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