File:  [gforth] / gforth / glocals.fs
Revision 1.1: download - view: text, annotated - select for diffs
Sat May 7 14:55:56 1994 UTC (27 years, 7 months ago) by anton
Branches: MAIN
CVS tags: HEAD
local variables
rewrote primitives2c.el in Forth (prims2x.el)
various small changes
Added Files:
 	from-cut-here gforth.el gforth.texi glocals.fs gray.fs
 	locals-test.fs prims2x.fs

    1: \ Local variables are quite important for writing readable programs, but
    2: \ IMO (anton) they are the worst part of the standard. There they are very
    3: \ restricted and have an ugly interface.
    4: 
    5: \ So, we implement the locals wordset, but do not recommend using
    6: \ locals-ext (which is a really bad user interface for locals).
    7: 
    8: \ We also have a nice and powerful user-interface for locals: locals are
    9: \ defined with
   10: 
   11: \ { local1 local2 ... }
   12: \ or
   13: \ { local1 local2 ... -- ... }
   14: \ (anything after the -- is just a comment)
   15: 
   16: \ Every local in this list consists of an optional type specification
   17: \ and a name. If there is only the name, it stands for a cell-sized
   18: \ value (i.e., you get the value of the local variable, not it's
   19: \ address). The following type specifiers stand before the name:
   20: 
   21: \ Specifier	Type	Access
   22: \ W:		Cell	value
   23: \ W^		Cell	address
   24: \ D:		Double	value
   25: \ D^		Double	address
   26: \ F:		Float	value
   27: \ F^		Float	address
   28: \ C:		Char	value
   29: \ C^		Char	address
   30: 
   31: \ The local variables are initialized with values from the appropriate
   32: \ stack. In contrast to the examples in the standard document our locals
   33: \ take the arguments in the expected way: The last local gets the top of
   34: \ stack, the second last gets the second stack item etc. An example:
   35: 
   36: \ : CX* { F: Ar  F: Ai  F: Br  F: Bi -- Cr Ci }
   37: \ \ complex multiplication
   38: \  Ar Br f* Ai Bi f* f-
   39: \  Ar Bi f* Ai Br f* f+ ;
   40: 
   41: \ There will also be a way to add user types, but it is not yet decided,
   42: \ how. Ideas are welcome.
   43: 
   44: \ Locals defined in this manner live until (!! see below). 
   45: \ Their names can be used during this time to get
   46: \ their value or address; The addresses produced in this way become
   47: \ invalid at the end of the lifetime.
   48: 
   49: \ Values can be changed with TO, but this is not recomended (TO is a
   50: \ kludge and words lose the single-assignment property, which makes them
   51: \ harder to analyse).
   52: 
   53: \ As for the internals, we use a special locals stack. This eliminates
   54: \ the problems and restrictions of reusing the return stack and allows
   55: \ to store floats as locals: the return stack is not guaranteed to be
   56: \ aligned correctly, but our locals stack must be float-aligned between
   57: \ words.
   58: 
   59: \ Other things about the internals are pretty unclear now.
   60: 
   61: \ Currently locals may only be
   62: \ defined at the outer level and TO is not supported.
   63: 
   64: include float.fs
   65: include search-order.fs
   66: 
   67: \ the locals stack grows downwards (see primitives)
   68: \ of the local variables of a group (in braces) the leftmost is on top,
   69: \ i.e. by going onto the locals stack the order is reversed.
   70: \ there are alignment gaps if necessary.
   71: \ lp must have the strictest alignment (usually float) across calls;
   72: \ for simplicity we align it strictly for every group.
   73: 
   74: vocabulary locals \ this contains the local variables
   75: ' locals >body Constant locals-list \ acts like a variable that contains
   76: 				    \ a linear list of locals names
   77: 
   78: create locals-buffer 1000 allot \ !! limited and unsafe
   79:     \ here the names of the local variables are stored
   80:     \ we would have problems storing them at the normal dp
   81: 
   82: variable locals-dp \ so here's the special dp for locals.
   83: 
   84: : alignlp-w ( n1 -- n2 )
   85:     \ cell-align size and generate the corresponding code for aligning lp
   86:     dup aligned tuck - compile-lp+!# ;
   87: 
   88: : alignlp-f ( n1 -- n2 )
   89:     dup faligned tuck - compile-lp+!# ;
   90: 
   91: \ a local declaration group (the braces stuff) is compiled by calling
   92: \ the appropriate compile-pushlocal for the locals, starting with the
   93: \ righmost local; the names are already created earlier, the
   94: \ compile-pushlocal just inserts the offsets from the frame base.
   95: 
   96: : compile-pushlocal-w ( a-addr -- ) ( run-time: w -- )
   97: \ compiles a push of a local variable, and adjusts locals-size
   98: \ stores the offset of the local variable to a-addr
   99:     locals-size @ alignlp-w cell+ dup locals-size !
  100:     swap !
  101:     postpone >l ;
  102: 
  103: : compile-pushlocal-f ( a-addr -- ) ( run-time: f -- )
  104:     locals-size @ alignlp-f float+ dup locals-size !
  105:     swap !
  106:     postpone f>l ;
  107: 
  108: : compile-pushlocal-d ( a-addr -- ) ( run-time: w1 w2 -- )
  109:     locals-size @ alignlp-w cell+ cell+ dup locals-size !
  110:     swap !
  111:     postpone swap postpone >l postpone >l ;
  112: 
  113: : compile-pushlocal-c ( a-addr -- ) ( run-time: w -- )
  114:     -1 chars compile-lp+!#
  115:     locals-size @ swap !
  116:     postpone lp@ postpone c! ;
  117: 
  118: : create-local ( " name" -- a-addr )
  119: 	\ defines the local "name"; the offset of the local shall be stored in a-addr
  120:     create
  121: 	immediate
  122: 	here 0 , ( place for the offset ) ;
  123: 
  124: : lp-offset, ( n -- )
  125: \ converts the offset from the frame start to an offset from lp and
  126: \ adds it as inline argument to a preceding locals primitive
  127: \ i.e., the address of the local is lp+locals_size-offset
  128:   locals-size @ swap - , ;
  129: 
  130: vocabulary locals-types \ this contains all the type specifyers, -- and }
  131: locals-types definitions
  132: 
  133: : W:
  134:     create-local ( "name" -- a-addr xt )
  135: 	\ xt produces the appropriate locals pushing code when executed
  136: 	['] compile-pushlocal-w
  137:     does> ( Compilation: -- ) ( Run-time: -- w )
  138:         \ compiles a local variable access
  139: 	postpone @local# @ lp-offset, ;
  140: 
  141: : W^
  142:     create-local ( "name" -- a-addr xt )
  143: 	['] compile-pushlocal-w
  144:     does> ( Compilation: -- ) ( Run-time: -- w )
  145: 	postpone laddr# @ lp-offset, ;
  146: 
  147: : F:
  148:     create-local ( "name" -- a-addr xt )
  149: 	['] compile-pushlocal-f
  150:     does> ( Compilation: -- ) ( Run-time: -- w )
  151: 	postpone f@local# @ lp-offset, ;
  152: 
  153: : F^
  154:     create-local ( "name" -- a-addr xt )
  155: 	['] compile-pushlocal-f
  156:     does> ( Compilation: -- ) ( Run-time: -- w )
  157: 	postpone laddr# @ lp-offset, ;
  158: 
  159: : D:
  160:     create-local ( "name" -- a-addr xt )
  161: 	['] compile-pushlocal-d
  162:     does> ( Compilation: -- ) ( Run-time: -- w )
  163: 	postpone laddr# @ lp-offset, postpone 2@ ;
  164: 
  165: : D^
  166:     create-local ( "name" -- a-addr xt )
  167: 	['] compile-pushlocal-d
  168:     does> ( Compilation: -- ) ( Run-time: -- w )
  169: 	postpone laddr# @ lp-offset, ;
  170: 
  171: : C:
  172:     create-local ( "name" -- a-addr xt )
  173: 	['] compile-pushlocal-c
  174:     does> ( Compilation: -- ) ( Run-time: -- w )
  175: 	postpone laddr# @ lp-offset, postpone c@ ;
  176: 
  177: : C^
  178:     create-local ( "name" -- a-addr xt )
  179: 	['] compile-pushlocal-c
  180:     does> ( Compilation: -- ) ( Run-time: -- w )
  181: 	postpone laddr# @ lp-offset, ;
  182: 
  183: \ you may want to make comments in a locals definitions group:
  184: ' \ alias \ immediate
  185: ' ( alias ( immediate
  186: 
  187: forth definitions
  188: 
  189: \ the following gymnastics are for declaring locals without type specifier.
  190: \ we exploit a feature of our dictionary: every wordlist
  191: \ has it's own methods for finding words etc.
  192: \ So we create a vocabulary new-locals, that creates a 'w:' local named x
  193: \ when it is asked if it contains x.
  194: 
  195: 0. 2constant last-local \ !! actually a 2value
  196: 
  197: also locals-types
  198: 
  199: : new-locals-find ( caddr u w -- nfa )
  200: \ this is the find method of the new-locals vocabulary
  201: \ make a new local with name caddr u; w is ignored
  202: \ the returned nfa denotes a word that produces what W: produces
  203: \ !! do the whole thing without nextname
  204:     drop nextname W: \ we don't want the thing that W: produces,
  205:     ['] last-local >body 2!   \ but the nfa of a word that produces that value: last-local
  206:     [ ' last-local >name ] Aliteral ;
  207: 
  208: previous
  209: 
  210: : new-locals-reveal ( -- )
  211:   true abort" this should not happen: new-locals-reveal" ;
  212: 
  213: create new-locals-map ' new-locals-find A, ' new-locals-reveal A,
  214: 
  215: vocabulary new-locals
  216: new-locals-map ' new-locals >body cell+ A! \ !! use special access words
  217: 
  218: variable old-dpp
  219: 
  220: \ and now, finally, the user interface words
  221: : { ( -- addr wid 0 )
  222:     dp old-dpp !
  223:     locals-dp dpp !
  224:     also new-locals
  225:     also get-current locals definitions  locals-types
  226:     0 TO locals-wordlist
  227:     0 postpone [ ; immediate
  228: 
  229: locals-types definitions
  230: 
  231: : } ( addr wid 0 a-addr1 xt1 ... -- )
  232:     \ ends locals definitions
  233:     ] old-dpp @ dpp !
  234:     begin
  235: 	dup
  236:     while
  237: 	execute
  238:     repeat
  239:     drop
  240:     locals-size @ alignlp-f locals-size ! \ the strictest alignment
  241:     set-current
  242:     previous previous
  243:     locals-list TO locals-wordlist ;
  244: 
  245: : -- ( addr wid 0 ... -- )
  246:     }
  247:     [char] } word drop ;
  248: 
  249: forth definitions
  250: 
  251: \ A few thoughts on automatic scopes for locals and how they can be
  252: \ implemented:
  253: 
  254: \ We have to combine locals with the control structures. My basic idea
  255: \ was to start the life of a local at the declaration point. The life
  256: \ would end at any control flow join (THEN, BEGIN etc.) where the local
  257: \ is lot live on both input flows (note that the local can still live in
  258: \ other, later parts of the control flow). This would make a local live
  259: \ as long as you expected and sometimes longer (e.g. a local declared in
  260: \ a BEGIN..UNTIL loop would still live after the UNTIL).
  261: 
  262: \ The following example illustrates the problems of this approach:
  263: 
  264: \ { z }
  265: \ if
  266: \   { x }
  267: \ begin
  268: \   { y }
  269: \ [ 1 cs-roll ] then
  270: \   ...
  271: \ until
  272: 
  273: \ x lives only until the BEGIN, but the compiler does not know this
  274: \ until it compiles the UNTIL (it can deduce it at the THEN, because at
  275: \ that point x lives in no thread, but that does not help much). This is
  276: \ solved by optimistically assuming at the BEGIN that x lives, but
  277: \ warning at the UNTIL that it does not. The user is then responsible
  278: \ for checking that x is only used where it lives.
  279: 
  280: \ The produced code might look like this (leaving out alignment code):
  281: 
  282: \ >l ( z )
  283: \ ?branch <then>
  284: \ >l ( x )
  285: \ <begin>:
  286: \ >l ( y )
  287: \ lp+!# 8 ( RIP: x,y )
  288: \ <then>:
  289: \ ...
  290: \ lp+!# -4 ( adjust lp to <begin> state )
  291: \ ?branch <begin>
  292: \ lp+!# 4 ( undo adjust )
  293: 
  294: \ The BEGIN problem also has another incarnation:
  295: 
  296: \ AHEAD
  297: \ BEGIN
  298: \   x
  299: \ [ 1 CS-ROLL ] THEN
  300: \   { x }
  301: \   ...
  302: \ UNTIL
  303: 
  304: \ should be legal: The BEGIN is not a control flow join in this case,
  305: \ since it cannot be entered from the top; therefore the definition of x
  306: \ dominates the use. But the compiler processes the use first, and since
  307: \ it does not look ahead to notice the definition, it will complain
  308: \ about it. Here's another variation of this problem:
  309: 
  310: \ IF
  311: \   { x }
  312: \ ELSE
  313: \   ...
  314: \ AHEAD
  315: \ BEGIN
  316: \   x
  317: \ [ 2 CS-ROLL ] THEN
  318: \   ...
  319: \ UNTIL
  320: 
  321: \ In this case x is defined before the use, and the definition dominates
  322: \ the use, but the compiler does not know this until it processes the
  323: \ UNTIL. So what should the compiler assume does live at the BEGIN, if
  324: \ the BEGIN is not a control flow join? The safest assumption would be
  325: \ the intersection of all locals lists on the control flow
  326: \ stack. However, our compiler assumes that the same variables are live
  327: \ as on the top of the control flow stack. This covers the following case:
  328: 
  329: \ { x }
  330: \ AHEAD
  331: \ BEGIN
  332: \   x
  333: \ [ 1 CS-ROLL ] THEN
  334: \   ...
  335: \ UNTIL
  336: 
  337: \ If this assumption is too optimistic, the compiler will warn the user.
  338: 
  339: \ Implementation:
  340: 
  341: \ orig, dest and do-sys have the following structure:
  342: \ address (of the branch or the instruction to be branched to) (TOS)
  343: \ locals-list (valid at address) (second)
  344: \ locals-size (at address; this could be computed from locals-list, but so what) (third)
  345: 
  346: 3 constant cs-item-size
  347: 
  348: : CS-PICK ( ... u -- ... destu )
  349:  1+ cs-item-size * 1- >r
  350:  r@ pick  r@ pick  r@ pick
  351:  rdrop ;
  352: 
  353: : CS-ROLL ( destu/origu .. dest0/orig0 u -- .. dest0/orig0 destu/origu )
  354:  1+ cs-item-size * 1- >r
  355:  r@ roll r@ roll r@ roll
  356:  rdrop ; 
  357: 
  358: : CS-PUSH ( -- dest/orig )
  359:  locals-size @
  360:  locals-list @
  361:  here ;
  362: 
  363: : BUT       sys? 1 cs-roll ;                      immediate restrict
  364: : YET       sys? 0 cs-pick ;                       immediate restrict
  365: 
  366: : common-list ( list1 list2 -- list3 )
  367: \ list1 and list2 are lists, where the heads are at higher addresses than
  368: \ the tail. list3 is the largest sublist of both lists.
  369:  begin
  370:    2dup u<>
  371:  while
  372:    2dup u>
  373:    if
  374:      swap
  375:    endif
  376:    @
  377:  repeat
  378:  drop ;
  379: 
  380: : sub-list? ( list1 list2 -- f )
  381: \ true iff list1 is a sublist of list2
  382:  begin
  383:    2dup u<
  384:  while
  385:    @
  386:  repeat
  387:  = ;
  388: 
  389: : list-size ( list -- u )
  390: \ size of the locals frame represented by list
  391:  0 ( list n )
  392:  begin
  393:    over 0<>
  394:  while
  395:    over
  396:    cell+ name> >body @ max
  397:    swap @ swap ( get next )
  398:  repeat
  399:  faligned nip ;
  400: 
  401: : x>mark ( -- orig )
  402:  cs-push 0 , ;
  403: 
  404: variable dead-code \ true if normal code at "here" would be dead
  405: 
  406: : unreachable ( -- )
  407: \ declares the current point of execution as unreachable and
  408: \ prepares the assumptions for a possible upcoming BEGIN
  409:  dead-code on
  410:  dup 0<> if
  411:    2 pick 2 pick
  412:  else
  413:    0 0
  414:  endif
  415:  locals-list !
  416:  locals-size ! ;
  417: 
  418: : check-begin ( list -- )
  419: \ warn if list is not a sublist of locals-list
  420:  locals-list @ sub-list? 0= if
  421:    \ !! print current position
  422:    ." compiler was overly optimistic about locals at a BEGIN" cr
  423:    \ !! print assumption and reality
  424:  endif ;
  425: 
  426: : xahead ( -- orig )
  427:  POSTPONE branch x>mark unreachable ; immediate
  428: 
  429: : xif ( -- orig )
  430:  POSTPONE ?branch x>mark ; immediate
  431: 
  432: \ THEN (another control flow from before joins the current one):
  433: \ The new locals-list is the intersection of the current locals-list and
  434: \ the orig-local-list. The new locals-size is the (alignment-adjusted)
  435: \ size of the new locals-list. The following code is generated:
  436: \ lp+!# (current-locals-size - orig-locals-size)
  437: \ <then>:
  438: \ lp+!# (orig-locals-size - new-locals-size)
  439: 
  440: \ Of course "lp+!# 0" is not generated. Still this is admittedly a bit
  441: \ inefficient, e.g. if there is a locals declaration between IF and
  442: \ ELSE. However, if ELSE generates an appropriate "lp+!#" before the
  443: \ branch, there will be none after the target <then>.
  444: : xthen ( orig -- )
  445:  sys? dup @ ?struc
  446:  dead-code @
  447:  if
  448:    >resolve
  449:    locals-list !
  450:    locals-size !
  451:  else
  452:    locals-size @ 3 roll - compile-lp+!#
  453:    >resolve
  454:    locals-list @ common-list locals-list !
  455:    locals-size @  locals-list @ list-size - compile-lp+!#
  456:  endif
  457:  dead-code off ; immediate
  458: 
  459: : scope ( -- dest )
  460:  cs-push ; immediate
  461: 
  462: : endscope ( dest -- )
  463:  drop
  464:  locals-list @ common-list locals-list !
  465:  locals-size @  locals-list @ list-size - compile-lp+!#
  466:  drop ; immediate
  467: 
  468: : xexit ( -- )
  469:     locals-size @ compile-lp+!# POSTPONE exit unreachable ; immediate
  470: 
  471: : x?exit ( -- )
  472:     POSTPONE xif POSTPONE xexit POSTPONE xthen ; immediate
  473: 
  474: : xelse ( orig1 -- orig2 )
  475:  sys?
  476:  POSTPONE xahead
  477:  1 cs-roll
  478:  POSTPONE xthen ; immediate
  479: 
  480: : xbegin ( -- dest )
  481:  cs-push dead-code off ; immediate
  482: 
  483: : xwhile ( dest -- orig dest )
  484:  sys?
  485:  POSTPONE xif
  486:  1 cs-roll ; immediate
  487: 
  488: \ AGAIN (the current control flow joins another, earlier one):
  489: \ If the dest-locals-list is not a subset of the current locals-list,
  490: \ issue a warning (see below). The following code is generated:
  491: \ lp+!# (current-local-size - dest-locals-size)
  492: \ branch <begin>
  493: : xagain ( dest -- )
  494:  sys?
  495:  locals-size @ 3 roll - compile-lp+!#
  496:  POSTPONE branch
  497:  <resolve
  498:  check-begin
  499:  unreachable ; immediate
  500: 
  501: \ UNTIL (the current control flow may join an earlier one or continue):
  502: \ Similar to AGAIN. The new locals-list and locals-size are the current
  503: \ ones. The following code is generated:
  504: \ lp+!# (current-local-size - dest-locals-size)
  505: \ ?branch <begin>
  506: \ lp+!# (dest-local-size - current-locals-size)
  507: \ (Another inefficiency. Maybe we should introduce a ?branch-lp+!#
  508: \ primitive. This would also solve the interrupt problem)
  509: : until-like ( dest xt -- )
  510:  >r
  511:  sys?
  512:  locals-size @ dup 4 roll - compile-lp+!# ( list dest-addr old-locals-size )
  513:  r> compile,
  514:  >r <resolve
  515:  check-begin
  516:  locals-size @ r> - compile-lp+!# ;
  517: 
  518: : xuntil ( dest -- )
  519:  ['] ?branch until-like ; immediate
  520: 
  521: : xrepeat ( orig dest -- )
  522:  3 pick 0= ?struc
  523:  postpone xagain
  524:  postpone xthen ; immediate
  525: 
  526: \ counted loops
  527: 
  528: \ leave poses a little problem here
  529: \ we have to store more than just the address of the branch, so the
  530: \ traditional linked list approach is no longer viable.
  531: \ This is solved by storing the information about the leavings in a
  532: \ special stack. The leavings of different DO-LOOPs are separated
  533: \ by a 0 entry
  534: 
  535: \ !! remove the fixed size limit. 'Tis easy.
  536: 20 constant leave-stack-size
  537: create leave-stack leave-stack-size cs-item-size * cells allot
  538: variable leave-sp  leave-stack leave-sp !
  539: 
  540: : clear-leave-stack ( -- )
  541:  leave-stack leave-sp ! ;
  542: 
  543: \ : leave-empty? ( -- f )
  544: \  leave-sp @ leave-stack = ;
  545: 
  546: : >leave ( orig -- )
  547: \ push on leave-stack
  548:  leave-sp @
  549:  dup [ leave-stack leave-stack-size cs-item-size * cells + ] Aliteral >=
  550:  if
  551:    abort" leave-stack full"
  552:  endif
  553:  tuck ! cell+
  554:  tuck ! cell+
  555:  tuck ! cell+
  556:  leave-sp ! ;
  557: 
  558: : leave> ( -- orig )
  559: \ pop from leave-stack
  560:  leave-sp @
  561:  dup leave-stack <= if
  562:    abort" leave-stack empty"
  563:  endif
  564:  cell - dup @ swap
  565:  cell - dup @ swap
  566:  cell - dup @ swap
  567:  leave-sp ! ;
  568: 
  569: : done ( -- )
  570: \ !! the original done had ( addr -- )
  571:  begin
  572:    leave>
  573:    dup
  574:  while
  575:    POSTPONE xthen
  576:  repeat
  577:  2drop drop ; immediate
  578: 
  579: : xleave ( -- )
  580:  POSTPONE xahead
  581:  >leave ; immediate
  582: 
  583: : x?leave ( -- )
  584:  POSTPONE 0= POSTPONE xif
  585:  >leave ; immediate
  586: 
  587: : xdo ( -- do-sys )
  588:  POSTPONE (do)
  589:  POSTPONE xbegin
  590:  0 0 0 >leave ; immediate
  591: 
  592: : x?do ( -- do-sys )
  593:  0 0 0 >leave
  594:  POSTPONE (?do)
  595:  x>mark >leave
  596:  POSTPONE xbegin ; immediate
  597: 
  598: : xfor ( -- do-sys )
  599:  POSTPONE (for)
  600:  POSTPONE xbegin
  601:  0 0 0 >leave ; immediate
  602: 
  603: \ LOOP etc. are just like UNTIL
  604: \ the generated code for ?DO ... LOOP with locals is inefficient, this 
  605: \ could be changed by introducing (loop)-lp+!# etc.
  606: 
  607: : loop-like ( do-sys xt -- )
  608:  until-like  POSTPONE done  POSTPONE unloop ;
  609: 
  610: : xloop ( do-sys -- )
  611:  ['] (loop) loop-like ; immediate
  612: 
  613: : x+loop ( do-sys -- )
  614:  ['] (+loop) loop-like ; immediate
  615: 
  616: : xs+loop ( do-sys -- )
  617:  ['] (s+loop) loop-like ; immediate
  618: 
  619: : locals-:-hook ( sys -- sys addr xt )
  620:     DEFERS :-hook
  621:     last @ lastcfa @
  622:     clear-leave-stack
  623:     0 locals-size !
  624:     locals-buffer locals-dp !
  625:     0 locals-list ! ; ( clear locals vocabulary )
  626: 
  627: : locals-;-hook ( sys addr xt -- sys )
  628:     0 TO locals-wordlist
  629:     locals-size @ compile-lp+!#
  630:     lastcfa ! last !
  631:     DEFERS ;-hook ;
  632: 
  633: ' locals-:-hook IS :-hook
  634: ' locals-;-hook IS ;-hook
  635: 
  636: \ The words in the locals dictionary space are not deleted until the end
  637: \ of the current word. This is a bit too conservative, but very simple.
  638: 
  639: \ There are a few cases to consider: (see above)
  640: 
  641: \ after AGAIN, AHEAD, EXIT (the current control flow is dead):
  642: \ We have to special-case the above cases against that. In this case the
  643: \ things above are not control flow joins. Everything should be taken
  644: \ over from the live flow. No lp+!# is generated.
  645: 
  646: \ !! The lp gymnastics for UNTIL are also a real problem: locals cannot be
  647: \ used in signal handlers (or anything else that may be called while
  648: \ locals live beyond the lp) without changing the locals stack.
  649: 
  650: \ About warning against uses of dead locals. There are several options:
  651: 
  652: \ 1) Do not complain (After all, this is Forth;-)
  653: 
  654: \ 2) Additional restrictions can be imposed so that the situation cannot
  655: \ arise; the programmer would have to introduce explicit scoping
  656: \ declarations in cases like the above one. I.e., complain if there are
  657: \ locals that are live before the BEGIN but not before the corresponding
  658: \ AGAIN (replace DO etc. for BEGIN and UNTIL etc. for AGAIN).
  659: 
  660: \ 3) The real thing: i.e. complain, iff a local lives at a BEGIN, is
  661: \ used on a path starting at the BEGIN, and does not live at the
  662: \ corresponding AGAIN. This is somewhat hard to implement. a) How does
  663: \ the compiler know when it is working on a path starting at a BEGIN
  664: \ (consider "{ x } if begin [ 1 cs-roll ] else x endif again")? b) How
  665: \ is the usage info stored?
  666: 
  667: \ For now I'll resort to alternative 2. When it produces warnings they
  668: \ will often be spurious, but warnings should be rare. And better
  669: \ spurious warnings now and then than days of bug-searching.
  670: 
  671: \ Explicit scoping of locals is implemented by cs-pushing the current
  672: \ locals-list and -size (and an unused cell, to make the size equal to
  673: \ the other entries) at the start of the scope, and restoring them at
  674: \ the end of the scope to the intersection, like THEN does.
  675: 
  676: 
  677: \ And here's finally the ANS standard stuff
  678: 
  679: : (local) ( addr u -- )
  680: \ a little space-inefficient, but well deserved ;-)
  681: \ In exchange, there are no restrictions whatsoever on using (local)
  682:  dup
  683:  if
  684:    nextname POSTPONE { [ also locals-types ] W: } [ previous ]
  685:  else
  686:    2drop
  687:  endif ;
  688: 
  689: \ \ !! untested
  690: \ : TO ( c|w|d|r "name" -- )
  691: \ \ !! state smart
  692: \  0 0 0. 0.0e0 { c: clocal w: wlocal d: dlocal f: flocal }
  693: \  ' dup >definer
  694: \  state @ 
  695: \  if
  696: \    case
  697: \      [ ' locals-wordlist >definer ] literal \ value
  698: \      OF >body POSTPONE Aliteral POSTPONE ! ENDOF
  699: \      [ ' clocal >definer ] literal
  700: \      OF POSTPONE laddr# >body @ lp-offset, POSTPONE c! ENDOF
  701: \      [ ' wlocal >definer ] literal
  702: \      OF POSTPONE laddr# >body @ lp-offset, POSTPONE ! ENDOF
  703: \      [ ' dlocal >definer ] literal
  704: \      OF POSTPONE laddr# >body @ lp-offset, POSTPONE d! ENDOF
  705: \      [ ' flocal >definer ] literal
  706: \      OF POSTPONE laddr# >body @ lp-offset, POSTPONE f! ENDOF
  707: \      abort" can only store TO value or local value"
  708: \    endcase
  709: \  else
  710: \    [ ' locals-wordlist >definer ] literal =
  711: \    if
  712: \      >body !
  713: \    else
  714: \      abort" can only store TO value"
  715: \    endif
  716: \  endif ;   
  717: 
  718: \ : locals|
  719: \ !! should lie around somewhere

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