1 | # Copyright 2004-2005 Elemental Security, Inc. All Rights Reserved.
|
2 | # Licensed to PSF under a Contributor Agreement.
|
3 |
|
4 | # Pgen imports
|
5 | #import grammar, token, tokenize
|
6 | # NOTE: Need these special versions of token/tokenize for BACKQUOTE and such.
|
7 | from . import grammar, token, tokenize
|
8 | from mycpp.mylib import log
|
9 |
|
10 | _ = log
|
11 |
|
12 |
|
13 | class PythonTokDef(object):
|
14 |
|
15 | def GetTerminalNum(self, label):
|
16 | """ e.g. NAME -> 1 """
|
17 | itoken = getattr(token, label, None)
|
18 | assert isinstance(itoken, int), label
|
19 | assert itoken in token.tok_name, label
|
20 | return itoken
|
21 |
|
22 | def GetKeywordNum(self, value):
|
23 | """
|
24 | e.g 'xor' -> Id.Expr_Xor
|
25 |
|
26 | Python doesn't have this, but Oil does. Returns None if not found.
|
27 | """
|
28 | return None
|
29 |
|
30 | def GetOpNum(self, op_str):
|
31 | """
|
32 | e.g '(' -> LPAR
|
33 |
|
34 | Raises an exception if it's not found.
|
35 | """
|
36 | return token.opmap[op_str]
|
37 |
|
38 |
|
39 | class ParserGenerator(object):
|
40 |
|
41 | def __init__(self, lexer):
|
42 | self.lexer = lexer
|
43 |
|
44 | def parse(self):
|
45 | self.gettoken() # Initialize lookahead
|
46 |
|
47 | dfas = {}
|
48 | startsymbol = None
|
49 | # MSTART: (NEWLINE | RULE)* ENDMARKER
|
50 | while self.type != token.ENDMARKER:
|
51 | while self.type == token.NEWLINE:
|
52 | self.gettoken()
|
53 | # RULE: NAME ':' RHS NEWLINE
|
54 | name = self.expect(token.NAME)
|
55 | self.expect(token.OP, ":")
|
56 | a, z = self.parse_rhs()
|
57 | self.expect(token.NEWLINE)
|
58 | #self.dump_nfa(name, a, z)
|
59 | dfa = self.make_dfa(a, z)
|
60 | #self.dump_dfa(name, dfa)
|
61 | oldlen = len(dfa)
|
62 | self.simplify_dfa(dfa)
|
63 | newlen = len(dfa)
|
64 | dfas[name] = dfa
|
65 | #print name, oldlen, newlen
|
66 | if startsymbol is None:
|
67 | startsymbol = name
|
68 | return dfas, startsymbol
|
69 |
|
70 | def make_dfa(self, start, finish):
|
71 | # To turn an NFA into a DFA, we define the states of the DFA
|
72 | # to correspond to *sets* of states of the NFA. Then do some
|
73 | # state reduction. Let's represent sets as dicts with 1 for
|
74 | # values.
|
75 | assert isinstance(start, NFAState)
|
76 | assert isinstance(finish, NFAState)
|
77 | def closure(state):
|
78 | base = {}
|
79 | addclosure(state, base)
|
80 | return base
|
81 | def addclosure(state, base):
|
82 | assert isinstance(state, NFAState)
|
83 | if state in base:
|
84 | return
|
85 | base[state] = 1
|
86 | for label, next in state.arcs:
|
87 | if label is None:
|
88 | addclosure(next, base)
|
89 | states = [DFAState(closure(start), finish)]
|
90 | for state in states: # NB states grows while we're iterating
|
91 | arcs = {}
|
92 | for nfastate in state.nfaset:
|
93 | for label, next in nfastate.arcs:
|
94 | if label is not None:
|
95 | addclosure(next, arcs.setdefault(label, {}))
|
96 | for label, nfaset in sorted(arcs.items()):
|
97 | for st in states:
|
98 | if st.nfaset == nfaset:
|
99 | break
|
100 | else:
|
101 | st = DFAState(nfaset, finish)
|
102 | states.append(st)
|
103 | state.addarc(st, label)
|
104 | return states # List of DFAState instances; first one is start
|
105 |
|
106 | def dump_nfa(self, name, start, finish):
|
107 | print("Dump of NFA for", name)
|
108 | todo = [start]
|
109 | for i, state in enumerate(todo):
|
110 | print(" State", i, state is finish and "(final)" or "")
|
111 | for label, next in state.arcs:
|
112 | if next in todo:
|
113 | j = todo.index(next)
|
114 | else:
|
115 | j = len(todo)
|
116 | todo.append(next)
|
117 | if label is None:
|
118 | print(" -> %d" % j)
|
119 | else:
|
120 | print(" %s -> %d" % (label, j))
|
121 |
|
122 | def dump_dfa(self, name, dfa):
|
123 | print("Dump of DFA for", name)
|
124 | for i, state in enumerate(dfa):
|
125 | print(" State", i, state.isfinal and "(final)" or "")
|
126 | for label, next in sorted(state.arcs.items()):
|
127 | print(" %s -> %d" % (label, dfa.index(next)))
|
128 |
|
129 | def simplify_dfa(self, dfa):
|
130 | # This is not theoretically optimal, but works well enough.
|
131 | # Algorithm: repeatedly look for two states that have the same
|
132 | # set of arcs (same labels pointing to the same nodes) and
|
133 | # unify them, until things stop changing.
|
134 |
|
135 | # dfa is a list of DFAState instances
|
136 | changes = True
|
137 | while changes:
|
138 | changes = False
|
139 | for i, state_i in enumerate(dfa):
|
140 | for j in range(i+1, len(dfa)):
|
141 | state_j = dfa[j]
|
142 | if state_i == state_j:
|
143 | #print " unify", i, j
|
144 | del dfa[j]
|
145 | for state in dfa:
|
146 | state.unifystate(state_j, state_i)
|
147 | changes = True
|
148 | break
|
149 |
|
150 | def parse_rhs(self):
|
151 | # RHS: ALT ('|' ALT)*
|
152 | a, z = self.parse_alt()
|
153 | if self.value != "|":
|
154 | return a, z
|
155 | else:
|
156 | aa = NFAState()
|
157 | zz = NFAState()
|
158 | aa.addarc(a)
|
159 | z.addarc(zz)
|
160 | while self.value == "|":
|
161 | self.gettoken()
|
162 | a, z = self.parse_alt()
|
163 | aa.addarc(a)
|
164 | z.addarc(zz)
|
165 | return aa, zz
|
166 |
|
167 | def parse_alt(self):
|
168 | # ALT: ITEM+
|
169 | a, b = self.parse_item()
|
170 | while (self.value in ("(", "[") or
|
171 | self.type in (token.NAME, token.STRING)):
|
172 | c, d = self.parse_item()
|
173 | b.addarc(c)
|
174 | b = d
|
175 | return a, b
|
176 |
|
177 | def parse_item(self):
|
178 | # ITEM: '[' RHS ']' | ATOM ['+' | '*']
|
179 | if self.value == "[":
|
180 | self.gettoken()
|
181 | a, z = self.parse_rhs()
|
182 | self.expect(token.OP, "]")
|
183 | a.addarc(z)
|
184 | return a, z
|
185 | else:
|
186 | a, z = self.parse_atom()
|
187 | value = self.value
|
188 | if value not in ("+", "*"):
|
189 | return a, z
|
190 | self.gettoken()
|
191 | z.addarc(a)
|
192 | if value == "+":
|
193 | return a, z
|
194 | else:
|
195 | return a, a
|
196 |
|
197 | def parse_atom(self):
|
198 | # ATOM: '(' RHS ')' | NAME | STRING
|
199 | if self.value == "(":
|
200 | self.gettoken()
|
201 | a, z = self.parse_rhs()
|
202 | self.expect(token.OP, ")")
|
203 | return a, z
|
204 | elif self.type in (token.NAME, token.STRING):
|
205 | a = NFAState()
|
206 | z = NFAState()
|
207 | a.addarc(z, self.value)
|
208 | self.gettoken()
|
209 | return a, z
|
210 | else:
|
211 | self.raise_error("expected (...) or NAME or STRING, got %s/%s",
|
212 | self.type, self.value)
|
213 |
|
214 | def expect(self, type, value=None):
|
215 | if self.type != type or (value is not None and self.value != value):
|
216 | self.raise_error("expected %s/%s, got %s/%s",
|
217 | type, value, self.type, self.value)
|
218 | value = self.value
|
219 | self.gettoken()
|
220 | return value
|
221 |
|
222 | def gettoken(self):
|
223 | tup = next(self.lexer)
|
224 | while tup[0] in (tokenize.COMMENT, tokenize.NL):
|
225 | tup = next(self.lexer)
|
226 | self.type, self.value, self.begin, self.end, self.line = tup
|
227 | #print token.tok_name[self.type], repr(self.value)
|
228 |
|
229 | def raise_error(self, msg, *args):
|
230 | if args:
|
231 | msg = msg % args
|
232 | raise SyntaxError(msg, ('<grammar>', self.end[0],
|
233 | self.end[1], self.line))
|
234 |
|
235 | class NFAState(object):
|
236 |
|
237 | def __init__(self):
|
238 | self.arcs = [] # list of (label, NFAState) pairs
|
239 |
|
240 | def addarc(self, next, label=None):
|
241 | assert label is None or isinstance(label, str)
|
242 | assert isinstance(next, NFAState)
|
243 | self.arcs.append((label, next))
|
244 |
|
245 | class DFAState(object):
|
246 |
|
247 | def __init__(self, nfaset, final):
|
248 | assert isinstance(nfaset, dict)
|
249 | assert isinstance(next(iter(nfaset)), NFAState)
|
250 | assert isinstance(final, NFAState)
|
251 | self.nfaset = nfaset
|
252 | self.isfinal = final in nfaset
|
253 | self.arcs = {} # map from label to DFAState
|
254 |
|
255 | def addarc(self, next, label):
|
256 | assert isinstance(label, str)
|
257 | assert label not in self.arcs
|
258 | assert isinstance(next, DFAState)
|
259 | self.arcs[label] = next
|
260 |
|
261 | def unifystate(self, old, new):
|
262 | for label, next in self.arcs.items():
|
263 | if next is old:
|
264 | self.arcs[label] = new
|
265 |
|
266 | def __eq__(self, other):
|
267 | # Equality test -- ignore the nfaset instance variable
|
268 | assert isinstance(other, DFAState)
|
269 | if self.isfinal != other.isfinal:
|
270 | return False
|
271 | # Can't just return self.arcs == other.arcs, because that
|
272 | # would invoke this method recursively, with cycles...
|
273 | if len(self.arcs) != len(other.arcs):
|
274 | return False
|
275 | for label, next in self.arcs.items():
|
276 | if next is not other.arcs.get(label):
|
277 | return False
|
278 | return True
|
279 |
|
280 | __hash__ = None # For Py3 compatibility.
|
281 |
|
282 |
|
283 | def calcfirst(dfas, first, name):
|
284 | """Recursive function that mutates first."""
|
285 | dfa = dfas[name]
|
286 | first[name] = None # dummy to detect left recursion
|
287 | state = dfa[0]
|
288 | totalset = {}
|
289 | overlapcheck = {}
|
290 | for label, _ in state.arcs.items():
|
291 | if label in dfas:
|
292 | if label in first:
|
293 | fset = first[label]
|
294 | if fset is None:
|
295 | raise ValueError("recursion for rule %r" % name)
|
296 | else:
|
297 | calcfirst(dfas, first, label)
|
298 | fset = first[label]
|
299 | totalset.update(fset)
|
300 | overlapcheck[label] = fset
|
301 | else:
|
302 | totalset[label] = 1
|
303 | overlapcheck[label] = {label: 1}
|
304 | inverse = {}
|
305 | for label, itsfirst in overlapcheck.items():
|
306 | for symbol in itsfirst:
|
307 | if symbol in inverse:
|
308 | raise ValueError("rule %s is ambiguous; %s is in the"
|
309 | " first sets of %s as well as %s" %
|
310 | (name, symbol, label, inverse[symbol]))
|
311 | inverse[symbol] = label
|
312 | first[name] = totalset
|
313 |
|
314 |
|
315 | def make_label(tok_def, gr, label):
|
316 | """Given a grammar item, return a unique integer representing it.
|
317 |
|
318 | It could be:
|
319 | 1. or_test - a non-terminal
|
320 | 2. Expr_Name - a terminal
|
321 | 3. 'for' - keyword (quotes)
|
322 | 4. '>=' - operator (quotes)
|
323 |
|
324 | Oil addition
|
325 | 5. Op_RBracket -- anything with _ is assumed to be in the Id namespace.
|
326 | """
|
327 | #log('make_label %r', label)
|
328 | # XXX Maybe this should be a method on a subclass of converter?
|
329 | ilabel = len(gr.labels)
|
330 | if label[0].isalpha():
|
331 | if label in gr.symbol2number: # NON-TERMINAL
|
332 | if label in gr.symbol2label:
|
333 | return gr.symbol2label[label]
|
334 | else:
|
335 | gr.labels.append(gr.symbol2number[label])
|
336 | gr.symbol2label[label] = ilabel
|
337 | return ilabel
|
338 | else: # TERMINAL like (NAME, NUMBER, STRING)
|
339 | itoken = tok_def.GetTerminalNum(label)
|
340 | if itoken in gr.tokens:
|
341 | return gr.tokens[itoken]
|
342 | else:
|
343 | gr.labels.append(itoken)
|
344 | #log('%s %d -> %s', token.tok_name[itoken], itoken, ilabel)
|
345 | gr.tokens[itoken] = ilabel
|
346 | return ilabel
|
347 | else:
|
348 | # Either a keyword or an operator
|
349 | assert label[0] in ('"', "'"), label
|
350 | value = eval(label)
|
351 |
|
352 | # Treat 'xor' just like '^'. TODO: I think this code can be
|
353 | # simplified.
|
354 | n = tok_def.GetKeywordNum(value) # int or None
|
355 |
|
356 | if value[0].isalpha() and n is None: # A word like 'for', 'xor'
|
357 | # Then look in the keywords automatically extracted from the
|
358 | # grammar.
|
359 | if value in gr.keywords:
|
360 | return gr.keywords[value]
|
361 | else:
|
362 | gr.labels.append(token.NAME) # arbitrary number < NT_OFFSET
|
363 | gr.keywords[value] = ilabel
|
364 | return ilabel
|
365 |
|
366 | else: # An operator e.g. '>='
|
367 | if n is None:
|
368 | itoken = tok_def.GetOpNum(value)
|
369 | else:
|
370 | itoken = n
|
371 |
|
372 | if itoken in gr.tokens:
|
373 | return gr.tokens[itoken]
|
374 | else:
|
375 | gr.labels.append(itoken)
|
376 | gr.tokens[itoken] = ilabel
|
377 | return ilabel
|
378 |
|
379 |
|
380 | def make_first(tok_def, rawfirst, gr):
|
381 | first = {}
|
382 | for label in sorted(rawfirst):
|
383 | ilabel = make_label(tok_def, gr, label)
|
384 | ##assert ilabel not in first # XXX failed on <> ... !=
|
385 | first[ilabel] = 1
|
386 | return first
|
387 |
|
388 |
|
389 | def MakeGrammar(f, tok_def=None):
|
390 | """Construct a Grammar object from a file."""
|
391 |
|
392 | lexer = tokenize.generate_tokens(f.readline)
|
393 | p = ParserGenerator(lexer)
|
394 | dfas, startsymbol = p.parse()
|
395 |
|
396 | first = {} # map from symbol name to set of tokens
|
397 | for name in sorted(dfas):
|
398 | if name not in first:
|
399 | calcfirst(dfas, first, name)
|
400 | #print name, self.first[name].keys()
|
401 |
|
402 | # TODO: startsymbol support could be removed. The call to p.setup() in
|
403 | # PushTokens() can always specify it explicitly.
|
404 | names = sorted(dfas)
|
405 | names.remove(startsymbol)
|
406 | names.insert(0, startsymbol)
|
407 |
|
408 | gr = grammar.Grammar()
|
409 | for name in names:
|
410 | i = token.NT_OFFSET + len(gr.symbol2number)
|
411 | gr.symbol2number[name] = i
|
412 | gr.number2symbol[i] = name
|
413 |
|
414 | tok_def = tok_def or PythonTokDef()
|
415 | for name in names:
|
416 | dfa = dfas[name]
|
417 | states = []
|
418 | for state in dfa:
|
419 | arcs = []
|
420 | for label, next_ in sorted(state.arcs.items()):
|
421 | arcs.append((make_label(tok_def, gr, label), dfa.index(next_)))
|
422 | if state.isfinal:
|
423 | arcs.append((0, dfa.index(state)))
|
424 | states.append(arcs)
|
425 | gr.states.append(states)
|
426 | fi = make_first(tok_def, first[name], gr)
|
427 | gr.dfas[gr.symbol2number[name]] = (states, fi)
|
428 |
|
429 | gr.start = gr.symbol2number[startsymbol]
|
430 | return gr
|