File:  [gforth] / gforth / doc / gforth.ds
Revision 1.20: download - view: text, annotated - select for diffs
Sun Dec 13 23:30:00 1998 UTC (23 years, 8 months ago) by pazsan
Branches: MAIN
CVS tags: v0-4-0, HEAD
Added some documentation (files stup, blocks stub, Mini-OOF implementation)
Added Benchres for my machine
made DOS and Win32 compile and run
New gforthmi.bat script for DOS - needs a temporary file for the commands
instead of the -e option.
Added select.o again for DOS (DJGPP's select is broken wrt timing)
Improved select.c
Bug with DOS: engine-ditc doesn't compile with optimization on. Maybe I need
to get a new GCC version for DOS?

    1: \input texinfo   @c -*-texinfo-*-
    2: @comment The source is gforth.ds, from which gforth.texi is generated
    3: @comment %**start of header (This is for running Texinfo on a region.)
    4: @setfilename gforth.info
    5: @settitle Gforth Manual
    6: @dircategory GNU programming tools
    7: @direntry
    8: * Gforth: (gforth).             A fast interpreter for the Forth language.
    9: @end direntry
   10: @comment @setchapternewpage odd
   11: @macro progstyle {}
   12: Programming style note:
   13: @end macro
   14: @comment %**end of header (This is for running Texinfo on a region.)
   15: 
   16: @include version.texi
   17: 
   18: @ifinfo
   19: This file documents Gforth @value{VERSION}
   20: 
   21: Copyright @copyright{} 1995-1998 Free Software Foundation, Inc.
   22: 
   23:      Permission is granted to make and distribute verbatim copies of
   24:      this manual provided the copyright notice and this permission notice
   25:      are preserved on all copies.
   26:      
   27: @ignore
   28:      Permission is granted to process this file through TeX and print the
   29:      results, provided the printed document carries a copying permission
   30:      notice identical to this one except for the removal of this paragraph
   31:      (this paragraph not being relevant to the printed manual).
   32:      
   33: @end ignore
   34:      Permission is granted to copy and distribute modified versions of this
   35:      manual under the conditions for verbatim copying, provided also that the
   36:      sections entitled "Distribution" and "General Public License" are
   37:      included exactly as in the original, and provided that the entire
   38:      resulting derived work is distributed under the terms of a permission
   39:      notice identical to this one.
   40:      
   41:      Permission is granted to copy and distribute translations of this manual
   42:      into another language, under the above conditions for modified versions,
   43:      except that the sections entitled "Distribution" and "General Public
   44:      License" may be included in a translation approved by the author instead
   45:      of in the original English.
   46: @end ifinfo
   47: 
   48: @finalout
   49: @titlepage
   50: @sp 10
   51: @center @titlefont{Gforth Manual}
   52: @sp 2
   53: @center for version @value{VERSION}
   54: @sp 2
   55: @center Anton Ertl
   56: @center Bernd Paysan
   57: @center Jens Wilke
   58: @sp 3
   59: @center This manual is permanently under construction
   60: 
   61: @comment  The following two commands start the copyright page.
   62: @page
   63: @vskip 0pt plus 1filll
   64: Copyright @copyright{} 1995--1998 Free Software Foundation, Inc.
   65: 
   66: @comment !! Published by ... or You can get a copy of this manual ...
   67: 
   68:      Permission is granted to make and distribute verbatim copies of
   69:      this manual provided the copyright notice and this permission notice
   70:      are preserved on all copies.
   71:      
   72:      Permission is granted to copy and distribute modified versions of this
   73:      manual under the conditions for verbatim copying, provided also that the
   74:      sections entitled "Distribution" and "General Public License" are
   75:      included exactly as in the original, and provided that the entire
   76:      resulting derived work is distributed under the terms of a permission
   77:      notice identical to this one.
   78:      
   79:      Permission is granted to copy and distribute translations of this manual
   80:      into another language, under the above conditions for modified versions,
   81:      except that the sections entitled "Distribution" and "General Public
   82:      License" may be included in a translation approved by the author instead
   83:      of in the original English.
   84: @end titlepage
   85: 
   86: 
   87: @node Top, License, (dir), (dir)
   88: @ifinfo
   89: Gforth is a free implementation of ANS Forth available on many
   90: personal machines. This manual corresponds to version @value{VERSION}.
   91: @end ifinfo
   92: 
   93: @menu
   94: * License::                     
   95: * Goals::                       About the Gforth Project
   96: * Other Books::                 Things you might want to read
   97: * Invoking Gforth::             Starting Gforth
   98: * Words::                       Forth words available in Gforth
   99: * Tools::                       Programming tools
  100: * ANS conformance::             Implementation-defined options etc.
  101: * Model::                       The abstract machine of Gforth
  102: * Integrating Gforth::          Forth as scripting language for applications
  103: * Emacs and Gforth::            The Gforth Mode
  104: * Image Files::                 @code{.fi} files contain compiled code
  105: * Engine::                      The inner interpreter and the primitives
  106: * Cross Compiler::              The Cross Compiler
  107: * Bugs::                        How to report them
  108: * Origin::                      Authors and ancestors of Gforth
  109: * Word Index::                  An item for each Forth word
  110: * Concept Index::               A menu covering many topics
  111: 
  112:  --- The Detailed Node Listing ---
  113: 
  114: Forth Words
  115: 
  116: * Notation::                    
  117: * Arithmetic::                  
  118: * Stack Manipulation::          
  119: * Memory::                      
  120: * Control Structures::          
  121: * Locals::                      
  122: * Defining Words::              
  123: * Structures::                  
  124: * Object-oriented Forth::       
  125: * Tokens for Words::            
  126: * Wordlists::                   
  127: * Files::                       
  128: * Including Files::             
  129: * Blocks::                      
  130: * Other I/O::                   
  131: * Programming Tools::           
  132: * Assembler and Code Words::    
  133: * Threading Words::             
  134: 
  135: Arithmetic
  136: 
  137: * Single precision::            
  138: * Bitwise operations::          
  139: * Mixed precision::             operations with single and double-cell integers
  140: * Double precision::            Double-cell integer arithmetic
  141: * Floating Point::              
  142: 
  143: Stack Manipulation
  144: 
  145: * Data stack::                  
  146: * Floating point stack::        
  147: * Return stack::                
  148: * Locals stack::                
  149: * Stack pointer manipulation::  
  150: 
  151: Memory
  152: 
  153: * Memory Access::      
  154: * Address arithmetic::          
  155: * Memory Blocks::         
  156: 
  157: Control Structures
  158: 
  159: * Selection::                   
  160: * Simple Loops::                
  161: * Counted Loops::               
  162: * Arbitrary control structures::  
  163: * Calls and returns::           
  164: * Exception Handling::          
  165: 
  166: Locals
  167: 
  168: * Gforth locals::               
  169: * ANS Forth locals::            
  170: 
  171: Gforth locals
  172: 
  173: * Where are locals visible by name?::  
  174: * How long do locals live?::    
  175: * Programming Style::           
  176: * Implementation::              
  177: 
  178: Defining Words
  179: 
  180: * Simple Defining Words::       
  181: * Colon Definitions::           
  182: * User-defined Defining Words::  
  183: * Supplying names::             
  184: * Interpretation and Compilation Semantics::  
  185: 
  186: Structures
  187: 
  188: * Why explicit structure support?::  
  189: * Structure Usage::             
  190: * Structure Naming Convention::  
  191: * Structure Implementation::    
  192: * Structure Glossary::          
  193: 
  194: Object-oriented Forth
  195: 
  196: * Objects::                     
  197: * OOF::                         
  198: * Mini-OOF::                    
  199: 
  200: Objects
  201: 
  202: * Properties of the Objects model::  
  203: * Why object-oriented programming?::  
  204: * Object-Oriented Terminology::  
  205: * Basic Objects Usage::         
  206: * The class Object::            
  207: * Creating objects::            
  208: * Object-Oriented Programming Style::  
  209: * Class Binding::               
  210: * Method conveniences::         
  211: * Classes and Scoping::         
  212: * Object Interfaces::           
  213: * Objects Implementation::      
  214: * Comparison with other object models::  
  215: * Objects Glossary::            
  216: 
  217: OOF
  218: 
  219: * Properties of the OOF model::
  220: * Basic OOF Usage::
  221: * The base class object::
  222: * Class Declaration::
  223: * Class Implementation::
  224: 
  225: Including Files
  226: 
  227: * Words for Including::         
  228: * Search Path::                 
  229: * Changing the Search Path::    
  230: * General Search Paths::        
  231: 
  232: Programming Tools
  233: 
  234: * Debugging::                   Simple and quick.
  235: * Assertions::                  Making your programs self-checking.
  236: * Singlestep Debugger::		Executing your program word by word.
  237: 
  238: Tools
  239: 
  240: * ANS Report::                  Report the words used, sorted by wordset.
  241: 
  242: ANS conformance
  243: 
  244: * The Core Words::              
  245: * The optional Block word set::  
  246: * The optional Double Number word set::  
  247: * The optional Exception word set::  
  248: * The optional Facility word set::  
  249: * The optional File-Access word set::  
  250: * The optional Floating-Point word set::  
  251: * The optional Locals word set::  
  252: * The optional Memory-Allocation word set::  
  253: * The optional Programming-Tools word set::  
  254: * The optional Search-Order word set::  
  255: 
  256: The Core Words
  257: 
  258: * core-idef::                   Implementation Defined Options                   
  259: * core-ambcond::                Ambiguous Conditions                
  260: * core-other::                  Other System Documentation                  
  261: 
  262: The optional Block word set
  263: 
  264: * block-idef::                  Implementation Defined Options
  265: * block-ambcond::               Ambiguous Conditions               
  266: * block-other::                 Other System Documentation                 
  267: 
  268: The optional Double Number word set
  269: 
  270: * double-ambcond::              Ambiguous Conditions              
  271: 
  272: The optional Exception word set
  273: 
  274: * exception-idef::              Implementation Defined Options              
  275: 
  276: The optional Facility word set
  277: 
  278: * facility-idef::               Implementation Defined Options               
  279: * facility-ambcond::            Ambiguous Conditions            
  280: 
  281: The optional File-Access word set
  282: 
  283: * file-idef::                   Implementation Defined Options
  284: * file-ambcond::                Ambiguous Conditions                
  285: 
  286: The optional Floating-Point word set
  287: 
  288: * floating-idef::               Implementation Defined Options
  289: * floating-ambcond::            Ambiguous Conditions            
  290: 
  291: The optional Locals word set
  292: 
  293: * locals-idef::                 Implementation Defined Options                 
  294: * locals-ambcond::              Ambiguous Conditions              
  295: 
  296: The optional Memory-Allocation word set
  297: 
  298: * memory-idef::                 Implementation Defined Options                 
  299: 
  300: The optional Programming-Tools word set
  301: 
  302: * programming-idef::            Implementation Defined Options            
  303: * programming-ambcond::         Ambiguous Conditions         
  304: 
  305: The optional Search-Order word set
  306: 
  307: * search-idef::                 Implementation Defined Options                 
  308: * search-ambcond::              Ambiguous Conditions              
  309: 
  310: Image Files
  311: 
  312: * Image File Background::          Why have image files?
  313: * Non-Relocatable Image Files::    don't always work.
  314: * Data-Relocatable Image Files::   are better.
  315: * Fully Relocatable Image Files::  better yet.
  316: * Stack and Dictionary Sizes::     Setting the default sizes for an image.
  317: * Running Image Files::            @code{gforth -i @var{file}} or @var{file}.
  318: * Modifying the Startup Sequence:: and turnkey applications.
  319: 
  320: Fully Relocatable Image Files
  321: 
  322: * gforthmi::            The normal way
  323: * cross.fs::                    The hard way
  324: 
  325: Engine
  326: 
  327: * Portability::                 
  328: * Threading::                   
  329: * Primitives::                  
  330: * Performance::                 
  331: 
  332: Threading
  333: 
  334: * Scheduling::                  
  335: * Direct or Indirect Threaded?::  
  336: * DOES>::                       
  337: 
  338: Primitives
  339: 
  340: * Automatic Generation::        
  341: * TOS Optimization::            
  342: * Produced code::               
  343: 
  344: System Libraries
  345: 
  346: * Binding to System Library::
  347: 
  348: Cross Compiler
  349: 
  350: * Using the Cross Compiler::
  351: * How the Cross Compiler Works::
  352: 
  353: @end menu
  354: 
  355: @node License, Goals, Top, Top
  356: @unnumbered GNU GENERAL PUBLIC LICENSE
  357: @center Version 2, June 1991
  358: 
  359: @display
  360: Copyright @copyright{} 1989, 1991 Free Software Foundation, Inc.
  361: 675 Mass Ave, Cambridge, MA 02139, USA
  362: 
  363: Everyone is permitted to copy and distribute verbatim copies
  364: of this license document, but changing it is not allowed.
  365: @end display
  366: 
  367: @unnumberedsec Preamble
  368: 
  369:   The licenses for most software are designed to take away your
  370: freedom to share and change it.  By contrast, the GNU General Public
  371: License is intended to guarantee your freedom to share and change free
  372: software---to make sure the software is free for all its users.  This
  373: General Public License applies to most of the Free Software
  374: Foundation's software and to any other program whose authors commit to
  375: using it.  (Some other Free Software Foundation software is covered by
  376: the GNU Library General Public License instead.)  You can apply it to
  377: your programs, too.
  378: 
  379:   When we speak of free software, we are referring to freedom, not
  380: price.  Our General Public Licenses are designed to make sure that you
  381: have the freedom to distribute copies of free software (and charge for
  382: this service if you wish), that you receive source code or can get it
  383: if you want it, that you can change the software or use pieces of it
  384: in new free programs; and that you know you can do these things.
  385: 
  386:   To protect your rights, we need to make restrictions that forbid
  387: anyone to deny you these rights or to ask you to surrender the rights.
  388: These restrictions translate to certain responsibilities for you if you
  389: distribute copies of the software, or if you modify it.
  390: 
  391:   For example, if you distribute copies of such a program, whether
  392: gratis or for a fee, you must give the recipients all the rights that
  393: you have.  You must make sure that they, too, receive or can get the
  394: source code.  And you must show them these terms so they know their
  395: rights.
  396: 
  397:   We protect your rights with two steps: (1) copyright the software, and
  398: (2) offer you this license which gives you legal permission to copy,
  399: distribute and/or modify the software.
  400: 
  401:   Also, for each author's protection and ours, we want to make certain
  402: that everyone understands that there is no warranty for this free
  403: software.  If the software is modified by someone else and passed on, we
  404: want its recipients to know that what they have is not the original, so
  405: that any problems introduced by others will not reflect on the original
  406: authors' reputations.
  407: 
  408:   Finally, any free program is threatened constantly by software
  409: patents.  We wish to avoid the danger that redistributors of a free
  410: program will individually obtain patent licenses, in effect making the
  411: program proprietary.  To prevent this, we have made it clear that any
  412: patent must be licensed for everyone's free use or not licensed at all.
  413: 
  414:   The precise terms and conditions for copying, distribution and
  415: modification follow.
  416: 
  417: @iftex
  418: @unnumberedsec TERMS AND CONDITIONS FOR COPYING, DISTRIBUTION AND MODIFICATION
  419: @end iftex
  420: @ifinfo
  421: @center TERMS AND CONDITIONS FOR COPYING, DISTRIBUTION AND MODIFICATION
  422: @end ifinfo
  423: 
  424: @enumerate 0
  425: @item
  426: This License applies to any program or other work which contains
  427: a notice placed by the copyright holder saying it may be distributed
  428: under the terms of this General Public License.  The ``Program'', below,
  429: refers to any such program or work, and a ``work based on the Program''
  430: means either the Program or any derivative work under copyright law:
  431: that is to say, a work containing the Program or a portion of it,
  432: either verbatim or with modifications and/or translated into another
  433: language.  (Hereinafter, translation is included without limitation in
  434: the term ``modification''.)  Each licensee is addressed as ``you''.
  435: 
  436: Activities other than copying, distribution and modification are not
  437: covered by this License; they are outside its scope.  The act of
  438: running the Program is not restricted, and the output from the Program
  439: is covered only if its contents constitute a work based on the
  440: Program (independent of having been made by running the Program).
  441: Whether that is true depends on what the Program does.
  442: 
  443: @item
  444: You may copy and distribute verbatim copies of the Program's
  445: source code as you receive it, in any medium, provided that you
  446: conspicuously and appropriately publish on each copy an appropriate
  447: copyright notice and disclaimer of warranty; keep intact all the
  448: notices that refer to this License and to the absence of any warranty;
  449: and give any other recipients of the Program a copy of this License
  450: along with the Program.
  451: 
  452: You may charge a fee for the physical act of transferring a copy, and
  453: you may at your option offer warranty protection in exchange for a fee.
  454: 
  455: @item
  456: You may modify your copy or copies of the Program or any portion
  457: of it, thus forming a work based on the Program, and copy and
  458: distribute such modifications or work under the terms of Section 1
  459: above, provided that you also meet all of these conditions:
  460: 
  461: @enumerate a
  462: @item
  463: You must cause the modified files to carry prominent notices
  464: stating that you changed the files and the date of any change.
  465: 
  466: @item
  467: You must cause any work that you distribute or publish, that in
  468: whole or in part contains or is derived from the Program or any
  469: part thereof, to be licensed as a whole at no charge to all third
  470: parties under the terms of this License.
  471: 
  472: @item
  473: If the modified program normally reads commands interactively
  474: when run, you must cause it, when started running for such
  475: interactive use in the most ordinary way, to print or display an
  476: announcement including an appropriate copyright notice and a
  477: notice that there is no warranty (or else, saying that you provide
  478: a warranty) and that users may redistribute the program under
  479: these conditions, and telling the user how to view a copy of this
  480: License.  (Exception: if the Program itself is interactive but
  481: does not normally print such an announcement, your work based on
  482: the Program is not required to print an announcement.)
  483: @end enumerate
  484: 
  485: These requirements apply to the modified work as a whole.  If
  486: identifiable sections of that work are not derived from the Program,
  487: and can be reasonably considered independent and separate works in
  488: themselves, then this License, and its terms, do not apply to those
  489: sections when you distribute them as separate works.  But when you
  490: distribute the same sections as part of a whole which is a work based
  491: on the Program, the distribution of the whole must be on the terms of
  492: this License, whose permissions for other licensees extend to the
  493: entire whole, and thus to each and every part regardless of who wrote it.
  494: 
  495: Thus, it is not the intent of this section to claim rights or contest
  496: your rights to work written entirely by you; rather, the intent is to
  497: exercise the right to control the distribution of derivative or
  498: collective works based on the Program.
  499: 
  500: In addition, mere aggregation of another work not based on the Program
  501: with the Program (or with a work based on the Program) on a volume of
  502: a storage or distribution medium does not bring the other work under
  503: the scope of this License.
  504: 
  505: @item
  506: You may copy and distribute the Program (or a work based on it,
  507: under Section 2) in object code or executable form under the terms of
  508: Sections 1 and 2 above provided that you also do one of the following:
  509: 
  510: @enumerate a
  511: @item
  512: Accompany it with the complete corresponding machine-readable
  513: source code, which must be distributed under the terms of Sections
  514: 1 and 2 above on a medium customarily used for software interchange; or,
  515: 
  516: @item
  517: Accompany it with a written offer, valid for at least three
  518: years, to give any third party, for a charge no more than your
  519: cost of physically performing source distribution, a complete
  520: machine-readable copy of the corresponding source code, to be
  521: distributed under the terms of Sections 1 and 2 above on a medium
  522: customarily used for software interchange; or,
  523: 
  524: @item
  525: Accompany it with the information you received as to the offer
  526: to distribute corresponding source code.  (This alternative is
  527: allowed only for noncommercial distribution and only if you
  528: received the program in object code or executable form with such
  529: an offer, in accord with Subsection b above.)
  530: @end enumerate
  531: 
  532: The source code for a work means the preferred form of the work for
  533: making modifications to it.  For an executable work, complete source
  534: code means all the source code for all modules it contains, plus any
  535: associated interface definition files, plus the scripts used to
  536: control compilation and installation of the executable.  However, as a
  537: special exception, the source code distributed need not include
  538: anything that is normally distributed (in either source or binary
  539: form) with the major components (compiler, kernel, and so on) of the
  540: operating system on which the executable runs, unless that component
  541: itself accompanies the executable.
  542: 
  543: If distribution of executable or object code is made by offering
  544: access to copy from a designated place, then offering equivalent
  545: access to copy the source code from the same place counts as
  546: distribution of the source code, even though third parties are not
  547: compelled to copy the source along with the object code.
  548: 
  549: @item
  550: You may not copy, modify, sublicense, or distribute the Program
  551: except as expressly provided under this License.  Any attempt
  552: otherwise to copy, modify, sublicense or distribute the Program is
  553: void, and will automatically terminate your rights under this License.
  554: However, parties who have received copies, or rights, from you under
  555: this License will not have their licenses terminated so long as such
  556: parties remain in full compliance.
  557: 
  558: @item
  559: You are not required to accept this License, since you have not
  560: signed it.  However, nothing else grants you permission to modify or
  561: distribute the Program or its derivative works.  These actions are
  562: prohibited by law if you do not accept this License.  Therefore, by
  563: modifying or distributing the Program (or any work based on the
  564: Program), you indicate your acceptance of this License to do so, and
  565: all its terms and conditions for copying, distributing or modifying
  566: the Program or works based on it.
  567: 
  568: @item
  569: Each time you redistribute the Program (or any work based on the
  570: Program), the recipient automatically receives a license from the
  571: original licensor to copy, distribute or modify the Program subject to
  572: these terms and conditions.  You may not impose any further
  573: restrictions on the recipients' exercise of the rights granted herein.
  574: You are not responsible for enforcing compliance by third parties to
  575: this License.
  576: 
  577: @item
  578: If, as a consequence of a court judgment or allegation of patent
  579: infringement or for any other reason (not limited to patent issues),
  580: conditions are imposed on you (whether by court order, agreement or
  581: otherwise) that contradict the conditions of this License, they do not
  582: excuse you from the conditions of this License.  If you cannot
  583: distribute so as to satisfy simultaneously your obligations under this
  584: License and any other pertinent obligations, then as a consequence you
  585: may not distribute the Program at all.  For example, if a patent
  586: license would not permit royalty-free redistribution of the Program by
  587: all those who receive copies directly or indirectly through you, then
  588: the only way you could satisfy both it and this License would be to
  589: refrain entirely from distribution of the Program.
  590: 
  591: If any portion of this section is held invalid or unenforceable under
  592: any particular circumstance, the balance of the section is intended to
  593: apply and the section as a whole is intended to apply in other
  594: circumstances.
  595: 
  596: It is not the purpose of this section to induce you to infringe any
  597: patents or other property right claims or to contest validity of any
  598: such claims; this section has the sole purpose of protecting the
  599: integrity of the free software distribution system, which is
  600: implemented by public license practices.  Many people have made
  601: generous contributions to the wide range of software distributed
  602: through that system in reliance on consistent application of that
  603: system; it is up to the author/donor to decide if he or she is willing
  604: to distribute software through any other system and a licensee cannot
  605: impose that choice.
  606: 
  607: This section is intended to make thoroughly clear what is believed to
  608: be a consequence of the rest of this License.
  609: 
  610: @item
  611: If the distribution and/or use of the Program is restricted in
  612: certain countries either by patents or by copyrighted interfaces, the
  613: original copyright holder who places the Program under this License
  614: may add an explicit geographical distribution limitation excluding
  615: those countries, so that distribution is permitted only in or among
  616: countries not thus excluded.  In such case, this License incorporates
  617: the limitation as if written in the body of this License.
  618: 
  619: @item
  620: The Free Software Foundation may publish revised and/or new versions
  621: of the General Public License from time to time.  Such new versions will
  622: be similar in spirit to the present version, but may differ in detail to
  623: address new problems or concerns.
  624: 
  625: Each version is given a distinguishing version number.  If the Program
  626: specifies a version number of this License which applies to it and ``any
  627: later version'', you have the option of following the terms and conditions
  628: either of that version or of any later version published by the Free
  629: Software Foundation.  If the Program does not specify a version number of
  630: this License, you may choose any version ever published by the Free Software
  631: Foundation.
  632: 
  633: @item
  634: If you wish to incorporate parts of the Program into other free
  635: programs whose distribution conditions are different, write to the author
  636: to ask for permission.  For software which is copyrighted by the Free
  637: Software Foundation, write to the Free Software Foundation; we sometimes
  638: make exceptions for this.  Our decision will be guided by the two goals
  639: of preserving the free status of all derivatives of our free software and
  640: of promoting the sharing and reuse of software generally.
  641: 
  642: @iftex
  643: @heading NO WARRANTY
  644: @end iftex
  645: @ifinfo
  646: @center NO WARRANTY
  647: @end ifinfo
  648: 
  649: @item
  650: BECAUSE THE PROGRAM IS LICENSED FREE OF CHARGE, THERE IS NO WARRANTY
  651: FOR THE PROGRAM, TO THE EXTENT PERMITTED BY APPLICABLE LAW.  EXCEPT WHEN
  652: OTHERWISE STATED IN WRITING THE COPYRIGHT HOLDERS AND/OR OTHER PARTIES
  653: PROVIDE THE PROGRAM ``AS IS'' WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED
  654: OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
  655: MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE.  THE ENTIRE RISK AS
  656: TO THE QUALITY AND PERFORMANCE OF THE PROGRAM IS WITH YOU.  SHOULD THE
  657: PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF ALL NECESSARY SERVICING,
  658: REPAIR OR CORRECTION.
  659: 
  660: @item
  661: IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING
  662: WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MAY MODIFY AND/OR
  663: REDISTRIBUTE THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES,
  664: INCLUDING ANY GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING
  665: OUT OF THE USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED
  666: TO LOSS OF DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY
  667: YOU OR THIRD PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER
  668: PROGRAMS), EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE
  669: POSSIBILITY OF SUCH DAMAGES.
  670: @end enumerate
  671: 
  672: @iftex
  673: @heading END OF TERMS AND CONDITIONS
  674: @end iftex
  675: @ifinfo
  676: @center END OF TERMS AND CONDITIONS
  677: @end ifinfo
  678: 
  679: @page
  680: @unnumberedsec How to Apply These Terms to Your New Programs
  681: 
  682:   If you develop a new program, and you want it to be of the greatest
  683: possible use to the public, the best way to achieve this is to make it
  684: free software which everyone can redistribute and change under these terms.
  685: 
  686:   To do so, attach the following notices to the program.  It is safest
  687: to attach them to the start of each source file to most effectively
  688: convey the exclusion of warranty; and each file should have at least
  689: the ``copyright'' line and a pointer to where the full notice is found.
  690: 
  691: @smallexample
  692: @var{one line to give the program's name and a brief idea of what it does.}
  693: Copyright (C) 19@var{yy}  @var{name of author}
  694: 
  695: This program is free software; you can redistribute it and/or modify 
  696: it under the terms of the GNU General Public License as published by 
  697: the Free Software Foundation; either version 2 of the License, or 
  698: (at your option) any later version.
  699: 
  700: This program is distributed in the hope that it will be useful,
  701: but WITHOUT ANY WARRANTY; without even the implied warranty of
  702: MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  703: GNU General Public License for more details.
  704: 
  705: You should have received a copy of the GNU General Public License
  706: along with this program; if not, write to the Free Software
  707: Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
  708: @end smallexample
  709: 
  710: Also add information on how to contact you by electronic and paper mail.
  711: 
  712: If the program is interactive, make it output a short notice like this
  713: when it starts in an interactive mode:
  714: 
  715: @smallexample
  716: Gnomovision version 69, Copyright (C) 19@var{yy} @var{name of author}
  717: Gnomovision comes with ABSOLUTELY NO WARRANTY; for details
  718: type `show w'.  
  719: This is free software, and you are welcome to redistribute it 
  720: under certain conditions; type `show c' for details.
  721: @end smallexample
  722: 
  723: The hypothetical commands @samp{show w} and @samp{show c} should show
  724: the appropriate parts of the General Public License.  Of course, the
  725: commands you use may be called something other than @samp{show w} and
  726: @samp{show c}; they could even be mouse-clicks or menu items---whatever
  727: suits your program.
  728: 
  729: You should also get your employer (if you work as a programmer) or your
  730: school, if any, to sign a ``copyright disclaimer'' for the program, if
  731: necessary.  Here is a sample; alter the names:
  732: 
  733: @smallexample
  734: Yoyodyne, Inc., hereby disclaims all copyright interest in the program
  735: `Gnomovision' (which makes passes at compilers) written by James Hacker.
  736: 
  737: @var{signature of Ty Coon}, 1 April 1989
  738: Ty Coon, President of Vice
  739: @end smallexample
  740: 
  741: This General Public License does not permit incorporating your program into
  742: proprietary programs.  If your program is a subroutine library, you may
  743: consider it more useful to permit linking proprietary applications with the
  744: library.  If this is what you want to do, use the GNU Library General
  745: Public License instead of this License.
  746: 
  747: @iftex
  748: @unnumbered Preface
  749: @cindex Preface
  750: This manual documents Gforth. The reader is expected to know
  751: Forth. This manual is primarily a reference manual. @xref{Other Books}
  752: for introductory material.
  753: @end iftex
  754: 
  755: @node    Goals, Other Books, License, Top
  756: @comment node-name,     next,           previous, up
  757: @chapter Goals of Gforth
  758: @cindex Goals
  759: The goal of the Gforth Project is to develop a standard model for
  760: ANS Forth. This can be split into several subgoals:
  761: 
  762: @itemize @bullet
  763: @item
  764: Gforth should conform to the Forth standard (ANS Forth).
  765: @item
  766: It should be a model, i.e. it should define all the
  767: implementation-dependent things.
  768: @item
  769: It should become standard, i.e. widely accepted and used. This goal
  770: is the most difficult one.
  771: @end itemize
  772: 
  773: To achieve these goals Gforth should be
  774: @itemize @bullet
  775: @item
  776: Similar to previous models (fig-Forth, F83)
  777: @item
  778: Powerful. It should provide for all the things that are considered
  779: necessary today and even some that are not yet considered necessary.
  780: @item
  781: Efficient. It should not get the reputation of being exceptionally
  782: slow.
  783: @item
  784: Free.
  785: @item
  786: Available on many machines/easy to port.
  787: @end itemize
  788: 
  789: Have we achieved these goals? Gforth conforms to the ANS Forth
  790: standard. It may be considered a model, but we have not yet documented
  791: which parts of the model are stable and which parts we are likely to
  792: change. It certainly has not yet become a de facto standard, but it
  793: appears to be quite popular. It has some similarities to and some
  794: differences from previous models. It has some powerful features, but not
  795: yet everything that we envisioned. We certainly have achieved our
  796: execution speed goals (@pxref{Performance}).  It is free and available
  797: on many machines.
  798: 
  799: @node Other Books, Invoking Gforth, Goals, Top
  800: @chapter Other books on ANS Forth
  801: @cindex books on Forth
  802: 
  803: As the standard is relatively new, there are not many books out yet. It
  804: is not recommended to learn Forth by using Gforth and a book that is not
  805: written for ANS Forth, as you will not know your mistakes from the
  806: deviations of the book. However, books based on the Forth-83 standard
  807: should be ok, because ANS Forth is primarily an extension of Forth-83.
  808: 
  809: @cindex standard document for ANS Forth
  810: @cindex ANS Forth document
  811: There is, of course, the standard, the definite reference if you want to
  812: write ANS Forth programs. It is available in printed form from the
  813: National Standards Institute Sales Department (Tel.: USA (212) 642-4900;
  814: Fax.: USA (212) 302-1286) as document @cite{X3.215-1994} for about $200. You
  815: can also get it from Global Engineering Documents (Tel.: USA (800)
  816: 854-7179; Fax.: (303) 843-9880) for about $300.
  817: 
  818: @cite{dpANS6}, the last draft of the standard, which was then submitted
  819: to ANSI for publication is available electronically and for free in some
  820: MS Word format, and it has been converted to HTML
  821: (@url{http://www.taygeta.com/forth/dpans.html}; this is my favourite
  822: format); this HTML version also includes the answers to Requests for
  823: Interpretation (RFIs). Some pointers to these versions can be found
  824: through @*@url{http://www.complang.tuwien.ac.at/projects/forth.html}.
  825: 
  826: @cindex introductory book
  827: @cindex book, introductory
  828: @cindex Woehr, Jack: @cite{Forth: The New Model}
  829: @cindex @cite{Forth: The new model} (book)
  830: @cite{Forth: The New Model} by Jack Woehr (Prentice-Hall, 1993) is an
  831: introductory book based on a draft version of the standard. It does not
  832: cover the whole standard. It also contains interesting background
  833: information (Jack Woehr was in the ANS Forth Technical Committee). It is
  834: not appropriate for complete newbies, but programmers experienced in
  835: other languages should find it ok.
  836: 
  837: @cindex Conklin, Edward K., and Elizabeth Rather: @cite{Forth Programmer's Handbook}
  838: @cindex Rather, Elizabeth and Edward K. Conklin: @cite{Forth Programmer's Handbook}
  839: @cindex @cite{Forth Programmer's Handbook} (book)
  840: @cite{Forth Programmer's Handbook} by Edward K. Conklin, Elizabeth
  841: D. Rather and the technical staff of Forth, Inc. (Forth, Inc., 1997;
  842: ISBN 0-9662156-0-5) contains little introductory material. The majority
  843: of the book is similar to @ref{Words}, but the book covers most of the
  844: standard words and some non-standard words (whereas this manual is
  845: quite incomplete). In addition, the book contains a chapter on
  846: programming style. The major drawback of this book is that it usually
  847: does not identify what is standard and what is specific to the Forth
  848: system described in the book (probably one of Forth, Inc.'s systems).
  849: Fortunately, many of the non-standard programming practices described in
  850: the book work in Gforth, too.  Still, this drawback makes the book
  851: hardly more useful than a pre-ANS book.
  852: 
  853: @node Invoking Gforth, Words, Other Books, Top
  854: @chapter Invoking Gforth
  855: @cindex invoking Gforth
  856: @cindex running Gforth
  857: @cindex command-line options
  858: @cindex options on the command line
  859: @cindex flags on the command line
  860: 
  861: You will usually just say @code{gforth}. In many other cases the default
  862: Gforth image will be invoked like this:
  863: @example
  864: gforth [files] [-e forth-code]
  865: @end example
  866: This interprets the contents of the files and the Forth code in the order they
  867: are given.
  868: 
  869: In general, the command line looks like this:
  870: 
  871: @example
  872: gforth [initialization options] [image-specific options]
  873: @end example
  874: 
  875: The initialization options must come before the rest of the command
  876: line. They are:
  877: 
  878: @table @code
  879: @cindex -i, command-line option
  880: @cindex --image-file, command-line option
  881: @item --image-file @var{file}
  882: @itemx -i @var{file}
  883: Loads the Forth image @var{file} instead of the default
  884: @file{gforth.fi} (@pxref{Image Files}).
  885: 
  886: @cindex --path, command-line option
  887: @cindex -p, command-line option
  888: @item --path @var{path}
  889: @itemx -p @var{path}
  890: Uses @var{path} for searching the image file and Forth source code files
  891: instead of the default in the environment variable @code{GFORTHPATH} or
  892: the path specified at installation time (e.g.,
  893: @file{/usr/local/share/gforth/0.2.0:.}). A path is given as a list of
  894: directories, separated by @samp{:} (on Unix) or @samp{;} (on other OSs).
  895: 
  896: @cindex --dictionary-size, command-line option
  897: @cindex -m, command-line option
  898: @cindex @var{size} parameters for command-line options
  899: @cindex size of the dictionary and the stacks
  900: @item --dictionary-size @var{size}
  901: @itemx -m @var{size}
  902: Allocate @var{size} space for the Forth dictionary space instead of
  903: using the default specified in the image (typically 256K). The
  904: @var{size} specification consists of an integer and a unit (e.g.,
  905: @code{4M}). The unit can be one of @code{b} (bytes), @code{e} (element
  906: size, in this case Cells), @code{k} (kilobytes), @code{M} (Megabytes),
  907: @code{G} (Gigabytes), and @code{T} (Terabytes). If no unit is specified,
  908: @code{e} is used.
  909: 
  910: @cindex --data-stack-size, command-line option
  911: @cindex -d, command-line option
  912: @item --data-stack-size @var{size}
  913: @itemx -d @var{size}
  914: Allocate @var{size} space for the data stack instead of using the
  915: default specified in the image (typically 16K).
  916: 
  917: @cindex --return-stack-size, command-line option
  918: @cindex -r, command-line option
  919: @item --return-stack-size @var{size}
  920: @itemx -r @var{size}
  921: Allocate @var{size} space for the return stack instead of using the
  922: default specified in the image (typically 15K).
  923: 
  924: @cindex --fp-stack-size, command-line option
  925: @cindex -f, command-line option
  926: @item --fp-stack-size @var{size}
  927: @itemx -f @var{size}
  928: Allocate @var{size} space for the floating point stack instead of
  929: using the default specified in the image (typically 15.5K). In this case
  930: the unit specifier @code{e} refers to floating point numbers.
  931: 
  932: @cindex --locals-stack-size, command-line option
  933: @cindex -l, command-line option
  934: @item --locals-stack-size @var{size}
  935: @itemx -l @var{size}
  936: Allocate @var{size} space for the locals stack instead of using the
  937: default specified in the image (typically 14.5K).
  938: 
  939: @cindex -h, command-line option
  940: @cindex --help, command-line option
  941: @item --help
  942: @itemx -h
  943: Print a message about the command-line options
  944: 
  945: @cindex -v, command-line option
  946: @cindex --version, command-line option
  947: @item --version
  948: @itemx -v
  949: Print version and exit
  950: 
  951: @cindex --debug, command-line option
  952: @item --debug
  953: Print some information useful for debugging on startup.
  954: 
  955: @cindex --offset-image, command-line option
  956: @item --offset-image
  957: Start the dictionary at a slightly different position than would be used
  958: otherwise (useful for creating data-relocatable images,
  959: @pxref{Data-Relocatable Image Files}).
  960: 
  961: @cindex --no-offset-im, command-line option
  962: @item --no-offset-im
  963: Start the dictionary at the normal position.
  964: 
  965: @cindex --clear-dictionary, command-line option
  966: @item --clear-dictionary
  967: Initialize all bytes in the dictionary to 0 before loading the image
  968: (@pxref{Data-Relocatable Image Files}).
  969: 
  970: @cindex --die-on-signal, command-line-option
  971: @item --die-on-signal
  972: Normally Gforth handles most signals (e.g., the user interrupt SIGINT,
  973: or the segmentation violation SIGSEGV) by translating it into a Forth
  974: @code{THROW}. With this option, Gforth exits if it receives such a
  975: signal. This option is useful when the engine and/or the image might be
  976: severely broken (such that it causes another signal before recovering
  977: from the first); this option avoids endless loops in such cases.
  978: @end table
  979: 
  980: @cindex loading files at startup
  981: @cindex executing code on startup
  982: @cindex batch processing with Gforth
  983: As explained above, the image-specific command-line arguments for the
  984: default image @file{gforth.fi} consist of a sequence of filenames and
  985: @code{-e @var{forth-code}} options that are interpreted in the sequence
  986: in which they are given. The @code{-e @var{forth-code}} or
  987: @code{--evaluate @var{forth-code}} option evaluates the forth
  988: code. This option takes only one argument; if you want to evaluate more
  989: Forth words, you have to quote them or use several @code{-e}s. To exit
  990: after processing the command line (instead of entering interactive mode)
  991: append @code{-e bye} to the command line.
  992: 
  993: @cindex versions, invoking other versions of Gforth
  994: If you have several versions of Gforth installed, @code{gforth} will
  995: invoke the version that was installed last. @code{gforth-@var{version}}
  996: invokes a specific version. You may want to use the option
  997: @code{--path}, if your environment contains the variable
  998: @code{GFORTHPATH}.
  999: 
 1000: Not yet implemented:
 1001: On startup the system first executes the system initialization file
 1002: (unless the option @code{--no-init-file} is given; note that the system
 1003: resulting from using this option may not be ANS Forth conformant). Then
 1004: the user initialization file @file{.gforth.fs} is executed, unless the
 1005: option @code{--no-rc} is given; this file is first searched in @file{.},
 1006: then in @file{~}, then in the normal path (see above).
 1007: 
 1008: @node Words, Tools, Invoking Gforth, Top
 1009: @chapter Forth Words
 1010: @cindex Words
 1011: 
 1012: @menu
 1013: * Notation::                    
 1014: * Arithmetic::                  
 1015: * Stack Manipulation::          
 1016: * Memory::                      
 1017: * Control Structures::          
 1018: * Locals::                      
 1019: * Defining Words::              
 1020: * Structures::                  
 1021: * Object-oriented Forth::       
 1022: * Tokens for Words::            
 1023: * Wordlists::                   
 1024: * Files::                       
 1025: * Including Files::             
 1026: * Blocks::                      
 1027: * Other I/O::                   
 1028: * Programming Tools::           
 1029: * Assembler and Code Words::    
 1030: * Threading Words::             
 1031: @end menu
 1032: 
 1033: @node Notation, Arithmetic, Words, Words
 1034: @section Notation
 1035: @cindex notation of glossary entries
 1036: @cindex format of glossary entries
 1037: @cindex glossary notation format
 1038: @cindex word glossary entry format
 1039: 
 1040: The Forth words are described in this section in the glossary notation
 1041: that has become a de-facto standard for Forth texts, i.e.,
 1042: 
 1043: @format
 1044: @var{word}     @var{Stack effect}   @var{wordset}   @var{pronunciation}
 1045: @end format
 1046: @var{Description}
 1047: 
 1048: @table @var
 1049: @item word
 1050: @cindex case insensitivity
 1051: The name of the word. BTW, Gforth is case insensitive, so you can
 1052: type the words in in lower case (However, @pxref{core-idef}).
 1053: 
 1054: @item Stack effect
 1055: @cindex stack effect
 1056: The stack effect is written in the notation @code{@var{before} --
 1057: @var{after}}, where @var{before} and @var{after} describe the top of
 1058: stack entries before and after the execution of the word. The rest of
 1059: the stack is not touched by the word. The top of stack is rightmost,
 1060: i.e., a stack sequence is written as it is typed in. Note that Gforth
 1061: uses a separate floating point stack, but a unified stack
 1062: notation. Also, return stack effects are not shown in @var{stack
 1063: effect}, but in @var{Description}. The name of a stack item describes
 1064: the type and/or the function of the item. See below for a discussion of
 1065: the types.
 1066: 
 1067: All words have two stack effects: A compile-time stack effect and a
 1068: run-time stack effect. The compile-time stack-effect of most words is
 1069: @var{ -- }. If the compile-time stack-effect of a word deviates from
 1070: this standard behaviour, or the word does other unusual things at
 1071: compile time, both stack effects are shown; otherwise only the run-time
 1072: stack effect is shown.
 1073: 
 1074: @cindex pronounciation of words
 1075: @item pronunciation
 1076: How the word is pronounced.
 1077: 
 1078: @cindex wordset
 1079: @item wordset
 1080: The ANS Forth standard is divided into several wordsets. A standard
 1081: system need not support all of them. So, the fewer wordsets your program
 1082: uses the more portable it will be in theory. However, we suspect that
 1083: most ANS Forth systems on personal machines will feature all
 1084: wordsets. Words that are not defined in the ANS standard have
 1085: @code{gforth} or @code{gforth-internal} as wordset. @code{gforth}
 1086: describes words that will work in future releases of Gforth;
 1087: @code{gforth-internal} words are more volatile. Environmental query
 1088: strings are also displayed like words; you can recognize them by the
 1089: @code{environment} in the wordset field.
 1090: 
 1091: @item Description
 1092: A description of the behaviour of the word.
 1093: @end table
 1094: 
 1095: @cindex types of stack items
 1096: @cindex stack item types
 1097: The type of a stack item is specified by the character(s) the name
 1098: starts with:
 1099: 
 1100: @table @code
 1101: @item f
 1102: @cindex @code{f}, stack item type
 1103: Boolean flags, i.e. @code{false} or @code{true}.
 1104: @item c
 1105: @cindex @code{c}, stack item type
 1106: Char
 1107: @item w
 1108: @cindex @code{w}, stack item type
 1109: Cell, can contain an integer or an address
 1110: @item n
 1111: @cindex @code{n}, stack item type
 1112: signed integer
 1113: @item u
 1114: @cindex @code{u}, stack item type
 1115: unsigned integer
 1116: @item d
 1117: @cindex @code{d}, stack item type
 1118: double sized signed integer
 1119: @item ud
 1120: @cindex @code{ud}, stack item type
 1121: double sized unsigned integer
 1122: @item r
 1123: @cindex @code{r}, stack item type
 1124: Float (on the FP stack)
 1125: @item a_
 1126: @cindex @code{a_}, stack item type
 1127: Cell-aligned address
 1128: @item c_
 1129: @cindex @code{c_}, stack item type
 1130: Char-aligned address (note that a Char may have two bytes in Windows NT)
 1131: @item f_
 1132: @cindex @code{f_}, stack item type
 1133: Float-aligned address
 1134: @item df_
 1135: @cindex @code{df_}, stack item type
 1136: Address aligned for IEEE double precision float
 1137: @item sf_
 1138: @cindex @code{sf_}, stack item type
 1139: Address aligned for IEEE single precision float
 1140: @item xt
 1141: @cindex @code{xt}, stack item type
 1142: Execution token, same size as Cell
 1143: @item wid
 1144: @cindex @code{wid}, stack item type
 1145: Wordlist ID, same size as Cell
 1146: @item f83name
 1147: @cindex @code{f83name}, stack item type
 1148: Pointer to a name structure
 1149: @item "
 1150: @cindex @code{"}, stack item type
 1151: string in the input stream (not on the stack). The terminating character
 1152: is a blank by default. If it is not a blank, it is shown in @code{<>}
 1153: quotes.
 1154: @end table
 1155: 
 1156: @node Arithmetic, Stack Manipulation, Notation, Words
 1157: @section Arithmetic
 1158: @cindex arithmetic words
 1159: 
 1160: @cindex division with potentially negative operands
 1161: Forth arithmetic is not checked, i.e., you will not hear about integer
 1162: overflow on addition or multiplication, you may hear about division by
 1163: zero if you are lucky. The operator is written after the operands, but
 1164: the operands are still in the original order. I.e., the infix @code{2-1}
 1165: corresponds to @code{2 1 -}. Forth offers a variety of division
 1166: operators. If you perform division with potentially negative operands,
 1167: you do not want to use @code{/} or @code{/mod} with its undefined
 1168: behaviour, but rather @code{fm/mod} or @code{sm/mod} (probably the
 1169: former, @pxref{Mixed precision}).
 1170: 
 1171: @menu
 1172: * Single precision::            
 1173: * Bitwise operations::          
 1174: * Mixed precision::             operations with single and double-cell integers
 1175: * Double precision::            Double-cell integer arithmetic
 1176: * Floating Point::              
 1177: @end menu
 1178: 
 1179: @node Single precision, Bitwise operations, Arithmetic, Arithmetic
 1180: @subsection Single precision
 1181: @cindex single precision arithmetic words
 1182: 
 1183: doc-+
 1184: doc--
 1185: doc-*
 1186: doc-/
 1187: doc-mod
 1188: doc-/mod
 1189: doc-negate
 1190: doc-abs
 1191: doc-min
 1192: doc-max
 1193: 
 1194: @node Bitwise operations, Mixed precision, Single precision, Arithmetic
 1195: @subsection Bitwise operations
 1196: @cindex bitwise operation words
 1197: 
 1198: doc-and
 1199: doc-or
 1200: doc-xor
 1201: doc-invert
 1202: doc-2*
 1203: doc-2/
 1204: 
 1205: @node Mixed precision, Double precision, Bitwise operations, Arithmetic
 1206: @subsection Mixed precision
 1207: @cindex mixed precision arithmetic words
 1208: 
 1209: doc-m+
 1210: doc-*/
 1211: doc-*/mod
 1212: doc-m*
 1213: doc-um*
 1214: doc-m*/
 1215: doc-um/mod
 1216: doc-fm/mod
 1217: doc-sm/rem
 1218: 
 1219: @node Double precision, Floating Point, Mixed precision, Arithmetic
 1220: @subsection Double precision
 1221: @cindex double precision arithmetic words
 1222: 
 1223: @cindex double-cell numbers, input format
 1224: @cindex input format for double-cell numbers
 1225: The outer (aka text) interpreter converts numbers containing a dot into
 1226: a double precision number. Note that only numbers with the dot as last
 1227: character are standard-conforming.
 1228: 
 1229: doc-d+
 1230: doc-d-
 1231: doc-dnegate
 1232: doc-dabs
 1233: doc-dmin
 1234: doc-dmax
 1235: 
 1236: @node Floating Point,  , Double precision, Arithmetic
 1237: @subsection Floating Point
 1238: @cindex floating point arithmetic words
 1239: 
 1240: @cindex floating-point numbers, input format
 1241: @cindex input format for floating-point numbers
 1242: The format of floating point numbers recognized by the outer (aka text)
 1243: interpreter is: a signed decimal number, possibly containing a decimal
 1244: point (@code{.}), followed by @code{E} or @code{e}, optionally followed
 1245: by a signed integer (the exponent). E.g., @code{1e} is the same as
 1246: @code{+1.0e+0}. Note that a number without @code{e} is not interpreted
 1247: as floating-point number, but as double (if the number contains a
 1248: @code{.}) or single precision integer. Also, conversions between string
 1249: and floating point numbers always use base 10, irrespective of the value
 1250: of @code{BASE} (in Gforth; for the standard this is an ambiguous
 1251: condition). If @code{BASE} contains a value greater then 14, the
 1252: @code{E} may be interpreted as digit and the number will be interpreted
 1253: as integer, unless it has a signed exponent (both @code{+} and @code{-}
 1254: are allowed as signs).
 1255: 
 1256: @cindex angles in trigonometric operations
 1257: @cindex trigonometric operations
 1258: Angles in floating point operations are given in radians (a full circle
 1259: has 2 pi radians). Note, that Gforth has a separate floating point
 1260: stack, but we use the unified notation.
 1261: 
 1262: @cindex floating-point arithmetic, pitfalls
 1263: Floating point numbers have a number of unpleasant surprises for the
 1264: unwary (e.g., floating point addition is not associative) and even a few
 1265: for the wary. You should not use them unless you know what you are doing
 1266: or you don't care that the results you get are totally bogus. If you
 1267: want to learn about the problems of floating point numbers (and how to
 1268: avoid them), you might start with @cite{David Goldberg, What Every
 1269: Computer Scientist Should Know About Floating-Point Arithmetic, ACM
 1270: Computing Surveys 23(1):5@minus{}48, March 1991}
 1271: (@url{http://www.validgh.com/goldberg/paper.ps}).
 1272: 
 1273: doc-f+
 1274: doc-f-
 1275: doc-f*
 1276: doc-f/
 1277: doc-fnegate
 1278: doc-fabs
 1279: doc-fmax
 1280: doc-fmin
 1281: doc-floor
 1282: doc-fround
 1283: doc-f**
 1284: doc-fsqrt
 1285: doc-fexp
 1286: doc-fexpm1
 1287: doc-fln
 1288: doc-flnp1
 1289: doc-flog
 1290: doc-falog
 1291: doc-fsin
 1292: doc-fcos
 1293: doc-fsincos
 1294: doc-ftan
 1295: doc-fasin
 1296: doc-facos
 1297: doc-fatan
 1298: doc-fatan2
 1299: doc-fsinh
 1300: doc-fcosh
 1301: doc-ftanh
 1302: doc-fasinh
 1303: doc-facosh
 1304: doc-fatanh
 1305: 
 1306: @node Stack Manipulation, Memory, Arithmetic, Words
 1307: @section Stack Manipulation
 1308: @cindex stack manipulation words
 1309: 
 1310: @cindex floating-point stack in the standard
 1311: Gforth has a data stack (aka parameter stack) for characters, cells,
 1312: addresses, and double cells, a floating point stack for floating point
 1313: numbers, a return stack for storing the return addresses of colon
 1314: definitions and other data, and a locals stack for storing local
 1315: variables. Note that while every sane Forth has a separate floating
 1316: point stack, this is not strictly required; an ANS Forth system could
 1317: theoretically keep floating point numbers on the data stack. As an
 1318: additional difficulty, you don't know how many cells a floating point
 1319: number takes. It is reportedly possible to write words in a way that
 1320: they work also for a unified stack model, but we do not recommend trying
 1321: it. Instead, just say that your program has an environmental dependency
 1322: on a separate FP stack.
 1323: 
 1324: @cindex return stack and locals
 1325: @cindex locals and return stack
 1326: Also, a Forth system is allowed to keep the local variables on the
 1327: return stack. This is reasonable, as local variables usually eliminate
 1328: the need to use the return stack explicitly. So, if you want to produce
 1329: a standard complying program and if you are using local variables in a
 1330: word, forget about return stack manipulations in that word (see the
 1331: standard document for the exact rules).
 1332: 
 1333: @menu
 1334: * Data stack::                  
 1335: * Floating point stack::        
 1336: * Return stack::                
 1337: * Locals stack::                
 1338: * Stack pointer manipulation::  
 1339: @end menu
 1340: 
 1341: @node Data stack, Floating point stack, Stack Manipulation, Stack Manipulation
 1342: @subsection Data stack
 1343: @cindex data stack manipulation words
 1344: @cindex stack manipulations words, data stack
 1345: 
 1346: doc-drop
 1347: doc-nip
 1348: doc-dup
 1349: doc-over
 1350: doc-tuck
 1351: doc-swap
 1352: doc-rot
 1353: doc--rot
 1354: doc-?dup
 1355: doc-pick
 1356: doc-roll
 1357: doc-2drop
 1358: doc-2nip
 1359: doc-2dup
 1360: doc-2over
 1361: doc-2tuck
 1362: doc-2swap
 1363: doc-2rot
 1364: 
 1365: @node Floating point stack, Return stack, Data stack, Stack Manipulation
 1366: @subsection Floating point stack
 1367: @cindex floating-point stack manipulation words
 1368: @cindex stack manipulation words, floating-point stack
 1369: 
 1370: doc-fdrop
 1371: doc-fnip
 1372: doc-fdup
 1373: doc-fover
 1374: doc-ftuck
 1375: doc-fswap
 1376: doc-frot
 1377: 
 1378: @node Return stack, Locals stack, Floating point stack, Stack Manipulation
 1379: @subsection Return stack
 1380: @cindex return stack manipulation words
 1381: @cindex stack manipulation words, return stack
 1382: 
 1383: doc->r
 1384: doc-r>
 1385: doc-r@
 1386: doc-rdrop
 1387: doc-2>r
 1388: doc-2r>
 1389: doc-2r@
 1390: doc-2rdrop
 1391: 
 1392: @node Locals stack, Stack pointer manipulation, Return stack, Stack Manipulation
 1393: @subsection Locals stack
 1394: 
 1395: @node Stack pointer manipulation,  , Locals stack, Stack Manipulation
 1396: @subsection Stack pointer manipulation
 1397: @cindex stack pointer manipulation words
 1398: 
 1399: doc-sp@
 1400: doc-sp!
 1401: doc-fp@
 1402: doc-fp!
 1403: doc-rp@
 1404: doc-rp!
 1405: doc-lp@
 1406: doc-lp!
 1407: 
 1408: @node Memory, Control Structures, Stack Manipulation, Words
 1409: @section Memory
 1410: @cindex Memory words
 1411: 
 1412: @menu
 1413: * Memory Access::      
 1414: * Address arithmetic::          
 1415: * Memory Blocks::         
 1416: @end menu
 1417: 
 1418: @node Memory Access, Address arithmetic, Memory, Memory
 1419: @subsection Memory Access
 1420: @cindex memory access words
 1421: 
 1422: doc-@
 1423: doc-!
 1424: doc-+!
 1425: doc-c@
 1426: doc-c!
 1427: doc-2@
 1428: doc-2!
 1429: doc-f@
 1430: doc-f!
 1431: doc-sf@
 1432: doc-sf!
 1433: doc-df@
 1434: doc-df!
 1435: 
 1436: @node Address arithmetic, Memory Blocks, Memory Access, Memory
 1437: @subsection Address arithmetic
 1438: @cindex address arithmetic words
 1439: 
 1440: ANS Forth does not specify the sizes of the data types. Instead, it
 1441: offers a number of words for computing sizes and doing address
 1442: arithmetic. Basically, address arithmetic is performed in terms of
 1443: address units (aus); on most systems the address unit is one byte. Note
 1444: that a character may have more than one au, so @code{chars} is no noop
 1445: (on systems where it is a noop, it compiles to nothing).
 1446: 
 1447: @cindex alignment of addresses for types
 1448: ANS Forth also defines words for aligning addresses for specific
 1449: types. Many computers require that accesses to specific data types
 1450: must only occur at specific addresses; e.g., that cells may only be
 1451: accessed at addresses divisible by 4. Even if a machine allows unaligned
 1452: accesses, it can usually perform aligned accesses faster. 
 1453: 
 1454: For the performance-conscious: alignment operations are usually only
 1455: necessary during the definition of a data structure, not during the
 1456: (more frequent) accesses to it.
 1457: 
 1458: ANS Forth defines no words for character-aligning addresses. This is not
 1459: an oversight, but reflects the fact that addresses that are not
 1460: char-aligned have no use in the standard and therefore will not be
 1461: created.
 1462: 
 1463: @cindex @code{CREATE} and alignment
 1464: The standard guarantees that addresses returned by @code{CREATE}d words
 1465: are cell-aligned; in addition, Gforth guarantees that these addresses
 1466: are aligned for all purposes.
 1467: 
 1468: Note that the standard defines a word @code{char}, which has nothing to
 1469: do with address arithmetic.
 1470: 
 1471: doc-chars
 1472: doc-char+
 1473: doc-cells
 1474: doc-cell+
 1475: doc-cell
 1476: doc-align
 1477: doc-aligned
 1478: doc-floats
 1479: doc-float+
 1480: doc-float
 1481: doc-falign
 1482: doc-faligned
 1483: doc-sfloats
 1484: doc-sfloat+
 1485: doc-sfalign
 1486: doc-sfaligned
 1487: doc-dfloats
 1488: doc-dfloat+
 1489: doc-dfalign
 1490: doc-dfaligned
 1491: doc-maxalign
 1492: doc-maxaligned
 1493: doc-cfalign
 1494: doc-cfaligned
 1495: doc-address-unit-bits
 1496: 
 1497: @node Memory Blocks,  , Address arithmetic, Memory
 1498: @subsection Memory Blocks
 1499: @cindex memory block words
 1500: 
 1501: doc-move
 1502: doc-erase
 1503: 
 1504: While the previous words work on address units, the rest works on
 1505: characters.
 1506: 
 1507: doc-cmove
 1508: doc-cmove>
 1509: doc-fill
 1510: doc-blank
 1511: 
 1512: @node Control Structures, Locals, Memory, Words
 1513: @section Control Structures
 1514: @cindex control structures
 1515: 
 1516: Control structures in Forth cannot be used in interpret state, only in
 1517: compile state@footnote{More precisely, they have no interpretation
 1518: semantics (@pxref{Interpretation and Compilation Semantics})}, i.e., in
 1519: a colon definition. We do not like this limitation, but have not seen a
 1520: satisfying way around it yet, although many schemes have been proposed.
 1521: 
 1522: @menu
 1523: * Selection::                   
 1524: * Simple Loops::                
 1525: * Counted Loops::               
 1526: * Arbitrary control structures::  
 1527: * Calls and returns::           
 1528: * Exception Handling::          
 1529: @end menu
 1530: 
 1531: @node Selection, Simple Loops, Control Structures, Control Structures
 1532: @subsection Selection
 1533: @cindex selection control structures
 1534: @cindex control structures for selection
 1535: 
 1536: @cindex @code{IF} control structure
 1537: @example
 1538: @var{flag}
 1539: IF
 1540:   @var{code}
 1541: ENDIF
 1542: @end example
 1543: or
 1544: @example
 1545: @var{flag}
 1546: IF
 1547:   @var{code1}
 1548: ELSE
 1549:   @var{code2}
 1550: ENDIF
 1551: @end example
 1552: 
 1553: You can use @code{THEN} instead of @code{ENDIF}. Indeed, @code{THEN} is
 1554: standard, and @code{ENDIF} is not, although it is quite popular. We
 1555: recommend using @code{ENDIF}, because it is less confusing for people
 1556: who also know other languages (and is not prone to reinforcing negative
 1557: prejudices against Forth in these people). Adding @code{ENDIF} to a
 1558: system that only supplies @code{THEN} is simple:
 1559: @example
 1560: : endif   POSTPONE then ; immediate
 1561: @end example
 1562: 
 1563: [According to @cite{Webster's New Encyclopedic Dictionary}, @dfn{then
 1564: (adv.)}  has the following meanings:
 1565: @quotation
 1566: ... 2b: following next after in order ... 3d: as a necessary consequence
 1567: (if you were there, then you saw them).
 1568: @end quotation
 1569: Forth's @code{THEN} has the meaning 2b, whereas @code{THEN} in Pascal
 1570: and many other programming languages has the meaning 3d.]
 1571: 
 1572: Gforth also provides the words @code{?dup-if} and @code{?dup-0=-if}, so
 1573: you can avoid using @code{?dup}. Using these alternatives is also more
 1574: efficient than using @code{?dup}. Definitions in plain standard Forth
 1575: for @code{ENDIF}, @code{?DUP-IF} and @code{?DUP-0=-IF} are provided in
 1576: @file{compat/control.fs}.
 1577: 
 1578: @cindex @code{CASE} control structure
 1579: @example
 1580: @var{n}
 1581: CASE
 1582:   @var{n1} OF @var{code1} ENDOF
 1583:   @var{n2} OF @var{code2} ENDOF
 1584:   @dots{}
 1585: ENDCASE
 1586: @end example
 1587: 
 1588: Executes the first @var{codei}, where the @var{ni} is equal to
 1589: @var{n}. A default case can be added by simply writing the code after
 1590: the last @code{ENDOF}. It may use @var{n}, which is on top of the stack,
 1591: but must not consume it.
 1592: 
 1593: @node Simple Loops, Counted Loops, Selection, Control Structures
 1594: @subsection Simple Loops
 1595: @cindex simple loops
 1596: @cindex loops without count 
 1597: 
 1598: @cindex @code{WHILE} loop
 1599: @example
 1600: BEGIN
 1601:   @var{code1}
 1602:   @var{flag}
 1603: WHILE
 1604:   @var{code2}
 1605: REPEAT
 1606: @end example
 1607: 
 1608: @var{code1} is executed and @var{flag} is computed. If it is true,
 1609: @var{code2} is executed and the loop is restarted; If @var{flag} is
 1610: false, execution continues after the @code{REPEAT}.
 1611: 
 1612: @cindex @code{UNTIL} loop
 1613: @example
 1614: BEGIN
 1615:   @var{code}
 1616:   @var{flag}
 1617: UNTIL
 1618: @end example
 1619: 
 1620: @var{code} is executed. The loop is restarted if @code{flag} is false.
 1621: 
 1622: @cindex endless loop
 1623: @cindex loops, endless
 1624: @example
 1625: BEGIN
 1626:   @var{code}
 1627: AGAIN
 1628: @end example
 1629: 
 1630: This is an endless loop.
 1631: 
 1632: @node Counted Loops, Arbitrary control structures, Simple Loops, Control Structures
 1633: @subsection Counted Loops
 1634: @cindex counted loops
 1635: @cindex loops, counted
 1636: @cindex @code{DO} loops
 1637: 
 1638: The basic counted loop is:
 1639: @example
 1640: @var{limit} @var{start}
 1641: ?DO
 1642:   @var{body}
 1643: LOOP
 1644: @end example
 1645: 
 1646: This performs one iteration for every integer, starting from @var{start}
 1647: and up to, but excluding @var{limit}. The counter, aka index, can be
 1648: accessed with @code{i}. E.g., the loop
 1649: @example
 1650: 10 0 ?DO
 1651:   i .
 1652: LOOP
 1653: @end example
 1654: prints
 1655: @example
 1656: 0 1 2 3 4 5 6 7 8 9
 1657: @end example
 1658: The index of the innermost loop can be accessed with @code{i}, the index
 1659: of the next loop with @code{j}, and the index of the third loop with
 1660: @code{k}.
 1661: 
 1662: doc-i
 1663: doc-j
 1664: doc-k
 1665: 
 1666: The loop control data are kept on the return stack, so there are some
 1667: restrictions on mixing return stack accesses and counted loop
 1668: words. E.g., if you put values on the return stack outside the loop, you
 1669: cannot read them inside the loop. If you put values on the return stack
 1670: within a loop, you have to remove them before the end of the loop and
 1671: before accessing the index of the loop.
 1672: 
 1673: There are several variations on the counted loop:
 1674: 
 1675: @code{LEAVE} leaves the innermost counted loop immediately.
 1676: 
 1677: If @var{start} is greater than @var{limit}, a @code{?DO} loop is entered
 1678: (and @code{LOOP} iterates until they become equal by wrap-around
 1679: arithmetic). This behaviour is usually not what you want. Therefore,
 1680: Gforth offers @code{+DO} and @code{U+DO} (as replacements for
 1681: @code{?DO}), which do not enter the loop if @var{start} is greater than
 1682: @var{limit}; @code{+DO} is for signed loop parameters, @code{U+DO} for
 1683: unsigned loop parameters.
 1684: 
 1685: @code{LOOP} can be replaced with @code{@var{n} +LOOP}; this updates the
 1686: index by @var{n} instead of by 1. The loop is terminated when the border
 1687: between @var{limit-1} and @var{limit} is crossed. E.g.:
 1688: 
 1689: @code{4 0 +DO  i .  2 +LOOP}   prints @code{0 2}
 1690: 
 1691: @code{4 1 +DO  i .  2 +LOOP}   prints @code{1 3}
 1692: 
 1693: @cindex negative increment for counted loops
 1694: @cindex counted loops with negative increment
 1695: The behaviour of @code{@var{n} +LOOP} is peculiar when @var{n} is negative:
 1696: 
 1697: @code{-1 0 ?DO  i .  -1 +LOOP}  prints @code{0 -1}
 1698: 
 1699: @code{ 0 0 ?DO  i .  -1 +LOOP}  prints nothing
 1700: 
 1701: Therefore we recommend avoiding @code{@var{n} +LOOP} with negative
 1702: @var{n}. One alternative is @code{@var{u} -LOOP}, which reduces the
 1703: index by @var{u} each iteration. The loop is terminated when the border
 1704: between @var{limit+1} and @var{limit} is crossed. Gforth also provides
 1705: @code{-DO} and @code{U-DO} for down-counting loops. E.g.:
 1706: 
 1707: @code{-2 0 -DO  i .  1 -LOOP}  prints @code{0 -1}
 1708: 
 1709: @code{-1 0 -DO  i .  1 -LOOP}  prints @code{0}
 1710: 
 1711: @code{ 0 0 -DO  i .  1 -LOOP}  prints nothing
 1712: 
 1713: Unfortunately, @code{+DO}, @code{U+DO}, @code{-DO}, @code{U-DO} and
 1714: @code{-LOOP} are not in the ANS Forth standard. However, an
 1715: implementation for these words that uses only standard words is provided
 1716: in @file{compat/loops.fs}.
 1717: 
 1718: @code{?DO} can also be replaced by @code{DO}. @code{DO} always enters
 1719: the loop, independent of the loop parameters. Do not use @code{DO}, even
 1720: if you know that the loop is entered in any case. Such knowledge tends
 1721: to become invalid during maintenance of a program, and then the
 1722: @code{DO} will make trouble.
 1723: 
 1724: @code{UNLOOP} is used to prepare for an abnormal loop exit, e.g., via
 1725: @code{EXIT}. @code{UNLOOP} removes the loop control parameters from the
 1726: return stack so @code{EXIT} can get to its return address.
 1727: 
 1728: @cindex @code{FOR} loops
 1729: Another counted loop is
 1730: @example
 1731: @var{n}
 1732: FOR
 1733:   @var{body}
 1734: NEXT
 1735: @end example
 1736: This is the preferred loop of native code compiler writers who are too
 1737: lazy to optimize @code{?DO} loops properly. In Gforth, this loop
 1738: iterates @var{n+1} times; @code{i} produces values starting with @var{n}
 1739: and ending with 0. Other Forth systems may behave differently, even if
 1740: they support @code{FOR} loops. To avoid problems, don't use @code{FOR}
 1741: loops.
 1742: 
 1743: @node Arbitrary control structures, Calls and returns, Counted Loops, Control Structures
 1744: @subsection Arbitrary control structures
 1745: @cindex control structures, user-defined
 1746: 
 1747: @cindex control-flow stack
 1748: ANS Forth permits and supports using control structures in a non-nested
 1749: way. Information about incomplete control structures is stored on the
 1750: control-flow stack. This stack may be implemented on the Forth data
 1751: stack, and this is what we have done in Gforth.
 1752: 
 1753: @cindex @code{orig}, control-flow stack item
 1754: @cindex @code{dest}, control-flow stack item
 1755: An @i{orig} entry represents an unresolved forward branch, a @i{dest}
 1756: entry represents a backward branch target. A few words are the basis for
 1757: building any control structure possible (except control structures that
 1758: need storage, like calls, coroutines, and backtracking).
 1759: 
 1760: doc-if
 1761: doc-ahead
 1762: doc-then
 1763: doc-begin
 1764: doc-until
 1765: doc-again
 1766: doc-cs-pick
 1767: doc-cs-roll
 1768: 
 1769: On many systems control-flow stack items take one word, in Gforth they
 1770: currently take three (this may change in the future). Therefore it is a
 1771: really good idea to manipulate the control flow stack with
 1772: @code{cs-pick} and @code{cs-roll}, not with data stack manipulation
 1773: words.
 1774: 
 1775: Some standard control structure words are built from these words:
 1776: 
 1777: doc-else
 1778: doc-while
 1779: doc-repeat
 1780: 
 1781: Gforth adds some more control-structure words:
 1782: 
 1783: doc-endif
 1784: doc-?dup-if
 1785: doc-?dup-0=-if
 1786: 
 1787: Counted loop words constitute a separate group of words:
 1788: 
 1789: doc-?do
 1790: doc-+do
 1791: doc-u+do
 1792: doc--do
 1793: doc-u-do
 1794: doc-do
 1795: doc-for
 1796: doc-loop
 1797: doc-+loop
 1798: doc--loop
 1799: doc-next
 1800: doc-leave
 1801: doc-?leave
 1802: doc-unloop
 1803: doc-done
 1804: 
 1805: The standard does not allow using @code{cs-pick} and @code{cs-roll} on
 1806: @i{do-sys}. Our system allows it, but it's your job to ensure that for
 1807: every @code{?DO} etc. there is exactly one @code{UNLOOP} on any path
 1808: through the definition (@code{LOOP} etc. compile an @code{UNLOOP} on the
 1809: fall-through path). Also, you have to ensure that all @code{LEAVE}s are
 1810: resolved (by using one of the loop-ending words or @code{DONE}).
 1811: 
 1812: Another group of control structure words are
 1813: 
 1814: doc-case
 1815: doc-endcase
 1816: doc-of
 1817: doc-endof
 1818: 
 1819: @i{case-sys} and @i{of-sys} cannot be processed using @code{cs-pick} and
 1820: @code{cs-roll}.
 1821: 
 1822: @subsubsection Programming Style
 1823: 
 1824: In order to ensure readability we recommend that you do not create
 1825: arbitrary control structures directly, but define new control structure
 1826: words for the control structure you want and use these words in your
 1827: program.
 1828: 
 1829: E.g., instead of writing
 1830: 
 1831: @example
 1832: begin
 1833:   ...
 1834: if [ 1 cs-roll ]
 1835:   ...
 1836: again then
 1837: @end example
 1838: 
 1839: we recommend defining control structure words, e.g.,
 1840: 
 1841: @example
 1842: : while ( dest -- orig dest )
 1843:  POSTPONE if
 1844:  1 cs-roll ; immediate
 1845: 
 1846: : repeat ( orig dest -- )
 1847:  POSTPONE again
 1848:  POSTPONE then ; immediate
 1849: @end example
 1850: 
 1851: and then using these to create the control structure:
 1852: 
 1853: @example
 1854: begin
 1855:   ...
 1856: while
 1857:   ...
 1858: repeat
 1859: @end example
 1860: 
 1861: That's much easier to read, isn't it? Of course, @code{REPEAT} and
 1862: @code{WHILE} are predefined, so in this example it would not be
 1863: necessary to define them.
 1864: 
 1865: @node Calls and returns, Exception Handling, Arbitrary control structures, Control Structures
 1866: @subsection Calls and returns
 1867: @cindex calling a definition
 1868: @cindex returning from a definition
 1869: 
 1870: @cindex recursive definitions
 1871: A definition can be called simply be writing the name of the definition
 1872: to be called. Note that normally a definition is invisible during its
 1873: definition. If you want to write a directly recursive definition, you
 1874: can use @code{recursive} to make the current definition visible.
 1875: 
 1876: doc-recursive
 1877: 
 1878: Another way to perform a recursive call is
 1879: 
 1880: doc-recurse
 1881: 
 1882: @quotation
 1883: @progstyle
 1884: I prefer using @code{recursive} to @code{recurse}, because calling the
 1885: definition by name is more descriptive (if the name is well-chosen) than
 1886: the somewhat cryptic @code{recurse}.  E.g., in a quicksort
 1887: implementation, it is much better to read (and think) ``now sort the
 1888: partitions'' than to read ``now do a recursive call''.
 1889: @end quotation
 1890: 
 1891: For mutual recursion, use @code{defer}red words, like this:
 1892: 
 1893: @example
 1894: defer foo
 1895: 
 1896: : bar ( ... -- ... )
 1897:  ... foo ... ;
 1898: 
 1899: :noname ( ... -- ... )
 1900:  ... bar ... ;
 1901: IS foo
 1902: @end example
 1903: 
 1904: When the end of the definition is reached, it returns. An earlier return
 1905: can be forced using
 1906: 
 1907: doc-exit
 1908: 
 1909: Don't forget to clean up the return stack and @code{UNLOOP} any
 1910: outstanding @code{?DO}...@code{LOOP}s before @code{EXIT}ing. The
 1911: primitive compiled by @code{EXIT} is
 1912: 
 1913: doc-;s
 1914: 
 1915: @node Exception Handling,  , Calls and returns, Control Structures
 1916: @subsection Exception Handling
 1917: @cindex Exceptions
 1918: 
 1919: doc-catch
 1920: doc-throw
 1921: 
 1922: @node Locals, Defining Words, Control Structures, Words
 1923: @section Locals
 1924: @cindex locals
 1925: 
 1926: Local variables can make Forth programming more enjoyable and Forth
 1927: programs easier to read. Unfortunately, the locals of ANS Forth are
 1928: laden with restrictions. Therefore, we provide not only the ANS Forth
 1929: locals wordset, but also our own, more powerful locals wordset (we
 1930: implemented the ANS Forth locals wordset through our locals wordset).
 1931: 
 1932: The ideas in this section have also been published in the paper
 1933: @cite{Automatic Scoping of Local Variables} by M. Anton Ertl, presented
 1934: at EuroForth '94; it is available at
 1935: @*@url{http://www.complang.tuwien.ac.at/papers/ertl94l.ps.gz}.
 1936: 
 1937: @menu
 1938: * Gforth locals::               
 1939: * ANS Forth locals::            
 1940: @end menu
 1941: 
 1942: @node Gforth locals, ANS Forth locals, Locals, Locals
 1943: @subsection Gforth locals
 1944: @cindex Gforth locals
 1945: @cindex locals, Gforth style
 1946: 
 1947: Locals can be defined with
 1948: 
 1949: @example
 1950: @{ local1 local2 ... -- comment @}
 1951: @end example
 1952: or
 1953: @example
 1954: @{ local1 local2 ... @}
 1955: @end example
 1956: 
 1957: E.g.,
 1958: @example
 1959: : max @{ n1 n2 -- n3 @}
 1960:  n1 n2 > if
 1961:    n1
 1962:  else
 1963:    n2
 1964:  endif ;
 1965: @end example
 1966: 
 1967: The similarity of locals definitions with stack comments is intended. A
 1968: locals definition often replaces the stack comment of a word. The order
 1969: of the locals corresponds to the order in a stack comment and everything
 1970: after the @code{--} is really a comment.
 1971: 
 1972: This similarity has one disadvantage: It is too easy to confuse locals
 1973: declarations with stack comments, causing bugs and making them hard to
 1974: find. However, this problem can be avoided by appropriate coding
 1975: conventions: Do not use both notations in the same program. If you do,
 1976: they should be distinguished using additional means, e.g. by position.
 1977: 
 1978: @cindex types of locals
 1979: @cindex locals types
 1980: The name of the local may be preceded by a type specifier, e.g.,
 1981: @code{F:} for a floating point value:
 1982: 
 1983: @example
 1984: : CX* @{ F: Ar F: Ai F: Br F: Bi -- Cr Ci @}
 1985: \ complex multiplication
 1986:  Ar Br f* Ai Bi f* f-
 1987:  Ar Bi f* Ai Br f* f+ ;
 1988: @end example
 1989: 
 1990: @cindex flavours of locals
 1991: @cindex locals flavours
 1992: @cindex value-flavoured locals
 1993: @cindex variable-flavoured locals
 1994: Gforth currently supports cells (@code{W:}, @code{W^}), doubles
 1995: (@code{D:}, @code{D^}), floats (@code{F:}, @code{F^}) and characters
 1996: (@code{C:}, @code{C^}) in two flavours: a value-flavoured local (defined
 1997: with @code{W:}, @code{D:} etc.) produces its value and can be changed
 1998: with @code{TO}. A variable-flavoured local (defined with @code{W^} etc.)
 1999: produces its address (which becomes invalid when the variable's scope is
 2000: left). E.g., the standard word @code{emit} can be defined in terms of
 2001: @code{type} like this:
 2002: 
 2003: @example
 2004: : emit @{ C^ char* -- @}
 2005:     char* 1 type ;
 2006: @end example
 2007: 
 2008: @cindex default type of locals
 2009: @cindex locals, default type
 2010: A local without type specifier is a @code{W:} local. Both flavours of
 2011: locals are initialized with values from the data or FP stack.
 2012: 
 2013: Currently there is no way to define locals with user-defined data
 2014: structures, but we are working on it.
 2015: 
 2016: Gforth allows defining locals everywhere in a colon definition. This
 2017: poses the following questions:
 2018: 
 2019: @menu
 2020: * Where are locals visible by name?::  
 2021: * How long do locals live?::    
 2022: * Programming Style::           
 2023: * Implementation::              
 2024: @end menu
 2025: 
 2026: @node Where are locals visible by name?, How long do locals live?, Gforth locals, Gforth locals
 2027: @subsubsection Where are locals visible by name?
 2028: @cindex locals visibility
 2029: @cindex visibility of locals
 2030: @cindex scope of locals
 2031: 
 2032: Basically, the answer is that locals are visible where you would expect
 2033: it in block-structured languages, and sometimes a little longer. If you
 2034: want to restrict the scope of a local, enclose its definition in
 2035: @code{SCOPE}...@code{ENDSCOPE}.
 2036: 
 2037: doc-scope
 2038: doc-endscope
 2039: 
 2040: These words behave like control structure words, so you can use them
 2041: with @code{CS-PICK} and @code{CS-ROLL} to restrict the scope in
 2042: arbitrary ways.
 2043: 
 2044: If you want a more exact answer to the visibility question, here's the
 2045: basic principle: A local is visible in all places that can only be
 2046: reached through the definition of the local@footnote{In compiler
 2047: construction terminology, all places dominated by the definition of the
 2048: local.}. In other words, it is not visible in places that can be reached
 2049: without going through the definition of the local. E.g., locals defined
 2050: in @code{IF}...@code{ENDIF} are visible until the @code{ENDIF}, locals
 2051: defined in @code{BEGIN}...@code{UNTIL} are visible after the
 2052: @code{UNTIL} (until, e.g., a subsequent @code{ENDSCOPE}).
 2053: 
 2054: The reasoning behind this solution is: We want to have the locals
 2055: visible as long as it is meaningful. The user can always make the
 2056: visibility shorter by using explicit scoping. In a place that can
 2057: only be reached through the definition of a local, the meaning of a
 2058: local name is clear. In other places it is not: How is the local
 2059: initialized at the control flow path that does not contain the
 2060: definition? Which local is meant, if the same name is defined twice in
 2061: two independent control flow paths?
 2062: 
 2063: This should be enough detail for nearly all users, so you can skip the
 2064: rest of this section. If you really must know all the gory details and
 2065: options, read on.
 2066: 
 2067: In order to implement this rule, the compiler has to know which places
 2068: are unreachable. It knows this automatically after @code{AHEAD},
 2069: @code{AGAIN}, @code{EXIT} and @code{LEAVE}; in other cases (e.g., after
 2070: most @code{THROW}s), you can use the word @code{UNREACHABLE} to tell the
 2071: compiler that the control flow never reaches that place. If
 2072: @code{UNREACHABLE} is not used where it could, the only consequence is
 2073: that the visibility of some locals is more limited than the rule above
 2074: says. If @code{UNREACHABLE} is used where it should not (i.e., if you
 2075: lie to the compiler), buggy code will be produced.
 2076: 
 2077: doc-unreachable
 2078: 
 2079: Another problem with this rule is that at @code{BEGIN}, the compiler
 2080: does not know which locals will be visible on the incoming
 2081: back-edge. All problems discussed in the following are due to this
 2082: ignorance of the compiler (we discuss the problems using @code{BEGIN}
 2083: loops as examples; the discussion also applies to @code{?DO} and other
 2084: loops). Perhaps the most insidious example is:
 2085: @example
 2086: AHEAD
 2087: BEGIN
 2088:   x
 2089: [ 1 CS-ROLL ] THEN
 2090:   @{ x @}
 2091:   ...
 2092: UNTIL
 2093: @end example
 2094: 
 2095: This should be legal according to the visibility rule. The use of
 2096: @code{x} can only be reached through the definition; but that appears
 2097: textually below the use.
 2098: 
 2099: From this example it is clear that the visibility rules cannot be fully
 2100: implemented without major headaches. Our implementation treats common
 2101: cases as advertised and the exceptions are treated in a safe way: The
 2102: compiler makes a reasonable guess about the locals visible after a
 2103: @code{BEGIN}; if it is too pessimistic, the
 2104: user will get a spurious error about the local not being defined; if the
 2105: compiler is too optimistic, it will notice this later and issue a
 2106: warning. In the case above the compiler would complain about @code{x}
 2107: being undefined at its use. You can see from the obscure examples in
 2108: this section that it takes quite unusual control structures to get the
 2109: compiler into trouble, and even then it will often do fine.
 2110: 
 2111: If the @code{BEGIN} is reachable from above, the most optimistic guess
 2112: is that all locals visible before the @code{BEGIN} will also be
 2113: visible after the @code{BEGIN}. This guess is valid for all loops that
 2114: are entered only through the @code{BEGIN}, in particular, for normal
 2115: @code{BEGIN}...@code{WHILE}...@code{REPEAT} and
 2116: @code{BEGIN}...@code{UNTIL} loops and it is implemented in our
 2117: compiler. When the branch to the @code{BEGIN} is finally generated by
 2118: @code{AGAIN} or @code{UNTIL}, the compiler checks the guess and
 2119: warns the user if it was too optimistic:
 2120: @example
 2121: IF
 2122:   @{ x @}
 2123: BEGIN
 2124:   \ x ? 
 2125: [ 1 cs-roll ] THEN
 2126:   ...
 2127: UNTIL
 2128: @end example
 2129: 
 2130: Here, @code{x} lives only until the @code{BEGIN}, but the compiler
 2131: optimistically assumes that it lives until the @code{THEN}. It notices
 2132: this difference when it compiles the @code{UNTIL} and issues a
 2133: warning. The user can avoid the warning, and make sure that @code{x}
 2134: is not used in the wrong area by using explicit scoping:
 2135: @example
 2136: IF
 2137:   SCOPE
 2138:   @{ x @}
 2139:   ENDSCOPE
 2140: BEGIN
 2141: [ 1 cs-roll ] THEN
 2142:   ...
 2143: UNTIL
 2144: @end example
 2145: 
 2146: Since the guess is optimistic, there will be no spurious error messages
 2147: about undefined locals.
 2148: 
 2149: If the @code{BEGIN} is not reachable from above (e.g., after
 2150: @code{AHEAD} or @code{EXIT}), the compiler cannot even make an
 2151: optimistic guess, as the locals visible after the @code{BEGIN} may be
 2152: defined later. Therefore, the compiler assumes that no locals are
 2153: visible after the @code{BEGIN}. However, the user can use
 2154: @code{ASSUME-LIVE} to make the compiler assume that the same locals are
 2155: visible at the BEGIN as at the point where the top control-flow stack
 2156: item was created.
 2157: 
 2158: doc-assume-live
 2159: 
 2160: E.g.,
 2161: @example
 2162: @{ x @}
 2163: AHEAD
 2164: ASSUME-LIVE
 2165: BEGIN
 2166:   x
 2167: [ 1 CS-ROLL ] THEN
 2168:   ...
 2169: UNTIL
 2170: @end example
 2171: 
 2172: Other cases where the locals are defined before the @code{BEGIN} can be
 2173: handled by inserting an appropriate @code{CS-ROLL} before the
 2174: @code{ASSUME-LIVE} (and changing the control-flow stack manipulation
 2175: behind the @code{ASSUME-LIVE}).
 2176: 
 2177: Cases where locals are defined after the @code{BEGIN} (but should be
 2178: visible immediately after the @code{BEGIN}) can only be handled by
 2179: rearranging the loop. E.g., the ``most insidious'' example above can be
 2180: arranged into:
 2181: @example
 2182: BEGIN
 2183:   @{ x @}
 2184:   ... 0=
 2185: WHILE
 2186:   x
 2187: REPEAT
 2188: @end example
 2189: 
 2190: @node How long do locals live?, Programming Style, Where are locals visible by name?, Gforth locals
 2191: @subsubsection How long do locals live?
 2192: @cindex locals lifetime
 2193: @cindex lifetime of locals
 2194: 
 2195: The right answer for the lifetime question would be: A local lives at
 2196: least as long as it can be accessed. For a value-flavoured local this
 2197: means: until the end of its visibility. However, a variable-flavoured
 2198: local could be accessed through its address far beyond its visibility
 2199: scope. Ultimately, this would mean that such locals would have to be
 2200: garbage collected. Since this entails un-Forth-like implementation
 2201: complexities, I adopted the same cowardly solution as some other
 2202: languages (e.g., C): The local lives only as long as it is visible;
 2203: afterwards its address is invalid (and programs that access it
 2204: afterwards are erroneous).
 2205: 
 2206: @node Programming Style, Implementation, How long do locals live?, Gforth locals
 2207: @subsubsection Programming Style
 2208: @cindex locals programming style
 2209: @cindex programming style, locals
 2210: 
 2211: The freedom to define locals anywhere has the potential to change
 2212: programming styles dramatically. In particular, the need to use the
 2213: return stack for intermediate storage vanishes. Moreover, all stack
 2214: manipulations (except @code{PICK}s and @code{ROLL}s with run-time
 2215: determined arguments) can be eliminated: If the stack items are in the
 2216: wrong order, just write a locals definition for all of them; then
 2217: write the items in the order you want.
 2218: 
 2219: This seems a little far-fetched and eliminating stack manipulations is
 2220: unlikely to become a conscious programming objective. Still, the number
 2221: of stack manipulations will be reduced dramatically if local variables
 2222: are used liberally (e.g., compare @code{max} in @ref{Gforth locals} with
 2223: a traditional implementation of @code{max}).
 2224: 
 2225: This shows one potential benefit of locals: making Forth programs more
 2226: readable. Of course, this benefit will only be realized if the
 2227: programmers continue to honour the principle of factoring instead of
 2228: using the added latitude to make the words longer.
 2229: 
 2230: @cindex single-assignment style for locals
 2231: Using @code{TO} can and should be avoided.  Without @code{TO},
 2232: every value-flavoured local has only a single assignment and many
 2233: advantages of functional languages apply to Forth. I.e., programs are
 2234: easier to analyse, to optimize and to read: It is clear from the
 2235: definition what the local stands for, it does not turn into something
 2236: different later.
 2237: 
 2238: E.g., a definition using @code{TO} might look like this:
 2239: @example
 2240: : strcmp @{ addr1 u1 addr2 u2 -- n @}
 2241:  u1 u2 min 0
 2242:  ?do
 2243:    addr1 c@@ addr2 c@@ -
 2244:    ?dup-if
 2245:      unloop exit
 2246:    then
 2247:    addr1 char+ TO addr1
 2248:    addr2 char+ TO addr2
 2249:  loop
 2250:  u1 u2 - ;
 2251: @end example
 2252: Here, @code{TO} is used to update @code{addr1} and @code{addr2} at
 2253: every loop iteration. @code{strcmp} is a typical example of the
 2254: readability problems of using @code{TO}. When you start reading
 2255: @code{strcmp}, you think that @code{addr1} refers to the start of the
 2256: string. Only near the end of the loop you realize that it is something
 2257: else.
 2258: 
 2259: This can be avoided by defining two locals at the start of the loop that
 2260: are initialized with the right value for the current iteration.
 2261: @example
 2262: : strcmp @{ addr1 u1 addr2 u2 -- n @}
 2263:  addr1 addr2
 2264:  u1 u2 min 0 
 2265:  ?do @{ s1 s2 @}
 2266:    s1 c@@ s2 c@@ -
 2267:    ?dup-if
 2268:      unloop exit
 2269:    then
 2270:    s1 char+ s2 char+
 2271:  loop
 2272:  2drop
 2273:  u1 u2 - ;
 2274: @end example
 2275: Here it is clear from the start that @code{s1} has a different value
 2276: in every loop iteration.
 2277: 
 2278: @node Implementation,  , Programming Style, Gforth locals
 2279: @subsubsection Implementation
 2280: @cindex locals implementation
 2281: @cindex implementation of locals
 2282: 
 2283: @cindex locals stack
 2284: Gforth uses an extra locals stack. The most compelling reason for
 2285: this is that the return stack is not float-aligned; using an extra stack
 2286: also eliminates the problems and restrictions of using the return stack
 2287: as locals stack. Like the other stacks, the locals stack grows toward
 2288: lower addresses. A few primitives allow an efficient implementation:
 2289: 
 2290: doc-@local#
 2291: doc-f@local#
 2292: doc-laddr#
 2293: doc-lp+!#
 2294: doc-lp!
 2295: doc->l
 2296: doc-f>l
 2297: 
 2298: In addition to these primitives, some specializations of these
 2299: primitives for commonly occurring inline arguments are provided for
 2300: efficiency reasons, e.g., @code{@@local0} as specialization of
 2301: @code{@@local#} for the inline argument 0. The following compiling words
 2302: compile the right specialized version, or the general version, as
 2303: appropriate:
 2304: 
 2305: doc-compile-@local
 2306: doc-compile-f@local
 2307: doc-compile-lp+!
 2308: 
 2309: Combinations of conditional branches and @code{lp+!#} like
 2310: @code{?branch-lp+!#} (the locals pointer is only changed if the branch
 2311: is taken) are provided for efficiency and correctness in loops.
 2312: 
 2313: A special area in the dictionary space is reserved for keeping the
 2314: local variable names. @code{@{} switches the dictionary pointer to this
 2315: area and @code{@}} switches it back and generates the locals
 2316: initializing code. @code{W:} etc.@ are normal defining words. This
 2317: special area is cleared at the start of every colon definition.
 2318: 
 2319: @cindex wordlist for defining locals
 2320: A special feature of Gforth's dictionary is used to implement the
 2321: definition of locals without type specifiers: every wordlist (aka
 2322: vocabulary) has its own methods for searching
 2323: etc. (@pxref{Wordlists}). For the present purpose we defined a wordlist
 2324: with a special search method: When it is searched for a word, it
 2325: actually creates that word using @code{W:}. @code{@{} changes the search
 2326: order to first search the wordlist containing @code{@}}, @code{W:} etc.,
 2327: and then the wordlist for defining locals without type specifiers.
 2328: 
 2329: The lifetime rules support a stack discipline within a colon
 2330: definition: The lifetime of a local is either nested with other locals
 2331: lifetimes or it does not overlap them.
 2332: 
 2333: At @code{BEGIN}, @code{IF}, and @code{AHEAD} no code for locals stack
 2334: pointer manipulation is generated. Between control structure words
 2335: locals definitions can push locals onto the locals stack. @code{AGAIN}
 2336: is the simplest of the other three control flow words. It has to
 2337: restore the locals stack depth of the corresponding @code{BEGIN}
 2338: before branching. The code looks like this:
 2339: @format
 2340: @code{lp+!#} current-locals-size @minus{} dest-locals-size
 2341: @code{branch} <begin>
 2342: @end format
 2343: 
 2344: @code{UNTIL} is a little more complicated: If it branches back, it
 2345: must adjust the stack just like @code{AGAIN}. But if it falls through,
 2346: the locals stack must not be changed. The compiler generates the
 2347: following code:
 2348: @format
 2349: @code{?branch-lp+!#} <begin> current-locals-size @minus{} dest-locals-size
 2350: @end format
 2351: The locals stack pointer is only adjusted if the branch is taken.
 2352: 
 2353: @code{THEN} can produce somewhat inefficient code:
 2354: @format
 2355: @code{lp+!#} current-locals-size @minus{} orig-locals-size
 2356: <orig target>:
 2357: @code{lp+!#} orig-locals-size @minus{} new-locals-size
 2358: @end format
 2359: The second @code{lp+!#} adjusts the locals stack pointer from the
 2360: level at the @var{orig} point to the level after the @code{THEN}. The
 2361: first @code{lp+!#} adjusts the locals stack pointer from the current
 2362: level to the level at the orig point, so the complete effect is an
 2363: adjustment from the current level to the right level after the
 2364: @code{THEN}.
 2365: 
 2366: @cindex locals information on the control-flow stack
 2367: @cindex control-flow stack items, locals information
 2368: In a conventional Forth implementation a dest control-flow stack entry
 2369: is just the target address and an orig entry is just the address to be
 2370: patched. Our locals implementation adds a wordlist to every orig or dest
 2371: item. It is the list of locals visible (or assumed visible) at the point
 2372: described by the entry. Our implementation also adds a tag to identify
 2373: the kind of entry, in particular to differentiate between live and dead
 2374: (reachable and unreachable) orig entries.
 2375: 
 2376: A few unusual operations have to be performed on locals wordlists:
 2377: 
 2378: doc-common-list
 2379: doc-sub-list?
 2380: doc-list-size
 2381: 
 2382: Several features of our locals wordlist implementation make these
 2383: operations easy to implement: The locals wordlists are organised as
 2384: linked lists; the tails of these lists are shared, if the lists
 2385: contain some of the same locals; and the address of a name is greater
 2386: than the address of the names behind it in the list.
 2387: 
 2388: Another important implementation detail is the variable
 2389: @code{dead-code}. It is used by @code{BEGIN} and @code{THEN} to
 2390: determine if they can be reached directly or only through the branch
 2391: that they resolve. @code{dead-code} is set by @code{UNREACHABLE},
 2392: @code{AHEAD}, @code{EXIT} etc., and cleared at the start of a colon
 2393: definition, by @code{BEGIN} and usually by @code{THEN}.
 2394: 
 2395: Counted loops are similar to other loops in most respects, but
 2396: @code{LEAVE} requires special attention: It performs basically the same
 2397: service as @code{AHEAD}, but it does not create a control-flow stack
 2398: entry. Therefore the information has to be stored elsewhere;
 2399: traditionally, the information was stored in the target fields of the
 2400: branches created by the @code{LEAVE}s, by organizing these fields into a
 2401: linked list. Unfortunately, this clever trick does not provide enough
 2402: space for storing our extended control flow information. Therefore, we
 2403: introduce another stack, the leave stack. It contains the control-flow
 2404: stack entries for all unresolved @code{LEAVE}s.
 2405: 
 2406: Local names are kept until the end of the colon definition, even if
 2407: they are no longer visible in any control-flow path. In a few cases
 2408: this may lead to increased space needs for the locals name area, but
 2409: usually less than reclaiming this space would cost in code size.
 2410: 
 2411: 
 2412: @node ANS Forth locals,  , Gforth locals, Locals
 2413: @subsection ANS Forth locals
 2414: @cindex locals, ANS Forth style
 2415: 
 2416: The ANS Forth locals wordset does not define a syntax for locals, but
 2417: words that make it possible to define various syntaxes. One of the
 2418: possible syntaxes is a subset of the syntax we used in the Gforth locals
 2419: wordset, i.e.:
 2420: 
 2421: @example
 2422: @{ local1 local2 ... -- comment @}
 2423: @end example
 2424: or
 2425: @example
 2426: @{ local1 local2 ... @}
 2427: @end example
 2428: 
 2429: The order of the locals corresponds to the order in a stack comment. The
 2430: restrictions are:
 2431: 
 2432: @itemize @bullet
 2433: @item
 2434: Locals can only be cell-sized values (no type specifiers are allowed).
 2435: @item
 2436: Locals can be defined only outside control structures.
 2437: @item
 2438: Locals can interfere with explicit usage of the return stack. For the
 2439: exact (and long) rules, see the standard. If you don't use return stack
 2440: accessing words in a definition using locals, you will be all right. The
 2441: purpose of this rule is to make locals implementation on the return
 2442: stack easier.
 2443: @item
 2444: The whole definition must be in one line.
 2445: @end itemize
 2446: 
 2447: Locals defined in this way behave like @code{VALUE}s (@xref{Simple
 2448: Defining Words}). I.e., they are initialized from the stack. Using their
 2449: name produces their value. Their value can be changed using @code{TO}.
 2450: 
 2451: Since this syntax is supported by Gforth directly, you need not do
 2452: anything to use it. If you want to port a program using this syntax to
 2453: another ANS Forth system, use @file{compat/anslocal.fs} to implement the
 2454: syntax on the other system.
 2455: 
 2456: Note that a syntax shown in the standard, section A.13 looks
 2457: similar, but is quite different in having the order of locals
 2458: reversed. Beware!
 2459: 
 2460: The ANS Forth locals wordset itself consists of the following word
 2461: 
 2462: doc-(local)
 2463: 
 2464: The ANS Forth locals extension wordset defines a syntax, but it is so
 2465: awful that we strongly recommend not to use it. We have implemented this
 2466: syntax to make porting to Gforth easy, but do not document it here. The
 2467: problem with this syntax is that the locals are defined in an order
 2468: reversed with respect to the standard stack comment notation, making
 2469: programs harder to read, and easier to misread and miswrite. The only
 2470: merit of this syntax is that it is easy to implement using the ANS Forth
 2471: locals wordset.
 2472: 
 2473: @node Defining Words, Structures, Locals, Words
 2474: @section Defining Words
 2475: @cindex defining words
 2476: 
 2477: @menu
 2478: * Simple Defining Words::       
 2479: * Colon Definitions::           
 2480: * User-defined Defining Words::  
 2481: * Supplying names::             
 2482: * Interpretation and Compilation Semantics::  
 2483: @end menu
 2484: 
 2485: @node Simple Defining Words, Colon Definitions, Defining Words, Defining Words
 2486: @subsection Simple Defining Words
 2487: @cindex simple defining words
 2488: @cindex defining words, simple
 2489: 
 2490: doc-constant
 2491: doc-2constant
 2492: doc-fconstant
 2493: doc-variable
 2494: doc-2variable
 2495: doc-fvariable
 2496: doc-create
 2497: doc-user
 2498: doc-value
 2499: doc-to
 2500: doc-defer
 2501: doc-is
 2502: 
 2503: @node Colon Definitions, User-defined Defining Words, Simple Defining Words, Defining Words
 2504: @subsection Colon Definitions
 2505: @cindex colon definitions
 2506: 
 2507: @example
 2508: : name ( ... -- ... )
 2509:     word1 word2 word3 ;
 2510: @end example
 2511: 
 2512: creates a word called @code{name}, that, upon execution, executes
 2513: @code{word1 word2 word3}. @code{name} is a @dfn{(colon) definition}.
 2514: 
 2515: The explanation above is somewhat superficial. @xref{Interpretation and
 2516: Compilation Semantics} for an in-depth discussion of some of the issues
 2517: involved.
 2518: 
 2519: doc-:
 2520: doc-;
 2521: 
 2522: @node User-defined Defining Words, Supplying names, Colon Definitions, Defining Words
 2523: @subsection User-defined Defining Words
 2524: @cindex user-defined defining words
 2525: @cindex defining words, user-defined
 2526: 
 2527: You can create new defining words simply by wrapping defining-time code
 2528: around existing defining words and putting the sequence in a colon
 2529: definition.
 2530: 
 2531: @cindex @code{CREATE} ... @code{DOES>}
 2532: If you want the words defined with your defining words to behave
 2533: differently from words defined with standard defining words, you can
 2534: write your defining word like this:
 2535: 
 2536: @example
 2537: : def-word ( "name" -- )
 2538:     Create @var{code1}
 2539: DOES> ( ... -- ... )
 2540:     @var{code2} ;
 2541: 
 2542: def-word name
 2543: @end example
 2544: 
 2545: Technically, this fragment defines a defining word @code{def-word}, and
 2546: a word @code{name}; when you execute @code{name}, the address of the
 2547: body of @code{name} is put on the data stack and @var{code2} is executed
 2548: (the address of the body of @code{name} is the address @code{HERE}
 2549: returns immediately after the @code{CREATE}).
 2550: 
 2551: In other words, if you make the following definitions:
 2552: 
 2553: @example
 2554: : def-word1 ( "name" -- )
 2555:     Create @var{code1} ;
 2556: 
 2557: : action1 ( ... -- ... )
 2558:     @var{code2} ;
 2559: 
 2560: def-word name1
 2561: @end example
 2562: 
 2563: Using @code{name1 action1} is equivalent to using @code{name}.
 2564: 
 2565: E.g., you can implement @code{Constant} in this way:
 2566: 
 2567: @example
 2568: : constant ( w "name" -- )
 2569:     create ,
 2570: DOES> ( -- w )
 2571:     @@ ;
 2572: @end example
 2573: 
 2574: When you create a constant with @code{5 constant five}, first a new word
 2575: @code{five} is created, then the value 5 is laid down in the body of
 2576: @code{five} with @code{,}. When @code{five} is invoked, the address of
 2577: the body is put on the stack, and @code{@@} retrieves the value 5.
 2578: 
 2579: @cindex stack effect of @code{DOES>}-parts
 2580: @cindex @code{DOES>}-parts, stack effect
 2581: In the example above the stack comment after the @code{DOES>} specifies
 2582: the stack effect of the defined words, not the stack effect of the
 2583: following code (the following code expects the address of the body on
 2584: the top of stack, which is not reflected in the stack comment). This is
 2585: the convention that I use and recommend (it clashes a bit with using
 2586: locals declarations for stack effect specification, though).
 2587: 
 2588: @subsubsection Applications of @code{CREATE..DOES>}
 2589: @cindex @code{CREATE} ... @code{DOES>}, applications
 2590: 
 2591: You may wonder how to use this feature. Here are some usage patterns:
 2592: 
 2593: @cindex factoring similar colon definitions
 2594: When you see a sequence of code occurring several times, and you can
 2595: identify a meaning, you will factor it out as a colon definition. When
 2596: you see similar colon definitions, you can factor them using
 2597: @code{CREATE..DOES>}. E.g., an assembler usually defines several words
 2598: that look very similar:
 2599: @example
 2600: : ori, ( reg-target reg-source n -- )
 2601:     0 asm-reg-reg-imm ;
 2602: : andi, ( reg-target reg-source n -- )
 2603:     1 asm-reg-reg-imm ;
 2604: @end example
 2605: 
 2606: This could be factored with:
 2607: @example
 2608: : reg-reg-imm ( op-code -- )
 2609:     create ,
 2610: DOES> ( reg-target reg-source n -- )
 2611:     @@ asm-reg-reg-imm ;
 2612: 
 2613: 0 reg-reg-imm ori,
 2614: 1 reg-reg-imm andi,
 2615: @end example
 2616: 
 2617: @cindex currying
 2618: Another view of @code{CREATE..DOES>} is to consider it as a crude way to
 2619: supply a part of the parameters for a word (known as @dfn{currying} in
 2620: the functional language community). E.g., @code{+} needs two
 2621: parameters. Creating versions of @code{+} with one parameter fixed can
 2622: be done like this:
 2623: @example
 2624: : curry+ ( n1 -- )
 2625:     create ,
 2626: DOES> ( n2 -- n1+n2 )
 2627:     @@ + ;
 2628: 
 2629:  3 curry+ 3+
 2630: -2 curry+ 2-
 2631: @end example
 2632: 
 2633: @subsubsection The gory details of @code{CREATE..DOES>}
 2634: @cindex @code{CREATE} ... @code{DOES>}, details
 2635: 
 2636: doc-does>
 2637: 
 2638: @cindex @code{DOES>} in a separate definition
 2639: This means that you need not use @code{CREATE} and @code{DOES>} in the
 2640: same definition; E.g., you can put the @code{DOES>}-part in a separate
 2641: definition. This allows us to, e.g., select among different DOES>-parts:
 2642: @example
 2643: : does1 
 2644: DOES> ( ... -- ... )
 2645:     ... ;
 2646: 
 2647: : does2
 2648: DOES> ( ... -- ... )
 2649:     ... ;
 2650: 
 2651: : def-word ( ... -- ... )
 2652:     create ...
 2653:     IF
 2654:        does1
 2655:     ELSE
 2656:        does2
 2657:     ENDIF ;
 2658: @end example
 2659: 
 2660: @cindex @code{DOES>} in interpretation state
 2661: In a standard program you can apply a @code{DOES>}-part only if the last
 2662: word was defined with @code{CREATE}. In Gforth, the @code{DOES>}-part
 2663: will override the behaviour of the last word defined in any case. In a
 2664: standard program, you can use @code{DOES>} only in a colon
 2665: definition. In Gforth, you can also use it in interpretation state, in a
 2666: kind of one-shot mode:
 2667: @example
 2668: CREATE name ( ... -- ... )
 2669:   @var{initialization}
 2670: DOES>
 2671:   @var{code} ;
 2672: @end example
 2673: This is equivalent to the standard
 2674: @example
 2675: :noname
 2676: DOES>
 2677:     @var{code} ;
 2678: CREATE name EXECUTE ( ... -- ... )
 2679:     @var{initialization}
 2680: @end example
 2681: 
 2682: You can get the address of the body of a word with
 2683: 
 2684: doc->body
 2685: 
 2686: @node Supplying names, Interpretation and Compilation Semantics, User-defined Defining Words, Defining Words
 2687: @subsection Supplying names for the defined words
 2688: @cindex names for defined words
 2689: @cindex defining words, name parameter
 2690: 
 2691: @cindex defining words, name given in a string
 2692: By default, defining words take the names for the defined words from the
 2693: input stream. Sometimes you want to supply the name from a string. You
 2694: can do this with
 2695: 
 2696: doc-nextname
 2697: 
 2698: E.g.,
 2699: 
 2700: @example
 2701: s" foo" nextname create
 2702: @end example
 2703: is equivalent to
 2704: @example
 2705: create foo
 2706: @end example
 2707: 
 2708: @cindex defining words without name
 2709: Sometimes you want to define a word without a name. You can do this with
 2710: 
 2711: doc-noname
 2712: 
 2713: @cindex execution token of last defined word
 2714: To make any use of the newly defined word, you need its execution
 2715: token. You can get it with
 2716: 
 2717: doc-lastxt
 2718: 
 2719: E.g., you can initialize a deferred word with an anonymous colon
 2720: definition:
 2721: @example
 2722: Defer deferred
 2723: noname : ( ... -- ... )
 2724:   ... ;
 2725: lastxt IS deferred
 2726: @end example
 2727: 
 2728: @code{lastxt} also works when the last word was not defined as
 2729: @code{noname}. 
 2730: 
 2731: The standard has also recognized the need for anonymous words and
 2732: provides
 2733: 
 2734: doc-:noname
 2735: 
 2736: This leaves the execution token for the word on the stack after the
 2737: closing @code{;}. You can rewrite the last example with @code{:noname}:
 2738: @example
 2739: Defer deferred
 2740: :noname ( ... -- ... )
 2741:   ... ;
 2742: IS deferred
 2743: @end example
 2744: 
 2745: @node Interpretation and Compilation Semantics,  , Supplying names, Defining Words
 2746: @subsection Interpretation and Compilation Semantics
 2747: @cindex semantics, interpretation and compilation
 2748: 
 2749: @cindex interpretation semantics
 2750: The @dfn{interpretation semantics} of a word are what the text
 2751: interpreter does when it encounters the word in interpret state. It also
 2752: appears in some other contexts, e.g., the execution token returned by
 2753: @code{' @var{word}} identifies the interpretation semantics of
 2754: @var{word} (in other words, @code{' @var{word} execute} is equivalent to
 2755: interpret-state text interpretation of @code{@var{word}}).
 2756: 
 2757: @cindex compilation semantics
 2758: The @dfn{compilation semantics} of a word are what the text interpreter
 2759: does when it encounters the word in compile state. It also appears in
 2760: other contexts, e.g, @code{POSTPONE @var{word}} compiles@footnote{In
 2761: standard terminology, ``appends to the current definition''.} the
 2762: compilation semantics of @var{word}.
 2763: 
 2764: @cindex execution semantics
 2765: The standard also talks about @dfn{execution semantics}. They are used
 2766: only for defining the interpretation and compilation semantics of many
 2767: words. By default, the interpretation semantics of a word are to
 2768: @code{execute} its execution semantics, and the compilation semantics of
 2769: a word are to @code{compile,} its execution semantics.@footnote{In
 2770: standard terminology: The default interpretation semantics are its
 2771: execution semantics; the default compilation semantics are to append its
 2772: execution semantics to the execution semantics of the current
 2773: definition.}
 2774: 
 2775: @cindex immediate words
 2776: You can change the compilation semantics into @code{execute}ing the
 2777: execution semantics with
 2778: 
 2779: doc-immediate
 2780: 
 2781: @cindex compile-only words
 2782: You can remove the interpretation semantics of a word with
 2783: 
 2784: doc-compile-only
 2785: doc-restrict
 2786: 
 2787: Note that ticking (@code{'}) compile-only words gives an error
 2788: (``Interpreting a compile-only word'').
 2789: 
 2790: Gforth also allows you to define words with arbitrary combinations of
 2791: interpretation and compilation semantics.
 2792: 
 2793: doc-interpret/compile:
 2794: 
 2795: This feature was introduced for implementing @code{TO} and @code{S"}. I
 2796: recommend that you do not define such words, as cute as they may be:
 2797: they make it hard to get at both parts of the word in some contexts.
 2798: E.g., assume you want to get an execution token for the compilation
 2799: part. Instead, define two words, one that embodies the interpretation
 2800: part, and one that embodies the compilation part.  Once you have done
 2801: that, you can define a combined word with @code{interpret/compile:} for
 2802: the convenience of your users.
 2803: 
 2804: You also might try to provide an optimizing implementation of the
 2805: default compilation semantics with this feature, like this:
 2806: 
 2807: @example
 2808: :noname
 2809:    foo bar ;
 2810: :noname
 2811:    POSTPONE foo POSTPONE bar ;
 2812: interpret/compile: foobar
 2813: @end example
 2814: 
 2815: as an optimizing version of
 2816: 
 2817: @example
 2818: : foobar
 2819:     foo bar ;
 2820: @end example
 2821: 
 2822: Unfortunately, this does not work correctly with @code{[compile]},
 2823: because @code{[compile]} assumes that the compilation semantics of all
 2824: @code{interpret/compile:} words are non-default. I.e., @code{[compile]
 2825: foobar} would compile the compilation semantics for the optimizing
 2826: @code{foobar}, whereas it would compile the interpretation semantics for
 2827: the non-optimizing @code{foobar}.
 2828: 
 2829: @cindex state-smart words are a bad idea
 2830: Some people try to use state-smart words to emulate the feature provided
 2831: by @code{interpret/compile:} (words are state-smart if they check
 2832: @code{STATE} during execution). E.g., they would try to code
 2833: @code{foobar} like this:
 2834: 
 2835: @example
 2836: : foobar
 2837:   STATE @@
 2838:   IF ( compilation state )
 2839:     POSTPONE foo POSTPONE bar
 2840:   ELSE
 2841:     foo bar
 2842:   ENDIF ; immediate
 2843: @end example
 2844: 
 2845: While this works if @code{foobar} is processed only by the text
 2846: interpreter, it does not work in other contexts (like @code{'} or
 2847: @code{POSTPONE}). E.g., @code{' foobar} will produce an execution token
 2848: for a state-smart word, not for the interpretation semantics of the
 2849: original @code{foobar}; when you execute this execution token (directly
 2850: with @code{EXECUTE} or indirectly through @code{COMPILE,}) in compile
 2851: state, the result will not be what you expected (i.e., it will not
 2852: perform @code{foo bar}). State-smart words are a bad idea. Simply don't
 2853: write them!
 2854: 
 2855: @cindex defining words with arbitrary semantics combinations
 2856: It is also possible to write defining words that define words with
 2857: arbitrary combinations of interpretation and compilation semantics. In
 2858: general, this looks like:
 2859: 
 2860: @example
 2861: : def-word
 2862:     create-interpret/compile
 2863:     @var{code1}
 2864: interpretation>
 2865:     @var{code2}
 2866: <interpretation
 2867: compilation>
 2868:     @var{code3}
 2869: <compilation ;
 2870: @end example
 2871: 
 2872: For a @var{word} defined with @code{def-word}, the interpretation
 2873: semantics are to push the address of the body of @var{word} and perform
 2874: @var{code2}, and the compilation semantics are to push the address of
 2875: the body of @var{word} and perform @var{code3}. E.g., @code{constant}
 2876: can also be defined like this (except that the defined constants don't
 2877: behave correctly when @code{[compile]}d):
 2878: 
 2879: @example
 2880: : constant ( n "name" -- )
 2881:     create-interpret/compile
 2882:     ,
 2883: interpretation> ( -- n )
 2884:     @@
 2885: <interpretation
 2886: compilation> ( compilation. -- ; run-time. -- n )
 2887:     @@ postpone literal
 2888: <compilation ;
 2889: @end example
 2890: 
 2891: doc-create-interpret/compile
 2892: doc-interpretation>
 2893: doc-<interpretation
 2894: doc-compilation>
 2895: doc-<compilation
 2896: 
 2897: Note that words defined with @code{interpret/compile:} and
 2898: @code{create-interpret/compile} have an extended header structure that
 2899: differs from other words; however, unless you try to access them with
 2900: plain address arithmetic, you should not notice this. Words for
 2901: accessing the header structure usually know how to deal with this; e.g.,
 2902: @code{' word >body} also gives you the body of a word created with
 2903: @code{create-interpret/compile}.
 2904: 
 2905: @c ----------------------------------------------------------
 2906: @node Structures, Object-oriented Forth, Defining Words, Words
 2907: @section  Structures
 2908: @cindex structures
 2909: @cindex records
 2910: 
 2911: This section presents the structure package that comes with Gforth. A
 2912: version of the package implemented in plain ANS Forth is available in
 2913: @file{compat/struct.fs}. This package was inspired by a posting on
 2914: comp.lang.forth in 1989 (unfortunately I don't remember, by whom;
 2915: possibly John Hayes). A version of this section has been published in
 2916: ???. Marcel Hendrix provided helpful comments.
 2917: 
 2918: @menu
 2919: * Why explicit structure support?::  
 2920: * Structure Usage::             
 2921: * Structure Naming Convention::  
 2922: * Structure Implementation::    
 2923: * Structure Glossary::          
 2924: @end menu
 2925: 
 2926: @node Why explicit structure support?, Structure Usage, Structures, Structures
 2927: @subsection Why explicit structure support?
 2928: 
 2929: @cindex address arithmetic for structures
 2930: @cindex structures using address arithmetic
 2931: If we want to use a structure containing several fields, we could simply
 2932: reserve memory for it, and access the fields using address arithmetic
 2933: (@pxref{Address arithmetic}). As an example, consider a structure with
 2934: the following fields
 2935: 
 2936: @table @code
 2937: @item a
 2938: is a float
 2939: @item b
 2940: is a cell
 2941: @item c
 2942: is a float
 2943: @end table
 2944: 
 2945: Given the (float-aligned) base address of the structure we get the
 2946: address of the field
 2947: 
 2948: @table @code
 2949: @item a
 2950: without doing anything further.
 2951: @item b
 2952: with @code{float+}
 2953: @item c
 2954: with @code{float+ cell+ faligned}
 2955: @end table
 2956: 
 2957: It is easy to see that this can become quite tiring. 
 2958: 
 2959: Moreover, it is not very readable, because seeing a
 2960: @code{cell+} tells us neither which kind of structure is
 2961: accessed nor what field is accessed; we have to somehow infer the kind
 2962: of structure, and then look up in the documentation, which field of
 2963: that structure corresponds to that offset.
 2964: 
 2965: Finally, this kind of address arithmetic also causes maintenance
 2966: troubles: If you add or delete a field somewhere in the middle of the
 2967: structure, you have to find and change all computations for the fields
 2968: afterwards.
 2969: 
 2970: So, instead of using @code{cell+} and friends directly, how
 2971: about storing the offsets in constants:
 2972: 
 2973: @example
 2974: 0 constant a-offset
 2975: 0 float+ constant b-offset
 2976: 0 float+ cell+ faligned c-offset
 2977: @end example
 2978: 
 2979: Now we can get the address of field @code{x} with @code{x-offset
 2980: +}. This is much better in all respects. Of course, you still
 2981: have to change all later offset definitions if you add a field. You can
 2982: fix this by declaring the offsets in the following way:
 2983: 
 2984: @example
 2985: 0 constant a-offset
 2986: a-offset float+ constant b-offset
 2987: b-offset cell+ faligned constant c-offset
 2988: @end example
 2989: 
 2990: Since we always use the offsets with @code{+}, using a defining
 2991: word @code{cfield} that includes the @code{+} in the
 2992: action of the defined word offers itself:
 2993: 
 2994: @example
 2995: : cfield ( n "name" -- )
 2996:     create ,
 2997: does> ( name execution: addr1 -- addr2 )
 2998:     @@ + ;
 2999: 
 3000: 0 cfield a
 3001: 0 a float+ cfield b
 3002: 0 b cell+ faligned cfield c
 3003: @end example
 3004: 
 3005: Instead of @code{x-offset +}, we now simply write @code{x}.
 3006: 
 3007: The structure field words now can be used quite nicely. However,
 3008: their definition is still a bit cumbersome: We have to repeat the
 3009: name, the information about size and alignment is distributed before
 3010: and after the field definitions etc.  The structure package presented
 3011: here addresses these problems.
 3012: 
 3013: @node Structure Usage, Structure Naming Convention, Why explicit structure support?, Structures
 3014: @subsection Structure Usage
 3015: @cindex structure usage
 3016: 
 3017: @cindex @code{field} usage
 3018: @cindex @code{struct} usage
 3019: @cindex @code{end-struct} usage
 3020: You can define a structure for a (data-less) linked list with
 3021: @example
 3022: struct
 3023:     cell% field list-next
 3024: end-struct list%
 3025: @end example
 3026: 
 3027: With the address of the list node on the stack, you can compute the
 3028: address of the field that contains the address of the next node with
 3029: @code{list-next}. E.g., you can determine the length of a list
 3030: with:
 3031: 
 3032: @example
 3033: : list-length ( list -- n )
 3034: \ "list" is a pointer to the first element of a linked list
 3035: \ "n" is the length of the list
 3036:     0 begin ( list1 n1 )
 3037:         over
 3038:     while ( list1 n1 )
 3039:         1+ swap list-next @@ swap
 3040:     repeat
 3041:     nip ;
 3042: @end example
 3043: 
 3044: You can reserve memory for a list node in the dictionary with
 3045: @code{list% %allot}, which leaves the address of the list node on the
 3046: stack. For the equivalent allocation on the heap you can use @code{list%
 3047: %alloc} (or, for an @code{allocate}-like stack effect (i.e., with ior),
 3048: use @code{list% %allocate}). You can also get the the size of a list
 3049: node with @code{list% %size} and it's alignment with @code{list%
 3050: %alignment}.
 3051: 
 3052: Note that in ANS Forth the body of a @code{create}d word is
 3053: @code{aligned} but not necessarily @code{faligned};
 3054: therefore, if you do a
 3055: @example
 3056: create @emph{name} foo% %allot
 3057: @end example
 3058: 
 3059: then the memory alloted for @code{foo%} is
 3060: guaranteed to start at the body of @code{@emph{name}} only if
 3061: @code{foo%} contains only character, cell and double fields.
 3062: 
 3063: @cindex strcutures containing structures
 3064: You can also include a structure @code{foo%} as field of
 3065: another structure, with:
 3066: @example
 3067: struct
 3068: ...
 3069:     foo% field ...
 3070: ...
 3071: end-struct ...
 3072: @end example
 3073: 
 3074: @cindex structure extension
 3075: @cindex extended records
 3076: Instead of starting with an empty structure, you can also extend an
 3077: existing structure. E.g., a plain linked list without data, as defined
 3078: above, is hardly useful; You can extend it to a linked list of integers,
 3079: like this:@footnote{This feature is also known as @emph{extended
 3080: records}. It is the main innovation in the Oberon language; in other
 3081: words, adding this feature to Modula-2 led Wirth to create a new
 3082: language, write a new compiler etc.  Adding this feature to Forth just
 3083: requires a few lines of code.}
 3084: 
 3085: @example
 3086: list%
 3087:     cell% field intlist-int
 3088: end-struct intlist%
 3089: @end example
 3090: 
 3091: @code{intlist%} is a structure with two fields:
 3092: @code{list-next} and @code{intlist-int}.
 3093: 
 3094: @cindex structures containing arrays
 3095: You can specify an array type containing @emph{n} elements of
 3096: type @code{foo%} like this:
 3097: 
 3098: @example
 3099: foo% @emph{n} *
 3100: @end example
 3101: 
 3102: You can use this array type in any place where you can use a normal
 3103: type, e.g., when defining a @code{field}, or with
 3104: @code{%allot}.
 3105: 
 3106: @cindex first field optimization
 3107: The first field is at the base address of a structure and the word
 3108: for this field (e.g., @code{list-next}) actually does not change
 3109: the address on the stack. You may be tempted to leave it away in the
 3110: interest of run-time and space efficiency. This is not necessary,
 3111: because the structure package optimizes this case and compiling such
 3112: words does not generate any code. So, in the interest of readability
 3113: and maintainability you should include the word for the field when
 3114: accessing the field.
 3115: 
 3116: @node Structure Naming Convention, Structure Implementation, Structure Usage, Structures
 3117: @subsection Structure Naming Convention
 3118: @cindex structure naming conventions
 3119: 
 3120: The field names that come to (my) mind are often quite generic, and,
 3121: if used, would cause frequent name clashes. E.g., many structures
 3122: probably contain a @code{counter} field. The structure names
 3123: that come to (my) mind are often also the logical choice for the names
 3124: of words that create such a structure.
 3125: 
 3126: Therefore, I have adopted the following naming conventions: 
 3127: 
 3128: @itemize @bullet
 3129: @cindex field naming convention
 3130: @item
 3131: The names of fields are of the form
 3132: @code{@emph{struct}-@emph{field}}, where
 3133: @code{@emph{struct}} is the basic name of the structure, and
 3134: @code{@emph{field}} is the basic name of the field. You can
 3135: think about field words as converting converts the (address of the)
 3136: structure into the (address of the) field.
 3137: 
 3138: @cindex structure naming convention
 3139: @item
 3140: The names of structures are of the form
 3141: @code{@emph{struct}%}, where
 3142: @code{@emph{struct}} is the basic name of the structure.
 3143: @end itemize
 3144: 
 3145: This naming convention does not work that well for fields of extended
 3146: structures; e.g., the integer list structure has a field
 3147: @code{intlist-int}, but has @code{list-next}, not
 3148: @code{intlist-next}.
 3149: 
 3150: @node Structure Implementation, Structure Glossary, Structure Naming Convention, Structures
 3151: @subsection Structure Implementation
 3152: @cindex structure implementation
 3153: @cindex implementation of structures
 3154: 
 3155: The central idea in the implementation is to pass the data about the
 3156: structure being built on the stack, not in some global
 3157: variable. Everything else falls into place naturally once this design
 3158: decision is made.
 3159: 
 3160: The type description on the stack is of the form @emph{align
 3161: size}. Keeping the size on the top-of-stack makes dealing with arrays
 3162: very simple.
 3163: 
 3164: @code{field} is a defining word that uses @code{create}
 3165: and @code{does>}. The body of the field contains the offset
 3166: of the field, and the normal @code{does>} action is
 3167: 
 3168: @example
 3169: @ +
 3170: @end example
 3171: 
 3172: i.e., add the offset to the address, giving the stack effect
 3173: @code{addr1 -- addr2} for a field.
 3174: 
 3175: @cindex first field optimization, implementation
 3176: This simple structure is slightly complicated by the optimization
 3177: for fields with offset 0, which requires a different
 3178: @code{does>}-part (because we cannot rely on there being
 3179: something on the stack if such a field is invoked during
 3180: compilation). Therefore, we put the different @code{does>}-parts
 3181: in separate words, and decide which one to invoke based on the
 3182: offset. For a zero offset, the field is basically a noop; it is
 3183: immediate, and therefore no code is generated when it is compiled.
 3184: 
 3185: @node Structure Glossary,  , Structure Implementation, Structures
 3186: @subsection Structure Glossary
 3187: @cindex structure glossary
 3188: 
 3189: doc-%align
 3190: doc-%alignment
 3191: doc-%alloc
 3192: doc-%allocate
 3193: doc-%allot
 3194: doc-cell%
 3195: doc-char%
 3196: doc-dfloat%
 3197: doc-double%
 3198: doc-end-struct
 3199: doc-field
 3200: doc-float%
 3201: doc-nalign
 3202: doc-sfloat%
 3203: doc-%size
 3204: doc-struct
 3205: 
 3206: @c -------------------------------------------------------------
 3207: @node Object-oriented Forth, Tokens for Words, Structures, Words
 3208: @section Object-oriented Forth
 3209: 
 3210: Gforth comes with three packets for object-oriented programming,
 3211: @file{objects.fs}, @file{oof.fs}, and @file{mini-oof.fs}; none of them
 3212: is preloaded, so you have to @code{include} them before use. The most
 3213: important differences between these packets (and others) are discussed
 3214: in @ref{Comparison with other object models}. All packets are written
 3215: in ANS Forth and can be used with any other ANS Forth.
 3216: 
 3217: @menu
 3218: * Objects::                     
 3219: * OOF::                         
 3220: * Mini-OOF::                    
 3221: @end menu
 3222: 
 3223: @node Objects, OOF, Object-oriented Forth, Object-oriented Forth
 3224: @subsection Objects
 3225: @cindex objects
 3226: @cindex object-oriented programming
 3227: 
 3228: @cindex @file{objects.fs}
 3229: @cindex @file{oof.fs}
 3230: 
 3231: This section describes the @file{objects.fs} packet. This material also has been published in @cite{Yet Another Forth Objects Package} by Anton Ertl and appeared in Forth Dimensions 19(2), pages 37--43 (@url{http://www.complang.tuwien.ac.at/forth/objects/objects.html}).
 3232: @c McKewan's and Zsoter's packages
 3233: 
 3234: This section assumes (in some places) that you have read @ref{Structures}.
 3235: 
 3236: @menu
 3237: * Properties of the Objects model::  
 3238: * Why object-oriented programming?::  
 3239: * Object-Oriented Terminology::  
 3240: * Basic Objects Usage::         
 3241: * The class Object::            
 3242: * Creating objects::            
 3243: * Object-Oriented Programming Style::  
 3244: * Class Binding::               
 3245: * Method conveniences::         
 3246: * Classes and Scoping::         
 3247: * Object Interfaces::           
 3248: * Objects Implementation::      
 3249: * Comparison with other object models::  
 3250: * Objects Glossary::            
 3251: @end menu
 3252: 
 3253: Marcel Hendrix provided helpful comments on this section. Andras Zsoter
 3254: and Bernd Paysan helped me with the related works section.
 3255: 
 3256: @node Properties of the Objects model, Why object-oriented programming?, Objects, Objects
 3257: @subsubsection Properties of the @file{objects.fs} model
 3258: @cindex @file{objects.fs} properties
 3259: 
 3260: @itemize @bullet
 3261: @item
 3262: It is straightforward to pass objects on the stack. Passing
 3263: selectors on the stack is a little less convenient, but possible.
 3264: 
 3265: @item
 3266: Objects are just data structures in memory, and are referenced by
 3267: their address. You can create words for objects with normal defining
 3268: words like @code{constant}. Likewise, there is no difference
 3269: between instance variables that contain objects and those
 3270: that contain other data.
 3271: 
 3272: @item
 3273: Late binding is efficient and easy to use.
 3274: 
 3275: @item
 3276: It avoids parsing, and thus avoids problems with state-smartness
 3277: and reduced extensibility; for convenience there are a few parsing
 3278: words, but they have non-parsing counterparts. There are also a few
 3279: defining words that parse. This is hard to avoid, because all standard
 3280: defining words parse (except @code{:noname}); however, such
 3281: words are not as bad as many other parsing words, because they are not
 3282: state-smart.
 3283: 
 3284: @item
 3285: It does not try to incorporate everything. It does a few things
 3286: and does them well (IMO). In particular, I did not intend to support
 3287: information hiding with this model (although it has features that may
 3288: help); you can use a separate package for achieving this.
 3289: 
 3290: @item
 3291: It is layered; you don't have to learn and use all features to use this
 3292: model. Only a few features are necessary (@xref{Basic Objects Usage},
 3293: @xref{The class Object}, @xref{Creating objects}.), the others
 3294: are optional and independent of each other.
 3295: 
 3296: @item
 3297: An implementation in ANS Forth is available.
 3298: 
 3299: @end itemize
 3300: 
 3301: I have used the technique, on which this model is based, for
 3302: implementing the parser generator Gray; we have also used this technique
 3303: in Gforth for implementing the various flavours of wordlists (hashed or
 3304: not, case-sensitive or not, special-purpose wordlists for locals etc.).
 3305: 
 3306: @node Why object-oriented programming?, Object-Oriented Terminology, Properties of the Objects model, Objects
 3307: @subsubsection Why object-oriented programming?
 3308: @cindex object-oriented programming motivation
 3309: @cindex motivation for object-oriented programming
 3310: 
 3311: Often we have to deal with several data structures (@emph{objects}),
 3312: that have to be treated similarly in some respects, but differ in
 3313: others. Graphical objects are the textbook example: circles,
 3314: triangles, dinosaurs, icons, and others, and we may want to add more
 3315: during program development. We want to apply some operations to any
 3316: graphical object, e.g., @code{draw} for displaying it on the
 3317: screen. However, @code{draw} has to do something different for
 3318: every kind of object.
 3319: 
 3320: We could implement @code{draw} as a big @code{CASE}
 3321: control structure that executes the appropriate code depending on the
 3322: kind of object to be drawn. This would be not be very elegant, and,
 3323: moreover, we would have to change @code{draw} every time we add
 3324: a new kind of graphical object (say, a spaceship).
 3325: 
 3326: What we would rather do is: When defining spaceships, we would tell
 3327: the system: "Here's how you @code{draw} a spaceship; you figure
 3328: out the rest."
 3329: 
 3330: This is the problem that all systems solve that (rightfully) call
 3331: themselves object-oriented, and the object-oriented package I present
 3332: here also solves this problem (and not much else).
 3333: 
 3334: @node Object-Oriented Terminology, Basic Objects Usage, Why object-oriented programming?, Objects
 3335: @subsubsection Object-Oriented Terminology
 3336: @cindex object-oriented terminology
 3337: @cindex terminology for object-oriented programming
 3338: 
 3339: This section is mainly for reference, so you don't have to understand
 3340: all of it right away.  The terminology is mainly Smalltalk-inspired.  In
 3341: short:
 3342: 
 3343: @table @emph
 3344: @cindex class
 3345: @item class
 3346: a data structure definition with some extras.
 3347: 
 3348: @cindex object
 3349: @item object
 3350: an instance of the data structure described by the class definition.
 3351: 
 3352: @cindex instance variables
 3353: @item instance variables
 3354: fields of the data structure.
 3355: 
 3356: @cindex selector
 3357: @cindex method selector
 3358: @cindex virtual function
 3359: @item selector
 3360: (or @emph{method selector}) a word (e.g.,
 3361: @code{draw}) for performing an operation on a variety of data
 3362: structures (classes). A selector describes @emph{what} operation to
 3363: perform. In C++ terminology: a (pure) virtual function.
 3364: 
 3365: @cindex method
 3366: @item method
 3367: the concrete definition that performs the operation
 3368: described by the selector for a specific class. A method specifies
 3369: @emph{how} the operation is performed for a specific class.
 3370: 
 3371: @cindex selector invocation
 3372: @cindex message send
 3373: @cindex invoking a selector
 3374: @item selector invocation
 3375: a call of a selector. One argument of the call (the TOS (top-of-stack))
 3376: is used for determining which method is used. In Smalltalk terminology:
 3377: a message (consisting of the selector and the other arguments) is sent
 3378: to the object.
 3379: 
 3380: @cindex receiving object
 3381: @item receiving object
 3382: the object used for determining the method executed by a selector
 3383: invocation. In our model it is the object that is on the TOS when the
 3384: selector is invoked. (@emph{Receiving} comes from Smalltalks
 3385: @emph{message} terminology.)
 3386: 
 3387: @cindex child class
 3388: @cindex parent class
 3389: @cindex inheritance
 3390: @item child class
 3391: a class that has (@emph{inherits}) all properties (instance variables,
 3392: selectors, methods) from a @emph{parent class}. In Smalltalk
 3393: terminology: The subclass inherits from the superclass. In C++
 3394: terminology: The derived class inherits from the base class.
 3395: 
 3396: @end table
 3397: 
 3398: @c If you wonder about the message sending terminology, it comes from
 3399: @c a time when each object had it's own task and objects communicated via
 3400: @c message passing; eventually the Smalltalk developers realized that
 3401: @c they can do most things through simple (indirect) calls. They kept the
 3402: @c terminology.
 3403: 
 3404: @node Basic Objects Usage, The class Object, Object-Oriented Terminology, Objects
 3405: @subsubsection Basic Objects Usage
 3406: @cindex basic objects usage
 3407: @cindex objects, basic usage
 3408: 
 3409: You can define a class for graphical objects like this:
 3410: 
 3411: @cindex @code{class} usage
 3412: @cindex @code{end-class} usage
 3413: @cindex @code{selector} usage
 3414: @example
 3415: object class \ "object" is the parent class
 3416:   selector draw ( x y graphical -- )
 3417: end-class graphical
 3418: @end example
 3419: 
 3420: This code defines a class @code{graphical} with an
 3421: operation @code{draw}.  We can perform the operation
 3422: @code{draw} on any @code{graphical} object, e.g.:
 3423: 
 3424: @example
 3425: 100 100 t-rex draw
 3426: @end example
 3427: 
 3428: where @code{t-rex} is a word (say, a constant) that produces a
 3429: graphical object.
 3430: 
 3431: @cindex abstract class
 3432: How do we create a graphical object? With the present definitions,
 3433: we cannot create a useful graphical object. The class
 3434: @code{graphical} describes graphical objects in general, but not
 3435: any concrete graphical object type (C++ users would call it an
 3436: @emph{abstract class}); e.g., there is no method for the selector
 3437: @code{draw} in the class @code{graphical}.
 3438: 
 3439: For concrete graphical objects, we define child classes of the
 3440: class @code{graphical}, e.g.:
 3441: 
 3442: @cindex @code{overrides} usage
 3443: @cindex @code{field} usage in class definition
 3444: @example
 3445: graphical class \ "graphical" is the parent class
 3446:   cell% field circle-radius
 3447: 
 3448: :noname ( x y circle -- )
 3449:   circle-radius @@ draw-circle ;
 3450: overrides draw
 3451: 
 3452: :noname ( n-radius circle -- )
 3453:   circle-radius ! ;
 3454: overrides construct
 3455: 
 3456: end-class circle
 3457: @end example
 3458: 
 3459: Here we define a class @code{circle} as a child of @code{graphical},
 3460: with a field @code{circle-radius} (which behaves just like a field in
 3461: @pxref{Structures}); it defines new methods for the selectors
 3462: @code{draw} and @code{construct} (@code{construct} is defined in
 3463: @code{object}, the parent class of @code{graphical}).
 3464: 
 3465: Now we can create a circle on the heap (i.e.,
 3466: @code{allocate}d memory) with
 3467: 
 3468: @cindex @code{heap-new} usage
 3469: @example
 3470: 50 circle heap-new constant my-circle
 3471: @end example
 3472: 
 3473: @code{heap-new} invokes @code{construct}, thus
 3474: initializing the field @code{circle-radius} with 50. We can draw
 3475: this new circle at (100,100) with
 3476: 
 3477: @example
 3478: 100 100 my-circle draw
 3479: @end example
 3480: 
 3481: @cindex selector invocation, restrictions
 3482: @cindex class definition, restrictions
 3483: Note: You can invoke a selector only if the object on the TOS
 3484: (the receiving object) belongs to the class where the selector was
 3485: defined or one of its descendents; e.g., you can invoke
 3486: @code{draw} only for objects belonging to @code{graphical}
 3487: or its descendents (e.g., @code{circle}).  Immediately before
 3488: @code{end-class}, the search order has to be the same as
 3489: immediately after @code{class}.
 3490: 
 3491: @node The class Object, Creating objects, Basic Objects Usage, Objects
 3492: @subsubsection The class @code{object}
 3493: @cindex @code{object} class
 3494: 
 3495: When you define a class, you have to specify a parent class.  So how do
 3496: you start defining classes? There is one class available from the start:
 3497: @code{object}. You can use it as ancestor for all classes. It is the
 3498: only class that has no parent. It has two selectors: @code{construct}
 3499: and @code{print}.
 3500: 
 3501: @node Creating objects, Object-Oriented Programming Style, The class Object, Objects
 3502: @subsubsection Creating objects
 3503: @cindex creating objects
 3504: @cindex object creation
 3505: @cindex object allocation options
 3506: 
 3507: @cindex @code{heap-new} discussion
 3508: @cindex @code{dict-new} discussion
 3509: @cindex @code{construct} discussion
 3510: You can create and initialize an object of a class on the heap with
 3511: @code{heap-new} ( ... class -- object ) and in the dictionary
 3512: (allocation with @code{allot}) with @code{dict-new} (
 3513: ... class -- object ). Both words invoke @code{construct}, which
 3514: consumes the stack items indicated by "..." above.
 3515: 
 3516: @cindex @code{init-object} discussion
 3517: @cindex @code{class-inst-size} discussion
 3518: If you want to allocate memory for an object yourself, you can get its
 3519: alignment and size with @code{class-inst-size 2@@} ( class --
 3520: align size ). Once you have memory for an object, you can initialize
 3521: it with @code{init-object} ( ... class object -- );
 3522: @code{construct} does only a part of the necessary work.
 3523: 
 3524: @node Object-Oriented Programming Style, Class Binding, Creating objects, Objects
 3525: @subsubsection Object-Oriented Programming Style
 3526: @cindex object-oriented programming style
 3527: 
 3528: This section is not exhaustive.
 3529: 
 3530: @cindex stack effects of selectors
 3531: @cindex selectors and stack effects
 3532: In general, it is a good idea to ensure that all methods for the
 3533: same selector have the same stack effect: when you invoke a selector,
 3534: you often have no idea which method will be invoked, so, unless all
 3535: methods have the same stack effect, you will not know the stack effect
 3536: of the selector invocation.
 3537: 
 3538: One exception to this rule is methods for the selector
 3539: @code{construct}. We know which method is invoked, because we
 3540: specify the class to be constructed at the same place. Actually, I
 3541: defined @code{construct} as a selector only to give the users a
 3542: convenient way to specify initialization. The way it is used, a
 3543: mechanism different from selector invocation would be more natural
 3544: (but probably would take more code and more space to explain).
 3545: 
 3546: @node Class Binding, Method conveniences, Object-Oriented Programming Style, Objects
 3547: @subsubsection Class Binding
 3548: @cindex class binding
 3549: @cindex early binding
 3550: 
 3551: @cindex late binding
 3552: Normal selector invocations determine the method at run-time depending
 3553: on the class of the receiving object (late binding).
 3554: 
 3555: Sometimes we want to invoke a different method. E.g., assume that
 3556: you want to use the simple method for @code{print}ing
 3557: @code{object}s instead of the possibly long-winded
 3558: @code{print} method of the receiver class. You can achieve this
 3559: by replacing the invocation of @code{print} with
 3560: 
 3561: @cindex @code{[bind]} usage
 3562: @example
 3563: [bind] object print
 3564: @end example
 3565: 
 3566: in compiled code or
 3567: 
 3568: @cindex @code{bind} usage
 3569: @example
 3570: bind object print
 3571: @end example
 3572: 
 3573: @cindex class binding, alternative to
 3574: in interpreted code. Alternatively, you can define the method with a
 3575: name (e.g., @code{print-object}), and then invoke it through the
 3576: name. Class binding is just a (often more convenient) way to achieve
 3577: the same effect; it avoids name clutter and allows you to invoke
 3578: methods directly without naming them first.
 3579: 
 3580: @cindex superclass binding
 3581: @cindex parent class binding
 3582: A frequent use of class binding is this: When we define a method
 3583: for a selector, we often want the method to do what the selector does
 3584: in the parent class, and a little more. There is a special word for
 3585: this purpose: @code{[parent]}; @code{[parent]
 3586: @emph{selector}} is equivalent to @code{[bind] @emph{parent
 3587: selector}}, where @code{@emph{parent}} is the parent
 3588: class of the current class. E.g., a method definition might look like:
 3589: 
 3590: @cindex @code{[parent]} usage
 3591: @example
 3592: :noname
 3593:   dup [parent] foo \ do parent's foo on the receiving object
 3594:   ... \ do some more
 3595: ; overrides foo
 3596: @end example
 3597: 
 3598: @cindex class binding as optimization
 3599: In @cite{Object-oriented programming in ANS Forth} (Forth Dimensions,
 3600: March 1997), Andrew McKewan presents class binding as an optimization
 3601: technique. I recommend not using it for this purpose unless you are in
 3602: an emergency. Late binding is pretty fast with this model anyway, so the
 3603: benefit of using class binding is small; the cost of using class binding
 3604: where it is not appropriate is reduced maintainability.
 3605: 
 3606: While we are at programming style questions: You should bind
 3607: selectors only to ancestor classes of the receiving object. E.g., say,
 3608: you know that the receiving object is of class @code{foo} or its
 3609: descendents; then you should bind only to @code{foo} and its
 3610: ancestors.
 3611: 
 3612: @node Method conveniences, Classes and Scoping, Class Binding, Objects
 3613: @subsubsection Method conveniences
 3614: @cindex method conveniences
 3615: 
 3616: In a method you usually access the receiving object pretty often.  If
 3617: you define the method as a plain colon definition (e.g., with
 3618: @code{:noname}), you may have to do a lot of stack
 3619: gymnastics. To avoid this, you can define the method with @code{m:
 3620: ... ;m}. E.g., you could define the method for
 3621: @code{draw}ing a @code{circle} with
 3622: 
 3623: @cindex @code{this} usage
 3624: @cindex @code{m:} usage
 3625: @cindex @code{;m} usage
 3626: @example
 3627: m: ( x y circle -- )
 3628:   ( x y ) this circle-radius @@ draw-circle ;m
 3629: @end example
 3630: 
 3631: @cindex @code{exit} in @code{m: ... ;m}
 3632: @cindex @code{exitm} discussion
 3633: @cindex @code{catch} in @code{m: ... ;m}
 3634: When this method is executed, the receiver object is removed from the
 3635: stack; you can access it with @code{this} (admittedly, in this
 3636: example the use of @code{m: ... ;m} offers no advantage). Note
 3637: that I specify the stack effect for the whole method (i.e. including
 3638: the receiver object), not just for the code between @code{m:}
 3639: and @code{;m}. You cannot use @code{exit} in
 3640: @code{m:...;m}; instead, use
 3641: @code{exitm}.@footnote{Moreover, for any word that calls
 3642: @code{catch} and was defined before loading
 3643: @code{objects.fs}, you have to redefine it like I redefined
 3644: @code{catch}: @code{: catch this >r catch r> to-this ;}}
 3645: 
 3646: @cindex @code{inst-var} usage
 3647: You will frequently use sequences of the form @code{this
 3648: @emph{field}} (in the example above: @code{this
 3649: circle-radius}). If you use the field only in this way, you can
 3650: define it with @code{inst-var} and eliminate the
 3651: @code{this} before the field name. E.g., the @code{circle}
 3652: class above could also be defined with:
 3653: 
 3654: @example
 3655: graphical class
 3656:   cell% inst-var radius
 3657: 
 3658: m: ( x y circle -- )
 3659:   radius @@ draw-circle ;m
 3660: overrides draw
 3661: 
 3662: m: ( n-radius circle -- )
 3663:   radius ! ;m
 3664: overrides construct
 3665: 
 3666: end-class circle
 3667: @end example
 3668: 
 3669: @code{radius} can only be used in @code{circle} and its
 3670: descendent classes and inside @code{m:...;m}.
 3671: 
 3672: @cindex @code{inst-value} usage
 3673: You can also define fields with @code{inst-value}, which is
 3674: to @code{inst-var} what @code{value} is to
 3675: @code{variable}.  You can change the value of such a field with
 3676: @code{[to-inst]}.  E.g., we could also define the class
 3677: @code{circle} like this:
 3678: 
 3679: @example
 3680: graphical class
 3681:   inst-value radius
 3682: 
 3683: m: ( x y circle -- )
 3684:   radius draw-circle ;m
 3685: overrides draw
 3686: 
 3687: m: ( n-radius circle -- )
 3688:   [to-inst] radius ;m
 3689: overrides construct
 3690: 
 3691: end-class circle
 3692: @end example
 3693: 
 3694: 
 3695: @node Classes and Scoping, Object Interfaces, Method conveniences, Objects
 3696: @subsubsection Classes and Scoping
 3697: @cindex classes and scoping
 3698: @cindex scoping and classes
 3699: 
 3700: Inheritance is frequent, unlike structure extension. This exacerbates
 3701: the problem with the field name convention (@pxref{Structure Naming
 3702: Convention}): One always has to remember in which class the field was
 3703: originally defined; changing a part of the class structure would require
 3704: changes for renaming in otherwise unaffected code.
 3705: 
 3706: @cindex @code{inst-var} visibility
 3707: @cindex @code{inst-value} visibility
 3708: To solve this problem, I added a scoping mechanism (which was not in my
 3709: original charter): A field defined with @code{inst-var} (or
 3710: @code{inst-value}) is visible only in the class where it is defined and in
 3711: the descendent classes of this class.  Using such fields only makes
 3712: sense in @code{m:}-defined methods in these classes anyway.
 3713: 
 3714: This scoping mechanism allows us to use the unadorned field name,
 3715: because name clashes with unrelated words become much less likely.
 3716: 
 3717: @cindex @code{protected} discussion
 3718: @cindex @code{private} discussion
 3719: Once we have this mechanism, we can also use it for controlling the
 3720: visibility of other words: All words defined after
 3721: @code{protected} are visible only in the current class and its
 3722: descendents. @code{public} restores the compilation
 3723: (i.e. @code{current}) wordlist that was in effect before. If you
 3724: have several @code{protected}s without an intervening
 3725: @code{public} or @code{set-current}, @code{public}
 3726: will restore the compilation wordlist in effect before the first of
 3727: these @code{protected}s.
 3728: 
 3729: @node Object Interfaces, Objects Implementation, Classes and Scoping, Objects
 3730: @subsubsection Object Interfaces
 3731: @cindex object interfaces
 3732: @cindex interfaces for objects
 3733: 
 3734: In this model you can only call selectors defined in the class of the
 3735: receiving objects or in one of its ancestors. If you call a selector
 3736: with a receiving object that is not in one of these classes, the
 3737: result is undefined; if you are lucky, the program crashes
 3738: immediately.
 3739: 
 3740: @cindex selectors common to hardly-related classes
 3741: Now consider the case when you want to have a selector (or several)
 3742: available in two classes: You would have to add the selector to a
 3743: common ancestor class, in the worst case to @code{object}. You
 3744: may not want to do this, e.g., because someone else is responsible for
 3745: this ancestor class.
 3746: 
 3747: The solution for this problem is interfaces. An interface is a
 3748: collection of selectors. If a class implements an interface, the
 3749: selectors become available to the class and its descendents. A class
 3750: can implement an unlimited number of interfaces. For the problem
 3751: discussed above, we would define an interface for the selector(s), and
 3752: both classes would implement the interface.
 3753: 
 3754: As an example, consider an interface @code{storage} for
 3755: writing objects to disk and getting them back, and a class
 3756: @code{foo} foo that implements it. The code for this would look
 3757: like this:
 3758: 
 3759: @cindex @code{interface} usage
 3760: @cindex @code{end-interface} usage
 3761: @cindex @code{implementation} usage
 3762: @example
 3763: interface
 3764:   selector write ( file object -- )
 3765:   selector read1 ( file object -- )
 3766: end-interface storage
 3767: 
 3768: bar class
 3769:   storage implementation
 3770: 
 3771: ... overrides write
 3772: ... overrides read
 3773: ...
 3774: end-class foo
 3775: @end example
 3776: 
 3777: (I would add a word @code{read} ( file -- object ) that uses
 3778: @code{read1} internally, but that's beyond the point illustrated
 3779: here.)
 3780: 
 3781: Note that you cannot use @code{protected} in an interface; and
 3782: of course you cannot define fields.
 3783: 
 3784: In the Neon model, all selectors are available for all classes;
 3785: therefore it does not need interfaces. The price you pay in this model
 3786: is slower late binding, and therefore, added complexity to avoid late
 3787: binding.
 3788: 
 3789: @node Objects Implementation, Comparison with other object models, Object Interfaces, Objects
 3790: @subsubsection @file{objects.fs} Implementation
 3791: @cindex @file{objects.fs} implementation
 3792: 
 3793: @cindex @code{object-map} discussion
 3794: An object is a piece of memory, like one of the data structures
 3795: described with @code{struct...end-struct}. It has a field
 3796: @code{object-map} that points to the method map for the object's
 3797: class.
 3798: 
 3799: @cindex method map
 3800: @cindex virtual function table
 3801: The @emph{method map}@footnote{This is Self terminology; in C++
 3802: terminology: virtual function table.} is an array that contains the
 3803: execution tokens (XTs) of the methods for the object's class. Each
 3804: selector contains an offset into the method maps.
 3805: 
 3806: @cindex @code{selector} implementation, class
 3807: @code{selector} is a defining word that uses
 3808: @code{create} and @code{does>}. The body of the
 3809: selector contains the offset; the @code{does>} action for a
 3810: class selector is, basically:
 3811: 
 3812: @example
 3813: ( object addr ) @@ over object-map @@ + @@ execute
 3814: @end example
 3815: 
 3816: Since @code{object-map} is the first field of the object, it
 3817: does not generate any code. As you can see, calling a selector has a
 3818: small, constant cost.
 3819: 
 3820: @cindex @code{current-interface} discussion
 3821: @cindex class implementation and representation
 3822: A class is basically a @code{struct} combined with a method
 3823: map. During the class definition the alignment and size of the class
 3824: are passed on the stack, just as with @code{struct}s, so
 3825: @code{field} can also be used for defining class
 3826: fields. However, passing more items on the stack would be
 3827: inconvenient, so @code{class} builds a data structure in memory,
 3828: which is accessed through the variable
 3829: @code{current-interface}. After its definition is complete, the
 3830: class is represented on the stack by a pointer (e.g., as parameter for
 3831: a child class definition).
 3832: 
 3833: At the start, a new class has the alignment and size of its parent,
 3834: and a copy of the parent's method map. Defining new fields extends the
 3835: size and alignment; likewise, defining new selectors extends the
 3836: method map. @code{overrides} just stores a new XT in the method
 3837: map at the offset given by the selector.
 3838: 
 3839: @cindex class binding, implementation
 3840: Class binding just gets the XT at the offset given by the selector
 3841: from the class's method map and @code{compile,}s (in the case of
 3842: @code{[bind]}) it.
 3843: 
 3844: @cindex @code{this} implementation
 3845: @cindex @code{catch} and @code{this}
 3846: @cindex @code{this} and @code{catch}
 3847: I implemented @code{this} as a @code{value}. At the
 3848: start of an @code{m:...;m} method the old @code{this} is
 3849: stored to the return stack and restored at the end; and the object on
 3850: the TOS is stored @code{TO this}. This technique has one
 3851: disadvantage: If the user does not leave the method via
 3852: @code{;m}, but via @code{throw} or @code{exit},
 3853: @code{this} is not restored (and @code{exit} may
 3854: crash). To deal with the @code{throw} problem, I have redefined
 3855: @code{catch} to save and restore @code{this}; the same
 3856: should be done with any word that can catch an exception. As for
 3857: @code{exit}, I simply forbid it (as a replacement, there is
 3858: @code{exitm}).
 3859: 
 3860: @cindex @code{inst-var} implementation
 3861: @code{inst-var} is just the same as @code{field}, with
 3862: a different @code{does>} action:
 3863: @example
 3864: @@ this +
 3865: @end example
 3866: Similar for @code{inst-value}.
 3867: 
 3868: @cindex class scoping implementation
 3869: Each class also has a wordlist that contains the words defined with
 3870: @code{inst-var} and @code{inst-value}, and its protected
 3871: words. It also has a pointer to its parent. @code{class} pushes
 3872: the wordlists of the class an all its ancestors on the search order,
 3873: and @code{end-class} drops them.
 3874: 
 3875: @cindex interface implementation
 3876: An interface is like a class without fields, parent and protected
 3877: words; i.e., it just has a method map. If a class implements an
 3878: interface, its method map contains a pointer to the method map of the
 3879: interface. The positive offsets in the map are reserved for class
 3880: methods, therefore interface map pointers have negative
 3881: offsets. Interfaces have offsets that are unique throughout the
 3882: system, unlike class selectors, whose offsets are only unique for the
 3883: classes where the selector is available (invokable).
 3884: 
 3885: This structure means that interface selectors have to perform one
 3886: indirection more than class selectors to find their method. Their body
 3887: contains the interface map pointer offset in the class method map, and
 3888: the method offset in the interface method map. The
 3889: @code{does>} action for an interface selector is, basically:
 3890: 
 3891: @example
 3892: ( object selector-body )
 3893: 2dup selector-interface @@ ( object selector-body object interface-offset )
 3894: swap object-map @@ + @@ ( object selector-body map )
 3895: swap selector-offset @@ + @@ execute
 3896: @end example
 3897: 
 3898: where @code{object-map} and @code{selector-offset} are
 3899: first fields and generate no code.
 3900: 
 3901: As a concrete example, consider the following code:
 3902: 
 3903: @example
 3904: interface
 3905:   selector if1sel1
 3906:   selector if1sel2
 3907: end-interface if1
 3908: 
 3909: object class
 3910:   if1 implementation
 3911:   selector cl1sel1
 3912:   cell% inst-var cl1iv1
 3913: 
 3914: ' m1 overrides construct
 3915: ' m2 overrides if1sel1
 3916: ' m3 overrides if1sel2
 3917: ' m4 overrides cl1sel2
 3918: end-class cl1
 3919: 
 3920: create obj1 object dict-new drop
 3921: create obj2 cl1    dict-new drop
 3922: @end example
 3923: 
 3924: The data structure created by this code (including the data structure
 3925: for @code{object}) is shown in the <a
 3926: href="objects-implementation.eps">figure</a>, assuming a cell size of 4.
 3927: 
 3928: @node Comparison with other object models, Objects Glossary, Objects Implementation, Objects
 3929: @subsubsection Comparison with other object models
 3930: @cindex comparison of object models
 3931: @cindex object models, comparison
 3932: 
 3933: Many object-oriented Forth extensions have been proposed (@cite{A survey
 3934: of object-oriented Forths} (SIGPLAN Notices, April 1996) by Bradford
 3935: J. Rodriguez and W. F. S. Poehlman lists 17). Here I'll discuss the
 3936: relation of @file{objects.fs} to two well-known and two closely-related
 3937: (by the use of method maps) models.
 3938: 
 3939: @cindex Neon model
 3940: The most popular model currently seems to be the Neon model (see
 3941: @cite{Object-oriented programming in ANS Forth} (Forth Dimensions, March
 3942: 1997) by Andrew McKewan). The Neon model uses a @code{@emph{selector
 3943: object}} syntax, which makes it unnatural to pass objects on the
 3944: stack. It also requires that the selector parses the input stream (at
 3945: compile time); this leads to reduced extensibility and to bugs that are
 3946: hard to find. Finally, it allows using every selector to every object;
 3947: this eliminates the need for classes, but makes it harder to create
 3948: efficient implementations. A longer version of this critique can be
 3949: found in @cite{On Standardizing Object-Oriented Forth Extensions} (Forth
 3950: Dimensions, May 1997) by Anton Ertl.
 3951: 
 3952: @cindex Pountain's object-oriented model
 3953: Another well-known publication is @cite{Object-Oriented Forth} (Academic
 3954: Press, London, 1987) by Dick Pountain. However, it is not really about
 3955: object-oriented programming, because it hardly deals with late
 3956: binding. Instead, it focuses on features like information hiding and
 3957: overloading that are characteristic of modular languages like Ada (83).
 3958: 
 3959: @cindex Zsoter's object-oriented model
 3960: In @cite{Does late binding have to be slow?} (Forth Dimensions ??? 1996)
 3961: Andras Zsoter describes a model that makes heavy use of an active object
 3962: (like @code{this} in @file{objects.fs}): The active object is not only
 3963: used for accessing all fields, but also specifies the receiving object
 3964: of every selector invocation; you have to change the active object
 3965: explicitly with @code{@{ ... @}}, whereas in @file{objects.fs} it
 3966: changes more or less implicitly at @code{m: ... ;m}. Such a change at
 3967: the method entry point is unnecessary with the Zsoter's model, because
 3968: the receiving object is the active object already; OTOH, the explicit
 3969: change is absolutely necessary in that model, because otherwise no one
 3970: could ever change the active object. An ANS Forth implementation of this
 3971: model is available at @url{http://www.forth.org/fig/oopf.html}.
 3972: 
 3973: @cindex @file{oof.fs}, differences to other models
 3974: The @file{oof.fs} model combines information hiding and overloading
 3975: resolution (by keeping names in various wordlists) with object-oriented
 3976: programming. It sets the active object implicitly on method entry, but
 3977: also allows explicit changing (with @code{>o...o>} or with
 3978: @code{with...endwith}). It uses parsing and state-smart objects and
 3979: classes for resolving overloading and for early binding: the object or
 3980: class parses the selector and determines the method from this. If the
 3981: selector is not parsed by an object or class, it performs a call to the
 3982: selector for the active object (late binding), like Zsoter's model.
 3983: Fields are always accessed through the active object. The big
 3984: disadvantage of this model is the parsing and the state-smartness, which
 3985: reduces extensibility and increases the opportunities for subtle bugs;
 3986: essentially, you are only safe if you never tick or @code{postpone} an
 3987: object or class (Bernd disagrees, but I (Anton) am not convinced).
 3988: 
 3989: @cindex @file{mini-oof.fs}, differences to other models
 3990: The Mini-OOF model is quite similar to a very stripped-down version of
 3991: the Objects model, but syntactically it is a mixture of the Objects and
 3992: the OOF model.
 3993: 
 3994: 
 3995: @node Objects Glossary,  , Comparison with other object models, Objects
 3996: @subsubsection @file{objects.fs} Glossary
 3997: @cindex @file{objects.fs} Glossary
 3998: 
 3999: doc---objects-bind
 4000: doc---objects-<bind>
 4001: doc---objects-bind'
 4002: doc---objects-[bind]
 4003: doc---objects-class
 4004: doc---objects-class->map
 4005: doc---objects-class-inst-size
 4006: doc---objects-class-override!
 4007: doc---objects-construct
 4008: doc---objects-current'
 4009: doc---objects-[current]
 4010: doc---objects-current-interface
 4011: doc---objects-dict-new
 4012: doc---objects-drop-order
 4013: doc---objects-end-class
 4014: doc---objects-end-class-noname
 4015: doc---objects-end-interface
 4016: doc---objects-end-interface-noname
 4017: doc---objects-exitm
 4018: doc---objects-heap-new
 4019: doc---objects-implementation
 4020: doc---objects-init-object
 4021: doc---objects-inst-value
 4022: doc---objects-inst-var
 4023: doc---objects-interface
 4024: doc---objects-;m
 4025: doc---objects-m:
 4026: doc---objects-method
 4027: doc---objects-object
 4028: doc---objects-overrides
 4029: doc---objects-[parent]
 4030: doc---objects-print
 4031: doc---objects-protected
 4032: doc---objects-public
 4033: doc---objects-push-order
 4034: doc---objects-selector
 4035: doc---objects-this
 4036: doc---objects-<to-inst>
 4037: doc---objects-[to-inst]
 4038: doc---objects-to-this
 4039: doc---objects-xt-new
 4040: 
 4041: @c -------------------------------------------------------------
 4042: @node OOF, Mini-OOF, Objects, Object-oriented Forth
 4043: @subsection OOF
 4044: @cindex oof
 4045: @cindex object-oriented programming
 4046: 
 4047: @cindex @file{objects.fs}
 4048: @cindex @file{oof.fs}
 4049: 
 4050: This section describes the @file{oof.fs} packet. This section uses the
 4051: same rationale why using object-oriented programming, and the same
 4052: terminology.
 4053: 
 4054: The packet described in this section is used in bigFORTH since 1991, and
 4055: used for two large applications: a chromatographic system used to
 4056: create new medicaments, and a graphic user interface library (MINOS).
 4057: 
 4058: You can find a description (in German) of @file{oof.fs} in @cite{Object
 4059: oriented bigFORTH} by Bernd Paysan, published in @cite{Vierte Dimension}
 4060: 10(2), 1994.
 4061: 
 4062: @menu
 4063: * Properties of the OOF model::
 4064: * Basic OOF Usage::
 4065: * The base class object::
 4066: * Class Declaration::
 4067: * Class Implementation::
 4068: @end menu
 4069: 
 4070: @node Properties of the OOF model, Basic OOF Usage, OOF, OOF
 4071: @subsubsection Properties of the OOF model
 4072: @cindex @file{oof.fs} properties
 4073: 
 4074: @itemize @bullet
 4075: @item
 4076: This model combines object oriented programming with information
 4077: hiding. It helps you writing large application, where scoping is
 4078: necessary, because it provides class-oriented scoping.
 4079: 
 4080: @item
 4081: Named objects, object pointers, and object arrays can be created,
 4082: selector invocation uses the "object selector" syntax. Selector invocation
 4083: to objects and/or selectors on the stack is a bit less convenient, but
 4084: possible.
 4085: 
 4086: @item
 4087: Selector invocation and instance variable usage of the active object is
 4088: straight forward, since both make use of the active object.
 4089: 
 4090: @item
 4091: Late binding is efficient and easy to use.
 4092: 
 4093: @item
 4094: State-smart objects parse selectors. However, extensibility is provided
 4095: using a (parsing) selector @code{postpone} and a selector @code{'}.
 4096: 
 4097: @item
 4098: An implementation in ANS Forth is available.
 4099: 
 4100: @end itemize
 4101: 
 4102: 
 4103: @node Basic OOF Usage, The base class object, Properties of the OOF model, OOF
 4104: @subsubsection Basic OOF Usage
 4105: @cindex @file{oof.fs} usage
 4106: 
 4107: Here, I use the same example as for @code{objects} (@pxref{Basic Objects Usage}).
 4108: 
 4109: You can define a class for graphical objects like this:
 4110: 
 4111: @cindex @code{class} usage
 4112: @cindex @code{class;} usage
 4113: @cindex @code{method} usage
 4114: @example
 4115: object class graphical \ "object" is the parent class
 4116:   method draw ( x y graphical -- )
 4117: class;
 4118: @end example
 4119: 
 4120: This code defines a class @code{graphical} with an
 4121: operation @code{draw}.  We can perform the operation
 4122: @code{draw} on any @code{graphical} object, e.g.:
 4123: 
 4124: @example
 4125: 100 100 t-rex draw
 4126: @end example
 4127: 
 4128: where @code{t-rex} is an object or object pointer, created with e.g.
 4129: @code{graphical : t-rex}.
 4130: 
 4131: @cindex abstract class
 4132: How do we create a graphical object? With the present definitions,
 4133: we cannot create a useful graphical object. The class
 4134: @code{graphical} describes graphical objects in general, but not
 4135: any concrete graphical object type (C++ users would call it an
 4136: @emph{abstract class}); e.g., there is no method for the selector
 4137: @code{draw} in the class @code{graphical}.
 4138: 
 4139: For concrete graphical objects, we define child classes of the
 4140: class @code{graphical}, e.g.:
 4141: 
 4142: @example
 4143: graphical class circle \ "graphical" is the parent class
 4144:   cell var circle-radius
 4145: how:
 4146:   : draw ( x y -- )
 4147:     circle-radius @@ draw-circle ;
 4148: 
 4149:   : init ( n-radius -- (
 4150:     circle-radius ! ;
 4151: class;
 4152: @end example
 4153: 
 4154: Here we define a class @code{circle} as a child of @code{graphical},
 4155: with a field @code{circle-radius}; it defines new methods for the
 4156: selectors @code{draw} and @code{init} (@code{init} is defined in
 4157: @code{object}, the parent class of @code{graphical}).
 4158: 
 4159: Now we can create a circle in the dictionary with
 4160: 
 4161: @example
 4162: 50 circle : my-circle
 4163: @end example
 4164: 
 4165: @code{:} invokes @code{init}, thus initializing the field
 4166: @code{circle-radius} with 50. We can draw this new circle at (100,100)
 4167: with
 4168: 
 4169: @example
 4170: 100 100 my-circle draw
 4171: @end example
 4172: 
 4173: @cindex selector invocation, restrictions
 4174: @cindex class definition, restrictions
 4175: Note: You can invoke a selector only if the receiving object belongs to
 4176: the class where the selector was defined or one of its descendents;
 4177: e.g., you can invoke @code{draw} only for objects belonging to
 4178: @code{graphical} or its descendents (e.g., @code{circle}). The scoping
 4179: mechanism will check if you try to invoke a selector that is not
 4180: defined in this class hierarchy, so you'll get an error at compilation
 4181: time.
 4182: 
 4183: 
 4184: @node The base class object, Class Declaration, Basic OOF Usage, OOF
 4185: @subsubsection The base class @file{object}
 4186: @cindex @file{oof.fs} base class
 4187: 
 4188: When you define a class, you have to specify a parent class.  So how do
 4189: you start defining classes? There is one class available from the start:
 4190: @code{object}. You have to use it as ancestor for all classes. It is the
 4191: only class that has no parent. Classes are also objects, except that
 4192: they don't have instance variables; class manipulation such as
 4193: inheritance or changing definitions of a class is handled through
 4194: selectors of the class @code{object}.
 4195: 
 4196: @code{object} provides a number of selectors:
 4197: 
 4198: @itemize @bullet
 4199: @item
 4200: @code{class} for subclassing, @code{definitions} to add definitions
 4201: later on, and @code{class?} to get type informations (is the class a
 4202: subclass of the class passed on the stack?).
 4203: doc---object-class
 4204: doc---object-definitions
 4205: doc---object-class?
 4206: 
 4207: @item
 4208: @code{init} and @code{dispose} as constructor and destroctor of the
 4209: object. @code{init} is invocated after the object's memory is allocated,
 4210: while @code{dispose} also handles deallocation. Thus if you redefine
 4211: @code{dispose}, you have to call the parent's dispose with @code{super
 4212: dispose}, too.
 4213: doc---object-init
 4214: doc---object-dispose
 4215: 
 4216: @item
 4217: @code{new}, @code{new[]}, @code{:}, @code{ptr}, @code{asptr}, and
 4218: @code{[]} to create named and unnamed objects and object arrays or
 4219: object pointers.
 4220: doc---object-new
 4221: doc---object-new[]
 4222: doc---object-:
 4223: doc---object-ptr
 4224: doc---object-asptr
 4225: doc---object-[]
 4226: 
 4227: @item
 4228: @code{::} and @code{super} for expicit scoping. You should use expicit
 4229: scoping only for super classes or classes with the same set of instance
 4230: variables. Explicit scoped selectors use early binding.
 4231: doc---object-::
 4232: doc---object-super
 4233: 
 4234: @item
 4235: @code{self} to get the address of the object
 4236: doc---object-self
 4237: 
 4238: @item
 4239: @code{bind}, @code{bound}, @code{link}, and @code{is} to assign object
 4240: pointers and instance defers.
 4241: doc---object-bind
 4242: doc---object-bound
 4243: doc---object-link
 4244: doc---object-is
 4245: 
 4246: @item
 4247: @code{'} to obtain selector tokens, @code{send} to invocate selectors
 4248: form the stack, and @code{postpone} to generate selector invocation code.
 4249: doc---object-'
 4250: doc---object-postpone
 4251: 
 4252: @item
 4253: @code{with} and @code{endwith} to select the active object from the
 4254: stack, and enabling it's scope. Using @code{with} and @code{endwith}
 4255: also allows to create code using selector @code{postpone} without being
 4256: trapped bye the state-smart objects.
 4257: doc---object-with
 4258: doc---object-endwith
 4259: 
 4260: @end itemize
 4261: 
 4262: @node Class Declaration, Class Implementation, The base class object, OOF
 4263: @subsubsection Class Declaration
 4264: @cindex class declaration
 4265: 
 4266: @itemize @bullet
 4267: @item
 4268: Instance variables
 4269: doc---oof-var
 4270: 
 4271: @item
 4272: Object pointers
 4273: doc---oof-ptr
 4274: doc---oof-asptr
 4275: 
 4276: @item
 4277: Instance defers
 4278: doc---oof-defer
 4279: 
 4280: @item
 4281: Method selectors
 4282: doc---oof-early
 4283: doc---oof-method
 4284: 
 4285: @item
 4286: Class wide variables
 4287: doc---oof-static
 4288: 
 4289: @item
 4290: End declaration
 4291: doc---oof-how:
 4292: doc---oof-class;
 4293: 
 4294: @end itemize
 4295: 
 4296: @c -------------------------------------------------------------
 4297: @node Class Implementation,  , Class Declaration, OOF
 4298: @subsubsection Class Implementation
 4299: @cindex class implementation
 4300: 
 4301: @c -------------------------------------------------------------
 4302: @node Mini-OOF, , OOF, Object-oriented Forth
 4303: @subsection Mini-OOF
 4304: @cindex mini-oof
 4305: 
 4306: Gforth's third object oriented Forth package is a 12-liner. It uses a
 4307: bit of a mixture of the @file{object.fs} and the @file{oof.fs} syntax,
 4308: and reduces to the bare minimum of features. This is based on a posting
 4309: of Bernd Paysan in comp.arch.
 4310: 
 4311: @menu
 4312: * Mini-OOF Usage::
 4313: * Mini-OOF Example::
 4314: * Mini-OOF Implementation::
 4315: @end menu
 4316: 
 4317: @c -------------------------------------------------------------
 4318: @node Mini-OOF Usage, Mini-OOF Example, , Mini-OOF
 4319: @subsubsection Usage
 4320: @cindex mini-oof usage
 4321: 
 4322: Basically, there are seven words, to define a method, a variable, a
 4323: class; to end a class, to define a method, to allocate an object, to
 4324: resolve binding, and the base class (which allocates one cell for the
 4325: object pointer).
 4326: 
 4327: doc-method
 4328: 
 4329: Defines a method
 4330: 
 4331: doc-var
 4332: 
 4333: Defines a variable with size bytes
 4334: 
 4335: doc-class
 4336: 
 4337: Starts the definition of a sub-class
 4338: 
 4339: doc-end-class
 4340: 
 4341: Ends the definition of a class
 4342: 
 4343: doc-defines
 4344: 
 4345: Binds the xt to the method name in the class
 4346: 
 4347: doc-new
 4348: 
 4349: Creates a new incarnation of the class
 4350: 
 4351: doc-::
 4352: 
 4353: Compiles the method name of the class (not immediate!)
 4354: 
 4355: doc-object
 4356: 
 4357: Is the base class of all objects
 4358: 
 4359: @c -------------------------------------------------------------
 4360: @node Mini-OOF Example, Mini-OOF Implementation, Mini-OOF Usage, Mini-OOF
 4361: @subsubsection Mini-OOF Example
 4362: @cindex mini-oof example
 4363: 
 4364: A short example shows how to use this package.
 4365: 
 4366: @example
 4367: object class
 4368:   method init
 4369:   method draw
 4370: end-class graphical
 4371: @end example
 4372: 
 4373: This code defines a class @code{graphical} with an
 4374: operation @code{draw}.  We can perform the operation
 4375: @code{draw} on any @code{graphical} object, e.g.:
 4376: 
 4377: @example
 4378: 100 100 t-rex draw
 4379: @end example
 4380: 
 4381: where @code{t-rex} is an object or object pointer, created with e.g.
 4382: @code{graphical new Constant t-rex}.
 4383: 
 4384: For concrete graphical objects, we define child classes of the
 4385: class @code{graphical}, e.g.:
 4386: 
 4387: @example
 4388: graphical class
 4389:   cell var circle-radius
 4390: end-class circle \ "graphical" is the parent class
 4391: 
 4392: :noname ( x y -- )
 4393:   circle-radius @@ draw-circle ; circle defines draw
 4394: :noname ( r -- )
 4395:   circle-radius ! ; circle defines init
 4396: @end example
 4397: 
 4398: There is no implicit init method, so we have to define one. The creation
 4399: code of the object now has to call init explicitely.
 4400: 
 4401: @example
 4402: circle new Constant my-circle
 4403: 50 my-circle init
 4404: @end example
 4405: 
 4406: It is also possible to add a function to create named objects with
 4407: automatic call of @code{init}, given that all objects have @code{init}
 4408: on the same place
 4409: 
 4410: @example
 4411: : new: ( .. o "name" -- )
 4412:     new dup Constant init ;
 4413: 80 circle new: large-circle
 4414: @end example
 4415: 
 4416: We can draw this new circle at (100,100)
 4417: with
 4418: 
 4419: @example
 4420: 100 100 my-circle draw
 4421: @end example
 4422: 
 4423: @node Mini-OOF Implementation, , Mini-OOF Example, Mini-OOF
 4424: @subsubsection Mini-OOF Implementation
 4425: 
 4426: Object oriented system with late binding typically use a
 4427: "vtable"-approach: the first variable in each object is a pointer to a
 4428: table, which contains the methods as function pointers. This vtable
 4429: may contain some other informations, too.
 4430: 
 4431: So first, let's declare methods:
 4432: 
 4433: @example
 4434: : method ( m v -- m' v ) Create  over , swap cell+ swap
 4435:   DOES> ( ... o -- ... ) @ over @ + @ execute ;
 4436: @end example
 4437: 
 4438: During method declaration, the number of methods and instance
 4439: variables is on the stack (in address units). @code{method} creates
 4440: one method and increments the method number. To execute a method, it
 4441: takes the object, fetches the vtable pointer, adds the offset, and
 4442: executes the xt stored there. Each method takes the object it is
 4443: invoked from as top of stack parameter. The method itself should
 4444: consume that object.
 4445: 
 4446: Now, we also have to declare instance variables
 4447: 
 4448: @example
 4449: : var ( m v size -- m v' ) Create  over , +
 4450:   DOES> ( o -- addr ) @ + ;
 4451: @end example
 4452: 
 4453: Same as above, a word is created with the current offset. Instance
 4454: variables can have different sizes (cells, floats, doubles, chars), so
 4455: all we do is take the size and add it to the offset. If your machine
 4456: has alignment restrictions, put the proper @code{aligned} or
 4457: @code{faligned} before the variable, it will adjust the variable
 4458: offset. That's why it is on the top of stack.
 4459: 
 4460: We need a starting point (the empty object) and some syntactic sugar:
 4461: 
 4462: @example
 4463: Create object  1 cells , 2 cells ,
 4464: : class ( class -- class methods vars ) dup 2@ ;
 4465: @end example
 4466: 
 4467: Now, for inheritance, the vtable of the parent object has to be
 4468: copied, when a new, derived class is declared. This gives all the
 4469: methods of the parent class, which can be overridden, though.
 4470: 
 4471: @example
 4472: : end-class  ( class methods vars -- )
 4473:   Create  here >r , dup , 2 cells ?DO ['] noop , 1 cells +LOOP
 4474:   cell+ dup cell+ r> rot @ 2 cells /string move ;
 4475: @end example
 4476: 
 4477: The first line creates the vtable, initialized with
 4478: @code{noop}s. The second line is the inheritance mechanism, it
 4479: copies the xts from the parent vtable.
 4480: 
 4481: We still have no way to define new methods, let's do that now:
 4482: 
 4483: @example
 4484: : defines ( xt class -- ) ' >body @ + ! ;
 4485: @end example
 4486: 
 4487: To allocate a new object, we need a word, too:
 4488: 
 4489: @example
 4490: : new ( class -- o )  here over @ allot swap over ! ;
 4491: @end example
 4492: 
 4493: And sometimes derived classes want to access the method of the
 4494: parent object. There are two ways to achieve this with this OOF:
 4495: first, you could use named words, and second, you could look up the
 4496: vtable of the parent object.
 4497: 
 4498: @example
 4499: : :: ( class "name" -- ) ' >body @ + @ compile, ;
 4500: @end example
 4501: 
 4502: <H2>An Example</H2>
 4503: 
 4504: Nothing can be more confusing than a good example, so here is
 4505: one. First let's declare a text object (further called
 4506: @code{button}), that stores text and position:
 4507: 
 4508: @example
 4509: object class
 4510:   cell var text
 4511:   cell var len
 4512:   cell var x
 4513:   cell var y
 4514:   method init
 4515:   method draw
 4516: end-class button
 4517: @end example
 4518: 
 4519: Now, implement the two methods, @code{draw} and @code{init}:
 4520: 
 4521: @example
 4522: :noname ( o -- ) >r
 4523:  r@ x @ r@ y @ at-xy  r@ text @ r> len @ type ;
 4524:  button defines draw
 4525: :noname ( addr u o -- ) >r
 4526:  0 r@ x ! 0 r@ y ! r@ len ! r> text ! ;
 4527:  button defines init
 4528: @end example
 4529: 
 4530: For inheritance, we define a class @code{bold-button}, with no
 4531: new data and no new methods.
 4532: 
 4533: @example
 4534: button class
 4535: end-class bold-button
 4536: 
 4537: : bold   27 emit ." [1m" ;
 4538: : normal 27 emit ." [0m" ;
 4539: 
 4540: :noname bold [ button :: draw ] normal ; bold-button defines draw
 4541: @end example
 4542: 
 4543: And finally, some code to demonstrate how to create objects and
 4544: apply methods:
 4545: 
 4546: @example
 4547: button new Constant foo
 4548: s" thin foo" foo init
 4549: page
 4550: foo draw
 4551: bold-button new Constant bar
 4552: s" fat bar" bar init
 4553: 1 bar y !
 4554: bar draw
 4555: @end example
 4556: 
 4557: @c -------------------------------------------------------------
 4558: @node Tokens for Words, Wordlists, Object-oriented Forth, Words
 4559: @section Tokens for Words
 4560: @cindex tokens for words
 4561: 
 4562: This chapter describes the creation and use of tokens that represent
 4563: words on the stack (and in data space).
 4564: 
 4565: Named words have interpretation and compilation semantics. Unnamed words
 4566: just have execution semantics.
 4567: 
 4568: @cindex execution token
 4569: An @dfn{execution token} represents the execution semantics of an
 4570: unnamed word. An execution token occupies one cell. As explained in
 4571: section @ref{Supplying names}, the execution token of the last words
 4572: defined can be produced with
 4573: 
 4574: short-lastxt
 4575: 
 4576: You can perform the semantics represented by an execution token with
 4577: doc-execute
 4578: You can compile the word with
 4579: doc-compile,
 4580: 
 4581: @cindex code field address
 4582: @cindex CFA
 4583: In Gforth, the abstract data type @emph{execution token} is implemented
 4584: as CFA (code field address).
 4585: 
 4586: The interpretation semantics of a named word are also represented by an
 4587: execution token. You can get it with
 4588: 
 4589: doc-[']
 4590: doc-'
 4591: 
 4592: For literals, you use @code{'} in interpreted code and @code{[']} in
 4593: compiled code. Gforth's @code{'} and @code{[']} behave somewhat unusual
 4594: by complaining about compile-only words. To get an execution token for a
 4595: compiling word @var{X}, use @code{COMP' @var{X} drop} or @code{[COMP']
 4596: @var{X} drop}.
 4597: 
 4598: @cindex compilation token
 4599: The compilation semantics are represented by a @dfn{compilation token}
 4600: consisting of two cells: @var{w xt}. The top cell @var{xt} is an
 4601: execution token. The compilation semantics represented by the
 4602: compilation token can be performed with @code{execute}, which consumes
 4603: the whole compilation token, with an additional stack effect determined
 4604: by the represented compilation semantics.
 4605: 
 4606: doc-[comp']
 4607: doc-comp'
 4608: 
 4609: You can compile the compilation semantics with @code{postpone,}. I.e.,
 4610: @code{COMP' @var{word} POSTPONE,} is equivalent to @code{POSTPONE
 4611: @var{word}}.
 4612: 
 4613: doc-postpone,
 4614: 
 4615: At present, the @var{w} part of a compilation token is an execution
 4616: token, and the @var{xt} part represents either @code{execute} or
 4617: @code{compile,}. However, don't rely on that knowledge, unless necessary;
 4618: we may introduce unusual compilation tokens in the future (e.g.,
 4619: compilation tokens representing the compilation semantics of literals).
 4620: 
 4621: @cindex name token
 4622: @cindex name field address
 4623: @cindex NFA
 4624: Named words are also represented by the @dfn{name token}. The abstract
 4625: data type @emph{name token} is implemented as NFA (name field address).
 4626: 
 4627: doc-find-name
 4628: doc-name>int
 4629: doc-name?int
 4630: doc-name>comp
 4631: doc-name>string
 4632: 
 4633: @node Wordlists, Files, Tokens for Words, Words
 4634: @section Wordlists
 4635: 
 4636: @node Files, Including Files, Wordlists, Words
 4637: @section Files
 4638: 
 4639: This chapter describes how to operate on files from Forth.
 4640: 
 4641: Files have the following types for opening and creating:
 4642: 
 4643: doc-r/o
 4644: doc-r/w
 4645: doc-w/o
 4646: doc-bin
 4647: 
 4648: Files are opened/created by name and type, and return a file
 4649: identifier.
 4650: 
 4651: doc-open-file
 4652: doc-create-file
 4653: 
 4654: This identifier is used for all other file commands.
 4655: 
 4656: doc-close-file
 4657: doc-delete-file
 4658: doc-rename-file
 4659: doc-read-file
 4660: doc-read-line
 4661: doc-write-file
 4662: doc-emit-file
 4663: doc-flush-file
 4664: 
 4665: doc-file-status
 4666: doc-file-position
 4667: doc-reposition-file
 4668: doc-file-size
 4669: doc-resize-file
 4670: 
 4671: @node Including Files, Blocks, Files, Words
 4672: @section Including Files
 4673: @cindex including files
 4674: 
 4675: @menu
 4676: * Words for Including::         
 4677: * Search Path::                 
 4678: * Changing the Search Path::    
 4679: * General Search Paths::        
 4680: @end menu
 4681: 
 4682: @node Words for Including, Search Path, Including Files, Including Files
 4683: @subsection Words for Including
 4684: 
 4685: doc-include-file
 4686: doc-included
 4687: doc-include
 4688: 
 4689: Usually you want to include a file only if it is not included already
 4690: (by, say, another source file):
 4691: 
 4692: doc-required
 4693: doc-require
 4694: doc-needs
 4695: 
 4696: @cindex stack effect of included files
 4697: @cindex including files, stack effect
 4698: I recommend that you write your source files such that interpreting them
 4699: does not change the stack. This allows using these files with
 4700: @code{required} and friends without complications. E.g.,
 4701: 
 4702: @example
 4703: 1 require foo.fs drop
 4704: @end example
 4705: 
 4706: @node Search Path, Changing the Search Path, Words for Including, Including Files
 4707: @subsection Search Path
 4708: @cindex path for @code{included}
 4709: @cindex file search path
 4710: @cindex include search path
 4711: @cindex search path for files
 4712: 
 4713: If you specify an absolute filename (i.e., a filename starting with
 4714: @file{/} or @file{~}, or with @file{:} in the second position (as in
 4715: @samp{C:...})) for @code{included} and friends, that file is included
 4716: just as you would expect.
 4717: 
 4718: For relative filenames, Gforth uses a search path similar to Forth's
 4719: search order (@pxref{Wordlists}). It tries to find the given filename in
 4720: the directories present in the path, and includes the first one it
 4721: finds.
 4722: 
 4723: If the search path contains the directory @file{.} (as it should), this
 4724: refers to the directory that the present file was @code{included}
 4725: from. This allows files to include other files relative to their own
 4726: position (irrespective of the current working directory or the absolute
 4727: position).  This feature is essential for libraries consisting of
 4728: several files, where a file may include other files from the library.
 4729: It corresponds to @code{#include "..."} in C. If the current input
 4730: source is not a file, @file{.} refers to the directory of the innermost
 4731: file being included, or, if there is no file being included, to the
 4732: current working directory.
 4733: 
 4734: Use @file{~+} to refer to the current working directory (as in the
 4735: @code{bash}).
 4736: 
 4737: If the filename starts with @file{./}, the search path is not searched
 4738: (just as with absolute filenames), and the @file{.} has the same meaning
 4739: as described above.
 4740: 
 4741: @node Changing the Search Path, General Search Paths, Search Path, Including Files
 4742: @subsection Changing the Search Path
 4743: @cindex search path, changes
 4744: 
 4745: The search path is initialized when you start Gforth (@pxref{Invoking
 4746: Gforth}). You can display it with
 4747: 
 4748: doc-.fpath
 4749: 
 4750: You can change it later with the following words:
 4751: 
 4752: doc-fpath+
 4753: doc-fpath=
 4754: 
 4755: Using fpath and require would look like:
 4756: 
 4757: @example
 4758: fpath= /usr/lib/forth/|./
 4759: 
 4760: require timer.fs
 4761: @end example
 4762: 
 4763: If you have the need to look for a file in the Forth search path, you could
 4764: use this Gforth feature in your application.
 4765: 
 4766: doc-open-fpath-file
 4767: 
 4768: @node General Search Paths,  , Changing the Search Path, Including Files
 4769: @subsection General Search Paths
 4770: @cindex search paths for user applications
 4771: 
 4772: Your application may need to search files in sevaral directories, like
 4773: @code{included} does. For this purpose you can define and use your own
 4774: search paths. Create a search path like this:
 4775: 
 4776: @example
 4777: 
 4778: Make a buffer for the path:
 4779: create mypath   100 chars ,     \ maximum length (is checked)
 4780:                 0 ,             \ real len
 4781:                 100 chars allot \ space for path
 4782: @end example
 4783: 
 4784: You have the same functions for the forth search path in a generic version
 4785: for different paths.
 4786: 
 4787: doc-path+
 4788: doc-path=
 4789: doc-.path
 4790: doc-open-path-file
 4791: 
 4792: 
 4793: @node Blocks, Other I/O, Including Files, Words
 4794: @section Blocks
 4795: 
 4796: This chapter describes how to use block files within Gforth.
 4797: 
 4798: Block files are traditionally means of data and source storage in
 4799: Forth. They have been very important in resource-starved computers
 4800: without OS in the past. Gforth doesn't encourage to use blocks as
 4801: source, and provides blocks only for backward compatibility. The ANS
 4802: standard requires blocks to be available when files are.
 4803: 
 4804: doc-open-blocks
 4805: doc-use
 4806: doc-get-block-fid
 4807: doc-block-position
 4808: doc-update
 4809: doc-save-buffer
 4810: doc-empty-buffer
 4811: doc-flush
 4812: doc-get-buffer
 4813: doc-block
 4814: doc-buffer
 4815: doc-updated?
 4816: doc-list
 4817: doc-load
 4818: doc-thru
 4819: doc-+load
 4820: doc-+thru
 4821: doc---block--->
 4822: doc-block-included
 4823: 
 4824: @node Other I/O, Programming Tools, Blocks, Words
 4825: @section Other I/O
 4826: 
 4827: @node Programming Tools, Assembler and Code Words, Other I/O, Words
 4828: @section Programming Tools
 4829: @cindex programming tools
 4830: 
 4831: @menu
 4832: * Debugging::                   Simple and quick.
 4833: * Assertions::                  Making your programs self-checking.
 4834: * Singlestep Debugger::		Executing your program word by word.
 4835: @end menu
 4836: 
 4837: @node Debugging, Assertions, Programming Tools, Programming Tools
 4838: @subsection Debugging
 4839: @cindex debugging
 4840: 
 4841: The simple debugging aids provided in @file{debugs.fs}
 4842: are meant to support a different style of debugging than the
 4843: tracing/stepping debuggers used in languages with long turn-around
 4844: times.
 4845: 
 4846: A much better (faster) way in fast-compiling languages is to add
 4847: printing code at well-selected places, let the program run, look at
 4848: the output, see where things went wrong, add more printing code, etc.,
 4849: until the bug is found.
 4850: 
 4851: The word @code{~~} is easy to insert. It just prints debugging
 4852: information (by default the source location and the stack contents). It
 4853: is also easy to remove (@kbd{C-x ~} in the Emacs Forth mode to
 4854: query-replace them with nothing). The deferred words
 4855: @code{printdebugdata} and @code{printdebugline} control the output of
 4856: @code{~~}. The default source location output format works well with
 4857: Emacs' compilation mode, so you can step through the program at the
 4858: source level using @kbd{C-x `} (the advantage over a stepping debugger
 4859: is that you can step in any direction and you know where the crash has
 4860: happened or where the strange data has occurred).
 4861: 
 4862: Note that the default actions clobber the contents of the pictured
 4863: numeric output string, so you should not use @code{~~}, e.g., between
 4864: @code{<#} and @code{#>}.
 4865: 
 4866: doc-~~
 4867: doc-printdebugdata
 4868: doc-printdebugline
 4869: 
 4870: @node Assertions, Singlestep Debugger, Debugging, Programming Tools
 4871: @subsection Assertions
 4872: @cindex assertions
 4873: 
 4874: It is a good idea to make your programs self-checking, in particular, if
 4875: you use an assumption (e.g., that a certain field of a data structure is
 4876: never zero) that may become wrong during maintenance. Gforth supports
 4877: assertions for this purpose. They are used like this:
 4878: 
 4879: @example
 4880: assert( @var{flag} )
 4881: @end example
 4882: 
 4883: The code between @code{assert(} and @code{)} should compute a flag, that
 4884: should be true if everything is alright and false otherwise. It should
 4885: not change anything else on the stack. The overall stack effect of the
 4886: assertion is @code{( -- )}. E.g.
 4887: 
 4888: @example
 4889: assert( 1 1 + 2 = ) \ what we learn in school
 4890: assert( dup 0<> ) \ assert that the top of stack is not zero
 4891: assert( false ) \ this code should not be reached
 4892: @end example
 4893: 
 4894: The need for assertions is different at different times. During
 4895: debugging, we want more checking, in production we sometimes care more
 4896: for speed. Therefore, assertions can be turned off, i.e., the assertion
 4897: becomes a comment. Depending on the importance of an assertion and the
 4898: time it takes to check it, you may want to turn off some assertions and
 4899: keep others turned on. Gforth provides several levels of assertions for
 4900: this purpose:
 4901: 
 4902: doc-assert0(
 4903: doc-assert1(
 4904: doc-assert2(
 4905: doc-assert3(
 4906: doc-assert(
 4907: doc-)
 4908: 
 4909: @code{Assert(} is the same as @code{assert1(}. The variable
 4910: @code{assert-level} specifies the highest assertions that are turned
 4911: on. I.e., at the default @code{assert-level} of one, @code{assert0(} and
 4912: @code{assert1(} assertions perform checking, while @code{assert2(} and
 4913: @code{assert3(} assertions are treated as comments.
 4914: 
 4915: Note that the @code{assert-level} is evaluated at compile-time, not at
 4916: run-time. I.e., you cannot turn assertions on or off at run-time, you
 4917: have to set the @code{assert-level} appropriately before compiling a
 4918: piece of code. You can compile several pieces of code at several
 4919: @code{assert-level}s (e.g., a trusted library at level 1 and newly
 4920: written code at level 3).
 4921: 
 4922: doc-assert-level
 4923: 
 4924: If an assertion fails, a message compatible with Emacs' compilation mode
 4925: is produced and the execution is aborted (currently with @code{ABORT"}.
 4926: If there is interest, we will introduce a special throw code. But if you
 4927: intend to @code{catch} a specific condition, using @code{throw} is
 4928: probably more appropriate than an assertion).
 4929: 
 4930: @node Singlestep Debugger, , Assertions, Programming Tools
 4931: @subsection Singlestep Debugger
 4932: @cindex singlestep Debugger
 4933: @cindex debugging Singlestep
 4934: @cindex @code{dbg}
 4935: @cindex @code{BREAK:}
 4936: @cindex @code{BREAK"}
 4937: 
 4938: When a new word is created there's often the need to check whether it behaves
 4939: correctly or not. You can do this by typing @code{dbg badword}. This might
 4940: look like:
 4941: @example
 4942: : badword 0 DO i . LOOP ;  ok
 4943: 2 dbg badword 
 4944: : badword  
 4945: Scanning code...
 4946: 
 4947: Nesting debugger ready!
 4948: 
 4949: 400D4738  8049BC4 0              -> [ 2 ] 00002 00000 
 4950: 400D4740  8049F68 DO             -> [ 0 ] 
 4951: 400D4744  804A0C8 i              -> [ 1 ] 00000 
 4952: 400D4748 400C5E60 .              -> 0 [ 0 ] 
 4953: 400D474C  8049D0C LOOP           -> [ 0 ] 
 4954: 400D4744  804A0C8 i              -> [ 1 ] 00001 
 4955: 400D4748 400C5E60 .              -> 1 [ 0 ] 
 4956: 400D474C  8049D0C LOOP           -> [ 0 ] 
 4957: 400D4758  804B384 ;              ->  ok
 4958: @end example
 4959: 
 4960: Each line displayed is one step. You always have to hit return to
 4961: execute the next word that is displayed. If you don't want to execute
 4962: the next word in a whole, you have to type @kbd{n} for @code{nest}. Here is
 4963: an overview what keys are available:
 4964: 
 4965: @table @i
 4966: 
 4967: @item <return>
 4968: Next; Execute the next word.
 4969: 
 4970: @item n
 4971: Nest; Single step through next word.
 4972: 
 4973: @item u
 4974: Unnest; Stop debugging and execute rest of word. If we got to this word
 4975: with nest, continue debugging with the calling word.
 4976: 
 4977: @item d
 4978: Done; Stop debugging and execute rest.
 4979: 
 4980: @item s
 4981: Stopp; Abort immediately.
 4982: 
 4983: @end table
 4984: 
 4985: Debugging large application with this mechanism is very difficult, because
 4986: you have to nest very deep into the program before the interesting part
 4987: begins. This takes a lot of time. 
 4988: 
 4989: To do it more directly put a @code{BREAK:} command into your source code.
 4990: When program execution reaches @code{BREAK:} the single step debugger is
 4991: invoked and you have all the features described above.
 4992: 
 4993: If you have more than one part to debug it is useful to know where the
 4994: program has stopped at the moment. You can do this by the 
 4995: @code{BREAK" string"} command. This behaves like @code{BREAK:} except that
 4996: string is typed out when the ``breakpoint'' is reached.
 4997: 
 4998: @node Assembler and Code Words, Threading Words, Programming Tools, Words
 4999: @section Assembler and Code Words
 5000: @cindex assembler
 5001: @cindex code words
 5002: 
 5003: Gforth provides some words for defining primitives (words written in
 5004: machine code), and for defining the the machine-code equivalent of
 5005: @code{DOES>}-based defining words. However, the machine-independent
 5006: nature of Gforth poses a few problems: First of all, Gforth runs on
 5007: several architectures, so it can provide no standard assembler. What's
 5008: worse is that the register allocation not only depends on the processor,
 5009: but also on the @code{gcc} version and options used.
 5010: 
 5011: The words that Gforth offers encapsulate some system dependences (e.g., the
 5012: header structure), so a system-independent assembler may be used in
 5013: Gforth. If you do not have an assembler, you can compile machine code
 5014: directly with @code{,} and @code{c,}.
 5015: 
 5016: doc-assembler
 5017: doc-code
 5018: doc-end-code
 5019: doc-;code
 5020: doc-flush-icache
 5021: 
 5022: If @code{flush-icache} does not work correctly, @code{code} words
 5023: etc. will not work (reliably), either.
 5024: 
 5025: These words are rarely used. Therefore they reside in @code{code.fs},
 5026: which is usually not loaded (except @code{flush-icache}, which is always
 5027: present). You can load them with @code{require code.fs}.
 5028: 
 5029: @cindex registers of the inner interpreter
 5030: In the assembly code you will want to refer to the inner interpreter's
 5031: registers (e.g., the data stack pointer) and you may want to use other
 5032: registers for temporary storage. Unfortunately, the register allocation
 5033: is installation-dependent.
 5034: 
 5035: The easiest solution is to use explicit register declarations
 5036: (@pxref{Explicit Reg Vars, , Variables in Specified Registers, gcc.info,
 5037: GNU C Manual}) for all of the inner interpreter's registers: You have to
 5038: compile Gforth with @code{-DFORCE_REG} (configure option
 5039: @code{--enable-force-reg}) and the appropriate declarations must be
 5040: present in the @code{machine.h} file (see @code{mips.h} for an example;
 5041: you can find a full list of all declarable register symbols with
 5042: @code{grep register engine.c}). If you give explicit registers to all
 5043: variables that are declared at the beginning of @code{engine()}, you
 5044: should be able to use the other caller-saved registers for temporary
 5045: storage. Alternatively, you can use the @code{gcc} option
 5046: @code{-ffixed-REG} (@pxref{Code Gen Options, , Options for Code
 5047: Generation Conventions, gcc.info, GNU C Manual}) to reserve a register
 5048: (however, this restriction on register allocation may slow Gforth
 5049: significantly).
 5050: 
 5051: If this solution is not viable (e.g., because @code{gcc} does not allow
 5052: you to explicitly declare all the registers you need), you have to find
 5053: out by looking at the code where the inner interpreter's registers
 5054: reside and which registers can be used for temporary storage. You can
 5055: get an assembly listing of the engine's code with @code{make engine.s}.
 5056: 
 5057: In any case, it is good practice to abstract your assembly code from the
 5058: actual register allocation. E.g., if the data stack pointer resides in
 5059: register @code{$17}, create an alias for this register called @code{sp},
 5060: and use that in your assembly code.
 5061: 
 5062: @cindex code words, portable
 5063: Another option for implementing normal and defining words efficiently
 5064: is: adding the wanted functionality to the source of Gforth. For normal
 5065: words you just have to edit @file{primitives} (@pxref{Automatic
 5066: Generation}), defining words (equivalent to @code{;CODE} words, for fast
 5067: defined words) may require changes in @file{engine.c}, @file{kernal.fs},
 5068: @file{prims2x.fs}, and possibly @file{cross.fs}.
 5069: 
 5070: 
 5071: @node Threading Words,  , Assembler and Code Words, Words
 5072: @section Threading Words
 5073: @cindex threading words
 5074: 
 5075: @cindex code address
 5076: These words provide access to code addresses and other threading stuff
 5077: in Gforth (and, possibly, other interpretive Forths). It more or less
 5078: abstracts away the differences between direct and indirect threading
 5079: (and, for direct threading, the machine dependences). However, at
 5080: present this wordset is still incomplete. It is also pretty low-level;
 5081: some day it will hopefully be made unnecessary by an internals wordset
 5082: that abstracts implementation details away completely.
 5083: 
 5084: doc->code-address
 5085: doc->does-code
 5086: doc-code-address!
 5087: doc-does-code!
 5088: doc-does-handler!
 5089: doc-/does-handler
 5090: 
 5091: The code addresses produced by various defining words are produced by
 5092: the following words:
 5093: 
 5094: doc-docol:
 5095: doc-docon:
 5096: doc-dovar:
 5097: doc-douser:
 5098: doc-dodefer:
 5099: doc-dofield:
 5100: 
 5101: You can recognize words defined by a @code{CREATE}...@code{DOES>} word
 5102: with @code{>DOES-CODE}. If the word was defined in that way, the value
 5103: returned is different from 0 and identifies the @code{DOES>} used by the
 5104: defining word.
 5105: 
 5106: @c ******************************************************************
 5107: @node Tools, ANS conformance, Words, Top
 5108: @chapter Tools
 5109: 
 5110: @menu
 5111: * ANS Report::                  Report the words used, sorted by wordset.
 5112: @end menu
 5113: 
 5114: See also @ref{Emacs and Gforth}.
 5115: 
 5116: @node ANS Report,  , Tools, Tools
 5117: @section @file{ans-report.fs}: Report the words used, sorted by wordset
 5118: @cindex @file{ans-report.fs}
 5119: @cindex report the words used in your program
 5120: @cindex words used in your program
 5121: 
 5122: If you want to label a Forth program as ANS Forth Program, you must
 5123: document which wordsets the program uses; for extension wordsets, it is
 5124: helpful to list the words the program requires from these wordsets
 5125: (because Forth systems are allowed to provide only some words of them).
 5126: 
 5127: The @file{ans-report.fs} tool makes it easy for you to determine which
 5128: words from which wordset and which non-ANS words your application
 5129: uses. You simply have to include @file{ans-report.fs} before loading the
 5130: program you want to check. After loading your program, you can get the
 5131: report with @code{print-ans-report}. A typical use is to run this as
 5132: batch job like this:
 5133: @example
 5134: gforth ans-report.fs myprog.fs -e "print-ans-report bye"
 5135: @end example
 5136: 
 5137: The output looks like this (for @file{compat/control.fs}):
 5138: @example
 5139: The program uses the following words
 5140: from CORE :
 5141: : POSTPONE THEN ; immediate ?dup IF 0= 
 5142: from BLOCK-EXT :
 5143: \ 
 5144: from FILE :
 5145: ( 
 5146: @end example
 5147: 
 5148: @subsection Caveats
 5149: 
 5150: Note that @file{ans-report.fs} just checks which words are used, not whether
 5151: they are used in an ANS Forth conforming way!
 5152: 
 5153: Some words are defined in several wordsets in the
 5154: standard. @file{ans-report.fs} reports them for only one of the
 5155: wordsets, and not necessarily the one you expect. It depends on usage
 5156: which wordset is the right one to specify. E.g., if you only use the
 5157: compilation semantics of @code{S"}, it is a Core word; if you also use
 5158: its interpretation semantics, it is a File word.
 5159: 
 5160: @c ******************************************************************
 5161: @node ANS conformance, Model, Tools, Top
 5162: @chapter ANS conformance
 5163: @cindex ANS conformance of Gforth
 5164: 
 5165: To the best of our knowledge, Gforth is an
 5166: 
 5167: ANS Forth System
 5168: @itemize @bullet
 5169: @item providing the Core Extensions word set
 5170: @item providing the Block word set
 5171: @item providing the Block Extensions word set
 5172: @item providing the Double-Number word set
 5173: @item providing the Double-Number Extensions word set
 5174: @item providing the Exception word set
 5175: @item providing the Exception Extensions word set
 5176: @item providing the Facility word set
 5177: @item providing @code{MS} and @code{TIME&DATE} from the Facility Extensions word set
 5178: @item providing the File Access word set
 5179: @item providing the File Access Extensions word set
 5180: @item providing the Floating-Point word set
 5181: @item providing the Floating-Point Extensions word set
 5182: @item providing the Locals word set
 5183: @item providing the Locals Extensions word set
 5184: @item providing the Memory-Allocation word set
 5185: @item providing the Memory-Allocation Extensions word set (that one's easy)
 5186: @item providing the Programming-Tools word set
 5187: @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
 5188: @item providing the Search-Order word set
 5189: @item providing the Search-Order Extensions word set
 5190: @item providing the String word set
 5191: @item providing the String Extensions word set (another easy one)
 5192: @end itemize
 5193: 
 5194: @cindex system documentation
 5195: In addition, ANS Forth systems are required to document certain
 5196: implementation choices. This chapter tries to meet these
 5197: requirements. In many cases it gives a way to ask the system for the
 5198: information instead of providing the information directly, in
 5199: particular, if the information depends on the processor, the operating
 5200: system or the installation options chosen, or if they are likely to
 5201: change during the maintenance of Gforth.
 5202: 
 5203: @comment The framework for the rest has been taken from pfe.
 5204: 
 5205: @menu
 5206: * The Core Words::              
 5207: * The optional Block word set::  
 5208: * The optional Double Number word set::  
 5209: * The optional Exception word set::  
 5210: * The optional Facility word set::  
 5211: * The optional File-Access word set::  
 5212: * The optional Floating-Point word set::  
 5213: * The optional Locals word set::  
 5214: * The optional Memory-Allocation word set::  
 5215: * The optional Programming-Tools word set::  
 5216: * The optional Search-Order word set::  
 5217: @end menu
 5218: 
 5219: 
 5220: @c =====================================================================
 5221: @node The Core Words, The optional Block word set, ANS conformance, ANS conformance
 5222: @comment  node-name,  next,  previous,  up
 5223: @section The Core Words
 5224: @c =====================================================================
 5225: @cindex core words, system documentation
 5226: @cindex system documentation, core words
 5227: 
 5228: @menu
 5229: * core-idef::                   Implementation Defined Options                   
 5230: * core-ambcond::                Ambiguous Conditions                
 5231: * core-other::                  Other System Documentation                  
 5232: @end menu
 5233: 
 5234: @c ---------------------------------------------------------------------
 5235: @node core-idef, core-ambcond, The Core Words, The Core Words
 5236: @subsection Implementation Defined Options
 5237: @c ---------------------------------------------------------------------
 5238: @cindex core words, implementation-defined options
 5239: @cindex implementation-defined options, core words
 5240: 
 5241: 
 5242: @table @i
 5243: @item (Cell) aligned addresses:
 5244: @cindex cell-aligned addresses
 5245: @cindex aligned addresses
 5246: processor-dependent. Gforth's alignment words perform natural alignment
 5247: (e.g., an address aligned for a datum of size 8 is divisible by
 5248: 8). Unaligned accesses usually result in a @code{-23 THROW}.
 5249: 
 5250: @item @code{EMIT} and non-graphic characters:
 5251: @cindex @code{EMIT} and non-graphic characters
 5252: @cindex non-graphic characters and @code{EMIT}
 5253: The character is output using the C library function (actually, macro)
 5254: @code{putc}.
 5255: 
 5256: @item character editing of @code{ACCEPT} and @code{EXPECT}:
 5257: @cindex character editing of @code{ACCEPT} and @code{EXPECT}
 5258: @cindex editing in @code{ACCEPT} and @code{EXPECT}
 5259: @cindex @code{ACCEPT}, editing
 5260: @cindex @code{EXPECT}, editing
 5261: This is modeled on the GNU readline library (@pxref{Readline
 5262: Interaction, , Command Line Editing, readline, The GNU Readline
 5263: Library}) with Emacs-like key bindings. @kbd{Tab} deviates a little by
 5264: producing a full word completion every time you type it (instead of
 5265: producing the common prefix of all completions).
 5266: 
 5267: @item character set:
 5268: @cindex character set
 5269: The character set of your computer and display device. Gforth is
 5270: 8-bit-clean (but some other component in your system may make trouble).
 5271: 
 5272: @item Character-aligned address requirements:
 5273: @cindex character-aligned address requirements
 5274: installation-dependent. Currently a character is represented by a C
 5275: @code{unsigned char}; in the future we might switch to @code{wchar_t}
 5276: (Comments on that requested).
 5277: 
 5278: @item character-set extensions and matching of names:
 5279: @cindex character-set extensions and matching of names
 5280: @cindex case sensitivity for name lookup
 5281: @cindex name lookup, case sensitivity
 5282: @cindex locale and case sensitivity
 5283: Any character except the ASCII NUL charcter can be used in a
 5284: name. Matching is case-insensitive (except in @code{TABLE}s). The
 5285: matching is performed using the C function @code{strncasecmp}, whose
 5286: function is probably influenced by the locale. E.g., the @code{C} locale
 5287: does not know about accents and umlauts, so they are matched
 5288: case-sensitively in that locale. For portability reasons it is best to
 5289: write programs such that they work in the @code{C} locale. Then one can
 5290: use libraries written by a Polish programmer (who might use words
 5291: containing ISO Latin-2 encoded characters) and by a French programmer
 5292: (ISO Latin-1) in the same program (of course, @code{WORDS} will produce
 5293: funny results for some of the words (which ones, depends on the font you
 5294: are using)). Also, the locale you prefer may not be available in other
 5295: operating systems. Hopefully, Unicode will solve these problems one day.
 5296: 
 5297: @item conditions under which control characters match a space delimiter:
 5298: @cindex space delimiters
 5299: @cindex control characters as delimiters
 5300: If @code{WORD} is called with the space character as a delimiter, all
 5301: white-space characters (as identified by the C macro @code{isspace()})
 5302: are delimiters. @code{PARSE}, on the other hand, treats space like other
 5303: delimiters. @code{PARSE-WORD} treats space like @code{WORD}, but behaves
 5304: like @code{PARSE} otherwise. @code{(NAME)}, which is used by the outer
 5305: interpreter (aka text interpreter) by default, treats all white-space
 5306: characters as delimiters.
 5307: 
 5308: @item format of the control flow stack:
 5309: @cindex control flow stack, format
 5310: The data stack is used as control flow stack. The size of a control flow
 5311: stack item in cells is given by the constant @code{cs-item-size}. At the
 5312: time of this writing, an item consists of a (pointer to a) locals list
 5313: (third), an address in the code (second), and a tag for identifying the
 5314: item (TOS). The following tags are used: @code{defstart},
 5315: @code{live-orig}, @code{dead-orig}, @code{dest}, @code{do-dest},
 5316: @code{scopestart}.
 5317: 
 5318: @item conversion of digits > 35
 5319: @cindex digits > 35
 5320: The characters @code{[\]^_'} are the digits with the decimal value
 5321: 36@minus{}41. There is no way to input many of the larger digits.
 5322: 
 5323: @item display after input terminates in @code{ACCEPT} and @code{EXPECT}:
 5324: @cindex @code{EXPECT}, display after end of input
 5325: @cindex @code{ACCEPT}, display after end of input
 5326: The cursor is moved to the end of the entered string. If the input is
 5327: terminated using the @kbd{Return} key, a space is typed.
 5328: 
 5329: @item exception abort sequence of @code{ABORT"}:
 5330: @cindex exception abort sequence of @code{ABORT"}
 5331: @cindex @code{ABORT"}, exception abort sequence
 5332: The error string is stored into the variable @code{"error} and a
 5333: @code{-2 throw} is performed.
 5334: 
 5335: @item input line terminator:
 5336: @cindex input line terminator
 5337: @cindex line terminator on input
 5338: @cindex newline charcter on input
 5339: For interactive input, @kbd{C-m} (CR) and @kbd{C-j} (LF) terminate
 5340: lines. One of these characters is typically produced when you type the
 5341: @kbd{Enter} or @kbd{Return} key.
 5342: 
 5343: @item maximum size of a counted string:
 5344: @cindex maximum size of a counted string
 5345: @cindex counted string, maximum size
 5346: @code{s" /counted-string" environment? drop .}. Currently 255 characters
 5347: on all ports, but this may change.
 5348: 
 5349: @item maximum size of a parsed string:
 5350: @cindex maximum size of a parsed string
 5351: @cindex parsed string, maximum size
 5352: Given by the constant @code{/line}. Currently 255 characters.
 5353: 
 5354: @item maximum size of a definition name, in characters:
 5355: @cindex maximum size of a definition name, in characters
 5356: @cindex name, maximum length
 5357: 31
 5358: 
 5359: @item maximum string length for @code{ENVIRONMENT?}, in characters:
 5360: @cindex maximum string length for @code{ENVIRONMENT?}, in characters
 5361: @cindex @code{ENVIRONMENT?} string length, maximum
 5362: 31
 5363: 
 5364: @item method of selecting the user input device:
 5365: @cindex user input device, method of selecting
 5366: The user input device is the standard input. There is currently no way to
 5367: change it from within Gforth. However, the input can typically be
 5368: redirected in the command line that starts Gforth.
 5369: 
 5370: @item method of selecting the user output device:
 5371: @cindex user output device, method of selecting
 5372: @code{EMIT} and @code{TYPE} output to the file-id stored in the value
 5373: @code{outfile-id} (@code{stdout} by default). Gforth uses unbuffered
 5374: output when the user output device is a terminal, otherwise the output
 5375: is buffered.
 5376: 
 5377: @item methods of dictionary compilation:
 5378: What are we expected to document here?
 5379: 
 5380: @item number of bits in one address unit:
 5381: @cindex number of bits in one address unit
 5382: @cindex address unit, size in bits
 5383: @code{s" address-units-bits" environment? drop .}. 8 in all current
 5384: ports.
 5385: 
 5386: @item number representation and arithmetic:
 5387: @cindex number representation and arithmetic
 5388: Processor-dependent. Binary two's complement on all current ports.
 5389: 
 5390: @item ranges for integer types:
 5391: @cindex ranges for integer types
 5392: @cindex integer types, ranges
 5393: Installation-dependent. Make environmental queries for @code{MAX-N},
 5394: @code{MAX-U}, @code{MAX-D} and @code{MAX-UD}. The lower bounds for
 5395: unsigned (and positive) types is 0. The lower bound for signed types on
 5396: two's complement and one's complement machines machines can be computed
 5397: by adding 1 to the upper bound.
 5398: 
 5399: @item read-only data space regions:
 5400: @cindex read-only data space regions
 5401: @cindex data-space, read-only regions
 5402: The whole Forth data space is writable.
 5403: 
 5404: @item size of buffer at @code{WORD}:
 5405: @cindex size of buffer at @code{WORD}
 5406: @cindex @code{WORD} buffer size
 5407: @code{PAD HERE - .}. 104 characters on 32-bit machines. The buffer is
 5408: shared with the pictured numeric output string. If overwriting
 5409: @code{PAD} is acceptable, it is as large as the remaining dictionary
 5410: space, although only as much can be sensibly used as fits in a counted
 5411: string.
 5412: 
 5413: @item size of one cell in address units:
 5414: @cindex cell size
 5415: @code{1 cells .}.
 5416: 
 5417: @item size of one character in address units:
 5418: @cindex char size
 5419: @code{1 chars .}. 1 on all current ports.
 5420: 
 5421: @item size of the keyboard terminal buffer:
 5422: @cindex size of the keyboard terminal buffer
 5423: @cindex terminal buffer, size
 5424: Varies. You can determine the size at a specific time using @code{lp@@
 5425: tib - .}. It is shared with the locals stack and TIBs of files that
 5426: include the current file. You can change the amount of space for TIBs
 5427: and locals stack at Gforth startup with the command line option
 5428: @code{-l}.
 5429: 
 5430: @item size of the pictured numeric output buffer:
 5431: @cindex size of the pictured numeric output buffer
 5432: @cindex pictured numeric output buffer, size
 5433: @code{PAD HERE - .}. 104 characters on 32-bit machines. The buffer is
 5434: shared with @code{WORD}.
 5435: 
 5436: @item size of the scratch area returned by @code{PAD}:
 5437: @cindex size of the scratch area returned by @code{PAD}
 5438: @cindex @code{PAD} size
 5439: The remainder of dictionary space. @code{unused pad here - - .}.
 5440: 
 5441: @item system case-sensitivity characteristics:
 5442: @cindex case-sensitivity characteristics
 5443: Dictionary searches are case insensitive (except in
 5444: @code{TABLE}s). However, as explained above under @i{character-set
 5445: extensions}, the matching for non-ASCII characters is determined by the
 5446: locale you are using. In the default @code{C} locale all non-ASCII
 5447: characters are matched case-sensitively.
 5448: 
 5449: @item system prompt:
 5450: @cindex system prompt
 5451: @cindex prompt
 5452: @code{ ok} in interpret state, @code{ compiled} in compile state.
 5453: 
 5454: @item division rounding:
 5455: @cindex division rounding
 5456: installation dependent. @code{s" floored" environment? drop .}. We leave
 5457: the choice to @code{gcc} (what to use for @code{/}) and to you (whether
 5458: to use @code{fm/mod}, @code{sm/rem} or simply @code{/}).
 5459: 
 5460: @item values of @code{STATE} when true:
 5461: @cindex @code{STATE} values
 5462: -1.
 5463: 
 5464: @item values returned after arithmetic overflow:
 5465: On two's complement machines, arithmetic is performed modulo
 5466: 2**bits-per-cell for single arithmetic and 4**bits-per-cell for double
 5467: arithmetic (with appropriate mapping for signed types). Division by zero
 5468: typically results in a @code{-55 throw} (Floating-point unidentified
 5469: fault), although a @code{-10 throw} (divide by zero) would be more
 5470: appropriate.
 5471: 
 5472: @item whether the current definition can be found after @t{DOES>}:
 5473: @cindex @t{DOES>}, visibility of current definition
 5474: No.
 5475: 
 5476: @end table
 5477: 
 5478: @c ---------------------------------------------------------------------
 5479: @node core-ambcond, core-other, core-idef, The Core Words
 5480: @subsection Ambiguous conditions
 5481: @c ---------------------------------------------------------------------
 5482: @cindex core words, ambiguous conditions
 5483: @cindex ambiguous conditions, core words
 5484: 
 5485: @table @i
 5486: 
 5487: @item a name is neither a word nor a number:
 5488: @cindex name not found
 5489: @cindex Undefined word
 5490: @code{-13 throw} (Undefined word). Actually, @code{-13 bounce}, which
 5491: preserves the data and FP stack, so you don't lose more work than
 5492: necessary.
 5493: 
 5494: @item a definition name exceeds the maximum length allowed:
 5495: @cindex Word name too long
 5496: @code{-19 throw} (Word name too long)
 5497: 
 5498: @item addressing a region not inside the various data spaces of the forth system:
 5499: @cindex Invalid memory address
 5500: The stacks, code space and name space are accessible. Machine code space is
 5501: typically readable. Accessing other addresses gives results dependent on
 5502: the operating system. On decent systems: @code{-9 throw} (Invalid memory
 5503: address).
 5504: 
 5505: @item argument type incompatible with parameter:
 5506: @cindex Argument type mismatch
 5507: This is usually not caught. Some words perform checks, e.g., the control
 5508: flow words, and issue a @code{ABORT"} or @code{-12 THROW} (Argument type
 5509: mismatch).
 5510: 
 5511: @item attempting to obtain the execution token of a word with undefined execution semantics:
 5512: @cindex Interpreting a compile-only word, for @code{'} etc.
 5513: @cindex execution token of words with undefined execution semantics
 5514: @code{-14 throw} (Interpreting a compile-only word). In some cases, you
 5515: get an execution token for @code{compile-only-error} (which performs a
 5516: @code{-14 throw} when executed).
 5517: 
 5518: @item dividing by zero:
 5519: @cindex dividing by zero
 5520: @cindex floating point unidentified fault, integer division
 5521: @cindex divide by zero
 5522: typically results in a @code{-55 throw} (floating point unidentified
 5523: fault), although a @code{-10 throw} (divide by zero) would be more
 5524: appropriate.
 5525: 
 5526: @item insufficient data stack or return stack space:
 5527: @cindex insufficient data stack or return stack space
 5528: @cindex stack overflow
 5529: @cindex Address alignment exception, stack overflow
 5530: @cindex Invalid memory address, stack overflow
 5531: Depending on the operating system, the installation, and the invocation
 5532: of Gforth, this is either checked by the memory management hardware, or
 5533: it is not checked. If it is checked, you typically get a @code{-9 throw}
 5534: (Invalid memory address) as soon as the overflow happens. If it is not
 5535: check, overflows typically result in mysterious illegal memory accesses,
 5536: producing @code{-9 throw} (Invalid memory address) or @code{-23 throw}
 5537: (Address alignment exception); they might also destroy the internal data
 5538: structure of @code{ALLOCATE} and friends, resulting in various errors in
 5539: these words.
 5540: 
 5541: @item insufficient space for loop control parameters:
 5542: @cindex insufficient space for loop control parameters
 5543: like other return stack overflows.
 5544: 
 5545: @item insufficient space in the dictionary:
 5546: @cindex insufficient space in the dictionary
 5547: @cindex dictionary overflow
 5548: If you try to allot (either directly with @code{allot}, or indirectly
 5549: with @code{,}, @code{create} etc.) more memory than available in the
 5550: dictionary, you get a @code{-8 throw} (Dictionary overflow). If you try
 5551: to access memory beyond the end of the dictionary, the results are
 5552: similar to stack overflows.
 5553: 
 5554: @item interpreting a word with undefined interpretation semantics:
 5555: @cindex interpreting a word with undefined interpretation semantics
 5556: @cindex Interpreting a compile-only word
 5557: For some words, we have defined interpretation semantics. For the
 5558: others: @code{-14 throw} (Interpreting a compile-only word).
 5559: 
 5560: @item modifying the contents of the input buffer or a string literal:
 5561: @cindex modifying the contents of the input buffer or a string literal
 5562: These are located in writable memory and can be modified.
 5563: 
 5564: @item overflow of the pictured numeric output string:
 5565: @cindex overflow of the pictured numeric output string
 5566: @cindex pictured numeric output string, overflow
 5567: Not checked. Runs into the dictionary and destroys it (at least,
 5568: partially).
 5569: 
 5570: @item parsed string overflow:
 5571: @cindex parsed string overflow
 5572: @code{PARSE} cannot overflow. @code{WORD} does not check for overflow.
 5573: 
 5574: @item producing a result out of range:
 5575: @cindex result out of range
 5576: On two's complement machines, arithmetic is performed modulo
 5577: 2**bits-per-cell for single arithmetic and 4**bits-per-cell for double
 5578: arithmetic (with appropriate mapping for signed types). Division by zero
 5579: typically results in a @code{-55 throw} (floatingpoint unidentified
 5580: fault), although a @code{-10 throw} (divide by zero) would be more
 5581: appropriate. @code{convert} and @code{>number} currently overflow
 5582: silently.
 5583: 
 5584: @item reading from an empty data or return stack:
 5585: @cindex stack empty
 5586: @cindex stack underflow
 5587: The data stack is checked by the outer (aka text) interpreter after
 5588: every word executed. If it has underflowed, a @code{-4 throw} (Stack
 5589: underflow) is performed. Apart from that, stacks may be checked or not,
 5590: depending on operating system, installation, and invocation. The
 5591: consequences of stack underflows are similar to the consequences of
 5592: stack overflows. Note that even if the system uses checking (through the
 5593: MMU), your program may have to underflow by a significant number of
 5594: stack items to trigger the reaction (the reason for this is that the
 5595: MMU, and therefore the checking, works with a page-size granularity).
 5596: 
 5597: @item unexpected end of the input buffer, resulting in an attempt to use a zero-length string as a name:
 5598: @cindex unexpected end of the input buffer
 5599: @cindex zero-length string as a name
 5600: @cindex Attempt to use zero-length string as a name
 5601: @code{Create} and its descendants perform a @code{-16 throw} (Attempt to
 5602: use zero-length string as a name). Words like @code{'} probably will not
 5603: find what they search. Note that it is possible to create zero-length
 5604: names with @code{nextname} (should it not?).
 5605: 
 5606: @item @code{>IN} greater than input buffer:
 5607: @cindex @code{>IN} greater than input buffer
 5608: The next invocation of a parsing word returns a string with length 0.
 5609: 
 5610: @item @code{RECURSE} appears after @code{DOES>}:
 5611: @cindex @code{RECURSE} appears after @code{DOES>}
 5612: Compiles a recursive call to the defining word, not to the defined word.
 5613: 
 5614: @item argument input source different than current input source for @code{RESTORE-INPUT}:
 5615: @cindex argument input source different than current input source for @code{RESTORE-INPUT}
 5616: @cindex Argument type mismatch, @code{RESTORE-INPUT}
 5617: @cindex @code{RESTORE-INPUT}, Argument type mismatch
 5618: @code{-12 THROW}. Note that, once an input file is closed (e.g., because
 5619: the end of the file was reached), its source-id may be
 5620: reused. Therefore, restoring an input source specification referencing a
 5621: closed file may lead to unpredictable results instead of a @code{-12
 5622: THROW}.
 5623: 
 5624: In the future, Gforth may be able to restore input source specifications
 5625: from other than the current input source.
 5626: 
 5627: @item data space containing definitions gets de-allocated:
 5628: @cindex data space containing definitions gets de-allocated
 5629: Deallocation with @code{allot} is not checked. This typically results in
 5630: memory access faults or execution of illegal instructions.
 5631: 
 5632: @item data space read/write with incorrect alignment:
 5633: @cindex data space read/write with incorrect alignment
 5634: @cindex alignment faults
 5635: @cindex Address alignment exception
 5636: Processor-dependent. Typically results in a @code{-23 throw} (Address
 5637: alignment exception). Under Linux-Intel on a 486 or later processor with
 5638: alignment turned on, incorrect alignment results in a @code{-9 throw}
 5639: (Invalid memory address). There are reportedly some processors with
 5640: alignment restrictions that do not report violations.
 5641: 
 5642: @item data space pointer not properly aligned, @code{,}, @code{C,}:
 5643: @cindex data space pointer not properly aligned, @code{,}, @code{C,}
 5644: Like other alignment errors.
 5645: 
 5646: @item less than u+2 stack items (@code{PICK} and @code{ROLL}):
 5647: Like other stack underflows.
 5648: 
 5649: @item loop control parameters not available:
 5650: @cindex loop control parameters not available
 5651: Not checked. The counted loop words simply assume that the top of return
 5652: stack items are loop control parameters and behave accordingly.
 5653: 
 5654: @item most recent definition does not have a name (@code{IMMEDIATE}):
 5655: @cindex most recent definition does not have a name (@code{IMMEDIATE})
 5656: @cindex last word was headerless
 5657: @code{abort" last word was headerless"}.
 5658: 
 5659: @item name not defined by @code{VALUE} used by @code{TO}:
 5660: @cindex name not defined by @code{VALUE} used by @code{TO}
 5661: @cindex @code{TO} on non-@code{VALUE}s
 5662: @cindex Invalid name argument, @code{TO}
 5663: @code{-32 throw} (Invalid name argument) (unless name is a local or was
 5664: defined by @code{CONSTANT}; in the latter case it just changes the constant).
 5665: 
 5666: @item name not found (@code{'}, @code{POSTPONE}, @code{[']}, @code{[COMPILE]}):
 5667: @cindex name not found (@code{'}, @code{POSTPONE}, @code{[']}, @code{[COMPILE]})
 5668: @cindex Undefined word, @code{'}, @code{POSTPONE}, @code{[']}, @code{[COMPILE]}
 5669: @code{-13 throw} (Undefined word)
 5670: 
 5671: @item parameters are not of the same type (@code{DO}, @code{?DO}, @code{WITHIN}):
 5672: @cindex parameters are not of the same type (@code{DO}, @code{?DO}, @code{WITHIN})
 5673: Gforth behaves as if they were of the same type. I.e., you can predict
 5674: the behaviour by interpreting all parameters as, e.g., signed.
 5675: 
 5676: @item @code{POSTPONE} or @code{[COMPILE]} applied to @code{TO}:
 5677: @cindex @code{POSTPONE} or @code{[COMPILE]} applied to @code{TO}
 5678: Assume @code{: X POSTPONE TO ; IMMEDIATE}. @code{X} performs the
 5679: compilation semantics of @code{TO}.
 5680: 
 5681: @item String longer than a counted string returned by @code{WORD}:
 5682: @cindex String longer than a counted string returned by @code{WORD}
 5683: @cindex @code{WORD}, string overflow
 5684: Not checked. The string will be ok, but the count will, of course,
 5685: contain only the least significant bits of the length.
 5686: 
 5687: @item u greater than or equal to the number of bits in a cell (@code{LSHIFT}, @code{RSHIFT}):
 5688: @cindex @code{LSHIFT}, large shift counts
 5689: @cindex @code{RSHIFT}, large shift counts
 5690: Processor-dependent. Typical behaviours are returning 0 and using only
 5691: the low bits of the shift count.
 5692: 
 5693: @item word not defined via @code{CREATE}:
 5694: @cindex @code{>BODY} of non-@code{CREATE}d words
 5695: @code{>BODY} produces the PFA of the word no matter how it was defined.
 5696: 
 5697: @cindex @code{DOES>} of non-@code{CREATE}d words
 5698: @code{DOES>} changes the execution semantics of the last defined word no
 5699: matter how it was defined. E.g., @code{CONSTANT DOES>} is equivalent to
 5700: @code{CREATE , DOES>}.
 5701: 
 5702: @item words improperly used outside @code{<#} and @code{#>}:
 5703: Not checked. As usual, you can expect memory faults.
 5704: 
 5705: @end table
 5706: 
 5707: 
 5708: @c ---------------------------------------------------------------------
 5709: @node core-other,  , core-ambcond, The Core Words
 5710: @subsection Other system documentation
 5711: @c ---------------------------------------------------------------------
 5712: @cindex other system documentation, core words
 5713: @cindex core words, other system documentation
 5714: 
 5715: @table @i
 5716: @item nonstandard words using @code{PAD}:
 5717: @cindex @code{PAD} use by nonstandard words
 5718: None.
 5719: 
 5720: @item operator's terminal facilities available:
 5721: @cindex operator's terminal facilities available
 5722: After processing the command line, Gforth goes into interactive mode,
 5723: and you can give commands to Gforth interactively. The actual facilities
 5724: available depend on how you invoke Gforth.
 5725: 
 5726: @item program data space available:
 5727: @cindex program data space available
 5728: @cindex data space available
 5729: @code{UNUSED .} gives the remaining dictionary space. The total
 5730: dictionary space can be specified with the @code{-m} switch
 5731: (@pxref{Invoking Gforth}) when Gforth starts up.
 5732: 
 5733: @item return stack space available:
 5734: @cindex return stack space available
 5735: You can compute the total return stack space in cells with
 5736: @code{s" RETURN-STACK-CELLS" environment? drop .}. You can specify it at
 5737: startup time with the @code{-r} switch (@pxref{Invoking Gforth}).
 5738: 
 5739: @item stack space available:
 5740: @cindex stack space available
 5741: You can compute the total data stack space in cells with
 5742: @code{s" STACK-CELLS" environment? drop .}. You can specify it at
 5743: startup time with the @code{-d} switch (@pxref{Invoking Gforth}).
 5744: 
 5745: @item system dictionary space required, in address units:
 5746: @cindex system dictionary space required, in address units
 5747: Type @code{here forthstart - .} after startup. At the time of this
 5748: writing, this gives 80080 (bytes) on a 32-bit system.
 5749: @end table
 5750: 
 5751: 
 5752: @c =====================================================================
 5753: @node The optional Block word set, The optional Double Number word set, The Core Words, ANS conformance
 5754: @section The optional Block word set
 5755: @c =====================================================================
 5756: @cindex system documentation, block words
 5757: @cindex block words, system documentation
 5758: 
 5759: @menu
 5760: * block-idef::                  Implementation Defined Options
 5761: * block-ambcond::               Ambiguous Conditions               
 5762: * block-other::                 Other System Documentation                 
 5763: @end menu
 5764: 
 5765: 
 5766: @c ---------------------------------------------------------------------
 5767: @node block-idef, block-ambcond, The optional Block word set, The optional Block word set
 5768: @subsection Implementation Defined Options
 5769: @c ---------------------------------------------------------------------
 5770: @cindex implementation-defined options, block words
 5771: @cindex block words, implementation-defined options
 5772: 
 5773: @table @i
 5774: @item the format for display by @code{LIST}:
 5775: @cindex @code{LIST} display format
 5776: First the screen number is displayed, then 16 lines of 64 characters,
 5777: each line preceded by the line number.
 5778: 
 5779: @item the length of a line affected by @code{\}:
 5780: @cindex length of a line affected by @code{\}
 5781: @cindex @code{\}, line length in blocks
 5782: 64 characters.
 5783: @end table
 5784: 
 5785: 
 5786: @c ---------------------------------------------------------------------
 5787: @node block-ambcond, block-other, block-idef, The optional Block word set
 5788: @subsection Ambiguous conditions
 5789: @c ---------------------------------------------------------------------
 5790: @cindex block words, ambiguous conditions
 5791: @cindex ambiguous conditions, block words
 5792: 
 5793: @table @i
 5794: @item correct block read was not possible:
 5795: @cindex block read not possible
 5796: Typically results in a @code{throw} of some OS-derived value (between
 5797: -512 and -2048). If the blocks file was just not long enough, blanks are
 5798: supplied for the missing portion.
 5799: 
 5800: @item I/O exception in block transfer:
 5801: @cindex I/O exception in block transfer
 5802: @cindex block transfer, I/O exception
 5803: Typically results in a @code{throw} of some OS-derived value (between
 5804: -512 and -2048).
 5805: 
 5806: @item invalid block number:
 5807: @cindex invalid block number
 5808: @cindex block number invalid
 5809: @code{-35 throw} (Invalid block number)
 5810: 
 5811: @item a program directly alters the contents of @code{BLK}:
 5812: @cindex @code{BLK}, altering @code{BLK}
 5813: The input stream is switched to that other block, at the same
 5814: position. If the storing to @code{BLK} happens when interpreting
 5815: non-block input, the system will get quite confused when the block ends.
 5816: 
 5817: @item no current block buffer for @code{UPDATE}:
 5818: @cindex @code{UPDATE}, no current block buffer
 5819: @code{UPDATE} has no effect.
 5820: 
 5821: @end table
 5822: 
 5823: @c ---------------------------------------------------------------------
 5824: @node block-other,  , block-ambcond, The optional Block word set
 5825: @subsection Other system documentation
 5826: @c ---------------------------------------------------------------------
 5827: @cindex other system documentation, block words
 5828: @cindex block words, other system documentation
 5829: 
 5830: @table @i
 5831: @item any restrictions a multiprogramming system places on the use of buffer addresses:
 5832: No restrictions (yet).
 5833: 
 5834: @item the number of blocks available for source and data:
 5835: depends on your disk space.
 5836: 
 5837: @end table
 5838: 
 5839: 
 5840: @c =====================================================================
 5841: @node The optional Double Number word set, The optional Exception word set, The optional Block word set, ANS conformance
 5842: @section The optional Double Number word set
 5843: @c =====================================================================
 5844: @cindex system documentation, double words
 5845: @cindex double words, system documentation
 5846: 
 5847: @menu
 5848: * double-ambcond::              Ambiguous Conditions              
 5849: @end menu
 5850: 
 5851: 
 5852: @c ---------------------------------------------------------------------
 5853: @node double-ambcond,  , The optional Double Number word set, The optional Double Number word set
 5854: @subsection Ambiguous conditions
 5855: @c ---------------------------------------------------------------------
 5856: @cindex double words, ambiguous conditions
 5857: @cindex ambiguous conditions, double words
 5858: 
 5859: @table @i
 5860: @item @var{d} outside of range of @var{n} in @code{D>S}:
 5861: @cindex @code{D>S}, @var{d} out of range of @var{n} 
 5862: The least significant cell of @var{d} is produced.
 5863: 
 5864: @end table
 5865: 
 5866: 
 5867: @c =====================================================================
 5868: @node The optional Exception word set, The optional Facility word set, The optional Double Number word set, ANS conformance
 5869: @section The optional Exception word set
 5870: @c =====================================================================
 5871: @cindex system documentation, exception words
 5872: @cindex exception words, system documentation
 5873: 
 5874: @menu
 5875: * exception-idef::              Implementation Defined Options              
 5876: @end menu
 5877: 
 5878: 
 5879: @c ---------------------------------------------------------------------
 5880: @node exception-idef,  , The optional Exception word set, The optional Exception word set
 5881: @subsection Implementation Defined Options
 5882: @c ---------------------------------------------------------------------
 5883: @cindex implementation-defined options, exception words
 5884: @cindex exception words, implementation-defined options
 5885: 
 5886: @table @i
 5887: @item @code{THROW}-codes used in the system:
 5888: @cindex @code{THROW}-codes used in the system
 5889: The codes -256@minus{}-511 are used for reporting signals. The mapping
 5890: from OS signal numbers to throw codes is -256@minus{}@var{signal}. The
 5891: codes -512@minus{}-2047 are used for OS errors (for file and memory
 5892: allocation operations). The mapping from OS error numbers to throw codes
 5893: is -512@minus{}@code{errno}. One side effect of this mapping is that
 5894: undefined OS errors produce a message with a strange number; e.g.,
 5895: @code{-1000 THROW} results in @code{Unknown error 488} on my system.
 5896: @end table
 5897: 
 5898: @c =====================================================================
 5899: @node The optional Facility word set, The optional File-Access word set, The optional Exception word set, ANS conformance
 5900: @section The optional Facility word set
 5901: @c =====================================================================
 5902: @cindex system documentation, facility words
 5903: @cindex facility words, system documentation
 5904: 
 5905: @menu
 5906: * facility-idef::               Implementation Defined Options               
 5907: * facility-ambcond::            Ambiguous Conditions            
 5908: @end menu
 5909: 
 5910: 
 5911: @c ---------------------------------------------------------------------
 5912: @node facility-idef, facility-ambcond, The optional Facility word set, The optional Facility word set
 5913: @subsection Implementation Defined Options
 5914: @c ---------------------------------------------------------------------
 5915: @cindex implementation-defined options, facility words
 5916: @cindex facility words, implementation-defined options
 5917: 
 5918: @table @i
 5919: @item encoding of keyboard events (@code{EKEY}):
 5920: @cindex keyboard events, encoding in @code{EKEY}
 5921: @cindex @code{EKEY}, encoding of keyboard events
 5922: Not yet implemented.
 5923: 
 5924: @item duration of a system clock tick:
 5925: @cindex duration of a system clock tick
 5926: @cindex clock tick duration
 5927: System dependent. With respect to @code{MS}, the time is specified in
 5928: microseconds. How well the OS and the hardware implement this, is
 5929: another question.
 5930: 
 5931: @item repeatability to be expected from the execution of @code{MS}:
 5932: @cindex repeatability to be expected from the execution of @code{MS}
 5933: @cindex @code{MS}, repeatability to be expected
 5934: System dependent. On Unix, a lot depends on load. If the system is
 5935: lightly loaded, and the delay is short enough that Gforth does not get
 5936: swapped out, the performance should be acceptable. Under MS-DOS and
 5937: other single-tasking systems, it should be good.
 5938: 
 5939: @end table
 5940: 
 5941: 
 5942: @c ---------------------------------------------------------------------
 5943: @node facility-ambcond,  , facility-idef, The optional Facility word set
 5944: @subsection Ambiguous conditions
 5945: @c ---------------------------------------------------------------------
 5946: @cindex facility words, ambiguous conditions
 5947: @cindex ambiguous conditions, facility words
 5948: 
 5949: @table @i
 5950: @item @code{AT-XY} can't be performed on user output device:
 5951: @cindex @code{AT-XY} can't be performed on user output device
 5952: Largely terminal dependent. No range checks are done on the arguments.
 5953: No errors are reported. You may see some garbage appearing, you may see
 5954: simply nothing happen.
 5955: 
 5956: @end table
 5957: 
 5958: 
 5959: @c =====================================================================
 5960: @node The optional File-Access word set, The optional Floating-Point word set, The optional Facility word set, ANS conformance
 5961: @section The optional File-Access word set
 5962: @c =====================================================================
 5963: @cindex system documentation, file words
 5964: @cindex file words, system documentation
 5965: 
 5966: @menu
 5967: * file-idef::                   Implementation Defined Options
 5968: * file-ambcond::                Ambiguous Conditions                
 5969: @end menu
 5970: 
 5971: @c ---------------------------------------------------------------------
 5972: @node file-idef, file-ambcond, The optional File-Access word set, The optional File-Access word set
 5973: @subsection Implementation Defined Options
 5974: @c ---------------------------------------------------------------------
 5975: @cindex implementation-defined options, file words
 5976: @cindex file words, implementation-defined options
 5977: 
 5978: @table @i
 5979: @item file access methods used:
 5980: @cindex file access methods used
 5981: @code{R/O}, @code{R/W} and @code{BIN} work as you would
 5982: expect. @code{W/O} translates into the C file opening mode @code{w} (or
 5983: @code{wb}): The file is cleared, if it exists, and created, if it does
 5984: not (with both @code{open-file} and @code{create-file}).  Under Unix
 5985: @code{create-file} creates a file with 666 permissions modified by your
 5986: umask.
 5987: 
 5988: @item file exceptions:
 5989: @cindex file exceptions
 5990: The file words do not raise exceptions (except, perhaps, memory access
 5991: faults when you pass illegal addresses or file-ids).
 5992: 
 5993: @item file line terminator:
 5994: @cindex file line terminator
 5995: System-dependent. Gforth uses C's newline character as line
 5996: terminator. What the actual character code(s) of this are is
 5997: system-dependent.
 5998: 
 5999: @item file name format:
 6000: @cindex file name format
 6001: System dependent. Gforth just uses the file name format of your OS.
 6002: 
 6003: @item information returned by @code{FILE-STATUS}:
 6004: @cindex @code{FILE-STATUS}, returned information
 6005: @code{FILE-STATUS} returns the most powerful file access mode allowed
 6006: for the file: Either @code{R/O}, @code{W/O} or @code{R/W}. If the file
 6007: cannot be accessed, @code{R/O BIN} is returned. @code{BIN} is applicable
 6008: along with the returned mode.
 6009: 
 6010: @item input file state after an exception when including source:
 6011: @cindex exception when including source
 6012: All files that are left via the exception are closed.
 6013: 
 6014: @item @var{ior} values and meaning:
 6015: @cindex @var{ior} values and meaning
 6016: The @var{ior}s returned by the file and memory allocation words are
 6017: intended as throw codes. They typically are in the range
 6018: -512@minus{}-2047 of OS errors.  The mapping from OS error numbers to
 6019: @var{ior}s is -512@minus{}@var{errno}.
 6020: 
 6021: @item maximum depth of file input nesting:
 6022: @cindex maximum depth of file input nesting
 6023: @cindex file input nesting, maximum depth
 6024: limited by the amount of return stack, locals/TIB stack, and the number
 6025: of open files available. This should not give you troubles.
 6026: 
 6027: @item maximum size of input line:
 6028: @cindex maximum size of input line
 6029: @cindex input line size, maximum
 6030: @code{/line}. Currently 255.
 6031: 
 6032: @item methods of mapping block ranges to files:
 6033: @cindex mapping block ranges to files
 6034: @cindex files containing blocks
 6035: @cindex blocks in files
 6036: By default, blocks are accessed in the file @file{blocks.fb} in the
 6037: current working directory. The file can be switched with @code{USE}.
 6038: 
 6039: @item number of string buffers provided by @code{S"}:
 6040: @cindex @code{S"}, number of string buffers
 6041: 1
 6042: 
 6043: @item size of string buffer used by @code{S"}:
 6044: @cindex @code{S"}, size of string buffer
 6045: @code{/line}. currently 255.
 6046: 
 6047: @end table
 6048: 
 6049: @c ---------------------------------------------------------------------
 6050: @node file-ambcond,  , file-idef, The optional File-Access word set
 6051: @subsection Ambiguous conditions
 6052: @c ---------------------------------------------------------------------
 6053: @cindex file words, ambiguous conditions
 6054: @cindex ambiguous conditions, file words
 6055: 
 6056: @table @i
 6057: @item attempting to position a file outside its boundaries:
 6058: @cindex @code{REPOSITION-FILE}, outside the file's boundaries
 6059: @code{REPOSITION-FILE} is performed as usual: Afterwards,
 6060: @code{FILE-POSITION} returns the value given to @code{REPOSITION-FILE}.
 6061: 
 6062: @item attempting to read from file positions not yet written:
 6063: @cindex reading from file positions not yet written
 6064: End-of-file, i.e., zero characters are read and no error is reported.
 6065: 
 6066: @item @var{file-id} is invalid (@code{INCLUDE-FILE}):
 6067: @cindex @code{INCLUDE-FILE}, @var{file-id} is invalid 
 6068: An appropriate exception may be thrown, but a memory fault or other
 6069: problem is more probable.
 6070: 
 6071: @item I/O exception reading or closing @var{file-id} (@code{INCLUDE-FILE}, @code{INCLUDED}):
 6072: @cindex @code{INCLUDE-FILE}, I/O exception reading or closing @var{file-id}
 6073: @cindex @code{INCLUDED}, I/O exception reading or closing @var{file-id}
 6074: The @var{ior} produced by the operation, that discovered the problem, is
 6075: thrown.
 6076: 
 6077: @item named file cannot be opened (@code{INCLUDED}):
 6078: @cindex @code{INCLUDED}, named file cannot be opened
 6079: The @var{ior} produced by @code{open-file} is thrown.
 6080: 
 6081: @item requesting an unmapped block number:
 6082: @cindex unmapped block numbers
 6083: There are no unmapped legal block numbers. On some operating systems,
 6084: writing a block with a large number may overflow the file system and
 6085: have an error message as consequence.
 6086: 
 6087: @item using @code{source-id} when @code{blk} is non-zero:
 6088: @cindex @code{SOURCE-ID}, behaviour when @code{BLK} is non-zero
 6089: @code{source-id} performs its function. Typically it will give the id of
 6090: the source which loaded the block. (Better ideas?)
 6091: 
 6092: @end table
 6093: 
 6094: 
 6095: @c =====================================================================
 6096: @node  The optional Floating-Point word set, The optional Locals word set, The optional File-Access word set, ANS conformance
 6097: @section The optional Floating-Point word set
 6098: @c =====================================================================
 6099: @cindex system documentation, floating-point words
 6100: @cindex floating-point words, system documentation
 6101: 
 6102: @menu
 6103: * floating-idef::               Implementation Defined Options
 6104: * floating-ambcond::            Ambiguous Conditions            
 6105: @end menu
 6106: 
 6107: 
 6108: @c ---------------------------------------------------------------------
 6109: @node floating-idef, floating-ambcond, The optional Floating-Point word set, The optional Floating-Point word set
 6110: @subsection Implementation Defined Options
 6111: @c ---------------------------------------------------------------------
 6112: @cindex implementation-defined options, floating-point words
 6113: @cindex floating-point words, implementation-defined options
 6114: 
 6115: @table @i
 6116: @item format and range of floating point numbers:
 6117: @cindex format and range of floating point numbers
 6118: @cindex floating point numbers, format and range
 6119: System-dependent; the @code{double} type of C.
 6120: 
 6121: @item results of @code{REPRESENT} when @var{float} is out of range:
 6122: @cindex  @code{REPRESENT}, results when @var{float} is out of range
 6123: System dependent; @code{REPRESENT} is implemented using the C library
 6124: function @code{ecvt()} and inherits its behaviour in this respect.
 6125: 
 6126: @item rounding or truncation of floating-point numbers:
 6127: @cindex rounding of floating-point numbers
 6128: @cindex truncation of floating-point numbers
 6129: @cindex floating-point numbers, rounding or truncation
 6130: System dependent; the rounding behaviour is inherited from the hosting C
 6131: compiler. IEEE-FP-based (i.e., most) systems by default round to
 6132: nearest, and break ties by rounding to even (i.e., such that the last
 6133: bit of the mantissa is 0).
 6134: 
 6135: @item size of floating-point stack:
 6136: @cindex floating-point stack size
 6137: @code{s" FLOATING-STACK" environment? drop .} gives the total size of
 6138: the floating-point stack (in floats). You can specify this on startup
 6139: with the command-line option @code{-f} (@pxref{Invoking Gforth}).
 6140: 
 6141: @item width of floating-point stack:
 6142: @cindex floating-point stack width 
 6143: @code{1 floats}.
 6144: 
 6145: @end table
 6146: 
 6147: 
 6148: @c ---------------------------------------------------------------------
 6149: @node floating-ambcond,  , floating-idef, The optional Floating-Point word set
 6150: @subsection Ambiguous conditions
 6151: @c ---------------------------------------------------------------------
 6152: @cindex floating-point words, ambiguous conditions
 6153: @cindex ambiguous conditions, floating-point words
 6154: 
 6155: @table @i
 6156: @item @code{df@@} or @code{df!} used with an address that is not double-float  aligned:
 6157: @cindex @code{df@@} or @code{df!} used with an address that is not double-float  aligned
 6158: System-dependent. Typically results in a @code{-23 THROW} like other
 6159: alignment violations.
 6160: 
 6161: @item @code{f@@} or @code{f!} used with an address that is not float  aligned:
 6162: @cindex @code{f@@} used with an address that is not float aligned
 6163: @cindex @code{f!} used with an address that is not float aligned
 6164: System-dependent. Typically results in a @code{-23 THROW} like other
 6165: alignment violations.
 6166: 
 6167: @item floating-point result out of range:
 6168: @cindex floating-point result out of range
 6169: System-dependent. Can result in a @code{-55 THROW} (Floating-point
 6170: unidentified fault), or can produce a special value representing, e.g.,
 6171: Infinity.
 6172: 
 6173: @item @code{sf@@} or @code{sf!} used with an address that is not single-float  aligned:
 6174: @cindex @code{sf@@} or @code{sf!} used with an address that is not single-float  aligned
 6175: System-dependent. Typically results in an alignment fault like other
 6176: alignment violations.
 6177: 
 6178: @item @code{BASE} is not decimal (@code{REPRESENT}, @code{F.}, @code{FE.}, @code{FS.}):
 6179: @cindex @code{BASE} is not decimal (@code{REPRESENT}, @code{F.}, @code{FE.}, @code{FS.})
 6180: The floating-point number is converted into decimal nonetheless.
 6181: 
 6182: @item Both arguments are equal to zero (@code{FATAN2}):
 6183: @cindex @code{FATAN2}, both arguments are equal to zero
 6184: System-dependent. @code{FATAN2} is implemented using the C library
 6185: function @code{atan2()}.
 6186: 
 6187: @item Using @code{FTAN} on an argument @var{r1} where cos(@var{r1}) is zero:
 6188: @cindex @code{FTAN} on an argument @var{r1} where cos(@var{r1}) is zero
 6189: System-dependent. Anyway, typically the cos of @var{r1} will not be zero
 6190: because of small errors and the tan will be a very large (or very small)
 6191: but finite number.
 6192: 
 6193: @item @var{d} cannot be presented precisely as a float in @code{D>F}:
 6194: @cindex @code{D>F}, @var{d} cannot be presented precisely as a float
 6195: The result is rounded to the nearest float.
 6196: 
 6197: @item dividing by zero:
 6198: @cindex dividing by zero, floating-point
 6199: @cindex floating-point dividing by zero
 6200: @cindex floating-point unidentified fault, FP divide-by-zero
 6201: @code{-55 throw} (Floating-point unidentified fault)
 6202: 
 6203: @item exponent too big for conversion (@code{DF!}, @code{DF@@}, @code{SF!}, @code{SF@@}):
 6204: @cindex exponent too big for conversion (@code{DF!}, @code{DF@@}, @code{SF!}, @code{SF@@})
 6205: System dependent. On IEEE-FP based systems the number is converted into
 6206: an infinity.
 6207: 
 6208: @item @var{float}<1 (@code{FACOSH}):
 6209: @cindex @code{FACOSH}, @var{float}<1
 6210: @cindex floating-point unidentified fault, @code{FACOSH}
 6211: @code{-55 throw} (Floating-point unidentified fault)
 6212: 
 6213: @item @var{float}=<-1 (@code{FLNP1}):
 6214: @cindex @code{FLNP1}, @var{float}=<-1
 6215: @cindex floating-point unidentified fault, @code{FLNP1}
 6216: @code{-55 throw} (Floating-point unidentified fault). On IEEE-FP systems
 6217: negative infinity is typically produced for @var{float}=-1.
 6218: 
 6219: @item @var{float}=<0 (@code{FLN}, @code{FLOG}):
 6220: @cindex @code{FLN}, @var{float}=<0
 6221: @cindex @code{FLOG}, @var{float}=<0
 6222: @cindex floating-point unidentified fault, @code{FLN} or @code{FLOG}
 6223: @code{-55 throw} (Floating-point unidentified fault). On IEEE-FP systems
 6224: negative infinity is typically produced for @var{float}=0.
 6225: 
 6226: @item @var{float}<0 (@code{FASINH}, @code{FSQRT}):
 6227: @cindex @code{FASINH}, @var{float}<0
 6228: @cindex @code{FSQRT}, @var{float}<0
 6229: @cindex floating-point unidentified fault, @code{FASINH} or @code{FSQRT}
 6230: @code{-55 throw} (Floating-point unidentified fault). @code{fasinh}
 6231: produces values for these inputs on my Linux box (Bug in the C library?)
 6232: 
 6233: @item |@var{float}|>1 (@code{FACOS}, @code{FASIN}, @code{FATANH}):
 6234: @cindex @code{FACOS}, |@var{float}|>1
 6235: @cindex @code{FASIN}, |@var{float}|>1
 6236: @cindex @code{FATANH}, |@var{float}|>1
 6237: @cindex floating-point unidentified fault, @code{FACOS}, @code{FASIN} or @code{FATANH}
 6238: @code{-55 throw} (Floating-point unidentified fault).
 6239: 
 6240: @item integer part of float cannot be represented by @var{d} in @code{F>D}:
 6241: @cindex @code{F>D}, integer part of float cannot be represented by @var{d}
 6242: @cindex floating-point unidentified fault, @code{F>D}
 6243: @code{-55 throw} (Floating-point unidentified fault).
 6244: 
 6245: @item string larger than pictured numeric output area (@code{f.}, @code{fe.}, @code{fs.}):
 6246: @cindex string larger than pictured numeric output area (@code{f.}, @code{fe.}, @code{fs.})
 6247: This does not happen.
 6248: @end table
 6249: 
 6250: @c =====================================================================
 6251: @node  The optional Locals word set, The optional Memory-Allocation word set, The optional Floating-Point word set, ANS conformance
 6252: @section The optional Locals word set
 6253: @c =====================================================================
 6254: @cindex system documentation, locals words
 6255: @cindex locals words, system documentation
 6256: 
 6257: @menu
 6258: * locals-idef::                 Implementation Defined Options                 
 6259: * locals-ambcond::              Ambiguous Conditions              
 6260: @end menu
 6261: 
 6262: 
 6263: @c ---------------------------------------------------------------------
 6264: @node locals-idef, locals-ambcond, The optional Locals word set, The optional Locals word set
 6265: @subsection Implementation Defined Options
 6266: @c ---------------------------------------------------------------------
 6267: @cindex implementation-defined options, locals words
 6268: @cindex locals words, implementation-defined options
 6269: 
 6270: @table @i
 6271: @item maximum number of locals in a definition:
 6272: @cindex maximum number of locals in a definition
 6273: @cindex locals, maximum number in a definition
 6274: @code{s" #locals" environment? drop .}. Currently 15. This is a lower
 6275: bound, e.g., on a 32-bit machine there can be 41 locals of up to 8
 6276: characters. The number of locals in a definition is bounded by the size
 6277: of locals-buffer, which contains the names of the locals.
 6278: 
 6279: @end table
 6280: 
 6281: 
 6282: @c ---------------------------------------------------------------------
 6283: @node locals-ambcond,  , locals-idef, The optional Locals word set
 6284: @subsection Ambiguous conditions
 6285: @c ---------------------------------------------------------------------
 6286: @cindex locals words, ambiguous conditions
 6287: @cindex ambiguous conditions, locals words
 6288: 
 6289: @table @i
 6290: @item executing a named local in interpretation state:
 6291: @cindex local in interpretation state
 6292: @cindex Interpreting a compile-only word, for a local
 6293: Locals have no interpretation semantics. If you try to perform the
 6294: interpretation semantics, you will get a @code{-14 throw} somewhere
 6295: (Interpreting a compile-only word). If you perform the compilation
 6296: semantics, the locals access will be compiled (irrespective of state).
 6297: 
 6298: @item @var{name} not defined by @code{VALUE} or @code{(LOCAL)} (@code{TO}):
 6299: @cindex name not defined by @code{VALUE} or @code{(LOCAL)} used by @code{TO}
 6300: @cindex @code{TO} on non-@code{VALUE}s and non-locals
 6301: @cindex Invalid name argument, @code{TO}
 6302: @code{-32 throw} (Invalid name argument)
 6303: 
 6304: @end table
 6305: 
 6306: 
 6307: @c =====================================================================
 6308: @node  The optional Memory-Allocation word set, The optional Programming-Tools word set, The optional Locals word set, ANS conformance
 6309: @section The optional Memory-Allocation word set
 6310: @c =====================================================================
 6311: @cindex system documentation, memory-allocation words
 6312: @cindex memory-allocation words, system documentation
 6313: 
 6314: @menu
 6315: * memory-idef::                 Implementation Defined Options                 
 6316: @end menu
 6317: 
 6318: 
 6319: @c ---------------------------------------------------------------------
 6320: @node memory-idef,  , The optional Memory-Allocation word set, The optional Memory-Allocation word set
 6321: @subsection Implementation Defined Options
 6322: @c ---------------------------------------------------------------------
 6323: @cindex implementation-defined options, memory-allocation words
 6324: @cindex memory-allocation words, implementation-defined options
 6325: 
 6326: @table @i
 6327: @item values and meaning of @var{ior}:
 6328: @cindex  @var{ior} values and meaning
 6329: The @var{ior}s returned by the file and memory allocation words are
 6330: intended as throw codes. They typically are in the range
 6331: -512@minus{}-2047 of OS errors.  The mapping from OS error numbers to
 6332: @var{ior}s is -512@minus{}@var{errno}.
 6333: 
 6334: @end table
 6335: 
 6336: @c =====================================================================
 6337: @node  The optional Programming-Tools word set, The optional Search-Order word set, The optional Memory-Allocation word set, ANS conformance
 6338: @section The optional Programming-Tools word set
 6339: @c =====================================================================
 6340: @cindex system documentation, programming-tools words
 6341: @cindex programming-tools words, system documentation
 6342: 
 6343: @menu
 6344: * programming-idef::            Implementation Defined Options            
 6345: * programming-ambcond::         Ambiguous Conditions         
 6346: @end menu
 6347: 
 6348: 
 6349: @c ---------------------------------------------------------------------
 6350: @node programming-idef, programming-ambcond, The optional Programming-Tools word set, The optional Programming-Tools word set
 6351: @subsection Implementation Defined Options
 6352: @c ---------------------------------------------------------------------
 6353: @cindex implementation-defined options, programming-tools words
 6354: @cindex programming-tools words, implementation-defined options
 6355: 
 6356: @table @i
 6357: @item ending sequence for input following @code{;CODE} and @code{CODE}:
 6358: @cindex @code{;CODE} ending sequence
 6359: @cindex @code{CODE} ending sequence
 6360: @code{END-CODE}
 6361: 
 6362: @item manner of processing input following @code{;CODE} and @code{CODE}:
 6363: @cindex @code{;CODE}, processing input
 6364: @cindex @code{CODE}, processing input
 6365: The @code{ASSEMBLER} vocabulary is pushed on the search order stack, and
 6366: the input is processed by the text interpreter, (starting) in interpret
 6367: state.
 6368: 
 6369: @item search order capability for @code{EDITOR} and @code{ASSEMBLER}:
 6370: @cindex @code{ASSEMBLER}, search order capability
 6371: The ANS Forth search order word set.
 6372: 
 6373: @item source and format of display by @code{SEE}:
 6374: @cindex @code{SEE}, source and format of output
 6375: The source for @code{see} is the intermediate code used by the inner
 6376: interpreter.  The current @code{see} tries to output Forth source code
 6377: as well as possible.
 6378: 
 6379: @end table
 6380: 
 6381: @c ---------------------------------------------------------------------
 6382: @node programming-ambcond,  , programming-idef, The optional Programming-Tools word set
 6383: @subsection Ambiguous conditions
 6384: @c ---------------------------------------------------------------------
 6385: @cindex programming-tools words, ambiguous conditions
 6386: @cindex ambiguous conditions, programming-tools words
 6387: 
 6388: @table @i
 6389: 
 6390: @item deleting the compilation wordlist (@code{FORGET}):
 6391: @cindex @code{FORGET}, deleting the compilation wordlist
 6392: Not implemented (yet).
 6393: 
 6394: @item fewer than @var{u}+1 items on the control flow stack (@code{CS-PICK}, @code{CS-ROLL}):
 6395: @cindex @code{CS-PICK}, fewer than @var{u}+1 items on the control flow stack
 6396: @cindex @code{CS-ROLL}, fewer than @var{u}+1 items on the control flow stack
 6397: @cindex control-flow stack underflow
 6398: This typically results in an @code{abort"} with a descriptive error
 6399: message (may change into a @code{-22 throw} (Control structure mismatch)
 6400: in the future). You may also get a memory access error. If you are
 6401: unlucky, this ambiguous condition is not caught.
 6402: 
 6403: @item @var{name} can't be found (@code{FORGET}):
 6404: @cindex @code{FORGET}, @var{name} can't be found
 6405: Not implemented (yet).
 6406: 
 6407: @item @var{name} not defined via @code{CREATE}:
 6408: @cindex @code{;CODE}, @var{name} not defined via @code{CREATE}
 6409: @code{;CODE} behaves like @code{DOES>} in this respect, i.e., it changes
 6410: the execution semantics of the last defined word no matter how it was
 6411: defined.
 6412: 
 6413: @item @code{POSTPONE} applied to @code{[IF]}:
 6414: @cindex @code{POSTPONE} applied to @code{[IF]}
 6415: @cindex @code{[IF]} and @code{POSTPONE}
 6416: After defining @code{: X POSTPONE [IF] ; IMMEDIATE}. @code{X} is
 6417: equivalent to @code{[IF]}.
 6418: 
 6419: @item reaching the end of the input source before matching @code{[ELSE]} or @code{[THEN]}:
 6420: @cindex @code{[IF]}, end of the input source before matching @code{[ELSE]} or @code{[THEN]}
 6421: Continue in the same state of conditional compilation in the next outer
 6422: input source. Currently there is no warning to the user about this.
 6423: 
 6424: @item removing a needed definition (@code{FORGET}):
 6425: @cindex @code{FORGET}, removing a needed definition
 6426: Not implemented (yet).
 6427: 
 6428: @end table
 6429: 
 6430: 
 6431: @c =====================================================================
 6432: @node  The optional Search-Order word set,  , The optional Programming-Tools word set, ANS conformance
 6433: @section The optional Search-Order word set
 6434: @c =====================================================================
 6435: @cindex system documentation, search-order words
 6436: @cindex search-order words, system documentation
 6437: 
 6438: @menu
 6439: * search-idef::                 Implementation Defined Options                 
 6440: * search-ambcond::              Ambiguous Conditions              
 6441: @end menu
 6442: 
 6443: 
 6444: @c ---------------------------------------------------------------------
 6445: @node search-idef, search-ambcond, The optional Search-Order word set, The optional Search-Order word set
 6446: @subsection Implementation Defined Options
 6447: @c ---------------------------------------------------------------------
 6448: @cindex implementation-defined options, search-order words
 6449: @cindex search-order words, implementation-defined options
 6450: 
 6451: @table @i
 6452: @item maximum number of word lists in search order:
 6453: @cindex maximum number of word lists in search order
 6454: @cindex search order, maximum depth
 6455: @code{s" wordlists" environment? drop .}. Currently 16.
 6456: 
 6457: @item minimum search order:
 6458: @cindex minimum search order
 6459: @cindex search order, minimum
 6460: @code{root root}.
 6461: 
 6462: @end table
 6463: 
 6464: @c ---------------------------------------------------------------------
 6465: @node search-ambcond,  , search-idef, The optional Search-Order word set
 6466: @subsection Ambiguous conditions
 6467: @c ---------------------------------------------------------------------
 6468: @cindex search-order words, ambiguous conditions
 6469: @cindex ambiguous conditions, search-order words
 6470: 
 6471: @table @i
 6472: @item changing the compilation wordlist (during compilation):
 6473: @cindex changing the compilation wordlist (during compilation)
 6474: @cindex compilation wordlist, change before definition ends
 6475: The word is entered into the wordlist that was the compilation wordlist
 6476: at the start of the definition. Any changes to the name field (e.g.,
 6477: @code{immediate}) or the code field (e.g., when executing @code{DOES>})
 6478: are applied to the latest defined word (as reported by @code{last} or
 6479: @code{lastxt}), if possible, irrespective of the compilation wordlist.
 6480: 
 6481: @item search order empty (@code{previous}):
 6482: @cindex @code{previous}, search order empty
 6483: @cindex Vocstack empty, @code{previous}
 6484: @code{abort" Vocstack empty"}.
 6485: 
 6486: @item too many word lists in search order (@code{also}):
 6487: @cindex @code{also}, too many word lists in search order
 6488: @cindex Vocstack full, @code{also}
 6489: @code{abort" Vocstack full"}.
 6490: 
 6491: @end table
 6492: 
 6493: @c ***************************************************************
 6494: @node Model, Integrating Gforth, ANS conformance, Top
 6495: @chapter Model
 6496: 
 6497: This chapter has yet to be written. It will contain information, on
 6498: which internal structures you can rely.
 6499: 
 6500: @c ***************************************************************
 6501: @node Integrating Gforth, Emacs and Gforth, Model, Top
 6502: @chapter Integrating Gforth into C programs
 6503: 
 6504: This is not yet implemented.
 6505: 
 6506: Several people like to use Forth as scripting language for applications
 6507: that are otherwise written in C, C++, or some other language.
 6508: 
 6509: The Forth system ATLAST provides facilities for embedding it into
 6510: applications; unfortunately it has several disadvantages: most
 6511: importantly, it is not based on ANS Forth, and it is apparently dead
 6512: (i.e., not developed further and not supported). The facilities
 6513: provided by Gforth in this area are inspired by ATLASTs facilities, so
 6514: making the switch should not be hard.
 6515: 
 6516: We also tried to design the interface such that it can easily be
 6517: implemented by other Forth systems, so that we may one day arrive at a
 6518: standardized interface. Such a standard interface would allow you to
 6519: replace the Forth system without having to rewrite C code.
 6520: 
 6521: You embed the Gforth interpreter by linking with the library
 6522: @code{libgforth.a} (give the compiler the option @code{-lgforth}).  All
 6523: global symbols in this library that belong to the interface, have the
 6524: prefix @code{forth_}. (Global symbols that are used internally have the
 6525: prefix @code{gforth_}).
 6526: 
 6527: You can include the declarations of Forth types and the functions and
 6528: variables of the interface with @code{#include <forth.h>}.
 6529: 
 6530: Types.
 6531: 
 6532: Variables.
 6533: 
 6534: Data and FP Stack pointer. Area sizes.
 6535: 
 6536: functions.
 6537: 
 6538: forth_init(imagefile)
 6539: forth_evaluate(string) exceptions?
 6540: forth_goto(address) (or forth_execute(xt)?)
 6541: forth_continue() (a corountining mechanism)
 6542: 
 6543: Adding primitives.
 6544: 
 6545: No checking.
 6546: 
 6547: Signals?
 6548: 
 6549: Accessing the Stacks
 6550: 
 6551: @node Emacs and Gforth, Image Files, Integrating Gforth, Top
 6552: @chapter Emacs and Gforth
 6553: @cindex Emacs and Gforth
 6554: 
 6555: @cindex @file{gforth.el}
 6556: @cindex @file{forth.el}
 6557: @cindex Rydqvist, Goran
 6558: @cindex comment editing commands
 6559: @cindex @code{\}, editing with Emacs
 6560: @cindex debug tracer editing commands
 6561: @cindex @code{~~}, removal with Emacs
 6562: @cindex Forth mode in Emacs
 6563: Gforth comes with @file{gforth.el}, an improved version of
 6564: @file{forth.el} by Goran Rydqvist (included in the TILE package). The
 6565: improvements are a better (but still not perfect) handling of
 6566: indentation. I have also added comment paragraph filling (@kbd{M-q}),
 6567: commenting (@kbd{C-x \}) and uncommenting (@kbd{C-u C-x \}) regions and
 6568: removing debugging tracers (@kbd{C-x ~}, @pxref{Debugging}). I left the
 6569: stuff I do not use alone, even though some of it only makes sense for
 6570: TILE. To get a description of these features, enter Forth mode and type
 6571: @kbd{C-h m}.
 6572: 
 6573: @cindex source location of error or debugging output in Emacs
 6574: @cindex error output, finding the source location in Emacs
 6575: @cindex debugging output, finding the source location in Emacs
 6576: In addition, Gforth supports Emacs quite well: The source code locations
 6577: given in error messages, debugging output (from @code{~~}) and failed
 6578: assertion messages are in the right format for Emacs' compilation mode
 6579: (@pxref{Compilation, , Running Compilations under Emacs, emacs, Emacs
 6580: Manual}) so the source location corresponding to an error or other
 6581: message is only a few keystrokes away (@kbd{C-x `} for the next error,
 6582: @kbd{C-c C-c} for the error under the cursor).
 6583: 
 6584: @cindex @file{TAGS} file
 6585: @cindex @file{etags.fs}
 6586: @cindex viewing the source of a word in Emacs
 6587: Also, if you @code{include} @file{etags.fs}, a new @file{TAGS} file
 6588: (@pxref{Tags, , Tags Tables, emacs, Emacs Manual}) will be produced that
 6589: contains the definitions of all words defined afterwards. You can then
 6590: find the source for a word using @kbd{M-.}. Note that emacs can use
 6591: several tags files at the same time (e.g., one for the Gforth sources
 6592: and one for your program, @pxref{Select Tags Table,,Selecting a Tags
 6593: Table,emacs, Emacs Manual}). The TAGS file for the preloaded words is
 6594: @file{$(datadir)/gforth/$(VERSION)/TAGS} (e.g.,
 6595: @file{/usr/local/share/gforth/0.2.0/TAGS}).
 6596: 
 6597: @cindex @file{.emacs}
 6598: To get all these benefits, add the following lines to your @file{.emacs}
 6599: file:
 6600: 
 6601: @example
 6602: (autoload 'forth-mode "gforth.el")
 6603: (setq auto-mode-alist (cons '("\\.fs\\'" . forth-mode) auto-mode-alist))
 6604: @end example
 6605: 
 6606: @node Image Files, Engine, Emacs and Gforth, Top
 6607: @chapter Image Files
 6608: @cindex image files
 6609: @cindex @code{.fi} files
 6610: @cindex precompiled Forth code
 6611: @cindex dictionary in persistent form
 6612: @cindex persistent form of dictionary
 6613: 
 6614: An image file is a file containing an image of the Forth dictionary,
 6615: i.e., compiled Forth code and data residing in the dictionary.  By
 6616: convention, we use the extension @code{.fi} for image files.
 6617: 
 6618: @menu
 6619: * Image Licensing Issues::      Distribution terms for images.
 6620: * Image File Background::       Why have image files?
 6621: * Non-Relocatable Image Files::  don't always work.
 6622: * Data-Relocatable Image Files::  are better.
 6623: * Fully Relocatable Image Files::  better yet.
 6624: * Stack and Dictionary Sizes::  Setting the default sizes for an image.
 6625: * Running Image Files::         @code{gforth -i @var{file}} or @var{file}.
 6626: * Modifying the Startup Sequence::  and turnkey applications.
 6627: @end menu
 6628: 
 6629: @node Image Licensing Issues, Image File Background, Image Files, Image Files
 6630: @section Image Licensing Issues
 6631: @cindex license for images
 6632: @cindex image license
 6633: 
 6634: An image created with @code{gforthmi} (@pxref{gforthmi}) or
 6635: @code{savesystem} (@pxref{Non-Relocatable Image Files}) includes the
 6636: original image; i.e., according to copyright law it is a derived work of
 6637: the original image.
 6638: 
 6639: Since Gforth is distributed under the GNU GPL, the newly created image
 6640: falls under the GNU GPL, too. In particular, this means that if you
 6641: distribute the image, you have to make all of the sources for the image
 6642: available, including those you wrote.  For details see @ref{License, ,
 6643: GNU General Public License (Section 3)}.
 6644: 
 6645: If you create an image with @code{cross} (@pxref{cross.fs}), the image
 6646: contains only code compiled from the sources you gave it; if none of
 6647: these sources is under the GPL, the terms discussed above do not apply
 6648: to the image. However, if your image needs an engine (a gforth binary)
 6649: that is under the GPL, you should make sure that you distribute both in
 6650: a way that is at most a @emph{mere aggregation}, if you don't want the
 6651: terms of the GPL to apply to the image.
 6652: 
 6653: @node Image File Background, Non-Relocatable Image Files, Image Licensing Issues, Image Files
 6654: @section Image File Background
 6655: @cindex image file background
 6656: 
 6657: Our Forth system consists not only of primitives, but also of
 6658: definitions written in Forth. Since the Forth compiler itself belongs to
 6659: those definitions, it is not possible to start the system with the
 6660: primitives and the Forth source alone. Therefore we provide the Forth
 6661: code as an image file in nearly executable form. At the start of the
 6662: system a C routine loads the image file into memory, optionally
 6663: relocates the addresses, then sets up the memory (stacks etc.) according
 6664: to information in the image file, and starts executing Forth code.
 6665: 
 6666: The image file variants represent different compromises between the
 6667: goals of making it easy to generate image files and making them
 6668: portable.
 6669: 
 6670: @cindex relocation at run-time
 6671: Win32Forth 3.4 and Mitch Bradleys @code{cforth} use relocation at
 6672: run-time. This avoids many of the complications discussed below (image
 6673: files are data relocatable without further ado), but costs performance
 6674: (one addition per memory access).
 6675: 
 6676: @cindex relocation at load-time
 6677: By contrast, our loader performs relocation at image load time. The
 6678: loader also has to replace tokens standing for primitive calls with the
 6679: appropriate code-field addresses (or code addresses in the case of
 6680: direct threading).
 6681: 
 6682: There are three kinds of image files, with different degrees of
 6683: relocatability: non-relocatable, data-relocatable, and fully relocatable
 6684: image files.
 6685: 
 6686: @cindex image file loader
 6687: @cindex relocating loader
 6688: @cindex loader for image files
 6689: These image file variants have several restrictions in common; they are
 6690: caused by the design of the image file loader:
 6691: 
 6692: @itemize @bullet
 6693: @item
 6694: There is only one segment; in particular, this means, that an image file
 6695: cannot represent @code{ALLOCATE}d memory chunks (and pointers to
 6696: them). And the contents of the stacks are not represented, either.
 6697: 
 6698: @item
 6699: The only kinds of relocation supported are: adding the same offset to
 6700: all cells that represent data addresses; and replacing special tokens
 6701: with code addresses or with pieces of machine code.
 6702: 
 6703: If any complex computations involving addresses are performed, the
 6704: results cannot be represented in the image file. Several applications that
 6705: use such computations come to mind:
 6706: @itemize @minus
 6707: @item
 6708: Hashing addresses (or data structures which contain addresses) for table
 6709: lookup. If you use Gforth's @code{table}s or @code{wordlist}s for this
 6710: purpose, you will have no problem, because the hash tables are
 6711: recomputed automatically when the system is started. If you use your own
 6712: hash tables, you will have to do something similar.
 6713: 
 6714: @item
 6715: There's a cute implementation of doubly-linked lists that uses
 6716: @code{XOR}ed addresses. You could represent such lists as singly-linked
 6717: in the image file, and restore the doubly-linked representation on
 6718: startup.@footnote{In my opinion, though, you should think thrice before
 6719: using a doubly-linked list (whatever implementation).}
 6720: 
 6721: @item
 6722: The code addresses of run-time routines like @code{docol:} cannot be
 6723: represented in the image file (because their tokens would be replaced by
 6724: machine code in direct threaded implementations). As a workaround,
 6725: compute these addresses at run-time with @code{>code-address} from the
 6726: executions tokens of appropriate words (see the definitions of
 6727: @code{docol:} and friends in @file{kernel.fs}).
 6728: 
 6729: @item
 6730: On many architectures addresses are represented in machine code in some
 6731: shifted or mangled form. You cannot put @code{CODE} words that contain
 6732: absolute addresses in this form in a relocatable image file. Workarounds
 6733: are representing the address in some relative form (e.g., relative to
 6734: the CFA, which is present in some register), or loading the address from
 6735: a place where it is stored in a non-mangled form.
 6736: @end itemize
 6737: @end itemize
 6738: 
 6739: @node  Non-Relocatable Image Files, Data-Relocatable Image Files, Image File Background, Image Files
 6740: @section Non-Relocatable Image Files
 6741: @cindex non-relocatable image files
 6742: @cindex image files, non-relocatable
 6743: 
 6744: These files are simple memory dumps of the dictionary. They are specific
 6745: to the executable (i.e., @file{gforth} file) they were created
 6746: with. What's worse, they are specific to the place on which the
 6747: dictionary resided when the image was created. Now, there is no
 6748: guarantee that the dictionary will reside at the same place the next
 6749: time you start Gforth, so there's no guarantee that a non-relocatable
 6750: image will work the next time (Gforth will complain instead of crashing,
 6751: though).
 6752: 
 6753: You can create a non-relocatable image file with
 6754: 
 6755: doc-savesystem
 6756: 
 6757: @node Data-Relocatable Image Files, Fully Relocatable Image Files, Non-Relocatable Image Files, Image Files
 6758: @section Data-Relocatable Image Files
 6759: @cindex data-relocatable image files
 6760: @cindex image files, data-relocatable
 6761: 
 6762: These files contain relocatable data addresses, but fixed code addresses
 6763: (instead of tokens). They are specific to the executable (i.e.,
 6764: @file{gforth} file) they were created with. For direct threading on some
 6765: architectures (e.g., the i386), data-relocatable images do not work. You
 6766: get a data-relocatable image, if you use @file{gforthmi} with a
 6767: Gforth binary that is not doubly indirect threaded (@pxref{Fully
 6768: Relocatable Image Files}).
 6769: 
 6770: @node Fully Relocatable Image Files, Stack and Dictionary Sizes, Data-Relocatable Image Files, Image Files
 6771: @section Fully Relocatable Image Files
 6772: @cindex fully relocatable image files
 6773: @cindex image files, fully relocatable
 6774: 
 6775: @cindex @file{kern*.fi}, relocatability
 6776: @cindex @file{gforth.fi}, relocatability
 6777: These image files have relocatable data addresses, and tokens for code
 6778: addresses. They can be used with different binaries (e.g., with and
 6779: without debugging) on the same machine, and even across machines with
 6780: the same data formats (byte order, cell size, floating point
 6781: format). However, they are usually specific to the version of Gforth
 6782: they were created with. The files @file{gforth.fi} and @file{kernl*.fi}
 6783: are fully relocatable.
 6784: 
 6785: There are two ways to create a fully relocatable image file:
 6786: 
 6787: @menu
 6788: * gforthmi::            The normal way
 6789: * cross.fs::                    The hard way
 6790: @end menu
 6791: 
 6792: @node gforthmi, cross.fs, Fully Relocatable Image Files, Fully Relocatable Image Files
 6793: @subsection @file{gforthmi}
 6794: @cindex @file{comp-i.fs}
 6795: @cindex @file{gforthmi}
 6796: 
 6797: You will usually use @file{gforthmi}. If you want to create an
 6798: image @var{file} that contains everything you would load by invoking
 6799: Gforth with @code{gforth @var{options}}, you simply say
 6800: @example
 6801: gforthmi @var{file} @var{options}
 6802: @end example
 6803: 
 6804: E.g., if you want to create an image @file{asm.fi} that has the file
 6805: @file{asm.fs} loaded in addition to the usual stuff, you could do it
 6806: like this:
 6807: 
 6808: @example
 6809: gforthmi asm.fi asm.fs
 6810: @end example
 6811: 
 6812: @file{gforthmi} works like this: It produces two non-relocatable
 6813: images for different addresses and then compares them. Its output
 6814: reflects this: first you see the output (if any) of the two Gforth
 6815: invocations that produce the nonrelocatable image files, then you see
 6816: the output of the comparing program: It displays the offset used for
 6817: data addresses and the offset used for code addresses;
 6818: moreover, for each cell that cannot be represented correctly in the
 6819: image files, it displays a line like the following one:
 6820: 
 6821: @example
 6822:      78DC         BFFFFA50         BFFFFA40
 6823: @end example
 6824: 
 6825: This means that at offset $78dc from @code{forthstart}, one input image
 6826: contains $bffffa50, and the other contains $bffffa40. Since these cells
 6827: cannot be represented correctly in the output image, you should examine
 6828: these places in the dictionary and verify that these cells are dead
 6829: (i.e., not read before they are written).
 6830: 
 6831: @cindex @code{savesystem} during @file{gforthmi}
 6832: @cindex @code{bye} during @file{gforthmi}
 6833: @cindex doubly indirect threaded code
 6834: @cindex environment variable @code{GFORTHD}
 6835: @cindex @code{GFORTHD} environment variable
 6836: @cindex @code{gforth-ditc}
 6837: There are a few wrinkles: After processing the passed @var{options}, the
 6838: words @code{savesystem} and @code{bye} must be visible. A special doubly
 6839: indirect threaded version of the @file{gforth} executable is used for
 6840: creating the nonrelocatable images; you can pass the exact filename of
 6841: this executable through the environment variable @code{GFORTHD}
 6842: (default: @file{gforth-ditc}); if you pass a version that is not doubly
 6843: indirect threaded, you will not get a fully relocatable image, but a
 6844: data-relocatable image (because there is no code address offset).
 6845: 
 6846: @node cross.fs,  , gforthmi, Fully Relocatable Image Files
 6847: @subsection @file{cross.fs}
 6848: @cindex @file{cross.fs}
 6849: @cindex cross-compiler
 6850: @cindex metacompiler
 6851: 
 6852: You can also use @code{cross}, a batch compiler that accepts a Forth-like
 6853: programming language. This @code{cross} language has to be documented
 6854: yet.
 6855: 
 6856: @cindex target compiler
 6857: @code{cross} also allows you to create image files for machines with
 6858: different data sizes and data formats than the one used for generating
 6859: the image file. You can also use it to create an application image that
 6860: does not contain a Forth compiler. These features are bought with
 6861: restrictions and inconveniences in programming. E.g., addresses have to
 6862: be stored in memory with special words (@code{A!}, @code{A,}, etc.) in
 6863: order to make the code relocatable.
 6864: 
 6865: 
 6866: @node Stack and Dictionary Sizes, Running Image Files, Fully Relocatable Image Files, Image Files
 6867: @section Stack and Dictionary Sizes
 6868: @cindex image file, stack and dictionary sizes
 6869: @cindex dictionary size default
 6870: @cindex stack size default
 6871: 
 6872: If you invoke Gforth with a command line flag for the size
 6873: (@pxref{Invoking Gforth}), the size you specify is stored in the
 6874: dictionary. If you save the dictionary with @code{savesystem} or create
 6875: an image with @file{gforthmi}, this size will become the default
 6876: for the resulting image file. E.g., the following will create a
 6877: fully relocatable version of gforth.fi with a 1MB dictionary:
 6878: 
 6879: @example
 6880: gforthmi gforth.fi -m 1M
 6881: @end example
 6882: 
 6883: In other words, if you want to set the default size for the dictionary
 6884: and the stacks of an image, just invoke @file{gforthmi} with the
 6885: appropriate options when creating the image.
 6886: 
 6887: @cindex stack size, cache-friendly
 6888: Note: For cache-friendly behaviour (i.e., good performance), you should
 6889: make the sizes of the stacks modulo, say, 2K, somewhat different. E.g.,
 6890: the default stack sizes are: data: 16k (mod 2k=0); fp: 15.5k (mod
 6891: 2k=1.5k); return: 15k(mod 2k=1k); locals: 14.5k (mod 2k=0.5k).
 6892: 
 6893: @node Running Image Files, Modifying the Startup Sequence, Stack and Dictionary Sizes, Image Files
 6894: @section Running Image Files
 6895: @cindex running image files
 6896: @cindex invoking image files
 6897: @cindex image file invocation
 6898: 
 6899: @cindex -i, invoke image file
 6900: @cindex --image file, invoke image file
 6901: You can invoke Gforth with an image file @var{image} instead of the
 6902: default @file{gforth.fi} with the @code{-i} flag (@pxref{Invoking Gforth}):
 6903: @example
 6904: gforth -i @var{image}
 6905: @end example
 6906: 
 6907: @cindex executable image file
 6908: @cindex image files, executable
 6909: If your operating system supports starting scripts with a line of the
 6910: form @code{#! ...}, you just have to type the image file name to start
 6911: Gforth with this image file (note that the file extension @code{.fi} is
 6912: just a convention). I.e., to run Gforth with the image file @var{image},
 6913: you can just type @var{image} instead of @code{gforth -i @var{image}}.
 6914: 
 6915: doc-#!
 6916: 
 6917: @node Modifying the Startup Sequence,  , Running Image Files, Image Files
 6918: @section Modifying the Startup Sequence
 6919: @cindex startup sequence for image file
 6920: @cindex image file initialization sequence
 6921: @cindex initialization sequence of image file
 6922: 
 6923: You can add your own initialization to the startup sequence through the
 6924: deferred word
 6925: 
 6926: doc-'cold
 6927: 
 6928: @code{'cold} is invoked just before the image-specific command line
 6929: processing (by default, loading files and evaluating (@code{-e}) strings)
 6930: starts.
 6931: 
 6932: A sequence for adding your initialization usually looks like this:
 6933: 
 6934: @example
 6935: :noname
 6936:     Defers 'cold \ do other initialization stuff (e.g., rehashing wordlists)
 6937:     ... \ your stuff
 6938: ; IS 'cold
 6939: @end example
 6940: 
 6941: @cindex turnkey image files
 6942: @cindex image files, turnkey applications
 6943: You can make a turnkey image by letting @code{'cold} execute a word
 6944: (your turnkey application) that never returns; instead, it exits Gforth
 6945: via @code{bye} or @code{throw}.
 6946: 
 6947: @cindex command-line arguments, access
 6948: @cindex arguments on the command line, access
 6949: You can access the (image-specific) command-line arguments through the
 6950: variables @code{argc} and @code{argv}. @code{arg} provides conventient
 6951: access to @code{argv}.
 6952: 
 6953: doc-argc
 6954: doc-argv
 6955: doc-arg
 6956: 
 6957: If @code{'cold} exits normally, Gforth processes the command-line
 6958: arguments as files to be loaded and strings to be evaluated.  Therefore,
 6959: @code{'cold} should remove the arguments it has used in this case.
 6960: 
 6961: @c ******************************************************************
 6962: @node Engine, Binding to System Library, Image Files, Top
 6963: @chapter Engine
 6964: @cindex engine
 6965: @cindex virtual machine
 6966: 
 6967: Reading this section is not necessary for programming with Gforth. It
 6968: may be helpful for finding your way in the Gforth sources.
 6969: 
 6970: The ideas in this section have also been published in the papers
 6971: @cite{ANS fig/GNU/??? Forth} (in German) by Bernd Paysan, presented at
 6972: the Forth-Tagung '93 and @cite{A Portable Forth Engine} by M. Anton
 6973: Ertl, presented at EuroForth '93; the latter is available at
 6974: @*@url{http://www.complang.tuwien.ac.at/papers/ertl93.ps.Z}.
 6975: 
 6976: @menu
 6977: * Portability::                 
 6978: * Threading::                   
 6979: * Primitives::                  
 6980: * Performance::                 
 6981: @end menu
 6982: 
 6983: @node Portability, Threading, Engine, Engine
 6984: @section Portability
 6985: @cindex engine portability
 6986: 
 6987: One of the main goals of the effort is availability across a wide range
 6988: of personal machines. fig-Forth, and, to a lesser extent, F83, achieved
 6989: this goal by manually coding the engine in assembly language for several
 6990: then-popular processors. This approach is very labor-intensive and the
 6991: results are short-lived due to progress in computer architecture.
 6992: 
 6993: @cindex C, using C for the engine
 6994: Others have avoided this problem by coding in C, e.g., Mitch Bradley
 6995: (cforth), Mikael Patel (TILE) and Dirk Zoller (pfe). This approach is
 6996: particularly popular for UNIX-based Forths due to the large variety of
 6997: architectures of UNIX machines. Unfortunately an implementation in C
 6998: does not mix well with the goals of efficiency and with using
 6999: traditional techniques: Indirect or direct threading cannot be expressed
 7000: in C, and switch threading, the fastest technique available in C, is
 7001: significantly slower. Another problem with C is that it is very
 7002: cumbersome to express double integer arithmetic.
 7003: 
 7004: @cindex GNU C for the engine
 7005: @cindex long long
 7006: Fortunately, there is a portable language that does not have these
 7007: limitations: GNU C, the version of C processed by the GNU C compiler
 7008: (@pxref{C Extensions, , Extensions to the C Language Family, gcc.info,
 7009: GNU C Manual}). Its labels as values feature (@pxref{Labels as Values, ,
 7010: Labels as Values, gcc.info, GNU C Manual}) makes direct and indirect
 7011: threading possible, its @code{long long} type (@pxref{Long Long, ,
 7012: Double-Word Integers, gcc.info, GNU C Manual}) corresponds to Forth's
 7013: double numbers@footnote{Unfortunately, long longs are not implemented
 7014: properly on all machines (e.g., on alpha-osf1, long longs are only 64
 7015: bits, the same size as longs (and pointers), but they should be twice as
 7016: long according to @pxref{Long Long, , Double-Word Integers, gcc.info, GNU
 7017: C Manual}). So, we had to implement doubles in C after all. Still, on
 7018: most machines we can use long longs and achieve better performance than
 7019: with the emulation package.}. GNU C is available for free on all
 7020: important (and many unimportant) UNIX machines, VMS, 80386s running
 7021: MS-DOS, the Amiga, and the Atari ST, so a Forth written in GNU C can run
 7022: on all these machines.
 7023: 
 7024: Writing in a portable language has the reputation of producing code that
 7025: is slower than assembly. For our Forth engine we repeatedly looked at
 7026: the code produced by the compiler and eliminated most compiler-induced
 7027: inefficiencies by appropriate changes in the source code.
 7028: 
 7029: @cindex explicit register declarations
 7030: @cindex --enable-force-reg, configuration flag
 7031: @cindex -DFORCE_REG
 7032: However, register allocation cannot be portably influenced by the
 7033: programmer, leading to some inefficiencies on register-starved
 7034: machines. We use explicit register declarations (@pxref{Explicit Reg
 7035: Vars, , Variables in Specified Registers, gcc.info, GNU C Manual}) to
 7036: improve the speed on some machines. They are turned on by using the
 7037: configuration flag @code{--enable-force-reg} (@code{gcc} switch
 7038: @code{-DFORCE_REG}). Unfortunately, this feature not only depends on the
 7039: machine, but also on the compiler version: On some machines some
 7040: compiler versions produce incorrect code when certain explicit register
 7041: declarations are used. So by default @code{-DFORCE_REG} is not used.
 7042: 
 7043: @node Threading, Primitives, Portability, Engine
 7044: @section Threading
 7045: @cindex inner interpreter implementation
 7046: @cindex threaded code implementation
 7047: 
 7048: @cindex labels as values
 7049: GNU C's labels as values extension (available since @code{gcc-2.0},
 7050: @pxref{Labels as Values, , Labels as Values, gcc.info, GNU C Manual})
 7051: makes it possible to take the address of @var{label} by writing
 7052: @code{&&@var{label}}.  This address can then be used in a statement like
 7053: @code{goto *@var{address}}. I.e., @code{goto *&&x} is the same as
 7054: @code{goto x}.
 7055: 
 7056: @cindex NEXT, indirect threaded
 7057: @cindex indirect threaded inner interpreter
 7058: @cindex inner interpreter, indirect threaded
 7059: With this feature an indirect threaded NEXT looks like:
 7060: @example
 7061: cfa = *ip++;
 7062: ca = *cfa;
 7063: goto *ca;
 7064: @end example
 7065: @cindex instruction pointer
 7066: For those unfamiliar with the names: @code{ip} is the Forth instruction
 7067: pointer; the @code{cfa} (code-field address) corresponds to ANS Forths
 7068: execution token and points to the code field of the next word to be
 7069: executed; The @code{ca} (code address) fetched from there points to some
 7070: executable code, e.g., a primitive or the colon definition handler
 7071: @code{docol}.
 7072: 
 7073: @cindex NEXT, direct threaded
 7074: @cindex direct threaded inner interpreter
 7075: @cindex inner interpreter, direct threaded
 7076: Direct threading is even simpler:
 7077: @example
 7078: ca = *ip++;
 7079: goto *ca;
 7080: @end example
 7081: 
 7082: Of course we have packaged the whole thing neatly in macros called
 7083: @code{NEXT} and @code{NEXT1} (the part of NEXT after fetching the cfa).
 7084: 
 7085: @menu
 7086: * Scheduling::                  
 7087: * Direct or Indirect Threaded?::  
 7088: * DOES>::                       
 7089: @end menu
 7090: 
 7091: @node Scheduling, Direct or Indirect Threaded?, Threading, Threading
 7092: @subsection Scheduling
 7093: @cindex inner interpreter optimization
 7094: 
 7095: There is a little complication: Pipelined and superscalar processors,
 7096: i.e., RISC and some modern CISC machines can process independent
 7097: instructions while waiting for the results of an instruction. The
 7098: compiler usually reorders (schedules) the instructions in a way that
 7099: achieves good usage of these delay slots. However, on our first tries
 7100: the compiler did not do well on scheduling primitives. E.g., for
 7101: @code{+} implemented as
 7102: @example
 7103: n=sp[0]+sp[1];
 7104: sp++;
 7105: sp[0]=n;
 7106: NEXT;
 7107: @end example
 7108: the NEXT comes strictly after the other code, i.e., there is nearly no
 7109: scheduling. After a little thought the problem becomes clear: The
 7110: compiler cannot know that sp and ip point to different addresses (and
 7111: the version of @code{gcc} we used would not know it even if it was
 7112: possible), so it could not move the load of the cfa above the store to
 7113: the TOS. Indeed the pointers could be the same, if code on or very near
 7114: the top of stack were executed. In the interest of speed we chose to
 7115: forbid this probably unused ``feature'' and helped the compiler in
 7116: scheduling: NEXT is divided into the loading part (@code{NEXT_P1}) and
 7117: the goto part (@code{NEXT_P2}). @code{+} now looks like:
 7118: @example
 7119: n=sp[0]+sp[1];
 7120: sp++;
 7121: NEXT_P1;
 7122: sp[0]=n;
 7123: NEXT_P2;
 7124: @end example
 7125: This can be scheduled optimally by the compiler.
 7126: 
 7127: This division can be turned off with the switch @code{-DCISC_NEXT}. This
 7128: switch is on by default on machines that do not profit from scheduling
 7129: (e.g., the 80386), in order to preserve registers.
 7130: 
 7131: @node Direct or Indirect Threaded?, DOES>, Scheduling, Threading
 7132: @subsection Direct or Indirect Threaded?
 7133: @cindex threading, direct or indirect?
 7134: 
 7135: @cindex -DDIRECT_THREADED
 7136: Both! After packaging the nasty details in macro definitions we
 7137: realized that we could switch between direct and indirect threading by
 7138: simply setting a compilation flag (@code{-DDIRECT_THREADED}) and
 7139: defining a few machine-specific macros for the direct-threading case.
 7140: On the Forth level we also offer access words that hide the
 7141: differences between the threading methods (@pxref{Threading Words}).
 7142: 
 7143: Indirect threading is implemented completely machine-independently.
 7144: Direct threading needs routines for creating jumps to the executable
 7145: code (e.g. to docol or dodoes). These routines are inherently
 7146: machine-dependent, but they do not amount to many source lines. I.e.,
 7147: even porting direct threading to a new machine is a small effort.
 7148: 
 7149: @cindex --enable-indirect-threaded, configuration flag
 7150: @cindex --enable-direct-threaded, configuration flag
 7151: The default threading method is machine-dependent. You can enforce a
 7152: specific threading method when building Gforth with the configuration
 7153: flag @code{--enable-direct-threaded} or
 7154: @code{--enable-indirect-threaded}. Note that direct threading is not
 7155: supported on all machines.
 7156: 
 7157: @node DOES>,  , Direct or Indirect Threaded?, Threading
 7158: @subsection DOES>
 7159: @cindex @code{DOES>} implementation
 7160: 
 7161: @cindex dodoes routine
 7162: @cindex DOES-code
 7163: One of the most complex parts of a Forth engine is @code{dodoes}, i.e.,
 7164: the chunk of code executed by every word defined by a
 7165: @code{CREATE}...@code{DOES>} pair. The main problem here is: How to find
 7166: the Forth code to be executed, i.e. the code after the
 7167: @code{DOES>} (the DOES-code)? There are two solutions:
 7168: 
 7169: In fig-Forth the code field points directly to the dodoes and the
 7170: DOES-code address is stored in the cell after the code address (i.e. at
 7171: @code{@var{cfa} cell+}). It may seem that this solution is illegal in
 7172: the Forth-79 and all later standards, because in fig-Forth this address
 7173: lies in the body (which is illegal in these standards). However, by
 7174: making the code field larger for all words this solution becomes legal
 7175: again. We use this approach for the indirect threaded version and for
 7176: direct threading on some machines. Leaving a cell unused in most words
 7177: is a bit wasteful, but on the machines we are targeting this is hardly a
 7178: problem. The other reason for having a code field size of two cells is
 7179: to avoid having different image files for direct and indirect threaded
 7180: systems (direct threaded systems require two-cell code fields on many
 7181: machines).
 7182: 
 7183: @cindex DOES-handler
 7184: The other approach is that the code field points or jumps to the cell
 7185: after @code{DOES}. In this variant there is a jump to @code{dodoes} at
 7186: this address (the DOES-handler). @code{dodoes} can then get the
 7187: DOES-code address by computing the code address, i.e., the address of
 7188: the jump to dodoes, and add the length of that jump field. A variant of
 7189: this is to have a call to @code{dodoes} after the @code{DOES>}; then the
 7190: return address (which can be found in the return register on RISCs) is
 7191: the DOES-code address. Since the two cells available in the code field
 7192: are used up by the jump to the code address in direct threading on many
 7193: architectures, we use this approach for direct threading on these
 7194: architectures. We did not want to add another cell to the code field.
 7195: 
 7196: @node Primitives, Performance, Threading, Engine
 7197: @section Primitives
 7198: @cindex primitives, implementation
 7199: @cindex virtual machine instructions, implementation
 7200: 
 7201: @menu
 7202: * Automatic Generation::        
 7203: * TOS Optimization::            
 7204: * Produced code::               
 7205: @end menu
 7206: 
 7207: @node Automatic Generation, TOS Optimization, Primitives, Primitives
 7208: @subsection Automatic Generation
 7209: @cindex primitives, automatic generation
 7210: 
 7211: @cindex @file{prims2x.fs}
 7212: Since the primitives are implemented in a portable language, there is no
 7213: longer any need to minimize the number of primitives. On the contrary,
 7214: having many primitives has an advantage: speed. In order to reduce the
 7215: number of errors in primitives and to make programming them easier, we
 7216: provide a tool, the primitive generator (@file{prims2x.fs}), that
 7217: automatically generates most (and sometimes all) of the C code for a
 7218: primitive from the stack effect notation.  The source for a primitive
 7219: has the following form:
 7220: 
 7221: @cindex primitive source format
 7222: @format
 7223: @var{Forth-name}	@var{stack-effect}	@var{category}	[@var{pronounc.}]
 7224: [@code{""}@var{glossary entry}@code{""}]
 7225: @var{C code}
 7226: [@code{:}
 7227: @var{Forth code}]
 7228: @end format
 7229: 
 7230: The items in brackets are optional. The category and glossary fields
 7231: are there for generating the documentation, the Forth code is there
 7232: for manual implementations on machines without GNU C. E.g., the source
 7233: for the primitive @code{+} is:
 7234: @example
 7235: +    n1 n2 -- n    core    plus
 7236: n = n1+n2;
 7237: @end example
 7238: 
 7239: This looks like a specification, but in fact @code{n = n1+n2} is C
 7240: code. Our primitive generation tool extracts a lot of information from
 7241: the stack effect notations@footnote{We use a one-stack notation, even
 7242: though we have separate data and floating-point stacks; The separate
 7243: notation can be generated easily from the unified notation.}: The number
 7244: of items popped from and pushed on the stack, their type, and by what
 7245: name they are referred to in the C code. It then generates a C code
 7246: prelude and postlude for each primitive. The final C code for @code{+}
 7247: looks like this:
 7248: 
 7249: @example
 7250: I_plus:	/* + ( n1 n2 -- n ) */  /* label, stack effect */
 7251: /*  */                          /* documentation */
 7252: @{
 7253: DEF_CA                          /* definition of variable ca (indirect threading) */
 7254: Cell n1;                        /* definitions of variables */
 7255: Cell n2;
 7256: Cell n;
 7257: n1 = (Cell) sp[1];              /* input */
 7258: n2 = (Cell) TOS;
 7259: sp += 1;                        /* stack adjustment */
 7260: NAME("+")                       /* debugging output (with -DDEBUG) */
 7261: @{
 7262: n = n1+n2;                      /* C code taken from the source */
 7263: @}
 7264: NEXT_P1;                        /* NEXT part 1 */
 7265: TOS = (Cell)n;                  /* output */
 7266: NEXT_P2;                        /* NEXT part 2 */
 7267: @}
 7268: @end example
 7269: 
 7270: This looks long and inefficient, but the GNU C compiler optimizes quite
 7271: well and produces optimal code for @code{+} on, e.g., the R3000 and the
 7272: HP RISC machines: Defining the @code{n}s does not produce any code, and
 7273: using them as intermediate storage also adds no cost.
 7274: 
 7275: There are also other optimizations, that are not illustrated by this
 7276: example: Assignments between simple variables are usually for free (copy
 7277: propagation). If one of the stack items is not used by the primitive
 7278: (e.g.  in @code{drop}), the compiler eliminates the load from the stack
 7279: (dead code elimination). On the other hand, there are some things that
 7280: the compiler does not do, therefore they are performed by
 7281: @file{prims2x.fs}: The compiler does not optimize code away that stores
 7282: a stack item to the place where it just came from (e.g., @code{over}).
 7283: 
 7284: While programming a primitive is usually easy, there are a few cases
 7285: where the programmer has to take the actions of the generator into
 7286: account, most notably @code{?dup}, but also words that do not (always)
 7287: fall through to NEXT.
 7288: 
 7289: @node TOS Optimization, Produced code, Automatic Generation, Primitives
 7290: @subsection TOS Optimization
 7291: @cindex TOS optimization for primitives
 7292: @cindex primitives, keeping the TOS in a register
 7293: 
 7294: An important optimization for stack machine emulators, e.g., Forth
 7295: engines, is keeping  one or more of the top stack items in
 7296: registers.  If a word has the stack effect @var{in1}...@var{inx} @code{--}
 7297: @var{out1}...@var{outy}, keeping the top @var{n} items in registers
 7298: @itemize @bullet
 7299: @item
 7300: is better than keeping @var{n-1} items, if @var{x>=n} and @var{y>=n},
 7301: due to fewer loads from and stores to the stack.
 7302: @item is slower than keeping @var{n-1} items, if @var{x<>y} and @var{x<n} and
 7303: @var{y<n}, due to additional moves between registers.
 7304: @end itemize
 7305: 
 7306: @cindex -DUSE_TOS
 7307: @cindex -DUSE_NO_TOS
 7308: In particular, keeping one item in a register is never a disadvantage,
 7309: if there are enough registers. Keeping two items in registers is a
 7310: disadvantage for frequent words like @code{?branch}, constants,
 7311: variables, literals and @code{i}. Therefore our generator only produces
 7312: code that keeps zero or one items in registers. The generated C code
 7313: covers both cases; the selection between these alternatives is made at
 7314: C-compile time using the switch @code{-DUSE_TOS}. @code{TOS} in the C
 7315: code for @code{+} is just a simple variable name in the one-item case,
 7316: otherwise it is a macro that expands into @code{sp[0]}. Note that the
 7317: GNU C compiler tries to keep simple variables like @code{TOS} in
 7318: registers, and it usually succeeds, if there are enough registers.
 7319: 
 7320: @cindex -DUSE_FTOS
 7321: @cindex -DUSE_NO_FTOS
 7322: The primitive generator performs the TOS optimization for the
 7323: floating-point stack, too (@code{-DUSE_FTOS}). For floating-point
 7324: operations the benefit of this optimization is even larger:
 7325: floating-point operations take quite long on most processors, but can be
 7326: performed in parallel with other operations as long as their results are
 7327: not used. If the FP-TOS is kept in a register, this works. If
 7328: it is kept on the stack, i.e., in memory, the store into memory has to
 7329: wait for the result of the floating-point operation, lengthening the
 7330: execution time of the primitive considerably.
 7331: 
 7332: The TOS optimization makes the automatic generation of primitives a
 7333: bit more complicated. Just replacing all occurrences of @code{sp[0]} by
 7334: @code{TOS} is not sufficient. There are some special cases to
 7335: consider:
 7336: @itemize @bullet
 7337: @item In the case of @code{dup ( w -- w w )} the generator must not
 7338: eliminate the store to the original location of the item on the stack,
 7339: if the TOS optimization is turned on.
 7340: @item Primitives with stack effects of the form @code{--}
 7341: @var{out1}...@var{outy} must store the TOS to the stack at the start.
 7342: Likewise, primitives with the stack effect @var{in1}...@var{inx} @code{--}
 7343: must load the TOS from the stack at the end. But for the null stack
 7344: effect @code{--} no stores or loads should be generated.
 7345: @end itemize
 7346: 
 7347: @node Produced code,  , TOS Optimization, Primitives
 7348: @subsection Produced code
 7349: @cindex primitives, assembly code listing
 7350: 
 7351: @cindex @file{engine.s}
 7352: To see what assembly code is produced for the primitives on your machine
 7353: with your compiler and your flag settings, type @code{make engine.s} and
 7354: look at the resulting file @file{engine.s}.
 7355: 
 7356: @node  Performance,  , Primitives, Engine
 7357: @section Performance
 7358: @cindex performance of some Forth interpreters
 7359: @cindex engine performance
 7360: @cindex benchmarking Forth systems
 7361: @cindex Gforth performance
 7362: 
 7363: On RISCs the Gforth engine is very close to optimal; i.e., it is usually
 7364: impossible to write a significantly faster engine.
 7365: 
 7366: On register-starved machines like the 386 architecture processors
 7367: improvements are possible, because @code{gcc} does not utilize the
 7368: registers as well as a human, even with explicit register declarations;
 7369: e.g., Bernd Beuster wrote a Forth system fragment in assembly language
 7370: and hand-tuned it for the 486; this system is 1.19 times faster on the
 7371: Sieve benchmark on a 486DX2/66 than Gforth compiled with
 7372: @code{gcc-2.6.3} with @code{-DFORCE_REG}.
 7373: 
 7374: @cindex Win32Forth performance
 7375: @cindex NT Forth performance
 7376: @cindex eforth performance
 7377: @cindex ThisForth performance
 7378: @cindex PFE performance
 7379: @cindex TILE performance
 7380: However, this potential advantage of assembly language implementations
 7381: is not necessarily realized in complete Forth systems: We compared
 7382: Gforth (direct threaded, compiled with @code{gcc-2.6.3} and
 7383: @code{-DFORCE_REG}) with Win32Forth 1.2093, LMI's NT Forth (Beta, May
 7384: 1994) and Eforth (with and without peephole (aka pinhole) optimization
 7385: of the threaded code); all these systems were written in assembly
 7386: language. We also compared Gforth with three systems written in C:
 7387: PFE-0.9.14 (compiled with @code{gcc-2.6.3} with the default
 7388: configuration for Linux: @code{-O2 -fomit-frame-pointer -DUSE_REGS
 7389: -DUNROLL_NEXT}), ThisForth Beta (compiled with gcc-2.6.3 -O3
 7390: -fomit-frame-pointer; ThisForth employs peephole optimization of the
 7391: threaded code) and TILE (compiled with @code{make opt}). We benchmarked
 7392: Gforth, PFE, ThisForth and TILE on a 486DX2/66 under Linux. Kenneth
 7393: O'Heskin kindly provided the results for Win32Forth and NT Forth on a
 7394: 486DX2/66 with similar memory performance under Windows NT. Marcel
 7395: Hendrix ported Eforth to Linux, then extended it to run the benchmarks,
 7396: added the peephole optimizer, ran the benchmarks and reported the
 7397: results.
 7398:  
 7399: We used four small benchmarks: the ubiquitous Sieve; bubble-sorting and
 7400: matrix multiplication come from the Stanford integer benchmarks and have
 7401: been translated into Forth by Martin Fraeman; we used the versions
 7402: included in the TILE Forth package, but with bigger data set sizes; and
 7403: a recursive Fibonacci number computation for benchmarking calling
 7404: performance. The following table shows the time taken for the benchmarks
 7405: scaled by the time taken by Gforth (in other words, it shows the speedup
 7406: factor that Gforth achieved over the other systems).
 7407: 
 7408: @example
 7409: relative      Win32-    NT       eforth       This-
 7410:   time  Gforth Forth Forth eforth  +opt   PFE Forth  TILE
 7411: sieve     1.00  1.39  1.14   1.39  0.85  1.58  3.18  8.58
 7412: bubble    1.00  1.31  1.41   1.48  0.88  1.50        3.88
 7413: matmul    1.00  1.47  1.35   1.46  0.74  1.58        4.09
 7414: fib       1.00  1.52  1.34   1.22  0.86  1.74  2.99  4.30
 7415: @end example
 7416: 
 7417: You may find the good performance of Gforth compared with the systems
 7418: written in assembly language quite surprising. One important reason for
 7419: the disappointing performance of these systems is probably that they are
 7420: not written optimally for the 486 (e.g., they use the @code{lods}
 7421: instruction). In addition, Win32Forth uses a comfortable, but costly
 7422: method for relocating the Forth image: like @code{cforth}, it computes
 7423: the actual addresses at run time, resulting in two address computations
 7424: per NEXT (@pxref{Image File Background}).
 7425: 
 7426: Only Eforth with the peephole optimizer performs comparable to
 7427: Gforth. The speedups achieved with peephole optimization of threaded
 7428: code are quite remarkable. Adding a peephole optimizer to Gforth should
 7429: cause similar speedups.
 7430: 
 7431: The speedup of Gforth over PFE, ThisForth and TILE can be easily
 7432: explained with the self-imposed restriction of the latter systems to
 7433: standard C, which makes efficient threading impossible (however, the
 7434: measured implementation of PFE uses a GNU C extension: @pxref{Global Reg
 7435: Vars, , Defining Global Register Variables, gcc.info, GNU C Manual}).
 7436: Moreover, current C compilers have a hard time optimizing other aspects
 7437: of the ThisForth and the TILE source.
 7438: 
 7439: Note that the performance of Gforth on 386 architecture processors
 7440: varies widely with the version of @code{gcc} used. E.g., @code{gcc-2.5.8}
 7441: failed to allocate any of the virtual machine registers into real
 7442: machine registers by itself and would not work correctly with explicit
 7443: register declarations, giving a 1.3 times slower engine (on a 486DX2/66
 7444: running the Sieve) than the one measured above.
 7445: 
 7446: Note also that there have been several releases of Win32Forth since the
 7447: release presented here, so the results presented here may have little
 7448: predictive value for the performance of Win32Forth today.
 7449: 
 7450: @cindex @file{Benchres}
 7451: In @cite{Translating Forth to Efficient C} by M. Anton Ertl and Martin
 7452: Maierhofer (presented at EuroForth '95), an indirect threaded version of
 7453: Gforth is compared with Win32Forth, NT Forth, PFE, and ThisForth; that
 7454: version of Gforth is 2%@minus{}8% slower on a 486 than the direct
 7455: threaded version used here. The paper available at
 7456: @*@url{http://www.complang.tuwien.ac.at/papers/ertl&maierhofer95.ps.gz};
 7457: it also contains numbers for some native code systems. You can find a
 7458: newer version of these measurements at
 7459: @url{http://www.complang.tuwien.ac.at/forth/performance.html}. You can
 7460: find numbers for Gforth on various machines in @file{Benchres}.
 7461: 
 7462: @node Binding to System Library, Cross Compiler, Engine, Top
 7463: @chapter Binding to System Library
 7464: 
 7465: @node Cross Compiler, Bugs, Binding to System Library, Top
 7466: @chapter Cross Compiler
 7467: 
 7468: Cross Compiler
 7469: 
 7470: @menu
 7471: * Using the Cross Compiler::
 7472: * How the Cross Compiler Works::
 7473: @end menu
 7474: 
 7475: @node Using the Cross Compiler, , How the Cross Compiler Works, Cross Compiler
 7476: @section Using the Cross Compiler
 7477: 
 7478: @node How the Cross Compiler Works, Using the Cross Compiler, , Cross Compiler
 7479: @section How the Cross Compiler Works
 7480: 
 7481: @node Bugs, Origin, Cross Compiler, Top
 7482: @chapter Bugs
 7483: @cindex bug reporting
 7484: 
 7485: Known bugs are described in the file BUGS in the Gforth distribution.
 7486: 
 7487: If you find a bug, please send a bug report to
 7488: @email{bug-gforth@@gnu.ai.mit.edu}. A bug report should
 7489: describe the Gforth version used (it is announced at the start of an
 7490: interactive Gforth session), the machine and operating system (on Unix
 7491: systems you can use @code{uname -a} to produce this information), the
 7492: installation options (send the @file{config.status} file), and a
 7493: complete list of changes you (or your installer) have made to the Gforth
 7494: sources (if any); it should contain a program (or a sequence of keyboard
 7495: commands) that reproduces the bug and a description of what you think
 7496: constitutes the buggy behaviour.
 7497: 
 7498: For a thorough guide on reporting bugs read @ref{Bug Reporting, , How
 7499: to Report Bugs, gcc.info, GNU C Manual}.
 7500: 
 7501: 
 7502: @node Origin, Word Index, Bugs, Top
 7503: @chapter Authors and Ancestors of Gforth
 7504: 
 7505: @section Authors and Contributors
 7506: @cindex authors of Gforth
 7507: @cindex contributors to Gforth
 7508: 
 7509: The Gforth project was started in mid-1992 by Bernd Paysan and Anton
 7510: Ertl. The third major author was Jens Wilke.  Lennart Benschop (who was
 7511: one of Gforth's first users, in mid-1993) and Stuart Ramsden inspired us
 7512: with their continuous feedback. Lennart Benshop contributed
 7513: @file{glosgen.fs}, while Stuart Ramsden has been working on automatic
 7514: support for calling C libraries. Helpful comments also came from Paul
 7515: Kleinrubatscher, Christian Pirker, Dirk Zoller, Marcel Hendrix, John
 7516: Wavrik, Barrie Stott, Marc de Groot, and Jorge Acerada. Since the
 7517: release of Gforth-0.2.1 there were also helpful comments from many
 7518: others; thank you all, sorry for not listing you here (but digging
 7519: through my mailbox to extract your names is on my to-do list).
 7520: 
 7521: Gforth also owes a lot to the authors of the tools we used (GCC, CVS,
 7522: and autoconf, among others), and to the creators of the Internet: Gforth
 7523: was developed across the Internet, and its authors have not met
 7524: physically for the first 4 years of development.
 7525: 
 7526: @section Pedigree
 7527: @cindex Pedigree of Gforth
 7528: 
 7529: Gforth descends from bigFORTH (1993) and fig-Forth. Gforth and PFE (by
 7530: Dirk Zoller) will cross-fertilize each other. Of course, a significant
 7531: part of the design of Gforth was prescribed by ANS Forth.
 7532: 
 7533: Bernd Paysan wrote bigFORTH, a descendent from TurboForth, an unreleased
 7534: 32 bit native code version of VolksForth for the Atari ST, written
 7535: mostly by Dietrich Weineck.
 7536: 
 7537: VolksForth descends from F83. It was written by Klaus Schleisiek, Bernd
 7538: Pennemann, Georg Rehfeld and Dietrich Weineck for the C64 (called
 7539: UltraForth there) in the mid-80s and ported to the Atari ST in 1986.
 7540: 
 7541: Henry Laxen and Mike Perry wrote F83 as a model implementation of the
 7542: Forth-83 standard. !! Pedigree? When?
 7543: 
 7544: A team led by Bill Ragsdale implemented fig-Forth on many processors in
 7545: 1979. Robert Selzer and Bill Ragsdale developed the original
 7546: implementation of fig-Forth for the 6502 based on microForth.
 7547: 
 7548: The principal architect of microForth was Dean Sanderson. microForth was
 7549: FORTH, Inc.'s first off-the-shelf product. It was developed in 1976 for
 7550: the 1802, and subsequently implemented on the 8080, the 6800 and the
 7551: Z80.
 7552: 
 7553: All earlier Forth systems were custom-made, usually by Charles Moore,
 7554: who discovered (as he puts it) Forth during the late 60s. The first full
 7555: Forth existed in 1971.
 7556: 
 7557: A part of the information in this section comes from @cite{The Evolution
 7558: of Forth} by Elizabeth D. Rather, Donald R. Colburn and Charles
 7559: H. Moore, presented at the HOPL-II conference and preprinted in SIGPLAN
 7560: Notices 28(3), 1993.  You can find more historical and genealogical
 7561: information about Forth there.
 7562: 
 7563: @node Word Index, Concept Index, Origin, Top
 7564: @unnumbered Word Index
 7565: 
 7566: This index is as incomplete as the manual. Each word is listed with
 7567: stack effect and wordset.
 7568: 
 7569: @printindex fn
 7570: 
 7571: @node Concept Index,  , Word Index, Top
 7572: @unnumbered Concept and Word Index
 7573: 
 7574: This index is as incomplete as the manual. Not all entries listed are
 7575: present verbatim in the text. Only the names are listed for the words
 7576: here.
 7577: 
 7578: @printindex cp
 7579: 
 7580: @contents
 7581: @bye
 7582: 

FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>