1 2 #define DPRINTF(p) /*nothing */ 3 #define DPRINTF(p) printf p 4 #define GETCHAR(c, eptr) c = *eptr; 5 #define GETCHARINC(c, eptr) c = *eptr++; 6 #define class pcre_class 7 #define match_condassert 0x01 /* Called to check a condition assertion */ 8 #define match_isgroup 0x02 /* Set if start of bracketed group */ 9 #else 10 #endif 11 #ifdef DEBUG /* Sigh. Some compilers never learn. */ 12 #ifdef DEBUG 13 #ifdef __cplusplus 14 #include "internal.h" 15 && length - re->max_match_size > start_offset) 16 ((*ecode++ == OP_BEG_WORD) ? prev_is_word : cur_is_word)) 17 ((md->ctypes[*eptr] & ctype_word) != 0); 18 ((md->ctypes[eptr[-1]] & ctype_word) != 0); 19 (eptr == md->end_subject - 1 && *eptr != '\n')) 20 (i.e. keep it out of the loop). Also we can test that there are at least 21 (md->ctypes[*eptr++] & ctype_digit) != 0) 22 (md->ctypes[*eptr++] & ctype_digit) == 0) 23 (md->ctypes[*eptr++] & ctype_space) != 0) 24 (md->ctypes[*eptr++] & ctype_space) == 0) 25 (md->ctypes[*eptr++] & ctype_word) != 0) 26 (md->ctypes[*eptr++] & ctype_word) == 0) 27 (offsetcount - 2) * sizeof (int)); 28 (offsets == NULL && offsetcount > 0)) 29 (pcre_free) (match_block.offset_vector); 30 (pcre_free) (save); 31 (re->tables + fcc_offset)[req_char] : req_char; 32 * Match a back-reference * 33 * Execute a Regular Expression * 34 * Match from current position * 35 * Debugging function to print chars * 36 * Perl-Compatible Regular Expressions * 37 * Macros and tables for character handling * 38 *************************************************/ 39 */ 40 *iptr = -1; 41 *iptr++ = -1; 42 *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT || 43 *prev == OP_ONCE) 44 ----------------------------------------------------------------------------- 45 -1 => failed to match 46 /* 47 /* "Once" brackets are like assertion brackets except that after a match, 48 /* ... else fall through */ 49 /* Advance to a possible match for an initial string after study */ 50 /* Allow compilation as C++ source code, should anybody want to do that. */ 51 /* Always fail if not enough characters left */ 52 /* An alternation is the end of a branch; scan along to find the end of the 53 /* Assert before internal newline if multiline, or before a terminating 54 /* Assertion brackets. Check the alternative branches in turn - the 55 /* At the start of a bracketed group, add the current subject pointer to the 56 /* BRAZERO and BRAMINZERO occur just before a bracket group, indicating 57 /* Caseful comparisons */ 58 /* Change option settings */ 59 /* Common code for all repeated single character type matches */ 60 /* Common code for all repeated single-character matches. We can give 61 /* Compute the minimum number of offsets that we need to reset each time. Doing 62 /* Conditional group: compilation checked that there are no more than 63 /* Continue as from after the assertion, updating the offsets high water 64 /* Continue from after the assertion, updating the offsets high water 65 /* Control never gets here */ 66 /* Control never reaches here */ 67 /* Copy the offset information from temporary store if necessary */ 68 /* Do a single test if no case difference is set up */ 69 /* Do not stick any code in here without much thought; it is assumed 70 /* End of a group, repeated or non-repeating. If we are at the end of 71 /* End of subject assertion (\z) */ 72 /* End of subject or ending \n assertion (\Z) */ 73 /* End of the pattern. If PCRE_NOTEMPTY is set, fail if we have matched 74 /* First, ensure the minimum number of matches are present. */ 75 /* First, ensure the minimum number of matches are present. Use inline 76 /* First, ensure the minimum number of matches are present. We get back 77 /* Flag bits for the match() function */ 78 /* For a non-repeating ket, just continue at this level. This also 79 /* For anchored or unanchored matches, there may be a "last known required 80 /* For extended extraction brackets (large number), we have to fish out 81 /* For extended extraction brackets (large number), we have to fish out the 82 /* For matches anchored to the end of the pattern, we can often avoid 83 /* If a back reference hasn't been set, the length that is passed is greater 84 /* If checking an assertion for a condition, return TRUE. */ 85 /* If hit the end of the group (which could be repeated), fail */ 86 /* If max == min we can continue with the main loop without the 87 /* If maximizing it is worth using inline code for speed, doing the type 88 /* If maximizing, find the longest possible run, then work backwards. */ 89 /* If maximizing, find the longest string and work backwards */ 90 /* If min = max, continue at the same level without recursing */ 91 /* If min = max, continue at the same level without recursion. 92 /* If minimizing, keep testing the rest of the expression and advancing 93 /* If minimizing, keep trying and advancing the pointer */ 94 /* If minimizing, we have to test the rest of the pattern before each 95 /* If req_char is set, we know that that character must appear in the subject 96 /* If the expression has got more back references than the offsets supplied can 97 /* If the length of the reference is zero, just continue with the 98 /* If the reference is unset, set the length to be longer than the amount 99 /* If we can't find the required character, break the matching loop */ 100 /* If we have found the required character, save the point where we 101 /* In all other cases except a conditional group we have to check the 102 /* In case the recursion has set more capturing values, save the final 103 /* Include the internals header, which itself includes Standard C headers plus 104 /* Insufficient room for saving captured contents */ 105 /* Loop for handling unanchored repeated matching attempts; for anchored regexs 106 /* Match a back reference, possibly repeatedly. Look past the end of the 107 /* Match a character class, possibly repeatedly. Look past the end of the 108 /* Match a negated single character */ 109 /* Match a negated single character repeatedly. This is almost a repeat of 110 /* Match a run of characters */ 111 /* Match a single character repeatedly; different opcodes share code. */ 112 /* Match a single character type repeatedly; several different opcodes 113 /* Match a single character type; inline for speed */ 114 /* Min and max values for the common repeats; for the maxima, 0 => infinity */ 115 /* Move the subject pointer back. This occurs only at the start of 116 /* Negative assertion: all branches must fail to match */ 117 /* Now start processing the operations. */ 118 /* OP_KETRMAX */ 119 /* On entry ecode points to the first opcode, and eptr to the first character 120 /* Opening capturing bracket. If there is space in the offset vector, save 121 /* Or to a non-unique first char after study */ 122 /* Or to a unique first char if possible */ 123 /* Or to just after \n for a multiline match if possible */ 124 /* Other types of node can be handled by a switch */ 125 /* Otherwise test for either case */ 126 /* Print a sequence of chars in printable format, stopping at the end of the 127 /* Recursion matches the current regex, nested. If there are any capturing 128 /* Reset the maximum number of extractions we might see. */ 129 /* Reset the value of the ims flags, in case they got changed during 130 /* Reset the working variable associated with each extraction. These should 131 /* Separate the caselesss case for speed */ 132 /* Set up for repetition, or handle the non-repeated case */ 133 /* Set up the first character to match, if available. The first_char value is 134 /* Skip over conditional reference data or large extraction number data if 135 /* Start of subject assertion */ 136 /* Start of subject unless notbol, or after internal newline if multiline */ 137 /* Structure for building a chain of data that actually lives on the 138 /* The code is duplicated for the caseless and caseful cases, for speed, 139 /* The condition is an assertion. Call match() to evaluate it - setting 140 /* The ims options can vary during the matching as a result of the presence 141 /* The repeating kets try the rest of the pattern or restart from the 142 /* There's been some horrible disaster. */ 143 /* This "while" is the end of the "do" above */ 144 /* This function applies a compiled re to a subject string and picks out 145 /* Use a macro for debugging printing, 'cause that limits the use of #ifdef 146 /* We don't need to repeat the search if we haven't yet reached the 147 /* When a match occurs, substrings will be set for all internal extractions; 148 /* Word boundary assertions */ 149 /************************************************* 150 1. This software is distributed in the hope that it will be useful, 151 2. The origin of this software must not be misrepresented, either by 152 3. Altered versions must be plainly marked as such, and must not be 153 4. If PCRE is embedded in any software that is released under the GNU 154 5.005. If there is an options reset, it will get obeyed in the normal 155 6 : 3 + (ecode[1] << 8) + ecode[2]), 156 < -1 => some kind of unexpected problem 157 = 0 => success, but offsets is not big enough 158 Arguments: 159 BOOL anchored; 160 BOOL cur_is_word = (eptr < md->end_subject) && 161 BOOL is_subject; 162 BOOL minimize = FALSE; 163 BOOL prev_is_word = (eptr != md->start_subject) && 164 BOOL rc; 165 BOOL startline; 166 BOOL using_temporary_offsets = FALSE; 167 Copyright (c) 1997-2000 University of Cambridge 168 DPRINTF ((">>>> returning %d\n", match_block.errorcode)); 169 DPRINTF ((">>>> returning %d\n", rc)); 170 DPRINTF (("Copied offsets from temporary memory\n")); 171 DPRINTF (("Freeing temporary memory\n")); 172 DPRINTF (("Got memory to hold back references\n")); 173 DPRINTF (("Unknown opcode %d\n", *ecode)); 174 DPRINTF (("bracket %d failed\n", number)); 175 DPRINTF (("bracket 0 failed\n")); 176 DPRINTF (("ims reset to %02lx\n", ims)); 177 DPRINTF (("ims set to %02lx at group repeat\n", ims)); 178 DPRINTF (("ims set to %02lx\n", ims)); 179 DPRINTF (("matching %c{%d,%d} against subject %.*s\n", c, min, max, 180 DPRINTF (("negative matching %c{%d,%d} against subject %.*s\n", c, min, max, 181 DPRINTF (("saving %d %d %d\n", save_offset1, save_offset2, save_offset3)); 182 DPRINTF (("start bracket 0\n")); 183 GETCHAR (c, eptr) /* Get character */ 184 GETCHARINC (c, eptr) /* Get character; increment eptr */ 185 General Purpose Licence (GPL), then the terms of that licence shall 186 However, if the referenced string is the empty string, always treat 187 If the bracket fails to match, we need to restore this value and also the 188 If there isn't enough space in the offset vector, treat this as if it were a 189 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. 190 Otherwise, we can use the vector supplied, rounding down its size to a multiple 191 Permission is granted to anyone to use this software for any purpose on any 192 REPEATCHAR: 193 REPEATNOTCHAR: 194 REPEATTYPE: 195 Returns: > 0 => success; value is the number of elements filled in 196 Returns: TRUE if matched 197 Returns: TRUE if matched 198 Returns: nothing 199 They are not both allowed to be zero. */ 200 This is a library of functions to support regular expressions whose syntax 201 This is the forcible breaking of infinite loops as implemented in Perl 202 Writing separate code makes it go faster, as does using an autoincrement and 203 Written by: Philip Hazel <ph10 (a] cam.ac.uk> 204 a move back into the brackets. Check the alternative branches in turn - the 205 address of eptr, so that eptr can be a register variable. */ 206 an assertion "group", stop matching and return TRUE, but record the 207 an empty string - recursion will then try other alternatives, if any. */ 208 an error. Save the top 15 values on the stack, and accept that the rest 209 an unanchored pattern, of course. If there's no first char and the pattern was 210 analyzing most of the pattern. length > re->max_match_size is 211 anchored = ((re->options | options) & PCRE_ANCHORED) != 0; 212 and advance one byte in the pattern code. */ 213 and reinstate them after the recursion. However, we don't know how many 214 and semantics are as close as possible to those of the Perl 5 language. See 215 and the required character in fact is caseful. */ 216 at run time, so we have to test for anchoring. The first char may be unset for 217 avoid duplicate testing (which takes significant time). This covers the vast 218 backing off on a match. */ 219 bmtable = extra->data.bmtable; 220 both cases of the character. Otherwise set the two values the same, which will 221 bracketed group and go to there. */ 222 brackets - for testing for empty matches 223 brackets started but not finished, we have to save their starting points 224 break; 225 but WITHOUT ANY WARRANTY; without even the implied warranty of 226 c != md->lcc[*eptr++]) 227 c = *ecode++ - OP_CRSTAR; 228 c = *ecode++ - OP_NOTSTAR; 229 c = *ecode++ - OP_STAR; 230 c = *ecode++ - OP_TYPESTAR; 231 c = *ecode++; 232 c = *eptr++; 233 c = 15; 234 c = max - min; 235 c = md->end_subject - eptr; 236 c = md->lcc[c]; 237 c = md->offset_max; 238 c == md->lcc[*eptr++]) 239 can't just fail here, because of the possibility of quantifiers with zero 240 case OP_ALT: 241 case OP_ANY: 242 case OP_ASSERT: 243 case OP_ASSERTBACK: 244 case OP_ASSERTBACK_NOT: 245 case OP_ASSERT_NOT: 246 case OP_BEG_WORD: 247 case OP_BRA: /* Non-capturing bracket: optimized */ 248 case OP_BRAMINZERO: 249 case OP_BRANUMBER: 250 case OP_BRAZERO: 251 case OP_CHARS: 252 case OP_CIRC: 253 case OP_CLASS: 254 case OP_COND: 255 case OP_CREF: 256 case OP_CRMINPLUS: 257 case OP_CRMINQUERY: 258 case OP_CRMINRANGE: 259 case OP_CRMINSTAR: 260 case OP_CRPLUS: 261 case OP_CRQUERY: 262 case OP_CRRANGE: 263 case OP_CRSTAR: 264 case OP_DIGIT: 265 case OP_DOLL: 266 case OP_END: 267 case OP_END_WORD: 268 case OP_EOD: 269 case OP_EODN: 270 case OP_EXACT: 271 case OP_KET: 272 case OP_KETRMAX: 273 case OP_KETRMIN: 274 case OP_MINPLUS: 275 case OP_MINQUERY: 276 case OP_MINSTAR: 277 case OP_MINUPTO: 278 case OP_NOT: 279 case OP_NOTEXACT: 280 case OP_NOTMINPLUS: 281 case OP_NOTMINQUERY: 282 case OP_NOTMINSTAR: 283 case OP_NOTMINUPTO: 284 case OP_NOTPLUS: 285 case OP_NOTQUERY: 286 case OP_NOTSTAR: 287 case OP_NOTUPTO: 288 case OP_NOT_DIGIT: 289 case OP_NOT_WHITESPACE: 290 case OP_NOT_WORDCHAR: 291 case OP_NOT_WORD_BOUNDARY: 292 case OP_ONCE: 293 case OP_OPT: 294 case OP_PLUS: 295 case OP_QUERY: 296 case OP_RECURSE: 297 case OP_REF: 298 case OP_REVERSE: 299 case OP_SOD: 300 case OP_STAR: 301 case OP_TYPEEXACT: 302 case OP_TYPEMINPLUS: 303 case OP_TYPEMINQUERY: 304 case OP_TYPEMINSTAR: 305 case OP_TYPEMINUPTO: 306 case OP_TYPEPLUS: 307 case OP_TYPEQUERY: 308 case OP_TYPESTAR: 309 case OP_TYPEUPTO: 310 case OP_UPTO: 311 case OP_WHITESPACE: 312 case OP_WORDCHAR: 313 case OP_WORD_BOUNDARY: 314 case matching may be when this character is hit, so test for it in both its 315 caselessly, or if there are any changes of this flag within the regex, set up 316 cases if necessary. However, the different cased versions will not be set up 317 character" set. If the PCRE_CASELESS is set, implying that the match starts 318 characters and work backwards. */ 319 code for maximizing the speed, and do the type test once at the start 320 code to character type repeats - written out again for speed. */ 321 commoning these up that doesn't require a test of the positive/negative 322 computer system, and to redistribute it freely, subject to the following 323 const char *subject; 324 const pcre *re; 325 const pcre_extra *extra; 326 const uschar *bmtable = NULL; 327 const uschar *data = ecode + 1; /* Save for matching */ 328 const uschar *end_subject; 329 const uschar *next = ecode + 1; 330 const uschar *p = md->start_subject + md->offset_vector[offset]; 331 const uschar *p; 332 const uschar *pp = eptr; 333 const uschar *prev = ecode - (ecode[1] << 8) - ecode[2]; 334 const uschar *prev = ecode; 335 const uschar *req_char_ptr = start_match - 1; 336 const uschar *saved_eptr = eptr; 337 const uschar *saved_eptr = eptrb->saved_eptr; 338 const uschar *saved_eptr; 339 const uschar *start_bits = NULL; 340 const uschar *start_match = (const uschar *) subject + start_offset; 341 continue; /* With the main loop */ 342 continue; 343 course of events. */ 344 ctype = *ecode++; /* Code for the character type */ 345 cur_is_word == prev_is_word : cur_is_word != prev_is_word) 346 current high water mark for use by positive assertions. Do this also 347 default: /* No repeat follows */ 348 default: 349 do 350 each branch of a lookbehind assertion. If we are too close to the start to 351 each substring: the offsets to the start and end of the substring. 352 ecode position in code 353 ecode + ((offset < offset_top && md->offset_vector[offset] >= 0) ? 354 ecode += (ecode[1] << 8) + ecode[2]; 355 ecode += 2; 356 ecode += 3 + (ecode[4] << 8) + ecode[5]; 357 ecode += 33; /* Advance past the item */ 358 ecode += 3; /* Advance past the item */ 359 ecode += 3; 360 ecode += 5; 361 ecode = next + 3; 362 ecode++; 363 else 364 else if ((extra->options & PCRE_STUDY_BM) != 0) 365 else if (first_char >= 0) 366 else if (start_bits != NULL) 367 else if (startline) 368 encountered */ 369 end_subject = match_block.end_subject; 370 eptr pointer in subject 371 eptr points into the subject 372 eptr += c; 373 eptr += length; 374 eptr += min; 375 eptr -= (ecode[1] << 8) + ecode[2]; 376 eptr -= length; 377 eptr = md->end_match_ptr; 378 eptr++; 379 eptrb pointer to chain of blocks containing eptr at start of 380 eptrb = &newptrb; 381 eptrb = eptrb->prev; /* Back up the stack of bracket start pointers */ 382 eptrblock *eptrb; 383 eptrblock newptrb; 384 eptrblock; 385 exactly what going to the ket would do. */ 386 explicit claim or by omission. 387 external_extra points to "hints" from pcre_study() or is NULL 388 external_re points to the compiled expression 389 extraction by setting the offsets and bumping the high water mark. */ 390 first_char = match_block.lcc[first_char]; 391 first_char = re->first_char; 392 flags can contain 393 for (;;) 394 for (i = 1; i <= c; i++) 395 for (i = 1; i <= min; i++) 396 for (i = min; i < max; i++) 397 for (i = min;; i++) 398 for the "once" (not-backup up) groups. */ 399 for the match to succeed. If the first character is set, req_char must be 400 found it, so that we don't search again next time round the loop if 401 from a previous iteration of this group, and be referred to by a reference 402 goto REPEATCHAR; 403 goto REPEATNOTCHAR; 404 goto REPEATTYPE; 405 group number back at the start and if necessary complete handling an 406 happens for a repeating ket if no characters were matched in the group. 407 here; that is handled in the code for KET. */ 408 hold, we get a temporary bit of working store to use during the matching. 409 i.e. it could be ()* or ()? in the pattern. Brackets with fixed upper 410 if (!anchored) 411 if (!match (start_match, re->code, 2, &match_block, ims, NULL, match_isgroup)) 412 if (!match_ref (offset, eptr, length, md, ims)) 413 if (!md->endonly) 414 if (!rc) 415 if (!startline && extra != NULL) 416 if ((*ecode++ == OP_WORD_BOUNDARY) ? 417 if ((data[c / 8] & (1 << (c & 7))) != 0) 418 if ((data[c / 8] & (1 << (c & 7))) == 0) 419 if ((extra->options & PCRE_STUDY_MAPPED) != 0) 420 if ((flags & match_condassert) != 0) 421 if ((flags & match_isgroup) != 0) 422 if ((ims & PCRE_CASELESS) != 0) 423 if ((ims & PCRE_DOTALL) == 0 && c == '\n') 424 if ((ims & PCRE_DOTALL) == 0 && eptr < md->end_subject && *eptr == '\n') 425 if ((ims & PCRE_DOTALL) == 0) 426 if ((ims & PCRE_MULTILINE) != 0) 427 if ((md->ctypes[*eptr++] & ctype_digit) != 0) 428 if ((md->ctypes[*eptr++] & ctype_digit) == 0) 429 if ((md->ctypes[*eptr++] & ctype_space) != 0) 430 if ((md->ctypes[*eptr++] & ctype_space) == 0) 431 if ((md->ctypes[*eptr++] & ctype_word) != 0) 432 if ((md->ctypes[*eptr++] & ctype_word) == 0) 433 if ((md->ctypes[c] & ctype_digit) != 0) 434 if ((md->ctypes[c] & ctype_digit) == 0) 435 if ((md->ctypes[c] & ctype_space) != 0) 436 if ((md->ctypes[c] & ctype_space) == 0) 437 if ((md->ctypes[c] & ctype_word) != 0) 438 if ((md->ctypes[c] & ctype_word) == 0) 439 if ((options & ~PUBLIC_EXEC_OPTIONS) != 0) 440 if ((re->options & PCRE_FIRSTSET) != 0) 441 if ((re->options & PCRE_REQCHSET) != 0) 442 if ((start_bits[c / 8] & (1 << (c & 7))) == 0) 443 if (*ecode != OP_ONCE && *ecode != OP_ALT) 444 if (*ecode == OP_KET || eptr == saved_eptr) 445 if (*ecode == OP_KET) 446 if (*ecode == OP_KETRMIN) 447 if (*ecode++ != *eptr++) 448 if (*ecode++ == *eptr++) 449 if (*eptr != '\n') 450 if (*eptr++ == '\n') 451 if (*p++ != *eptr++) 452 if (*p++ == req_char) 453 if (*prev != OP_COND) 454 if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT || 455 if (bmtable != NULL) 456 if (bmtable[*start_match]) 457 if (c != *eptr++) 458 if (c != md->lcc[*eptr++]) 459 if (c < 16) 460 if (c == *eptr++) 461 if (c == md->lcc[*eptr++]) 462 if (c > md->end_subject - eptr) 463 if (cur_is_word == prev_is_word || 464 if (ecode[3] == OP_CREF) /* Condition is extraction test */ 465 if (ecode[3] == OP_OPT) 466 if (eptr != md->start_subject && eptr[-1] != '\n') 467 if (eptr != md->start_subject) 468 if (eptr < md->end_subject - 1 || 469 if (eptr < md->end_subject) 470 if (eptr < md->start_subject) 471 if (eptr >= md->end_subject || 472 if (eptr >= md->end_subject || (md->ctypes[*eptr] & ctype_digit) != 0) 473 if (eptr >= md->end_subject || (md->ctypes[*eptr] & ctype_digit) == 0) 474 if (eptr >= md->end_subject || (md->ctypes[*eptr] & ctype_space) != 0) 475 if (eptr >= md->end_subject || (md->ctypes[*eptr] & ctype_space) == 0) 476 if (eptr >= md->end_subject || (md->ctypes[*eptr] & ctype_word) != 0) 477 if (eptr >= md->end_subject || (md->ctypes[*eptr] & ctype_word) == 0) 478 if (eptr >= md->end_subject || *eptr == '\n') 479 if (eptr >= md->end_subject || c != *eptr) 480 if (eptr >= md->end_subject || c != md->lcc[*eptr]) 481 if (eptr >= md->end_subject || c == *eptr) 482 if (eptr >= md->end_subject || c == md->lcc[*eptr]) 483 if (eptr >= md->end_subject) 484 if (eptr++ >= md->end_subject) 485 if (i >= max || !match_ref (offset, eptr, length, md, ims)) 486 if (i >= max || eptr >= md->end_subject || 487 if (i >= max || eptr >= md->end_subject || c != *eptr++) 488 if (i >= max || eptr >= md->end_subject || c == *eptr++) 489 if (i >= max || eptr >= md->end_subject) 490 if (is_subject && length > md->end_subject - p) 491 if (isprint (c = *(p++))) 492 if (length == 0) 493 if (length > md->end_subject - eptr) 494 if (match (eptr, ecode + 3, offset_top, md, ims, NULL, 495 if (match (eptr, ecode + 3, offset_top, md, ims, NULL, match_isgroup)) 496 if (match (eptr, ecode + 3, offset_top, md, ims, eptrb, 0) || 497 if (match (eptr, ecode + 3, offset_top, md, ims, eptrb, match_isgroup)) 498 if (match (eptr, ecode, offset_top, md, ims, eptrb, 0)) 499 if (match (eptr, next + 3, offset_top, md, ims, eptrb, match_isgroup)) 500 if (match (eptr, next, offset_top, md, ims, eptrb, match_isgroup)) 501 if (match (eptr, prev, offset_top, md, ims, eptrb, match_isgroup) || 502 if (match (eptr--, ecode, offset_top, md, ims, eptrb, 0)) 503 if (match_block.end_offset_top > offsetcount) 504 if (match_block.offset_vector != NULL) 505 if (match_block.offset_vector == NULL) 506 if (max == 0) 507 if (md->lcc[*ecode++] != md->lcc[*eptr++]) 508 if (md->lcc[*ecode++] == md->lcc[*eptr++]) 509 if (md->lcc[*p++] != md->lcc[*eptr++]) 510 if (md->notbol && eptr == md->start_subject) 511 if (md->notempty && eptr == md->start_match) 512 if (md->noteol) 513 if (min == max) 514 if (min > 0) 515 if (min > md->end_subject - eptr) 516 if (minimize) 517 if (number > 0) 518 if (number > EXTRACT_BASIC_MAX) 519 if (offset < md->offset_max) 520 if (offset >= md->offset_max) 521 if (offset_top <= offset) 522 if (offsetcount < 2) 523 if (offsetcount >= 4) 524 if (op > OP_BRA) 525 if (p > req_char_ptr) 526 if (p >= end_subject) 527 if (pp == req_char || pp == req_char2) 528 if (re == NULL || subject == NULL || 529 if (re->magic_number != MAGIC_NUMBER) 530 if (re->max_match_size >= 0 531 if (re->top_backref > 0 && re->top_backref >= ocount / 3) 532 if (req_char == req_char2) 533 if (req_char >= 0) 534 if (resetcount > offsetcount) 535 if (save != stacksave) 536 if (save == NULL) 537 if (skipped_chars) 538 if (start_match + bmtable[256] > end_subject) 539 if (start_match > match_block.start_subject + start_offset) 540 if (using_temporary_offsets) 541 if certain parts of the pattern were not used. */ 542 if the malloc fails ... there is no way of returning to the top level with 543 implied in the second condition, because start_offset > 0. */ 544 ims current /i, /m, and /s options 545 ims the ims flags 546 ims = (ims & ~PCRE_IMS) | ecode[4]; 547 ims = ecode[1]; 548 ims = original_ims; 549 ims = re->options & (PCRE_CASELESS | PCRE_MULTILINE | PCRE_DOTALL); 550 in the pattern. */ 551 in the subject string, while eptrb holds the value of eptr at the start of the 552 initialize them to avoid reading uninitialized locations. */ 553 inline, and there are *still* stupid compilers about that don't like indented 554 inside the group. 555 int 556 int *offsets; 557 int *save; 558 int c; 559 int first_char = -1; 560 int flags; 561 int length; 562 int min, max, ctype; 563 int number = *prev - OP_BRA; 564 int number = op - OP_BRA; 565 int offset = (ecode[1] << 9) | (ecode[2] << 1); /* Doubled reference number */ 566 int offset = (ecode[4] << 9) | (ecode[5] << 1); /* Doubled reference number */ 567 int offset; 568 int offset_top; 569 int offsetcount; 570 int op = (int) *ecode; 571 int options; 572 int rc; 573 int req_char = -1; 574 int req_char2 = -1; 575 int resetcount, ocount; 576 int save_offset1 = md->offset_vector[offset]; 577 int save_offset2 = md->offset_vector[offset + 1]; 578 int save_offset3 = md->offset_vector[md->offset_end - number]; 579 int skipped_chars = 0; 580 int stacksave[15]; 581 int start_offset; 582 is a bit large to put on the stack, but using malloc for small numbers 583 is_subject TRUE if printing from within md->start_subject 584 it as matched, any number of times (otherwise there could be infinite 585 item to see if there is repeat information following. The code is similar 586 item to see if there is repeat information following. Then obey similar 587 last bracketed group - used for breaking infinite loops matching zero-length 588 later in the subject; otherwise the test starts at the match point. This 589 length length of subject string (may contain binary zeros) 590 length length to be matched 591 length number to print 592 length = (offset >= offset_top || md->offset_vector[offset] < 0) ? 593 length = md->end_subject - p; 594 level without recursing. Otherwise, if minimizing, keep trying the rest of 595 loop. */ 596 loops). */ 597 main loop. */ 598 majority of cases. It will be suboptimal when the case flag changes in a regex 599 mark, since extracts may have been taken during the assertion. */ 600 mark, since extracts may have been taken. */ 601 match (eptr, ecode + 3, offset_top, md, ims, eptrb, 0)) 602 match (eptr, ecode, offset_top, md, ims, eptrb, flags) 603 match (eptr, prev, offset_top, md, ims, eptrb, match_isgroup)) 604 match_block.ctypes = re->tables + ctypes_offset; 605 match_block.end_subject = match_block.start_subject + length; 606 match_block.endonly = (re->options & PCRE_DOLLAR_ENDONLY) != 0; 607 match_block.errorcode = PCRE_ERROR_NOMATCH; /* Default error */ 608 match_block.errorcode == PCRE_ERROR_NOMATCH && 609 match_block.lcc = re->tables + lcc_offset; 610 match_block.lcc[*start_match] != first_char) 611 match_block.notbol = (options & PCRE_NOTBOL) != 0; 612 match_block.notempty = (options & PCRE_NOTEMPTY) != 0; 613 match_block.noteol = (options & PCRE_NOTEOL) != 0; 614 match_block.offset_end = ocount; 615 match_block.offset_max = (2 * ocount) / 3; 616 match_block.offset_overflow = FALSE; 617 match_block.offset_overflow = TRUE; 618 match_block.offset_vector = (int *) (pcre_malloc) (ocount * sizeof (int)); 619 match_block.offset_vector = offsets; 620 match_block.start_match = start_match; 621 match_block.start_pattern = re->code; 622 match_block.start_subject = (const uschar *) subject; 623 match_condassert - this is an assertion condition 624 match_condassert | match_isgroup)) 625 match_data *md; 626 match_data match_block; 627 match_isgroup - this is the start of a bracketed group 628 match_isgroup); 629 match_ref (offset, eptr, length, md, ims) 630 matches, we carry on as at the end of a normal bracket, leaving the subject 631 matching won't pass the KET for an assertion. If any one branch matches, 632 matching won't pass the KET for this kind of subpattern. If any one branch 633 max = (ecode[1] << 8) + ecode[2]; 634 max = (ecode[3] << 8) + ecode[4]; 635 max = INT_MAX; 636 max = rep_max[c]; /* zero for max => infinity */ 637 max, eptr)); 638 maximum. Alternatively, if maximizing, find the maximum number of 639 may be wrong. */ 640 md pointer to "static" info for the match 641 md pointer to matching data block, if is_subject is TRUE 642 md points to match data block 643 md->end_match_ptr = eptr; /* For ONCE */ 644 md->end_match_ptr = eptr; /* Record where we ended */ 645 md->end_offset_top = offset_top; /* and how many extracts were taken */ 646 md->end_offset_top = offset_top; 647 md->end_subject - eptr + 1 : 648 md->errorcode = PCRE_ERROR_UNKNOWN_NODE; 649 md->offset_overflow = TRUE; 650 md->offset_vector[md->offset_end - i] = save[i]; 651 md->offset_vector[md->offset_end - number] = eptr - md->start_subject; 652 md->offset_vector[md->offset_end - number] = save_offset3; 653 md->offset_vector[md->offset_end - number]; 654 md->offset_vector[offset + 1] - md->offset_vector[offset]; 655 md->offset_vector[offset + 1] = eptr - md->start_subject; 656 md->offset_vector[offset + 1] = save_offset2; 657 md->offset_vector[offset] = 658 md->offset_vector[offset] = save_offset1; 659 memcpy (offsets + 2, match_block.offset_vector + 2, 660 min = (ecode[1] << 8) + ecode[2]; 661 min = 0; 662 min = max = (ecode[1] << 8) + ecode[2]; 663 min = max = 1; 664 min = rep_min[c]; /* Pick up values from tables; */ 665 minima. */ 666 minimize = (*ecode == OP_CRMINRANGE); 667 minimize = (c & 1) != 0; 668 minimize = *ecode == OP_MINUPTO; 669 minimize = *ecode == OP_NOTMINUPTO; 670 minimize = *ecode == OP_TYPEMINUPTO; 671 minimize = TRUE; 672 minimum number of matches are present. If min = max, continue at the same 673 misrepresented as being the original software. 674 move back, this match function fails. */ 675 mustn't change the current values of the data slot, because they may be set 676 need to recurse. */ 677 never be used unless previously set, but they get saved and restored, and so we 678 never set for an anchored regular expression, but the anchoring may be forced 679 newline unless endonly is set, else end of subject unless noteol is set. */ 680 newptrb.prev = eptrb; 681 newptrb.saved_eptr = eptr; 682 next += (next[1] << 8) + next[2]; 683 non-capturing bracket. Don't worry about setting the flag for the error case 684 number = (ecode[4] << 8) | ecode[5]; 685 number = (prev[4] << 8) | prev[5]; 686 number from a dummy opcode at the start. */ 687 number, then move along the subject till after the recursive match, 688 ocount = offsetcount - (offsetcount % 3); 689 ocount = re->top_backref * 3 + 3; 690 of (?ims) items in the pattern. They are kept in a local variable so that 691 of 3. */ 692 of subject left; this ensures that every attempt at a match fails. We 693 offset index into the offset vector 694 offset = number << 1; 695 offset_top current top pointer 696 offset_top = md->end_offset_top; 697 offset_top = offset + 2; 698 offset_top, md, ims, eptrb, match_isgroup); 699 offsetcount the number of elements in the vector 700 offsets points to a vector of ints to be filled in with offsets 701 offsets[0] = start_match - match_block.start_subject; 702 offsets[1] = match_block.end_match_ptr - match_block.start_subject; 703 op = OP_BRA; 704 opcode. */ 705 optimization can save a huge amount of backtracking in patterns with nested 706 option for each character match. Maybe that wouldn't add very much to the 707 options option bits 708 p points to characters 709 p--; 710 past the end if there is only one branch, but that's OK because that is 711 pchars (ecode, length, FALSE, md); 712 pchars (eptr, 16, TRUE, md); 713 pchars (eptr, length, TRUE, md); 714 pchars (p, length, FALSE, md); 715 pchars (p, length, is_subject, md) 716 pchars (start_match, end_subject - start_match, TRUE, &match_block); 717 pcre_exec (re, extra, subject, length, start_offset, options, offsets, offsetcount) 718 place we found it at last time. */ 719 pointer. */ 720 portions of the string if it matches. Two elements in the vector are set for 721 pre-processor statements. I suppose it's only been 10 years... */ 722 preceded by BRAZERO or BRAMINZERO. */ 723 preceding bracket, in the appropriate order. */ 724 preceding bracket, in the appropriate order. We need to reset any options 725 printf (" against backref "); 726 printf (" against pattern "); 727 printf ("%c", c); 728 printf (">>>> Match against: "); 729 printf (">>>>> Skipped %d chars to reach first character\n", 730 printf ("\\x%02x", c); 731 printf ("\n"); 732 printf ("end bracket %d", number); 733 printf ("matching subject "); 734 printf ("matching subject <null> against pattern "); 735 printf ("matching subject <null>"); 736 printf ("start bracket %d subject=", number); 737 rc = 0; 738 rc = match (eptr, md->start_pattern, offset_top, md, ims, eptrb, 739 rc = match_block.offset_overflow ? 0 : match_block.end_offset_top / 2; 740 register const uschar *ecode; 741 register const uschar *eptr; 742 register const uschar *p = start_match + ((first_char >= 0) ? 1 : 0); 743 register int *iend = iptr + resetcount; 744 register int *iend = iptr - resetcount / 2 + 1; 745 register int *iptr = match_block.offset_vector + ocount; 746 register int *iptr = match_block.offset_vector; 747 register int c = *start_match; 748 register int c; 749 register int i; 750 register int length = ecode[1]; 751 register int pp = *p++; 752 repeat it in the interests of efficiency. */ 753 repeat limits are compiled as a number of copies, with the optional ones 754 req_char = re->req_char; 755 req_char2 = ((re->options & (PCRE_CASELESS | PCRE_ICHANGED)) != 0) ? 756 req_char_ptr = p; 757 resetcount = 2 + re->top_bracket * 2; 758 resetcount = ocount; 759 restoring at the exit of a group is easy. */ 760 restrictions: 761 return FALSE; 762 return PCRE_ERROR_BADMAGIC; 763 return PCRE_ERROR_BADOPTION; 764 return PCRE_ERROR_NOMATCH; 765 return PCRE_ERROR_NOMEMORY; 766 return PCRE_ERROR_NULL; 767 return TRUE; 768 return match (eptr, 769 return match (eptr, ecode + 3, offset_top, md, ims, eptrb, match_isgroup); 770 return match_block.errorcode; 771 return rc; 772 save = (int *) (pcre_malloc) ((c + 1) * sizeof (int)); 773 save = stacksave; 774 save[i] = md->offset_vector[md->offset_end - i]; 775 seems expensive. As a compromise, the stack is used when there are fewer 776 share code. This is very similar to the code for single characters, but we 777 similar code to character type repeats - written out again for speed. 778 since matching characters is likely to be quite common. First, ensure the 779 skipped_chars += bmtable[*start_match], 780 skipped_chars += bmtable[256] - 1; 781 skipped_chars -= bmtable[256] - 1; 782 skipped_chars); 783 skipped_chars++, 784 stack of such pointers, to be re-instated at the end of the group when we hit 785 stack, for holding the values of the subject pointer at the start of each 786 start of each branch to move the current point backwards, so the code at 787 start_bits = extra->data.start_bits; 788 start_match += bmtable[*start_match]; 789 start_match += bmtable[256] - 1; 790 start_match -= bmtable[256] - 1; 791 start_match = (const uschar *) subject + length - re->max_match_size; 792 start_match++ < end_subject); 793 start_match++; 794 start_offset where to start in the subject string 795 startline = (re->options & PCRE_STARTLINE) != 0; 796 static BOOL 797 static const char rep_max[] = 798 static const char rep_min[] = 799 static void 800 strings. 801 struct eptrblock *prev; 802 studied, there may be a bitmap of possible first characters. */ 803 subject points to the subject string 804 subject if the requested. 805 subpattern - to break infinite loops. */ 806 subpattern, so as to detect when an empty string has been matched by a 807 subsequent match. */ 808 such there are (offset_top records the completed total) so we just have 809 supersede any condition above with which it is incompatible. 810 switch (*ecode) 811 switch (ctype) 812 switch (op) 813 test once at the start (i.e. keep it out of the loop). */ 814 than 16 values to store; otherwise malloc is used. A problem is what to do 815 than the number of characters left in the string, so the match fails. 816 that "continue" in the code above comes out to here to repeat the main 817 that changed within the bracket before re-running it, so check the next 818 that it may occur zero times. It may repeat infinitely, or not at all - 819 the assertion is true. Lookbehind assertions have an OP_REVERSE item at the 820 the closing ket. When match() is called in other circumstances, we don't add to 821 the code for a repeated single character, but I haven't found a nice way of 822 the current subject position in the working slot at the top of the vector. We 823 the expression and advancing one matching character if failing, up to the 824 the external pcre header. */ 825 the file Tech.Notes for some information on the internals. 826 the final argument TRUE causes it to stop at the end of an assertion. */ 827 the group. */ 828 the length of the reference string explicitly rather than passing the 829 the loop runs just once. */ 830 the minimum number of bytes before we start. */ 831 the number from a dummy opcode at the start. */ 832 the point in the subject string is not moved back. Thus there can never be 833 the pointer while it matches the class. */ 834 the same bracket. 835 the stack. */ 836 the start hasn't passed this character yet. */ 837 the subject. */ 838 there were too many extractions, set the return code to zero. In the case 839 this level is identical to the lookahead case. */ 840 this makes a huge difference to execution time when there aren't many brackets 841 those back references that we can. In this case there need not be overflow 842 time taken, but character matching *is* what this is all about... */ 843 to save all the potential data. There may be up to 99 such values, which 844 to that for character classes, but repeated for efficiency. Then obey 845 two branches. If the condition is false, skipping the first branch takes us 846 typedef struct eptrblock 847 unless PCRE_CASELESS was given or the casing state changes within the regex. 848 unlimited repeats that aren't going to match. We don't know what the state of 849 unsigned long int ims = 0; 850 unsigned long int ims; 851 unsigned long int original_ims = ims; /* Save for resetting on ')' */ 852 up quickly if there are fewer than the minimum number of characters left in 853 using_temporary_offsets = TRUE; 854 values of the final offsets, in case they were set by a previous iteration of 855 we just need to set up the whole thing as substring 0 before returning. If 856 where we had to get some local store to hold offsets for backreferences, copy 857 while (!anchored && 858 while (*ecode == OP_ALT) 859 while (*ecode == OP_ALT); 860 while (*next == OP_ALT); 861 while (--iptr >= iend) 862 while (eptr >= pp) 863 while (iptr < iend) 864 while (length-- > 0) 865 while (p < end_subject) 866 while (start_match < end_subject && 867 while (start_match < end_subject && *start_match != first_char) 868 while (start_match < end_subject && start_match[-1] != '\n') 869 while (start_match < end_subject) 870 { 871 {0, 0, 0, 0, 1, 1}; 872 {0, 0, 1, 1, 0, 0}; 873 } /* End of main loop */ 874 } 875
