| 1 |
# |
| 2 |
# Secret Labs' Regular Expression Engine |
| 3 |
# |
| 4 |
# convert template to internal format |
| 5 |
# |
| 6 |
# Copyright (c) 1997-2001 by Secret Labs AB. All rights reserved. |
| 7 |
# |
| 8 |
# See the __init__.py file for information on usage and redistribution. |
| 9 |
# |
| 10 |
|
| 11 |
"""Internal support module for sre""" |
| 12 |
|
| 13 |
import _sre |
| 14 |
from . import _parser |
| 15 |
from ._constants import * |
| 16 |
from ._casefix import _EXTRA_CASES |
| 17 |
|
| 18 |
assert _sre.MAGIC == MAGIC, "SRE module mismatch" |
| 19 |
|
| 20 |
_LITERAL_CODES = {LITERAL, NOT_LITERAL} |
| 21 |
_SUCCESS_CODES = {SUCCESS, FAILURE} |
| 22 |
_ASSERT_CODES = {ASSERT, ASSERT_NOT} |
| 23 |
_UNIT_CODES = _LITERAL_CODES | {ANY, IN} |
| 24 |
|
| 25 |
_REPEATING_CODES = { |
| 26 |
MIN_REPEAT: (REPEAT, MIN_UNTIL, MIN_REPEAT_ONE), |
| 27 |
MAX_REPEAT: (REPEAT, MAX_UNTIL, REPEAT_ONE), |
| 28 |
POSSESSIVE_REPEAT: (POSSESSIVE_REPEAT, SUCCESS, POSSESSIVE_REPEAT_ONE), |
| 29 |
} |
| 30 |
|
| 31 |
_CHARSET_ALL = [(NEGATE, None)] |
| 32 |
|
| 33 |
def _combine_flags(flags, add_flags, del_flags, |
| 34 |
TYPE_FLAGS=_parser.TYPE_FLAGS): |
| 35 |
if add_flags & TYPE_FLAGS: |
| 36 |
flags &= ~TYPE_FLAGS |
| 37 |
return (flags | add_flags) & ~del_flags |
| 38 |
|
| 39 |
def _compile(code, pattern, flags): |
| 40 |
# internal: compile a (sub)pattern |
| 41 |
emit = code.append |
| 42 |
_len = len |
| 43 |
LITERAL_CODES = _LITERAL_CODES |
| 44 |
REPEATING_CODES = _REPEATING_CODES |
| 45 |
SUCCESS_CODES = _SUCCESS_CODES |
| 46 |
ASSERT_CODES = _ASSERT_CODES |
| 47 |
iscased = None |
| 48 |
tolower = None |
| 49 |
fixes = None |
| 50 |
if flags & SRE_FLAG_IGNORECASE and not flags & SRE_FLAG_LOCALE: |
| 51 |
if flags & SRE_FLAG_UNICODE: |
| 52 |
iscased = _sre.unicode_iscased |
| 53 |
tolower = _sre.unicode_tolower |
| 54 |
fixes = _EXTRA_CASES |
| 55 |
else: |
| 56 |
iscased = _sre.ascii_iscased |
| 57 |
tolower = _sre.ascii_tolower |
| 58 |
for op, av in pattern: |
| 59 |
if op in LITERAL_CODES: |
| 60 |
if not flags & SRE_FLAG_IGNORECASE: |
| 61 |
emit(op) |
| 62 |
emit(av) |
| 63 |
elif flags & SRE_FLAG_LOCALE: |
| 64 |
emit(OP_LOCALE_IGNORE[op]) |
| 65 |
emit(av) |
| 66 |
elif not iscased(av): |
| 67 |
emit(op) |
| 68 |
emit(av) |
| 69 |
else: |
| 70 |
lo = tolower(av) |
| 71 |
if not fixes: # ascii |
| 72 |
emit(OP_IGNORE[op]) |
| 73 |
emit(lo) |
| 74 |
elif lo not in fixes: |
| 75 |
emit(OP_UNICODE_IGNORE[op]) |
| 76 |
emit(lo) |
| 77 |
else: |
| 78 |
emit(IN_UNI_IGNORE) |
| 79 |
skip = _len(code); emit(0) |
| 80 |
if op is NOT_LITERAL: |
| 81 |
emit(NEGATE) |
| 82 |
for k in (lo,) + fixes[lo]: |
| 83 |
emit(LITERAL) |
| 84 |
emit(k) |
| 85 |
emit(FAILURE) |
| 86 |
code[skip] = _len(code) - skip |
| 87 |
elif op is IN: |
| 88 |
charset, hascased = _optimize_charset(av, iscased, tolower, fixes) |
| 89 |
if not charset: |
| 90 |
emit(FAILURE) |
| 91 |
elif charset == _CHARSET_ALL: |
| 92 |
emit(ANY_ALL) |
| 93 |
else: |
| 94 |
if flags & SRE_FLAG_IGNORECASE and flags & SRE_FLAG_LOCALE: |
| 95 |
emit(IN_LOC_IGNORE) |
| 96 |
elif not hascased: |
| 97 |
emit(IN) |
| 98 |
elif not fixes: # ascii |
| 99 |
emit(IN_IGNORE) |
| 100 |
else: |
| 101 |
emit(IN_UNI_IGNORE) |
| 102 |
skip = _len(code); emit(0) |
| 103 |
_compile_charset(charset, flags, code) |
| 104 |
code[skip] = _len(code) - skip |
| 105 |
elif op is ANY: |
| 106 |
if flags & SRE_FLAG_DOTALL: |
| 107 |
emit(ANY_ALL) |
| 108 |
else: |
| 109 |
emit(ANY) |
| 110 |
elif op in REPEATING_CODES: |
| 111 |
if _simple(av[2]): |
| 112 |
emit(REPEATING_CODES[op][2]) |
| 113 |
skip = _len(code); emit(0) |
| 114 |
emit(av[0]) |
| 115 |
emit(av[1]) |
| 116 |
_compile(code, av[2], flags) |
| 117 |
emit(SUCCESS) |
| 118 |
code[skip] = _len(code) - skip |
| 119 |
else: |
| 120 |
emit(REPEATING_CODES[op][0]) |
| 121 |
skip = _len(code); emit(0) |
| 122 |
emit(av[0]) |
| 123 |
emit(av[1]) |
| 124 |
_compile(code, av[2], flags) |
| 125 |
code[skip] = _len(code) - skip |
| 126 |
emit(REPEATING_CODES[op][1]) |
| 127 |
elif op is SUBPATTERN: |
| 128 |
group, add_flags, del_flags, p = av |
| 129 |
if group: |
| 130 |
emit(MARK) |
| 131 |
emit((group-1)*2) |
| 132 |
# _compile_info(code, p, _combine_flags(flags, add_flags, del_flags)) |
| 133 |
_compile(code, p, _combine_flags(flags, add_flags, del_flags)) |
| 134 |
if group: |
| 135 |
emit(MARK) |
| 136 |
emit((group-1)*2+1) |
| 137 |
elif op is ATOMIC_GROUP: |
| 138 |
# Atomic Groups are handled by starting with an Atomic |
| 139 |
# Group op code, then putting in the atomic group pattern |
| 140 |
# and finally a success op code to tell any repeat |
| 141 |
# operations within the Atomic Group to stop eating and |
| 142 |
# pop their stack if they reach it |
| 143 |
emit(ATOMIC_GROUP) |
| 144 |
skip = _len(code); emit(0) |
| 145 |
_compile(code, av, flags) |
| 146 |
emit(SUCCESS) |
| 147 |
code[skip] = _len(code) - skip |
| 148 |
elif op in SUCCESS_CODES: |
| 149 |
emit(op) |
| 150 |
elif op in ASSERT_CODES: |
| 151 |
emit(op) |
| 152 |
skip = _len(code); emit(0) |
| 153 |
if av[0] >= 0: |
| 154 |
emit(0) # look ahead |
| 155 |
else: |
| 156 |
lo, hi = av[1].getwidth() |
| 157 |
if lo > MAXCODE: |
| 158 |
raise error("looks too much behind") |
| 159 |
if lo != hi: |
| 160 |
raise PatternError("look-behind requires fixed-width pattern") |
| 161 |
emit(lo) # look behind |
| 162 |
_compile(code, av[1], flags) |
| 163 |
emit(SUCCESS) |
| 164 |
code[skip] = _len(code) - skip |
| 165 |
elif op is AT: |
| 166 |
emit(op) |
| 167 |
if flags & SRE_FLAG_MULTILINE: |
| 168 |
av = AT_MULTILINE.get(av, av) |
| 169 |
if flags & SRE_FLAG_LOCALE: |
| 170 |
av = AT_LOCALE.get(av, av) |
| 171 |
elif flags & SRE_FLAG_UNICODE: |
| 172 |
av = AT_UNICODE.get(av, av) |
| 173 |
emit(av) |
| 174 |
elif op is BRANCH: |
| 175 |
emit(op) |
| 176 |
tail = [] |
| 177 |
tailappend = tail.append |
| 178 |
for av in av[1]: |
| 179 |
skip = _len(code); emit(0) |
| 180 |
# _compile_info(code, av, flags) |
| 181 |
_compile(code, av, flags) |
| 182 |
emit(JUMP) |
| 183 |
tailappend(_len(code)); emit(0) |
| 184 |
code[skip] = _len(code) - skip |
| 185 |
emit(FAILURE) # end of branch |
| 186 |
for tail in tail: |
| 187 |
code[tail] = _len(code) - tail |
| 188 |
elif op is CATEGORY: |
| 189 |
emit(op) |
| 190 |
if flags & SRE_FLAG_LOCALE: |
| 191 |
av = CH_LOCALE[av] |
| 192 |
elif flags & SRE_FLAG_UNICODE: |
| 193 |
av = CH_UNICODE[av] |
| 194 |
emit(av) |
| 195 |
elif op is GROUPREF: |
| 196 |
if not flags & SRE_FLAG_IGNORECASE: |
| 197 |
emit(op) |
| 198 |
elif flags & SRE_FLAG_LOCALE: |
| 199 |
emit(GROUPREF_LOC_IGNORE) |
| 200 |
elif not fixes: # ascii |
| 201 |
emit(GROUPREF_IGNORE) |
| 202 |
else: |
| 203 |
emit(GROUPREF_UNI_IGNORE) |
| 204 |
emit(av-1) |
| 205 |
elif op is GROUPREF_EXISTS: |
| 206 |
emit(op) |
| 207 |
emit(av[0]-1) |
| 208 |
skipyes = _len(code); emit(0) |
| 209 |
_compile(code, av[1], flags) |
| 210 |
if av[2]: |
| 211 |
emit(JUMP) |
| 212 |
skipno = _len(code); emit(0) |
| 213 |
code[skipyes] = _len(code) - skipyes + 1 |
| 214 |
_compile(code, av[2], flags) |
| 215 |
code[skipno] = _len(code) - skipno |
| 216 |
else: |
| 217 |
code[skipyes] = _len(code) - skipyes + 1 |
| 218 |
else: |
| 219 |
raise PatternError(f"internal: unsupported operand type {op!r}") |
| 220 |
|
| 221 |
def _compile_charset(charset, flags, code): |
| 222 |
# compile charset subprogram |
| 223 |
emit = code.append |
| 224 |
for op, av in charset: |
| 225 |
emit(op) |
| 226 |
if op is NEGATE: |
| 227 |
pass |
| 228 |
elif op is LITERAL: |
| 229 |
emit(av) |
| 230 |
elif op is RANGE or op is RANGE_UNI_IGNORE: |
| 231 |
emit(av[0]) |
| 232 |
emit(av[1]) |
| 233 |
elif op is CHARSET: |
| 234 |
code.extend(av) |
| 235 |
elif op is BIGCHARSET: |
| 236 |
code.extend(av) |
| 237 |
elif op is CATEGORY: |
| 238 |
if flags & SRE_FLAG_LOCALE: |
| 239 |
emit(CH_LOCALE[av]) |
| 240 |
elif flags & SRE_FLAG_UNICODE: |
| 241 |
emit(CH_UNICODE[av]) |
| 242 |
else: |
| 243 |
emit(av) |
| 244 |
else: |
| 245 |
raise PatternError(f"internal: unsupported set operator {op!r}") |
| 246 |
emit(FAILURE) |
| 247 |
|
| 248 |
def _optimize_charset(charset, iscased=None, fixup=None, fixes=None): |
| 249 |
# internal: optimize character set |
| 250 |
out = [] |
| 251 |
tail = [] |
| 252 |
charmap = bytearray(256) |
| 253 |
hascased = False |
| 254 |
for op, av in charset: |
| 255 |
while True: |
| 256 |
try: |
| 257 |
if op is LITERAL: |
| 258 |
if fixup: # IGNORECASE and not LOCALE |
| 259 |
av = fixup(av) |
| 260 |
charmap[av] = 1 |
| 261 |
if fixes and av in fixes: |
| 262 |
for k in fixes[av]: |
| 263 |
charmap[k] = 1 |
| 264 |
if not hascased and iscased(av): |
| 265 |
hascased = True |
| 266 |
else: |
| 267 |
charmap[av] = 1 |
| 268 |
elif op is RANGE: |
| 269 |
r = range(av[0], av[1]+1) |
| 270 |
if fixup: # IGNORECASE and not LOCALE |
| 271 |
if fixes: |
| 272 |
for i in map(fixup, r): |
| 273 |
charmap[i] = 1 |
| 274 |
if i in fixes: |
| 275 |
for k in fixes[i]: |
| 276 |
charmap[k] = 1 |
| 277 |
else: |
| 278 |
for i in map(fixup, r): |
| 279 |
charmap[i] = 1 |
| 280 |
if not hascased: |
| 281 |
hascased = any(map(iscased, r)) |
| 282 |
else: |
| 283 |
for i in r: |
| 284 |
charmap[i] = 1 |
| 285 |
elif op is NEGATE: |
| 286 |
out.append((op, av)) |
| 287 |
elif op is CATEGORY and tail and (CATEGORY, CH_NEGATE[av]) in tail: |
| 288 |
# Optimize [\s\S] etc. |
| 289 |
out = [] if out else _CHARSET_ALL |
| 290 |
return out, False |
| 291 |
else: |
| 292 |
tail.append((op, av)) |
| 293 |
except IndexError: |
| 294 |
if len(charmap) == 256: |
| 295 |
# character set contains non-UCS1 character codes |
| 296 |
charmap += b'\0' * 0xff00 |
| 297 |
continue |
| 298 |
# Character set contains non-BMP character codes. |
| 299 |
# For range, all BMP characters in the range are already |
| 300 |
# proceeded. |
| 301 |
if fixup: # IGNORECASE and not LOCALE |
| 302 |
# For now, IN_UNI_IGNORE+LITERAL and |
| 303 |
# IN_UNI_IGNORE+RANGE_UNI_IGNORE work for all non-BMP |
| 304 |
# characters, because two characters (at least one of |
| 305 |
# which is not in the BMP) match case-insensitively |
| 306 |
# if and only if: |
| 307 |
# 1) c1.lower() == c2.lower() |
| 308 |
# 2) c1.lower() == c2 or c1.lower().upper() == c2 |
| 309 |
# Also, both c.lower() and c.lower().upper() are single |
| 310 |
# characters for every non-BMP character. |
| 311 |
if op is RANGE: |
| 312 |
if fixes: # not ASCII |
| 313 |
op = RANGE_UNI_IGNORE |
| 314 |
hascased = True |
| 315 |
else: |
| 316 |
assert op is LITERAL |
| 317 |
if not hascased and iscased(av): |
| 318 |
hascased = True |
| 319 |
tail.append((op, av)) |
| 320 |
break |
| 321 |
|
| 322 |
# compress character map |
| 323 |
runs = [] |
| 324 |
q = 0 |
| 325 |
while True: |
| 326 |
p = charmap.find(1, q) |
| 327 |
if p < 0: |
| 328 |
break |
| 329 |
if len(runs) >= 2: |
| 330 |
runs = None |
| 331 |
break |
| 332 |
q = charmap.find(0, p) |
| 333 |
if q < 0: |
| 334 |
runs.append((p, len(charmap))) |
| 335 |
break |
| 336 |
runs.append((p, q)) |
| 337 |
if runs is not None: |
| 338 |
# use literal/range |
| 339 |
for p, q in runs: |
| 340 |
if q - p == 1: |
| 341 |
out.append((LITERAL, p)) |
| 342 |
else: |
| 343 |
out.append((RANGE, (p, q - 1))) |
| 344 |
out += tail |
| 345 |
# if the case was changed or new representation is more compact |
| 346 |
if hascased or len(out) < len(charset): |
| 347 |
return out, hascased |
| 348 |
# else original character set is good enough |
| 349 |
return charset, hascased |
| 350 |
|
| 351 |
# use bitmap |
| 352 |
if len(charmap) == 256: |
| 353 |
data = _mk_bitmap(charmap) |
| 354 |
out.append((CHARSET, data)) |
| 355 |
out += tail |
| 356 |
return out, hascased |
| 357 |
|
| 358 |
# To represent a big charset, first a bitmap of all characters in the |
| 359 |
# set is constructed. Then, this bitmap is sliced into chunks of 256 |
| 360 |
# characters, duplicate chunks are eliminated, and each chunk is |
| 361 |
# given a number. In the compiled expression, the charset is |
| 362 |
# represented by a 32-bit word sequence, consisting of one word for |
| 363 |
# the number of different chunks, a sequence of 256 bytes (64 words) |
| 364 |
# of chunk numbers indexed by their original chunk position, and a |
| 365 |
# sequence of 256-bit chunks (8 words each). |
| 366 |
|
| 367 |
# Compression is normally good: in a typical charset, large ranges of |
| 368 |
# Unicode will be either completely excluded (e.g. if only cyrillic |
| 369 |
# letters are to be matched), or completely included (e.g. if large |
| 370 |
# subranges of Kanji match). These ranges will be represented by |
| 371 |
# chunks of all one-bits or all zero-bits. |
| 372 |
|
| 373 |
# Matching can be also done efficiently: the more significant byte of |
| 374 |
# the Unicode character is an index into the chunk number, and the |
| 375 |
# less significant byte is a bit index in the chunk (just like the |
| 376 |
# CHARSET matching). |
| 377 |
|
| 378 |
charmap = bytes(charmap) # should be hashable |
| 379 |
comps = {} |
| 380 |
mapping = bytearray(256) |
| 381 |
block = 0 |
| 382 |
data = bytearray() |
| 383 |
for i in range(0, 65536, 256): |
| 384 |
chunk = charmap[i: i + 256] |
| 385 |
if chunk in comps: |
| 386 |
mapping[i // 256] = comps[chunk] |
| 387 |
else: |
| 388 |
mapping[i // 256] = comps[chunk] = block |
| 389 |
block += 1 |
| 390 |
data += chunk |
| 391 |
data = _mk_bitmap(data) |
| 392 |
data[0:0] = [block] + _bytes_to_codes(mapping) |
| 393 |
out.append((BIGCHARSET, data)) |
| 394 |
out += tail |
| 395 |
return out, hascased |
| 396 |
|
| 397 |
_CODEBITS = _sre.CODESIZE * 8 |
| 398 |
MAXCODE = (1 << _CODEBITS) - 1 |
| 399 |
_BITS_TRANS = b'0' + b'1' * 255 |
| 400 |
def _mk_bitmap(bits, _CODEBITS=_CODEBITS, _int=int): |
| 401 |
s = bits.translate(_BITS_TRANS)[::-1] |
| 402 |
return [_int(s[i - _CODEBITS: i], 2) |
| 403 |
for i in range(len(s), 0, -_CODEBITS)] |
| 404 |
|
| 405 |
def _bytes_to_codes(b): |
| 406 |
# Convert block indices to word array |
| 407 |
a = memoryview(b).cast('I') |
| 408 |
assert a.itemsize == _sre.CODESIZE |
| 409 |
assert len(a) * a.itemsize == len(b) |
| 410 |
return a.tolist() |
| 411 |
|
| 412 |
def _simple(p): |
| 413 |
# check if this subpattern is a "simple" operator |
| 414 |
if len(p) != 1: |
| 415 |
return False |
| 416 |
op, av = p[0] |
| 417 |
if op is SUBPATTERN: |
| 418 |
return av[0] is None and _simple(av[-1]) |
| 419 |
return op in _UNIT_CODES |
| 420 |
|
| 421 |
def _generate_overlap_table(prefix): |
| 422 |
""" |
| 423 |
Generate an overlap table for the following prefix. |
| 424 |
An overlap table is a table of the same size as the prefix which |
| 425 |
informs about the potential self-overlap for each index in the prefix: |
| 426 |
- if overlap[i] == 0, prefix[i:] can't overlap prefix[0:...] |
| 427 |
- if overlap[i] == k with 0 < k <= i, prefix[i-k+1:i+1] overlaps with |
| 428 |
prefix[0:k] |
| 429 |
""" |
| 430 |
table = [0] * len(prefix) |
| 431 |
for i in range(1, len(prefix)): |
| 432 |
idx = table[i - 1] |
| 433 |
while prefix[i] != prefix[idx]: |
| 434 |
if idx == 0: |
| 435 |
table[i] = 0 |
| 436 |
break |
| 437 |
idx = table[idx - 1] |
| 438 |
else: |
| 439 |
table[i] = idx + 1 |
| 440 |
return table |
| 441 |
|
| 442 |
def _get_iscased(flags): |
| 443 |
if not flags & SRE_FLAG_IGNORECASE: |
| 444 |
return None |
| 445 |
elif flags & SRE_FLAG_UNICODE: |
| 446 |
return _sre.unicode_iscased |
| 447 |
else: |
| 448 |
return _sre.ascii_iscased |
| 449 |
|
| 450 |
def _get_literal_prefix(pattern, flags): |
| 451 |
# look for literal prefix |
| 452 |
prefix = [] |
| 453 |
prefixappend = prefix.append |
| 454 |
prefix_skip = None |
| 455 |
iscased = _get_iscased(flags) |
| 456 |
for op, av in pattern.data: |
| 457 |
if op is LITERAL: |
| 458 |
if iscased and iscased(av): |
| 459 |
break |
| 460 |
prefixappend(av) |
| 461 |
elif op is SUBPATTERN: |
| 462 |
group, add_flags, del_flags, p = av |
| 463 |
flags1 = _combine_flags(flags, add_flags, del_flags) |
| 464 |
if flags1 & SRE_FLAG_IGNORECASE and flags1 & SRE_FLAG_LOCALE: |
| 465 |
break |
| 466 |
prefix1, prefix_skip1, got_all = _get_literal_prefix(p, flags1) |
| 467 |
if prefix_skip is None: |
| 468 |
if group is not None: |
| 469 |
prefix_skip = len(prefix) |
| 470 |
elif prefix_skip1 is not None: |
| 471 |
prefix_skip = len(prefix) + prefix_skip1 |
| 472 |
prefix.extend(prefix1) |
| 473 |
if not got_all: |
| 474 |
break |
| 475 |
else: |
| 476 |
break |
| 477 |
else: |
| 478 |
return prefix, prefix_skip, True |
| 479 |
return prefix, prefix_skip, False |
| 480 |
|
| 481 |
def _get_charset_prefix(pattern, flags): |
| 482 |
while True: |
| 483 |
if not pattern.data: |
| 484 |
return None |
| 485 |
op, av = pattern.data[0] |
| 486 |
if op is not SUBPATTERN: |
| 487 |
break |
| 488 |
group, add_flags, del_flags, pattern = av |
| 489 |
flags = _combine_flags(flags, add_flags, del_flags) |
| 490 |
if flags & SRE_FLAG_IGNORECASE and flags & SRE_FLAG_LOCALE: |
| 491 |
return None |
| 492 |
|
| 493 |
iscased = _get_iscased(flags) |
| 494 |
if op is LITERAL: |
| 495 |
if iscased and iscased(av): |
| 496 |
return None |
| 497 |
return [(op, av)] |
| 498 |
elif op is BRANCH: |
| 499 |
charset = [] |
| 500 |
charsetappend = charset.append |
| 501 |
for p in av[1]: |
| 502 |
if not p: |
| 503 |
return None |
| 504 |
op, av = p[0] |
| 505 |
if op is LITERAL and not (iscased and iscased(av)): |
| 506 |
charsetappend((op, av)) |
| 507 |
else: |
| 508 |
return None |
| 509 |
return charset |
| 510 |
elif op is IN: |
| 511 |
charset = av |
| 512 |
if iscased: |
| 513 |
for op, av in charset: |
| 514 |
if op is LITERAL: |
| 515 |
if iscased(av): |
| 516 |
return None |
| 517 |
elif op is RANGE: |
| 518 |
if av[1] > 0xffff: |
| 519 |
return None |
| 520 |
if any(map(iscased, range(av[0], av[1]+1))): |
| 521 |
return None |
| 522 |
return charset |
| 523 |
return None |
| 524 |
|
| 525 |
def _compile_info(code, pattern, flags): |
| 526 |
# internal: compile an info block. in the current version, |
| 527 |
# this contains min/max pattern width, and an optional literal |
| 528 |
# prefix or a character map |
| 529 |
lo, hi = pattern.getwidth() |
| 530 |
if hi > MAXCODE: |
| 531 |
hi = MAXCODE |
| 532 |
if lo == 0: |
| 533 |
code.extend([INFO, 4, 0, lo, hi]) |
| 534 |
return |
| 535 |
# look for a literal prefix |
| 536 |
prefix = [] |
| 537 |
prefix_skip = 0 |
| 538 |
charset = None # not used |
| 539 |
if not (flags & SRE_FLAG_IGNORECASE and flags & SRE_FLAG_LOCALE): |
| 540 |
# look for literal prefix |
| 541 |
prefix, prefix_skip, got_all = _get_literal_prefix(pattern, flags) |
| 542 |
# if no prefix, look for charset prefix |
| 543 |
if not prefix: |
| 544 |
charset = _get_charset_prefix(pattern, flags) |
| 545 |
if charset: |
| 546 |
charset, hascased = _optimize_charset(charset) |
| 547 |
assert not hascased |
| 548 |
if charset == _CHARSET_ALL: |
| 549 |
charset = None |
| 550 |
## if prefix: |
| 551 |
## print("*** PREFIX", prefix, prefix_skip) |
| 552 |
## if charset: |
| 553 |
## print("*** CHARSET", charset) |
| 554 |
# add an info block |
| 555 |
emit = code.append |
| 556 |
emit(INFO) |
| 557 |
skip = len(code); emit(0) |
| 558 |
# literal flag |
| 559 |
mask = 0 |
| 560 |
if prefix: |
| 561 |
mask = SRE_INFO_PREFIX |
| 562 |
if prefix_skip is None and got_all: |
| 563 |
mask = mask | SRE_INFO_LITERAL |
| 564 |
elif charset: |
| 565 |
mask = mask | SRE_INFO_CHARSET |
| 566 |
emit(mask) |
| 567 |
# pattern length |
| 568 |
if lo < MAXCODE: |
| 569 |
emit(lo) |
| 570 |
else: |
| 571 |
emit(MAXCODE) |
| 572 |
prefix = prefix[:MAXCODE] |
| 573 |
emit(hi) |
| 574 |
# add literal prefix |
| 575 |
if prefix: |
| 576 |
emit(len(prefix)) # length |
| 577 |
if prefix_skip is None: |
| 578 |
prefix_skip = len(prefix) |
| 579 |
emit(prefix_skip) # skip |
| 580 |
code.extend(prefix) |
| 581 |
# generate overlap table |
| 582 |
code.extend(_generate_overlap_table(prefix)) |
| 583 |
elif charset: |
| 584 |
_compile_charset(charset, flags, code) |
| 585 |
code[skip] = len(code) - skip |
| 586 |
|
| 587 |
def isstring(obj): |
| 588 |
return isinstance(obj, (str, bytes)) |
| 589 |
|
| 590 |
def _code(p, flags): |
| 591 |
|
| 592 |
flags = p.state.flags | flags |
| 593 |
code = [] |
| 594 |
|
| 595 |
# compile info block |
| 596 |
_compile_info(code, p, flags) |
| 597 |
|
| 598 |
# compile the pattern |
| 599 |
_compile(code, p.data, flags) |
| 600 |
|
| 601 |
code.append(SUCCESS) |
| 602 |
|
| 603 |
return code |
| 604 |
|
| 605 |
def _hex_code(code): |
| 606 |
return '[%s]' % ', '.join('%#0*x' % (_sre.CODESIZE*2+2, x) for x in code) |
| 607 |
|
| 608 |
def dis(code): |
| 609 |
import sys |
| 610 |
|
| 611 |
labels = set() |
| 612 |
level = 0 |
| 613 |
offset_width = len(str(len(code) - 1)) |
| 614 |
|
| 615 |
def dis_(start, end): |
| 616 |
def print_(*args, to=None): |
| 617 |
if to is not None: |
| 618 |
labels.add(to) |
| 619 |
args += ('(to %d)' % (to,),) |
| 620 |
print('%*d%s ' % (offset_width, start, ':' if start in labels else '.'), |
| 621 |
end=' '*(level-1)) |
| 622 |
print(*args) |
| 623 |
|
| 624 |
def print_2(*args): |
| 625 |
print(end=' '*(offset_width + 2*level)) |
| 626 |
print(*args) |
| 627 |
|
| 628 |
nonlocal level |
| 629 |
level += 1 |
| 630 |
i = start |
| 631 |
while i < end: |
| 632 |
start = i |
| 633 |
op = code[i] |
| 634 |
i += 1 |
| 635 |
op = OPCODES[op] |
| 636 |
if op in (SUCCESS, FAILURE, ANY, ANY_ALL, |
| 637 |
MAX_UNTIL, MIN_UNTIL, NEGATE): |
| 638 |
print_(op) |
| 639 |
elif op in (LITERAL, NOT_LITERAL, |
| 640 |
LITERAL_IGNORE, NOT_LITERAL_IGNORE, |
| 641 |
LITERAL_UNI_IGNORE, NOT_LITERAL_UNI_IGNORE, |
| 642 |
LITERAL_LOC_IGNORE, NOT_LITERAL_LOC_IGNORE): |
| 643 |
arg = code[i] |
| 644 |
i += 1 |
| 645 |
print_(op, '%#02x (%r)' % (arg, chr(arg))) |
| 646 |
elif op is AT: |
| 647 |
arg = code[i] |
| 648 |
i += 1 |
| 649 |
arg = str(ATCODES[arg]) |
| 650 |
assert arg[:3] == 'AT_' |
| 651 |
print_(op, arg[3:]) |
| 652 |
elif op is CATEGORY: |
| 653 |
arg = code[i] |
| 654 |
i += 1 |
| 655 |
arg = str(CHCODES[arg]) |
| 656 |
assert arg[:9] == 'CATEGORY_' |
| 657 |
print_(op, arg[9:]) |
| 658 |
elif op in (IN, IN_IGNORE, IN_UNI_IGNORE, IN_LOC_IGNORE): |
| 659 |
skip = code[i] |
| 660 |
print_(op, skip, to=i+skip) |
| 661 |
dis_(i+1, i+skip) |
| 662 |
i += skip |
| 663 |
elif op in (RANGE, RANGE_UNI_IGNORE): |
| 664 |
lo, hi = code[i: i+2] |
| 665 |
i += 2 |
| 666 |
print_(op, '%#02x %#02x (%r-%r)' % (lo, hi, chr(lo), chr(hi))) |
| 667 |
elif op is CHARSET: |
| 668 |
print_(op, _hex_code(code[i: i + 256//_CODEBITS])) |
| 669 |
i += 256//_CODEBITS |
| 670 |
elif op is BIGCHARSET: |
| 671 |
arg = code[i] |
| 672 |
i += 1 |
| 673 |
mapping = list(b''.join(x.to_bytes(_sre.CODESIZE, sys.byteorder) |
| 674 |
for x in code[i: i + 256//_sre.CODESIZE])) |
| 675 |
print_(op, arg, mapping) |
| 676 |
i += 256//_sre.CODESIZE |
| 677 |
level += 1 |
| 678 |
for j in range(arg): |
| 679 |
print_2(_hex_code(code[i: i + 256//_CODEBITS])) |
| 680 |
i += 256//_CODEBITS |
| 681 |
level -= 1 |
| 682 |
elif op in (MARK, GROUPREF, GROUPREF_IGNORE, GROUPREF_UNI_IGNORE, |
| 683 |
GROUPREF_LOC_IGNORE): |
| 684 |
arg = code[i] |
| 685 |
i += 1 |
| 686 |
print_(op, arg) |
| 687 |
elif op is JUMP: |
| 688 |
skip = code[i] |
| 689 |
print_(op, skip, to=i+skip) |
| 690 |
i += 1 |
| 691 |
elif op is BRANCH: |
| 692 |
skip = code[i] |
| 693 |
print_(op, skip, to=i+skip) |
| 694 |
while skip: |
| 695 |
dis_(i+1, i+skip) |
| 696 |
i += skip |
| 697 |
start = i |
| 698 |
skip = code[i] |
| 699 |
if skip: |
| 700 |
print_('branch', skip, to=i+skip) |
| 701 |
else: |
| 702 |
print_(FAILURE) |
| 703 |
i += 1 |
| 704 |
elif op in (REPEAT, REPEAT_ONE, MIN_REPEAT_ONE, |
| 705 |
POSSESSIVE_REPEAT, POSSESSIVE_REPEAT_ONE): |
| 706 |
skip, min, max = code[i: i+3] |
| 707 |
if max == MAXREPEAT: |
| 708 |
max = 'MAXREPEAT' |
| 709 |
print_(op, skip, min, max, to=i+skip) |
| 710 |
dis_(i+3, i+skip) |
| 711 |
i += skip |
| 712 |
elif op is GROUPREF_EXISTS: |
| 713 |
arg, skip = code[i: i+2] |
| 714 |
print_(op, arg, skip, to=i+skip) |
| 715 |
i += 2 |
| 716 |
elif op in (ASSERT, ASSERT_NOT): |
| 717 |
skip, arg = code[i: i+2] |
| 718 |
print_(op, skip, arg, to=i+skip) |
| 719 |
dis_(i+2, i+skip) |
| 720 |
i += skip |
| 721 |
elif op is ATOMIC_GROUP: |
| 722 |
skip = code[i] |
| 723 |
print_(op, skip, to=i+skip) |
| 724 |
dis_(i+1, i+skip) |
| 725 |
i += skip |
| 726 |
elif op is INFO: |
| 727 |
skip, flags, min, max = code[i: i+4] |
| 728 |
if max == MAXREPEAT: |
| 729 |
max = 'MAXREPEAT' |
| 730 |
print_(op, skip, bin(flags), min, max, to=i+skip) |
| 731 |
start = i+4 |
| 732 |
if flags & SRE_INFO_PREFIX: |
| 733 |
prefix_len, prefix_skip = code[i+4: i+6] |
| 734 |
print_2(' prefix_skip', prefix_skip) |
| 735 |
start = i + 6 |
| 736 |
prefix = code[start: start+prefix_len] |
| 737 |
print_2(' prefix', |
| 738 |
'[%s]' % ', '.join('%#02x' % x for x in prefix), |
| 739 |
'(%r)' % ''.join(map(chr, prefix))) |
| 740 |
start += prefix_len |
| 741 |
print_2(' overlap', code[start: start+prefix_len]) |
| 742 |
start += prefix_len |
| 743 |
if flags & SRE_INFO_CHARSET: |
| 744 |
level += 1 |
| 745 |
print_2('in') |
| 746 |
dis_(start, i+skip) |
| 747 |
level -= 1 |
| 748 |
i += skip |
| 749 |
else: |
| 750 |
raise ValueError(op) |
| 751 |
|
| 752 |
level -= 1 |
| 753 |
|
| 754 |
dis_(0, len(code)) |
| 755 |
|
| 756 |
|
| 757 |
def compile(p, flags=0): |
| 758 |
# internal: convert pattern list to internal format |
| 759 |
|
| 760 |
if isstring(p): |
| 761 |
pattern = p |
| 762 |
p = _parser.parse(p, flags) |
| 763 |
else: |
| 764 |
pattern = None |
| 765 |
|
| 766 |
code = _code(p, flags) |
| 767 |
|
| 768 |
if flags & SRE_FLAG_DEBUG: |
| 769 |
print() |
| 770 |
dis(code) |
| 771 |
|
| 772 |
# map in either direction |
| 773 |
groupindex = p.state.groupdict |
| 774 |
indexgroup = [None] * p.state.groups |
| 775 |
for k, i in groupindex.items(): |
| 776 |
indexgroup[i] = k |
| 777 |
|
| 778 |
return _sre.compile( |
| 779 |
pattern, flags | p.state.flags, code, |
| 780 |
p.state.groups-1, |
| 781 |
groupindex, tuple(indexgroup) |
| 782 |
) |