| 4710c53d |
1 | #\r |
| 2 | # Secret Labs' Regular Expression Engine\r |
| 3 | #\r |
| 4 | # convert template to internal format\r |
| 5 | #\r |
| 6 | # Copyright (c) 1997-2001 by Secret Labs AB. All rights reserved.\r |
| 7 | #\r |
| 8 | # See the sre.py file for information on usage and redistribution.\r |
| 9 | #\r |
| 10 | \r |
| 11 | """Internal support module for sre"""\r |
| 12 | \r |
| 13 | import _sre, sys\r |
| 14 | import sre_parse\r |
| 15 | from sre_constants import *\r |
| 16 | \r |
| 17 | assert _sre.MAGIC == MAGIC, "SRE module mismatch"\r |
| 18 | \r |
| 19 | if _sre.CODESIZE == 2:\r |
| 20 | MAXCODE = 65535\r |
| 21 | else:\r |
| 22 | MAXCODE = 0xFFFFFFFFL\r |
| 23 | \r |
| 24 | def _identityfunction(x):\r |
| 25 | return x\r |
| 26 | \r |
| 27 | _LITERAL_CODES = set([LITERAL, NOT_LITERAL])\r |
| 28 | _REPEATING_CODES = set([REPEAT, MIN_REPEAT, MAX_REPEAT])\r |
| 29 | _SUCCESS_CODES = set([SUCCESS, FAILURE])\r |
| 30 | _ASSERT_CODES = set([ASSERT, ASSERT_NOT])\r |
| 31 | \r |
| 32 | def _compile(code, pattern, flags):\r |
| 33 | # internal: compile a (sub)pattern\r |
| 34 | emit = code.append\r |
| 35 | _len = len\r |
| 36 | LITERAL_CODES = _LITERAL_CODES\r |
| 37 | REPEATING_CODES = _REPEATING_CODES\r |
| 38 | SUCCESS_CODES = _SUCCESS_CODES\r |
| 39 | ASSERT_CODES = _ASSERT_CODES\r |
| 40 | for op, av in pattern:\r |
| 41 | if op in LITERAL_CODES:\r |
| 42 | if flags & SRE_FLAG_IGNORECASE:\r |
| 43 | emit(OPCODES[OP_IGNORE[op]])\r |
| 44 | emit(_sre.getlower(av, flags))\r |
| 45 | else:\r |
| 46 | emit(OPCODES[op])\r |
| 47 | emit(av)\r |
| 48 | elif op is IN:\r |
| 49 | if flags & SRE_FLAG_IGNORECASE:\r |
| 50 | emit(OPCODES[OP_IGNORE[op]])\r |
| 51 | def fixup(literal, flags=flags):\r |
| 52 | return _sre.getlower(literal, flags)\r |
| 53 | else:\r |
| 54 | emit(OPCODES[op])\r |
| 55 | fixup = _identityfunction\r |
| 56 | skip = _len(code); emit(0)\r |
| 57 | _compile_charset(av, flags, code, fixup)\r |
| 58 | code[skip] = _len(code) - skip\r |
| 59 | elif op is ANY:\r |
| 60 | if flags & SRE_FLAG_DOTALL:\r |
| 61 | emit(OPCODES[ANY_ALL])\r |
| 62 | else:\r |
| 63 | emit(OPCODES[ANY])\r |
| 64 | elif op in REPEATING_CODES:\r |
| 65 | if flags & SRE_FLAG_TEMPLATE:\r |
| 66 | raise error, "internal: unsupported template operator"\r |
| 67 | emit(OPCODES[REPEAT])\r |
| 68 | skip = _len(code); emit(0)\r |
| 69 | emit(av[0])\r |
| 70 | emit(av[1])\r |
| 71 | _compile(code, av[2], flags)\r |
| 72 | emit(OPCODES[SUCCESS])\r |
| 73 | code[skip] = _len(code) - skip\r |
| 74 | elif _simple(av) and op is not REPEAT:\r |
| 75 | if op is MAX_REPEAT:\r |
| 76 | emit(OPCODES[REPEAT_ONE])\r |
| 77 | else:\r |
| 78 | emit(OPCODES[MIN_REPEAT_ONE])\r |
| 79 | skip = _len(code); emit(0)\r |
| 80 | emit(av[0])\r |
| 81 | emit(av[1])\r |
| 82 | _compile(code, av[2], flags)\r |
| 83 | emit(OPCODES[SUCCESS])\r |
| 84 | code[skip] = _len(code) - skip\r |
| 85 | else:\r |
| 86 | emit(OPCODES[REPEAT])\r |
| 87 | skip = _len(code); emit(0)\r |
| 88 | emit(av[0])\r |
| 89 | emit(av[1])\r |
| 90 | _compile(code, av[2], flags)\r |
| 91 | code[skip] = _len(code) - skip\r |
| 92 | if op is MAX_REPEAT:\r |
| 93 | emit(OPCODES[MAX_UNTIL])\r |
| 94 | else:\r |
| 95 | emit(OPCODES[MIN_UNTIL])\r |
| 96 | elif op is SUBPATTERN:\r |
| 97 | if av[0]:\r |
| 98 | emit(OPCODES[MARK])\r |
| 99 | emit((av[0]-1)*2)\r |
| 100 | # _compile_info(code, av[1], flags)\r |
| 101 | _compile(code, av[1], flags)\r |
| 102 | if av[0]:\r |
| 103 | emit(OPCODES[MARK])\r |
| 104 | emit((av[0]-1)*2+1)\r |
| 105 | elif op in SUCCESS_CODES:\r |
| 106 | emit(OPCODES[op])\r |
| 107 | elif op in ASSERT_CODES:\r |
| 108 | emit(OPCODES[op])\r |
| 109 | skip = _len(code); emit(0)\r |
| 110 | if av[0] >= 0:\r |
| 111 | emit(0) # look ahead\r |
| 112 | else:\r |
| 113 | lo, hi = av[1].getwidth()\r |
| 114 | if lo != hi:\r |
| 115 | raise error, "look-behind requires fixed-width pattern"\r |
| 116 | emit(lo) # look behind\r |
| 117 | _compile(code, av[1], flags)\r |
| 118 | emit(OPCODES[SUCCESS])\r |
| 119 | code[skip] = _len(code) - skip\r |
| 120 | elif op is CALL:\r |
| 121 | emit(OPCODES[op])\r |
| 122 | skip = _len(code); emit(0)\r |
| 123 | _compile(code, av, flags)\r |
| 124 | emit(OPCODES[SUCCESS])\r |
| 125 | code[skip] = _len(code) - skip\r |
| 126 | elif op is AT:\r |
| 127 | emit(OPCODES[op])\r |
| 128 | if flags & SRE_FLAG_MULTILINE:\r |
| 129 | av = AT_MULTILINE.get(av, av)\r |
| 130 | if flags & SRE_FLAG_LOCALE:\r |
| 131 | av = AT_LOCALE.get(av, av)\r |
| 132 | elif flags & SRE_FLAG_UNICODE:\r |
| 133 | av = AT_UNICODE.get(av, av)\r |
| 134 | emit(ATCODES[av])\r |
| 135 | elif op is BRANCH:\r |
| 136 | emit(OPCODES[op])\r |
| 137 | tail = []\r |
| 138 | tailappend = tail.append\r |
| 139 | for av in av[1]:\r |
| 140 | skip = _len(code); emit(0)\r |
| 141 | # _compile_info(code, av, flags)\r |
| 142 | _compile(code, av, flags)\r |
| 143 | emit(OPCODES[JUMP])\r |
| 144 | tailappend(_len(code)); emit(0)\r |
| 145 | code[skip] = _len(code) - skip\r |
| 146 | emit(0) # end of branch\r |
| 147 | for tail in tail:\r |
| 148 | code[tail] = _len(code) - tail\r |
| 149 | elif op is CATEGORY:\r |
| 150 | emit(OPCODES[op])\r |
| 151 | if flags & SRE_FLAG_LOCALE:\r |
| 152 | av = CH_LOCALE[av]\r |
| 153 | elif flags & SRE_FLAG_UNICODE:\r |
| 154 | av = CH_UNICODE[av]\r |
| 155 | emit(CHCODES[av])\r |
| 156 | elif op is GROUPREF:\r |
| 157 | if flags & SRE_FLAG_IGNORECASE:\r |
| 158 | emit(OPCODES[OP_IGNORE[op]])\r |
| 159 | else:\r |
| 160 | emit(OPCODES[op])\r |
| 161 | emit(av-1)\r |
| 162 | elif op is GROUPREF_EXISTS:\r |
| 163 | emit(OPCODES[op])\r |
| 164 | emit(av[0]-1)\r |
| 165 | skipyes = _len(code); emit(0)\r |
| 166 | _compile(code, av[1], flags)\r |
| 167 | if av[2]:\r |
| 168 | emit(OPCODES[JUMP])\r |
| 169 | skipno = _len(code); emit(0)\r |
| 170 | code[skipyes] = _len(code) - skipyes + 1\r |
| 171 | _compile(code, av[2], flags)\r |
| 172 | code[skipno] = _len(code) - skipno\r |
| 173 | else:\r |
| 174 | code[skipyes] = _len(code) - skipyes + 1\r |
| 175 | else:\r |
| 176 | raise ValueError, ("unsupported operand type", op)\r |
| 177 | \r |
| 178 | def _compile_charset(charset, flags, code, fixup=None):\r |
| 179 | # compile charset subprogram\r |
| 180 | emit = code.append\r |
| 181 | if fixup is None:\r |
| 182 | fixup = _identityfunction\r |
| 183 | for op, av in _optimize_charset(charset, fixup):\r |
| 184 | emit(OPCODES[op])\r |
| 185 | if op is NEGATE:\r |
| 186 | pass\r |
| 187 | elif op is LITERAL:\r |
| 188 | emit(fixup(av))\r |
| 189 | elif op is RANGE:\r |
| 190 | emit(fixup(av[0]))\r |
| 191 | emit(fixup(av[1]))\r |
| 192 | elif op is CHARSET:\r |
| 193 | code.extend(av)\r |
| 194 | elif op is BIGCHARSET:\r |
| 195 | code.extend(av)\r |
| 196 | elif op is CATEGORY:\r |
| 197 | if flags & SRE_FLAG_LOCALE:\r |
| 198 | emit(CHCODES[CH_LOCALE[av]])\r |
| 199 | elif flags & SRE_FLAG_UNICODE:\r |
| 200 | emit(CHCODES[CH_UNICODE[av]])\r |
| 201 | else:\r |
| 202 | emit(CHCODES[av])\r |
| 203 | else:\r |
| 204 | raise error, "internal: unsupported set operator"\r |
| 205 | emit(OPCODES[FAILURE])\r |
| 206 | \r |
| 207 | def _optimize_charset(charset, fixup):\r |
| 208 | # internal: optimize character set\r |
| 209 | out = []\r |
| 210 | outappend = out.append\r |
| 211 | charmap = [0]*256\r |
| 212 | try:\r |
| 213 | for op, av in charset:\r |
| 214 | if op is NEGATE:\r |
| 215 | outappend((op, av))\r |
| 216 | elif op is LITERAL:\r |
| 217 | charmap[fixup(av)] = 1\r |
| 218 | elif op is RANGE:\r |
| 219 | for i in range(fixup(av[0]), fixup(av[1])+1):\r |
| 220 | charmap[i] = 1\r |
| 221 | elif op is CATEGORY:\r |
| 222 | # XXX: could append to charmap tail\r |
| 223 | return charset # cannot compress\r |
| 224 | except IndexError:\r |
| 225 | # character set contains unicode characters\r |
| 226 | return _optimize_unicode(charset, fixup)\r |
| 227 | # compress character map\r |
| 228 | i = p = n = 0\r |
| 229 | runs = []\r |
| 230 | runsappend = runs.append\r |
| 231 | for c in charmap:\r |
| 232 | if c:\r |
| 233 | if n == 0:\r |
| 234 | p = i\r |
| 235 | n = n + 1\r |
| 236 | elif n:\r |
| 237 | runsappend((p, n))\r |
| 238 | n = 0\r |
| 239 | i = i + 1\r |
| 240 | if n:\r |
| 241 | runsappend((p, n))\r |
| 242 | if len(runs) <= 2:\r |
| 243 | # use literal/range\r |
| 244 | for p, n in runs:\r |
| 245 | if n == 1:\r |
| 246 | outappend((LITERAL, p))\r |
| 247 | else:\r |
| 248 | outappend((RANGE, (p, p+n-1)))\r |
| 249 | if len(out) < len(charset):\r |
| 250 | return out\r |
| 251 | else:\r |
| 252 | # use bitmap\r |
| 253 | data = _mk_bitmap(charmap)\r |
| 254 | outappend((CHARSET, data))\r |
| 255 | return out\r |
| 256 | return charset\r |
| 257 | \r |
| 258 | def _mk_bitmap(bits):\r |
| 259 | data = []\r |
| 260 | dataappend = data.append\r |
| 261 | if _sre.CODESIZE == 2:\r |
| 262 | start = (1, 0)\r |
| 263 | else:\r |
| 264 | start = (1L, 0L)\r |
| 265 | m, v = start\r |
| 266 | for c in bits:\r |
| 267 | if c:\r |
| 268 | v = v + m\r |
| 269 | m = m + m\r |
| 270 | if m > MAXCODE:\r |
| 271 | dataappend(v)\r |
| 272 | m, v = start\r |
| 273 | return data\r |
| 274 | \r |
| 275 | # To represent a big charset, first a bitmap of all characters in the\r |
| 276 | # set is constructed. Then, this bitmap is sliced into chunks of 256\r |
| 277 | # characters, duplicate chunks are eliminated, and each chunk is\r |
| 278 | # given a number. In the compiled expression, the charset is\r |
| 279 | # represented by a 16-bit word sequence, consisting of one word for\r |
| 280 | # the number of different chunks, a sequence of 256 bytes (128 words)\r |
| 281 | # of chunk numbers indexed by their original chunk position, and a\r |
| 282 | # sequence of chunks (16 words each).\r |
| 283 | \r |
| 284 | # Compression is normally good: in a typical charset, large ranges of\r |
| 285 | # Unicode will be either completely excluded (e.g. if only cyrillic\r |
| 286 | # letters are to be matched), or completely included (e.g. if large\r |
| 287 | # subranges of Kanji match). These ranges will be represented by\r |
| 288 | # chunks of all one-bits or all zero-bits.\r |
| 289 | \r |
| 290 | # Matching can be also done efficiently: the more significant byte of\r |
| 291 | # the Unicode character is an index into the chunk number, and the\r |
| 292 | # less significant byte is a bit index in the chunk (just like the\r |
| 293 | # CHARSET matching).\r |
| 294 | \r |
| 295 | # In UCS-4 mode, the BIGCHARSET opcode still supports only subsets\r |
| 296 | # of the basic multilingual plane; an efficient representation\r |
| 297 | # for all of UTF-16 has not yet been developed. This means,\r |
| 298 | # in particular, that negated charsets cannot be represented as\r |
| 299 | # bigcharsets.\r |
| 300 | \r |
| 301 | def _optimize_unicode(charset, fixup):\r |
| 302 | try:\r |
| 303 | import array\r |
| 304 | except ImportError:\r |
| 305 | return charset\r |
| 306 | charmap = [0]*65536\r |
| 307 | negate = 0\r |
| 308 | try:\r |
| 309 | for op, av in charset:\r |
| 310 | if op is NEGATE:\r |
| 311 | negate = 1\r |
| 312 | elif op is LITERAL:\r |
| 313 | charmap[fixup(av)] = 1\r |
| 314 | elif op is RANGE:\r |
| 315 | for i in xrange(fixup(av[0]), fixup(av[1])+1):\r |
| 316 | charmap[i] = 1\r |
| 317 | elif op is CATEGORY:\r |
| 318 | # XXX: could expand category\r |
| 319 | return charset # cannot compress\r |
| 320 | except IndexError:\r |
| 321 | # non-BMP characters\r |
| 322 | return charset\r |
| 323 | if negate:\r |
| 324 | if sys.maxunicode != 65535:\r |
| 325 | # XXX: negation does not work with big charsets\r |
| 326 | return charset\r |
| 327 | for i in xrange(65536):\r |
| 328 | charmap[i] = not charmap[i]\r |
| 329 | comps = {}\r |
| 330 | mapping = [0]*256\r |
| 331 | block = 0\r |
| 332 | data = []\r |
| 333 | for i in xrange(256):\r |
| 334 | chunk = tuple(charmap[i*256:(i+1)*256])\r |
| 335 | new = comps.setdefault(chunk, block)\r |
| 336 | mapping[i] = new\r |
| 337 | if new == block:\r |
| 338 | block = block + 1\r |
| 339 | data = data + _mk_bitmap(chunk)\r |
| 340 | header = [block]\r |
| 341 | if _sre.CODESIZE == 2:\r |
| 342 | code = 'H'\r |
| 343 | else:\r |
| 344 | code = 'I'\r |
| 345 | # Convert block indices to byte array of 256 bytes\r |
| 346 | mapping = array.array('b', mapping).tostring()\r |
| 347 | # Convert byte array to word array\r |
| 348 | mapping = array.array(code, mapping)\r |
| 349 | assert mapping.itemsize == _sre.CODESIZE\r |
| 350 | header = header + mapping.tolist()\r |
| 351 | data[0:0] = header\r |
| 352 | return [(BIGCHARSET, data)]\r |
| 353 | \r |
| 354 | def _simple(av):\r |
| 355 | # check if av is a "simple" operator\r |
| 356 | lo, hi = av[2].getwidth()\r |
| 357 | if lo == 0 and hi == MAXREPEAT:\r |
| 358 | raise error, "nothing to repeat"\r |
| 359 | return lo == hi == 1 and av[2][0][0] != SUBPATTERN\r |
| 360 | \r |
| 361 | def _compile_info(code, pattern, flags):\r |
| 362 | # internal: compile an info block. in the current version,\r |
| 363 | # this contains min/max pattern width, and an optional literal\r |
| 364 | # prefix or a character map\r |
| 365 | lo, hi = pattern.getwidth()\r |
| 366 | if lo == 0:\r |
| 367 | return # not worth it\r |
| 368 | # look for a literal prefix\r |
| 369 | prefix = []\r |
| 370 | prefixappend = prefix.append\r |
| 371 | prefix_skip = 0\r |
| 372 | charset = [] # not used\r |
| 373 | charsetappend = charset.append\r |
| 374 | if not (flags & SRE_FLAG_IGNORECASE):\r |
| 375 | # look for literal prefix\r |
| 376 | for op, av in pattern.data:\r |
| 377 | if op is LITERAL:\r |
| 378 | if len(prefix) == prefix_skip:\r |
| 379 | prefix_skip = prefix_skip + 1\r |
| 380 | prefixappend(av)\r |
| 381 | elif op is SUBPATTERN and len(av[1]) == 1:\r |
| 382 | op, av = av[1][0]\r |
| 383 | if op is LITERAL:\r |
| 384 | prefixappend(av)\r |
| 385 | else:\r |
| 386 | break\r |
| 387 | else:\r |
| 388 | break\r |
| 389 | # if no prefix, look for charset prefix\r |
| 390 | if not prefix and pattern.data:\r |
| 391 | op, av = pattern.data[0]\r |
| 392 | if op is SUBPATTERN and av[1]:\r |
| 393 | op, av = av[1][0]\r |
| 394 | if op is LITERAL:\r |
| 395 | charsetappend((op, av))\r |
| 396 | elif op is BRANCH:\r |
| 397 | c = []\r |
| 398 | cappend = c.append\r |
| 399 | for p in av[1]:\r |
| 400 | if not p:\r |
| 401 | break\r |
| 402 | op, av = p[0]\r |
| 403 | if op is LITERAL:\r |
| 404 | cappend((op, av))\r |
| 405 | else:\r |
| 406 | break\r |
| 407 | else:\r |
| 408 | charset = c\r |
| 409 | elif op is BRANCH:\r |
| 410 | c = []\r |
| 411 | cappend = c.append\r |
| 412 | for p in av[1]:\r |
| 413 | if not p:\r |
| 414 | break\r |
| 415 | op, av = p[0]\r |
| 416 | if op is LITERAL:\r |
| 417 | cappend((op, av))\r |
| 418 | else:\r |
| 419 | break\r |
| 420 | else:\r |
| 421 | charset = c\r |
| 422 | elif op is IN:\r |
| 423 | charset = av\r |
| 424 | ## if prefix:\r |
| 425 | ## print "*** PREFIX", prefix, prefix_skip\r |
| 426 | ## if charset:\r |
| 427 | ## print "*** CHARSET", charset\r |
| 428 | # add an info block\r |
| 429 | emit = code.append\r |
| 430 | emit(OPCODES[INFO])\r |
| 431 | skip = len(code); emit(0)\r |
| 432 | # literal flag\r |
| 433 | mask = 0\r |
| 434 | if prefix:\r |
| 435 | mask = SRE_INFO_PREFIX\r |
| 436 | if len(prefix) == prefix_skip == len(pattern.data):\r |
| 437 | mask = mask + SRE_INFO_LITERAL\r |
| 438 | elif charset:\r |
| 439 | mask = mask + SRE_INFO_CHARSET\r |
| 440 | emit(mask)\r |
| 441 | # pattern length\r |
| 442 | if lo < MAXCODE:\r |
| 443 | emit(lo)\r |
| 444 | else:\r |
| 445 | emit(MAXCODE)\r |
| 446 | prefix = prefix[:MAXCODE]\r |
| 447 | if hi < MAXCODE:\r |
| 448 | emit(hi)\r |
| 449 | else:\r |
| 450 | emit(0)\r |
| 451 | # add literal prefix\r |
| 452 | if prefix:\r |
| 453 | emit(len(prefix)) # length\r |
| 454 | emit(prefix_skip) # skip\r |
| 455 | code.extend(prefix)\r |
| 456 | # generate overlap table\r |
| 457 | table = [-1] + ([0]*len(prefix))\r |
| 458 | for i in xrange(len(prefix)):\r |
| 459 | table[i+1] = table[i]+1\r |
| 460 | while table[i+1] > 0 and prefix[i] != prefix[table[i+1]-1]:\r |
| 461 | table[i+1] = table[table[i+1]-1]+1\r |
| 462 | code.extend(table[1:]) # don't store first entry\r |
| 463 | elif charset:\r |
| 464 | _compile_charset(charset, flags, code)\r |
| 465 | code[skip] = len(code) - skip\r |
| 466 | \r |
| 467 | try:\r |
| 468 | unicode\r |
| 469 | except NameError:\r |
| 470 | STRING_TYPES = (type(""),)\r |
| 471 | else:\r |
| 472 | STRING_TYPES = (type(""), type(unicode("")))\r |
| 473 | \r |
| 474 | def isstring(obj):\r |
| 475 | for tp in STRING_TYPES:\r |
| 476 | if isinstance(obj, tp):\r |
| 477 | return 1\r |
| 478 | return 0\r |
| 479 | \r |
| 480 | def _code(p, flags):\r |
| 481 | \r |
| 482 | flags = p.pattern.flags | flags\r |
| 483 | code = []\r |
| 484 | \r |
| 485 | # compile info block\r |
| 486 | _compile_info(code, p, flags)\r |
| 487 | \r |
| 488 | # compile the pattern\r |
| 489 | _compile(code, p.data, flags)\r |
| 490 | \r |
| 491 | code.append(OPCODES[SUCCESS])\r |
| 492 | \r |
| 493 | return code\r |
| 494 | \r |
| 495 | def compile(p, flags=0):\r |
| 496 | # internal: convert pattern list to internal format\r |
| 497 | \r |
| 498 | if isstring(p):\r |
| 499 | pattern = p\r |
| 500 | p = sre_parse.parse(p, flags)\r |
| 501 | else:\r |
| 502 | pattern = None\r |
| 503 | \r |
| 504 | code = _code(p, flags)\r |
| 505 | \r |
| 506 | # print code\r |
| 507 | \r |
| 508 | # XXX: <fl> get rid of this limitation!\r |
| 509 | if p.pattern.groups > 100:\r |
| 510 | raise AssertionError(\r |
| 511 | "sorry, but this version only supports 100 named groups"\r |
| 512 | )\r |
| 513 | \r |
| 514 | # map in either direction\r |
| 515 | groupindex = p.pattern.groupdict\r |
| 516 | indexgroup = [None] * p.pattern.groups\r |
| 517 | for k, i in groupindex.items():\r |
| 518 | indexgroup[i] = k\r |
| 519 | \r |
| 520 | return _sre.compile(\r |
| 521 | pattern, flags | p.pattern.flags, code,\r |
| 522 | p.pattern.groups-1,\r |
| 523 | groupindex, indexgroup\r |
| 524 | )\r |
projects / mirror_edk2.git / blame