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