Olivier Deprez | f4ef2d0 | 2021-04-20 13:36:24 +0200 | [diff] [blame^] | 1 | # |
| 2 | # Secret Labs' Regular Expression Engine |
| 3 | # |
| 4 | # convert re-style regular expression to sre pattern |
| 5 | # |
| 6 | # Copyright (c) 1998-2001 by Secret Labs AB. All rights reserved. |
| 7 | # |
| 8 | # See the sre.py file for information on usage and redistribution. |
| 9 | # |
| 10 | |
| 11 | """Internal support module for sre""" |
| 12 | |
| 13 | # XXX: show string offset and offending character for all errors |
| 14 | |
| 15 | from sre_constants import * |
| 16 | |
| 17 | SPECIAL_CHARS = ".\\[{()*+?^$|" |
| 18 | REPEAT_CHARS = "*+?{" |
| 19 | |
| 20 | DIGITS = frozenset("0123456789") |
| 21 | |
| 22 | OCTDIGITS = frozenset("01234567") |
| 23 | HEXDIGITS = frozenset("0123456789abcdefABCDEF") |
| 24 | ASCIILETTERS = frozenset("abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ") |
| 25 | |
| 26 | WHITESPACE = frozenset(" \t\n\r\v\f") |
| 27 | |
| 28 | _REPEATCODES = frozenset({MIN_REPEAT, MAX_REPEAT}) |
| 29 | _UNITCODES = frozenset({ANY, RANGE, IN, LITERAL, NOT_LITERAL, CATEGORY}) |
| 30 | |
| 31 | ESCAPES = { |
| 32 | r"\a": (LITERAL, ord("\a")), |
| 33 | r"\b": (LITERAL, ord("\b")), |
| 34 | r"\f": (LITERAL, ord("\f")), |
| 35 | r"\n": (LITERAL, ord("\n")), |
| 36 | r"\r": (LITERAL, ord("\r")), |
| 37 | r"\t": (LITERAL, ord("\t")), |
| 38 | r"\v": (LITERAL, ord("\v")), |
| 39 | r"\\": (LITERAL, ord("\\")) |
| 40 | } |
| 41 | |
| 42 | CATEGORIES = { |
| 43 | r"\A": (AT, AT_BEGINNING_STRING), # start of string |
| 44 | r"\b": (AT, AT_BOUNDARY), |
| 45 | r"\B": (AT, AT_NON_BOUNDARY), |
| 46 | r"\d": (IN, [(CATEGORY, CATEGORY_DIGIT)]), |
| 47 | r"\D": (IN, [(CATEGORY, CATEGORY_NOT_DIGIT)]), |
| 48 | r"\s": (IN, [(CATEGORY, CATEGORY_SPACE)]), |
| 49 | r"\S": (IN, [(CATEGORY, CATEGORY_NOT_SPACE)]), |
| 50 | r"\w": (IN, [(CATEGORY, CATEGORY_WORD)]), |
| 51 | r"\W": (IN, [(CATEGORY, CATEGORY_NOT_WORD)]), |
| 52 | r"\Z": (AT, AT_END_STRING), # end of string |
| 53 | } |
| 54 | |
| 55 | FLAGS = { |
| 56 | # standard flags |
| 57 | "i": SRE_FLAG_IGNORECASE, |
| 58 | "L": SRE_FLAG_LOCALE, |
| 59 | "m": SRE_FLAG_MULTILINE, |
| 60 | "s": SRE_FLAG_DOTALL, |
| 61 | "x": SRE_FLAG_VERBOSE, |
| 62 | # extensions |
| 63 | "a": SRE_FLAG_ASCII, |
| 64 | "t": SRE_FLAG_TEMPLATE, |
| 65 | "u": SRE_FLAG_UNICODE, |
| 66 | } |
| 67 | |
| 68 | TYPE_FLAGS = SRE_FLAG_ASCII | SRE_FLAG_LOCALE | SRE_FLAG_UNICODE |
| 69 | GLOBAL_FLAGS = SRE_FLAG_DEBUG | SRE_FLAG_TEMPLATE |
| 70 | |
| 71 | class Verbose(Exception): |
| 72 | pass |
| 73 | |
| 74 | class State: |
| 75 | # keeps track of state for parsing |
| 76 | def __init__(self): |
| 77 | self.flags = 0 |
| 78 | self.groupdict = {} |
| 79 | self.groupwidths = [None] # group 0 |
| 80 | self.lookbehindgroups = None |
| 81 | @property |
| 82 | def groups(self): |
| 83 | return len(self.groupwidths) |
| 84 | def opengroup(self, name=None): |
| 85 | gid = self.groups |
| 86 | self.groupwidths.append(None) |
| 87 | if self.groups > MAXGROUPS: |
| 88 | raise error("too many groups") |
| 89 | if name is not None: |
| 90 | ogid = self.groupdict.get(name, None) |
| 91 | if ogid is not None: |
| 92 | raise error("redefinition of group name %r as group %d; " |
| 93 | "was group %d" % (name, gid, ogid)) |
| 94 | self.groupdict[name] = gid |
| 95 | return gid |
| 96 | def closegroup(self, gid, p): |
| 97 | self.groupwidths[gid] = p.getwidth() |
| 98 | def checkgroup(self, gid): |
| 99 | return gid < self.groups and self.groupwidths[gid] is not None |
| 100 | |
| 101 | def checklookbehindgroup(self, gid, source): |
| 102 | if self.lookbehindgroups is not None: |
| 103 | if not self.checkgroup(gid): |
| 104 | raise source.error('cannot refer to an open group') |
| 105 | if gid >= self.lookbehindgroups: |
| 106 | raise source.error('cannot refer to group defined in the same ' |
| 107 | 'lookbehind subpattern') |
| 108 | |
| 109 | class SubPattern: |
| 110 | # a subpattern, in intermediate form |
| 111 | def __init__(self, state, data=None): |
| 112 | self.state = state |
| 113 | if data is None: |
| 114 | data = [] |
| 115 | self.data = data |
| 116 | self.width = None |
| 117 | |
| 118 | def dump(self, level=0): |
| 119 | nl = True |
| 120 | seqtypes = (tuple, list) |
| 121 | for op, av in self.data: |
| 122 | print(level*" " + str(op), end='') |
| 123 | if op is IN: |
| 124 | # member sublanguage |
| 125 | print() |
| 126 | for op, a in av: |
| 127 | print((level+1)*" " + str(op), a) |
| 128 | elif op is BRANCH: |
| 129 | print() |
| 130 | for i, a in enumerate(av[1]): |
| 131 | if i: |
| 132 | print(level*" " + "OR") |
| 133 | a.dump(level+1) |
| 134 | elif op is GROUPREF_EXISTS: |
| 135 | condgroup, item_yes, item_no = av |
| 136 | print('', condgroup) |
| 137 | item_yes.dump(level+1) |
| 138 | if item_no: |
| 139 | print(level*" " + "ELSE") |
| 140 | item_no.dump(level+1) |
| 141 | elif isinstance(av, seqtypes): |
| 142 | nl = False |
| 143 | for a in av: |
| 144 | if isinstance(a, SubPattern): |
| 145 | if not nl: |
| 146 | print() |
| 147 | a.dump(level+1) |
| 148 | nl = True |
| 149 | else: |
| 150 | if not nl: |
| 151 | print(' ', end='') |
| 152 | print(a, end='') |
| 153 | nl = False |
| 154 | if not nl: |
| 155 | print() |
| 156 | else: |
| 157 | print('', av) |
| 158 | def __repr__(self): |
| 159 | return repr(self.data) |
| 160 | def __len__(self): |
| 161 | return len(self.data) |
| 162 | def __delitem__(self, index): |
| 163 | del self.data[index] |
| 164 | def __getitem__(self, index): |
| 165 | if isinstance(index, slice): |
| 166 | return SubPattern(self.state, self.data[index]) |
| 167 | return self.data[index] |
| 168 | def __setitem__(self, index, code): |
| 169 | self.data[index] = code |
| 170 | def insert(self, index, code): |
| 171 | self.data.insert(index, code) |
| 172 | def append(self, code): |
| 173 | self.data.append(code) |
| 174 | def getwidth(self): |
| 175 | # determine the width (min, max) for this subpattern |
| 176 | if self.width is not None: |
| 177 | return self.width |
| 178 | lo = hi = 0 |
| 179 | for op, av in self.data: |
| 180 | if op is BRANCH: |
| 181 | i = MAXREPEAT - 1 |
| 182 | j = 0 |
| 183 | for av in av[1]: |
| 184 | l, h = av.getwidth() |
| 185 | i = min(i, l) |
| 186 | j = max(j, h) |
| 187 | lo = lo + i |
| 188 | hi = hi + j |
| 189 | elif op is CALL: |
| 190 | i, j = av.getwidth() |
| 191 | lo = lo + i |
| 192 | hi = hi + j |
| 193 | elif op is SUBPATTERN: |
| 194 | i, j = av[-1].getwidth() |
| 195 | lo = lo + i |
| 196 | hi = hi + j |
| 197 | elif op in _REPEATCODES: |
| 198 | i, j = av[2].getwidth() |
| 199 | lo = lo + i * av[0] |
| 200 | hi = hi + j * av[1] |
| 201 | elif op in _UNITCODES: |
| 202 | lo = lo + 1 |
| 203 | hi = hi + 1 |
| 204 | elif op is GROUPREF: |
| 205 | i, j = self.state.groupwidths[av] |
| 206 | lo = lo + i |
| 207 | hi = hi + j |
| 208 | elif op is GROUPREF_EXISTS: |
| 209 | i, j = av[1].getwidth() |
| 210 | if av[2] is not None: |
| 211 | l, h = av[2].getwidth() |
| 212 | i = min(i, l) |
| 213 | j = max(j, h) |
| 214 | else: |
| 215 | i = 0 |
| 216 | lo = lo + i |
| 217 | hi = hi + j |
| 218 | elif op is SUCCESS: |
| 219 | break |
| 220 | self.width = min(lo, MAXREPEAT - 1), min(hi, MAXREPEAT) |
| 221 | return self.width |
| 222 | |
| 223 | class Tokenizer: |
| 224 | def __init__(self, string): |
| 225 | self.istext = isinstance(string, str) |
| 226 | self.string = string |
| 227 | if not self.istext: |
| 228 | string = str(string, 'latin1') |
| 229 | self.decoded_string = string |
| 230 | self.index = 0 |
| 231 | self.next = None |
| 232 | self.__next() |
| 233 | def __next(self): |
| 234 | index = self.index |
| 235 | try: |
| 236 | char = self.decoded_string[index] |
| 237 | except IndexError: |
| 238 | self.next = None |
| 239 | return |
| 240 | if char == "\\": |
| 241 | index += 1 |
| 242 | try: |
| 243 | char += self.decoded_string[index] |
| 244 | except IndexError: |
| 245 | raise error("bad escape (end of pattern)", |
| 246 | self.string, len(self.string) - 1) from None |
| 247 | self.index = index + 1 |
| 248 | self.next = char |
| 249 | def match(self, char): |
| 250 | if char == self.next: |
| 251 | self.__next() |
| 252 | return True |
| 253 | return False |
| 254 | def get(self): |
| 255 | this = self.next |
| 256 | self.__next() |
| 257 | return this |
| 258 | def getwhile(self, n, charset): |
| 259 | result = '' |
| 260 | for _ in range(n): |
| 261 | c = self.next |
| 262 | if c not in charset: |
| 263 | break |
| 264 | result += c |
| 265 | self.__next() |
| 266 | return result |
| 267 | def getuntil(self, terminator, name): |
| 268 | result = '' |
| 269 | while True: |
| 270 | c = self.next |
| 271 | self.__next() |
| 272 | if c is None: |
| 273 | if not result: |
| 274 | raise self.error("missing " + name) |
| 275 | raise self.error("missing %s, unterminated name" % terminator, |
| 276 | len(result)) |
| 277 | if c == terminator: |
| 278 | if not result: |
| 279 | raise self.error("missing " + name, 1) |
| 280 | break |
| 281 | result += c |
| 282 | return result |
| 283 | @property |
| 284 | def pos(self): |
| 285 | return self.index - len(self.next or '') |
| 286 | def tell(self): |
| 287 | return self.index - len(self.next or '') |
| 288 | def seek(self, index): |
| 289 | self.index = index |
| 290 | self.__next() |
| 291 | |
| 292 | def error(self, msg, offset=0): |
| 293 | return error(msg, self.string, self.tell() - offset) |
| 294 | |
| 295 | def _class_escape(source, escape): |
| 296 | # handle escape code inside character class |
| 297 | code = ESCAPES.get(escape) |
| 298 | if code: |
| 299 | return code |
| 300 | code = CATEGORIES.get(escape) |
| 301 | if code and code[0] is IN: |
| 302 | return code |
| 303 | try: |
| 304 | c = escape[1:2] |
| 305 | if c == "x": |
| 306 | # hexadecimal escape (exactly two digits) |
| 307 | escape += source.getwhile(2, HEXDIGITS) |
| 308 | if len(escape) != 4: |
| 309 | raise source.error("incomplete escape %s" % escape, len(escape)) |
| 310 | return LITERAL, int(escape[2:], 16) |
| 311 | elif c == "u" and source.istext: |
| 312 | # unicode escape (exactly four digits) |
| 313 | escape += source.getwhile(4, HEXDIGITS) |
| 314 | if len(escape) != 6: |
| 315 | raise source.error("incomplete escape %s" % escape, len(escape)) |
| 316 | return LITERAL, int(escape[2:], 16) |
| 317 | elif c == "U" and source.istext: |
| 318 | # unicode escape (exactly eight digits) |
| 319 | escape += source.getwhile(8, HEXDIGITS) |
| 320 | if len(escape) != 10: |
| 321 | raise source.error("incomplete escape %s" % escape, len(escape)) |
| 322 | c = int(escape[2:], 16) |
| 323 | chr(c) # raise ValueError for invalid code |
| 324 | return LITERAL, c |
| 325 | elif c == "N" and source.istext: |
| 326 | import unicodedata |
| 327 | # named unicode escape e.g. \N{EM DASH} |
| 328 | if not source.match('{'): |
| 329 | raise source.error("missing {") |
| 330 | charname = source.getuntil('}', 'character name') |
| 331 | try: |
| 332 | c = ord(unicodedata.lookup(charname)) |
| 333 | except KeyError: |
| 334 | raise source.error("undefined character name %r" % charname, |
| 335 | len(charname) + len(r'\N{}')) |
| 336 | return LITERAL, c |
| 337 | elif c in OCTDIGITS: |
| 338 | # octal escape (up to three digits) |
| 339 | escape += source.getwhile(2, OCTDIGITS) |
| 340 | c = int(escape[1:], 8) |
| 341 | if c > 0o377: |
| 342 | raise source.error('octal escape value %s outside of ' |
| 343 | 'range 0-0o377' % escape, len(escape)) |
| 344 | return LITERAL, c |
| 345 | elif c in DIGITS: |
| 346 | raise ValueError |
| 347 | if len(escape) == 2: |
| 348 | if c in ASCIILETTERS: |
| 349 | raise source.error('bad escape %s' % escape, len(escape)) |
| 350 | return LITERAL, ord(escape[1]) |
| 351 | except ValueError: |
| 352 | pass |
| 353 | raise source.error("bad escape %s" % escape, len(escape)) |
| 354 | |
| 355 | def _escape(source, escape, state): |
| 356 | # handle escape code in expression |
| 357 | code = CATEGORIES.get(escape) |
| 358 | if code: |
| 359 | return code |
| 360 | code = ESCAPES.get(escape) |
| 361 | if code: |
| 362 | return code |
| 363 | try: |
| 364 | c = escape[1:2] |
| 365 | if c == "x": |
| 366 | # hexadecimal escape |
| 367 | escape += source.getwhile(2, HEXDIGITS) |
| 368 | if len(escape) != 4: |
| 369 | raise source.error("incomplete escape %s" % escape, len(escape)) |
| 370 | return LITERAL, int(escape[2:], 16) |
| 371 | elif c == "u" and source.istext: |
| 372 | # unicode escape (exactly four digits) |
| 373 | escape += source.getwhile(4, HEXDIGITS) |
| 374 | if len(escape) != 6: |
| 375 | raise source.error("incomplete escape %s" % escape, len(escape)) |
| 376 | return LITERAL, int(escape[2:], 16) |
| 377 | elif c == "U" and source.istext: |
| 378 | # unicode escape (exactly eight digits) |
| 379 | escape += source.getwhile(8, HEXDIGITS) |
| 380 | if len(escape) != 10: |
| 381 | raise source.error("incomplete escape %s" % escape, len(escape)) |
| 382 | c = int(escape[2:], 16) |
| 383 | chr(c) # raise ValueError for invalid code |
| 384 | return LITERAL, c |
| 385 | elif c == "N" and source.istext: |
| 386 | import unicodedata |
| 387 | # named unicode escape e.g. \N{EM DASH} |
| 388 | if not source.match('{'): |
| 389 | raise source.error("missing {") |
| 390 | charname = source.getuntil('}', 'character name') |
| 391 | try: |
| 392 | c = ord(unicodedata.lookup(charname)) |
| 393 | except KeyError: |
| 394 | raise source.error("undefined character name %r" % charname, |
| 395 | len(charname) + len(r'\N{}')) |
| 396 | return LITERAL, c |
| 397 | elif c == "0": |
| 398 | # octal escape |
| 399 | escape += source.getwhile(2, OCTDIGITS) |
| 400 | return LITERAL, int(escape[1:], 8) |
| 401 | elif c in DIGITS: |
| 402 | # octal escape *or* decimal group reference (sigh) |
| 403 | if source.next in DIGITS: |
| 404 | escape += source.get() |
| 405 | if (escape[1] in OCTDIGITS and escape[2] in OCTDIGITS and |
| 406 | source.next in OCTDIGITS): |
| 407 | # got three octal digits; this is an octal escape |
| 408 | escape += source.get() |
| 409 | c = int(escape[1:], 8) |
| 410 | if c > 0o377: |
| 411 | raise source.error('octal escape value %s outside of ' |
| 412 | 'range 0-0o377' % escape, |
| 413 | len(escape)) |
| 414 | return LITERAL, c |
| 415 | # not an octal escape, so this is a group reference |
| 416 | group = int(escape[1:]) |
| 417 | if group < state.groups: |
| 418 | if not state.checkgroup(group): |
| 419 | raise source.error("cannot refer to an open group", |
| 420 | len(escape)) |
| 421 | state.checklookbehindgroup(group, source) |
| 422 | return GROUPREF, group |
| 423 | raise source.error("invalid group reference %d" % group, len(escape) - 1) |
| 424 | if len(escape) == 2: |
| 425 | if c in ASCIILETTERS: |
| 426 | raise source.error("bad escape %s" % escape, len(escape)) |
| 427 | return LITERAL, ord(escape[1]) |
| 428 | except ValueError: |
| 429 | pass |
| 430 | raise source.error("bad escape %s" % escape, len(escape)) |
| 431 | |
| 432 | def _uniq(items): |
| 433 | return list(dict.fromkeys(items)) |
| 434 | |
| 435 | def _parse_sub(source, state, verbose, nested): |
| 436 | # parse an alternation: a|b|c |
| 437 | |
| 438 | items = [] |
| 439 | itemsappend = items.append |
| 440 | sourcematch = source.match |
| 441 | start = source.tell() |
| 442 | while True: |
| 443 | itemsappend(_parse(source, state, verbose, nested + 1, |
| 444 | not nested and not items)) |
| 445 | if not sourcematch("|"): |
| 446 | break |
| 447 | |
| 448 | if len(items) == 1: |
| 449 | return items[0] |
| 450 | |
| 451 | subpattern = SubPattern(state) |
| 452 | |
| 453 | # check if all items share a common prefix |
| 454 | while True: |
| 455 | prefix = None |
| 456 | for item in items: |
| 457 | if not item: |
| 458 | break |
| 459 | if prefix is None: |
| 460 | prefix = item[0] |
| 461 | elif item[0] != prefix: |
| 462 | break |
| 463 | else: |
| 464 | # all subitems start with a common "prefix". |
| 465 | # move it out of the branch |
| 466 | for item in items: |
| 467 | del item[0] |
| 468 | subpattern.append(prefix) |
| 469 | continue # check next one |
| 470 | break |
| 471 | |
| 472 | # check if the branch can be replaced by a character set |
| 473 | set = [] |
| 474 | for item in items: |
| 475 | if len(item) != 1: |
| 476 | break |
| 477 | op, av = item[0] |
| 478 | if op is LITERAL: |
| 479 | set.append((op, av)) |
| 480 | elif op is IN and av[0][0] is not NEGATE: |
| 481 | set.extend(av) |
| 482 | else: |
| 483 | break |
| 484 | else: |
| 485 | # we can store this as a character set instead of a |
| 486 | # branch (the compiler may optimize this even more) |
| 487 | subpattern.append((IN, _uniq(set))) |
| 488 | return subpattern |
| 489 | |
| 490 | subpattern.append((BRANCH, (None, items))) |
| 491 | return subpattern |
| 492 | |
| 493 | def _parse(source, state, verbose, nested, first=False): |
| 494 | # parse a simple pattern |
| 495 | subpattern = SubPattern(state) |
| 496 | |
| 497 | # precompute constants into local variables |
| 498 | subpatternappend = subpattern.append |
| 499 | sourceget = source.get |
| 500 | sourcematch = source.match |
| 501 | _len = len |
| 502 | _ord = ord |
| 503 | |
| 504 | while True: |
| 505 | |
| 506 | this = source.next |
| 507 | if this is None: |
| 508 | break # end of pattern |
| 509 | if this in "|)": |
| 510 | break # end of subpattern |
| 511 | sourceget() |
| 512 | |
| 513 | if verbose: |
| 514 | # skip whitespace and comments |
| 515 | if this in WHITESPACE: |
| 516 | continue |
| 517 | if this == "#": |
| 518 | while True: |
| 519 | this = sourceget() |
| 520 | if this is None or this == "\n": |
| 521 | break |
| 522 | continue |
| 523 | |
| 524 | if this[0] == "\\": |
| 525 | code = _escape(source, this, state) |
| 526 | subpatternappend(code) |
| 527 | |
| 528 | elif this not in SPECIAL_CHARS: |
| 529 | subpatternappend((LITERAL, _ord(this))) |
| 530 | |
| 531 | elif this == "[": |
| 532 | here = source.tell() - 1 |
| 533 | # character set |
| 534 | set = [] |
| 535 | setappend = set.append |
| 536 | ## if sourcematch(":"): |
| 537 | ## pass # handle character classes |
| 538 | if source.next == '[': |
| 539 | import warnings |
| 540 | warnings.warn( |
| 541 | 'Possible nested set at position %d' % source.tell(), |
| 542 | FutureWarning, stacklevel=nested + 6 |
| 543 | ) |
| 544 | negate = sourcematch("^") |
| 545 | # check remaining characters |
| 546 | while True: |
| 547 | this = sourceget() |
| 548 | if this is None: |
| 549 | raise source.error("unterminated character set", |
| 550 | source.tell() - here) |
| 551 | if this == "]" and set: |
| 552 | break |
| 553 | elif this[0] == "\\": |
| 554 | code1 = _class_escape(source, this) |
| 555 | else: |
| 556 | if set and this in '-&~|' and source.next == this: |
| 557 | import warnings |
| 558 | warnings.warn( |
| 559 | 'Possible set %s at position %d' % ( |
| 560 | 'difference' if this == '-' else |
| 561 | 'intersection' if this == '&' else |
| 562 | 'symmetric difference' if this == '~' else |
| 563 | 'union', |
| 564 | source.tell() - 1), |
| 565 | FutureWarning, stacklevel=nested + 6 |
| 566 | ) |
| 567 | code1 = LITERAL, _ord(this) |
| 568 | if sourcematch("-"): |
| 569 | # potential range |
| 570 | that = sourceget() |
| 571 | if that is None: |
| 572 | raise source.error("unterminated character set", |
| 573 | source.tell() - here) |
| 574 | if that == "]": |
| 575 | if code1[0] is IN: |
| 576 | code1 = code1[1][0] |
| 577 | setappend(code1) |
| 578 | setappend((LITERAL, _ord("-"))) |
| 579 | break |
| 580 | if that[0] == "\\": |
| 581 | code2 = _class_escape(source, that) |
| 582 | else: |
| 583 | if that == '-': |
| 584 | import warnings |
| 585 | warnings.warn( |
| 586 | 'Possible set difference at position %d' % ( |
| 587 | source.tell() - 2), |
| 588 | FutureWarning, stacklevel=nested + 6 |
| 589 | ) |
| 590 | code2 = LITERAL, _ord(that) |
| 591 | if code1[0] != LITERAL or code2[0] != LITERAL: |
| 592 | msg = "bad character range %s-%s" % (this, that) |
| 593 | raise source.error(msg, len(this) + 1 + len(that)) |
| 594 | lo = code1[1] |
| 595 | hi = code2[1] |
| 596 | if hi < lo: |
| 597 | msg = "bad character range %s-%s" % (this, that) |
| 598 | raise source.error(msg, len(this) + 1 + len(that)) |
| 599 | setappend((RANGE, (lo, hi))) |
| 600 | else: |
| 601 | if code1[0] is IN: |
| 602 | code1 = code1[1][0] |
| 603 | setappend(code1) |
| 604 | |
| 605 | set = _uniq(set) |
| 606 | # XXX: <fl> should move set optimization to compiler! |
| 607 | if _len(set) == 1 and set[0][0] is LITERAL: |
| 608 | # optimization |
| 609 | if negate: |
| 610 | subpatternappend((NOT_LITERAL, set[0][1])) |
| 611 | else: |
| 612 | subpatternappend(set[0]) |
| 613 | else: |
| 614 | if negate: |
| 615 | set.insert(0, (NEGATE, None)) |
| 616 | # charmap optimization can't be added here because |
| 617 | # global flags still are not known |
| 618 | subpatternappend((IN, set)) |
| 619 | |
| 620 | elif this in REPEAT_CHARS: |
| 621 | # repeat previous item |
| 622 | here = source.tell() |
| 623 | if this == "?": |
| 624 | min, max = 0, 1 |
| 625 | elif this == "*": |
| 626 | min, max = 0, MAXREPEAT |
| 627 | |
| 628 | elif this == "+": |
| 629 | min, max = 1, MAXREPEAT |
| 630 | elif this == "{": |
| 631 | if source.next == "}": |
| 632 | subpatternappend((LITERAL, _ord(this))) |
| 633 | continue |
| 634 | |
| 635 | min, max = 0, MAXREPEAT |
| 636 | lo = hi = "" |
| 637 | while source.next in DIGITS: |
| 638 | lo += sourceget() |
| 639 | if sourcematch(","): |
| 640 | while source.next in DIGITS: |
| 641 | hi += sourceget() |
| 642 | else: |
| 643 | hi = lo |
| 644 | if not sourcematch("}"): |
| 645 | subpatternappend((LITERAL, _ord(this))) |
| 646 | source.seek(here) |
| 647 | continue |
| 648 | |
| 649 | if lo: |
| 650 | min = int(lo) |
| 651 | if min >= MAXREPEAT: |
| 652 | raise OverflowError("the repetition number is too large") |
| 653 | if hi: |
| 654 | max = int(hi) |
| 655 | if max >= MAXREPEAT: |
| 656 | raise OverflowError("the repetition number is too large") |
| 657 | if max < min: |
| 658 | raise source.error("min repeat greater than max repeat", |
| 659 | source.tell() - here) |
| 660 | else: |
| 661 | raise AssertionError("unsupported quantifier %r" % (char,)) |
| 662 | # figure out which item to repeat |
| 663 | if subpattern: |
| 664 | item = subpattern[-1:] |
| 665 | else: |
| 666 | item = None |
| 667 | if not item or item[0][0] is AT: |
| 668 | raise source.error("nothing to repeat", |
| 669 | source.tell() - here + len(this)) |
| 670 | if item[0][0] in _REPEATCODES: |
| 671 | raise source.error("multiple repeat", |
| 672 | source.tell() - here + len(this)) |
| 673 | if item[0][0] is SUBPATTERN: |
| 674 | group, add_flags, del_flags, p = item[0][1] |
| 675 | if group is None and not add_flags and not del_flags: |
| 676 | item = p |
| 677 | if sourcematch("?"): |
| 678 | subpattern[-1] = (MIN_REPEAT, (min, max, item)) |
| 679 | else: |
| 680 | subpattern[-1] = (MAX_REPEAT, (min, max, item)) |
| 681 | |
| 682 | elif this == ".": |
| 683 | subpatternappend((ANY, None)) |
| 684 | |
| 685 | elif this == "(": |
| 686 | start = source.tell() - 1 |
| 687 | group = True |
| 688 | name = None |
| 689 | add_flags = 0 |
| 690 | del_flags = 0 |
| 691 | if sourcematch("?"): |
| 692 | # options |
| 693 | char = sourceget() |
| 694 | if char is None: |
| 695 | raise source.error("unexpected end of pattern") |
| 696 | if char == "P": |
| 697 | # python extensions |
| 698 | if sourcematch("<"): |
| 699 | # named group: skip forward to end of name |
| 700 | name = source.getuntil(">", "group name") |
| 701 | if not name.isidentifier(): |
| 702 | msg = "bad character in group name %r" % name |
| 703 | raise source.error(msg, len(name) + 1) |
| 704 | elif sourcematch("="): |
| 705 | # named backreference |
| 706 | name = source.getuntil(")", "group name") |
| 707 | if not name.isidentifier(): |
| 708 | msg = "bad character in group name %r" % name |
| 709 | raise source.error(msg, len(name) + 1) |
| 710 | gid = state.groupdict.get(name) |
| 711 | if gid is None: |
| 712 | msg = "unknown group name %r" % name |
| 713 | raise source.error(msg, len(name) + 1) |
| 714 | if not state.checkgroup(gid): |
| 715 | raise source.error("cannot refer to an open group", |
| 716 | len(name) + 1) |
| 717 | state.checklookbehindgroup(gid, source) |
| 718 | subpatternappend((GROUPREF, gid)) |
| 719 | continue |
| 720 | |
| 721 | else: |
| 722 | char = sourceget() |
| 723 | if char is None: |
| 724 | raise source.error("unexpected end of pattern") |
| 725 | raise source.error("unknown extension ?P" + char, |
| 726 | len(char) + 2) |
| 727 | elif char == ":": |
| 728 | # non-capturing group |
| 729 | group = None |
| 730 | elif char == "#": |
| 731 | # comment |
| 732 | while True: |
| 733 | if source.next is None: |
| 734 | raise source.error("missing ), unterminated comment", |
| 735 | source.tell() - start) |
| 736 | if sourceget() == ")": |
| 737 | break |
| 738 | continue |
| 739 | |
| 740 | elif char in "=!<": |
| 741 | # lookahead assertions |
| 742 | dir = 1 |
| 743 | if char == "<": |
| 744 | char = sourceget() |
| 745 | if char is None: |
| 746 | raise source.error("unexpected end of pattern") |
| 747 | if char not in "=!": |
| 748 | raise source.error("unknown extension ?<" + char, |
| 749 | len(char) + 2) |
| 750 | dir = -1 # lookbehind |
| 751 | lookbehindgroups = state.lookbehindgroups |
| 752 | if lookbehindgroups is None: |
| 753 | state.lookbehindgroups = state.groups |
| 754 | p = _parse_sub(source, state, verbose, nested + 1) |
| 755 | if dir < 0: |
| 756 | if lookbehindgroups is None: |
| 757 | state.lookbehindgroups = None |
| 758 | if not sourcematch(")"): |
| 759 | raise source.error("missing ), unterminated subpattern", |
| 760 | source.tell() - start) |
| 761 | if char == "=": |
| 762 | subpatternappend((ASSERT, (dir, p))) |
| 763 | else: |
| 764 | subpatternappend((ASSERT_NOT, (dir, p))) |
| 765 | continue |
| 766 | |
| 767 | elif char == "(": |
| 768 | # conditional backreference group |
| 769 | condname = source.getuntil(")", "group name") |
| 770 | if condname.isidentifier(): |
| 771 | condgroup = state.groupdict.get(condname) |
| 772 | if condgroup is None: |
| 773 | msg = "unknown group name %r" % condname |
| 774 | raise source.error(msg, len(condname) + 1) |
| 775 | else: |
| 776 | try: |
| 777 | condgroup = int(condname) |
| 778 | if condgroup < 0: |
| 779 | raise ValueError |
| 780 | except ValueError: |
| 781 | msg = "bad character in group name %r" % condname |
| 782 | raise source.error(msg, len(condname) + 1) from None |
| 783 | if not condgroup: |
| 784 | raise source.error("bad group number", |
| 785 | len(condname) + 1) |
| 786 | if condgroup >= MAXGROUPS: |
| 787 | msg = "invalid group reference %d" % condgroup |
| 788 | raise source.error(msg, len(condname) + 1) |
| 789 | state.checklookbehindgroup(condgroup, source) |
| 790 | item_yes = _parse(source, state, verbose, nested + 1) |
| 791 | if source.match("|"): |
| 792 | item_no = _parse(source, state, verbose, nested + 1) |
| 793 | if source.next == "|": |
| 794 | raise source.error("conditional backref with more than two branches") |
| 795 | else: |
| 796 | item_no = None |
| 797 | if not source.match(")"): |
| 798 | raise source.error("missing ), unterminated subpattern", |
| 799 | source.tell() - start) |
| 800 | subpatternappend((GROUPREF_EXISTS, (condgroup, item_yes, item_no))) |
| 801 | continue |
| 802 | |
| 803 | elif char in FLAGS or char == "-": |
| 804 | # flags |
| 805 | flags = _parse_flags(source, state, char) |
| 806 | if flags is None: # global flags |
| 807 | if not first or subpattern: |
| 808 | import warnings |
| 809 | warnings.warn( |
| 810 | 'Flags not at the start of the expression %r%s' % ( |
| 811 | source.string[:20], # truncate long regexes |
| 812 | ' (truncated)' if len(source.string) > 20 else '', |
| 813 | ), |
| 814 | DeprecationWarning, stacklevel=nested + 6 |
| 815 | ) |
| 816 | if (state.flags & SRE_FLAG_VERBOSE) and not verbose: |
| 817 | raise Verbose |
| 818 | continue |
| 819 | |
| 820 | add_flags, del_flags = flags |
| 821 | group = None |
| 822 | else: |
| 823 | raise source.error("unknown extension ?" + char, |
| 824 | len(char) + 1) |
| 825 | |
| 826 | # parse group contents |
| 827 | if group is not None: |
| 828 | try: |
| 829 | group = state.opengroup(name) |
| 830 | except error as err: |
| 831 | raise source.error(err.msg, len(name) + 1) from None |
| 832 | sub_verbose = ((verbose or (add_flags & SRE_FLAG_VERBOSE)) and |
| 833 | not (del_flags & SRE_FLAG_VERBOSE)) |
| 834 | p = _parse_sub(source, state, sub_verbose, nested + 1) |
| 835 | if not source.match(")"): |
| 836 | raise source.error("missing ), unterminated subpattern", |
| 837 | source.tell() - start) |
| 838 | if group is not None: |
| 839 | state.closegroup(group, p) |
| 840 | subpatternappend((SUBPATTERN, (group, add_flags, del_flags, p))) |
| 841 | |
| 842 | elif this == "^": |
| 843 | subpatternappend((AT, AT_BEGINNING)) |
| 844 | |
| 845 | elif this == "$": |
| 846 | subpatternappend((AT, AT_END)) |
| 847 | |
| 848 | else: |
| 849 | raise AssertionError("unsupported special character %r" % (char,)) |
| 850 | |
| 851 | # unpack non-capturing groups |
| 852 | for i in range(len(subpattern))[::-1]: |
| 853 | op, av = subpattern[i] |
| 854 | if op is SUBPATTERN: |
| 855 | group, add_flags, del_flags, p = av |
| 856 | if group is None and not add_flags and not del_flags: |
| 857 | subpattern[i: i+1] = p |
| 858 | |
| 859 | return subpattern |
| 860 | |
| 861 | def _parse_flags(source, state, char): |
| 862 | sourceget = source.get |
| 863 | add_flags = 0 |
| 864 | del_flags = 0 |
| 865 | if char != "-": |
| 866 | while True: |
| 867 | flag = FLAGS[char] |
| 868 | if source.istext: |
| 869 | if char == 'L': |
| 870 | msg = "bad inline flags: cannot use 'L' flag with a str pattern" |
| 871 | raise source.error(msg) |
| 872 | else: |
| 873 | if char == 'u': |
| 874 | msg = "bad inline flags: cannot use 'u' flag with a bytes pattern" |
| 875 | raise source.error(msg) |
| 876 | add_flags |= flag |
| 877 | if (flag & TYPE_FLAGS) and (add_flags & TYPE_FLAGS) != flag: |
| 878 | msg = "bad inline flags: flags 'a', 'u' and 'L' are incompatible" |
| 879 | raise source.error(msg) |
| 880 | char = sourceget() |
| 881 | if char is None: |
| 882 | raise source.error("missing -, : or )") |
| 883 | if char in ")-:": |
| 884 | break |
| 885 | if char not in FLAGS: |
| 886 | msg = "unknown flag" if char.isalpha() else "missing -, : or )" |
| 887 | raise source.error(msg, len(char)) |
| 888 | if char == ")": |
| 889 | state.flags |= add_flags |
| 890 | return None |
| 891 | if add_flags & GLOBAL_FLAGS: |
| 892 | raise source.error("bad inline flags: cannot turn on global flag", 1) |
| 893 | if char == "-": |
| 894 | char = sourceget() |
| 895 | if char is None: |
| 896 | raise source.error("missing flag") |
| 897 | if char not in FLAGS: |
| 898 | msg = "unknown flag" if char.isalpha() else "missing flag" |
| 899 | raise source.error(msg, len(char)) |
| 900 | while True: |
| 901 | flag = FLAGS[char] |
| 902 | if flag & TYPE_FLAGS: |
| 903 | msg = "bad inline flags: cannot turn off flags 'a', 'u' and 'L'" |
| 904 | raise source.error(msg) |
| 905 | del_flags |= flag |
| 906 | char = sourceget() |
| 907 | if char is None: |
| 908 | raise source.error("missing :") |
| 909 | if char == ":": |
| 910 | break |
| 911 | if char not in FLAGS: |
| 912 | msg = "unknown flag" if char.isalpha() else "missing :" |
| 913 | raise source.error(msg, len(char)) |
| 914 | assert char == ":" |
| 915 | if del_flags & GLOBAL_FLAGS: |
| 916 | raise source.error("bad inline flags: cannot turn off global flag", 1) |
| 917 | if add_flags & del_flags: |
| 918 | raise source.error("bad inline flags: flag turned on and off", 1) |
| 919 | return add_flags, del_flags |
| 920 | |
| 921 | def fix_flags(src, flags): |
| 922 | # Check and fix flags according to the type of pattern (str or bytes) |
| 923 | if isinstance(src, str): |
| 924 | if flags & SRE_FLAG_LOCALE: |
| 925 | raise ValueError("cannot use LOCALE flag with a str pattern") |
| 926 | if not flags & SRE_FLAG_ASCII: |
| 927 | flags |= SRE_FLAG_UNICODE |
| 928 | elif flags & SRE_FLAG_UNICODE: |
| 929 | raise ValueError("ASCII and UNICODE flags are incompatible") |
| 930 | else: |
| 931 | if flags & SRE_FLAG_UNICODE: |
| 932 | raise ValueError("cannot use UNICODE flag with a bytes pattern") |
| 933 | if flags & SRE_FLAG_LOCALE and flags & SRE_FLAG_ASCII: |
| 934 | raise ValueError("ASCII and LOCALE flags are incompatible") |
| 935 | return flags |
| 936 | |
| 937 | def parse(str, flags=0, state=None): |
| 938 | # parse 're' pattern into list of (opcode, argument) tuples |
| 939 | |
| 940 | source = Tokenizer(str) |
| 941 | |
| 942 | if state is None: |
| 943 | state = State() |
| 944 | state.flags = flags |
| 945 | state.str = str |
| 946 | |
| 947 | try: |
| 948 | p = _parse_sub(source, state, flags & SRE_FLAG_VERBOSE, 0) |
| 949 | except Verbose: |
| 950 | # the VERBOSE flag was switched on inside the pattern. to be |
| 951 | # on the safe side, we'll parse the whole thing again... |
| 952 | state = State() |
| 953 | state.flags = flags | SRE_FLAG_VERBOSE |
| 954 | state.str = str |
| 955 | source.seek(0) |
| 956 | p = _parse_sub(source, state, True, 0) |
| 957 | |
| 958 | p.state.flags = fix_flags(str, p.state.flags) |
| 959 | |
| 960 | if source.next is not None: |
| 961 | assert source.next == ")" |
| 962 | raise source.error("unbalanced parenthesis") |
| 963 | |
| 964 | if flags & SRE_FLAG_DEBUG: |
| 965 | p.dump() |
| 966 | |
| 967 | return p |
| 968 | |
| 969 | def parse_template(source, state): |
| 970 | # parse 're' replacement string into list of literals and |
| 971 | # group references |
| 972 | s = Tokenizer(source) |
| 973 | sget = s.get |
| 974 | groups = [] |
| 975 | literals = [] |
| 976 | literal = [] |
| 977 | lappend = literal.append |
| 978 | def addgroup(index, pos): |
| 979 | if index > state.groups: |
| 980 | raise s.error("invalid group reference %d" % index, pos) |
| 981 | if literal: |
| 982 | literals.append(''.join(literal)) |
| 983 | del literal[:] |
| 984 | groups.append((len(literals), index)) |
| 985 | literals.append(None) |
| 986 | groupindex = state.groupindex |
| 987 | while True: |
| 988 | this = sget() |
| 989 | if this is None: |
| 990 | break # end of replacement string |
| 991 | if this[0] == "\\": |
| 992 | # group |
| 993 | c = this[1] |
| 994 | if c == "g": |
| 995 | name = "" |
| 996 | if not s.match("<"): |
| 997 | raise s.error("missing <") |
| 998 | name = s.getuntil(">", "group name") |
| 999 | if name.isidentifier(): |
| 1000 | try: |
| 1001 | index = groupindex[name] |
| 1002 | except KeyError: |
| 1003 | raise IndexError("unknown group name %r" % name) |
| 1004 | else: |
| 1005 | try: |
| 1006 | index = int(name) |
| 1007 | if index < 0: |
| 1008 | raise ValueError |
| 1009 | except ValueError: |
| 1010 | raise s.error("bad character in group name %r" % name, |
| 1011 | len(name) + 1) from None |
| 1012 | if index >= MAXGROUPS: |
| 1013 | raise s.error("invalid group reference %d" % index, |
| 1014 | len(name) + 1) |
| 1015 | addgroup(index, len(name) + 1) |
| 1016 | elif c == "0": |
| 1017 | if s.next in OCTDIGITS: |
| 1018 | this += sget() |
| 1019 | if s.next in OCTDIGITS: |
| 1020 | this += sget() |
| 1021 | lappend(chr(int(this[1:], 8) & 0xff)) |
| 1022 | elif c in DIGITS: |
| 1023 | isoctal = False |
| 1024 | if s.next in DIGITS: |
| 1025 | this += sget() |
| 1026 | if (c in OCTDIGITS and this[2] in OCTDIGITS and |
| 1027 | s.next in OCTDIGITS): |
| 1028 | this += sget() |
| 1029 | isoctal = True |
| 1030 | c = int(this[1:], 8) |
| 1031 | if c > 0o377: |
| 1032 | raise s.error('octal escape value %s outside of ' |
| 1033 | 'range 0-0o377' % this, len(this)) |
| 1034 | lappend(chr(c)) |
| 1035 | if not isoctal: |
| 1036 | addgroup(int(this[1:]), len(this) - 1) |
| 1037 | else: |
| 1038 | try: |
| 1039 | this = chr(ESCAPES[this][1]) |
| 1040 | except KeyError: |
| 1041 | if c in ASCIILETTERS: |
| 1042 | raise s.error('bad escape %s' % this, len(this)) |
| 1043 | lappend(this) |
| 1044 | else: |
| 1045 | lappend(this) |
| 1046 | if literal: |
| 1047 | literals.append(''.join(literal)) |
| 1048 | if not isinstance(source, str): |
| 1049 | # The tokenizer implicitly decodes bytes objects as latin-1, we must |
| 1050 | # therefore re-encode the final representation. |
| 1051 | literals = [None if s is None else s.encode('latin-1') for s in literals] |
| 1052 | return groups, literals |
| 1053 | |
| 1054 | def expand_template(template, match): |
| 1055 | g = match.group |
| 1056 | empty = match.string[:0] |
| 1057 | groups, literals = template |
| 1058 | literals = literals[:] |
| 1059 | try: |
| 1060 | for index, group in groups: |
| 1061 | literals[index] = g(group) or empty |
| 1062 | except IndexError: |
| 1063 | raise error("invalid group reference %d" % index) |
| 1064 | return empty.join(literals) |