Move a Misc/NEWS entry to right section.
[python.git] / Parser / spark.py
blobb064d62ec68bb7378c8e4705350724c348029f90
1 # Copyright (c) 1998-2002 John Aycock
3 # Permission is hereby granted, free of charge, to any person obtaining
4 # a copy of this software and associated documentation files (the
5 # "Software"), to deal in the Software without restriction, including
6 # without limitation the rights to use, copy, modify, merge, publish,
7 # distribute, sublicense, and/or sell copies of the Software, and to
8 # permit persons to whom the Software is furnished to do so, subject to
9 # the following conditions:
11 # The above copyright notice and this permission notice shall be
12 # included in all copies or substantial portions of the Software.
14 # THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
15 # EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
16 # MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
17 # IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY
18 # CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
19 # TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
20 # SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
22 __version__ = 'SPARK-0.7 (pre-alpha-5)'
24 import re
25 import string
27 def _namelist(instance):
28 namelist, namedict, classlist = [], {}, [instance.__class__]
29 for c in classlist:
30 for b in c.__bases__:
31 classlist.append(b)
32 for name in c.__dict__.keys():
33 if not namedict.has_key(name):
34 namelist.append(name)
35 namedict[name] = 1
36 return namelist
38 class GenericScanner:
39 def __init__(self, flags=0):
40 pattern = self.reflect()
41 self.re = re.compile(pattern, re.VERBOSE|flags)
43 self.index2func = {}
44 for name, number in self.re.groupindex.items():
45 self.index2func[number-1] = getattr(self, 't_' + name)
47 def makeRE(self, name):
48 doc = getattr(self, name).__doc__
49 rv = '(?P<%s>%s)' % (name[2:], doc)
50 return rv
52 def reflect(self):
53 rv = []
54 for name in _namelist(self):
55 if name[:2] == 't_' and name != 't_default':
56 rv.append(self.makeRE(name))
58 rv.append(self.makeRE('t_default'))
59 return string.join(rv, '|')
61 def error(self, s, pos):
62 print "Lexical error at position %s" % pos
63 raise SystemExit
65 def tokenize(self, s):
66 pos = 0
67 n = len(s)
68 while pos < n:
69 m = self.re.match(s, pos)
70 if m is None:
71 self.error(s, pos)
73 groups = m.groups()
74 for i in range(len(groups)):
75 if groups[i] and self.index2func.has_key(i):
76 self.index2func[i](groups[i])
77 pos = m.end()
79 def t_default(self, s):
80 r'( . | \n )+'
81 print "Specification error: unmatched input"
82 raise SystemExit
85 # Extracted from GenericParser and made global so that [un]picking works.
87 class _State:
88 def __init__(self, stateno, items):
89 self.T, self.complete, self.items = [], [], items
90 self.stateno = stateno
92 class GenericParser:
94 # An Earley parser, as per J. Earley, "An Efficient Context-Free
95 # Parsing Algorithm", CACM 13(2), pp. 94-102. Also J. C. Earley,
96 # "An Efficient Context-Free Parsing Algorithm", Ph.D. thesis,
97 # Carnegie-Mellon University, August 1968. New formulation of
98 # the parser according to J. Aycock, "Practical Earley Parsing
99 # and the SPARK Toolkit", Ph.D. thesis, University of Victoria,
100 # 2001, and J. Aycock and R. N. Horspool, "Practical Earley
101 # Parsing", unpublished paper, 2001.
104 def __init__(self, start):
105 self.rules = {}
106 self.rule2func = {}
107 self.rule2name = {}
108 self.collectRules()
109 self.augment(start)
110 self.ruleschanged = 1
112 _NULLABLE = '\e_'
113 _START = 'START'
114 _BOF = '|-'
117 # When pickling, take the time to generate the full state machine;
118 # some information is then extraneous, too. Unfortunately we
119 # can't save the rule2func map.
121 def __getstate__(self):
122 if self.ruleschanged:
124 # XXX - duplicated from parse()
126 self.computeNull()
127 self.newrules = {}
128 self.new2old = {}
129 self.makeNewRules()
130 self.ruleschanged = 0
131 self.edges, self.cores = {}, {}
132 self.states = { 0: self.makeState0() }
133 self.makeState(0, self._BOF)
135 # XXX - should find a better way to do this..
137 changes = 1
138 while changes:
139 changes = 0
140 for k, v in self.edges.items():
141 if v is None:
142 state, sym = k
143 if self.states.has_key(state):
144 self.goto(state, sym)
145 changes = 1
146 rv = self.__dict__.copy()
147 for s in self.states.values():
148 del s.items
149 del rv['rule2func']
150 del rv['nullable']
151 del rv['cores']
152 return rv
154 def __setstate__(self, D):
155 self.rules = {}
156 self.rule2func = {}
157 self.rule2name = {}
158 self.collectRules()
159 start = D['rules'][self._START][0][1][1] # Blech.
160 self.augment(start)
161 D['rule2func'] = self.rule2func
162 D['makeSet'] = self.makeSet_fast
163 self.__dict__ = D
166 # A hook for GenericASTBuilder and GenericASTMatcher. Mess
167 # thee not with this; nor shall thee toucheth the _preprocess
168 # argument to addRule.
170 def preprocess(self, rule, func): return rule, func
172 def addRule(self, doc, func, _preprocess=1):
173 fn = func
174 rules = string.split(doc)
176 index = []
177 for i in range(len(rules)):
178 if rules[i] == '::=':
179 index.append(i-1)
180 index.append(len(rules))
182 for i in range(len(index)-1):
183 lhs = rules[index[i]]
184 rhs = rules[index[i]+2:index[i+1]]
185 rule = (lhs, tuple(rhs))
187 if _preprocess:
188 rule, fn = self.preprocess(rule, func)
190 if self.rules.has_key(lhs):
191 self.rules[lhs].append(rule)
192 else:
193 self.rules[lhs] = [ rule ]
194 self.rule2func[rule] = fn
195 self.rule2name[rule] = func.__name__[2:]
196 self.ruleschanged = 1
198 def collectRules(self):
199 for name in _namelist(self):
200 if name[:2] == 'p_':
201 func = getattr(self, name)
202 doc = func.__doc__
203 self.addRule(doc, func)
205 def augment(self, start):
206 rule = '%s ::= %s %s' % (self._START, self._BOF, start)
207 self.addRule(rule, lambda args: args[1], 0)
209 def computeNull(self):
210 self.nullable = {}
211 tbd = []
213 for rulelist in self.rules.values():
214 lhs = rulelist[0][0]
215 self.nullable[lhs] = 0
216 for rule in rulelist:
217 rhs = rule[1]
218 if len(rhs) == 0:
219 self.nullable[lhs] = 1
220 continue
222 # We only need to consider rules which
223 # consist entirely of nonterminal symbols.
224 # This should be a savings on typical
225 # grammars.
227 for sym in rhs:
228 if not self.rules.has_key(sym):
229 break
230 else:
231 tbd.append(rule)
232 changes = 1
233 while changes:
234 changes = 0
235 for lhs, rhs in tbd:
236 if self.nullable[lhs]:
237 continue
238 for sym in rhs:
239 if not self.nullable[sym]:
240 break
241 else:
242 self.nullable[lhs] = 1
243 changes = 1
245 def makeState0(self):
246 s0 = _State(0, [])
247 for rule in self.newrules[self._START]:
248 s0.items.append((rule, 0))
249 return s0
251 def finalState(self, tokens):
253 # Yuck.
255 if len(self.newrules[self._START]) == 2 and len(tokens) == 0:
256 return 1
257 start = self.rules[self._START][0][1][1]
258 return self.goto(1, start)
260 def makeNewRules(self):
261 worklist = []
262 for rulelist in self.rules.values():
263 for rule in rulelist:
264 worklist.append((rule, 0, 1, rule))
266 for rule, i, candidate, oldrule in worklist:
267 lhs, rhs = rule
268 n = len(rhs)
269 while i < n:
270 sym = rhs[i]
271 if not self.rules.has_key(sym) or \
272 not self.nullable[sym]:
273 candidate = 0
274 i = i + 1
275 continue
277 newrhs = list(rhs)
278 newrhs[i] = self._NULLABLE+sym
279 newrule = (lhs, tuple(newrhs))
280 worklist.append((newrule, i+1,
281 candidate, oldrule))
282 candidate = 0
283 i = i + 1
284 else:
285 if candidate:
286 lhs = self._NULLABLE+lhs
287 rule = (lhs, rhs)
288 if self.newrules.has_key(lhs):
289 self.newrules[lhs].append(rule)
290 else:
291 self.newrules[lhs] = [ rule ]
292 self.new2old[rule] = oldrule
294 def typestring(self, token):
295 return None
297 def error(self, token):
298 print "Syntax error at or near `%s' token" % token
299 raise SystemExit
301 def parse(self, tokens):
302 sets = [ [(1,0), (2,0)] ]
303 self.links = {}
305 if self.ruleschanged:
306 self.computeNull()
307 self.newrules = {}
308 self.new2old = {}
309 self.makeNewRules()
310 self.ruleschanged = 0
311 self.edges, self.cores = {}, {}
312 self.states = { 0: self.makeState0() }
313 self.makeState(0, self._BOF)
315 for i in xrange(len(tokens)):
316 sets.append([])
318 if sets[i] == []:
319 break
320 self.makeSet(tokens[i], sets, i)
321 else:
322 sets.append([])
323 self.makeSet(None, sets, len(tokens))
325 #_dump(tokens, sets, self.states)
327 finalitem = (self.finalState(tokens), 0)
328 if finalitem not in sets[-2]:
329 if len(tokens) > 0:
330 self.error(tokens[i-1])
331 else:
332 self.error(None)
334 return self.buildTree(self._START, finalitem,
335 tokens, len(sets)-2)
337 def isnullable(self, sym):
339 # For symbols in G_e only. If we weren't supporting 1.5,
340 # could just use sym.startswith().
342 return self._NULLABLE == sym[0:len(self._NULLABLE)]
344 def skip(self, (lhs, rhs), pos=0):
345 n = len(rhs)
346 while pos < n:
347 if not self.isnullable(rhs[pos]):
348 break
349 pos = pos + 1
350 return pos
352 def makeState(self, state, sym):
353 assert sym is not None
355 # Compute \epsilon-kernel state's core and see if
356 # it exists already.
358 kitems = []
359 for rule, pos in self.states[state].items:
360 lhs, rhs = rule
361 if rhs[pos:pos+1] == (sym,):
362 kitems.append((rule, self.skip(rule, pos+1)))
363 core = kitems
365 core.sort()
366 tcore = tuple(core)
367 if self.cores.has_key(tcore):
368 return self.cores[tcore]
370 # Nope, doesn't exist. Compute it and the associated
371 # \epsilon-nonkernel state together; we'll need it right away.
373 k = self.cores[tcore] = len(self.states)
374 K, NK = _State(k, kitems), _State(k+1, [])
375 self.states[k] = K
376 predicted = {}
378 edges = self.edges
379 rules = self.newrules
380 for X in K, NK:
381 worklist = X.items
382 for item in worklist:
383 rule, pos = item
384 lhs, rhs = rule
385 if pos == len(rhs):
386 X.complete.append(rule)
387 continue
389 nextSym = rhs[pos]
390 key = (X.stateno, nextSym)
391 if not rules.has_key(nextSym):
392 if not edges.has_key(key):
393 edges[key] = None
394 X.T.append(nextSym)
395 else:
396 edges[key] = None
397 if not predicted.has_key(nextSym):
398 predicted[nextSym] = 1
399 for prule in rules[nextSym]:
400 ppos = self.skip(prule)
401 new = (prule, ppos)
402 NK.items.append(new)
404 # Problem: we know K needs generating, but we
405 # don't yet know about NK. Can't commit anything
406 # regarding NK to self.edges until we're sure. Should
407 # we delay committing on both K and NK to avoid this
408 # hacky code? This creates other problems..
410 if X is K:
411 edges = {}
413 if NK.items == []:
414 return k
417 # Check for \epsilon-nonkernel's core. Unfortunately we
418 # need to know the entire set of predicted nonterminals
419 # to do this without accidentally duplicating states.
421 core = predicted.keys()
422 core.sort()
423 tcore = tuple(core)
424 if self.cores.has_key(tcore):
425 self.edges[(k, None)] = self.cores[tcore]
426 return k
428 nk = self.cores[tcore] = self.edges[(k, None)] = NK.stateno
429 self.edges.update(edges)
430 self.states[nk] = NK
431 return k
433 def goto(self, state, sym):
434 key = (state, sym)
435 if not self.edges.has_key(key):
437 # No transitions from state on sym.
439 return None
441 rv = self.edges[key]
442 if rv is None:
444 # Target state isn't generated yet. Remedy this.
446 rv = self.makeState(state, sym)
447 self.edges[key] = rv
448 return rv
450 def gotoT(self, state, t):
451 return [self.goto(state, t)]
453 def gotoST(self, state, st):
454 rv = []
455 for t in self.states[state].T:
456 if st == t:
457 rv.append(self.goto(state, t))
458 return rv
460 def add(self, set, item, i=None, predecessor=None, causal=None):
461 if predecessor is None:
462 if item not in set:
463 set.append(item)
464 else:
465 key = (item, i)
466 if item not in set:
467 self.links[key] = []
468 set.append(item)
469 self.links[key].append((predecessor, causal))
471 def makeSet(self, token, sets, i):
472 cur, next = sets[i], sets[i+1]
474 ttype = token is not None and self.typestring(token) or None
475 if ttype is not None:
476 fn, arg = self.gotoT, ttype
477 else:
478 fn, arg = self.gotoST, token
480 for item in cur:
481 ptr = (item, i)
482 state, parent = item
483 add = fn(state, arg)
484 for k in add:
485 if k is not None:
486 self.add(next, (k, parent), i+1, ptr)
487 nk = self.goto(k, None)
488 if nk is not None:
489 self.add(next, (nk, i+1))
491 if parent == i:
492 continue
494 for rule in self.states[state].complete:
495 lhs, rhs = rule
496 for pitem in sets[parent]:
497 pstate, pparent = pitem
498 k = self.goto(pstate, lhs)
499 if k is not None:
500 why = (item, i, rule)
501 pptr = (pitem, parent)
502 self.add(cur, (k, pparent),
503 i, pptr, why)
504 nk = self.goto(k, None)
505 if nk is not None:
506 self.add(cur, (nk, i))
508 def makeSet_fast(self, token, sets, i):
510 # Call *only* when the entire state machine has been built!
511 # It relies on self.edges being filled in completely, and
512 # then duplicates and inlines code to boost speed at the
513 # cost of extreme ugliness.
515 cur, next = sets[i], sets[i+1]
516 ttype = token is not None and self.typestring(token) or None
518 for item in cur:
519 ptr = (item, i)
520 state, parent = item
521 if ttype is not None:
522 k = self.edges.get((state, ttype), None)
523 if k is not None:
524 #self.add(next, (k, parent), i+1, ptr)
525 #INLINED --v
526 new = (k, parent)
527 key = (new, i+1)
528 if new not in next:
529 self.links[key] = []
530 next.append(new)
531 self.links[key].append((ptr, None))
532 #INLINED --^
533 #nk = self.goto(k, None)
534 nk = self.edges.get((k, None), None)
535 if nk is not None:
536 #self.add(next, (nk, i+1))
537 #INLINED --v
538 new = (nk, i+1)
539 if new not in next:
540 next.append(new)
541 #INLINED --^
542 else:
543 add = self.gotoST(state, token)
544 for k in add:
545 if k is not None:
546 self.add(next, (k, parent), i+1, ptr)
547 #nk = self.goto(k, None)
548 nk = self.edges.get((k, None), None)
549 if nk is not None:
550 self.add(next, (nk, i+1))
552 if parent == i:
553 continue
555 for rule in self.states[state].complete:
556 lhs, rhs = rule
557 for pitem in sets[parent]:
558 pstate, pparent = pitem
559 #k = self.goto(pstate, lhs)
560 k = self.edges.get((pstate, lhs), None)
561 if k is not None:
562 why = (item, i, rule)
563 pptr = (pitem, parent)
564 #self.add(cur, (k, pparent),
565 # i, pptr, why)
566 #INLINED --v
567 new = (k, pparent)
568 key = (new, i)
569 if new not in cur:
570 self.links[key] = []
571 cur.append(new)
572 self.links[key].append((pptr, why))
573 #INLINED --^
574 #nk = self.goto(k, None)
575 nk = self.edges.get((k, None), None)
576 if nk is not None:
577 #self.add(cur, (nk, i))
578 #INLINED --v
579 new = (nk, i)
580 if new not in cur:
581 cur.append(new)
582 #INLINED --^
584 def predecessor(self, key, causal):
585 for p, c in self.links[key]:
586 if c == causal:
587 return p
588 assert 0
590 def causal(self, key):
591 links = self.links[key]
592 if len(links) == 1:
593 return links[0][1]
594 choices = []
595 rule2cause = {}
596 for p, c in links:
597 rule = c[2]
598 choices.append(rule)
599 rule2cause[rule] = c
600 return rule2cause[self.ambiguity(choices)]
602 def deriveEpsilon(self, nt):
603 if len(self.newrules[nt]) > 1:
604 rule = self.ambiguity(self.newrules[nt])
605 else:
606 rule = self.newrules[nt][0]
607 #print rule
609 rhs = rule[1]
610 attr = [None] * len(rhs)
612 for i in range(len(rhs)-1, -1, -1):
613 attr[i] = self.deriveEpsilon(rhs[i])
614 return self.rule2func[self.new2old[rule]](attr)
616 def buildTree(self, nt, item, tokens, k):
617 state, parent = item
619 choices = []
620 for rule in self.states[state].complete:
621 if rule[0] == nt:
622 choices.append(rule)
623 rule = choices[0]
624 if len(choices) > 1:
625 rule = self.ambiguity(choices)
626 #print rule
628 rhs = rule[1]
629 attr = [None] * len(rhs)
631 for i in range(len(rhs)-1, -1, -1):
632 sym = rhs[i]
633 if not self.newrules.has_key(sym):
634 if sym != self._BOF:
635 attr[i] = tokens[k-1]
636 key = (item, k)
637 item, k = self.predecessor(key, None)
638 #elif self.isnullable(sym):
639 elif self._NULLABLE == sym[0:len(self._NULLABLE)]:
640 attr[i] = self.deriveEpsilon(sym)
641 else:
642 key = (item, k)
643 why = self.causal(key)
644 attr[i] = self.buildTree(sym, why[0],
645 tokens, why[1])
646 item, k = self.predecessor(key, why)
647 return self.rule2func[self.new2old[rule]](attr)
649 def ambiguity(self, rules):
651 # XXX - problem here and in collectRules() if the same rule
652 # appears in >1 method. Also undefined results if rules
653 # causing the ambiguity appear in the same method.
655 sortlist = []
656 name2index = {}
657 for i in range(len(rules)):
658 lhs, rhs = rule = rules[i]
659 name = self.rule2name[self.new2old[rule]]
660 sortlist.append((len(rhs), name))
661 name2index[name] = i
662 sortlist.sort()
663 list = map(lambda (a,b): b, sortlist)
664 return rules[name2index[self.resolve(list)]]
666 def resolve(self, list):
668 # Resolve ambiguity in favor of the shortest RHS.
669 # Since we walk the tree from the top down, this
670 # should effectively resolve in favor of a "shift".
672 return list[0]
675 # GenericASTBuilder automagically constructs a concrete/abstract syntax tree
676 # for a given input. The extra argument is a class (not an instance!)
677 # which supports the "__setslice__" and "__len__" methods.
679 # XXX - silently overrides any user code in methods.
682 class GenericASTBuilder(GenericParser):
683 def __init__(self, AST, start):
684 GenericParser.__init__(self, start)
685 self.AST = AST
687 def preprocess(self, rule, func):
688 rebind = lambda lhs, self=self: \
689 lambda args, lhs=lhs, self=self: \
690 self.buildASTNode(args, lhs)
691 lhs, rhs = rule
692 return rule, rebind(lhs)
694 def buildASTNode(self, args, lhs):
695 children = []
696 for arg in args:
697 if isinstance(arg, self.AST):
698 children.append(arg)
699 else:
700 children.append(self.terminal(arg))
701 return self.nonterminal(lhs, children)
703 def terminal(self, token): return token
705 def nonterminal(self, type, args):
706 rv = self.AST(type)
707 rv[:len(args)] = args
708 return rv
711 # GenericASTTraversal is a Visitor pattern according to Design Patterns. For
712 # each node it attempts to invoke the method n_<node type>, falling
713 # back onto the default() method if the n_* can't be found. The preorder
714 # traversal also looks for an exit hook named n_<node type>_exit (no default
715 # routine is called if it's not found). To prematurely halt traversal
716 # of a subtree, call the prune() method -- this only makes sense for a
717 # preorder traversal. Node type is determined via the typestring() method.
720 class GenericASTTraversalPruningException:
721 pass
723 class GenericASTTraversal:
724 def __init__(self, ast):
725 self.ast = ast
727 def typestring(self, node):
728 return node.type
730 def prune(self):
731 raise GenericASTTraversalPruningException
733 def preorder(self, node=None):
734 if node is None:
735 node = self.ast
737 try:
738 name = 'n_' + self.typestring(node)
739 if hasattr(self, name):
740 func = getattr(self, name)
741 func(node)
742 else:
743 self.default(node)
744 except GenericASTTraversalPruningException:
745 return
747 for kid in node:
748 self.preorder(kid)
750 name = name + '_exit'
751 if hasattr(self, name):
752 func = getattr(self, name)
753 func(node)
755 def postorder(self, node=None):
756 if node is None:
757 node = self.ast
759 for kid in node:
760 self.postorder(kid)
762 name = 'n_' + self.typestring(node)
763 if hasattr(self, name):
764 func = getattr(self, name)
765 func(node)
766 else:
767 self.default(node)
770 def default(self, node):
771 pass
774 # GenericASTMatcher. AST nodes must have "__getitem__" and "__cmp__"
775 # implemented.
777 # XXX - makes assumptions about how GenericParser walks the parse tree.
780 class GenericASTMatcher(GenericParser):
781 def __init__(self, start, ast):
782 GenericParser.__init__(self, start)
783 self.ast = ast
785 def preprocess(self, rule, func):
786 rebind = lambda func, self=self: \
787 lambda args, func=func, self=self: \
788 self.foundMatch(args, func)
789 lhs, rhs = rule
790 rhslist = list(rhs)
791 rhslist.reverse()
793 return (lhs, tuple(rhslist)), rebind(func)
795 def foundMatch(self, args, func):
796 func(args[-1])
797 return args[-1]
799 def match_r(self, node):
800 self.input.insert(0, node)
801 children = 0
803 for child in node:
804 if children == 0:
805 self.input.insert(0, '(')
806 children = children + 1
807 self.match_r(child)
809 if children > 0:
810 self.input.insert(0, ')')
812 def match(self, ast=None):
813 if ast is None:
814 ast = self.ast
815 self.input = []
817 self.match_r(ast)
818 self.parse(self.input)
820 def resolve(self, list):
822 # Resolve ambiguity in favor of the longest RHS.
824 return list[-1]
826 def _dump(tokens, sets, states):
827 for i in range(len(sets)):
828 print 'set', i
829 for item in sets[i]:
830 print '\t', item
831 for (lhs, rhs), pos in states[item[0]].items:
832 print '\t\t', lhs, '::=',
833 print string.join(rhs[:pos]),
834 print '.',
835 print string.join(rhs[pos:])
836 if i < len(tokens):
837 print
838 print 'token', str(tokens[i])
839 print