1 /* regexp.h 2 * 3 * Copyright (C) 1993, 1994, 1996, 1997, 1999, 2000, 2001, 2003, 4 * 2005, 2006, 2007, 2008 by Larry Wall and others 5 * 6 * You may distribute under the terms of either the GNU General Public 7 * License or the Artistic License, as specified in the README file. 8 * 9 */ 10 11 /* 12 * Definitions etc. for regexp(3) routines. 13 * 14 * Caveat: this is V8 regexp(3) [actually, a reimplementation thereof], 15 * not the System V one. 16 */ 17 #ifndef PLUGGABLE_RE_EXTENSION 18 /* we don't want to include this stuff if we are inside of 19 an external regex engine based on the core one - like re 'debug'*/ 20 21 #include "utf8.h" 22 23 typedef SSize_t regnode_offset; 24 25 struct regnode { 26 U8 flags; 27 U8 type; 28 U16 next_off; 29 }; 30 31 typedef struct regnode regnode; 32 33 struct reg_substr_data; 34 35 struct reg_data; 36 37 struct regexp_engine; 38 struct regexp; 39 40 struct reg_substr_datum { 41 SSize_t min_offset; /* min pos (in chars) that substr must appear */ 42 SSize_t max_offset /* max pos (in chars) that substr must appear */; 43 SV *substr; /* non-utf8 variant */ 44 SV *utf8_substr; /* utf8 variant */ 45 SSize_t end_shift; /* how many fixed chars must end the string */ 46 }; 47 struct reg_substr_data { 48 U8 check_ix; /* index into data[] of check substr */ 49 struct reg_substr_datum data[3]; /* Actual array */ 50 }; 51 52 #ifdef PERL_ANY_COW 53 #define SV_SAVED_COPY SV *saved_copy; /* If non-NULL, SV which is COW from original */ 54 #else 55 #define SV_SAVED_COPY 56 #endif 57 58 /* offsets within a string of a particular /(.)/ capture */ 59 60 typedef struct regexp_paren_pair { 61 SSize_t start; 62 SSize_t end; 63 /* 'start_tmp' records a new opening position before the matching end 64 * has been found, so that the old start and end values are still 65 * valid, e.g. 66 * "abc" =~ /(.(?{print "[$1]"}))+/ 67 *outputs [][a][b] 68 * This field is not part of the API. */ 69 SSize_t start_tmp; 70 } regexp_paren_pair; 71 72 #if defined(PERL_IN_REGCOMP_C) || defined(PERL_IN_UTF8_C) 73 #define _invlist_union(a, b, output) _invlist_union_maybe_complement_2nd(a, b, FALSE, output) 74 #define _invlist_intersection(a, b, output) _invlist_intersection_maybe_complement_2nd(a, b, FALSE, output) 75 76 /* Subtracting b from a leaves in a everything that was there that isn't in b, 77 * that is the intersection of a with b's complement */ 78 #define _invlist_subtract(a, b, output) _invlist_intersection_maybe_complement_2nd(a, b, TRUE, output) 79 #endif 80 81 /* record the position of a (?{...}) within a pattern */ 82 83 struct reg_code_block { 84 STRLEN start; 85 STRLEN end; 86 OP *block; 87 REGEXP *src_regex; 88 }; 89 90 /* array of reg_code_block's plus header info */ 91 92 struct reg_code_blocks { 93 int refcnt; /* we may be pointed to from a regex and from the savestack */ 94 int count; /* how many code blocks */ 95 struct reg_code_block *cb; /* array of reg_code_block's */ 96 }; 97 98 99 /* 100 The regexp/REGEXP struct, see L<perlreapi> for further documentation 101 on the individual fields. The struct is ordered so that the most 102 commonly used fields are placed at the start. 103 104 Any patch that adds items to this struct will need to include 105 changes to F<sv.c> (C<Perl_re_dup()>) and F<regcomp.c> 106 (C<pregfree()>). This involves freeing or cloning items in the 107 regexp's data array based on the data item's type. 108 */ 109 110 typedef struct regexp { 111 _XPV_HEAD; 112 const struct regexp_engine* engine; /* what engine created this regexp? */ 113 REGEXP *mother_re; /* what re is this a lightweight copy of? */ 114 HV *paren_names; /* Optional hash of paren names */ 115 116 /*---------------------------------------------------------------------- 117 * Information about the match that the perl core uses to manage things 118 */ 119 120 U32 extflags; /* Flags used both externally and internally */ 121 SSize_t minlen; /* minimum possible number of chars in string to match */ 122 SSize_t minlenret; /* mininum possible number of chars in $& */ 123 STRLEN gofs; /* chars left of pos that we search from */ 124 /* substring data about strings that must appear in 125 * the final match, used for optimisations */ 126 struct reg_substr_data *substrs; 127 U32 nparens; /* number of capture buffers */ 128 129 /* private engine specific data */ 130 131 U32 intflags; /* Engine Specific Internal flags */ 132 void *pprivate; /* Data private to the regex engine which 133 * created this object. */ 134 135 /*---------------------------------------------------------------------- 136 * Data about the last/current match. These are modified during matching 137 */ 138 139 U32 lastparen; /* highest close paren matched ($+) */ 140 U32 lastcloseparen; /* last close paren matched ($^N) */ 141 regexp_paren_pair *offs; /* Array of offsets for (@-) and (@+) */ 142 char **recurse_locinput; /* used to detect infinite recursion, XXX: move to internal */ 143 144 /*---------------------------------------------------------------------- */ 145 146 char *subbeg; /* saved or original string so \digit works forever. */ 147 SV_SAVED_COPY /* If non-NULL, SV which is COW from original */ 148 SSize_t sublen; /* Length of string pointed by subbeg */ 149 SSize_t suboffset; /* byte offset of subbeg from logical start of str */ 150 SSize_t subcoffset; /* suboffset equiv, but in chars (for @-/@+) */ 151 152 /* Information about the match that isn't often used */ 153 154 SSize_t maxlen; /* minimum possible number of chars in string to match */ 155 156 /*---------------------------------------------------------------------- */ 157 158 /* offset from wrapped to the start of precomp */ 159 PERL_BITFIELD32 pre_prefix:4; 160 161 /* original flags used to compile the pattern, may differ from 162 * extflags in various ways */ 163 PERL_BITFIELD32 compflags:9; 164 165 /*---------------------------------------------------------------------- */ 166 167 CV *qr_anoncv; /* the anon sub wrapped round qr/(?{..})/ */ 168 } regexp; 169 170 171 #define RXp_PAREN_NAMES(rx) ((rx)->paren_names) 172 173 /* used for high speed searches */ 174 typedef struct re_scream_pos_data_s 175 { 176 char **scream_olds; /* match pos */ 177 SSize_t *scream_pos; /* Internal iterator of scream. */ 178 } re_scream_pos_data; 179 180 /* regexp_engine structure. This is the dispatch table for regexes. 181 * Any regex engine implementation must be able to build one of these. 182 */ 183 typedef struct regexp_engine { 184 REGEXP* (*comp) (pTHX_ SV * const pattern, U32 flags); 185 I32 (*exec) (pTHX_ REGEXP * const rx, char* stringarg, char* strend, 186 char* strbeg, SSize_t minend, SV* sv, 187 void* data, U32 flags); 188 char* (*intuit) (pTHX_ 189 REGEXP * const rx, 190 SV *sv, 191 const char * const strbeg, 192 char *strpos, 193 char *strend, 194 const U32 flags, 195 re_scream_pos_data *data); 196 SV* (*checkstr) (pTHX_ REGEXP * const rx); 197 void (*rxfree) (pTHX_ REGEXP * const rx); 198 void (*numbered_buff_FETCH) (pTHX_ REGEXP * const rx, const I32 paren, 199 SV * const sv); 200 void (*numbered_buff_STORE) (pTHX_ REGEXP * const rx, const I32 paren, 201 SV const * const value); 202 I32 (*numbered_buff_LENGTH) (pTHX_ REGEXP * const rx, const SV * const sv, 203 const I32 paren); 204 SV* (*named_buff) (pTHX_ REGEXP * const rx, SV * const key, 205 SV * const value, const U32 flags); 206 SV* (*named_buff_iter) (pTHX_ REGEXP * const rx, const SV * const lastkey, 207 const U32 flags); 208 SV* (*qr_package)(pTHX_ REGEXP * const rx); 209 #ifdef USE_ITHREADS 210 void* (*dupe) (pTHX_ REGEXP * const rx, CLONE_PARAMS *param); 211 #endif 212 REGEXP* (*op_comp) (pTHX_ SV ** const patternp, int pat_count, 213 OP *expr, const struct regexp_engine* eng, 214 REGEXP *old_re, 215 bool *is_bare_re, U32 orig_rx_flags, U32 pm_flags); 216 } regexp_engine; 217 218 /* 219 These are passed to the numbered capture variable callbacks as the 220 paren name. >= 1 is reserved for actual numbered captures, i.e. $1, 221 $2 etc. 222 */ 223 #define RX_BUFF_IDX_CARET_PREMATCH -5 /* ${^PREMATCH} */ 224 #define RX_BUFF_IDX_CARET_POSTMATCH -4 /* ${^POSTMATCH} */ 225 #define RX_BUFF_IDX_CARET_FULLMATCH -3 /* ${^MATCH} */ 226 #define RX_BUFF_IDX_PREMATCH -2 /* $` */ 227 #define RX_BUFF_IDX_POSTMATCH -1 /* $' */ 228 #define RX_BUFF_IDX_FULLMATCH 0 /* $& */ 229 230 /* 231 Flags that are passed to the named_buff and named_buff_iter 232 callbacks above. Those routines are called from universal.c via the 233 Tie::Hash::NamedCapture interface for %+ and %- and the re:: 234 functions in the same file. 235 */ 236 237 /* The Tie::Hash::NamedCapture operation this is part of, if any */ 238 #define RXapif_FETCH 0x0001 239 #define RXapif_STORE 0x0002 240 #define RXapif_DELETE 0x0004 241 #define RXapif_CLEAR 0x0008 242 #define RXapif_EXISTS 0x0010 243 #define RXapif_SCALAR 0x0020 244 #define RXapif_FIRSTKEY 0x0040 245 #define RXapif_NEXTKEY 0x0080 246 247 /* Whether %+ or %- is being operated on */ 248 #define RXapif_ONE 0x0100 /* %+ */ 249 #define RXapif_ALL 0x0200 /* %- */ 250 251 /* Whether this is being called from a re:: function */ 252 #define RXapif_REGNAME 0x0400 253 #define RXapif_REGNAMES 0x0800 254 #define RXapif_REGNAMES_COUNT 0x1000 255 256 /* 257 =head1 REGEXP Functions 258 259 =for apidoc Am|REGEXP *|SvRX|SV *sv 260 261 Convenience macro to get the REGEXP from a SV. This is approximately 262 equivalent to the following snippet: 263 264 if (SvMAGICAL(sv)) 265 mg_get(sv); 266 if (SvROK(sv)) 267 sv = MUTABLE_SV(SvRV(sv)); 268 if (SvTYPE(sv) == SVt_REGEXP) 269 return (REGEXP*) sv; 270 271 C<NULL> will be returned if a REGEXP* is not found. 272 273 =for apidoc Am|bool|SvRXOK|SV* sv 274 275 Returns a boolean indicating whether the SV (or the one it references) 276 is a REGEXP. 277 278 If you want to do something with the REGEXP* later use SvRX instead 279 and check for NULL. 280 281 =cut 282 */ 283 284 #define SvRX(sv) (Perl_get_re_arg(aTHX_ sv)) 285 #define SvRXOK(sv) cBOOL(Perl_get_re_arg(aTHX_ sv)) 286 287 288 /* Flags stored in regexp->extflags 289 * These are used by code external to the regexp engine 290 * 291 * Note that the flags whose names start with RXf_PMf_ are defined in 292 * op_reg_common.h, being copied from the parallel flags of op_pmflags 293 * 294 * NOTE: if you modify any RXf flags you should run regen.pl or 295 * regen/regcomp.pl so that regnodes.h is updated with the changes. 296 * 297 */ 298 299 #include "op_reg_common.h" 300 301 #define RXf_PMf_STD_PMMOD (RXf_PMf_MULTILINE|RXf_PMf_SINGLELINE|RXf_PMf_FOLD|RXf_PMf_EXTENDED|RXf_PMf_EXTENDED_MORE|RXf_PMf_NOCAPTURE) 302 303 #define CASE_STD_PMMOD_FLAGS_PARSE_SET(pmfl, x_count) \ 304 case IGNORE_PAT_MOD: *(pmfl) |= RXf_PMf_FOLD; break; \ 305 case MULTILINE_PAT_MOD: *(pmfl) |= RXf_PMf_MULTILINE; break; \ 306 case SINGLE_PAT_MOD: *(pmfl) |= RXf_PMf_SINGLELINE; break; \ 307 case XTENDED_PAT_MOD: if (x_count == 0) { \ 308 *(pmfl) |= RXf_PMf_EXTENDED; \ 309 *(pmfl) &= ~RXf_PMf_EXTENDED_MORE; \ 310 } \ 311 else { \ 312 *(pmfl) |= RXf_PMf_EXTENDED \ 313 |RXf_PMf_EXTENDED_MORE; \ 314 } \ 315 (x_count)++; break; \ 316 case NOCAPTURE_PAT_MOD: *(pmfl) |= RXf_PMf_NOCAPTURE; break; 317 318 /* Note, includes charset ones, assumes 0 is the default for them */ 319 #define STD_PMMOD_FLAGS_CLEAR(pmfl) \ 320 *(pmfl) &= ~(RXf_PMf_FOLD|RXf_PMf_MULTILINE|RXf_PMf_SINGLELINE|RXf_PMf_EXTENDED|RXf_PMf_EXTENDED_MORE|RXf_PMf_CHARSET|RXf_PMf_NOCAPTURE) 321 322 /* chars and strings used as regex pattern modifiers 323 * Singular is a 'c'har, plural is a "string" 324 * 325 * NOTE, KEEPCOPY was originally 'k', but was changed to 'p' for preserve 326 * for compatibility reasons with Regexp::Common which highjacked (?k:...) 327 * for its own uses. So 'k' is out as well. 328 */ 329 #define DEFAULT_PAT_MOD '^' /* Short for all the default modifiers */ 330 #define EXEC_PAT_MOD 'e' 331 #define KEEPCOPY_PAT_MOD 'p' 332 #define NOCAPTURE_PAT_MOD 'n' 333 #define ONCE_PAT_MOD 'o' 334 #define GLOBAL_PAT_MOD 'g' 335 #define CONTINUE_PAT_MOD 'c' 336 #define MULTILINE_PAT_MOD 'm' 337 #define SINGLE_PAT_MOD 's' 338 #define IGNORE_PAT_MOD 'i' 339 #define XTENDED_PAT_MOD 'x' 340 #define NONDESTRUCT_PAT_MOD 'r' 341 #define LOCALE_PAT_MOD 'l' 342 #define UNICODE_PAT_MOD 'u' 343 #define DEPENDS_PAT_MOD 'd' 344 #define ASCII_RESTRICT_PAT_MOD 'a' 345 346 #define ONCE_PAT_MODS "o" 347 #define KEEPCOPY_PAT_MODS "p" 348 #define NOCAPTURE_PAT_MODS "n" 349 #define EXEC_PAT_MODS "e" 350 #define LOOP_PAT_MODS "gc" 351 #define NONDESTRUCT_PAT_MODS "r" 352 #define LOCALE_PAT_MODS "l" 353 #define UNICODE_PAT_MODS "u" 354 #define DEPENDS_PAT_MODS "d" 355 #define ASCII_RESTRICT_PAT_MODS "a" 356 #define ASCII_MORE_RESTRICT_PAT_MODS "aa" 357 358 /* This string is expected by regcomp.c to be ordered so that the first 359 * character is the flag in bit RXf_PMf_STD_PMMOD_SHIFT of extflags; the next 360 * character is bit +1, etc. */ 361 #define STD_PAT_MODS "msixxn" 362 363 #define CHARSET_PAT_MODS ASCII_RESTRICT_PAT_MODS DEPENDS_PAT_MODS LOCALE_PAT_MODS UNICODE_PAT_MODS 364 365 /* This string is expected by XS_re_regexp_pattern() in universal.c to be ordered 366 * so that the first character is the flag in bit RXf_PMf_STD_PMMOD_SHIFT of 367 * extflags; the next character is in bit +1, etc. */ 368 #define INT_PAT_MODS STD_PAT_MODS KEEPCOPY_PAT_MODS 369 370 #define EXT_PAT_MODS ONCE_PAT_MODS KEEPCOPY_PAT_MODS NOCAPTURE_PAT_MODS 371 #define QR_PAT_MODS STD_PAT_MODS EXT_PAT_MODS CHARSET_PAT_MODS 372 #define M_PAT_MODS QR_PAT_MODS LOOP_PAT_MODS 373 #define S_PAT_MODS M_PAT_MODS EXEC_PAT_MODS NONDESTRUCT_PAT_MODS 374 375 /* 376 * NOTE: if you modify any RXf flags you should run regen.pl or 377 * regen/regcomp.pl so that regnodes.h is updated with the changes. 378 * 379 */ 380 381 /* 382 Set in Perl_pmruntime for a split. Will be used by regex engines to 383 check whether they should set RXf_SKIPWHITE 384 */ 385 #define RXf_SPLIT RXf_PMf_SPLIT 386 387 /* Currently the regex flags occupy a single 32-bit word. Not all bits are 388 * currently used. The lower bits are shared with their corresponding PMf flag 389 * bits, up to but not including _RXf_PMf_SHIFT_NEXT. The unused bits 390 * immediately follow; finally the used RXf-only (unshared) bits, so that the 391 * highest bit in the word is used. This gathers all the unused bits as a pool 392 * in the middle, like so: 11111111111111110000001111111111 393 * where the '1's represent used bits, and the '0's unused. This design allows 394 * us to allocate off one end of the pool if we need to add a shared bit, and 395 * off the other end if we need a non-shared bit, without disturbing the other 396 * bits. This maximizes the likelihood of being able to change things without 397 * breaking binary compatibility. 398 * 399 * To add shared bits, do so in op_reg_common.h. This should change 400 * _RXf_PMf_SHIFT_NEXT so that things won't compile. Then come to regexp.h and 401 * op.h and adjust the constant adders in the definitions of RXf_BASE_SHIFT and 402 * Pmf_BASE_SHIFT down by the number of shared bits you added. That's it. 403 * Things should be binary compatible. But if either of these gets to having 404 * to subtract rather than add, leave at 0 and instead adjust all the entries 405 * that are in terms of it. But if the first one of those is already 406 * RXf_BASE_SHIFT+0, there are no bits left, and a redesign is in order. 407 * 408 * To remove unshared bits, just delete its entry. If you're where breaking 409 * binary compatibility is ok to do, you might want to adjust things to move 410 * the newly opened space so that it gets absorbed into the common pool. 411 * 412 * To add unshared bits, first use up any gaps in the middle. Otherwise, 413 * allocate off the low end until you get to RXf_BASE_SHIFT+0. If that isn't 414 * enough, move RXf_BASE_SHIFT down (if possible) and add the new bit at the 415 * other end instead; this preserves binary compatibility. 416 * 417 * For the regexp bits, PL_reg_extflags_name[] in regnodes.h has a comment 418 * giving which bits are used/unused */ 419 420 #define RXf_BASE_SHIFT (_RXf_PMf_SHIFT_NEXT + 2) 421 422 /* What we have seen */ 423 #define RXf_NO_INPLACE_SUBST (1U<<(RXf_BASE_SHIFT+2)) 424 #define RXf_EVAL_SEEN (1U<<(RXf_BASE_SHIFT+3)) 425 426 /* Special */ 427 #define RXf_UNBOUNDED_QUANTIFIER_SEEN (1U<<(RXf_BASE_SHIFT+4)) 428 #define RXf_CHECK_ALL (1U<<(RXf_BASE_SHIFT+5)) 429 430 /* UTF8 related */ 431 #define RXf_MATCH_UTF8 (1U<<(RXf_BASE_SHIFT+6)) /* $1 etc are utf8 */ 432 433 /* Intuit related */ 434 #define RXf_USE_INTUIT_NOML (1U<<(RXf_BASE_SHIFT+7)) 435 #define RXf_USE_INTUIT_ML (1U<<(RXf_BASE_SHIFT+8)) 436 #define RXf_INTUIT_TAIL (1U<<(RXf_BASE_SHIFT+9)) 437 #define RXf_USE_INTUIT (RXf_USE_INTUIT_NOML|RXf_USE_INTUIT_ML) 438 439 /* Do we have some sort of anchor? */ 440 #define RXf_IS_ANCHORED (1U<<(RXf_BASE_SHIFT+10)) 441 442 /* Copy and tainted info */ 443 #define RXf_COPY_DONE (1U<<(RXf_BASE_SHIFT+11)) 444 445 /* post-execution: $1 et al are tainted */ 446 #define RXf_TAINTED_SEEN (1U<<(RXf_BASE_SHIFT+12)) 447 /* this pattern was tainted during compilation */ 448 #define RXf_TAINTED (1U<<(RXf_BASE_SHIFT+13)) 449 450 /* Flags indicating special patterns */ 451 #define RXf_START_ONLY (1U<<(RXf_BASE_SHIFT+14)) /* Pattern is /^/ */ 452 #define RXf_SKIPWHITE (1U<<(RXf_BASE_SHIFT+15)) /* Pattern is for a */ 453 /* split " " */ 454 #define RXf_WHITE (1U<<(RXf_BASE_SHIFT+16)) /* Pattern is /\s+/ */ 455 #define RXf_NULL (1U<<(RXf_BASE_SHIFT+17)) /* Pattern is // */ 456 457 /* See comments at the beginning of these defines about adding bits. The 458 * highest bit position should be used, so that if RXf_BASE_SHIFT gets 459 * increased, the #error below will be triggered so that you will be reminded 460 * to adjust things at the other end to keep the bit positions unchanged */ 461 #if RXf_BASE_SHIFT+17 > 31 462 # error Too many RXf_PMf bits used. See comments at beginning of these for what to do 463 #endif 464 465 /* 466 * NOTE: if you modify any RXf flags you should run regen.pl or 467 * regen/regcomp.pl so that regnodes.h is updated with the changes. 468 * 469 */ 470 471 #ifdef NO_TAINT_SUPPORT 472 # define RX_ISTAINTED(rx_sv) 0 473 # define RXp_ISTAINTED(prog) 0 474 # define RX_TAINT_on(rx_sv) NOOP 475 # define RXp_MATCH_TAINTED(prog) 0 476 # define RX_MATCH_TAINTED(rx_sv) 0 477 # define RXp_MATCH_TAINTED_on(prog) NOOP 478 # define RX_MATCH_TAINTED_on(rx_sv) NOOP 479 # define RXp_MATCH_TAINTED_off(prog) NOOP 480 # define RX_MATCH_TAINTED_off(rx_sv) NOOP 481 #else 482 # define RX_ISTAINTED(rx_sv) (RX_EXTFLAGS(rx_sv) & RXf_TAINTED) 483 # define RXp_ISTAINTED(prog) (RXp_EXTFLAGS(prog) & RXf_TAINTED) 484 # define RX_TAINT_on(rx_sv) (RX_EXTFLAGS(rx_sv) |= RXf_TAINTED) 485 # define RXp_MATCH_TAINTED(prog) (RXp_EXTFLAGS(prog) & RXf_TAINTED_SEEN) 486 # define RX_MATCH_TAINTED(rx_sv) (RX_EXTFLAGS(rx_sv) & RXf_TAINTED_SEEN) 487 # define RXp_MATCH_TAINTED_on(prog) (RXp_EXTFLAGS(prog) |= RXf_TAINTED_SEEN) 488 # define RX_MATCH_TAINTED_on(rx_sv) (RX_EXTFLAGS(rx_sv) |= RXf_TAINTED_SEEN) 489 # define RXp_MATCH_TAINTED_off(prog) (RXp_EXTFLAGS(prog) &= ~RXf_TAINTED_SEEN) 490 # define RX_MATCH_TAINTED_off(rx_sv) (RX_EXTFLAGS(rx_sv) &= ~RXf_TAINTED_SEEN) 491 #endif 492 493 #define RXp_HAS_CUTGROUP(prog) ((prog)->intflags & PREGf_CUTGROUP_SEEN) 494 495 #define RX_MATCH_TAINTED_set(rx_sv, t) ((t) \ 496 ? RX_MATCH_TAINTED_on(rx_sv) \ 497 : RX_MATCH_TAINTED_off(rx_sv)) 498 499 #define RXp_MATCH_COPIED(prog) (RXp_EXTFLAGS(prog) & RXf_COPY_DONE) 500 #define RX_MATCH_COPIED(rx_sv) (RX_EXTFLAGS(rx_sv) & RXf_COPY_DONE) 501 #define RXp_MATCH_COPIED_on(prog) (RXp_EXTFLAGS(prog) |= RXf_COPY_DONE) 502 #define RX_MATCH_COPIED_on(rx_sv) (RX_EXTFLAGS(rx_sv) |= RXf_COPY_DONE) 503 #define RXp_MATCH_COPIED_off(prog) (RXp_EXTFLAGS(prog) &= ~RXf_COPY_DONE) 504 #define RX_MATCH_COPIED_off(rx_sv) (RX_EXTFLAGS(rx_sv) &= ~RXf_COPY_DONE) 505 #define RX_MATCH_COPIED_set(rx_sv,t) ((t) \ 506 ? RX_MATCH_COPIED_on(rx_sv) \ 507 : RX_MATCH_COPIED_off(rx_sv)) 508 509 #define RXp_EXTFLAGS(rx) ((rx)->extflags) 510 #define RXp_COMPFLAGS(rx) ((rx)->compflags) 511 512 /* For source compatibility. We used to store these explicitly. */ 513 #define RX_PRECOMP(rx_sv) (RX_WRAPPED(rx_sv) \ 514 + ReANY(rx_sv)->pre_prefix) 515 #define RX_PRECOMP_const(rx_sv) (RX_WRAPPED_const(rx_sv) \ 516 + ReANY(rx_sv)->pre_prefix) 517 /* FIXME? Are we hardcoding too much here and constraining plugin extension 518 writers? Specifically, the value 1 assumes that the wrapped version always 519 has exactly one character at the end, a ')'. Will that always be true? */ 520 #define RX_PRELEN(rx_sv) (RX_WRAPLEN(rx_sv) \ 521 - ReANY(rx_sv)->pre_prefix - 1) 522 523 #define RX_WRAPPED(rx_sv) SvPVX(rx_sv) 524 #define RX_WRAPPED_const(rx_sv) SvPVX_const(rx_sv) 525 #define RX_WRAPLEN(rx_sv) SvCUR(rx_sv) 526 #define RX_CHECK_SUBSTR(rx_sv) (ReANY(rx_sv)->check_substr) 527 #define RX_REFCNT(rx_sv) SvREFCNT(rx_sv) 528 #define RX_EXTFLAGS(rx_sv) RXp_EXTFLAGS(ReANY(rx_sv)) 529 #define RX_COMPFLAGS(rx_sv) RXp_COMPFLAGS(ReANY(rx_sv)) 530 #define RXp_ENGINE(prog) ((prog)->engine) 531 #define RX_ENGINE(rx_sv) (RXp_ENGINE(ReANY(rx_sv))) 532 #define RXp_SUBBEG(prog) (prog->subbeg) 533 #define RX_SUBBEG(rx_sv) (RXp_SUBBEG(ReANY(rx_sv))) 534 #define RXp_SUBOFFSET(prog) (prog->suboffset) 535 #define RX_SUBOFFSET(rx_sv) (RXp_SUBOFFSET(ReANY(rx_sv))) 536 #define RX_SUBCOFFSET(rx_sv) (ReANY(rx_sv)->subcoffset) 537 #define RXp_OFFS(prog) (prog->offs) 538 #define RX_OFFS(rx_sv) (RXp_OFFS(ReANY(rx_sv))) 539 #define RXp_NPARENS(prog) (prog->nparens) 540 #define RX_NPARENS(rx_sv) (RXp_NPARENS(ReANY(rx_sv))) 541 #define RX_SUBLEN(rx_sv) (ReANY(rx_sv)->sublen) 542 #define RXp_MINLEN(prog) (prog->minlen) 543 #define RX_MINLEN(rx_sv) (RXp_MINLEN(ReANY(rx_sv))) 544 #define RXp_MINLENRET(prog) (prog->minlenret) 545 #define RX_MINLENRET(rx_sv) (RXp_MINLENRET(ReANY(rx_sv))) 546 #define RXp_GOFS(prog) (prog->gofs) 547 #define RX_GOFS(rx_sv) (RXp_GOFS(ReANY(rx_sv))) 548 #define RX_LASTPAREN(rx_sv) (ReANY(rx_sv)->lastparen) 549 #define RX_LASTCLOSEPAREN(rx_sv) (ReANY(rx_sv)->lastcloseparen) 550 #define RXp_SAVED_COPY(prog) (prog->saved_copy) 551 #define RX_SAVED_COPY(rx_sv) (RXp_SAVED_COPY(ReANY(rx_sv))) 552 /* last match was zero-length */ 553 #define RXp_ZERO_LEN(prog) \ 554 (RXp_OFFS(prog)[0].start + (SSize_t)RXp_GOFS(prog) \ 555 == RXp_OFFS(prog)[0].end) 556 #define RX_ZERO_LEN(rx_sv) (RXp_ZERO_LEN(ReANY(rx_sv))) 557 558 #endif /* PLUGGABLE_RE_EXTENSION */ 559 560 /* Stuff that needs to be included in the pluggable extension goes below here */ 561 562 #ifdef PERL_ANY_COW 563 #define RXp_MATCH_COPY_FREE(prog) \ 564 STMT_START {if (RXp_SAVED_COPY(prog)) { \ 565 SV_CHECK_THINKFIRST_COW_DROP(RXp_SAVED_COPY(prog)); \ 566 } \ 567 if (RXp_MATCH_COPIED(prog)) { \ 568 Safefree(RXp_SUBBEG(prog)); \ 569 RXp_MATCH_COPIED_off(prog); \ 570 }} STMT_END 571 #else 572 #define RXp_MATCH_COPY_FREE(prog) \ 573 STMT_START {if (RXp_MATCH_COPIED(prog)) { \ 574 Safefree(RXp_SUBBEG(prog)); \ 575 RXp_MATCH_COPIED_off(prog); \ 576 }} STMT_END 577 #endif 578 #define RX_MATCH_COPY_FREE(rx_sv) RXp_MATCH_COPY_FREE(ReANY(rx_sv)) 579 580 #define RXp_MATCH_UTF8(prog) (RXp_EXTFLAGS(prog) & RXf_MATCH_UTF8) 581 #define RX_MATCH_UTF8(rx_sv) (RX_EXTFLAGS(rx_sv) & RXf_MATCH_UTF8) 582 #define RXp_MATCH_UTF8_on(prog) (RXp_EXTFLAGS(prog) |= RXf_MATCH_UTF8) 583 #define RX_MATCH_UTF8_on(rx_sv) (RXp_MATCH_UTF8_on(ReANY(rx_sv))) 584 #define RXp_MATCH_UTF8_off(prog) (RXp_EXTFLAGS(prog) &= ~RXf_MATCH_UTF8) 585 #define RX_MATCH_UTF8_off(rx_sv) (RXp_MATCH_UTF8_off(ReANY(rx_sv)) 586 #define RXp_MATCH_UTF8_set(prog, t) ((t) \ 587 ? RXp_MATCH_UTF8_on(prog) \ 588 : RXp_MATCH_UTF8_off(prog)) 589 #define RX_MATCH_UTF8_set(rx_sv, t) (RXp_MATCH_UTF8_set(ReANY(rx_sv), t)) 590 591 /* Whether the pattern stored at RX_WRAPPED is in UTF-8 */ 592 #define RX_UTF8(rx_sv) SvUTF8(rx_sv) 593 594 595 /* bits in flags arg of Perl_regexec_flags() */ 596 597 #define REXEC_COPY_STR 0x01 /* Need to copy the string for captures. */ 598 #define REXEC_CHECKED 0x02 /* re_intuit_start() already called. */ 599 #define REXEC_SCREAM 0x04 /* currently unused. */ 600 #define REXEC_IGNOREPOS 0x08 /* use stringarg, not pos(), for \G match */ 601 #define REXEC_NOT_FIRST 0x10 /* This is another iteration of //g: 602 no need to copy string again */ 603 604 /* under REXEC_COPY_STR, it's ok for the 605 engine (modulo PL_sawamperand etc) 606 to skip copying: ... */ 607 #define REXEC_COPY_SKIP_PRE 0x20 /* ...the $` part of the string, or */ 608 #define REXEC_COPY_SKIP_POST 0x40 /* ...the $' part of the string */ 609 #define REXEC_FAIL_ON_UNDERFLOW 0x80 /* fail the match if $& would start before 610 the start pos (so s/.\G// would fail 611 on second iteration */ 612 613 #if defined(__GNUC__) && !defined(PERL_GCC_BRACE_GROUPS_FORBIDDEN) 614 # define ReREFCNT_inc(re) \ 615 ({ \ 616 /* This is here to generate a casting warning if incorrect. */ \ 617 REGEXP *const _rerefcnt_inc = (re); \ 618 assert(SvTYPE(_rerefcnt_inc) == SVt_REGEXP); \ 619 SvREFCNT_inc(_rerefcnt_inc); \ 620 _rerefcnt_inc; \ 621 }) 622 # define ReREFCNT_dec(re) \ 623 ({ \ 624 /* This is here to generate a casting warning if incorrect. */ \ 625 REGEXP *const _rerefcnt_dec = (re); \ 626 SvREFCNT_dec(_rerefcnt_dec); \ 627 }) 628 #else 629 # define ReREFCNT_dec(re) SvREFCNT_dec(re) 630 # define ReREFCNT_inc(re) ((REGEXP *) SvREFCNT_inc(re)) 631 #endif 632 #define ReANY(re) S_ReANY((const REGEXP *)(re)) 633 634 /* FIXME for plugins. */ 635 636 #define FBMcf_TAIL_DOLLAR 1 637 #define FBMcf_TAIL_DOLLARM 2 638 #define FBMcf_TAIL_Z 4 639 #define FBMcf_TAIL_z 8 640 #define FBMcf_TAIL (FBMcf_TAIL_DOLLAR|FBMcf_TAIL_DOLLARM|FBMcf_TAIL_Z|FBMcf_TAIL_z) 641 642 #define FBMrf_MULTILINE 1 643 644 struct regmatch_state; 645 struct regmatch_slab; 646 647 /* like regmatch_info_aux, but contains extra fields only needed if the 648 * pattern contains (?{}). If used, is snuck into the second slot in the 649 * regmatch_state stack at the start of execution */ 650 651 typedef struct { 652 regexp *rex; 653 PMOP *curpm; /* saved PL_curpm */ 654 #ifdef PERL_ANY_COW 655 SV *saved_copy; /* saved saved_copy field from rex */ 656 #endif 657 char *subbeg; /* saved subbeg field from rex */ 658 STRLEN sublen; /* saved sublen field from rex */ 659 STRLEN suboffset; /* saved suboffset field from rex */ 660 STRLEN subcoffset; /* saved subcoffset field from rex */ 661 SV *sv; /* $_ during (?{}) */ 662 MAGIC *pos_magic; /* pos() magic attached to $_ */ 663 SSize_t pos; /* the original value of pos() in pos_magic */ 664 U8 pos_flags; /* flags to be restored; currently only MGf_BYTES*/ 665 } regmatch_info_aux_eval; 666 667 668 /* fields that logically live in regmatch_info, but which need cleaning 669 * up on croak(), and so are instead are snuck into the first slot in 670 * the regmatch_state stack at the start of execution */ 671 672 typedef struct { 673 regmatch_info_aux_eval *info_aux_eval; 674 struct regmatch_state *old_regmatch_state; /* saved PL_regmatch_state */ 675 struct regmatch_slab *old_regmatch_slab; /* saved PL_regmatch_slab */ 676 char *poscache; /* S-L cache of fail positions of WHILEMs */ 677 } regmatch_info_aux; 678 679 680 /* some basic information about the current match that is created by 681 * Perl_regexec_flags and then passed to regtry(), regmatch() etc. 682 * It is allocated as a local var on the stack, so nothing should be 683 * stored in it that needs preserving or clearing up on croak(). 684 * For that, see the aux_info and aux_info_eval members of the 685 * regmatch_state union. */ 686 687 typedef struct { 688 REGEXP *prog; /* the regex being executed */ 689 const char * strbeg; /* real start of string */ 690 char *strend; /* one byte beyond last char of match string */ 691 char *till; /* matches shorter than this fail (see minlen arg) */ 692 SV *sv; /* the SV string currently being matched */ 693 char *ganch; /* position of \G anchor */ 694 char *cutpoint; /* (*COMMIT) position (if any) */ 695 regmatch_info_aux *info_aux; /* extra fields that need cleanup */ 696 regmatch_info_aux_eval *info_aux_eval; /* extra saved state for (?{}) */ 697 I32 poscache_maxiter; /* how many whilems todo before S-L cache kicks in */ 698 I32 poscache_iter; /* current countdown from _maxiter to zero */ 699 STRLEN poscache_size; /* size of regmatch_info_aux.poscache */ 700 bool intuit; /* re_intuit_start() is the top-level caller */ 701 bool is_utf8_pat; /* regex is utf8 */ 702 bool is_utf8_target; /* string being matched is utf8 */ 703 bool warned; /* we have issued a recursion warning; no need for more */ 704 } regmatch_info; 705 706 707 /* structures for holding and saving the state maintained by regmatch() */ 708 709 #ifndef MAX_RECURSE_EVAL_NOCHANGE_DEPTH 710 #define MAX_RECURSE_EVAL_NOCHANGE_DEPTH 10 711 #endif 712 713 typedef I32 CHECKPOINT; 714 715 typedef struct regmatch_state { 716 int resume_state; /* where to jump to on return */ 717 char *locinput; /* where to backtrack in string on failure */ 718 char *loceol; 719 U8 *sr0; /* position of start of script run, or NULL */ 720 721 union { 722 723 /* the 'info_aux' and 'info_aux_eval' union members are cuckoos in 724 * the nest. They aren't saved backtrack state; rather they 725 * represent one or two extra chunks of data that need allocating 726 * at the start of a match. These fields would logically live in 727 * the regmatch_info struct, except that is allocated on the 728 * C stack, and these fields are all things that require cleanup 729 * after a croak(), when the stack is lost. 730 * As a convenience, we just use the first 1 or 2 regmatch_state 731 * slots to store this info, as we will be allocating a slab of 732 * these anyway. Otherwise we'd have to malloc and then free them, 733 * or allocate them on the save stack (where they will get 734 * realloced if the save stack grows). 735 * info_aux contains the extra fields that are always needed; 736 * info_aux_eval contains extra fields that only needed if 737 * the pattern contains code blocks 738 * We split them into two separate structs to avoid increasing 739 * the size of the union. 740 */ 741 742 regmatch_info_aux info_aux; 743 744 regmatch_info_aux_eval info_aux_eval; 745 746 /* this is a fake union member that matches the first element 747 * of each member that needs to store positive backtrack 748 * information */ 749 struct { 750 struct regmatch_state *prev_yes_state; 751 } yes; 752 753 /* branchlike members */ 754 /* this is a fake union member that matches the first elements 755 * of each member that needs to behave like a branch */ 756 struct { 757 /* this first element must match u.yes */ 758 struct regmatch_state *prev_yes_state; 759 U32 lastparen; 760 U32 lastcloseparen; 761 CHECKPOINT cp; 762 763 } branchlike; 764 765 struct { 766 /* the first elements must match u.branchlike */ 767 struct regmatch_state *prev_yes_state; 768 U32 lastparen; 769 U32 lastcloseparen; 770 CHECKPOINT cp; 771 772 regnode *next_branch; /* next branch node */ 773 } branch; 774 775 struct { 776 /* the first elements must match u.branchlike */ 777 struct regmatch_state *prev_yes_state; 778 U32 lastparen; 779 U32 lastcloseparen; 780 CHECKPOINT cp; 781 782 U32 accepted; /* how many accepting states left */ 783 bool longfold;/* saw a fold with a 1->n char mapping */ 784 U16 *jump; /* positive offsets from me */ 785 regnode *me; /* Which node am I - needed for jump tries*/ 786 U8 *firstpos;/* pos in string of first trie match */ 787 U32 firstchars;/* len in chars of firstpos from start */ 788 U16 nextword;/* next word to try */ 789 U16 topword; /* longest accepted word */ 790 } trie; 791 792 /* special types - these members are used to store state for special 793 regops like eval, if/then, lookaround and the markpoint state */ 794 struct { 795 /* this first element must match u.yes */ 796 struct regmatch_state *prev_yes_state; 797 struct regmatch_state *prev_curlyx; 798 struct regmatch_state *prev_eval; 799 REGEXP *prev_rex; 800 CHECKPOINT cp; /* remember current savestack indexes */ 801 CHECKPOINT lastcp; 802 U32 close_paren; /* which close bracket is our end (+1) */ 803 regnode *B; /* the node following us */ 804 char *prev_recurse_locinput; 805 } eval; 806 807 struct { 808 /* this first element must match u.yes */ 809 struct regmatch_state *prev_yes_state; 810 I32 wanted; 811 I32 logical; /* saved copy of 'logical' var */ 812 U8 count; /* number of beginning positions */ 813 char *start; 814 char *end; 815 regnode *me; /* the IFMATCH/SUSPEND/UNLESSM node */ 816 } ifmatch; /* and SUSPEND/UNLESSM */ 817 818 struct { 819 /* this first element must match u.yes */ 820 struct regmatch_state *prev_yes_state; 821 struct regmatch_state *prev_mark; 822 SV* mark_name; 823 char *mark_loc; 824 } mark; 825 826 struct { 827 int val; 828 } keeper; 829 830 /* quantifiers - these members are used for storing state for 831 for the regops used to implement quantifiers */ 832 struct { 833 /* this first element must match u.yes */ 834 struct regmatch_state *prev_yes_state; 835 struct regmatch_state *prev_curlyx; /* previous cur_curlyx */ 836 regnode *me; /* the CURLYX node */ 837 regnode *B; /* the B node in /A*B/ */ 838 CHECKPOINT cp; /* remember current savestack index */ 839 bool minmod; 840 int parenfloor;/* how far back to strip paren data */ 841 842 /* these two are modified by WHILEM */ 843 int count; /* how many instances of A we've matched */ 844 char *lastloc;/* where previous A matched (0-len detect) */ 845 } curlyx; 846 847 struct { 848 /* this first element must match u.yes */ 849 struct regmatch_state *prev_yes_state; 850 struct regmatch_state *save_curlyx; 851 CHECKPOINT cp; /* remember current savestack indexes */ 852 CHECKPOINT lastcp; 853 char *save_lastloc; /* previous curlyx.lastloc */ 854 I32 cache_offset; 855 I32 cache_mask; 856 } whilem; 857 858 struct { 859 /* this first element must match u.yes */ 860 struct regmatch_state *prev_yes_state; 861 int c1, c2; /* case fold search */ 862 CHECKPOINT cp; 863 U32 lastparen; 864 U32 lastcloseparen; 865 I32 alen; /* length of first-matched A string */ 866 I32 count; 867 bool minmod; 868 regnode *A, *B; /* the nodes corresponding to /A*B/ */ 869 regnode *me; /* the curlym node */ 870 U8 c1_utf8[UTF8_MAXBYTES+1]; /* */ 871 U8 c2_utf8[UTF8_MAXBYTES+1]; 872 } curlym; 873 874 struct { 875 U32 paren; 876 CHECKPOINT cp; 877 U32 lastparen; 878 U32 lastcloseparen; 879 int c1, c2; /* case fold search */ 880 char *maxpos; /* highest possible point in string to match */ 881 char *oldloc; /* the previous locinput */ 882 int count; 883 int min, max; /* {m,n} */ 884 regnode *A, *B; /* the nodes corresponding to /A*B/ */ 885 U8 c1_utf8[UTF8_MAXBYTES+1]; /* */ 886 U8 c2_utf8[UTF8_MAXBYTES+1]; 887 } curly; /* and CURLYN/PLUS/STAR */ 888 889 } u; 890 } regmatch_state; 891 892 893 894 /* how many regmatch_state structs to allocate as a single slab. 895 * We do it in 4K blocks for efficiency. The "3" is 2 for the next/prev 896 * pointers, plus 1 for any mythical malloc overhead. */ 897 898 #define PERL_REGMATCH_SLAB_SLOTS \ 899 ((4096 - 3 * sizeof (void*)) / sizeof(regmatch_state)) 900 901 typedef struct regmatch_slab { 902 regmatch_state states[PERL_REGMATCH_SLAB_SLOTS]; 903 struct regmatch_slab *prev, *next; 904 } regmatch_slab; 905 906 907 908 /* 909 * ex: set ts=8 sts=4 sw=4 et: 910 */ 911