cindex motivation for object-oriented programming
4770:
4771: Often we have to deal with several data structures (@emph{objects}),
4772: that have to be treated similarly in some respects, but differ in
4773: others. Graphical objects are the textbook example: circles,
4774: triangles, dinosaurs, icons, and others, and we may want to add more
4775: during program development. We want to apply some operations to any
4776: graphical object, e.g., @code{draw} for displaying it on the
4777: screen. However, @code{draw} has to do something different for
4778: every kind of object.
4779:
4780: We could implement @code{draw} as a big @code{CASE}
4781: control structure that executes the appropriate code depending on the
4782: kind of object to be drawn. This would be not be very elegant, and,
4783: moreover, we would have to change @code{draw} every time we add
4784: a new kind of graphical object (say, a spaceship).
4785:
4786: What we would rather do is: When defining spaceships, we would tell
4787: the system: "Here's how you @code{draw} a spaceship; you figure
4788: out the rest."
4789:
4790: This is the problem that all systems solve that (rightfully) call
4791: themselves object-oriented, and the object-oriented package I present
4792: here also solves this problem (and not much else).
4793:
4794: @node Object-Oriented Terminology, Basic Objects Usage, Why object-oriented programming?, Objects
1.12 anton 4795: @subsubsection Object-Oriented Terminology
1.5 anton 4796: @cindex object-oriented terminology
4797: @cindex terminology for object-oriented programming
4798:
4799: This section is mainly for reference, so you don't have to understand
4800: all of it right away. The terminology is mainly Smalltalk-inspired. In
4801: short:
4802:
4803: @table @emph
4804: @cindex class
4805: @item class
4806: a data structure definition with some extras.
4807:
4808: @cindex object
4809: @item object
4810: an instance of the data structure described by the class definition.
4811:
4812: @cindex instance variables
4813: @item instance variables
4814: fields of the data structure.
4815:
4816: @cindex selector
4817: @cindex method selector
4818: @cindex virtual function
4819: @item selector
4820: (or @emph{method selector}) a word (e.g.,
4821: @code{draw}) for performing an operation on a variety of data
4822: structures (classes). A selector describes @emph{what} operation to
4823: perform. In C++ terminology: a (pure) virtual function.
4824:
4825: @cindex method
4826: @item method
4827: the concrete definition that performs the operation
4828: described by the selector for a specific class. A method specifies
4829: @emph{how} the operation is performed for a specific class.
4830:
4831: @cindex selector invocation
4832: @cindex message send
4833: @cindex invoking a selector
4834: @item selector invocation
4835: a call of a selector. One argument of the call (the TOS (top-of-stack))
4836: is used for determining which method is used. In Smalltalk terminology:
4837: a message (consisting of the selector and the other arguments) is sent
4838: to the object.
4839:
4840: @cindex receiving object
4841: @item receiving object
4842: the object used for determining the method executed by a selector
4843: invocation. In our model it is the object that is on the TOS when the
4844: selector is invoked. (@emph{Receiving} comes from Smalltalks
4845: @emph{message} terminology.)
4846:
4847: @cindex child class
4848: @cindex parent class
4849: @cindex inheritance
4850: @item child class
4851: a class that has (@emph{inherits}) all properties (instance variables,
4852: selectors, methods) from a @emph{parent class}. In Smalltalk
4853: terminology: The subclass inherits from the superclass. In C++
4854: terminology: The derived class inherits from the base class.
4855:
4856: @end table
4857:
4858: @c If you wonder about the message sending terminology, it comes from
4859: @c a time when each object had it's own task and objects communicated via
4860: @c message passing; eventually the Smalltalk developers realized that
4861: @c they can do most things through simple (indirect) calls. They kept the
4862: @c terminology.
4863:
4864: @node Basic Objects Usage, The class Object, Object-Oriented Terminology, Objects
1.12 anton 4865: @subsubsection Basic Objects Usage
1.5 anton 4866: @cindex basic objects usage
4867: @cindex objects, basic usage
4868:
4869: You can define a class for graphical objects like this:
4870:
4871: @cindex @code{class} usage
4872: @cindex @code{end-class} usage
4873: @cindex @code{selector} usage
4874: @example
4875: object class \ "object" is the parent class
4876: selector draw ( x y graphical -- )
4877: end-class graphical
4878: @end example
4879:
4880: This code defines a class @code{graphical} with an
4881: operation @code{draw}. We can perform the operation
4882: @code{draw} on any @code{graphical} object, e.g.:
4883:
4884: @example
4885: 100 100 t-rex draw
4886: @end example
4887:
4888: where @code{t-rex} is a word (say, a constant) that produces a
4889: graphical object.
4890:
4891: @cindex abstract class
4892: How do we create a graphical object? With the present definitions,
4893: we cannot create a useful graphical object. The class
4894: @code{graphical} describes graphical objects in general, but not
4895: any concrete graphical object type (C++ users would call it an
4896: @emph{abstract class}); e.g., there is no method for the selector
4897: @code{draw} in the class @code{graphical}.
4898:
4899: For concrete graphical objects, we define child classes of the
4900: class @code{graphical}, e.g.:
4901:
4902: @cindex @code{overrides} usage
4903: @cindex @code{field} usage in class definition
4904: @example
4905: graphical class \ "graphical" is the parent class
4906: cell% field circle-radius
4907:
4908: :noname ( x y circle -- )
4909: circle-radius @@ draw-circle ;
4910: overrides draw
4911:
4912: :noname ( n-radius circle -- )
4913: circle-radius ! ;
4914: overrides construct
4915:
4916: end-class circle
4917: @end example
4918:
4919: Here we define a class @code{circle} as a child of @code{graphical},
4920: with a field @code{circle-radius} (which behaves just like a field in
4921: @pxref{Structures}); it defines new methods for the selectors
4922: @code{draw} and @code{construct} (@code{construct} is defined in
4923: @code{object}, the parent class of @code{graphical}).
4924:
4925: Now we can create a circle on the heap (i.e.,
4926: @code{allocate}d memory) with
4927:
4928: @cindex @code{heap-new} usage
4929: @example
4930: 50 circle heap-new constant my-circle
4931: @end example
4932:
4933: @code{heap-new} invokes @code{construct}, thus
4934: initializing the field @code{circle-radius} with 50. We can draw
4935: this new circle at (100,100) with
4936:
4937: @example
4938: 100 100 my-circle draw
4939: @end example
4940:
4941: @cindex selector invocation, restrictions
4942: @cindex class definition, restrictions
4943: Note: You can invoke a selector only if the object on the TOS
4944: (the receiving object) belongs to the class where the selector was
4945: defined or one of its descendents; e.g., you can invoke
4946: @code{draw} only for objects belonging to @code{graphical}
4947: or its descendents (e.g., @code{circle}). Immediately before
4948: @code{end-class}, the search order has to be the same as
4949: immediately after @code{class}.
4950:
4951: @node The class Object, Creating objects, Basic Objects Usage, Objects
1.12 anton 4952: @subsubsection The class @code{object}
1.5 anton 4953: @cindex @code{object} class
4954:
4955: When you define a class, you have to specify a parent class. So how do
4956: you start defining classes? There is one class available from the start:
4957: @code{object}. You can use it as ancestor for all classes. It is the
4958: only class that has no parent. It has two selectors: @code{construct}
4959: and @code{print}.
4960:
4961: @node Creating objects, Object-Oriented Programming Style, The class Object, Objects
1.12 anton 4962: @subsubsection Creating objects
1.5 anton 4963: @cindex creating objects
4964: @cindex object creation
4965: @cindex object allocation options
4966:
4967: @cindex @code{heap-new} discussion
4968: @cindex @code{dict-new} discussion
4969: @cindex @code{construct} discussion
4970: You can create and initialize an object of a class on the heap with
4971: @code{heap-new} ( ... class -- object ) and in the dictionary
4972: (allocation with @code{allot}) with @code{dict-new} (
4973: ... class -- object ). Both words invoke @code{construct}, which
4974: consumes the stack items indicated by "..." above.
4975:
4976: @cindex @code{init-object} discussion
4977: @cindex @code{class-inst-size} discussion
4978: If you want to allocate memory for an object yourself, you can get its
4979: alignment and size with @code{class-inst-size 2@@} ( class --
4980: align size ). Once you have memory for an object, you can initialize
4981: it with @code{init-object} ( ... class object -- );
4982: @code{construct} does only a part of the necessary work.
4983:
4984: @node Object-Oriented Programming Style, Class Binding, Creating objects, Objects
1.12 anton 4985: @subsubsection Object-Oriented Programming Style
1.5 anton 4986: @cindex object-oriented programming style
4987:
4988: This section is not exhaustive.
4989:
4990: @cindex stack effects of selectors
4991: @cindex selectors and stack effects
4992: In general, it is a good idea to ensure that all methods for the
4993: same selector have the same stack effect: when you invoke a selector,
4994: you often have no idea which method will be invoked, so, unless all
4995: methods have the same stack effect, you will not know the stack effect
4996: of the selector invocation.
4997:
4998: One exception to this rule is methods for the selector
4999: @code{construct}. We know which method is invoked, because we
5000: specify the class to be constructed at the same place. Actually, I
5001: defined @code{construct} as a selector only to give the users a
5002: convenient way to specify initialization. The way it is used, a
5003: mechanism different from selector invocation would be more natural
5004: (but probably would take more code and more space to explain).
5005:
5006: @node Class Binding, Method conveniences, Object-Oriented Programming Style, Objects
1.12 anton 5007: @subsubsection Class Binding
1.5 anton 5008: @cindex class binding
5009: @cindex early binding
5010:
5011: @cindex late binding
5012: Normal selector invocations determine the method at run-time depending
5013: on the class of the receiving object (late binding).
5014:
5015: Sometimes we want to invoke a different method. E.g., assume that
5016: you want to use the simple method for @code{print}ing
5017: @code{object}s instead of the possibly long-winded
5018: @code{print} method of the receiver class. You can achieve this
5019: by replacing the invocation of @code{print} with
5020:
5021: @cindex @code{[bind]} usage
5022: @example
5023: [bind] object print
5024: @end example
5025:
5026: in compiled code or
5027:
5028: @cindex @code{bind} usage
5029: @example
5030: bind object print
5031: @end example
5032:
5033: @cindex class binding, alternative to
5034: in interpreted code. Alternatively, you can define the method with a
5035: name (e.g., @code{print-object}), and then invoke it through the
5036: name. Class binding is just a (often more convenient) way to achieve
5037: the same effect; it avoids name clutter and allows you to invoke
5038: methods directly without naming them first.
5039:
5040: @cindex superclass binding
5041: @cindex parent class binding
5042: A frequent use of class binding is this: When we define a method
5043: for a selector, we often want the method to do what the selector does
5044: in the parent class, and a little more. There is a special word for
5045: this purpose: @code{[parent]}; @code{[parent]
5046: @emph{selector}} is equivalent to @code{[bind] @emph{parent
5047: selector}}, where @code{@emph{parent}} is the parent
5048: class of the current class. E.g., a method definition might look like:
5049:
5050: @cindex @code{[parent]} usage
5051: @example
5052: :noname
5053: dup [parent] foo \ do parent's foo on the receiving object
5054: ... \ do some more
5055: ; overrides foo
5056: @end example
5057:
5058: @cindex class binding as optimization
5059: In @cite{Object-oriented programming in ANS Forth} (Forth Dimensions,
5060: March 1997), Andrew McKewan presents class binding as an optimization
5061: technique. I recommend not using it for this purpose unless you are in
5062: an emergency. Late binding is pretty fast with this model anyway, so the
5063: benefit of using class binding is small; the cost of using class binding
5064: where it is not appropriate is reduced maintainability.
5065:
5066: While we are at programming style questions: You should bind
5067: selectors only to ancestor classes of the receiving object. E.g., say,
5068: you know that the receiving object is of class @code{foo} or its
5069: descendents; then you should bind only to @code{foo} and its
5070: ancestors.
5071:
5072: @node Method conveniences, Classes and Scoping, Class Binding, Objects
1.12 anton 5073: @subsubsection Method conveniences
1.5 anton 5074: @cindex method conveniences
5075:
5076: In a method you usually access the receiving object pretty often. If
5077: you define the method as a plain colon definition (e.g., with
5078: @code{:noname}), you may have to do a lot of stack
5079: gymnastics. To avoid this, you can define the method with @code{m:
5080: ... ;m}. E.g., you could define the method for
5081: @code{draw}ing a @code{circle} with
5082:
5083: @cindex @code{this} usage
5084: @cindex @code{m:} usage
5085: @cindex @code{;m} usage
5086: @example
5087: m: ( x y circle -- )
5088: ( x y ) this circle-radius @@ draw-circle ;m
5089: @end example
5090:
5091: @cindex @code{exit} in @code{m: ... ;m}
5092: @cindex @code{exitm} discussion
5093: @cindex @code{catch} in @code{m: ... ;m}
5094: When this method is executed, the receiver object is removed from the
5095: stack; you can access it with @code{this} (admittedly, in this
5096: example the use of @code{m: ... ;m} offers no advantage). Note
5097: that I specify the stack effect for the whole method (i.e. including
5098: the receiver object), not just for the code between @code{m:}
5099: and @code{;m}. You cannot use @code{exit} in
5100: @code{m:...;m}; instead, use
5101: @code{exitm}.@footnote{Moreover, for any word that calls
5102: @code{catch} and was defined before loading
5103: @code{objects.fs}, you have to redefine it like I redefined
5104: @code{catch}: @code{: catch this >r catch r> to-this ;}}
5105:
5106: @cindex @code{inst-var} usage
5107: You will frequently use sequences of the form @code{this
5108: @emph{field}} (in the example above: @code{this
5109: circle-radius}). If you use the field only in this way, you can
5110: define it with @code{inst-var} and eliminate the
5111: @code{this} before the field name. E.g., the @code{circle}
5112: class above could also be defined with:
5113:
5114: @example
5115: graphical class
5116: cell% inst-var radius
5117:
5118: m: ( x y circle -- )
5119: radius @@ draw-circle ;m
5120: overrides draw
5121:
5122: m: ( n-radius circle -- )
5123: radius ! ;m
5124: overrides construct
5125:
5126: end-class circle
5127: @end example
5128:
5129: @code{radius} can only be used in @code{circle} and its
5130: descendent classes and inside @code{m:...;m}.
5131:
5132: @cindex @code{inst-value} usage
5133: You can also define fields with @code{inst-value}, which is
5134: to @code{inst-var} what @code{value} is to
5135: @code{variable}. You can change the value of such a field with
5136: @code{[to-inst]}. E.g., we could also define the class
5137: @code{circle} like this:
5138:
5139: @example
5140: graphical class
5141: inst-value radius
5142:
5143: m: ( x y circle -- )
5144: radius draw-circle ;m
5145: overrides draw
5146:
5147: m: ( n-radius circle -- )
5148: [to-inst] radius ;m
5149: overrides construct
5150:
5151: end-class circle
5152: @end example
5153:
5154:
5155: @node Classes and Scoping, Object Interfaces, Method conveniences, Objects
1.12 anton 5156: @subsubsection Classes and Scoping
1.5 anton 5157: @cindex classes and scoping
5158: @cindex scoping and classes
5159:
5160: Inheritance is frequent, unlike structure extension. This exacerbates
5161: the problem with the field name convention (@pxref{Structure Naming
5162: Convention}): One always has to remember in which class the field was
5163: originally defined; changing a part of the class structure would require
5164: changes for renaming in otherwise unaffected code.
5165:
5166: @cindex @code{inst-var} visibility
5167: @cindex @code{inst-value} visibility
5168: To solve this problem, I added a scoping mechanism (which was not in my
5169: original charter): A field defined with @code{inst-var} (or
5170: @code{inst-value}) is visible only in the class where it is defined and in
5171: the descendent classes of this class. Using such fields only makes
5172: sense in @code{m:}-defined methods in these classes anyway.
5173:
5174: This scoping mechanism allows us to use the unadorned field name,
5175: because name clashes with unrelated words become much less likely.
5176:
5177: @cindex @code{protected} discussion
5178: @cindex @code{private} discussion
5179: Once we have this mechanism, we can also use it for controlling the
5180: visibility of other words: All words defined after
5181: @code{protected} are visible only in the current class and its
5182: descendents. @code{public} restores the compilation
1.21 crook 5183: (i.e. @code{current}) word list that was in effect before. If you
1.5 anton 5184: have several @code{protected}s without an intervening
5185: @code{public} or @code{set-current}, @code{public}
1.21 crook 5186: will restore the compilation word list in effect before the first of
1.5 anton 5187: these @code{protected}s.
5188:
5189: @node Object Interfaces, Objects Implementation, Classes and Scoping, Objects
1.12 anton 5190: @subsubsection Object Interfaces
1.5 anton 5191: @cindex object interfaces
5192: @cindex interfaces for objects
5193:
5194: In this model you can only call selectors defined in the class of the
5195: receiving objects or in one of its ancestors. If you call a selector
5196: with a receiving object that is not in one of these classes, the
5197: result is undefined; if you are lucky, the program crashes
5198: immediately.
5199:
5200: @cindex selectors common to hardly-related classes
5201: Now consider the case when you want to have a selector (or several)
5202: available in two classes: You would have to add the selector to a
5203: common ancestor class, in the worst case to @code{object}. You
5204: may not want to do this, e.g., because someone else is responsible for
5205: this ancestor class.
5206:
5207: The solution for this problem is interfaces. An interface is a
5208: collection of selectors. If a class implements an interface, the
5209: selectors become available to the class and its descendents. A class
5210: can implement an unlimited number of interfaces. For the problem
5211: discussed above, we would define an interface for the selector(s), and
5212: both classes would implement the interface.
5213:
5214: As an example, consider an interface @code{storage} for
5215: writing objects to disk and getting them back, and a class
5216: @code{foo} foo that implements it. The code for this would look
5217: like this:
5218:
5219: @cindex @code{interface} usage
5220: @cindex @code{end-interface} usage
5221: @cindex @code{implementation} usage
5222: @example
5223: interface
5224: selector write ( file object -- )
5225: selector read1 ( file object -- )
5226: end-interface storage
5227:
5228: bar class
5229: storage implementation
5230:
5231: ... overrides write
5232: ... overrides read
5233: ...
5234: end-class foo
5235: @end example
5236:
5237: (I would add a word @code{read} ( file -- object ) that uses
5238: @code{read1} internally, but that's beyond the point illustrated
5239: here.)
5240:
5241: Note that you cannot use @code{protected} in an interface; and
5242: of course you cannot define fields.
5243:
5244: In the Neon model, all selectors are available for all classes;
5245: therefore it does not need interfaces. The price you pay in this model
5246: is slower late binding, and therefore, added complexity to avoid late
5247: binding.
5248:
5249: @node Objects Implementation, Comparison with other object models, Object Interfaces, Objects
1.12 anton 5250: @subsubsection @file{objects.fs} Implementation
1.5 anton 5251: @cindex @file{objects.fs} implementation
5252:
5253: @cindex @code{object-map} discussion
5254: An object is a piece of memory, like one of the data structures
5255: described with @code{struct...end-struct}. It has a field
5256: @code{object-map} that points to the method map for the object's
5257: class.
5258:
5259: @cindex method map
5260: @cindex virtual function table
5261: The @emph{method map}@footnote{This is Self terminology; in C++
5262: terminology: virtual function table.} is an array that contains the
5263: execution tokens (XTs) of the methods for the object's class. Each
5264: selector contains an offset into the method maps.
5265:
5266: @cindex @code{selector} implementation, class
5267: @code{selector} is a defining word that uses
5268: @code{create} and @code{does>}. The body of the
5269: selector contains the offset; the @code{does>} action for a
5270: class selector is, basically:
5271:
5272: @example
5273: ( object addr ) @@ over object-map @@ + @@ execute
5274: @end example
5275:
5276: Since @code{object-map} is the first field of the object, it
5277: does not generate any code. As you can see, calling a selector has a
5278: small, constant cost.
5279:
5280: @cindex @code{current-interface} discussion
5281: @cindex class implementation and representation
5282: A class is basically a @code{struct} combined with a method
5283: map. During the class definition the alignment and size of the class
5284: are passed on the stack, just as with @code{struct}s, so
5285: @code{field} can also be used for defining class
5286: fields. However, passing more items on the stack would be
5287: inconvenient, so @code{class} builds a data structure in memory,
5288: which is accessed through the variable
5289: @code{current-interface}. After its definition is complete, the
5290: class is represented on the stack by a pointer (e.g., as parameter for
5291: a child class definition).
5292:
5293: At the start, a new class has the alignment and size of its parent,
5294: and a copy of the parent's method map. Defining new fields extends the
5295: size and alignment; likewise, defining new selectors extends the
5296: method map. @code{overrides} just stores a new XT in the method
5297: map at the offset given by the selector.
5298:
5299: @cindex class binding, implementation
5300: Class binding just gets the XT at the offset given by the selector
5301: from the class's method map and @code{compile,}s (in the case of
5302: @code{[bind]}) it.
5303:
5304: @cindex @code{this} implementation
5305: @cindex @code{catch} and @code{this}
5306: @cindex @code{this} and @code{catch}
5307: I implemented @code{this} as a @code{value}. At the
5308: start of an @code{m:...;m} method the old @code{this} is
5309: stored to the return stack and restored at the end; and the object on
5310: the TOS is stored @code{TO this}. This technique has one
5311: disadvantage: If the user does not leave the method via
5312: @code{;m}, but via @code{throw} or @code{exit},
5313: @code{this} is not restored (and @code{exit} may
5314: crash). To deal with the @code{throw} problem, I have redefined
5315: @code{catch} to save and restore @code{this}; the same
5316: should be done with any word that can catch an exception. As for
5317: @code{exit}, I simply forbid it (as a replacement, there is
5318: @code{exitm}).
5319:
5320: @cindex @code{inst-var} implementation
5321: @code{inst-var} is just the same as @code{field}, with
5322: a different @code{does>} action:
5323: @example
5324: @@ this +
5325: @end example
5326: Similar for @code{inst-value}.
5327:
5328: @cindex class scoping implementation
1.21 crook 5329: Each class also has a word list that contains the words defined with
1.5 anton 5330: @code{inst-var} and @code{inst-value}, and its protected
5331: words. It also has a pointer to its parent. @code{class} pushes
1.21 crook 5332: the word lists of the class an all its ancestors on the search order,
1.5 anton 5333: and @code{end-class} drops them.
5334:
5335: @cindex interface implementation
5336: An interface is like a class without fields, parent and protected
5337: words; i.e., it just has a method map. If a class implements an
5338: interface, its method map contains a pointer to the method map of the
5339: interface. The positive offsets in the map are reserved for class
5340: methods, therefore interface map pointers have negative
5341: offsets. Interfaces have offsets that are unique throughout the
5342: system, unlike class selectors, whose offsets are only unique for the
5343: classes where the selector is available (invokable).
5344:
5345: This structure means that interface selectors have to perform one
5346: indirection more than class selectors to find their method. Their body
5347: contains the interface map pointer offset in the class method map, and
5348: the method offset in the interface method map. The
5349: @code{does>} action for an interface selector is, basically:
5350:
5351: @example
5352: ( object selector-body )
5353: 2dup selector-interface @@ ( object selector-body object interface-offset )
5354: swap object-map @@ + @@ ( object selector-body map )
5355: swap selector-offset @@ + @@ execute
5356: @end example
5357:
5358: where @code{object-map} and @code{selector-offset} are
5359: first fields and generate no code.
5360:
5361: As a concrete example, consider the following code:
5362:
5363: @example
5364: interface
5365: selector if1sel1
5366: selector if1sel2
5367: end-interface if1
5368:
5369: object class
5370: if1 implementation
5371: selector cl1sel1
5372: cell% inst-var cl1iv1
5373:
5374: ' m1 overrides construct
5375: ' m2 overrides if1sel1
5376: ' m3 overrides if1sel2
5377: ' m4 overrides cl1sel2
5378: end-class cl1
5379:
5380: create obj1 object dict-new drop
5381: create obj2 cl1 dict-new drop
5382: @end example
5383:
5384: The data structure created by this code (including the data structure
5385: for @code{object}) is shown in the <a
5386: href="objects-implementation.eps">figure</a>, assuming a cell size of 4.
5387:
5388: @node Comparison with other object models, Objects Glossary, Objects Implementation, Objects
1.12 anton 5389: @subsubsection Comparison with other object models
1.5 anton 5390: @cindex comparison of object models
5391: @cindex object models, comparison
5392:
5393: Many object-oriented Forth extensions have been proposed (@cite{A survey
5394: of object-oriented Forths} (SIGPLAN Notices, April 1996) by Bradford
5395: J. Rodriguez and W. F. S. Poehlman lists 17). Here I'll discuss the
5396: relation of @file{objects.fs} to two well-known and two closely-related
5397: (by the use of method maps) models.
5398:
5399: @cindex Neon model
5400: The most popular model currently seems to be the Neon model (see
5401: @cite{Object-oriented programming in ANS Forth} (Forth Dimensions, March
5402: 1997) by Andrew McKewan). The Neon model uses a @code{@emph{selector
5403: object}} syntax, which makes it unnatural to pass objects on the
5404: stack. It also requires that the selector parses the input stream (at
5405: compile time); this leads to reduced extensibility and to bugs that are
5406: hard to find. Finally, it allows using every selector to every object;
5407: this eliminates the need for classes, but makes it harder to create
5408: efficient implementations. A longer version of this critique can be
5409: found in @cite{On Standardizing Object-Oriented Forth Extensions} (Forth
5410: Dimensions, May 1997) by Anton Ertl.
5411:
5412: @cindex Pountain's object-oriented model
5413: Another well-known publication is @cite{Object-Oriented Forth} (Academic
5414: Press, London, 1987) by Dick Pountain. However, it is not really about
5415: object-oriented programming, because it hardly deals with late
5416: binding. Instead, it focuses on features like information hiding and
5417: overloading that are characteristic of modular languages like Ada (83).
5418:
5419: @cindex Zsoter's object-oriented model
5420: In @cite{Does late binding have to be slow?} (Forth Dimensions ??? 1996)
5421: Andras Zsoter describes a model that makes heavy use of an active object
5422: (like @code{this} in @file{objects.fs}): The active object is not only
5423: used for accessing all fields, but also specifies the receiving object
5424: of every selector invocation; you have to change the active object
5425: explicitly with @code{@{ ... @}}, whereas in @file{objects.fs} it
5426: changes more or less implicitly at @code{m: ... ;m}. Such a change at
5427: the method entry point is unnecessary with the Zsoter's model, because
5428: the receiving object is the active object already; OTOH, the explicit
5429: change is absolutely necessary in that model, because otherwise no one
5430: could ever change the active object. An ANS Forth implementation of this
5431: model is available at @url{http://www.forth.org/fig/oopf.html}.
5432:
1.12 anton 5433: @cindex @file{oof.fs}, differences to other models
1.5 anton 5434: The @file{oof.fs} model combines information hiding and overloading
1.21 crook 5435: resolution (by keeping names in various word lists) with object-oriented
1.5 anton 5436: programming. It sets the active object implicitly on method entry, but
5437: also allows explicit changing (with @code{>o...o>} or with
5438: @code{with...endwith}). It uses parsing and state-smart objects and
5439: classes for resolving overloading and for early binding: the object or
5440: class parses the selector and determines the method from this. If the
5441: selector is not parsed by an object or class, it performs a call to the
5442: selector for the active object (late binding), like Zsoter's model.
5443: Fields are always accessed through the active object. The big
5444: disadvantage of this model is the parsing and the state-smartness, which
5445: reduces extensibility and increases the opportunities for subtle bugs;
5446: essentially, you are only safe if you never tick or @code{postpone} an
1.12 anton 5447: object or class (Bernd disagrees, but I (Anton) am not convinced).
5448:
5449: @cindex @file{mini-oof.fs}, differences to other models
5450: The Mini-OOF model is quite similar to a very stripped-down version of
5451: the Objects model, but syntactically it is a mixture of the Objects and
5452: the OOF model.
5453:
1.5 anton 5454:
5455: @node Objects Glossary, , Comparison with other object models, Objects
1.12 anton 5456: @subsubsection @file{objects.fs} Glossary
1.5 anton 5457: @cindex @file{objects.fs} Glossary
5458:
1.19 anton 5459: doc---objects-bind
5460: doc---objects-<bind>
5461: doc---objects-bind'
5462: doc---objects-[bind]
5463: doc---objects-class
5464: doc---objects-class->map
5465: doc---objects-class-inst-size
5466: doc---objects-class-override!
5467: doc---objects-construct
5468: doc---objects-current'
5469: doc---objects-[current]
5470: doc---objects-current-interface
5471: doc---objects-dict-new
5472: doc---objects-drop-order
5473: doc---objects-end-class
5474: doc---objects-end-class-noname
5475: doc---objects-end-interface
5476: doc---objects-end-interface-noname
5477: doc---objects-exitm
5478: doc---objects-heap-new
5479: doc---objects-implementation
5480: doc---objects-init-object
5481: doc---objects-inst-value
5482: doc---objects-inst-var
5483: doc---objects-interface
5484: doc---objects-;m
5485: doc---objects-m:
5486: doc---objects-method
5487: doc---objects-object
5488: doc---objects-overrides
5489: doc---objects-[parent]
5490: doc---objects-print
5491: doc---objects-protected
5492: doc---objects-public
5493: doc---objects-push-order
5494: doc---objects-selector
5495: doc---objects-this
5496: doc---objects-<to-inst>
5497: doc---objects-[to-inst]
5498: doc---objects-to-this
5499: doc---objects-xt-new
1.5 anton 5500:
5501: @c -------------------------------------------------------------
1.12 anton 5502: @node OOF, Mini-OOF, Objects, Object-oriented Forth
5503: @subsection OOF
1.6 pazsan 5504: @cindex oof
5505: @cindex object-oriented programming
5506:
5507: @cindex @file{objects.fs}
5508: @cindex @file{oof.fs}
1.12 anton 5509:
5510: This section describes the @file{oof.fs} packet. This section uses the
5511: same rationale why using object-oriented programming, and the same
1.6 pazsan 5512: terminology.
5513:
5514: The packet described in this section is used in bigFORTH since 1991, and
5515: used for two large applications: a chromatographic system used to
5516: create new medicaments, and a graphic user interface library (MINOS).
5517:
1.12 anton 5518: You can find a description (in German) of @file{oof.fs} in @cite{Object
5519: oriented bigFORTH} by Bernd Paysan, published in @cite{Vierte Dimension}
5520: 10(2), 1994.
5521:
1.6 pazsan 5522: @menu
5523: * Properties of the OOF model::
5524: * Basic OOF Usage::
5525: * The base class object::
1.7 pazsan 5526: * Class Declaration::
5527: * Class Implementation::
1.6 pazsan 5528: @end menu
5529:
1.12 anton 5530: @node Properties of the OOF model, Basic OOF Usage, OOF, OOF
5531: @subsubsection Properties of the OOF model
1.6 pazsan 5532: @cindex @file{oof.fs} properties
5533:
5534: @itemize @bullet
5535: @item
5536: This model combines object oriented programming with information
5537: hiding. It helps you writing large application, where scoping is
5538: necessary, because it provides class-oriented scoping.
5539:
5540: @item
5541: Named objects, object pointers, and object arrays can be created,
5542: selector invocation uses the "object selector" syntax. Selector invocation
5543: to objects and/or selectors on the stack is a bit less convenient, but
5544: possible.
5545:
5546: @item
5547: Selector invocation and instance variable usage of the active object is
5548: straight forward, since both make use of the active object.
5549:
5550: @item
5551: Late binding is efficient and easy to use.
5552:
5553: @item
5554: State-smart objects parse selectors. However, extensibility is provided
5555: using a (parsing) selector @code{postpone} and a selector @code{'}.
5556:
5557: @item
5558: An implementation in ANS Forth is available.
5559:
5560: @end itemize
5561:
5562:
1.12 anton 5563: @node Basic OOF Usage, The base class object, Properties of the OOF model, OOF
5564: @subsubsection Basic OOF Usage
1.6 pazsan 5565: @cindex @file{oof.fs} usage
5566:
5567: Here, I use the same example as for @code{objects} (@pxref{Basic Objects Usage}).
5568:
5569: You can define a class for graphical objects like this:
5570:
5571: @cindex @code{class} usage
5572: @cindex @code{class;} usage
5573: @cindex @code{method} usage
5574: @example
5575: object class graphical \ "object" is the parent class
5576: method draw ( x y graphical -- )
5577: class;
5578: @end example
5579:
5580: This code defines a class @code{graphical} with an
5581: operation @code{draw}. We can perform the operation
5582: @code{draw} on any @code{graphical} object, e.g.:
5583:
5584: @example
5585: 100 100 t-rex draw
5586: @end example
5587:
5588: where @code{t-rex} is an object or object pointer, created with e.g.
1.13 pazsan 5589: @code{graphical : t-rex}.
1.6 pazsan 5590:
5591: @cindex abstract class
5592: How do we create a graphical object? With the present definitions,
5593: we cannot create a useful graphical object. The class
5594: @code{graphical} describes graphical objects in general, but not
5595: any concrete graphical object type (C++ users would call it an
5596: @emph{abstract class}); e.g., there is no method for the selector
5597: @code{draw} in the class @code{graphical}.
5598:
5599: For concrete graphical objects, we define child classes of the
5600: class @code{graphical}, e.g.:
5601:
5602: @example
5603: graphical class circle \ "graphical" is the parent class
5604: cell var circle-radius
5605: how:
5606: : draw ( x y -- )
5607: circle-radius @@ draw-circle ;
5608:
5609: : init ( n-radius -- (
5610: circle-radius ! ;
5611: class;
5612: @end example
5613:
5614: Here we define a class @code{circle} as a child of @code{graphical},
5615: with a field @code{circle-radius}; it defines new methods for the
5616: selectors @code{draw} and @code{init} (@code{init} is defined in
5617: @code{object}, the parent class of @code{graphical}).
5618:
5619: Now we can create a circle in the dictionary with
5620:
5621: @example
5622: 50 circle : my-circle
5623: @end example
5624:
5625: @code{:} invokes @code{init}, thus initializing the field
5626: @code{circle-radius} with 50. We can draw this new circle at (100,100)
5627: with
5628:
5629: @example
5630: 100 100 my-circle draw
5631: @end example
5632:
5633: @cindex selector invocation, restrictions
5634: @cindex class definition, restrictions
5635: Note: You can invoke a selector only if the receiving object belongs to
5636: the class where the selector was defined or one of its descendents;
5637: e.g., you can invoke @code{draw} only for objects belonging to
5638: @code{graphical} or its descendents (e.g., @code{circle}). The scoping
1.7 pazsan 5639: mechanism will check if you try to invoke a selector that is not
1.6 pazsan 5640: defined in this class hierarchy, so you'll get an error at compilation
5641: time.
5642:
5643:
1.12 anton 5644: @node The base class object, Class Declaration, Basic OOF Usage, OOF
5645: @subsubsection The base class @file{object}
1.6 pazsan 5646: @cindex @file{oof.fs} base class
5647:
5648: When you define a class, you have to specify a parent class. So how do
5649: you start defining classes? There is one class available from the start:
5650: @code{object}. You have to use it as ancestor for all classes. It is the
5651: only class that has no parent. Classes are also objects, except that
5652: they don't have instance variables; class manipulation such as
5653: inheritance or changing definitions of a class is handled through
5654: selectors of the class @code{object}.
5655:
5656: @code{object} provides a number of selectors:
5657:
5658: @itemize @bullet
5659: @item
5660: @code{class} for subclassing, @code{definitions} to add definitions
5661: later on, and @code{class?} to get type informations (is the class a
5662: subclass of the class passed on the stack?).
1.7 pazsan 5663: doc---object-class
5664: doc---object-definitions
5665: doc---object-class?
1.6 pazsan 5666:
5667: @item
5668: @code{init} and @code{dispose} as constructor and destroctor of the
5669: object. @code{init} is invocated after the object's memory is allocated,
5670: while @code{dispose} also handles deallocation. Thus if you redefine
5671: @code{dispose}, you have to call the parent's dispose with @code{super
5672: dispose}, too.
1.7 pazsan 5673: doc---object-init
5674: doc---object-dispose
1.6 pazsan 5675:
5676: @item
1.7 pazsan 5677: @code{new}, @code{new[]}, @code{:}, @code{ptr}, @code{asptr}, and
5678: @code{[]} to create named and unnamed objects and object arrays or
5679: object pointers.
5680: doc---object-new
5681: doc---object-new[]
5682: doc---object-:
5683: doc---object-ptr
5684: doc---object-asptr
5685: doc---object-[]
1.6 pazsan 5686:
5687: @item
1.21 crook 5688: @code{::} and @code{super} for explicit scoping. You should use expicit
1.6 pazsan 5689: scoping only for super classes or classes with the same set of instance
5690: variables. Explicit scoped selectors use early binding.
1.7 pazsan 5691: doc---object-::
5692: doc---object-super
1.6 pazsan 5693:
5694: @item
5695: @code{self} to get the address of the object
1.7 pazsan 5696: doc---object-self
1.6 pazsan 5697:
5698: @item
5699: @code{bind}, @code{bound}, @code{link}, and @code{is} to assign object
5700: pointers and instance defers.
1.7 pazsan 5701: doc---object-bind
5702: doc---object-bound
5703: doc---object-link
5704: doc---object-is
1.6 pazsan 5705:
5706: @item
5707: @code{'} to obtain selector tokens, @code{send} to invocate selectors
5708: form the stack, and @code{postpone} to generate selector invocation code.
1.7 pazsan 5709: doc---object-'
5710: doc---object-postpone
1.6 pazsan 5711:
5712: @item
5713: @code{with} and @code{endwith} to select the active object from the
5714: stack, and enabling it's scope. Using @code{with} and @code{endwith}
5715: also allows to create code using selector @code{postpone} without being
5716: trapped bye the state-smart objects.
1.7 pazsan 5717: doc---object-with
5718: doc---object-endwith
1.6 pazsan 5719:
5720: @end itemize
5721:
1.12 anton 5722: @node Class Declaration, Class Implementation, The base class object, OOF
5723: @subsubsection Class Declaration
1.7 pazsan 5724: @cindex class declaration
5725:
5726: @itemize @bullet
5727: @item
5728: Instance variables
5729: doc---oof-var
5730:
5731: @item
5732: Object pointers
5733: doc---oof-ptr
5734: doc---oof-asptr
5735:
5736: @item
5737: Instance defers
5738: doc---oof-defer
5739:
5740: @item
5741: Method selectors
5742: doc---oof-early
5743: doc---oof-method
5744:
5745: @item
5746: Class wide variables
5747: doc---oof-static
5748:
5749: @item
5750: End declaration
5751: doc---oof-how:
5752: doc---oof-class;
5753:
5754: @end itemize
5755:
1.13 pazsan 5756: @c -------------------------------------------------------------
1.12 anton 5757: @node Class Implementation, , Class Declaration, OOF
5758: @subsubsection Class Implementation
1.7 pazsan 5759: @cindex class implementation
5760:
1.13 pazsan 5761: @c -------------------------------------------------------------
5762: @node Mini-OOF, , OOF, Object-oriented Forth
1.12 anton 5763: @subsection Mini-OOF
1.8 pazsan 5764: @cindex mini-oof
5765:
5766: Gforth's third object oriented Forth package is a 12-liner. It uses a
5767: bit of a mixture of the @file{object.fs} and the @file{oof.fs} syntax,
1.13 pazsan 5768: and reduces to the bare minimum of features. This is based on a posting
5769: of Bernd Paysan in comp.arch.
5770:
5771: @menu
5772: * Mini-OOF Usage::
5773: * Mini-OOF Example::
1.20 pazsan 5774: * Mini-OOF Implementation::
1.13 pazsan 5775: @end menu
5776:
5777: @c -------------------------------------------------------------
5778: @node Mini-OOF Usage, Mini-OOF Example, , Mini-OOF
5779: @subsubsection Usage
5780: @cindex mini-oof usage
5781:
5782: Basically, there are seven words, to define a method, a variable, a
5783: class; to end a class, to define a method, to allocate an object, to
5784: resolve binding, and the base class (which allocates one cell for the
5785: object pointer).
5786:
5787: doc-method
5788:
5789: Defines a method
5790:
5791: doc-var
5792:
5793: Defines a variable with size bytes
5794:
5795: doc-class
5796:
5797: Starts the definition of a sub-class
5798:
5799: doc-end-class
5800:
5801: Ends the definition of a class
5802:
5803: doc-defines
5804:
5805: Binds the xt to the method name in the class
5806:
5807: doc-new
5808:
5809: Creates a new incarnation of the class
5810:
5811: doc-::
5812:
5813: Compiles the method name of the class (not immediate!)
5814:
5815: doc-object
5816:
5817: Is the base class of all objects
5818:
5819: @c -------------------------------------------------------------
1.20 pazsan 5820: @node Mini-OOF Example, Mini-OOF Implementation, Mini-OOF Usage, Mini-OOF
1.13 pazsan 5821: @subsubsection Mini-OOF Example
5822: @cindex mini-oof example
5823:
5824: A short example shows how to use this package.
5825:
5826: @example
5827: object class
5828: method init
5829: method draw
5830: end-class graphical
5831: @end example
5832:
5833: This code defines a class @code{graphical} with an
5834: operation @code{draw}. We can perform the operation
5835: @code{draw} on any @code{graphical} object, e.g.:
5836:
5837: @example
5838: 100 100 t-rex draw
5839: @end example
5840:
5841: where @code{t-rex} is an object or object pointer, created with e.g.
5842: @code{graphical new Constant t-rex}.
5843:
5844: For concrete graphical objects, we define child classes of the
5845: class @code{graphical}, e.g.:
1.8 pazsan 5846:
5847: @example
1.13 pazsan 5848: graphical class
5849: cell var circle-radius
5850: end-class circle \ "graphical" is the parent class
5851:
5852: :noname ( x y -- )
5853: circle-radius @@ draw-circle ; circle defines draw
5854: :noname ( r -- )
5855: circle-radius ! ; circle defines init
5856: @end example
5857:
5858: There is no implicit init method, so we have to define one. The creation
5859: code of the object now has to call init explicitely.
5860:
5861: @example
5862: circle new Constant my-circle
5863: 50 my-circle init
5864: @end example
5865:
5866: It is also possible to add a function to create named objects with
5867: automatic call of @code{init}, given that all objects have @code{init}
5868: on the same place
5869:
5870: @example
5871: : new: ( .. o "name" -- )
5872: new dup Constant init ;
5873: 80 circle new: large-circle
5874: @end example
5875:
5876: We can draw this new circle at (100,100)
5877: with
5878:
5879: @example
5880: 100 100 my-circle draw
1.8 pazsan 5881: @end example
5882:
1.20 pazsan 5883: @node Mini-OOF Implementation, , Mini-OOF Example, Mini-OOF
5884: @subsubsection Mini-OOF Implementation
5885:
5886: Object oriented system with late binding typically use a
5887: "vtable"-approach: the first variable in each object is a pointer to a
5888: table, which contains the methods as function pointers. This vtable
5889: may contain some other informations, too.
5890:
5891: So first, let's declare methods:
5892:
5893: @example
5894: : method ( m v -- m' v ) Create over , swap cell+ swap
5895: DOES> ( ... o -- ... ) @ over @ + @ execute ;
5896: @end example
5897:
5898: During method declaration, the number of methods and instance
5899: variables is on the stack (in address units). @code{method} creates
5900: one method and increments the method number. To execute a method, it
5901: takes the object, fetches the vtable pointer, adds the offset, and
5902: executes the xt stored there. Each method takes the object it is
5903: invoked from as top of stack parameter. The method itself should
5904: consume that object.
5905:
5906: Now, we also have to declare instance variables
5907:
5908: @example
5909: : var ( m v size -- m v' ) Create over , +
5910: DOES> ( o -- addr ) @ + ;
5911: @end example
5912:
5913: Same as above, a word is created with the current offset. Instance
5914: variables can have different sizes (cells, floats, doubles, chars), so
5915: all we do is take the size and add it to the offset. If your machine
5916: has alignment restrictions, put the proper @code{aligned} or
5917: @code{faligned} before the variable, it will adjust the variable
5918: offset. That's why it is on the top of stack.
5919:
5920: We need a starting point (the empty object) and some syntactic sugar:
5921:
5922: @example
5923: Create object 1 cells , 2 cells ,
5924: : class ( class -- class methods vars ) dup 2@ ;
5925: @end example
5926:
5927: Now, for inheritance, the vtable of the parent object has to be
5928: copied, when a new, derived class is declared. This gives all the
5929: methods of the parent class, which can be overridden, though.
5930:
5931: @example
5932: : end-class ( class methods vars -- )
5933: Create here >r , dup , 2 cells ?DO ['] noop , 1 cells +LOOP
5934: cell+ dup cell+ r> rot @ 2 cells /string move ;
5935: @end example
5936:
5937: The first line creates the vtable, initialized with
5938: @code{noop}s. The second line is the inheritance mechanism, it
5939: copies the xts from the parent vtable.
5940:
5941: We still have no way to define new methods, let's do that now:
5942:
5943: @example
5944: : defines ( xt class -- ) ' >body @ + ! ;
5945: @end example
5946:
5947: To allocate a new object, we need a word, too:
5948:
5949: @example
5950: : new ( class -- o ) here over @ allot swap over ! ;
5951: @end example
5952:
5953: And sometimes derived classes want to access the method of the
5954: parent object. There are two ways to achieve this with this OOF:
5955: first, you could use named words, and second, you could look up the
5956: vtable of the parent object.
5957:
5958: @example
5959: : :: ( class "name" -- ) ' >body @ + @ compile, ;
5960: @end example
5961:
5962: <H2>An Example</H2>
5963:
5964: Nothing can be more confusing than a good example, so here is
5965: one. First let's declare a text object (further called
5966: @code{button}), that stores text and position:
5967:
5968: @example
5969: object class
5970: cell var text
5971: cell var len
5972: cell var x
5973: cell var y
5974: method init
5975: method draw
5976: end-class button
5977: @end example
5978:
5979: Now, implement the two methods, @code{draw} and @code{init}:
5980:
5981: @example
5982: :noname ( o -- ) >r
5983: r@ x @ r@ y @ at-xy r@ text @ r> len @ type ;
5984: button defines draw
5985: :noname ( addr u o -- ) >r
5986: 0 r@ x ! 0 r@ y ! r@ len ! r> text ! ;
5987: button defines init
5988: @end example
5989:
5990: For inheritance, we define a class @code{bold-button}, with no
5991: new data and no new methods.
5992:
5993: @example
5994: button class
5995: end-class bold-button
5996:
5997: : bold 27 emit ." [1m" ;
5998: : normal 27 emit ." [0m" ;
5999:
6000: :noname bold [ button :: draw ] normal ; bold-button defines draw
6001: @end example
6002:
6003: And finally, some code to demonstrate how to create objects and
6004: apply methods:
6005:
6006: @example
6007: button new Constant foo
6008: s" thin foo" foo init
6009: page
6010: foo draw
6011: bold-button new Constant bar
6012: s" fat bar" bar init
6013: 1 bar y !
6014: bar draw
6015: @end example
6016:
1.6 pazsan 6017: @c -------------------------------------------------------------
1.21 crook 6018: @node Tokens for Words, Word Lists, Object-oriented Forth, Words
1.1 anton 6019: @section Tokens for Words
6020: @cindex tokens for words
6021:
6022: This chapter describes the creation and use of tokens that represent
6023: words on the stack (and in data space).
6024:
6025: Named words have interpretation and compilation semantics. Unnamed words
6026: just have execution semantics.
6027:
1.21 crook 6028: @comment TODO ?normally interpretation semantics are the execution semantics.
6029: @comment this should all be covered in earlier ss
6030:
1.1 anton 6031: @cindex execution token
6032: An @dfn{execution token} represents the execution semantics of an
6033: unnamed word. An execution token occupies one cell. As explained in
1.21 crook 6034: @ref{Supplying names}, the execution token of the last word
6035: defined can be produced with @code{lastxt}.
1.1 anton 6036:
1.21 crook 6037: You can perform the semantics represented by an execution token with:
1.1 anton 6038: doc-execute
1.21 crook 6039: You can compile the word with:
1.1 anton 6040: doc-compile,
6041:
6042: @cindex code field address
6043: @cindex CFA
6044: In Gforth, the abstract data type @emph{execution token} is implemented
6045: as CFA (code field address).
1.21 crook 6046: @comment TODO note that the standard does not say what it represents..
6047: @comment and you cannot necessarily compile it in all Forths (eg native
6048: @comment compilers?).
1.1 anton 6049:
6050: The interpretation semantics of a named word are also represented by an
6051: execution token. You can get it with
6052:
6053: doc-[']
6054: doc-'
6055:
6056: For literals, you use @code{'} in interpreted code and @code{[']} in
6057: compiled code. Gforth's @code{'} and @code{[']} behave somewhat unusual
6058: by complaining about compile-only words. To get an execution token for a
6059: compiling word @var{X}, use @code{COMP' @var{X} drop} or @code{[COMP']
6060: @var{X} drop}.
6061:
6062: @cindex compilation token
6063: The compilation semantics are represented by a @dfn{compilation token}
6064: consisting of two cells: @var{w xt}. The top cell @var{xt} is an
6065: execution token. The compilation semantics represented by the
6066: compilation token can be performed with @code{execute}, which consumes
6067: the whole compilation token, with an additional stack effect determined
6068: by the represented compilation semantics.
6069:
6070: doc-[comp']
6071: doc-comp'
6072:
6073: You can compile the compilation semantics with @code{postpone,}. I.e.,
6074: @code{COMP' @var{word} POSTPONE,} is equivalent to @code{POSTPONE
6075: @var{word}}.
6076:
6077: doc-postpone,
6078:
6079: At present, the @var{w} part of a compilation token is an execution
6080: token, and the @var{xt} part represents either @code{execute} or
6081: @code{compile,}. However, don't rely on that knowledge, unless necessary;
6082: we may introduce unusual compilation tokens in the future (e.g.,
6083: compilation tokens representing the compilation semantics of literals).
6084:
6085: @cindex name token
6086: @cindex name field address
6087: @cindex NFA
6088: Named words are also represented by the @dfn{name token}. The abstract
6089: data type @emph{name token} is implemented as NFA (name field address).
6090:
6091: doc-find-name
6092: doc-name>int
6093: doc-name?int
6094: doc-name>comp
6095: doc-name>string
6096:
1.21 crook 6097: @node Word Lists, Environmental Queries, Tokens for Words, Words
6098: @section Word Lists
6099: @cindex word lists
6100: @cindex name dictionary
6101:
6102: @cindex wid
6103: All definitions other than those created by @code{:noname} have an entry
6104: in the name dictionary. The name dictionary is fragmented into a number
6105: of parts, called @var{word lists}. A word list is identified by a
6106: cell-sized word list identifier (@var{wid}) in much the same way as a
6107: file is identified by a file handle. The numerical value of the wid has
6108: no (portable) meaning, and might change from session to session.
6109:
6110: @cindex compilation word list
6111: At any one time, a single word list is defined as the word list to which
6112: all new definitions will be added -- this is called the @var{compilation
6113: word list}. When Gforth is started, the compilation word list is the
6114: word list called @code{FORTH-WORDLIST}.
6115:
6116: @cindex search order stack
6117: Forth maintains a stack of word lists, representing the @var{search
6118: order}. When the name dictionary is searched (for example, when
6119: attempting to find a word's execution token during compilation), only
6120: those word lists that are currently in the search order are
6121: searched. The most recently-defined word in the word list at the top of
6122: the word list stack is searched first, and the search proceeds until
6123: either the word is located or the oldest definition in the word list at
6124: the bottom of the stack is reached. Definitions of the word may exist in
6125: more than one word lists; the search order determines which version will
6126: be found.
6127:
6128: The ANS Forth Standard "Search order" word set is intended to provide a
6129: set of low-level tools that allow various different schemes to be
6130: implemented. Gforth provides @code{vocabulary}, a traditional Forth
6131: word. @file{compat/vocabulary.fs} provides an implementation in ANS
6132: Standard Forth.
6133:
6134: TODO: locals section refers to here, saying that every word list (aka
6135: vocabulary) has its own methods for searching etc. Need to document that.
6136:
6137: doc-forth-wordlist
6138: doc-definitions
6139: doc-get-current
6140: doc-set-current
6141:
6142: @comment TODO when a defn (like set-order) is instanced twice, the second instance gets documented.
6143: @comment In general that might be fine, but in this example (search.fs) the second instance is an
6144: @comment alias, so it would not naturally have documentation
6145:
6146: doc-get-order
6147: doc-set-order
6148: doc-wordlist
6149: doc-also
6150: doc-forth
6151: doc-only
6152: doc-order
6153: doc-previous
6154:
6155: doc-find
6156: doc-search-wordlist
6157:
6158: doc-words
6159: doc-vlist
6160:
6161: doc-mappedwordlist
6162: doc-root
6163: doc-vocabulary
6164: doc-seal
6165: doc-vocs
6166: doc-current
6167: doc-context
6168:
6169: @menu
6170: * Why use word lists?::
6171: * Word list examples::
6172: @end menu
6173:
6174: @node Why use word lists?, Word list examples, Word Lists, Word Lists
6175: @subsection Why use word lists?
6176: @cindex word lists - why use them?
6177:
6178: There are several reasons for using multiple word lists:
6179:
6180: @itemize @bullet
6181: @item
6182: To improve compilation speed by reducing the number of name dictionary
6183: entries that must be searched. This is achieved by creating a new
6184: word list that contains all of the definitions that are used in the
6185: definition of a Forth system but which would not usually be used by
6186: programs running on that system. That word list would be on the search
6187: list when the Forth system was compiled but would be removed from the
6188: search list for normal operation. This can be a useful technique for
6189: low-performance systems (for example, 8-bit processors in embedded
6190: systems) but is unlikely to be necessary in high-performance desktop
6191: systems.
6192: @item
6193: To prevent a set of words from being used outside the context in which
6194: they are valid. Two classic examples of this are an integrated editor
6195: (all of the edit commands are defined in a separate word list; the
6196: search order is set to the editor word list when the editor is invoked;
6197: the old search order is restored when the editor is terminated) and an
6198: integrated assembler (the op-codes for the machine are defined in a
6199: separate word list which is used when a @code{CODE} word is defined).
6200: @item
6201: To prevent a name-space clash between multiple definitions with the same
6202: name. For example, when building a cross-compiler you might have a word
6203: @code{IF} that generates conditional code for your target system. By
6204: placing this definition in a different word list you can control whether
6205: the host system's @code{IF} or the target system's @code{IF} get used in
6206: any particular context by controlling the order of the word lists on the
6207: search order stack.
6208: @end itemize
6209:
6210: @node Word list examples, ,Why use word lists?, Word Lists
6211: @subsection Word list examples
6212: @cindex word lists - examples
6213:
6214: Here is an example of creating and using a new wordlist using ANS
6215: Standard words:
6216:
6217: @example
6218: wordlist constant my-new-words-wordlist
6219: : my-new-words get-order nip my-new-words-wordlist swap set-order ;
6220:
6221: \ add it to the search order
6222: also my-new-words
6223:
6224: \ alternatively, add it to the search order and make it
6225: \ the compilation word list
6226: also my-new-words definitions
6227: \ type "order" to see the problem
6228: @end example
6229:
6230: The problem with this example is that @code{order} has no way to
6231: associate the name @code{my-new-words} with the wid of the word list (in
6232: Gforth, @code{order} and @code{vocs} will display @code{???} for a wid
6233: that has no associated name). There is no Standard way of associating a
6234: name with a wid.
6235:
6236: In Gforth, this example can be re-coded using @code{vocabulary}, which
6237: associates a name with a wid:
6238:
6239: @example
6240: vocabulary my-new-words
6241:
6242: \ add it to the search order
6243: my-new-words
6244:
6245: \ alternatively, add it to the search order and make it
6246: \ the compilation word list
6247: my-new-words definitions
6248: \ type "order" to see that the problem is solved
6249: @end example
6250:
6251:
6252: @node Environmental Queries, Files, Word Lists, Words
6253: @section Environmental Queries
6254: @cindex environmental queries
6255: @comment TODO more index entries
6256:
6257: The ANS Standard introduced the idea of "environmental queries" as a way
6258: for a program running on a system to determine certain characteristics of the system.
6259: The Standard specifies a number of strings that might be recognised by a system.
6260:
6261: The Standard requires that the name space used for environmental queries
6262: be distinct from the name space used for definitions.
6263:
6264: Typically, environmental queries are supported by creating a set of
6265: definitions in a word set that is @var{only} used during environmental
6266: queries; that is what Gforth does. There is no Standard way of adding
6267: definitions to the set of recognised environmental queries, but any
6268: implementation that supports the loading of optional word sets must have
6269: some mechanism for doing this (after loading the word set, the
6270: associated environmental query string must return @code{true}). In
6271: Gforth, the word set used to honour environmental queries can be
6272: manipulated just like any other word set.
6273:
6274: doc-environment?
6275: doc-environment-wordlist
6276:
6277: doc-gforth
6278: doc-os-class
6279:
6280: Note that, whilst the documentation for (eg) @code{gforth} shows it
6281: returning two items on the stack, querying it using @code{environment?}
6282: will return an additional item; the @code{true} flag that shows that the
6283: string was recognised.
1.1 anton 6284:
1.21 crook 6285: TODO Document the standard strings or note where they are documented herein
6286:
6287: Here are some examples of using environmental queries:
6288:
6289: @example
6290: s" address-unit-bits" environment? 0=
6291: [IF]
6292: cr .( environmental attribute address-units-bits unknown... ) cr
6293: [THEN]
6294:
6295: s" block" environment? [IF] DROP include block.fs [THEN]
6296:
6297: s" gforth" environment? [IF] 2DROP include compat/vocabulary.fs [THEN]
6298:
6299: s" gforth" environment? [IF] .( Gforth version ) TYPE [ELSE] .( Not Gforth..) [THEN]
6300:
6301: @end example
6302:
6303:
6304: Here is an example of adding a definition to the environment word list:
6305:
6306: @example
6307: get-current environment-wordlist set-current
6308: true constant block
6309: true constant block-ext
6310: set-current
6311: @end example
6312:
6313: You can see what definitions are in the environment word list like this:
6314:
6315: @example
6316: get-order 1+ environment-wordlist swap set-order words previous
6317: @end example
6318:
6319:
6320:
6321: @node Files, Including Files, Environmental Queries, Words
1.1 anton 6322: @section Files
6323:
1.20 pazsan 6324: This chapter describes how to operate on files from Forth.
6325:
1.21 crook 6326: Files are opened/created by name and type. The following types are
6327: recognised:
1.20 pazsan 6328:
6329: doc-r/o
6330: doc-r/w
6331: doc-w/o
6332: doc-bin
6333:
1.21 crook 6334: When a file is opened/created, it returns a file identifier,
6335: @var{wfileid} that is used for all other file commands. All file
6336: commands also return a status value, @var{wior}, that is 0 for a
6337: successful operation and an implementation-defined non-zero value in the
6338: case of an error.
1.20 pazsan 6339:
6340: doc-open-file
6341: doc-create-file
6342:
6343: doc-close-file
6344: doc-delete-file
6345: doc-rename-file
6346: doc-read-file
6347: doc-read-line
6348: doc-write-file
1.21 crook 6349: doc-write-line
1.20 pazsan 6350: doc-emit-file
6351: doc-flush-file
6352:
6353: doc-file-status
6354: doc-file-position
6355: doc-reposition-file
6356: doc-file-size
6357: doc-resize-file
6358:
1.12 anton 6359: @node Including Files, Blocks, Files, Words
6360: @section Including Files
6361: @cindex including files
6362:
6363: @menu
6364: * Words for Including::
6365: * Search Path::
1.21 crook 6366: * Forth Search Paths::
1.12 anton 6367: * General Search Paths::
6368: @end menu
6369:
6370: @node Words for Including, Search Path, Including Files, Including Files
6371: @subsection Words for Including
6372:
6373: doc-include-file
6374: doc-included
6375: doc-include
6376:
6377: Usually you want to include a file only if it is not included already
6378: (by, say, another source file):
1.21 crook 6379: @comment TODO describe what happens on error. Describes how the require
6380: @comment stuff works and describe how to clear/reset the history (eg
6381: @comment for debug). Might want to include that in the MARKER example.
1.12 anton 6382:
6383: doc-required
6384: doc-require
6385: doc-needs
6386:
1.21 crook 6387: A definition in ANS Standard Forth for @code{required} is provided in
6388: @file{compat/required.fs}.
6389:
1.12 anton 6390: @cindex stack effect of included files
6391: @cindex including files, stack effect
6392: I recommend that you write your source files such that interpreting them
6393: does not change the stack. This allows using these files with
6394: @code{required} and friends without complications. E.g.,
6395:
6396: @example
6397: 1 require foo.fs drop
6398: @end example
6399:
1.21 crook 6400: @node Search Path, Forth Search Paths, Words for Including, Including Files
1.12 anton 6401: @subsection Search Path
6402: @cindex path for @code{included}
6403: @cindex file search path
6404: @cindex include search path
6405: @cindex search path for files
6406:
1.21 crook 6407: @comment what uses these search paths.. just inc;lude and friends?
1.12 anton 6408: If you specify an absolute filename (i.e., a filename starting with
6409: @file{/} or @file{~}, or with @file{:} in the second position (as in
6410: @samp{C:...})) for @code{included} and friends, that file is included
6411: just as you would expect.
6412:
6413: For relative filenames, Gforth uses a search path similar to Forth's
1.21 crook 6414: search order (@pxref{Word Lists}). It tries to find the given filename in
1.12 anton 6415: the directories present in the path, and includes the first one it
6416: finds.
6417:
6418: If the search path contains the directory @file{.} (as it should), this
6419: refers to the directory that the present file was @code{included}
6420: from. This allows files to include other files relative to their own
6421: position (irrespective of the current working directory or the absolute
6422: position). This feature is essential for libraries consisting of
6423: several files, where a file may include other files from the library.
6424: It corresponds to @code{#include "..."} in C. If the current input
6425: source is not a file, @file{.} refers to the directory of the innermost
6426: file being included, or, if there is no file being included, to the
6427: current working directory.
6428:
6429: Use @file{~+} to refer to the current working directory (as in the
6430: @code{bash}).
6431:
6432: If the filename starts with @file{./}, the search path is not searched
6433: (just as with absolute filenames), and the @file{.} has the same meaning
6434: as described above.
6435:
1.21 crook 6436: @node Forth Search Paths, General Search Paths, Search Path, Including Files
6437: @subsection Forth Search Paths
6438: @cindex search path control - forth
1.12 anton 6439:
6440: The search path is initialized when you start Gforth (@pxref{Invoking
6441: Gforth}). You can display it with
6442:
6443: doc-.fpath
6444:
6445: You can change it later with the following words:
6446:
6447: doc-fpath+
6448: doc-fpath=
6449:
6450: Using fpath and require would look like:
6451:
6452: @example
6453: fpath= /usr/lib/forth/|./
6454:
6455: require timer.fs
6456: @end example
6457:
6458: If you have the need to look for a file in the Forth search path, you could
1.21 crook 6459: use this Gforth feature in your application:
1.12 anton 6460:
6461: doc-open-fpath-file
6462:
1.21 crook 6463: @node General Search Paths, , Forth Search Paths, Including Files
1.12 anton 6464: @subsection General Search Paths
1.21 crook 6465: @cindex search path control - for user applications
1.12 anton 6466:
6467: Your application may need to search files in sevaral directories, like
6468: @code{included} does. For this purpose you can define and use your own
6469: search paths. Create a search path like this:
6470:
6471: @example
1.21 crook 6472: \ Make a buffer for the path:
1.12 anton 6473: create mypath 100 chars , \ maximum length (is checked)
6474: 0 , \ real len
6475: 100 chars allot \ space for path
6476: @end example
6477:
6478: You have the same functions for the forth search path in a generic version
6479: for different paths.
6480:
1.21 crook 6481: Gforth also provides generic equivalents of the Forth search path words:
6482:
6483: doc-.path
1.12 anton 6484: doc-path+
6485: doc-path=
6486: doc-open-path-file
6487:
6488:
6489: @node Blocks, Other I/O, Including Files, Words
1.1 anton 6490: @section Blocks
6491:
1.20 pazsan 6492: This chapter describes how to use block files within Gforth.
6493:
6494: Block files are traditionally means of data and source storage in
6495: Forth. They have been very important in resource-starved computers
6496: without OS in the past. Gforth doesn't encourage to use blocks as
6497: source, and provides blocks only for backward compatibility. The ANS
6498: standard requires blocks to be available when files are.
6499:
1.21 crook 6500: @comment TODO what about errors on open-blocks?
1.20 pazsan 6501: doc-open-blocks
6502: doc-use
1.21 crook 6503: doc-scr
6504: doc-blk
1.20 pazsan 6505: doc-get-block-fid
6506: doc-block-position
6507: doc-update
1.21 crook 6508: doc-save-buffers
1.20 pazsan 6509: doc-save-buffer
1.21 crook 6510: doc-empty-buffers
1.20 pazsan 6511: doc-empty-buffer
6512: doc-flush
6513: doc-get-buffer
1.21 crook 6514: doc---block-block
1.20 pazsan 6515: doc-buffer
6516: doc-updated?
6517: doc-list
6518: doc-load
6519: doc-thru
6520: doc-+load
6521: doc-+thru
6522: doc---block--->
6523: doc-block-included
6524:
1.1 anton 6525: @node Other I/O, Programming Tools, Blocks, Words
6526: @section Other I/O
1.21 crook 6527: @comment TODO more index entries
6528:
6529: @menu
6530: * Simple numeric output:: Predefined formats
6531: * Formatted numeric output:: Formatted (pictured) output
6532: * String Formats:: How Forth stores strings in memory
6533: * Displaying characters and strings:: Other stuff
6534: * Input:: Input
6535: @end menu
6536:
6537: @node Simple numeric output, Formatted numeric output, Other I/O, Other I/O
6538: @subsection Simple numeric output
6539: @cindex Simple numeric output
6540: @comment TODO more index entries
6541:
6542: The simplest output functions are those that display numbers from the
6543: data or floating-point stacks. Floating-point output is always displayed
6544: using base 10. Numbers displayed from the data stack use the value stored
6545: in @code{base}.
6546:
6547: doc-.
6548: doc-dec.
6549: doc-hex.
6550: doc-u.
6551: doc-.r
6552: doc-u.r
6553: doc-d.
6554: doc-ud.
6555: doc-d.r
6556: doc-ud.r
6557: doc-f.
6558: doc-fe.
6559: doc-fs.
6560:
6561: Examples of printing the number 1234.5678E23 in the different floating-point output
6562: formats are shown below:
6563:
6564: @example
6565: f. 123456779999999000000000000.
6566: fe. 123.456779999999E24
6567: fs. 1.23456779999999E26
6568: @end example
6569:
6570:
6571: @node Formatted numeric output, String Formats, Simple numeric output, Other I/O
6572: @subsection Formatted numeric output
6573: @cindex Formatted numeric output
6574: @cindex pictured numeric output
6575: @comment TODO more index entries
6576:
6577: Forth traditionally uses a technique called @var{pictured numeric
6578: output} for formatted printing of integers. In this technique,
6579: digits are extracted from the number (using the current output radix
6580: defined by @code{base}), converted to ASCII codes and appended to a
6581: string that is built in a scratch-pad area of memory
6582: (@pxref{core-idef,Implementation-defined options}). During the extraction
6583: sequence, other arbitrary characters can be appended to the string. The
6584: completed string is specified by an address and length and can
6585: be manipulated (@code{TYPE}ed, copied, modified) under program control.
6586:
6587: All of the words described in the previous section for simple numeric
6588: output are implemented in Gforth using pictured numeric output.
6589:
6590: Three important things to remember about Pictured Numeric Output:
6591:
6592: @itemize @bullet
6593: @item
6594: It always operates on double-precision numbers; to display a single-precision number,
6595: convert it first (@pxref{Double precision} for ways of doing this).
6596: @item
6597: It always treats the double-precision number as though it were unsigned. Refer to
6598: the examples below for ways of printing signed numbers.
6599: @item
6600: The string is built up from right to left; least significant digit first.
6601: @end itemize
6602:
6603: doc-<#
6604: doc-#
6605: doc-#s
6606: doc-hold
6607: doc-sign
6608: doc-#>
6609:
6610: doc-represent
6611:
6612: Here are some examples of using pictured numeric output:
6613:
6614: @example
6615: : my-u. ( u -- )
6616: \ Simplest use of pns.. behaves like Standard u.
6617: 0 \ convert to unsigned double
6618: <# \ start conversion
6619: #s \ convert all digits
6620: #> \ complete conversion
6621: TYPE SPACE ; \ display, with trailing space
6622:
6623: : cents-only ( u -- )
6624: 0 \ convert to unsigned double
6625: <# \ start conversion
6626: # # \ convert two least-significant digits
6627: #> \ complete conversion, discard other digits
6628: TYPE SPACE ; \ display, with trailing space
6629:
6630: : dollars-and-cents ( u -- )
6631: 0 \ convert to unsigned double
6632: <# \ start conversion
6633: # # \ convert two least-significant digits
6634: [char] . hold \ insert decimal point
6635: #s \ convert remaining digits
6636: [char] $ hold \ append currency symbol
6637: #> \ complete conversion
6638: TYPE SPACE ; \ display, with trailing space
6639:
6640: : my-. ( n -- )
6641: \ handling negatives.. behaves like Standard .
6642: s>d \ convert to signed double
6643: swap over dabs \ leave sign byte followed by unsigned double
6644: <# \ start conversion
6645: #s \ convert all digits
6646: rot sign \ get at sign byte, append "-" if needed
6647: #> \ complete conversion
6648: TYPE SPACE ; \ display, with trailing space
6649:
6650: : account. ( n -- )
6651: \ accountants don't like minus signs, they use braces
6652: \ for negative numbers
6653: s>d \ convert to signed double
6654: swap over dabs \ leave sign byte followed by unsigned double
6655: <# \ start conversion
6656: 2 pick \ get copy of sign byte
6657: 0< IF [char] ) hold THEN \ right-most character of output
6658: #s \ convert all digits
6659: rot \ get at sign byte
6660: 0< IF [char] ( hold THEN
6661: #> \ complete conversion
6662: TYPE SPACE ; \ display, with trailing space
6663: @end example
6664:
6665: Here are some examples of using these words:
6666:
6667: @example
6668: 1 my-u. 1
6669: hex -1 my-u. decimal FFFFFFFF
6670: 1 cents-only 01
6671: 1234 cents-only 34
6672: 2 dollars-and-cents $0.02
6673: 1234 dollars-and-cents $12.34
6674: 123 my-. 123
6675: -123 my. -123
6676: 123 account. 123
6677: -456 account. (456)
6678: @end example
6679:
6680:
6681: @node String Formats, Displaying characters and strings, Formatted numeric output, Other I/O
6682: @subsection String Formats
6683: @cindex string formats
6684:
6685: @comment TODO more index entries
6686:
6687: Forth commonly uses two different methods for representing a string:
6688:
6689: @itemize @bullet
6690: @item
6691: @cindex address of counted string
6692: As a @var{counted string}, represented by a c-addr. The char addressed
6693: by c-addr contains a character-count, n, of the string and the string
6694: occupies the subsequent n char addresses in memory.
6695: @item
6696: As cell pair on the stack; c-addr u, where u is the length of the string
6697: in characters, and c-addr is the address of the first byte of the string.
6698: @end itemize
6699:
6700: The ANS Forth Standard encourages the use of the second format when
6701: representing strings on the stack, whilst conceeding that the counted
6702: string format remains useful as a way of storing strings in memory.
6703:
6704: doc-count
6705:
6706: @xref{Memory Blocks} for words that move, copy and search
6707: for strings. @xref{Displaying characters and strings,} for words that
6708: display characters and strings.
6709:
6710:
6711: @node Displaying characters and strings, Input, String Formats, Other I/O
6712: @subsection Displaying characters and strings
6713: @cindex displaying characters and strings
6714: @cindex compiling characters and strings
6715: @cindex cursor control
6716:
6717: @comment TODO more index entries
6718:
6719: This section starts with a glossary of Forth words and ends with a set
6720: of examples.
6721:
6722: doc-bl
6723: doc-space
6724: doc-spaces
6725: doc-emit
6726: doc-."
6727: doc-.(
6728: doc-type
6729: doc-cr
6730: doc-at-xy
6731: doc-page
6732: doc-s"
6733: doc-c"
6734: doc-char
6735: doc-[char]
6736: doc-sliteral
6737:
6738: As an example, consider the following text, stored in a file @file{test.fs}:
6739:
6740: @example
6741: .( text-1)
6742: : my-word
6743: ." text-2" cr
6744: .( text-3)
6745: ;
6746:
6747: ." text-4"
6748:
6749: : my-char
6750: [char] ALPHABET emit
6751: char emit
6752: ;
6753: @end example
6754:
6755: When you load this code into Gforth, the following output is generated:
6756:
6757: @example
6758: @kbd{include test.fs} text-1text-3text-4 ok
6759: @end example
6760:
6761: @itemize @bullet
6762: @item
6763: Messages @code{text-1} and @code{text-3} are displayed because @code{.(}
6764: is an immediate word; it behaves in the same way whether it is used inside
6765: or outside a colon definition.
6766: @item
6767: Message @code{text-4} is displayed because of Gforth's added interpretation
6768: semantics for @code{."}.
6769: @item
6770: Message @code{text-2} is @var{not} displayed, because the text interpreter
6771: performs the compilation semantics for @code{."} within the definition of
6772: @code{my-word}.
6773: @end itemize
6774:
6775: Here are some examples of executing @code{my-word} and @code{my-char}:
6776:
6777: @example
6778: my-word text-2
6779: ok
6780: @kbd{my-char fred} Af ok
6781: @kbd{my-char jim} Aj ok
6782: @end example
6783:
6784: @itemize @bullet
6785: @item
6786: Message @code{text-2} is displayed because of the run-time behaviour of
6787: @code{."}.
6788: @item
6789: @code{[char]} compiles the "A" from "ALPHABET" and puts its display code
6790: on the stack at run-time. @code{emit} always displays the character
6791: when @code{my-char} is executed.
6792: @item
6793: @code{char} parses a string at run-time and the second @code{emit} displays
6794: the first character of the string.
6795: @item
6796: If you type @code{see my-char} you can see that @code{[char]} discarded
6797: the text "LPHABET" and only compiled the display code for "A" into the
6798: definition of @code{my-char}.
6799: @end itemize
6800:
6801:
6802:
6803: @node Input, , Displaying characters and strings, Other I/O
6804: @subsection Input
6805: @cindex Input
6806: @comment TODO more index entries
6807:
6808: Blah on traditional and recommended string formats.
6809:
6810: doc-tib
6811: doc-#tib
6812: doc--trailing
6813: doc-/string
6814: doc-convert
6815: doc->number
6816: doc->float
6817: doc-accept
6818: doc-query
6819: doc-expect
6820: doc-evaluate
6821: doc-key
6822: doc-key?
6823:
6824: TODO reference the block move stuff elsewhere
6825:
6826: TODO convert and >number might be better in the numeric input section.
6827:
6828: TODO maybe some of these shouldn't be here but should be in a "parsing" section
6829:
1.1 anton 6830:
1.7 pazsan 6831: @node Programming Tools, Assembler and Code Words, Other I/O, Words
1.1 anton 6832: @section Programming Tools
6833: @cindex programming tools
6834:
6835: @menu
6836: * Debugging:: Simple and quick.
6837: * Assertions:: Making your programs self-checking.
1.6 pazsan 6838: * Singlestep Debugger:: Executing your program word by word.
1.1 anton 6839: @end menu
6840:
6841: @node Debugging, Assertions, Programming Tools, Programming Tools
6842: @subsection Debugging
6843: @cindex debugging
6844:
1.21 crook 6845: Languages with a slow edit/compile/link/test development loop tend to
6846: require sophisticated tracing/stepping debuggers to facilate
6847: productive debugging.
1.1 anton 6848:
6849: A much better (faster) way in fast-compiling languages is to add
6850: printing code at well-selected places, let the program run, look at
6851: the output, see where things went wrong, add more printing code, etc.,
6852: until the bug is found.
6853:
1.21 crook 6854: The simple debugging aids provided in @file{debugs.fs}
6855: are meant to support this style of debugging. In addition, there are
6856: words for non-destructively inspecting the stack and memory:
6857:
6858: doc-.s
6859: doc-f.s
6860:
6861: There is a word @code{.r} but it does @var{not} display the return
6862: stack! It is used for formatted numeric output.
6863:
6864: doc-depth
6865: doc-fdepth
6866: doc-clearstack
6867: doc-?
6868: doc-dump
6869:
6870: The word @code{~~} prints debugging information (by default the source
6871: location and the stack contents). It is easy to insert. If you use Emacs
6872: it is also easy to remove (@kbd{C-x ~} in the Emacs Forth mode to
1.1 anton 6873: query-replace them with nothing). The deferred words
6874: @code{printdebugdata} and @code{printdebugline} control the output of
6875: @code{~~}. The default source location output format works well with
6876: Emacs' compilation mode, so you can step through the program at the
6877: source level using @kbd{C-x `} (the advantage over a stepping debugger
6878: is that you can step in any direction and you know where the crash has
6879: happened or where the strange data has occurred).
6880:
6881: Note that the default actions clobber the contents of the pictured
6882: numeric output string, so you should not use @code{~~}, e.g., between
6883: @code{<#} and @code{#>}.
6884:
6885: doc-~~
6886: doc-printdebugdata
6887: doc-printdebugline
6888:
1.21 crook 6889: doc-see
6890: doc-marker
6891:
6892: Here's an example of using @code{marker} at the start of a source file
6893: that you are debugging; it ensures that you only ever have one copy of
6894: the file's definitions compiled at any time:
6895:
6896: @example
6897: [IFDEF] my-code
6898: my-code
6899: [ENDIF]
6900:
6901: marker my-code
6902:
6903: \ .. definitions start here
6904: \ .
6905: \ .
6906: \ end
6907: @end example
6908:
6909:
6910:
1.2 jwilke 6911: @node Assertions, Singlestep Debugger, Debugging, Programming Tools
1.1 anton 6912: @subsection Assertions
6913: @cindex assertions
6914:
6915: It is a good idea to make your programs self-checking, in particular, if
6916: you use an assumption (e.g., that a certain field of a data structure is
6917: never zero) that may become wrong during maintenance. Gforth supports
6918: assertions for this purpose. They are used like this:
6919:
6920: @example
6921: assert( @var{flag} )
6922: @end example
6923:
6924: The code between @code{assert(} and @code{)} should compute a flag, that
6925: should be true if everything is alright and false otherwise. It should
6926: not change anything else on the stack. The overall stack effect of the
6927: assertion is @code{( -- )}. E.g.
6928:
6929: @example
6930: assert( 1 1 + 2 = ) \ what we learn in school
6931: assert( dup 0<> ) \ assert that the top of stack is not zero
6932: assert( false ) \ this code should not be reached
6933: @end example
6934:
6935: The need for assertions is different at different times. During
6936: debugging, we want more checking, in production we sometimes care more
6937: for speed. Therefore, assertions can be turned off, i.e., the assertion
6938: becomes a comment. Depending on the importance of an assertion and the
6939: time it takes to check it, you may want to turn off some assertions and
6940: keep others turned on. Gforth provides several levels of assertions for
6941: this purpose:
6942:
6943: doc-assert0(
6944: doc-assert1(
6945: doc-assert2(
6946: doc-assert3(
6947: doc-assert(
6948: doc-)
6949:
6950: @code{Assert(} is the same as @code{assert1(}. The variable
6951: @code{assert-level} specifies the highest assertions that are turned
6952: on. I.e., at the default @code{assert-level} of one, @code{assert0(} and
6953: @code{assert1(} assertions perform checking, while @code{assert2(} and
6954: @code{assert3(} assertions are treated as comments.
6955:
6956: Note that the @code{assert-level} is evaluated at compile-time, not at
6957: run-time. I.e., you cannot turn assertions on or off at run-time, you
6958: have to set the @code{assert-level} appropriately before compiling a
6959: piece of code. You can compile several pieces of code at several
6960: @code{assert-level}s (e.g., a trusted library at level 1 and newly
6961: written code at level 3).
6962:
6963: doc-assert-level
6964:
6965: If an assertion fails, a message compatible with Emacs' compilation mode
6966: is produced and the execution is aborted (currently with @code{ABORT"}.
6967: If there is interest, we will introduce a special throw code. But if you
6968: intend to @code{catch} a specific condition, using @code{throw} is
6969: probably more appropriate than an assertion).
6970:
1.21 crook 6971: Definitions in ANS Standard Forth for these assertion words are provided
6972: in @file{compat/assert.fs}.
6973:
6974:
1.2 jwilke 6975: @node Singlestep Debugger, , Assertions, Programming Tools
6976: @subsection Singlestep Debugger
6977: @cindex singlestep Debugger
6978: @cindex debugging Singlestep
6979: @cindex @code{dbg}
6980: @cindex @code{BREAK:}
6981: @cindex @code{BREAK"}
6982:
6983: When a new word is created there's often the need to check whether it behaves
1.21 crook 6984: correctly or not. You can do this by typing @code{dbg badword}.
6985:
6986: doc-dbg
6987:
6988: This might look like:
6989:
1.2 jwilke 6990: @example
6991: : badword 0 DO i . LOOP ; ok
6992: 2 dbg badword
6993: : badword
6994: Scanning code...
6995:
6996: Nesting debugger ready!
6997:
6998: 400D4738 8049BC4 0 -> [ 2 ] 00002 00000
6999: 400D4740 8049F68 DO -> [ 0 ]
7000: 400D4744 804A0C8 i -> [ 1 ] 00000
7001: 400D4748 400C5E60 . -> 0 [ 0 ]
7002: 400D474C 8049D0C LOOP -> [ 0 ]
7003: 400D4744 804A0C8 i -> [ 1 ] 00001
7004: 400D4748 400C5E60 . -> 1 [ 0 ]
7005: 400D474C 8049D0C LOOP -> [ 0 ]
7006: 400D4758 804B384 ; -> ok
7007: @end example
7008:
1.5 anton 7009: Each line displayed is one step. You always have to hit return to
7010: execute the next word that is displayed. If you don't want to execute
7011: the next word in a whole, you have to type @kbd{n} for @code{nest}. Here is
7012: an overview what keys are available:
1.2 jwilke 7013:
7014: @table @i
7015:
1.4 anton 7016: @item <return>
1.5 anton 7017: Next; Execute the next word.
1.2 jwilke 7018:
7019: @item n
1.5 anton 7020: Nest; Single step through next word.
1.2 jwilke 7021:
7022: @item u
1.5 anton 7023: Unnest; Stop debugging and execute rest of word. If we got to this word
7024: with nest, continue debugging with the calling word.
1.2 jwilke 7025:
7026: @item d
1.5 anton 7027: Done; Stop debugging and execute rest.
1.2 jwilke 7028:
7029: @item s
1.5 anton 7030: Stopp; Abort immediately.
1.2 jwilke 7031:
7032: @end table
7033:
7034: Debugging large application with this mechanism is very difficult, because
7035: you have to nest very deep into the program before the interesting part
7036: begins. This takes a lot of time.
7037:
7038: To do it more directly put a @code{BREAK:} command into your source code.
7039: When program execution reaches @code{BREAK:} the single step debugger is
7040: invoked and you have all the features described above.
7041:
7042: If you have more than one part to debug it is useful to know where the
7043: program has stopped at the moment. You can do this by the
7044: @code{BREAK" string"} command. This behaves like @code{BREAK:} except that
7045: string is typed out when the ``breakpoint'' is reached.
7046:
1.7 pazsan 7047: @node Assembler and Code Words, Threading Words, Programming Tools, Words
7048: @section Assembler and Code Words
1.1 anton 7049: @cindex assembler
7050: @cindex code words
7051:
7052: Gforth provides some words for defining primitives (words written in
7053: machine code), and for defining the the machine-code equivalent of
7054: @code{DOES>}-based defining words. However, the machine-independent
7055: nature of Gforth poses a few problems: First of all, Gforth runs on
7056: several architectures, so it can provide no standard assembler. What's
7057: worse is that the register allocation not only depends on the processor,
7058: but also on the @code{gcc} version and options used.
7059:
7060: The words that Gforth offers encapsulate some system dependences (e.g., the
7061: header structure), so a system-independent assembler may be used in
7062: Gforth. If you do not have an assembler, you can compile machine code
7063: directly with @code{,} and @code{c,}.
7064:
7065: doc-assembler
7066: doc-code
7067: doc-end-code
7068: doc-;code
7069: doc-flush-icache
7070:
7071: If @code{flush-icache} does not work correctly, @code{code} words
7072: etc. will not work (reliably), either.
7073:
7074: These words are rarely used. Therefore they reside in @code{code.fs},
7075: which is usually not loaded (except @code{flush-icache}, which is always
7076: present). You can load them with @code{require code.fs}.
7077:
7078: @cindex registers of the inner interpreter
7079: In the assembly code you will want to refer to the inner interpreter's
7080: registers (e.g., the data stack pointer) and you may want to use other
7081: registers for temporary storage. Unfortunately, the register allocation
7082: is installation-dependent.
7083:
7084: The easiest solution is to use explicit register declarations
7085: (@pxref{Explicit Reg Vars, , Variables in Specified Registers, gcc.info,
7086: GNU C Manual}) for all of the inner interpreter's registers: You have to
7087: compile Gforth with @code{-DFORCE_REG} (configure option
7088: @code{--enable-force-reg}) and the appropriate declarations must be
7089: present in the @code{machine.h} file (see @code{mips.h} for an example;
7090: you can find a full list of all declarable register symbols with
7091: @code{grep register engine.c}). If you give explicit registers to all
7092: variables that are declared at the beginning of @code{engine()}, you
7093: should be able to use the other caller-saved registers for temporary
7094: storage. Alternatively, you can use the @code{gcc} option
7095: @code{-ffixed-REG} (@pxref{Code Gen Options, , Options for Code
7096: Generation Conventions, gcc.info, GNU C Manual}) to reserve a register
7097: (however, this restriction on register allocation may slow Gforth
7098: significantly).
7099:
7100: If this solution is not viable (e.g., because @code{gcc} does not allow
7101: you to explicitly declare all the registers you need), you have to find
7102: out by looking at the code where the inner interpreter's registers
7103: reside and which registers can be used for temporary storage. You can
7104: get an assembly listing of the engine's code with @code{make engine.s}.
7105:
7106: In any case, it is good practice to abstract your assembly code from the
7107: actual register allocation. E.g., if the data stack pointer resides in
7108: register @code{$17}, create an alias for this register called @code{sp},
7109: and use that in your assembly code.
7110:
7111: @cindex code words, portable
7112: Another option for implementing normal and defining words efficiently
7113: is: adding the wanted functionality to the source of Gforth. For normal
7114: words you just have to edit @file{primitives} (@pxref{Automatic
7115: Generation}), defining words (equivalent to @code{;CODE} words, for fast
1.21 crook 7116: defined words) may require changes in @file{engine.c}, @file{kernel.fs},
1.1 anton 7117: @file{prims2x.fs}, and possibly @file{cross.fs}.
7118:
7119:
1.21 crook 7120: @node Threading Words, Passing Commands to the OS, Assembler and Code Words, Words
1.1 anton 7121: @section Threading Words
7122: @cindex threading words
7123:
7124: @cindex code address
7125: These words provide access to code addresses and other threading stuff
7126: in Gforth (and, possibly, other interpretive Forths). It more or less
7127: abstracts away the differences between direct and indirect threading
7128: (and, for direct threading, the machine dependences). However, at
7129: present this wordset is still incomplete. It is also pretty low-level;
7130: some day it will hopefully be made unnecessary by an internals wordset
7131: that abstracts implementation details away completely.
7132:
1.21 crook 7133: doc-threading-method
1.1 anton 7134: doc->code-address
7135: doc->does-code
7136: doc-code-address!
7137: doc-does-code!
7138: doc-does-handler!
7139: doc-/does-handler
7140:
7141: The code addresses produced by various defining words are produced by
7142: the following words:
7143:
7144: doc-docol:
7145: doc-docon:
7146: doc-dovar:
7147: doc-douser:
7148: doc-dodefer:
7149: doc-dofield:
7150:
7151: You can recognize words defined by a @code{CREATE}...@code{DOES>} word
7152: with @code{>DOES-CODE}. If the word was defined in that way, the value
7153: returned is different from 0 and identifies the @code{DOES>} used by the
7154: defining word.
1.21 crook 7155: @comment TODO should that be "identifies the xt of the DOES> ??
7156:
7157: @node Passing Commands to the OS, Miscellaneous Words, Threading Words, Words
7158: @section Passing Commands to the Operating System
7159: @cindex operating system - passing commands
7160: @cindex shell commands
7161:
7162: Gforth allows you to pass an arbitrary string to the host operating
7163: system shell (if such a thing exists) for execution.
7164:
7165: doc-sh
7166: doc-system
7167: doc-$?
7168:
7169:
7170: @node Miscellaneous Words, , Passing Commands to the OS, Words
7171: @section Miscellaneous Words
7172: @cindex miscellaneous words
7173:
7174: These section lists the ANS Standard Forth words that are not documented
7175: elsewhere in this manual. Ultimately, they all need proper homes.
7176:
7177: doc-,
7178: doc-allocate
7179: doc-allot
7180: doc-c,
7181: doc-here
7182: doc-ms
7183: doc-pad
7184: doc-parse
7185: doc-postpone
7186: doc-resize
7187: doc-restore-input
7188: doc-save-input
7189: doc-source
7190: doc-source-id
7191: doc-span
7192: doc-time&date
7193: doc-unused
7194: doc-word
7195: doc-[compile]
7196:
7197: These ANS Standard Forth words are not currently implemented in Gforth
7198: (see TODO section on dependencies)
7199:
7200: The following ANS Standard Forth words are not currently supported by Gforth
7201: (@pxref{ANS conformance})
7202:
7203: @code{EDITOR}
7204: @code{EKEY}
7205: @code{EKEY>CHAR}
7206: @code{EKEY?}
7207: @code{EMIT?}
7208: @code{FORGET}
7209:
1.2 jwilke 7210:
1.5 anton 7211: @c ******************************************************************
1.1 anton 7212: @node Tools, ANS conformance, Words, Top
7213: @chapter Tools
7214:
7215: @menu
7216: * ANS Report:: Report the words used, sorted by wordset.
7217: @end menu
7218:
7219: See also @ref{Emacs and Gforth}.
7220:
7221: @node ANS Report, , Tools, Tools
7222: @section @file{ans-report.fs}: Report the words used, sorted by wordset
7223: @cindex @file{ans-report.fs}
7224: @cindex report the words used in your program
7225: @cindex words used in your program
7226:
7227: If you want to label a Forth program as ANS Forth Program, you must
7228: document which wordsets the program uses; for extension wordsets, it is
7229: helpful to list the words the program requires from these wordsets
7230: (because Forth systems are allowed to provide only some words of them).
7231:
7232: The @file{ans-report.fs} tool makes it easy for you to determine which
7233: words from which wordset and which non-ANS words your application
7234: uses. You simply have to include @file{ans-report.fs} before loading the
7235: program you want to check. After loading your program, you can get the
7236: report with @code{print-ans-report}. A typical use is to run this as
7237: batch job like this:
7238: @example
7239: gforth ans-report.fs myprog.fs -e "print-ans-report bye"
7240: @end example
7241:
7242: The output looks like this (for @file{compat/control.fs}):
7243: @example
7244: The program uses the following words
7245: from CORE :
7246: : POSTPONE THEN ; immediate ?dup IF 0=
7247: from BLOCK-EXT :
7248: \
7249: from FILE :
7250: (
7251: @end example
7252:
7253: @subsection Caveats
7254:
7255: Note that @file{ans-report.fs} just checks which words are used, not whether
7256: they are used in an ANS Forth conforming way!
7257:
7258: Some words are defined in several wordsets in the
7259: standard. @file{ans-report.fs} reports them for only one of the
7260: wordsets, and not necessarily the one you expect. It depends on usage
7261: which wordset is the right one to specify. E.g., if you only use the
7262: compilation semantics of @code{S"}, it is a Core word; if you also use
7263: its interpretation semantics, it is a File word.
7264:
7265: @c ******************************************************************
7266: @node ANS conformance, Model, Tools, Top
7267: @chapter ANS conformance
7268: @cindex ANS conformance of Gforth
7269:
7270: To the best of our knowledge, Gforth is an
7271:
7272: ANS Forth System
7273: @itemize @bullet
7274: @item providing the Core Extensions word set
7275: @item providing the Block word set
7276: @item providing the Block Extensions word set
7277: @item providing the Double-Number word set
7278: @item providing the Double-Number Extensions word set
7279: @item providing the Exception word set
7280: @item providing the Exception Extensions word set
7281: @item providing the Facility word set
7282: @item providing @code{MS} and @code{TIME&DATE} from the Facility Extensions word set
7283: @item providing the File Access word set
7284: @item providing the File Access Extensions word set
7285: @item providing the Floating-Point word set
7286: @item providing the Floating-Point Extensions word set
7287: @item providing the Locals word set
7288: @item providing the Locals Extensions word set
7289: @item providing the Memory-Allocation word set
7290: @item providing the Memory-Allocation Extensions word set (that one's easy)
7291: @item providing the Programming-Tools word set
7292: @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
7293: @item providing the Search-Order word set
7294: @item providing the Search-Order Extensions word set
7295: @item providing the String word set
7296: @item providing the String Extensions word set (another easy one)
7297: @end itemize
7298:
7299: @cindex system documentation
7300: In addition, ANS Forth systems are required to document certain
7301: implementation choices. This chapter tries to meet these
7302: requirements. In many cases it gives a way to ask the system for the
7303: information instead of providing the information directly, in
7304: particular, if the information depends on the processor, the operating
7305: system or the installation options chosen, or if they are likely to
7306: change during the maintenance of Gforth.
7307:
7308: @comment The framework for the rest has been taken from pfe.
7309:
7310: @menu
7311: * The Core Words::
7312: * The optional Block word set::
7313: * The optional Double Number word set::
7314: * The optional Exception word set::
7315: * The optional Facility word set::
7316: * The optional File-Access word set::
7317: * The optional Floating-Point word set::
7318: * The optional Locals word set::
7319: * The optional Memory-Allocation word set::
7320: * The optional Programming-Tools word set::
7321: * The optional Search-Order word set::
7322: @end menu
7323:
7324:
7325: @c =====================================================================
7326: @node The Core Words, The optional Block word set, ANS conformance, ANS conformance
7327: @comment node-name, next, previous, up
7328: @section The Core Words
7329: @c =====================================================================
7330: @cindex core words, system documentation
7331: @cindex system documentation, core words
7332:
7333: @menu
7334: * core-idef:: Implementation Defined Options
7335: * core-ambcond:: Ambiguous Conditions
7336: * core-other:: Other System Documentation
7337: @end menu
7338:
7339: @c ---------------------------------------------------------------------
7340: @node core-idef, core-ambcond, The Core Words, The Core Words
7341: @subsection Implementation Defined Options
7342: @c ---------------------------------------------------------------------
7343: @cindex core words, implementation-defined options
7344: @cindex implementation-defined options, core words
7345:
7346:
7347: @table @i
7348: @item (Cell) aligned addresses:
7349: @cindex cell-aligned addresses
7350: @cindex aligned addresses
7351: processor-dependent. Gforth's alignment words perform natural alignment
7352: (e.g., an address aligned for a datum of size 8 is divisible by
7353: 8). Unaligned accesses usually result in a @code{-23 THROW}.
7354:
7355: @item @code{EMIT} and non-graphic characters:
7356: @cindex @code{EMIT} and non-graphic characters
7357: @cindex non-graphic characters and @code{EMIT}
7358: The character is output using the C library function (actually, macro)
7359: @code{putc}.
7360:
7361: @item character editing of @code{ACCEPT} and @code{EXPECT}:
7362: @cindex character editing of @code{ACCEPT} and @code{EXPECT}
7363: @cindex editing in @code{ACCEPT} and @code{EXPECT}
7364: @cindex @code{ACCEPT}, editing
7365: @cindex @code{EXPECT}, editing
7366: This is modeled on the GNU readline library (@pxref{Readline
7367: Interaction, , Command Line Editing, readline, The GNU Readline
7368: Library}) with Emacs-like key bindings. @kbd{Tab} deviates a little by
7369: producing a full word completion every time you type it (instead of
7370: producing the common prefix of all completions).
7371:
7372: @item character set:
7373: @cindex character set
7374: The character set of your computer and display device. Gforth is
7375: 8-bit-clean (but some other component in your system may make trouble).
7376:
7377: @item Character-aligned address requirements:
7378: @cindex character-aligned address requirements
7379: installation-dependent. Currently a character is represented by a C
7380: @code{unsigned char}; in the future we might switch to @code{wchar_t}
7381: (Comments on that requested).
7382:
7383: @item character-set extensions and matching of names:
7384: @cindex character-set extensions and matching of names
7385: @cindex case sensitivity for name lookup
7386: @cindex name lookup, case sensitivity
7387: @cindex locale and case sensitivity
1.21 crook 7388: Any character except the ASCII NUL character can be used in a
1.1 anton 7389: name. Matching is case-insensitive (except in @code{TABLE}s). The
7390: matching is performed using the C function @code{strncasecmp}, whose
7391: function is probably influenced by the locale. E.g., the @code{C} locale
7392: does not know about accents and umlauts, so they are matched
7393: case-sensitively in that locale. For portability reasons it is best to
7394: write programs such that they work in the @code{C} locale. Then one can
7395: use libraries written by a Polish programmer (who might use words
7396: containing ISO Latin-2 encoded characters) and by a French programmer
7397: (ISO Latin-1) in the same program (of course, @code{WORDS} will produce
7398: funny results for some of the words (which ones, depends on the font you
7399: are using)). Also, the locale you prefer may not be available in other
7400: operating systems. Hopefully, Unicode will solve these problems one day.
7401:
7402: @item conditions under which control characters match a space delimiter:
7403: @cindex space delimiters
7404: @cindex control characters as delimiters
7405: If @code{WORD} is called with the space character as a delimiter, all
7406: white-space characters (as identified by the C macro @code{isspace()})
7407: are delimiters. @code{PARSE}, on the other hand, treats space like other
7408: delimiters. @code{PARSE-WORD} treats space like @code{WORD}, but behaves
7409: like @code{PARSE} otherwise. @code{(NAME)}, which is used by the outer
7410: interpreter (aka text interpreter) by default, treats all white-space
7411: characters as delimiters.
7412:
7413: @item format of the control flow stack:
7414: @cindex control flow stack, format
7415: The data stack is used as control flow stack. The size of a control flow
7416: stack item in cells is given by the constant @code{cs-item-size}. At the
7417: time of this writing, an item consists of a (pointer to a) locals list
7418: (third), an address in the code (second), and a tag for identifying the
7419: item (TOS). The following tags are used: @code{defstart},
7420: @code{live-orig}, @code{dead-orig}, @code{dest}, @code{do-dest},
7421: @code{scopestart}.
7422:
7423: @item conversion of digits > 35
7424: @cindex digits > 35
7425: The characters @code{[\]^_'} are the digits with the decimal value
7426: 36@minus{}41. There is no way to input many of the larger digits.
7427:
7428: @item display after input terminates in @code{ACCEPT} and @code{EXPECT}:
7429: @cindex @code{EXPECT}, display after end of input
7430: @cindex @code{ACCEPT}, display after end of input
7431: The cursor is moved to the end of the entered string. If the input is
7432: terminated using the @kbd{Return} key, a space is typed.
7433:
7434: @item exception abort sequence of @code{ABORT"}:
7435: @cindex exception abort sequence of @code{ABORT"}
7436: @cindex @code{ABORT"}, exception abort sequence
7437: The error string is stored into the variable @code{"error} and a
7438: @code{-2 throw} is performed.
7439:
7440: @item input line terminator:
7441: @cindex input line terminator
7442: @cindex line terminator on input
7443: @cindex newline charcter on input
7444: For interactive input, @kbd{C-m} (CR) and @kbd{C-j} (LF) terminate
7445: lines. One of these characters is typically produced when you type the
7446: @kbd{Enter} or @kbd{Return} key.
7447:
7448: @item maximum size of a counted string:
7449: @cindex maximum size of a counted string
7450: @cindex counted string, maximum size
7451: @code{s" /counted-string" environment? drop .}. Currently 255 characters
7452: on all ports, but this may change.
7453:
7454: @item maximum size of a parsed string:
7455: @cindex maximum size of a parsed string
7456: @cindex parsed string, maximum size
7457: Given by the constant @code{/line}. Currently 255 characters.
7458:
7459: @item maximum size of a definition name, in characters:
7460: @cindex maximum size of a definition name, in characters
7461: @cindex name, maximum length
7462: 31
7463:
7464: @item maximum string length for @code{ENVIRONMENT?}, in characters:
7465: @cindex maximum string length for @code{ENVIRONMENT?}, in characters
7466: @cindex @code{ENVIRONMENT?} string length, maximum
7467: 31
7468:
7469: @item method of selecting the user input device:
7470: @cindex user input device, method of selecting
7471: The user input device is the standard input. There is currently no way to
7472: change it from within Gforth. However, the input can typically be
7473: redirected in the command line that starts Gforth.
7474:
7475: @item method of selecting the user output device:
7476: @cindex user output device, method of selecting
7477: @code{EMIT} and @code{TYPE} output to the file-id stored in the value
1.10 anton 7478: @code{outfile-id} (@code{stdout} by default). Gforth uses unbuffered
7479: output when the user output device is a terminal, otherwise the output
7480: is buffered.
1.1 anton 7481:
7482: @item methods of dictionary compilation:
7483: What are we expected to document here?
7484:
7485: @item number of bits in one address unit:
7486: @cindex number of bits in one address unit
7487: @cindex address unit, size in bits
7488: @code{s" address-units-bits" environment? drop .}. 8 in all current
7489: ports.
7490:
7491: @item number representation and arithmetic:
7492: @cindex number representation and arithmetic
7493: Processor-dependent. Binary two's complement on all current ports.
7494:
7495: @item ranges for integer types:
7496: @cindex ranges for integer types
7497: @cindex integer types, ranges
7498: Installation-dependent. Make environmental queries for @code{MAX-N},
7499: @code{MAX-U}, @code{MAX-D} and @code{MAX-UD}. The lower bounds for
7500: unsigned (and positive) types is 0. The lower bound for signed types on
7501: two's complement and one's complement machines machines can be computed
7502: by adding 1 to the upper bound.
7503:
7504: @item read-only data space regions:
7505: @cindex read-only data space regions
7506: @cindex data-space, read-only regions
7507: The whole Forth data space is writable.
7508:
7509: @item size of buffer at @code{WORD}:
7510: @cindex size of buffer at @code{WORD}
7511: @cindex @code{WORD} buffer size
7512: @code{PAD HERE - .}. 104 characters on 32-bit machines. The buffer is
7513: shared with the pictured numeric output string. If overwriting
7514: @code{PAD} is acceptable, it is as large as the remaining dictionary
7515: space, although only as much can be sensibly used as fits in a counted
7516: string.
7517:
7518: @item size of one cell in address units:
7519: @cindex cell size
7520: @code{1 cells .}.
7521:
7522: @item size of one character in address units:
7523: @cindex char size
7524: @code{1 chars .}. 1 on all current ports.
7525:
7526: @item size of the keyboard terminal buffer:
7527: @cindex size of the keyboard terminal buffer
7528: @cindex terminal buffer, size
7529: Varies. You can determine the size at a specific time using @code{lp@@
7530: tib - .}. It is shared with the locals stack and TIBs of files that
7531: include the current file. You can change the amount of space for TIBs
7532: and locals stack at Gforth startup with the command line option
7533: @code{-l}.
7534:
7535: @item size of the pictured numeric output buffer:
7536: @cindex size of the pictured numeric output buffer
7537: @cindex pictured numeric output buffer, size
7538: @code{PAD HERE - .}. 104 characters on 32-bit machines. The buffer is
7539: shared with @code{WORD}.
7540:
7541: @item size of the scratch area returned by @code{PAD}:
7542: @cindex size of the scratch area returned by @code{PAD}
7543: @cindex @code{PAD} size
7544: The remainder of dictionary space. @code{unused pad here - - .}.
7545:
7546: @item system case-sensitivity characteristics:
7547: @cindex case-sensitivity characteristics
7548: Dictionary searches are case insensitive (except in
7549: @code{TABLE}s). However, as explained above under @i{character-set
7550: extensions}, the matching for non-ASCII characters is determined by the
7551: locale you are using. In the default @code{C} locale all non-ASCII
7552: characters are matched case-sensitively.
7553:
7554: @item system prompt:
7555: @cindex system prompt
7556: @cindex prompt
7557: @code{ ok} in interpret state, @code{ compiled} in compile state.
7558:
7559: @item division rounding:
7560: @cindex division rounding
7561: installation dependent. @code{s" floored" environment? drop .}. We leave
7562: the choice to @code{gcc} (what to use for @code{/}) and to you (whether
7563: to use @code{fm/mod}, @code{sm/rem} or simply @code{/}).
7564:
7565: @item values of @code{STATE} when true:
7566: @cindex @code{STATE} values
7567: -1.
7568:
7569: @item values returned after arithmetic overflow:
7570: On two's complement machines, arithmetic is performed modulo
7571: 2**bits-per-cell for single arithmetic and 4**bits-per-cell for double
7572: arithmetic (with appropriate mapping for signed types). Division by zero
7573: typically results in a @code{-55 throw} (Floating-point unidentified
7574: fault), although a @code{-10 throw} (divide by zero) would be more
7575: appropriate.
7576:
7577: @item whether the current definition can be found after @t{DOES>}:
7578: @cindex @t{DOES>}, visibility of current definition
7579: No.
7580:
7581: @end table
7582:
7583: @c ---------------------------------------------------------------------
7584: @node core-ambcond, core-other, core-idef, The Core Words
7585: @subsection Ambiguous conditions
7586: @c ---------------------------------------------------------------------
7587: @cindex core words, ambiguous conditions
7588: @cindex ambiguous conditions, core words
7589:
7590: @table @i
7591:
7592: @item a name is neither a word nor a number:
7593: @cindex name not found
7594: @cindex Undefined word
7595: @code{-13 throw} (Undefined word). Actually, @code{-13 bounce}, which
7596: preserves the data and FP stack, so you don't lose more work than
7597: necessary.
7598:
7599: @item a definition name exceeds the maximum length allowed:
7600: @cindex Word name too long
7601: @code{-19 throw} (Word name too long)
7602:
7603: @item addressing a region not inside the various data spaces of the forth system:
7604: @cindex Invalid memory address
7605: The stacks, code space and name space are accessible. Machine code space is
7606: typically readable. Accessing other addresses gives results dependent on
7607: the operating system. On decent systems: @code{-9 throw} (Invalid memory
7608: address).
7609:
7610: @item argument type incompatible with parameter:
7611: @cindex Argument type mismatch
7612: This is usually not caught. Some words perform checks, e.g., the control
7613: flow words, and issue a @code{ABORT"} or @code{-12 THROW} (Argument type
7614: mismatch).
7615:
7616: @item attempting to obtain the execution token of a word with undefined execution semantics:
7617: @cindex Interpreting a compile-only word, for @code{'} etc.
7618: @cindex execution token of words with undefined execution semantics
7619: @code{-14 throw} (Interpreting a compile-only word). In some cases, you
7620: get an execution token for @code{compile-only-error} (which performs a
7621: @code{-14 throw} when executed).
7622:
7623: @item dividing by zero:
7624: @cindex dividing by zero
7625: @cindex floating point unidentified fault, integer division
7626: @cindex divide by zero
7627: typically results in a @code{-55 throw} (floating point unidentified
7628: fault), although a @code{-10 throw} (divide by zero) would be more
7629: appropriate.
7630:
7631: @item insufficient data stack or return stack space:
7632: @cindex insufficient data stack or return stack space
7633: @cindex stack overflow
7634: @cindex Address alignment exception, stack overflow
7635: @cindex Invalid memory address, stack overflow
7636: Depending on the operating system, the installation, and the invocation
7637: of Gforth, this is either checked by the memory management hardware, or
7638: it is not checked. If it is checked, you typically get a @code{-9 throw}
7639: (Invalid memory address) as soon as the overflow happens. If it is not
1.21 crook 7640: checked, overflows typically result in mysterious illegal memory accesses,
1.1 anton 7641: producing @code{-9 throw} (Invalid memory address) or @code{-23 throw}
7642: (Address alignment exception); they might also destroy the internal data
7643: structure of @code{ALLOCATE} and friends, resulting in various errors in
7644: these words.
7645:
7646: @item insufficient space for loop control parameters:
7647: @cindex insufficient space for loop control parameters
7648: like other return stack overflows.
7649:
7650: @item insufficient space in the dictionary:
7651: @cindex insufficient space in the dictionary
7652: @cindex dictionary overflow
1.12 anton 7653: If you try to allot (either directly with @code{allot}, or indirectly
7654: with @code{,}, @code{create} etc.) more memory than available in the
7655: dictionary, you get a @code{-8 throw} (Dictionary overflow). If you try
7656: to access memory beyond the end of the dictionary, the results are
7657: similar to stack overflows.
1.1 anton 7658:
7659: @item interpreting a word with undefined interpretation semantics:
7660: @cindex interpreting a word with undefined interpretation semantics
7661: @cindex Interpreting a compile-only word
7662: For some words, we have defined interpretation semantics. For the
7663: others: @code{-14 throw} (Interpreting a compile-only word).
7664:
7665: @item modifying the contents of the input buffer or a string literal:
7666: @cindex modifying the contents of the input buffer or a string literal
7667: These are located in writable memory and can be modified.
7668:
7669: @item overflow of the pictured numeric output string:
7670: @cindex overflow of the pictured numeric output string
7671: @cindex pictured numeric output string, overflow
7672: Not checked. Runs into the dictionary and destroys it (at least,
7673: partially).
7674:
7675: @item parsed string overflow:
7676: @cindex parsed string overflow
7677: @code{PARSE} cannot overflow. @code{WORD} does not check for overflow.
7678:
7679: @item producing a result out of range:
7680: @cindex result out of range
7681: On two's complement machines, arithmetic is performed modulo
7682: 2**bits-per-cell for single arithmetic and 4**bits-per-cell for double
7683: arithmetic (with appropriate mapping for signed types). Division by zero
7684: typically results in a @code{-55 throw} (floatingpoint unidentified
7685: fault), although a @code{-10 throw} (divide by zero) would be more
7686: appropriate. @code{convert} and @code{>number} currently overflow
7687: silently.
7688:
7689: @item reading from an empty data or return stack:
7690: @cindex stack empty
7691: @cindex stack underflow
7692: The data stack is checked by the outer (aka text) interpreter after
7693: every word executed. If it has underflowed, a @code{-4 throw} (Stack
7694: underflow) is performed. Apart from that, stacks may be checked or not,
7695: depending on operating system, installation, and invocation. The
7696: consequences of stack underflows are similar to the consequences of
7697: stack overflows. Note that even if the system uses checking (through the
7698: MMU), your program may have to underflow by a significant number of
7699: stack items to trigger the reaction (the reason for this is that the
7700: MMU, and therefore the checking, works with a page-size granularity).
7701:
7702: @item unexpected end of the input buffer, resulting in an attempt to use a zero-length string as a name:
7703: @cindex unexpected end of the input buffer
7704: @cindex zero-length string as a name
7705: @cindex Attempt to use zero-length string as a name
7706: @code{Create} and its descendants perform a @code{-16 throw} (Attempt to
7707: use zero-length string as a name). Words like @code{'} probably will not
7708: find what they search. Note that it is possible to create zero-length
7709: names with @code{nextname} (should it not?).
7710:
7711: @item @code{>IN} greater than input buffer:
7712: @cindex @code{>IN} greater than input buffer
7713: The next invocation of a parsing word returns a string with length 0.
7714:
7715: @item @code{RECURSE} appears after @code{DOES>}:
7716: @cindex @code{RECURSE} appears after @code{DOES>}
7717: Compiles a recursive call to the defining word, not to the defined word.
7718:
7719: @item argument input source different than current input source for @code{RESTORE-INPUT}:
7720: @cindex argument input source different than current input source for @code{RESTORE-INPUT}
7721: @cindex Argument type mismatch, @code{RESTORE-INPUT}
7722: @cindex @code{RESTORE-INPUT}, Argument type mismatch
7723: @code{-12 THROW}. Note that, once an input file is closed (e.g., because
7724: the end of the file was reached), its source-id may be
7725: reused. Therefore, restoring an input source specification referencing a
7726: closed file may lead to unpredictable results instead of a @code{-12
7727: THROW}.
7728:
7729: In the future, Gforth may be able to restore input source specifications
7730: from other than the current input source.
7731:
7732: @item data space containing definitions gets de-allocated:
7733: @cindex data space containing definitions gets de-allocated
7734: Deallocation with @code{allot} is not checked. This typically results in
7735: memory access faults or execution of illegal instructions.
7736:
7737: @item data space read/write with incorrect alignment:
7738: @cindex data space read/write with incorrect alignment
7739: @cindex alignment faults
7740: @cindex Address alignment exception
7741: Processor-dependent. Typically results in a @code{-23 throw} (Address
1.12 anton 7742: alignment exception). Under Linux-Intel on a 486 or later processor with
1.1 anton 7743: alignment turned on, incorrect alignment results in a @code{-9 throw}
7744: (Invalid memory address). There are reportedly some processors with
1.12 anton 7745: alignment restrictions that do not report violations.
1.1 anton 7746:
7747: @item data space pointer not properly aligned, @code{,}, @code{C,}:
7748: @cindex data space pointer not properly aligned, @code{,}, @code{C,}
7749: Like other alignment errors.
7750:
7751: @item less than u+2 stack items (@code{PICK} and @code{ROLL}):
7752: Like other stack underflows.
7753:
7754: @item loop control parameters not available:
7755: @cindex loop control parameters not available
7756: Not checked. The counted loop words simply assume that the top of return
7757: stack items are loop control parameters and behave accordingly.
7758:
7759: @item most recent definition does not have a name (@code{IMMEDIATE}):
7760: @cindex most recent definition does not have a name (@code{IMMEDIATE})
7761: @cindex last word was headerless
7762: @code{abort" last word was headerless"}.
7763:
7764: @item name not defined by @code{VALUE} used by @code{TO}:
7765: @cindex name not defined by @code{VALUE} used by @code{TO}
7766: @cindex @code{TO} on non-@code{VALUE}s
7767: @cindex Invalid name argument, @code{TO}
7768: @code{-32 throw} (Invalid name argument) (unless name is a local or was
7769: defined by @code{CONSTANT}; in the latter case it just changes the constant).
7770:
7771: @item name not found (@code{'}, @code{POSTPONE}, @code{[']}, @code{[COMPILE]}):
7772: @cindex name not found (@code{'}, @code{POSTPONE}, @code{[']}, @code{[COMPILE]})
7773: @cindex Undefined word, @code{'}, @code{POSTPONE}, @code{[']}, @code{[COMPILE]}
7774: @code{-13 throw} (Undefined word)
7775:
7776: @item parameters are not of the same type (@code{DO}, @code{?DO}, @code{WITHIN}):
7777: @cindex parameters are not of the same type (@code{DO}, @code{?DO}, @code{WITHIN})
7778: Gforth behaves as if they were of the same type. I.e., you can predict
7779: the behaviour by interpreting all parameters as, e.g., signed.
7780:
7781: @item @code{POSTPONE} or @code{[COMPILE]} applied to @code{TO}:
7782: @cindex @code{POSTPONE} or @code{[COMPILE]} applied to @code{TO}
7783: Assume @code{: X POSTPONE TO ; IMMEDIATE}. @code{X} performs the
7784: compilation semantics of @code{TO}.
7785:
7786: @item String longer than a counted string returned by @code{WORD}:
7787: @cindex String longer than a counted string returned by @code{WORD}
7788: @cindex @code{WORD}, string overflow
7789: Not checked. The string will be ok, but the count will, of course,
7790: contain only the least significant bits of the length.
7791:
7792: @item u greater than or equal to the number of bits in a cell (@code{LSHIFT}, @code{RSHIFT}):
7793: @cindex @code{LSHIFT}, large shift counts
7794: @cindex @code{RSHIFT}, large shift counts
7795: Processor-dependent. Typical behaviours are returning 0 and using only
7796: the low bits of the shift count.
7797:
7798: @item word not defined via @code{CREATE}:
7799: @cindex @code{>BODY} of non-@code{CREATE}d words
7800: @code{>BODY} produces the PFA of the word no matter how it was defined.
7801:
7802: @cindex @code{DOES>} of non-@code{CREATE}d words
7803: @code{DOES>} changes the execution semantics of the last defined word no
7804: matter how it was defined. E.g., @code{CONSTANT DOES>} is equivalent to
7805: @code{CREATE , DOES>}.
7806:
7807: @item words improperly used outside @code{<#} and @code{#>}:
7808: Not checked. As usual, you can expect memory faults.
7809:
7810: @end table
7811:
7812:
7813: @c ---------------------------------------------------------------------
7814: @node core-other, , core-ambcond, The Core Words
7815: @subsection Other system documentation
7816: @c ---------------------------------------------------------------------
7817: @cindex other system documentation, core words
7818: @cindex core words, other system documentation
7819:
7820: @table @i
7821: @item nonstandard words using @code{PAD}:
7822: @cindex @code{PAD} use by nonstandard words
7823: None.
7824:
7825: @item operator's terminal facilities available:
7826: @cindex operator's terminal facilities available
7827: After processing the command line, Gforth goes into interactive mode,
7828: and you can give commands to Gforth interactively. The actual facilities
7829: available depend on how you invoke Gforth.
7830:
7831: @item program data space available:
7832: @cindex program data space available
7833: @cindex data space available
7834: @code{UNUSED .} gives the remaining dictionary space. The total
7835: dictionary space can be specified with the @code{-m} switch
7836: (@pxref{Invoking Gforth}) when Gforth starts up.
7837:
7838: @item return stack space available:
7839: @cindex return stack space available
7840: You can compute the total return stack space in cells with
7841: @code{s" RETURN-STACK-CELLS" environment? drop .}. You can specify it at
7842: startup time with the @code{-r} switch (@pxref{Invoking Gforth}).
7843:
7844: @item stack space available:
7845: @cindex stack space available
7846: You can compute the total data stack space in cells with
7847: @code{s" STACK-CELLS" environment? drop .}. You can specify it at
7848: startup time with the @code{-d} switch (@pxref{Invoking Gforth}).
7849:
7850: @item system dictionary space required, in address units:
7851: @cindex system dictionary space required, in address units
7852: Type @code{here forthstart - .} after startup. At the time of this
7853: writing, this gives 80080 (bytes) on a 32-bit system.
7854: @end table
7855:
7856:
7857: @c =====================================================================
7858: @node The optional Block word set, The optional Double Number word set, The Core Words, ANS conformance
7859: @section The optional Block word set
7860: @c =====================================================================
7861: @cindex system documentation, block words
7862: @cindex block words, system documentation
7863:
7864: @menu
7865: * block-idef:: Implementation Defined Options
7866: * block-ambcond:: Ambiguous Conditions
7867: * block-other:: Other System Documentation
7868: @end menu
7869:
7870:
7871: @c ---------------------------------------------------------------------
7872: @node block-idef, block-ambcond, The optional Block word set, The optional Block word set
7873: @subsection Implementation Defined Options
7874: @c ---------------------------------------------------------------------
7875: @cindex implementation-defined options, block words
7876: @cindex block words, implementation-defined options
7877:
7878: @table @i
7879: @item the format for display by @code{LIST}:
7880: @cindex @code{LIST} display format
7881: First the screen number is displayed, then 16 lines of 64 characters,
7882: each line preceded by the line number.
7883:
7884: @item the length of a line affected by @code{\}:
7885: @cindex length of a line affected by @code{\}
7886: @cindex @code{\}, line length in blocks
7887: 64 characters.
7888: @end table
7889:
7890:
7891: @c ---------------------------------------------------------------------
7892: @node block-ambcond, block-other, block-idef, The optional Block word set
7893: @subsection Ambiguous conditions
7894: @c ---------------------------------------------------------------------
7895: @cindex block words, ambiguous conditions
7896: @cindex ambiguous conditions, block words
7897:
7898: @table @i
7899: @item correct block read was not possible:
7900: @cindex block read not possible
7901: Typically results in a @code{throw} of some OS-derived value (between
7902: -512 and -2048). If the blocks file was just not long enough, blanks are
7903: supplied for the missing portion.
7904:
7905: @item I/O exception in block transfer:
7906: @cindex I/O exception in block transfer
7907: @cindex block transfer, I/O exception
7908: Typically results in a @code{throw} of some OS-derived value (between
7909: -512 and -2048).
7910:
7911: @item invalid block number:
7912: @cindex invalid block number
7913: @cindex block number invalid
7914: @code{-35 throw} (Invalid block number)
7915:
7916: @item a program directly alters the contents of @code{BLK}:
7917: @cindex @code{BLK}, altering @code{BLK}
7918: The input stream is switched to that other block, at the same
7919: position. If the storing to @code{BLK} happens when interpreting
7920: non-block input, the system will get quite confused when the block ends.
7921:
7922: @item no current block buffer for @code{UPDATE}:
7923: @cindex @code{UPDATE}, no current block buffer
7924: @code{UPDATE} has no effect.
7925:
7926: @end table
7927:
7928: @c ---------------------------------------------------------------------
7929: @node block-other, , block-ambcond, The optional Block word set
7930: @subsection Other system documentation
7931: @c ---------------------------------------------------------------------
7932: @cindex other system documentation, block words
7933: @cindex block words, other system documentation
7934:
7935: @table @i
7936: @item any restrictions a multiprogramming system places on the use of buffer addresses:
7937: No restrictions (yet).
7938:
7939: @item the number of blocks available for source and data:
7940: depends on your disk space.
7941:
7942: @end table
7943:
7944:
7945: @c =====================================================================
7946: @node The optional Double Number word set, The optional Exception word set, The optional Block word set, ANS conformance
7947: @section The optional Double Number word set
7948: @c =====================================================================
7949: @cindex system documentation, double words
7950: @cindex double words, system documentation
7951:
7952: @menu
7953: * double-ambcond:: Ambiguous Conditions
7954: @end menu
7955:
7956:
7957: @c ---------------------------------------------------------------------
7958: @node double-ambcond, , The optional Double Number word set, The optional Double Number word set
7959: @subsection Ambiguous conditions
7960: @c ---------------------------------------------------------------------
7961: @cindex double words, ambiguous conditions
7962: @cindex ambiguous conditions, double words
7963:
7964: @table @i
7965: @item @var{d} outside of range of @var{n} in @code{D>S}:
7966: @cindex @code{D>S}, @var{d} out of range of @var{n}
7967: The least significant cell of @var{d} is produced.
7968:
7969: @end table
7970:
7971:
7972: @c =====================================================================
7973: @node The optional Exception word set, The optional Facility word set, The optional Double Number word set, ANS conformance
7974: @section The optional Exception word set
7975: @c =====================================================================
7976: @cindex system documentation, exception words
7977: @cindex exception words, system documentation
7978:
7979: @menu
7980: * exception-idef:: Implementation Defined Options
7981: @end menu
7982:
7983:
7984: @c ---------------------------------------------------------------------
7985: @node exception-idef, , The optional Exception word set, The optional Exception word set
7986: @subsection Implementation Defined Options
7987: @c ---------------------------------------------------------------------
7988: @cindex implementation-defined options, exception words
7989: @cindex exception words, implementation-defined options
7990:
7991: @table @i
7992: @item @code{THROW}-codes used in the system:
7993: @cindex @code{THROW}-codes used in the system
7994: The codes -256@minus{}-511 are used for reporting signals. The mapping
7995: from OS signal numbers to throw codes is -256@minus{}@var{signal}. The
7996: codes -512@minus{}-2047 are used for OS errors (for file and memory
7997: allocation operations). The mapping from OS error numbers to throw codes
7998: is -512@minus{}@code{errno}. One side effect of this mapping is that
7999: undefined OS errors produce a message with a strange number; e.g.,
8000: @code{-1000 THROW} results in @code{Unknown error 488} on my system.
8001: @end table
8002:
8003: @c =====================================================================
8004: @node The optional Facility word set, The optional File-Access word set, The optional Exception word set, ANS conformance
8005: @section The optional Facility word set
8006: @c =====================================================================
8007: @cindex system documentation, facility words
8008: @cindex facility words, system documentation
8009:
8010: @menu
8011: * facility-idef:: Implementation Defined Options
8012: * facility-ambcond:: Ambiguous Conditions
8013: @end menu
8014:
8015:
8016: @c ---------------------------------------------------------------------
8017: @node facility-idef, facility-ambcond, The optional Facility word set, The optional Facility word set
8018: @subsection Implementation Defined Options
8019: @c ---------------------------------------------------------------------
8020: @cindex implementation-defined options, facility words
8021: @cindex facility words, implementation-defined options
8022:
8023: @table @i
8024: @item encoding of keyboard events (@code{EKEY}):
8025: @cindex keyboard events, encoding in @code{EKEY}
8026: @cindex @code{EKEY}, encoding of keyboard events
8027: Not yet implemented.
8028:
8029: @item duration of a system clock tick:
8030: @cindex duration of a system clock tick
8031: @cindex clock tick duration
8032: System dependent. With respect to @code{MS}, the time is specified in
8033: microseconds. How well the OS and the hardware implement this, is
8034: another question.
8035:
8036: @item repeatability to be expected from the execution of @code{MS}:
8037: @cindex repeatability to be expected from the execution of @code{MS}
8038: @cindex @code{MS}, repeatability to be expected
8039: System dependent. On Unix, a lot depends on load. If the system is
8040: lightly loaded, and the delay is short enough that Gforth does not get
8041: swapped out, the performance should be acceptable. Under MS-DOS and
8042: other single-tasking systems, it should be good.
8043:
8044: @end table
8045:
8046:
8047: @c ---------------------------------------------------------------------
8048: @node facility-ambcond, , facility-idef, The optional Facility word set
8049: @subsection Ambiguous conditions
8050: @c ---------------------------------------------------------------------
8051: @cindex facility words, ambiguous conditions
8052: @cindex ambiguous conditions, facility words
8053:
8054: @table @i
8055: @item @code{AT-XY} can't be performed on user output device:
8056: @cindex @code{AT-XY} can't be performed on user output device
8057: Largely terminal dependent. No range checks are done on the arguments.
8058: No errors are reported. You may see some garbage appearing, you may see
8059: simply nothing happen.
8060:
8061: @end table
8062:
8063:
8064: @c =====================================================================
8065: @node The optional File-Access word set, The optional Floating-Point word set, The optional Facility word set, ANS conformance
8066: @section The optional File-Access word set
8067: @c =====================================================================
8068: @cindex system documentation, file words
8069: @cindex file words, system documentation
8070:
8071: @menu
8072: * file-idef:: Implementation Defined Options
8073: * file-ambcond:: Ambiguous Conditions
8074: @end menu
8075:
8076: @c ---------------------------------------------------------------------
8077: @node file-idef, file-ambcond, The optional File-Access word set, The optional File-Access word set
8078: @subsection Implementation Defined Options
8079: @c ---------------------------------------------------------------------
8080: @cindex implementation-defined options, file words
8081: @cindex file words, implementation-defined options
8082:
8083: @table @i
8084: @item file access methods used:
8085: @cindex file access methods used
8086: @code{R/O}, @code{R/W} and @code{BIN} work as you would
8087: expect. @code{W/O} translates into the C file opening mode @code{w} (or
8088: @code{wb}): The file is cleared, if it exists, and created, if it does
8089: not (with both @code{open-file} and @code{create-file}). Under Unix
8090: @code{create-file} creates a file with 666 permissions modified by your
8091: umask.
8092:
8093: @item file exceptions:
8094: @cindex file exceptions
8095: The file words do not raise exceptions (except, perhaps, memory access
8096: faults when you pass illegal addresses or file-ids).
8097:
8098: @item file line terminator:
8099: @cindex file line terminator
8100: System-dependent. Gforth uses C's newline character as line
8101: terminator. What the actual character code(s) of this are is
8102: system-dependent.
8103:
8104: @item file name format:
8105: @cindex file name format
8106: System dependent. Gforth just uses the file name format of your OS.
8107:
8108: @item information returned by @code{FILE-STATUS}:
8109: @cindex @code{FILE-STATUS}, returned information
8110: @code{FILE-STATUS} returns the most powerful file access mode allowed
8111: for the file: Either @code{R/O}, @code{W/O} or @code{R/W}. If the file
8112: cannot be accessed, @code{R/O BIN} is returned. @code{BIN} is applicable
8113: along with the returned mode.
8114:
8115: @item input file state after an exception when including source:
8116: @cindex exception when including source
8117: All files that are left via the exception are closed.
8118:
8119: @item @var{ior} values and meaning:
8120: @cindex @var{ior} values and meaning
8121: The @var{ior}s returned by the file and memory allocation words are
8122: intended as throw codes. They typically are in the range
8123: -512@minus{}-2047 of OS errors. The mapping from OS error numbers to
8124: @var{ior}s is -512@minus{}@var{errno}.
8125:
8126: @item maximum depth of file input nesting:
8127: @cindex maximum depth of file input nesting
8128: @cindex file input nesting, maximum depth
8129: limited by the amount of return stack, locals/TIB stack, and the number
8130: of open files available. This should not give you troubles.
8131:
8132: @item maximum size of input line:
8133: @cindex maximum size of input line
8134: @cindex input line size, maximum
8135: @code{/line}. Currently 255.
8136:
8137: @item methods of mapping block ranges to files:
8138: @cindex mapping block ranges to files
8139: @cindex files containing blocks
8140: @cindex blocks in files
8141: By default, blocks are accessed in the file @file{blocks.fb} in the
8142: current working directory. The file can be switched with @code{USE}.
8143:
8144: @item number of string buffers provided by @code{S"}:
8145: @cindex @code{S"}, number of string buffers
8146: 1
8147:
8148: @item size of string buffer used by @code{S"}:
8149: @cindex @code{S"}, size of string buffer
8150: @code{/line}. currently 255.
8151:
8152: @end table
8153:
8154: @c ---------------------------------------------------------------------
8155: @node file-ambcond, , file-idef, The optional File-Access word set
8156: @subsection Ambiguous conditions
8157: @c ---------------------------------------------------------------------
8158: @cindex file words, ambiguous conditions
8159: @cindex ambiguous conditions, file words
8160:
8161: @table @i
8162: @item attempting to position a file outside its boundaries:
8163: @cindex @code{REPOSITION-FILE}, outside the file's boundaries
8164: @code{REPOSITION-FILE} is performed as usual: Afterwards,
8165: @code{FILE-POSITION} returns the value given to @code{REPOSITION-FILE}.
8166:
8167: @item attempting to read from file positions not yet written:
8168: @cindex reading from file positions not yet written
8169: End-of-file, i.e., zero characters are read and no error is reported.
8170:
8171: @item @var{file-id} is invalid (@code{INCLUDE-FILE}):
8172: @cindex @code{INCLUDE-FILE}, @var{file-id} is invalid
8173: An appropriate exception may be thrown, but a memory fault or other
8174: problem is more probable.
8175:
8176: @item I/O exception reading or closing @var{file-id} (@code{INCLUDE-FILE}, @code{INCLUDED}):
8177: @cindex @code{INCLUDE-FILE}, I/O exception reading or closing @var{file-id}
8178: @cindex @code{INCLUDED}, I/O exception reading or closing @var{file-id}
8179: The @var{ior} produced by the operation, that discovered the problem, is
8180: thrown.
8181:
8182: @item named file cannot be opened (@code{INCLUDED}):
8183: @cindex @code{INCLUDED}, named file cannot be opened
8184: The @var{ior} produced by @code{open-file} is thrown.
8185:
8186: @item requesting an unmapped block number:
8187: @cindex unmapped block numbers
8188: There are no unmapped legal block numbers. On some operating systems,
8189: writing a block with a large number may overflow the file system and
8190: have an error message as consequence.
8191:
8192: @item using @code{source-id} when @code{blk} is non-zero:
8193: @cindex @code{SOURCE-ID}, behaviour when @code{BLK} is non-zero
8194: @code{source-id} performs its function. Typically it will give the id of
8195: the source which loaded the block. (Better ideas?)
8196:
8197: @end table
8198:
8199:
8200: @c =====================================================================
8201: @node The optional Floating-Point word set, The optional Locals word set, The optional File-Access word set, ANS conformance
8202: @section The optional Floating-Point word set
8203: @c =====================================================================
8204: @cindex system documentation, floating-point words
8205: @cindex floating-point words, system documentation
8206:
8207: @menu
8208: * floating-idef:: Implementation Defined Options
8209: * floating-ambcond:: Ambiguous Conditions
8210: @end menu
8211:
8212:
8213: @c ---------------------------------------------------------------------
8214: @node floating-idef, floating-ambcond, The optional Floating-Point word set, The optional Floating-Point word set
8215: @subsection Implementation Defined Options
8216: @c ---------------------------------------------------------------------
8217: @cindex implementation-defined options, floating-point words
8218: @cindex floating-point words, implementation-defined options
8219:
8220: @table @i
8221: @item format and range of floating point numbers:
8222: @cindex format and range of floating point numbers
8223: @cindex floating point numbers, format and range
8224: System-dependent; the @code{double} type of C.
8225:
8226: @item results of @code{REPRESENT} when @var{float} is out of range:
8227: @cindex @code{REPRESENT}, results when @var{float} is out of range
8228: System dependent; @code{REPRESENT} is implemented using the C library
8229: function @code{ecvt()} and inherits its behaviour in this respect.
8230:
8231: @item rounding or truncation of floating-point numbers:
8232: @cindex rounding of floating-point numbers
8233: @cindex truncation of floating-point numbers
8234: @cindex floating-point numbers, rounding or truncation
8235: System dependent; the rounding behaviour is inherited from the hosting C
8236: compiler. IEEE-FP-based (i.e., most) systems by default round to
8237: nearest, and break ties by rounding to even (i.e., such that the last
8238: bit of the mantissa is 0).
8239:
8240: @item size of floating-point stack:
8241: @cindex floating-point stack size
8242: @code{s" FLOATING-STACK" environment? drop .} gives the total size of
8243: the floating-point stack (in floats). You can specify this on startup
8244: with the command-line option @code{-f} (@pxref{Invoking Gforth}).
8245:
8246: @item width of floating-point stack:
8247: @cindex floating-point stack width
8248: @code{1 floats}.
8249:
8250: @end table
8251:
8252:
8253: @c ---------------------------------------------------------------------
8254: @node floating-ambcond, , floating-idef, The optional Floating-Point word set
8255: @subsection Ambiguous conditions
8256: @c ---------------------------------------------------------------------
8257: @cindex floating-point words, ambiguous conditions
8258: @cindex ambiguous conditions, floating-point words
8259:
8260: @table @i
8261: @item @code{df@@} or @code{df!} used with an address that is not double-float aligned:
8262: @cindex @code{df@@} or @code{df!} used with an address that is not double-float aligned
8263: System-dependent. Typically results in a @code{-23 THROW} like other
8264: alignment violations.
8265:
8266: @item @code{f@@} or @code{f!} used with an address that is not float aligned:
8267: @cindex @code{f@@} used with an address that is not float aligned
8268: @cindex @code{f!} used with an address that is not float aligned
8269: System-dependent. Typically results in a @code{-23 THROW} like other
8270: alignment violations.
8271:
8272: @item floating-point result out of range:
8273: @cindex floating-point result out of range
8274: System-dependent. Can result in a @code{-55 THROW} (Floating-point
8275: unidentified fault), or can produce a special value representing, e.g.,
8276: Infinity.
8277:
8278: @item @code{sf@@} or @code{sf!} used with an address that is not single-float aligned:
8279: @cindex @code{sf@@} or @code{sf!} used with an address that is not single-float aligned
8280: System-dependent. Typically results in an alignment fault like other
8281: alignment violations.
8282:
8283: @item @code{BASE} is not decimal (@code{REPRESENT}, @code{F.}, @code{FE.}, @code{FS.}):
8284: @cindex @code{BASE} is not decimal (@code{REPRESENT}, @code{F.}, @code{FE.}, @code{FS.})
8285: The floating-point number is converted into decimal nonetheless.
8286:
8287: @item Both arguments are equal to zero (@code{FATAN2}):
8288: @cindex @code{FATAN2}, both arguments are equal to zero
8289: System-dependent. @code{FATAN2} is implemented using the C library
8290: function @code{atan2()}.
8291:
8292: @item Using @code{FTAN} on an argument @var{r1} where cos(@var{r1}) is zero:
8293: @cindex @code{FTAN} on an argument @var{r1} where cos(@var{r1}) is zero
8294: System-dependent. Anyway, typically the cos of @var{r1} will not be zero
8295: because of small errors and the tan will be a very large (or very small)
8296: but finite number.
8297:
8298: @item @var{d} cannot be presented precisely as a float in @code{D>F}:
8299: @cindex @code{D>F}, @var{d} cannot be presented precisely as a float
8300: The result is rounded to the nearest float.
8301:
8302: @item dividing by zero:
8303: @cindex dividing by zero, floating-point
8304: @cindex floating-point dividing by zero
8305: @cindex floating-point unidentified fault, FP divide-by-zero
8306: @code{-55 throw} (Floating-point unidentified fault)
8307:
8308: @item exponent too big for conversion (@code{DF!}, @code{DF@@}, @code{SF!}, @code{SF@@}):
8309: @cindex exponent too big for conversion (@code{DF!}, @code{DF@@}, @code{SF!}, @code{SF@@})
8310: System dependent. On IEEE-FP based systems the number is converted into
8311: an infinity.
8312:
8313: @item @var{float}<1 (@code{FACOSH}):
8314: @cindex @code{FACOSH}, @var{float}<1
8315: @cindex floating-point unidentified fault, @code{FACOSH}
8316: @code{-55 throw} (Floating-point unidentified fault)
8317:
8318: @item @var{float}=<-1 (@code{FLNP1}):
8319: @cindex @code{FLNP1}, @var{float}=<-1
8320: @cindex floating-point unidentified fault, @code{FLNP1}
8321: @code{-55 throw} (Floating-point unidentified fault). On IEEE-FP systems
8322: negative infinity is typically produced for @var{float}=-1.
8323:
8324: @item @var{float}=<0 (@code{FLN}, @code{FLOG}):
8325: @cindex @code{FLN}, @var{float}=<0
8326: @cindex @code{FLOG}, @var{float}=<0
8327: @cindex floating-point unidentified fault, @code{FLN} or @code{FLOG}
8328: @code{-55 throw} (Floating-point unidentified fault). On IEEE-FP systems
8329: negative infinity is typically produced for @var{float}=0.
8330:
8331: @item @var{float}<0 (@code{FASINH}, @code{FSQRT}):
8332: @cindex @code{FASINH}, @var{float}<0
8333: @cindex @code{FSQRT}, @var{float}<0
8334: @cindex floating-point unidentified fault, @code{FASINH} or @code{FSQRT}
8335: @code{-55 throw} (Floating-point unidentified fault). @code{fasinh}
8336: produces values for these inputs on my Linux box (Bug in the C library?)
8337:
8338: @item |@var{float}|>1 (@code{FACOS}, @code{FASIN}, @code{FATANH}):
8339: @cindex @code{FACOS}, |@var{float}|>1
8340: @cindex @code{FASIN}, |@var{float}|>1
8341: @cindex @code{FATANH}, |@var{float}|>1
8342: @cindex floating-point unidentified fault, @code{FACOS}, @code{FASIN} or @code{FATANH}
8343: @code{-55 throw} (Floating-point unidentified fault).
8344:
8345: @item integer part of float cannot be represented by @var{d} in @code{F>D}:
8346: @cindex @code{F>D}, integer part of float cannot be represented by @var{d}
8347: @cindex floating-point unidentified fault, @code{F>D}
8348: @code{-55 throw} (Floating-point unidentified fault).
8349:
8350: @item string larger than pictured numeric output area (@code{f.}, @code{fe.}, @code{fs.}):
8351: @cindex string larger than pictured numeric output area (@code{f.}, @code{fe.}, @code{fs.})
8352: This does not happen.
8353: @end table
8354:
8355: @c =====================================================================
8356: @node The optional Locals word set, The optional Memory-Allocation word set, The optional Floating-Point word set, ANS conformance
8357: @section The optional Locals word set
8358: @c =====================================================================
8359: @cindex system documentation, locals words
8360: @cindex locals words, system documentation
8361:
8362: @menu
8363: * locals-idef:: Implementation Defined Options
8364: * locals-ambcond:: Ambiguous Conditions
8365: @end menu
8366:
8367:
8368: @c ---------------------------------------------------------------------
8369: @node locals-idef, locals-ambcond, The optional Locals word set, The optional Locals word set
8370: @subsection Implementation Defined Options
8371: @c ---------------------------------------------------------------------
8372: @cindex implementation-defined options, locals words
8373: @cindex locals words, implementation-defined options
8374:
8375: @table @i
8376: @item maximum number of locals in a definition:
8377: @cindex maximum number of locals in a definition
8378: @cindex locals, maximum number in a definition
8379: @code{s" #locals" environment? drop .}. Currently 15. This is a lower
8380: bound, e.g., on a 32-bit machine there can be 41 locals of up to 8
8381: characters. The number of locals in a definition is bounded by the size
8382: of locals-buffer, which contains the names of the locals.
8383:
8384: @end table
8385:
8386:
8387: @c ---------------------------------------------------------------------
8388: @node locals-ambcond, , locals-idef, The optional Locals word set
8389: @subsection Ambiguous conditions
8390: @c ---------------------------------------------------------------------
8391: @cindex locals words, ambiguous conditions
8392: @cindex ambiguous conditions, locals words
8393:
8394: @table @i
8395: @item executing a named local in interpretation state:
8396: @cindex local in interpretation state
8397: @cindex Interpreting a compile-only word, for a local
8398: Locals have no interpretation semantics. If you try to perform the
8399: interpretation semantics, you will get a @code{-14 throw} somewhere
8400: (Interpreting a compile-only word). If you perform the compilation
8401: semantics, the locals access will be compiled (irrespective of state).
8402:
8403: @item @var{name} not defined by @code{VALUE} or @code{(LOCAL)} (@code{TO}):
8404: @cindex name not defined by @code{VALUE} or @code{(LOCAL)} used by @code{TO}
8405: @cindex @code{TO} on non-@code{VALUE}s and non-locals
8406: @cindex Invalid name argument, @code{TO}
8407: @code{-32 throw} (Invalid name argument)
8408:
8409: @end table
8410:
8411:
8412: @c =====================================================================
8413: @node The optional Memory-Allocation word set, The optional Programming-Tools word set, The optional Locals word set, ANS conformance
8414: @section The optional Memory-Allocation word set
8415: @c =====================================================================
8416: @cindex system documentation, memory-allocation words
8417: @cindex memory-allocation words, system documentation
8418:
8419: @menu
8420: * memory-idef:: Implementation Defined Options
8421: @end menu
8422:
8423:
8424: @c ---------------------------------------------------------------------
8425: @node memory-idef, , The optional Memory-Allocation word set, The optional Memory-Allocation word set
8426: @subsection Implementation Defined Options
8427: @c ---------------------------------------------------------------------
8428: @cindex implementation-defined options, memory-allocation words
8429: @cindex memory-allocation words, implementation-defined options
8430:
8431: @table @i
8432: @item values and meaning of @var{ior}:
8433: @cindex @var{ior} values and meaning
8434: The @var{ior}s returned by the file and memory allocation words are
8435: intended as throw codes. They typically are in the range
8436: -512@minus{}-2047 of OS errors. The mapping from OS error numbers to
8437: @var{ior}s is -512@minus{}@var{errno}.
8438:
8439: @end table
8440:
8441: @c =====================================================================
8442: @node The optional Programming-Tools word set, The optional Search-Order word set, The optional Memory-Allocation word set, ANS conformance
8443: @section The optional Programming-Tools word set
8444: @c =====================================================================
8445: @cindex system documentation, programming-tools words
8446: @cindex programming-tools words, system documentation
8447:
8448: @menu
8449: * programming-idef:: Implementation Defined Options
8450: * programming-ambcond:: Ambiguous Conditions
8451: @end menu
8452:
8453:
8454: @c ---------------------------------------------------------------------
8455: @node programming-idef, programming-ambcond, The optional Programming-Tools word set, The optional Programming-Tools word set
8456: @subsection Implementation Defined Options
8457: @c ---------------------------------------------------------------------
8458: @cindex implementation-defined options, programming-tools words
8459: @cindex programming-tools words, implementation-defined options
8460:
8461: @table @i
8462: @item ending sequence for input following @code{;CODE} and @code{CODE}:
8463: @cindex @code{;CODE} ending sequence
8464: @cindex @code{CODE} ending sequence
8465: @code{END-CODE}
8466:
8467: @item manner of processing input following @code{;CODE} and @code{CODE}:
8468: @cindex @code{;CODE}, processing input
8469: @cindex @code{CODE}, processing input
8470: The @code{ASSEMBLER} vocabulary is pushed on the search order stack, and
8471: the input is processed by the text interpreter, (starting) in interpret
8472: state.
8473:
8474: @item search order capability for @code{EDITOR} and @code{ASSEMBLER}:
8475: @cindex @code{ASSEMBLER}, search order capability
8476: The ANS Forth search order word set.
8477:
8478: @item source and format of display by @code{SEE}:
8479: @cindex @code{SEE}, source and format of output
8480: The source for @code{see} is the intermediate code used by the inner
8481: interpreter. The current @code{see} tries to output Forth source code
8482: as well as possible.
8483:
8484: @end table
8485:
8486: @c ---------------------------------------------------------------------
8487: @node programming-ambcond, , programming-idef, The optional Programming-Tools word set
8488: @subsection Ambiguous conditions
8489: @c ---------------------------------------------------------------------
8490: @cindex programming-tools words, ambiguous conditions
8491: @cindex ambiguous conditions, programming-tools words
8492:
8493: @table @i
8494:
1.21 crook 8495: @item deleting the compilation word list (@code{FORGET}):
8496: @cindex @code{FORGET}, deleting the compilation word list
1.1 anton 8497: Not implemented (yet).
8498:
8499: @item fewer than @var{u}+1 items on the control flow stack (@code{CS-PICK}, @code{CS-ROLL}):
8500: @cindex @code{CS-PICK}, fewer than @var{u}+1 items on the control flow stack
8501: @cindex @code{CS-ROLL}, fewer than @var{u}+1 items on the control flow stack
8502: @cindex control-flow stack underflow
8503: This typically results in an @code{abort"} with a descriptive error
8504: message (may change into a @code{-22 throw} (Control structure mismatch)
8505: in the future). You may also get a memory access error. If you are
8506: unlucky, this ambiguous condition is not caught.
8507:
8508: @item @var{name} can't be found (@code{FORGET}):
8509: @cindex @code{FORGET}, @var{name} can't be found
8510: Not implemented (yet).
8511:
8512: @item @var{name} not defined via @code{CREATE}:
8513: @cindex @code{;CODE}, @var{name} not defined via @code{CREATE}
8514: @code{;CODE} behaves like @code{DOES>} in this respect, i.e., it changes
8515: the execution semantics of the last defined word no matter how it was
8516: defined.
8517:
8518: @item @code{POSTPONE} applied to @code{[IF]}:
8519: @cindex @code{POSTPONE} applied to @code{[IF]}
8520: @cindex @code{[IF]} and @code{POSTPONE}
8521: After defining @code{: X POSTPONE [IF] ; IMMEDIATE}. @code{X} is
8522: equivalent to @code{[IF]}.
8523:
8524: @item reaching the end of the input source before matching @code{[ELSE]} or @code{[THEN]}:
8525: @cindex @code{[IF]}, end of the input source before matching @code{[ELSE]} or @code{[THEN]}
8526: Continue in the same state of conditional compilation in the next outer
8527: input source. Currently there is no warning to the user about this.
8528:
8529: @item removing a needed definition (@code{FORGET}):
8530: @cindex @code{FORGET}, removing a needed definition
8531: Not implemented (yet).
8532:
8533: @end table
8534:
8535:
8536: @c =====================================================================
8537: @node The optional Search-Order word set, , The optional Programming-Tools word set, ANS conformance
8538: @section The optional Search-Order word set
8539: @c =====================================================================
8540: @cindex system documentation, search-order words
8541: @cindex search-order words, system documentation
8542:
8543: @menu
8544: * search-idef:: Implementation Defined Options
8545: * search-ambcond:: Ambiguous Conditions
8546: @end menu
8547:
8548:
8549: @c ---------------------------------------------------------------------
8550: @node search-idef, search-ambcond, The optional Search-Order word set, The optional Search-Order word set
8551: @subsection Implementation Defined Options
8552: @c ---------------------------------------------------------------------
8553: @cindex implementation-defined options, search-order words
8554: @cindex search-order words, implementation-defined options
8555:
8556: @table @i
8557: @item maximum number of word lists in search order:
8558: @cindex maximum number of word lists in search order
8559: @cindex search order, maximum depth
8560: @code{s" wordlists" environment? drop .}. Currently 16.
8561:
8562: @item minimum search order:
8563: @cindex minimum search order
8564: @cindex search order, minimum
8565: @code{root root}.
8566:
8567: @end table
8568:
8569: @c ---------------------------------------------------------------------
8570: @node search-ambcond, , search-idef, The optional Search-Order word set
8571: @subsection Ambiguous conditions
8572: @c ---------------------------------------------------------------------
8573: @cindex search-order words, ambiguous conditions
8574: @cindex ambiguous conditions, search-order words
8575:
8576: @table @i
1.21 crook 8577: @item changing the compilation word list (during compilation):
8578: @cindex changing the compilation word list (during compilation)
8579: @cindex compilation word list, change before definition ends
8580: The word is entered into the word list that was the compilation word list
1.1 anton 8581: at the start of the definition. Any changes to the name field (e.g.,
8582: @code{immediate}) or the code field (e.g., when executing @code{DOES>})
8583: are applied to the latest defined word (as reported by @code{last} or
1.21 crook 8584: @code{lastxt}), if possible, irrespective of the compilation word list.
1.1 anton 8585:
8586: @item search order empty (@code{previous}):
8587: @cindex @code{previous}, search order empty
8588: @cindex Vocstack empty, @code{previous}
8589: @code{abort" Vocstack empty"}.
8590:
8591: @item too many word lists in search order (@code{also}):
8592: @cindex @code{also}, too many word lists in search order
8593: @cindex Vocstack full, @code{also}
8594: @code{abort" Vocstack full"}.
8595:
8596: @end table
8597:
8598: @c ***************************************************************
8599: @node Model, Integrating Gforth, ANS conformance, Top
8600: @chapter Model
8601:
8602: This chapter has yet to be written. It will contain information, on
8603: which internal structures you can rely.
8604:
8605: @c ***************************************************************
8606: @node Integrating Gforth, Emacs and Gforth, Model, Top
8607: @chapter Integrating Gforth into C programs
8608:
8609: This is not yet implemented.
8610:
8611: Several people like to use Forth as scripting language for applications
8612: that are otherwise written in C, C++, or some other language.
8613:
8614: The Forth system ATLAST provides facilities for embedding it into
8615: applications; unfortunately it has several disadvantages: most
8616: importantly, it is not based on ANS Forth, and it is apparently dead
8617: (i.e., not developed further and not supported). The facilities
1.21 crook 8618: provided by Gforth in this area are inspired by ATLAST's facilities, so
1.1 anton 8619: making the switch should not be hard.
8620:
8621: We also tried to design the interface such that it can easily be
8622: implemented by other Forth systems, so that we may one day arrive at a
8623: standardized interface. Such a standard interface would allow you to
8624: replace the Forth system without having to rewrite C code.
8625:
8626: You embed the Gforth interpreter by linking with the library
8627: @code{libgforth.a} (give the compiler the option @code{-lgforth}). All
8628: global symbols in this library that belong to the interface, have the
8629: prefix @code{forth_}. (Global symbols that are used internally have the
8630: prefix @code{gforth_}).
8631:
8632: You can include the declarations of Forth types and the functions and
8633: variables of the interface with @code{#include <forth.h>}.
8634:
8635: Types.
8636:
8637: Variables.
8638:
8639: Data and FP Stack pointer. Area sizes.
8640:
8641: functions.
8642:
8643: forth_init(imagefile)
8644: forth_evaluate(string) exceptions?
8645: forth_goto(address) (or forth_execute(xt)?)
8646: forth_continue() (a corountining mechanism)
8647:
8648: Adding primitives.
8649:
8650: No checking.
8651:
8652: Signals?
8653:
8654: Accessing the Stacks
8655:
8656: @node Emacs and Gforth, Image Files, Integrating Gforth, Top
8657: @chapter Emacs and Gforth
8658: @cindex Emacs and Gforth
8659:
8660: @cindex @file{gforth.el}
8661: @cindex @file{forth.el}
8662: @cindex Rydqvist, Goran
8663: @cindex comment editing commands
8664: @cindex @code{\}, editing with Emacs
8665: @cindex debug tracer editing commands
8666: @cindex @code{~~}, removal with Emacs
8667: @cindex Forth mode in Emacs
8668: Gforth comes with @file{gforth.el}, an improved version of
8669: @file{forth.el} by Goran Rydqvist (included in the TILE package). The
8670: improvements are a better (but still not perfect) handling of
8671: indentation. I have also added comment paragraph filling (@kbd{M-q}),
8672: commenting (@kbd{C-x \}) and uncommenting (@kbd{C-u C-x \}) regions and
8673: removing debugging tracers (@kbd{C-x ~}, @pxref{Debugging}). I left the
8674: stuff I do not use alone, even though some of it only makes sense for
8675: TILE. To get a description of these features, enter Forth mode and type
8676: @kbd{C-h m}.
8677:
8678: @cindex source location of error or debugging output in Emacs
8679: @cindex error output, finding the source location in Emacs
8680: @cindex debugging output, finding the source location in Emacs
8681: In addition, Gforth supports Emacs quite well: The source code locations
8682: given in error messages, debugging output (from @code{~~}) and failed
8683: assertion messages are in the right format for Emacs' compilation mode
8684: (@pxref{Compilation, , Running Compilations under Emacs, emacs, Emacs
8685: Manual}) so the source location corresponding to an error or other
8686: message is only a few keystrokes away (@kbd{C-x `} for the next error,
8687: @kbd{C-c C-c} for the error under the cursor).
8688:
8689: @cindex @file{TAGS} file
8690: @cindex @file{etags.fs}
8691: @cindex viewing the source of a word in Emacs
8692: Also, if you @code{include} @file{etags.fs}, a new @file{TAGS} file
8693: (@pxref{Tags, , Tags Tables, emacs, Emacs Manual}) will be produced that
8694: contains the definitions of all words defined afterwards. You can then
8695: find the source for a word using @kbd{M-.}. Note that emacs can use
8696: several tags files at the same time (e.g., one for the Gforth sources
8697: and one for your program, @pxref{Select Tags Table,,Selecting a Tags
8698: Table,emacs, Emacs Manual}). The TAGS file for the preloaded words is
8699: @file{$(datadir)/gforth/$(VERSION)/TAGS} (e.g.,
8700: @file{/usr/local/share/gforth/0.2.0/TAGS}).
8701:
8702: @cindex @file{.emacs}
8703: To get all these benefits, add the following lines to your @file{.emacs}
8704: file:
8705:
8706: @example
8707: (autoload 'forth-mode "gforth.el")
8708: (setq auto-mode-alist (cons '("\\.fs\\'" . forth-mode) auto-mode-alist))
8709: @end example
8710:
8711: @node Image Files, Engine, Emacs and Gforth, Top
8712: @chapter Image Files
8713: @cindex image files
8714: @cindex @code{.fi} files
8715: @cindex precompiled Forth code
8716: @cindex dictionary in persistent form
8717: @cindex persistent form of dictionary
8718:
8719: An image file is a file containing an image of the Forth dictionary,
8720: i.e., compiled Forth code and data residing in the dictionary. By
8721: convention, we use the extension @code{.fi} for image files.
8722:
8723: @menu
1.18 anton 8724: * Image Licensing Issues:: Distribution terms for images.
8725: * Image File Background:: Why have image files?
8726: * Non-Relocatable Image Files:: don't always work.
8727: * Data-Relocatable Image Files:: are better.
1.1 anton 8728: * Fully Relocatable Image Files:: better yet.
1.18 anton 8729: * Stack and Dictionary Sizes:: Setting the default sizes for an image.
8730: * Running Image Files:: @code{gforth -i @var{file}} or @var{file}.
8731: * Modifying the Startup Sequence:: and turnkey applications.
1.1 anton 8732: @end menu
8733:
1.18 anton 8734: @node Image Licensing Issues, Image File Background, Image Files, Image Files
8735: @section Image Licensing Issues
8736: @cindex license for images
8737: @cindex image license
8738:
8739: An image created with @code{gforthmi} (@pxref{gforthmi}) or
8740: @code{savesystem} (@pxref{Non-Relocatable Image Files}) includes the
8741: original image; i.e., according to copyright law it is a derived work of
8742: the original image.
8743:
8744: Since Gforth is distributed under the GNU GPL, the newly created image
8745: falls under the GNU GPL, too. In particular, this means that if you
8746: distribute the image, you have to make all of the sources for the image
8747: available, including those you wrote. For details see @ref{License, ,
8748: GNU General Public License (Section 3)}.
8749:
8750: If you create an image with @code{cross} (@pxref{cross.fs}), the image
8751: contains only code compiled from the sources you gave it; if none of
8752: these sources is under the GPL, the terms discussed above do not apply
8753: to the image. However, if your image needs an engine (a gforth binary)
8754: that is under the GPL, you should make sure that you distribute both in
8755: a way that is at most a @emph{mere aggregation}, if you don't want the
8756: terms of the GPL to apply to the image.
8757:
8758: @node Image File Background, Non-Relocatable Image Files, Image Licensing Issues, Image Files
1.1 anton 8759: @section Image File Background
8760: @cindex image file background
8761:
8762: Our Forth system consists not only of primitives, but also of
8763: definitions written in Forth. Since the Forth compiler itself belongs to
8764: those definitions, it is not possible to start the system with the
8765: primitives and the Forth source alone. Therefore we provide the Forth
8766: code as an image file in nearly executable form. At the start of the
8767: system a C routine loads the image file into memory, optionally
8768: relocates the addresses, then sets up the memory (stacks etc.) according
8769: to information in the image file, and starts executing Forth code.
8770:
8771: The image file variants represent different compromises between the
8772: goals of making it easy to generate image files and making them
8773: portable.
8774:
8775: @cindex relocation at run-time
8776: Win32Forth 3.4 and Mitch Bradleys @code{cforth} use relocation at
8777: run-time. This avoids many of the complications discussed below (image
8778: files are data relocatable without further ado), but costs performance
8779: (one addition per memory access).
8780:
8781: @cindex relocation at load-time
8782: By contrast, our loader performs relocation at image load time. The
8783: loader also has to replace tokens standing for primitive calls with the
8784: appropriate code-field addresses (or code addresses in the case of
8785: direct threading).
8786:
8787: There are three kinds of image files, with different degrees of
8788: relocatability: non-relocatable, data-relocatable, and fully relocatable
8789: image files.
8790:
8791: @cindex image file loader
8792: @cindex relocating loader
8793: @cindex loader for image files
8794: These image file variants have several restrictions in common; they are
8795: caused by the design of the image file loader:
8796:
8797: @itemize @bullet
8798: @item
8799: There is only one segment; in particular, this means, that an image file
8800: cannot represent @code{ALLOCATE}d memory chunks (and pointers to
8801: them). And the contents of the stacks are not represented, either.
8802:
8803: @item
8804: The only kinds of relocation supported are: adding the same offset to
8805: all cells that represent data addresses; and replacing special tokens
8806: with code addresses or with pieces of machine code.
8807:
8808: If any complex computations involving addresses are performed, the
8809: results cannot be represented in the image file. Several applications that
8810: use such computations come to mind:
8811: @itemize @minus
8812: @item
8813: Hashing addresses (or data structures which contain addresses) for table
8814: lookup. If you use Gforth's @code{table}s or @code{wordlist}s for this
8815: purpose, you will have no problem, because the hash tables are
8816: recomputed automatically when the system is started. If you use your own
8817: hash tables, you will have to do something similar.
8818:
8819: @item
8820: There's a cute implementation of doubly-linked lists that uses
8821: @code{XOR}ed addresses. You could represent such lists as singly-linked
8822: in the image file, and restore the doubly-linked representation on
8823: startup.@footnote{In my opinion, though, you should think thrice before
8824: using a doubly-linked list (whatever implementation).}
8825:
8826: @item
8827: The code addresses of run-time routines like @code{docol:} cannot be
8828: represented in the image file (because their tokens would be replaced by
8829: machine code in direct threaded implementations). As a workaround,
8830: compute these addresses at run-time with @code{>code-address} from the
8831: executions tokens of appropriate words (see the definitions of
8832: @code{docol:} and friends in @file{kernel.fs}).
8833:
8834: @item
8835: On many architectures addresses are represented in machine code in some
8836: shifted or mangled form. You cannot put @code{CODE} words that contain
8837: absolute addresses in this form in a relocatable image file. Workarounds
8838: are representing the address in some relative form (e.g., relative to
8839: the CFA, which is present in some register), or loading the address from
8840: a place where it is stored in a non-mangled form.
8841: @end itemize
8842: @end itemize
8843:
8844: @node Non-Relocatable Image Files, Data-Relocatable Image Files, Image File Background, Image Files
8845: @section Non-Relocatable Image Files
8846: @cindex non-relocatable image files
8847: @cindex image files, non-relocatable
8848:
8849: These files are simple memory dumps of the dictionary. They are specific
8850: to the executable (i.e., @file{gforth} file) they were created
8851: with. What's worse, they are specific to the place on which the
8852: dictionary resided when the image was created. Now, there is no
8853: guarantee that the dictionary will reside at the same place the next
8854: time you start Gforth, so there's no guarantee that a non-relocatable
8855: image will work the next time (Gforth will complain instead of crashing,
8856: though).
8857:
8858: You can create a non-relocatable image file with
8859:
8860: doc-savesystem
8861:
8862: @node Data-Relocatable Image Files, Fully Relocatable Image Files, Non-Relocatable Image Files, Image Files
8863: @section Data-Relocatable Image Files
8864: @cindex data-relocatable image files
8865: @cindex image files, data-relocatable
8866:
8867: These files contain relocatable data addresses, but fixed code addresses
8868: (instead of tokens). They are specific to the executable (i.e.,
8869: @file{gforth} file) they were created with. For direct threading on some
8870: architectures (e.g., the i386), data-relocatable images do not work. You
8871: get a data-relocatable image, if you use @file{gforthmi} with a
8872: Gforth binary that is not doubly indirect threaded (@pxref{Fully
8873: Relocatable Image Files}).
8874:
8875: @node Fully Relocatable Image Files, Stack and Dictionary Sizes, Data-Relocatable Image Files, Image Files
8876: @section Fully Relocatable Image Files
8877: @cindex fully relocatable image files
8878: @cindex image files, fully relocatable
8879:
8880: @cindex @file{kern*.fi}, relocatability
8881: @cindex @file{gforth.fi}, relocatability
8882: These image files have relocatable data addresses, and tokens for code
8883: addresses. They can be used with different binaries (e.g., with and
8884: without debugging) on the same machine, and even across machines with
8885: the same data formats (byte order, cell size, floating point
8886: format). However, they are usually specific to the version of Gforth
8887: they were created with. The files @file{gforth.fi} and @file{kernl*.fi}
8888: are fully relocatable.
8889:
8890: There are two ways to create a fully relocatable image file:
8891:
8892: @menu
8893: * gforthmi:: The normal way
8894: * cross.fs:: The hard way
8895: @end menu
8896:
8897: @node gforthmi, cross.fs, Fully Relocatable Image Files, Fully Relocatable Image Files
8898: @subsection @file{gforthmi}
8899: @cindex @file{comp-i.fs}
8900: @cindex @file{gforthmi}
8901:
8902: You will usually use @file{gforthmi}. If you want to create an
8903: image @var{file} that contains everything you would load by invoking
8904: Gforth with @code{gforth @var{options}}, you simply say
8905: @example
8906: gforthmi @var{file} @var{options}
8907: @end example
8908:
8909: E.g., if you want to create an image @file{asm.fi} that has the file
8910: @file{asm.fs} loaded in addition to the usual stuff, you could do it
8911: like this:
8912:
8913: @example
8914: gforthmi asm.fi asm.fs
8915: @end example
8916:
8917: @file{gforthmi} works like this: It produces two non-relocatable
8918: images for different addresses and then compares them. Its output
8919: reflects this: first you see the output (if any) of the two Gforth
8920: invocations that produce the nonrelocatable image files, then you see
8921: the output of the comparing program: It displays the offset used for
8922: data addresses and the offset used for code addresses;
8923: moreover, for each cell that cannot be represented correctly in the
8924: image files, it displays a line like the following one:
8925:
8926: @example
8927: 78DC BFFFFA50 BFFFFA40
8928: @end example
8929:
8930: This means that at offset $78dc from @code{forthstart}, one input image
8931: contains $bffffa50, and the other contains $bffffa40. Since these cells
8932: cannot be represented correctly in the output image, you should examine
8933: these places in the dictionary and verify that these cells are dead
8934: (i.e., not read before they are written).
8935:
8936: @cindex @code{savesystem} during @file{gforthmi}
8937: @cindex @code{bye} during @file{gforthmi}
8938: @cindex doubly indirect threaded code
8939: @cindex environment variable @code{GFORTHD}
8940: @cindex @code{GFORTHD} environment variable
8941: @cindex @code{gforth-ditc}
8942: There are a few wrinkles: After processing the passed @var{options}, the
8943: words @code{savesystem} and @code{bye} must be visible. A special doubly
8944: indirect threaded version of the @file{gforth} executable is used for
8945: creating the nonrelocatable images; you can pass the exact filename of
8946: this executable through the environment variable @code{GFORTHD}
8947: (default: @file{gforth-ditc}); if you pass a version that is not doubly
8948: indirect threaded, you will not get a fully relocatable image, but a
8949: data-relocatable image (because there is no code address offset).
8950:
8951: @node cross.fs, , gforthmi, Fully Relocatable Image Files
8952: @subsection @file{cross.fs}
8953: @cindex @file{cross.fs}
8954: @cindex cross-compiler
8955: @cindex metacompiler
8956:
8957: You can also use @code{cross}, a batch compiler that accepts a Forth-like
8958: programming language. This @code{cross} language has to be documented
8959: yet.
8960:
8961: @cindex target compiler
8962: @code{cross} also allows you to create image files for machines with
8963: different data sizes and data formats than the one used for generating
8964: the image file. You can also use it to create an application image that
8965: does not contain a Forth compiler. These features are bought with
8966: restrictions and inconveniences in programming. E.g., addresses have to
8967: be stored in memory with special words (@code{A!}, @code{A,}, etc.) in
8968: order to make the code relocatable.
8969:
8970:
8971: @node Stack and Dictionary Sizes, Running Image Files, Fully Relocatable Image Files, Image Files
8972: @section Stack and Dictionary Sizes
8973: @cindex image file, stack and dictionary sizes
8974: @cindex dictionary size default
8975: @cindex stack size default
8976:
8977: If you invoke Gforth with a command line flag for the size
8978: (@pxref{Invoking Gforth}), the size you specify is stored in the
8979: dictionary. If you save the dictionary with @code{savesystem} or create
8980: an image with @file{gforthmi}, this size will become the default
8981: for the resulting image file. E.g., the following will create a
1.21 crook 8982: fully relocatable version of @file{gforth.fi} with a 1MB dictionary:
1.1 anton 8983:
8984: @example
8985: gforthmi gforth.fi -m 1M
8986: @end example
8987:
8988: In other words, if you want to set the default size for the dictionary
8989: and the stacks of an image, just invoke @file{gforthmi} with the
8990: appropriate options when creating the image.
8991:
8992: @cindex stack size, cache-friendly
8993: Note: For cache-friendly behaviour (i.e., good performance), you should
8994: make the sizes of the stacks modulo, say, 2K, somewhat different. E.g.,
8995: the default stack sizes are: data: 16k (mod 2k=0); fp: 15.5k (mod
8996: 2k=1.5k); return: 15k(mod 2k=1k); locals: 14.5k (mod 2k=0.5k).
8997:
8998: @node Running Image Files, Modifying the Startup Sequence, Stack and Dictionary Sizes, Image Files
8999: @section Running Image Files
9000: @cindex running image files
9001: @cindex invoking image files
9002: @cindex image file invocation
9003:
9004: @cindex -i, invoke image file
9005: @cindex --image file, invoke image file
9006: You can invoke Gforth with an image file @var{image} instead of the
9007: default @file{gforth.fi} with the @code{-i} flag (@pxref{Invoking Gforth}):
9008: @example
9009: gforth -i @var{image}
9010: @end example
9011:
9012: @cindex executable image file
9013: @cindex image files, executable
9014: If your operating system supports starting scripts with a line of the
9015: form @code{#! ...}, you just have to type the image file name to start
9016: Gforth with this image file (note that the file extension @code{.fi} is
9017: just a convention). I.e., to run Gforth with the image file @var{image},
9018: you can just type @var{image} instead of @code{gforth -i @var{image}}.
9019:
9020: doc-#!
9021:
9022: @node Modifying the Startup Sequence, , Running Image Files, Image Files
9023: @section Modifying the Startup Sequence
9024: @cindex startup sequence for image file
9025: @cindex image file initialization sequence
9026: @cindex initialization sequence of image file
9027:
9028: You can add your own initialization to the startup sequence through the
9029: deferred word
9030:
9031: doc-'cold
9032:
9033: @code{'cold} is invoked just before the image-specific command line
9034: processing (by default, loading files and evaluating (@code{-e}) strings)
9035: starts.
9036:
9037: A sequence for adding your initialization usually looks like this:
9038:
9039: @example
9040: :noname
9041: Defers 'cold \ do other initialization stuff (e.g., rehashing wordlists)
9042: ... \ your stuff
9043: ; IS 'cold
9044: @end example
9045:
9046: @cindex turnkey image files
9047: @cindex image files, turnkey applications
9048: You can make a turnkey image by letting @code{'cold} execute a word
9049: (your turnkey application) that never returns; instead, it exits Gforth
9050: via @code{bye} or @code{throw}.
9051:
9052: @cindex command-line arguments, access
9053: @cindex arguments on the command line, access
9054: You can access the (image-specific) command-line arguments through the
9055: variables @code{argc} and @code{argv}. @code{arg} provides conventient
9056: access to @code{argv}.
9057:
9058: doc-argc
9059: doc-argv
9060: doc-arg
9061:
9062: If @code{'cold} exits normally, Gforth processes the command-line
9063: arguments as files to be loaded and strings to be evaluated. Therefore,
9064: @code{'cold} should remove the arguments it has used in this case.
9065:
9066: @c ******************************************************************
1.13 pazsan 9067: @node Engine, Binding to System Library, Image Files, Top
1.1 anton 9068: @chapter Engine
9069: @cindex engine
9070: @cindex virtual machine
9071:
9072: Reading this section is not necessary for programming with Gforth. It
9073: may be helpful for finding your way in the Gforth sources.
9074:
9075: The ideas in this section have also been published in the papers
9076: @cite{ANS fig/GNU/??? Forth} (in German) by Bernd Paysan, presented at
9077: the Forth-Tagung '93 and @cite{A Portable Forth Engine} by M. Anton
9078: Ertl, presented at EuroForth '93; the latter is available at
9079: @*@url{http://www.complang.tuwien.ac.at/papers/ertl93.ps.Z}.
9080:
9081: @menu
9082: * Portability::
9083: * Threading::
9084: * Primitives::
9085: * Performance::
9086: @end menu
9087:
9088: @node Portability, Threading, Engine, Engine
9089: @section Portability
9090: @cindex engine portability
9091:
9092: One of the main goals of the effort is availability across a wide range
9093: of personal machines. fig-Forth, and, to a lesser extent, F83, achieved
9094: this goal by manually coding the engine in assembly language for several
9095: then-popular processors. This approach is very labor-intensive and the
9096: results are short-lived due to progress in computer architecture.
9097:
9098: @cindex C, using C for the engine
9099: Others have avoided this problem by coding in C, e.g., Mitch Bradley
9100: (cforth), Mikael Patel (TILE) and Dirk Zoller (pfe). This approach is
9101: particularly popular for UNIX-based Forths due to the large variety of
9102: architectures of UNIX machines. Unfortunately an implementation in C
9103: does not mix well with the goals of efficiency and with using
9104: traditional techniques: Indirect or direct threading cannot be expressed
9105: in C, and switch threading, the fastest technique available in C, is
9106: significantly slower. Another problem with C is that it is very
9107: cumbersome to express double integer arithmetic.
9108:
9109: @cindex GNU C for the engine
9110: @cindex long long
9111: Fortunately, there is a portable language that does not have these
9112: limitations: GNU C, the version of C processed by the GNU C compiler
9113: (@pxref{C Extensions, , Extensions to the C Language Family, gcc.info,
9114: GNU C Manual}). Its labels as values feature (@pxref{Labels as Values, ,
9115: Labels as Values, gcc.info, GNU C Manual}) makes direct and indirect
9116: threading possible, its @code{long long} type (@pxref{Long Long, ,
9117: Double-Word Integers, gcc.info, GNU C Manual}) corresponds to Forth's
9118: double numbers@footnote{Unfortunately, long longs are not implemented
9119: properly on all machines (e.g., on alpha-osf1, long longs are only 64
9120: bits, the same size as longs (and pointers), but they should be twice as
1.4 anton 9121: long according to @pxref{Long Long, , Double-Word Integers, gcc.info, GNU
1.1 anton 9122: C Manual}). So, we had to implement doubles in C after all. Still, on
9123: most machines we can use long longs and achieve better performance than
9124: with the emulation package.}. GNU C is available for free on all
9125: important (and many unimportant) UNIX machines, VMS, 80386s running
9126: MS-DOS, the Amiga, and the Atari ST, so a Forth written in GNU C can run
9127: on all these machines.
9128:
9129: Writing in a portable language has the reputation of producing code that
9130: is slower than assembly. For our Forth engine we repeatedly looked at
9131: the code produced by the compiler and eliminated most compiler-induced
9132: inefficiencies by appropriate changes in the source code.
9133:
9134: @cindex explicit register declarations
9135: @cindex --enable-force-reg, configuration flag
9136: @cindex -DFORCE_REG
9137: However, register allocation cannot be portably influenced by the
9138: programmer, leading to some inefficiencies on register-starved
9139: machines. We use explicit register declarations (@pxref{Explicit Reg
9140: Vars, , Variables in Specified Registers, gcc.info, GNU C Manual}) to
9141: improve the speed on some machines. They are turned on by using the
9142: configuration flag @code{--enable-force-reg} (@code{gcc} switch
9143: @code{-DFORCE_REG}). Unfortunately, this feature not only depends on the
9144: machine, but also on the compiler version: On some machines some
9145: compiler versions produce incorrect code when certain explicit register
9146: declarations are used. So by default @code{-DFORCE_REG} is not used.
9147:
9148: @node Threading, Primitives, Portability, Engine
9149: @section Threading
9150: @cindex inner interpreter implementation
9151: @cindex threaded code implementation
9152:
9153: @cindex labels as values
9154: GNU C's labels as values extension (available since @code{gcc-2.0},
9155: @pxref{Labels as Values, , Labels as Values, gcc.info, GNU C Manual})
9156: makes it possible to take the address of @var{label} by writing
9157: @code{&&@var{label}}. This address can then be used in a statement like
9158: @code{goto *@var{address}}. I.e., @code{goto *&&x} is the same as
9159: @code{goto x}.
9160:
9161: @cindex NEXT, indirect threaded
9162: @cindex indirect threaded inner interpreter
9163: @cindex inner interpreter, indirect threaded
9164: With this feature an indirect threaded NEXT looks like:
9165: @example
9166: cfa = *ip++;
9167: ca = *cfa;
9168: goto *ca;
9169: @end example
9170: @cindex instruction pointer
9171: For those unfamiliar with the names: @code{ip} is the Forth instruction
9172: pointer; the @code{cfa} (code-field address) corresponds to ANS Forths
9173: execution token and points to the code field of the next word to be
9174: executed; The @code{ca} (code address) fetched from there points to some
9175: executable code, e.g., a primitive or the colon definition handler
9176: @code{docol}.
9177:
9178: @cindex NEXT, direct threaded
9179: @cindex direct threaded inner interpreter
9180: @cindex inner interpreter, direct threaded
9181: Direct threading is even simpler:
9182: @example
9183: ca = *ip++;
9184: goto *ca;
9185: @end example
9186:
9187: Of course we have packaged the whole thing neatly in macros called
9188: @code{NEXT} and @code{NEXT1} (the part of NEXT after fetching the cfa).
9189:
9190: @menu
9191: * Scheduling::
9192: * Direct or Indirect Threaded?::
9193: * DOES>::
9194: @end menu
9195:
9196: @node Scheduling, Direct or Indirect Threaded?, Threading, Threading
9197: @subsection Scheduling
9198: @cindex inner interpreter optimization
9199:
9200: There is a little complication: Pipelined and superscalar processors,
9201: i.e., RISC and some modern CISC machines can process independent
9202: instructions while waiting for the results of an instruction. The
9203: compiler usually reorders (schedules) the instructions in a way that
9204: achieves good usage of these delay slots. However, on our first tries
9205: the compiler did not do well on scheduling primitives. E.g., for
9206: @code{+} implemented as
9207: @example
9208: n=sp[0]+sp[1];
9209: sp++;
9210: sp[0]=n;
9211: NEXT;
9212: @end example
9213: the NEXT comes strictly after the other code, i.e., there is nearly no
9214: scheduling. After a little thought the problem becomes clear: The
1.21 crook 9215: compiler cannot know that @code{sp} and @code{ip} point to different
9216: addresses (and the version of @code{gcc} we used would not know it even
9217: if it was possible), so it could not move the load of the cfa above the
9218: store to the TOS. Indeed the pointers could be the same, if code on or
9219: very near the top of stack were executed. In the interest of speed we
9220: chose to forbid this probably unused ``feature'' and helped the compiler
9221: in scheduling: NEXT is divided into the loading part (@code{NEXT_P1})
9222: and the goto part (@code{NEXT_P2}). @code{+} now looks like:
1.1 anton 9223: @example
9224: n=sp[0]+sp[1];
9225: sp++;
9226: NEXT_P1;
9227: sp[0]=n;
9228: NEXT_P2;
9229: @end example
9230: This can be scheduled optimally by the compiler.
9231:
9232: This division can be turned off with the switch @code{-DCISC_NEXT}. This
9233: switch is on by default on machines that do not profit from scheduling
9234: (e.g., the 80386), in order to preserve registers.
9235:
9236: @node Direct or Indirect Threaded?, DOES>, Scheduling, Threading
9237: @subsection Direct or Indirect Threaded?
9238: @cindex threading, direct or indirect?
9239:
9240: @cindex -DDIRECT_THREADED
9241: Both! After packaging the nasty details in macro definitions we
9242: realized that we could switch between direct and indirect threading by
9243: simply setting a compilation flag (@code{-DDIRECT_THREADED}) and
9244: defining a few machine-specific macros for the direct-threading case.
9245: On the Forth level we also offer access words that hide the
9246: differences between the threading methods (@pxref{Threading Words}).
9247:
9248: Indirect threading is implemented completely machine-independently.
9249: Direct threading needs routines for creating jumps to the executable
1.21 crook 9250: code (e.g. to @code{docol} or @code{dodoes}). These routines are inherently
9251: machine-dependent, but they do not amount to many source lines. Therefore,
9252: even porting direct threading to a new machine requires little effort.
1.1 anton 9253:
9254: @cindex --enable-indirect-threaded, configuration flag
9255: @cindex --enable-direct-threaded, configuration flag
9256: The default threading method is machine-dependent. You can enforce a
9257: specific threading method when building Gforth with the configuration
9258: flag @code{--enable-direct-threaded} or
9259: @code{--enable-indirect-threaded}. Note that direct threading is not
9260: supported on all machines.
9261:
9262: @node DOES>, , Direct or Indirect Threaded?, Threading
9263: @subsection DOES>
9264: @cindex @code{DOES>} implementation
9265:
9266: @cindex dodoes routine
9267: @cindex DOES-code
9268: One of the most complex parts of a Forth engine is @code{dodoes}, i.e.,
9269: the chunk of code executed by every word defined by a
9270: @code{CREATE}...@code{DOES>} pair. The main problem here is: How to find
9271: the Forth code to be executed, i.e. the code after the
9272: @code{DOES>} (the DOES-code)? There are two solutions:
9273:
1.21 crook 9274: In fig-Forth the code field points directly to the @code{dodoes} and the
1.1 anton 9275: DOES-code address is stored in the cell after the code address (i.e. at
9276: @code{@var{cfa} cell+}). It may seem that this solution is illegal in
9277: the Forth-79 and all later standards, because in fig-Forth this address
9278: lies in the body (which is illegal in these standards). However, by
9279: making the code field larger for all words this solution becomes legal
9280: again. We use this approach for the indirect threaded version and for
9281: direct threading on some machines. Leaving a cell unused in most words
9282: is a bit wasteful, but on the machines we are targeting this is hardly a
9283: problem. The other reason for having a code field size of two cells is
9284: to avoid having different image files for direct and indirect threaded
9285: systems (direct threaded systems require two-cell code fields on many
9286: machines).
9287:
9288: @cindex DOES-handler
9289: The other approach is that the code field points or jumps to the cell
9290: after @code{DOES}. In this variant there is a jump to @code{dodoes} at
9291: this address (the DOES-handler). @code{dodoes} can then get the
9292: DOES-code address by computing the code address, i.e., the address of
9293: the jump to dodoes, and add the length of that jump field. A variant of
9294: this is to have a call to @code{dodoes} after the @code{DOES>}; then the
9295: return address (which can be found in the return register on RISCs) is
9296: the DOES-code address. Since the two cells available in the code field
9297: are used up by the jump to the code address in direct threading on many
9298: architectures, we use this approach for direct threading on these
9299: architectures. We did not want to add another cell to the code field.
9300:
9301: @node Primitives, Performance, Threading, Engine
9302: @section Primitives
9303: @cindex primitives, implementation
9304: @cindex virtual machine instructions, implementation
9305:
9306: @menu
9307: * Automatic Generation::
9308: * TOS Optimization::
9309: * Produced code::
9310: @end menu
9311:
9312: @node Automatic Generation, TOS Optimization, Primitives, Primitives
9313: @subsection Automatic Generation
9314: @cindex primitives, automatic generation
9315:
9316: @cindex @file{prims2x.fs}
9317: Since the primitives are implemented in a portable language, there is no
9318: longer any need to minimize the number of primitives. On the contrary,
9319: having many primitives has an advantage: speed. In order to reduce the
9320: number of errors in primitives and to make programming them easier, we
9321: provide a tool, the primitive generator (@file{prims2x.fs}), that
9322: automatically generates most (and sometimes all) of the C code for a
9323: primitive from the stack effect notation. The source for a primitive
9324: has the following form:
9325:
9326: @cindex primitive source format
9327: @format
9328: @var{Forth-name} @var{stack-effect} @var{category} [@var{pronounc.}]
9329: [@code{""}@var{glossary entry}@code{""}]
9330: @var{C code}
9331: [@code{:}
9332: @var{Forth code}]
9333: @end format
9334:
9335: The items in brackets are optional. The category and glossary fields
9336: are there for generating the documentation, the Forth code is there
9337: for manual implementations on machines without GNU C. E.g., the source
9338: for the primitive @code{+} is:
9339: @example
9340: + n1 n2 -- n core plus
9341: n = n1+n2;
9342: @end example
9343:
9344: This looks like a specification, but in fact @code{n = n1+n2} is C
9345: code. Our primitive generation tool extracts a lot of information from
9346: the stack effect notations@footnote{We use a one-stack notation, even
9347: though we have separate data and floating-point stacks; The separate
9348: notation can be generated easily from the unified notation.}: The number
9349: of items popped from and pushed on the stack, their type, and by what
9350: name they are referred to in the C code. It then generates a C code
9351: prelude and postlude for each primitive. The final C code for @code{+}
9352: looks like this:
9353:
9354: @example
9355: I_plus: /* + ( n1 n2 -- n ) */ /* label, stack effect */
9356: /* */ /* documentation */
9357: @{
9358: DEF_CA /* definition of variable ca (indirect threading) */
9359: Cell n1; /* definitions of variables */
9360: Cell n2;
9361: Cell n;
9362: n1 = (Cell) sp[1]; /* input */
9363: n2 = (Cell) TOS;
9364: sp += 1; /* stack adjustment */
9365: NAME("+") /* debugging output (with -DDEBUG) */
9366: @{
9367: n = n1+n2; /* C code taken from the source */
9368: @}
9369: NEXT_P1; /* NEXT part 1 */
9370: TOS = (Cell)n; /* output */
9371: NEXT_P2; /* NEXT part 2 */
9372: @}
9373: @end example
9374:
9375: This looks long and inefficient, but the GNU C compiler optimizes quite
9376: well and produces optimal code for @code{+} on, e.g., the R3000 and the
9377: HP RISC machines: Defining the @code{n}s does not produce any code, and
9378: using them as intermediate storage also adds no cost.
9379:
9380: There are also other optimizations, that are not illustrated by this
9381: example: Assignments between simple variables are usually for free (copy
9382: propagation). If one of the stack items is not used by the primitive
9383: (e.g. in @code{drop}), the compiler eliminates the load from the stack
9384: (dead code elimination). On the other hand, there are some things that
9385: the compiler does not do, therefore they are performed by
9386: @file{prims2x.fs}: The compiler does not optimize code away that stores
9387: a stack item to the place where it just came from (e.g., @code{over}).
9388:
9389: While programming a primitive is usually easy, there are a few cases
9390: where the programmer has to take the actions of the generator into
9391: account, most notably @code{?dup}, but also words that do not (always)
9392: fall through to NEXT.
9393:
9394: @node TOS Optimization, Produced code, Automatic Generation, Primitives
9395: @subsection TOS Optimization
9396: @cindex TOS optimization for primitives
9397: @cindex primitives, keeping the TOS in a register
9398:
9399: An important optimization for stack machine emulators, e.g., Forth
9400: engines, is keeping one or more of the top stack items in
9401: registers. If a word has the stack effect @var{in1}...@var{inx} @code{--}
9402: @var{out1}...@var{outy}, keeping the top @var{n} items in registers
9403: @itemize @bullet
9404: @item
9405: is better than keeping @var{n-1} items, if @var{x>=n} and @var{y>=n},
9406: due to fewer loads from and stores to the stack.
9407: @item is slower than keeping @var{n-1} items, if @var{x<>y} and @var{x<n} and
9408: @var{y<n}, due to additional moves between registers.
9409: @end itemize
9410:
9411: @cindex -DUSE_TOS
9412: @cindex -DUSE_NO_TOS
9413: In particular, keeping one item in a register is never a disadvantage,
9414: if there are enough registers. Keeping two items in registers is a
9415: disadvantage for frequent words like @code{?branch}, constants,
9416: variables, literals and @code{i}. Therefore our generator only produces
9417: code that keeps zero or one items in registers. The generated C code
9418: covers both cases; the selection between these alternatives is made at
9419: C-compile time using the switch @code{-DUSE_TOS}. @code{TOS} in the C
9420: code for @code{+} is just a simple variable name in the one-item case,
9421: otherwise it is a macro that expands into @code{sp[0]}. Note that the
9422: GNU C compiler tries to keep simple variables like @code{TOS} in
9423: registers, and it usually succeeds, if there are enough registers.
9424:
9425: @cindex -DUSE_FTOS
9426: @cindex -DUSE_NO_FTOS
9427: The primitive generator performs the TOS optimization for the
9428: floating-point stack, too (@code{-DUSE_FTOS}). For floating-point
9429: operations the benefit of this optimization is even larger:
9430: floating-point operations take quite long on most processors, but can be
9431: performed in parallel with other operations as long as their results are
9432: not used. If the FP-TOS is kept in a register, this works. If
9433: it is kept on the stack, i.e., in memory, the store into memory has to
9434: wait for the result of the floating-point operation, lengthening the
9435: execution time of the primitive considerably.
9436:
9437: The TOS optimization makes the automatic generation of primitives a
9438: bit more complicated. Just replacing all occurrences of @code{sp[0]} by
9439: @code{TOS} is not sufficient. There are some special cases to
9440: consider:
9441: @itemize @bullet
9442: @item In the case of @code{dup ( w -- w w )} the generator must not
9443: eliminate the store to the original location of the item on the stack,
9444: if the TOS optimization is turned on.
9445: @item Primitives with stack effects of the form @code{--}
9446: @var{out1}...@var{outy} must store the TOS to the stack at the start.
9447: Likewise, primitives with the stack effect @var{in1}...@var{inx} @code{--}
9448: must load the TOS from the stack at the end. But for the null stack
9449: effect @code{--} no stores or loads should be generated.
9450: @end itemize
9451:
9452: @node Produced code, , TOS Optimization, Primitives
9453: @subsection Produced code
9454: @cindex primitives, assembly code listing
9455:
9456: @cindex @file{engine.s}
9457: To see what assembly code is produced for the primitives on your machine
9458: with your compiler and your flag settings, type @code{make engine.s} and
9459: look at the resulting file @file{engine.s}.
9460:
9461: @node Performance, , Primitives, Engine
9462: @section Performance
9463: @cindex performance of some Forth interpreters
9464: @cindex engine performance
9465: @cindex benchmarking Forth systems
9466: @cindex Gforth performance
9467:
9468: On RISCs the Gforth engine is very close to optimal; i.e., it is usually
9469: impossible to write a significantly faster engine.
9470:
9471: On register-starved machines like the 386 architecture processors
9472: improvements are possible, because @code{gcc} does not utilize the
9473: registers as well as a human, even with explicit register declarations;
9474: e.g., Bernd Beuster wrote a Forth system fragment in assembly language
9475: and hand-tuned it for the 486; this system is 1.19 times faster on the
9476: Sieve benchmark on a 486DX2/66 than Gforth compiled with
9477: @code{gcc-2.6.3} with @code{-DFORCE_REG}.
9478:
9479: @cindex Win32Forth performance
9480: @cindex NT Forth performance
9481: @cindex eforth performance
9482: @cindex ThisForth performance
9483: @cindex PFE performance
9484: @cindex TILE performance
9485: However, this potential advantage of assembly language implementations
9486: is not necessarily realized in complete Forth systems: We compared
9487: Gforth (direct threaded, compiled with @code{gcc-2.6.3} and
9488: @code{-DFORCE_REG}) with Win32Forth 1.2093, LMI's NT Forth (Beta, May
9489: 1994) and Eforth (with and without peephole (aka pinhole) optimization
9490: of the threaded code); all these systems were written in assembly
9491: language. We also compared Gforth with three systems written in C:
9492: PFE-0.9.14 (compiled with @code{gcc-2.6.3} with the default
9493: configuration for Linux: @code{-O2 -fomit-frame-pointer -DUSE_REGS
1.21 crook 9494: -DUNROLL_NEXT}), ThisForth Beta (compiled with @code{gcc-2.6.3 -O3
9495: -fomit-frame-pointer}; ThisForth employs peephole optimization of the
1.1 anton 9496: threaded code) and TILE (compiled with @code{make opt}). We benchmarked
9497: Gforth, PFE, ThisForth and TILE on a 486DX2/66 under Linux. Kenneth
9498: O'Heskin kindly provided the results for Win32Forth and NT Forth on a
9499: 486DX2/66 with similar memory performance under Windows NT. Marcel
9500: Hendrix ported Eforth to Linux, then extended it to run the benchmarks,
9501: added the peephole optimizer, ran the benchmarks and reported the
9502: results.
9503:
9504: We used four small benchmarks: the ubiquitous Sieve; bubble-sorting and
9505: matrix multiplication come from the Stanford integer benchmarks and have
9506: been translated into Forth by Martin Fraeman; we used the versions
9507: included in the TILE Forth package, but with bigger data set sizes; and
9508: a recursive Fibonacci number computation for benchmarking calling
9509: performance. The following table shows the time taken for the benchmarks
9510: scaled by the time taken by Gforth (in other words, it shows the speedup
9511: factor that Gforth achieved over the other systems).
9512:
9513: @example
9514: relative Win32- NT eforth This-
9515: time Gforth Forth Forth eforth +opt PFE Forth TILE
9516: sieve 1.00 1.39 1.14 1.39 0.85 1.58 3.18 8.58
9517: bubble 1.00 1.31 1.41 1.48 0.88 1.50 3.88
9518: matmul 1.00 1.47 1.35 1.46 0.74 1.58 4.09
9519: fib 1.00 1.52 1.34 1.22 0.86 1.74 2.99 4.30
9520: @end example
9521:
9522: You may find the good performance of Gforth compared with the systems
9523: written in assembly language quite surprising. One important reason for
9524: the disappointing performance of these systems is probably that they are
9525: not written optimally for the 486 (e.g., they use the @code{lods}
9526: instruction). In addition, Win32Forth uses a comfortable, but costly
9527: method for relocating the Forth image: like @code{cforth}, it computes
9528: the actual addresses at run time, resulting in two address computations
9529: per NEXT (@pxref{Image File Background}).
9530:
9531: Only Eforth with the peephole optimizer performs comparable to
9532: Gforth. The speedups achieved with peephole optimization of threaded
9533: code are quite remarkable. Adding a peephole optimizer to Gforth should
9534: cause similar speedups.
9535:
9536: The speedup of Gforth over PFE, ThisForth and TILE can be easily
9537: explained with the self-imposed restriction of the latter systems to
9538: standard C, which makes efficient threading impossible (however, the
1.4 anton 9539: measured implementation of PFE uses a GNU C extension: @pxref{Global Reg
1.1 anton 9540: Vars, , Defining Global Register Variables, gcc.info, GNU C Manual}).
9541: Moreover, current C compilers have a hard time optimizing other aspects
9542: of the ThisForth and the TILE source.
9543:
9544: Note that the performance of Gforth on 386 architecture processors
9545: varies widely with the version of @code{gcc} used. E.g., @code{gcc-2.5.8}
9546: failed to allocate any of the virtual machine registers into real
9547: machine registers by itself and would not work correctly with explicit
9548: register declarations, giving a 1.3 times slower engine (on a 486DX2/66
9549: running the Sieve) than the one measured above.
9550:
9551: Note also that there have been several releases of Win32Forth since the
9552: release presented here, so the results presented here may have little
9553: predictive value for the performance of Win32Forth today.
9554:
9555: @cindex @file{Benchres}
9556: In @cite{Translating Forth to Efficient C} by M. Anton Ertl and Martin
9557: Maierhofer (presented at EuroForth '95), an indirect threaded version of
9558: Gforth is compared with Win32Forth, NT Forth, PFE, and ThisForth; that
9559: version of Gforth is 2%@minus{}8% slower on a 486 than the direct
9560: threaded version used here. The paper available at
9561: @*@url{http://www.complang.tuwien.ac.at/papers/ertl&maierhofer95.ps.gz};
9562: it also contains numbers for some native code systems. You can find a
9563: newer version of these measurements at
9564: @url{http://www.complang.tuwien.ac.at/forth/performance.html}. You can
9565: find numbers for Gforth on various machines in @file{Benchres}.
9566:
1.13 pazsan 9567: @node Binding to System Library, Cross Compiler, Engine, Top
1.14 pazsan 9568: @chapter Binding to System Library
1.13 pazsan 9569:
9570: @node Cross Compiler, Bugs, Binding to System Library, Top
1.14 pazsan 9571: @chapter Cross Compiler
1.13 pazsan 9572:
9573: Cross Compiler
9574:
9575: @menu
9576: * Using the Cross Compiler::
9577: * How the Cross Compiler Works::
9578: @end menu
9579:
1.21 crook 9580: @node Using the Cross Compiler, How the Cross Compiler Works, Cross Compiler, Cross Compiler
1.14 pazsan 9581: @section Using the Cross Compiler
1.13 pazsan 9582:
1.21 crook 9583: @node How the Cross Compiler Works, , Using the Cross Compiler, Cross Compiler
1.14 pazsan 9584: @section How the Cross Compiler Works
1.13 pazsan 9585:
9586: @node Bugs, Origin, Cross Compiler, Top
1.21 crook 9587: @appendix Bugs
1.1 anton 9588: @cindex bug reporting
9589:
1.21 crook 9590: Known bugs are described in the file @file{BUGS} in the Gforth distribution.
1.1 anton 9591:
9592: If you find a bug, please send a bug report to
1.21 crook 9593: @email{bug-gforth@@gnu.ai.mit.edu}. A bug report should include this
9594: information:
9595:
9596: @itemize @bullet
9597: @item
9598: The Gforth version used (it is announced at the start of an
9599: interactive Gforth session).
9600: @item
9601: The machine and operating system (on Unix
9602: systems @code{uname -a} will report this information).
9603: @item
9604: The installation options (send the file @file{config.status}).
9605: @item
9606: A complete list of changes (if any) you (or your installer) have made to the
9607: Gforth sources.
9608: @item
9609: A program (or a sequence of keyboard commands) that reproduces the bug.
9610: @item
9611: A description of what you think constitutes the buggy behaviour.
9612: @end itemize
1.1 anton 9613:
9614: For a thorough guide on reporting bugs read @ref{Bug Reporting, , How
9615: to Report Bugs, gcc.info, GNU C Manual}.
9616:
9617:
1.21 crook 9618: @node Origin, Forth-related information, Bugs, Top
9619: @appendix Authors and Ancestors of Gforth
1.1 anton 9620:
9621: @section Authors and Contributors
9622: @cindex authors of Gforth
9623: @cindex contributors to Gforth
9624:
9625: The Gforth project was started in mid-1992 by Bernd Paysan and Anton
9626: Ertl. The third major author was Jens Wilke. Lennart Benschop (who was
9627: one of Gforth's first users, in mid-1993) and Stuart Ramsden inspired us
9628: with their continuous feedback. Lennart Benshop contributed
9629: @file{glosgen.fs}, while Stuart Ramsden has been working on automatic
9630: support for calling C libraries. Helpful comments also came from Paul
9631: Kleinrubatscher, Christian Pirker, Dirk Zoller, Marcel Hendrix, John
1.12 anton 9632: Wavrik, Barrie Stott, Marc de Groot, and Jorge Acerada. Since the
9633: release of Gforth-0.2.1 there were also helpful comments from many
9634: others; thank you all, sorry for not listing you here (but digging
9635: through my mailbox to extract your names is on my to-do list).
1.1 anton 9636:
9637: Gforth also owes a lot to the authors of the tools we used (GCC, CVS,
9638: and autoconf, among others), and to the creators of the Internet: Gforth
1.21 crook 9639: was developed across the Internet, and its authors did not meet
1.20 pazsan 9640: physically for the first 4 years of development.
1.1 anton 9641:
9642: @section Pedigree
9643: @cindex Pedigree of Gforth
9644:
1.20 pazsan 9645: Gforth descends from bigFORTH (1993) and fig-Forth. Gforth and PFE (by
1.1 anton 9646: Dirk Zoller) will cross-fertilize each other. Of course, a significant
9647: part of the design of Gforth was prescribed by ANS Forth.
9648:
1.20 pazsan 9649: Bernd Paysan wrote bigFORTH, a descendent from TurboForth, an unreleased
1.1 anton 9650: 32 bit native code version of VolksForth for the Atari ST, written
9651: mostly by Dietrich Weineck.
9652:
9653: VolksForth descends from F83. It was written by Klaus Schleisiek, Bernd
9654: Pennemann, Georg Rehfeld and Dietrich Weineck for the C64 (called
9655: UltraForth there) in the mid-80s and ported to the Atari ST in 1986.
9656:
9657: Henry Laxen and Mike Perry wrote F83 as a model implementation of the
9658: Forth-83 standard. !! Pedigree? When?
9659:
9660: A team led by Bill Ragsdale implemented fig-Forth on many processors in
9661: 1979. Robert Selzer and Bill Ragsdale developed the original
9662: implementation of fig-Forth for the 6502 based on microForth.
9663:
9664: The principal architect of microForth was Dean Sanderson. microForth was
9665: FORTH, Inc.'s first off-the-shelf product. It was developed in 1976 for
9666: the 1802, and subsequently implemented on the 8080, the 6800 and the
9667: Z80.
9668:
9669: All earlier Forth systems were custom-made, usually by Charles Moore,
9670: who discovered (as he puts it) Forth during the late 60s. The first full
9671: Forth existed in 1971.
9672:
9673: A part of the information in this section comes from @cite{The Evolution
9674: of Forth} by Elizabeth D. Rather, Donald R. Colburn and Charles
9675: H. Moore, presented at the HOPL-II conference and preprinted in SIGPLAN
9676: Notices 28(3), 1993. You can find more historical and genealogical
9677: information about Forth there.
9678:
1.21 crook 9679: @node Forth-related information, Word Index, Origin, Top
9680: @appendix Other Forth-related information
9681: @cindex Forth-related information
9682:
9683: @menu
9684: * Internet resources::
9685: * Books::
9686: * The Forth Interest Group::
9687: * Conferences::
9688: @end menu
9689:
9690:
9691: @node Internet resources, Books, Forth-related information, Forth-related information
9692: @section Internet resources
9693: @cindex Internet resources
9694:
9695: @cindex comp.lang.forth
9696: @cindex frequently asked questions
9697: There is an active newsgroup (comp.lang.forth) discussing Forth and
9698: Forth-related issues. A frequently-asked-questions (FAQ) list
9699: is posted to the newsgroup regulary, and archived at these sites:
9700:
9701: @itemize @bullet
9702: @item
9703: @url{ftp://rtfm.mit.edu/pub/usenet-by-group/comp.lang.forth/}
9704: @item
9705: @url{ftp://ftp.forth.org/pub/Forth/FAQ/}
9706: @end itemize
9707:
9708: The FAQ list should be considered mandatory reading before posting to
9709: the newsgroup.
9710:
9711: Here are some other web sites holding Forth-related material:
9712:
9713: @itemize @bullet
9714: @item
9715: @url{http://www.taygeta.com/forth.html} -- Skip Carter's Forth pages.
9716: @item
9717: @url{http://www.jwdt.com/~paysan/gforth.html} -- the Gforth home page.
9718: @item
9719: @url{http://www.minerva.com/uathena.htm} -- home of ANS Forth Standard.
9720: @item
9721: @url{http://dec.bournemouth.ac.uk/forth/index.html} -- the Forth
9722: Research page, including links to the Journal of Forth Application and
9723: Research (JFAR) and a searchable Forth bibliography.
9724: @end itemize
9725:
9726:
9727: @node Books, The Forth Interest Group, Internet resources, Forth-related information
9728: @section Books
9729: @cindex Books
9730:
9731: As the Standard is relatively new, there are not many books out yet. It
9732: is not recommended to learn Forth by using Gforth and a book that is not
9733: written for ANS Forth, as you will not know your mistakes from the
9734: deviations of the book. However, books based on the Forth-83 standard
9735: should be ok, because ANS Forth is primarily an extension of Forth-83.
9736:
9737: @cindex standard document for ANS Forth
9738: @cindex ANS Forth document
9739: The definite reference if you want to write ANS Forth programs is, of
9740: course, the ANS Forth Standard. It is available in printed form from the
9741: National Standards Institute Sales Department (Tel.: USA (212) 642-4900;
9742: Fax.: USA (212) 302-1286) as document @cite{X3.215-1994} for about
9743: $200. You can also get it from Global Engineering Documents (Tel.: USA
9744: (800) 854-7179; Fax.: (303) 843-9880) for about $300.
9745:
9746: @cite{dpANS6}, the last draft of the standard, which was then submitted
9747: to ANSI for publication is available electronically and for free in some
9748: MS Word format, and it has been converted to HTML
9749: (@url{http://www.taygeta.com/forth/dpans.html}; this is my favourite
9750: format); this HTML version also includes the answers to Requests for
9751: Interpretation (RFIs). Some pointers to these versions can be found
9752: through @*@url{http://www.complang.tuwien.ac.at/projects/forth.html}.
9753:
9754: @cindex introductory book
9755: @cindex book, introductory
9756: @cindex Woehr, Jack: @cite{Forth: The New Model}
9757: @cindex @cite{Forth: The new model} (book)
9758: @cite{Forth: The New Model} by Jack Woehr (Prentice-Hall, 1993) is an
9759: introductory book based on a draft version of the standard. It does not
9760: cover the whole standard. It also contains interesting background
9761: information (Jack Woehr was in the ANS Forth Technical Committee). It is
9762: not appropriate for complete newbies, but programmers experienced in
9763: other languages should find it ok.
9764:
9765: @cindex Conklin, Edward K., and Elizabeth Rather: @cite{Forth Programmer's Handbook}
9766: @cindex Rather, Elizabeth and Edward K. Conklin: @cite{Forth Programmer's Handbook}
9767: @cindex @cite{Forth Programmer's Handbook} (book)
9768: @cite{Forth Programmer's Handbook} by Edward K. Conklin, Elizabeth
9769: D. Rather and the technical staff of Forth, Inc. (Forth, Inc., 1997;
9770: ISBN 0-9662156-0-5) contains little introductory material. The majority
9771: of the book is similar to @ref{Words}, but the book covers most of the
9772: standard words and some non-standard words (whereas this manual is
9773: quite incomplete). In addition, the book contains a chapter on
9774: programming style. The major drawback of this book is that it usually
9775: does not identify what is standard and what is specific to the Forth
9776: system described in the book (probably one of Forth, Inc.'s systems).
9777: Fortunately, many of the non-standard programming practices described in
9778: the book work in Gforth, too. Still, this drawback makes the book
9779: hardly more useful than a pre-ANS book.
9780:
9781: @node The Forth Interest Group, Conferences, Books, Forth-related information
9782: @section The Forth Interest Group
9783: @cindex Forth interest group (FIG)
9784:
9785: The Forth Interest Group (FIG) is a world-wide, non-profit,
9786: member-supported organisation. It publishes a regular magazine and
9787: offers other benefits of membership. You can contact the FIG through
9788: their office email address: @email{office@@forth.org} or by visiting
9789: their web site at @url{http://www.forth.org/}. This web site also
9790: includes links to FIG chapters in other countries and American cities
9791: (@url{http://www.forth.org/chapters.html}).
9792:
9793: @node Conferences, , The Forth Interest Group, Forth-related information
9794: @section Conferences
9795: @cindex Conferences
9796:
9797: There are several regular conferences related to Forth. They are all
9798: well-publicised in FIG magazine and on the comp.lang.forth news group:
9799:
9800: @itemize @bullet
9801: @item
9802: FORML -- the Forth modification laboratory convenes every year near
9803: Monterey, California.
9804: @item
9805: The Rochester Forth Conference -- an annual conference traditionally
9806: held in Rochester, New York.
9807: @item
9808: EuroForth -- this European conference takes place annually.
9809: @end itemize
9810:
9811:
9812: @node Word Index, Concept Index, Forth-related information, Top
1.1 anton 9813: @unnumbered Word Index
9814:
9815: This index is as incomplete as the manual. Each word is listed with
9816: stack effect and wordset.
9817:
9818: @printindex fn
9819:
9820: @node Concept Index, , Word Index, Top
9821: @unnumbered Concept and Word Index
9822:
9823: This index is as incomplete as the manual. Not all entries listed are
9824: present verbatim in the text. Only the names are listed for the words
9825: here.
9826:
9827: @printindex cp
9828:
9829: @contents
9830: @bye
9831:
FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>