1 ! Copyright (C) 2007, 2008 Chris Double.
2 ! See http://factorcode.org/license.txt for BSD license.
4 USING: accessors arrays assocs classes combinators
5 combinators.short-circuit compiler.units effects.parser fry
6 generalizations kernel locals make math math.order namespaces
7 quotations sequences sets splitting unicode vectors
12 TUPLE: parse-result remaining ast ;
13 TUPLE: parse-error position got messages ;
14 TUPLE: parser peg compiled id ;
16 M: parser equal? { [ [ class-of ] same? ] [ [ id>> ] same? ] } 2&& ;
17 M: parser hashcode* id>> hashcode* ;
19 C: <parse-result> parse-result
20 C: <parse-error> parse-error
24 : merge-overlapping-errors ( a b -- c )
25 dupd [ messages>> ] bi@ union [ [ position>> ] [ got>> ] bi ] dip
28 : (merge-errors) ( a b -- c )
30 { [ over position>> not ] [ nip ] }
31 { [ dup position>> not ] [ drop ] }
33 2dup [ position>> ] compare {
36 { +eq+ [ merge-overlapping-errors ] }
42 error-stack get dup length 1 > [
43 [ pop ] [ pop swap (merge-errors) ] [ ] tri push
48 : add-error ( position got message -- )
49 <parse-error> error-stack get push ;
53 : packrat ( id -- cache )
54 ! The packrat cache is a mapping of parser-id->cache.
55 ! For each parser it maps to a cache holding a mapping
56 ! of position->result. The packrat cache therefore keeps
57 ! track of all parses that have occurred at each position
58 ! of the input string and the results obtained from that
60 \ packrat get [ drop H{ } clone ] cache ;
68 ! A mapping from position->peg-head. It maps a
69 ! position in the input string being parsed to
70 ! the head of the left recursion which is currently
71 ! being grown. It is 'f' at any position where
72 ! left recursion growth is not underway.
75 : failed? ( obj -- ? )
78 : peg-cache ( -- cache )
79 ! Holds a hashtable mapping a peg tuple to
80 ! the parser tuple for that peg. The parser tuple
81 ! holds a unique id and the compiled form of that peg.
82 \ peg-cache get-global [
83 H{ } clone dup \ peg-cache set-global
87 H{ } clone \ peg-cache set-global ;
91 ! An entry in the table of memoized parse results
92 ! ast = an AST produced from the parse
93 ! or the symbol 'fail'
94 ! or a left-recursion object
95 ! pos = the position in the input string of this entry
96 TUPLE: memo-entry ans pos ;
98 TUPLE: left-recursion seed rule-id head next ;
99 TUPLE: peg-head rule-id involved-set eval-set ;
101 : rule-id ( word -- id )
102 ! A rule is the parser compiled down to a word. It has
103 ! a "peg-id" property containing the id of the original parser.
106 : input-slice ( -- slice )
107 ! Return a slice of the input from the current parse position
108 input get pos get tail-slice ;
110 : input-from ( input -- n )
111 ! Return the index from the original string that the
112 ! input slice is based on.
113 dup slice? [ from>> ] [ drop 0 ] if ;
115 : process-rule-result ( p result -- result )
117 nip [ ast>> ] [ remaining>> ] bi input-from pos namespaces:set
119 pos namespaces:set fail
122 : eval-rule ( rule -- ast )
123 ! Evaluate a rule, return an ast resulting from it.
124 ! Return fail if the rule failed. The rule has
125 ! stack effect ( -- parse-result )
126 pos get swap execute( -- parse-result ) process-rule-result ; inline
128 : memo ( pos id -- memo-entry )
129 ! Return the result from the memo cache.
132 : set-memo ( memo-entry pos id -- )
133 ! Store an entry in the cache
136 : update-m ( ast m -- )
137 swap >>ans pos get >>pos drop ;
139 : stop-growth? ( ast m -- ? )
140 [ failed? pos get ] dip pos>> <= or ;
142 : setup-growth ( h p -- )
143 pos namespaces:set dup involved-set>> clone >>eval-set drop ;
145 : (grow-lr) ( h p r: ( -- result ) m -- )
146 [ [ setup-growth ] 2keep ] 2dip
147 [ dup eval-rule ] dip swap
148 dup pick stop-growth? [
153 ] if ; inline recursive
155 : grow-lr ( h p r m -- ast )
156 [ [ heads set-at ] 2keep ] 2dip
157 pick over [ (grow-lr) ] 2dip
159 dup pos>> pos namespaces:set ans>>
162 :: (setup-lr) ( l s -- )
164 s left-recursion? [ s throw ] unless
165 s head>> l head>> eq? [
167 l head>> [ s rule-id>> suffix ] change-involved-set drop
168 l s next>> (setup-lr)
172 :: setup-lr ( r l -- )
174 r rule-id V{ } clone V{ } clone peg-head boa l head<<
176 l lrstack get (setup-lr) ;
178 :: lr-answer ( r p m -- ast )
180 h rule-id>> r rule-id eq? [
181 m ans>> seed>> m ans<<
191 :: recall ( r p -- memo-entry )
192 p r rule-id memo :> m
195 m r rule-id h involved-set>> h rule-id>> suffix member? not and [
196 fail p memo-entry boa
198 r rule-id h eval-set>> member? [
199 h [ r rule-id swap remove ] change-eval-set drop
211 :: apply-non-memo-rule ( r p -- ast )
212 fail r rule-id f lrstack get left-recursion boa :> lr
213 lr lrstack namespaces:set lr p memo-entry boa dup p r rule-id set-memo :> m
215 lrstack get next>> lrstack namespaces:set
218 m ans>> left-recursion? [
227 : apply-memo-rule ( r m -- ast )
228 [ ans>> ] [ pos>> ] bi pos namespaces:set
229 dup left-recursion? [
230 [ setup-lr ] keep seed>>
235 : apply-rule ( r p -- ast )
242 : with-packrat ( input quot -- result )
243 ! Run the quotation with a packrat cache active.
248 V{ } clone error-stack ,,
249 H{ } clone \ heads ,,
250 H{ } clone \ packrat ,,
251 ] H{ } make swap with-variables ; inline
253 GENERIC: (compile) ( peg -- quot )
255 : process-parser-result ( result -- result )
259 input-slice swap <parse-result>
262 : execute-parser ( word -- result )
263 pos get apply-rule process-parser-result ;
265 : preset-parser-word ( parser -- parser word )
266 gensym [ >>compiled ] keep ;
268 : define-parser-word ( parser word -- )
269 ! Return the body of the word that is the compiled version
271 2dup swap peg>> (compile) ( -- result ) define-declared
272 swap id>> "peg-id" set-word-prop ;
274 : compile-parser ( parser -- word )
275 ! Look to see if the given parser has been compiled.
276 ! If not, compile it to a temporary word, cache it,
277 ! and return it. Otherwise return the existing one.
278 ! Circular parsers are supported by getting the word
279 ! name and storing it in the cache, before compiling,
280 ! so it is picked up when re-entered.
284 preset-parser-word [ define-parser-word ] keep
287 : compile-parser-quot ( parser -- quot )
288 compile-parser [ execute-parser ] curry ;
292 : fixup-delayed ( -- )
293 ! Work through all delayed parsers and recompile their
294 ! words to have the correct bodies.
296 call( -- parser ) compile-parser-quot ( -- result ) define-declared
299 : compile ( parser -- word )
302 compile-parser-quot ( -- result ) define-temp fixup-delayed
304 ] with-compilation-unit ;
306 : compiled-parse ( state word -- result )
309 [ error-stack get ?first [ throw ]
310 [ pos get input get f <parse-error> throw ] if* ] unless*
313 : (parse) ( input parser -- result )
314 dup word? [ compile ] unless compiled-parse ;
316 : parse ( input parser -- ast )
322 ! Return the next unique id for a parser
325 : wrap-peg ( peg -- parser )
326 ! Wrap a parser tuple around the peg object.
327 ! Look for an existing parser tuple for that
329 peg-cache [ f next-id parser boa ] cache ;
331 TUPLE: token-parser symbol ;
333 : parse-token ( input string -- result )
334 ! Parse the string, returning a parse result
335 [ ?head-slice ] keep swap [
338 [ seq>> pos get swap ] dip "'" "'" surround 1vector add-error f
341 M: token-parser (compile) ( peg -- quot )
342 symbol>> '[ input-slice _ parse-token ] ;
344 TUPLE: satisfy-parser quot ;
346 :: parse-satisfy ( input quot -- result/f )
348 unclip-slice dup quot call [
355 M: satisfy-parser (compile)
356 quot>> '[ input-slice _ parse-satisfy ] ;
358 TUPLE: range-parser min max ;
360 :: parse-range ( input min max -- result/f )
362 dup first min max between? [
363 unclip-slice <parse-result>
369 M: range-parser (compile)
370 [ min>> ] [ max>> ] bi '[ input-slice _ _ parse-range ] ;
372 TUPLE: seq-parser parsers ;
374 : calc-seq-result ( prev-result current-result -- next-result )
376 [ remaining>> swap remaining<< ] 2keep
380 swap [ ast>> push ] keep
386 : parse-seq-element ( result quot -- result )
387 '[ @ calc-seq-result ] [ f ] if* ; inline
389 M: seq-parser (compile)
391 [ input-slice V{ } clone <parse-result> ] %
393 parsers>> unclip compile-parser-quot [ parse-seq-element ] curry ,
394 [ compile-parser-quot [ merge-errors ] compose [ parse-seq-element ] curry , ] each
398 TUPLE: choice-parser parsers ;
400 M: choice-parser (compile)
403 parsers>> [ compile-parser-quot ] map
404 unclip , [ [ merge-errors ] compose , ] each
408 TUPLE: repeat0-parser parser ;
410 : (repeat) ( quot: ( -- result ) result -- result )
412 [ remaining>> swap remaining<< ] 2keep
413 ast>> swap [ ast>> push ] keep
417 ] if* ; inline recursive
419 M: repeat0-parser (compile)
420 parser>> compile-parser-quot '[
421 input-slice V{ } clone <parse-result> _ swap (repeat)
424 TUPLE: repeat1-parser parser ;
426 : repeat1-empty-check ( result -- result )
427 [ dup ast>> empty? [ drop f ] when ] [ f ] if* ;
429 M: repeat1-parser (compile)
430 parser>> compile-parser-quot '[
431 input-slice V{ } clone <parse-result> _ swap (repeat)
435 TUPLE: optional-parser parser ;
437 : check-optional ( result -- result )
438 [ input-slice f <parse-result> ] unless* ;
440 M: optional-parser (compile)
441 parser>> compile-parser-quot '[ @ check-optional ] ;
443 TUPLE: semantic-parser parser quot ;
445 : check-semantic ( result quot -- result )
446 dupd '[ dup ast>> @ [ drop f ] unless ] when ; inline
448 M: semantic-parser (compile)
449 [ parser>> compile-parser-quot ] [ quot>> ] bi
450 '[ @ _ check-semantic ] ;
452 TUPLE: ensure-parser parser ;
454 : check-ensure ( old-input result -- result )
455 [ ignore <parse-result> ] [ drop f ] if ;
457 M: ensure-parser (compile)
458 parser>> compile-parser-quot '[ input-slice @ check-ensure ] ;
460 TUPLE: ensure-not-parser parser ;
462 : check-ensure-not ( old-input result -- result )
463 [ drop f ] [ ignore <parse-result> ] if ;
465 M: ensure-not-parser (compile)
466 parser>> compile-parser-quot '[ input-slice @ check-ensure-not ] ;
468 TUPLE: action-parser parser quot ;
470 : check-action ( result quot -- result )
471 dupd '[ [ _ call( ast -- ast ) ] change-ast ] when ;
473 M: action-parser (compile)
474 [ parser>> compile-parser-quot ] [ quot>> ] bi
475 '[ @ _ check-action ] ;
477 TUPLE: sp-parser parser ;
479 M: sp-parser (compile)
480 parser>> compile-parser-quot '[
481 input-slice [ blank? ] trim-head-slice input-from pos namespaces:set @
484 TUPLE: delay-parser quot ;
486 M: delay-parser (compile)
487 ! For efficiency we memoize the quotation.
488 ! This way it is run only once and the
489 ! parser constructed once at run time.
490 quot>> gensym [ delayed get set-at ] keep 1quotation ;
492 TUPLE: box-parser quot ;
494 M: box-parser (compile)
495 ! Calls the quotation at compile time
496 ! to produce the parser to be compiled.
497 ! This differs from 'delay' which calls
499 quot>> call( -- parser ) compile-parser-quot ;
503 : token ( string -- parser )
504 token-parser boa wrap-peg ;
506 : satisfy ( quot -- parser )
507 satisfy-parser boa wrap-peg ;
509 : range ( min max -- parser )
510 range-parser boa wrap-peg ;
512 : seq ( seq -- parser )
513 seq-parser boa wrap-peg ;
515 : 2seq ( parser1 parser2 -- parser )
518 : 3seq ( parser1 parser2 parser3 -- parser )
521 : 4seq ( parser1 parser2 parser3 parser4 -- parser )
524 : seq* ( quot -- paser )
525 { } make seq ; inline
527 : choice ( seq -- parser )
528 choice-parser boa wrap-peg ;
530 : 2choice ( parser1 parser2 -- parser )
533 : 3choice ( parser1 parser2 parser3 -- parser )
536 : 4choice ( parser1 parser2 parser3 parser4 -- parser )
539 : choice* ( quot -- paser )
540 { } make choice ; inline
542 : repeat0 ( parser -- parser )
543 repeat0-parser boa wrap-peg ;
545 : repeat1 ( parser -- parser )
546 repeat1-parser boa wrap-peg ;
548 : optional ( parser -- parser )
549 optional-parser boa wrap-peg ;
551 : semantic ( parser quot -- parser )
552 semantic-parser boa wrap-peg ;
554 : ensure ( parser -- parser )
555 ensure-parser boa wrap-peg ;
557 : ensure-not ( parser -- parser )
558 ensure-not-parser boa wrap-peg ;
560 : action ( parser quot -- parser )
561 action-parser boa wrap-peg ;
563 : sp ( parser -- parser )
564 sp-parser boa wrap-peg ;
566 : hide ( parser -- parser )
567 [ drop ignore ] action ;
569 : delay ( quot -- parser )
570 delay-parser boa wrap-peg ;
572 : box ( quot -- parser )
573 ! because a box has its quotation run at compile time
574 ! it must always have a new parser wrapper created,
575 ! not a cached one. This is because the same box,
576 ! compiled twice can have a different compiled word
577 ! due to running at compile time.
578 ! Why the [ ] action at the end? Box parsers don't get
579 ! memoized during parsing due to all box parsers being
580 ! unique. This breaks left recursion detection during the
581 ! parse. The action adds an indirection with a parser type
582 ! that gets memoized and fixes this. Need to rethink how
583 ! to fix boxes so this isn't needed...
584 box-parser boa f next-id parser boa [ ] action ;
586 ERROR: parse-failed input word ;
590 (:) :> ( word def effect )
593 def call compile :> compiled-def
595 dup compiled-def compiled-parse
596 [ ast>> ] [ word parse-failed ] ?if
597 ] effect define-declared
598 ] with-compilation-unit
602 { "debugger" "peg" } "peg.debugger" require-when