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 struct regnode { 24 U8 flags; 25 U8 type; 26 U16 next_off; 27 }; 28 29 typedef struct regnode regnode; 30 31 struct reg_substr_data; 32 33 struct reg_data; 34 35 struct regexp_engine; 36 struct regexp; 37 38 struct reg_substr_datum { 39 SSize_t min_offset; /* min pos (in chars) that substr must appear */ 40 SSize_t max_offset /* max pos (in chars) that substr must appear */; 41 SV *substr; /* non-utf8 variant */ 42 SV *utf8_substr; /* utf8 variant */ 43 SSize_t end_shift; /* how many fixed chars must end the string */ 44 }; 45 struct reg_substr_data { 46 U8 check_ix; /* index into data[] of check substr */ 47 struct reg_substr_datum data[3]; /* Actual array */ 48 }; 49 50 #ifdef PERL_ANY_COW 51 #define SV_SAVED_COPY SV *saved_copy; /* If non-NULL, SV which is COW from original */ 52 #else 53 #define SV_SAVED_COPY 54 #endif 55 56 /* offsets within a string of a particular /(.)/ capture */ 57 58 typedef struct regexp_paren_pair { 59 SSize_t start; 60 SSize_t end; 61 /* 'start_tmp' records a new opening position before the matching end 62 * has been found, so that the old start and end values are still 63 * valid, e.g. 64 * "abc" =~ /(.(?{print "[$1]"}))+/ 65 *outputs [][a][b] 66 * This field is not part of the API. */ 67 SSize_t start_tmp; 68 } regexp_paren_pair; 69 70 #if defined(PERL_IN_REGCOMP_C) || defined(PERL_IN_UTF8_C) 71 #define _invlist_union(a, b, output) _invlist_union_maybe_complement_2nd(a, b, FALSE, output) 72 #define _invlist_intersection(a, b, output) _invlist_intersection_maybe_complement_2nd(a, b, FALSE, output) 73 74 /* Subtracting b from a leaves in a everything that was there that isn't in b, 75 * that is the intersection of a with b's complement */ 76 #define _invlist_subtract(a, b, output) _invlist_intersection_maybe_complement_2nd(a, b, TRUE, output) 77 #endif 78 79 /* record the position of a (?{...}) within a pattern */ 80 81 struct reg_code_block { 82 STRLEN start; 83 STRLEN end; 84 OP *block; 85 REGEXP *src_regex; 86 }; 87 88 89 /* 90 The regexp/REGEXP struct, see L<perlreapi> for further documentation 91 on the individual fields. The struct is ordered so that the most 92 commonly used fields are placed at the start. 93 94 Any patch that adds items to this struct will need to include 95 changes to F<sv.c> (C<Perl_re_dup()>) and F<regcomp.c> 96 (C<pregfree()>). This involves freeing or cloning items in the 97 regexp's data array based on the data item's type. 98 */ 99 100 #define _REGEXP_COMMON \ 101 /* what engine created this regexp? */ \ 102 const struct regexp_engine* engine; \ 103 REGEXP *mother_re; /* what re is this a lightweight copy of? */ \ 104 HV *paren_names; /* Optional hash of paren names */ \ 105 /*--------------------------------------------------------*/ \ 106 /* Information about the match that the perl core uses to */ \ 107 /* manage things */ \ 108 U32 extflags; /* Flags used both externally and internally */ \ 109 SSize_t minlen; /* mininum possible number of chars in string to match */\ 110 SSize_t minlenret; /* mininum possible number of chars in $& */ \ 111 STRLEN gofs; /* chars left of pos that we search from */ \ 112 /* substring data about strings that must appear in the */ \ 113 /* final match, used for optimisations */ \ 114 struct reg_substr_data *substrs; \ 115 U32 nparens; /* number of capture buffers */ \ 116 /* private engine specific data */ \ 117 U32 intflags; /* Engine Specific Internal flags */ \ 118 void *pprivate; /* Data private to the regex engine which */ \ 119 /* created this object. */ \ 120 /*--------------------------------------------------------*/ \ 121 /* Data about the last/current match. These are modified */ \ 122 /* during matching */ \ 123 U32 lastparen; /* last open paren matched */ \ 124 U32 lastcloseparen; /* last close paren matched */ \ 125 /* Array of offsets for (@-) and (@+) */ \ 126 regexp_paren_pair *offs; \ 127 char **recurse_locinput; /* used to detect infinite recursion, XXX: move to internal */ \ 128 /*--------------------------------------------------------*/ \ 129 /* saved or original string so \digit works forever. */ \ 130 char *subbeg; \ 131 SV_SAVED_COPY /* If non-NULL, SV which is COW from original */\ 132 SSize_t sublen; /* Length of string pointed by subbeg */ \ 133 SSize_t suboffset; /* byte offset of subbeg from logical start of str */ \ 134 SSize_t subcoffset; /* suboffset equiv, but in chars (for @-/@+) */ \ 135 /* Information about the match that isn't often used */ \ 136 SSize_t maxlen; /* mininum possible number of chars in string to match */\ 137 /*--------------------------------------------------------*/ \ 138 /* offset from wrapped to the start of precomp */ \ 139 PERL_BITFIELD32 pre_prefix:4; \ 140 /* original flags used to compile the pattern, may differ */ \ 141 /* from extflags in various ways */ \ 142 PERL_BITFIELD32 compflags:9; \ 143 /*--------------------------------------------------------*/ \ 144 CV *qr_anoncv /* the anon sub wrapped round qr/(?{..})/ */ 145 146 typedef struct regexp { 147 _XPV_HEAD; 148 _REGEXP_COMMON; 149 } regexp; 150 151 #define RXp_PAREN_NAMES(rx) ((rx)->paren_names) 152 153 /* used for high speed searches */ 154 typedef struct re_scream_pos_data_s 155 { 156 char **scream_olds; /* match pos */ 157 SSize_t *scream_pos; /* Internal iterator of scream. */ 158 } re_scream_pos_data; 159 160 /* regexp_engine structure. This is the dispatch table for regexes. 161 * Any regex engine implementation must be able to build one of these. 162 */ 163 typedef struct regexp_engine { 164 REGEXP* (*comp) (pTHX_ SV * const pattern, U32 flags); 165 I32 (*exec) (pTHX_ REGEXP * const rx, char* stringarg, char* strend, 166 char* strbeg, SSize_t minend, SV* sv, 167 void* data, U32 flags); 168 char* (*intuit) (pTHX_ 169 REGEXP * const rx, 170 SV *sv, 171 const char * const strbeg, 172 char *strpos, 173 char *strend, 174 const U32 flags, 175 re_scream_pos_data *data); 176 SV* (*checkstr) (pTHX_ REGEXP * const rx); 177 void (*rxfree) (pTHX_ REGEXP * const rx); 178 void (*numbered_buff_FETCH) (pTHX_ REGEXP * const rx, const I32 paren, 179 SV * const sv); 180 void (*numbered_buff_STORE) (pTHX_ REGEXP * const rx, const I32 paren, 181 SV const * const value); 182 I32 (*numbered_buff_LENGTH) (pTHX_ REGEXP * const rx, const SV * const sv, 183 const I32 paren); 184 SV* (*named_buff) (pTHX_ REGEXP * const rx, SV * const key, 185 SV * const value, const U32 flags); 186 SV* (*named_buff_iter) (pTHX_ REGEXP * const rx, const SV * const lastkey, 187 const U32 flags); 188 SV* (*qr_package)(pTHX_ REGEXP * const rx); 189 #ifdef USE_ITHREADS 190 void* (*dupe) (pTHX_ REGEXP * const rx, CLONE_PARAMS *param); 191 #endif 192 REGEXP* (*op_comp) (pTHX_ SV ** const patternp, int pat_count, 193 OP *expr, const struct regexp_engine* eng, 194 REGEXP *old_re, 195 bool *is_bare_re, U32 orig_rx_flags, U32 pm_flags); 196 } regexp_engine; 197 198 /* 199 These are passed to the numbered capture variable callbacks as the 200 paren name. >= 1 is reserved for actual numbered captures, i.e. $1, 201 $2 etc. 202 */ 203 #define RX_BUFF_IDX_CARET_PREMATCH -5 /* ${^PREMATCH} */ 204 #define RX_BUFF_IDX_CARET_POSTMATCH -4 /* ${^POSTMATCH} */ 205 #define RX_BUFF_IDX_CARET_FULLMATCH -3 /* ${^MATCH} */ 206 #define RX_BUFF_IDX_PREMATCH -2 /* $` */ 207 #define RX_BUFF_IDX_POSTMATCH -1 /* $' */ 208 #define RX_BUFF_IDX_FULLMATCH 0 /* $& */ 209 210 /* 211 Flags that are passed to the named_buff and named_buff_iter 212 callbacks above. Those routines are called from universal.c via the 213 Tie::Hash::NamedCapture interface for %+ and %- and the re:: 214 functions in the same file. 215 */ 216 217 /* The Tie::Hash::NamedCapture operation this is part of, if any */ 218 #define RXapif_FETCH 0x0001 219 #define RXapif_STORE 0x0002 220 #define RXapif_DELETE 0x0004 221 #define RXapif_CLEAR 0x0008 222 #define RXapif_EXISTS 0x0010 223 #define RXapif_SCALAR 0x0020 224 #define RXapif_FIRSTKEY 0x0040 225 #define RXapif_NEXTKEY 0x0080 226 227 /* Whether %+ or %- is being operated on */ 228 #define RXapif_ONE 0x0100 /* %+ */ 229 #define RXapif_ALL 0x0200 /* %- */ 230 231 /* Whether this is being called from a re:: function */ 232 #define RXapif_REGNAME 0x0400 233 #define RXapif_REGNAMES 0x0800 234 #define RXapif_REGNAMES_COUNT 0x1000 235 236 /* 237 =head1 REGEXP Functions 238 239 =for apidoc Am|REGEXP *|SvRX|SV *sv 240 241 Convenience macro to get the REGEXP from a SV. This is approximately 242 equivalent to the following snippet: 243 244 if (SvMAGICAL(sv)) 245 mg_get(sv); 246 if (SvROK(sv)) 247 sv = MUTABLE_SV(SvRV(sv)); 248 if (SvTYPE(sv) == SVt_REGEXP) 249 return (REGEXP*) sv; 250 251 C<NULL> will be returned if a REGEXP* is not found. 252 253 =for apidoc Am|bool|SvRXOK|SV* sv 254 255 Returns a boolean indicating whether the SV (or the one it references) 256 is a REGEXP. 257 258 If you want to do something with the REGEXP* later use SvRX instead 259 and check for NULL. 260 261 =cut 262 */ 263 264 #define SvRX(sv) (Perl_get_re_arg(aTHX_ sv)) 265 #define SvRXOK(sv) (Perl_get_re_arg(aTHX_ sv) ? TRUE : FALSE) 266 267 268 /* Flags stored in regexp->extflags 269 * These are used by code external to the regexp engine 270 * 271 * Note that the flags whose names start with RXf_PMf_ are defined in 272 * op_reg_common.h, being copied from the parallel flags of op_pmflags 273 * 274 * NOTE: if you modify any RXf flags you should run regen.pl or 275 * regen/regcomp.pl so that regnodes.h is updated with the changes. 276 * 277 */ 278 279 #include "op_reg_common.h" 280 281 #define RXf_PMf_STD_PMMOD (RXf_PMf_MULTILINE|RXf_PMf_SINGLELINE|RXf_PMf_FOLD|RXf_PMf_EXTENDED|RXf_PMf_NOCAPTURE) 282 283 #define CASE_STD_PMMOD_FLAGS_PARSE_SET(pmfl, x_count) \ 284 case IGNORE_PAT_MOD: *(pmfl) |= RXf_PMf_FOLD; break; \ 285 case MULTILINE_PAT_MOD: *(pmfl) |= RXf_PMf_MULTILINE; break; \ 286 case SINGLE_PAT_MOD: *(pmfl) |= RXf_PMf_SINGLELINE; break; \ 287 case XTENDED_PAT_MOD: *(pmfl) |= RXf_PMf_EXTENDED; (x_count)++; break;\ 288 case NOCAPTURE_PAT_MOD: *(pmfl) |= RXf_PMf_NOCAPTURE; break; 289 290 #define STD_PMMOD_FLAGS_PARSE_X_WARN(x_count) \ 291 if (UNLIKELY((x_count) > 1)) { \ 292 Perl_ck_warner_d(aTHX_ packWARN2(WARN_DEPRECATED, WARN_REGEXP), \ 293 "Having more than one /%c regexp modifier is deprecated", \ 294 XTENDED_PAT_MOD); \ 295 } 296 297 /* Note, includes charset ones, assumes 0 is the default for them */ 298 #define STD_PMMOD_FLAGS_CLEAR(pmfl) \ 299 *(pmfl) &= ~(RXf_PMf_FOLD|RXf_PMf_MULTILINE|RXf_PMf_SINGLELINE|RXf_PMf_EXTENDED|RXf_PMf_CHARSET|RXf_PMf_NOCAPTURE) 300 301 /* chars and strings used as regex pattern modifiers 302 * Singular is a 'c'har, plural is a "string" 303 * 304 * NOTE, KEEPCOPY was originally 'k', but was changed to 'p' for preserve 305 * for compatibility reasons with Regexp::Common which highjacked (?k:...) 306 * for its own uses. So 'k' is out as well. 307 */ 308 #define DEFAULT_PAT_MOD '^' /* Short for all the default modifiers */ 309 #define EXEC_PAT_MOD 'e' 310 #define KEEPCOPY_PAT_MOD 'p' 311 #define NOCAPTURE_PAT_MOD 'n' 312 #define ONCE_PAT_MOD 'o' 313 #define GLOBAL_PAT_MOD 'g' 314 #define CONTINUE_PAT_MOD 'c' 315 #define MULTILINE_PAT_MOD 'm' 316 #define SINGLE_PAT_MOD 's' 317 #define IGNORE_PAT_MOD 'i' 318 #define XTENDED_PAT_MOD 'x' 319 #define NONDESTRUCT_PAT_MOD 'r' 320 #define LOCALE_PAT_MOD 'l' 321 #define UNICODE_PAT_MOD 'u' 322 #define DEPENDS_PAT_MOD 'd' 323 #define ASCII_RESTRICT_PAT_MOD 'a' 324 325 #define ONCE_PAT_MODS "o" 326 #define KEEPCOPY_PAT_MODS "p" 327 #define NOCAPTURE_PAT_MODS "n" 328 #define EXEC_PAT_MODS "e" 329 #define LOOP_PAT_MODS "gc" 330 #define NONDESTRUCT_PAT_MODS "r" 331 #define LOCALE_PAT_MODS "l" 332 #define UNICODE_PAT_MODS "u" 333 #define DEPENDS_PAT_MODS "d" 334 #define ASCII_RESTRICT_PAT_MODS "a" 335 #define ASCII_MORE_RESTRICT_PAT_MODS "aa" 336 337 /* This string is expected by regcomp.c to be ordered so that the first 338 * character is the flag in bit RXf_PMf_STD_PMMOD_SHIFT of extflags; the next 339 * character is bit +1, etc. */ 340 #define STD_PAT_MODS "msixxn" 341 342 #define CHARSET_PAT_MODS ASCII_RESTRICT_PAT_MODS DEPENDS_PAT_MODS LOCALE_PAT_MODS UNICODE_PAT_MODS 343 344 /* This string is expected by XS_re_regexp_pattern() in universal.c to be ordered 345 * so that the first character is the flag in bit RXf_PMf_STD_PMMOD_SHIFT of 346 * extflags; the next character is in bit +1, etc. */ 347 #define INT_PAT_MODS STD_PAT_MODS KEEPCOPY_PAT_MODS 348 349 #define EXT_PAT_MODS ONCE_PAT_MODS KEEPCOPY_PAT_MODS NOCAPTURE_PAT_MODS 350 #define QR_PAT_MODS STD_PAT_MODS EXT_PAT_MODS CHARSET_PAT_MODS 351 #define M_PAT_MODS QR_PAT_MODS LOOP_PAT_MODS 352 #define S_PAT_MODS M_PAT_MODS EXEC_PAT_MODS NONDESTRUCT_PAT_MODS 353 354 /* 355 * NOTE: if you modify any RXf flags you should run regen.pl or 356 * regen/regcomp.pl so that regnodes.h is updated with the changes. 357 * 358 */ 359 360 /* 361 Set in Perl_pmruntime if op_flags & OPf_SPECIAL, i.e. split. Will 362 be used by regex engines to check whether they should set 363 RXf_SKIPWHITE 364 */ 365 #define RXf_SPLIT RXf_PMf_SPLIT 366 367 /* Currently the regex flags occupy a single 32-bit word. Not all bits are 368 * currently used. The lower bits are shared with their corresponding PMf flag 369 * bits, up to but not including _RXf_PMf_SHIFT_NEXT. The unused bits 370 * immediately follow; finally the used RXf-only (unshared) bits, so that the 371 * highest bit in the word is used. This gathers all the unused bits as a pool 372 * in the middle, like so: 11111111111111110000001111111111 373 * where the '1's represent used bits, and the '0's unused. This design allows 374 * us to allocate off one end of the pool if we need to add a shared bit, and 375 * off the other end if we need a non-shared bit, without disturbing the other 376 * bits. This maximizes the likelihood of being able to change things without 377 * breaking binary compatibility. 378 * 379 * To add shared bits, do so in op_reg_common.h. This should change 380 * _RXf_PMf_SHIFT_NEXT so that things won't compile. Then come to regexp.h and 381 * op.h and adjust the constant adders in the definitions of RXf_BASE_SHIFT and 382 * Pmf_BASE_SHIFT down by the number of shared bits you added. That's it. 383 * Things should be binary compatible. But if either of these gets to having 384 * to subtract rather than add, leave at 0 and instead adjust all the entries 385 * that are in terms of it. But if the first one of those is already 386 * RXf_BASE_SHIFT+0, there are no bits left, and a redesign is in order. 387 * 388 * To remove unshared bits, just delete its entry. If you're where breaking 389 * binary compatibility is ok to do, you might want to adjust things to move 390 * the newly opened space so that it gets absorbed into the common pool. 391 * 392 * To add unshared bits, first use up any gaps in the middle. Otherwise, 393 * allocate off the low end until you get to RXf_BASE_SHIFT+0. If that isn't 394 * enough, move RXf_BASE_SHIFT down (if possible) and add the new bit at the 395 * other end instead; this preserves binary compatibility. 396 * 397 * For the regexp bits, PL_reg_extflags_name[] in regnodes.h has a comment 398 * giving which bits are used/unused */ 399 400 #define RXf_BASE_SHIFT (_RXf_PMf_SHIFT_NEXT + 2) 401 402 /* What we have seen */ 403 #define RXf_NO_INPLACE_SUBST (1U<<(RXf_BASE_SHIFT+2)) 404 #define RXf_EVAL_SEEN (1U<<(RXf_BASE_SHIFT+3)) 405 406 /* Special */ 407 #define RXf_UNBOUNDED_QUANTIFIER_SEEN (1U<<(RXf_BASE_SHIFT+4)) 408 #define RXf_CHECK_ALL (1U<<(RXf_BASE_SHIFT+5)) 409 410 /* UTF8 related */ 411 #define RXf_MATCH_UTF8 (1U<<(RXf_BASE_SHIFT+6)) /* $1 etc are utf8 */ 412 413 /* Intuit related */ 414 #define RXf_USE_INTUIT_NOML (1U<<(RXf_BASE_SHIFT+7)) 415 #define RXf_USE_INTUIT_ML (1U<<(RXf_BASE_SHIFT+8)) 416 #define RXf_INTUIT_TAIL (1U<<(RXf_BASE_SHIFT+9)) 417 #define RXf_USE_INTUIT (RXf_USE_INTUIT_NOML|RXf_USE_INTUIT_ML) 418 419 /* Do we have some sort of anchor? */ 420 #define RXf_IS_ANCHORED (1U<<(RXf_BASE_SHIFT+10)) 421 422 /* Copy and tainted info */ 423 #define RXf_COPY_DONE (1U<<(RXf_BASE_SHIFT+11)) 424 425 /* post-execution: $1 et al are tainted */ 426 #define RXf_TAINTED_SEEN (1U<<(RXf_BASE_SHIFT+12)) 427 /* this pattern was tainted during compilation */ 428 #define RXf_TAINTED (1U<<(RXf_BASE_SHIFT+13)) 429 430 /* Flags indicating special patterns */ 431 #define RXf_START_ONLY (1U<<(RXf_BASE_SHIFT+14)) /* Pattern is /^/ */ 432 #define RXf_SKIPWHITE (1U<<(RXf_BASE_SHIFT+15)) /* Pattern is for a */ 433 /* split " " */ 434 #define RXf_WHITE (1U<<(RXf_BASE_SHIFT+16)) /* Pattern is /\s+/ */ 435 #define RXf_NULL (1U<<(RXf_BASE_SHIFT+17)) /* Pattern is // */ 436 437 /* See comments at the beginning of these defines about adding bits. The 438 * highest bit position should be used, so that if RXf_BASE_SHIFT gets 439 * increased, the #error below will be triggered so that you will be reminded 440 * to adjust things at the other end to keep the bit positions unchanged */ 441 #if RXf_BASE_SHIFT+17 > 31 442 # error Too many RXf_PMf bits used. See comments at beginning of these for what to do 443 #endif 444 445 /* 446 * NOTE: if you modify any RXf flags you should run regen.pl or 447 * regen/regcomp.pl so that regnodes.h is updated with the changes. 448 * 449 */ 450 451 #ifdef NO_TAINT_SUPPORT 452 # define RX_ISTAINTED(prog) 0 453 # define RX_TAINT_on(prog) NOOP 454 # define RXp_MATCH_TAINTED(prog) 0 455 # define RX_MATCH_TAINTED(prog) 0 456 # define RXp_MATCH_TAINTED_on(prog) NOOP 457 # define RX_MATCH_TAINTED_on(prog) NOOP 458 # define RX_MATCH_TAINTED_off(prog) NOOP 459 #else 460 # define RX_ISTAINTED(prog) (RX_EXTFLAGS(prog) & RXf_TAINTED) 461 # define RX_TAINT_on(prog) (RX_EXTFLAGS(prog) |= RXf_TAINTED) 462 # define RXp_MATCH_TAINTED(prog) (RXp_EXTFLAGS(prog) & RXf_TAINTED_SEEN) 463 # define RX_MATCH_TAINTED(prog) (RX_EXTFLAGS(prog) & RXf_TAINTED_SEEN) 464 # define RXp_MATCH_TAINTED_on(prog) (RXp_EXTFLAGS(prog) |= RXf_TAINTED_SEEN) 465 # define RX_MATCH_TAINTED_on(prog) (RX_EXTFLAGS(prog) |= RXf_TAINTED_SEEN) 466 # define RX_MATCH_TAINTED_off(prog) (RX_EXTFLAGS(prog) &= ~RXf_TAINTED_SEEN) 467 #endif 468 469 #define RX_HAS_CUTGROUP(prog) ((prog)->intflags & PREGf_CUTGROUP_SEEN) 470 #define RX_MATCH_TAINTED_set(prog, t) ((t) \ 471 ? RX_MATCH_TAINTED_on(prog) \ 472 : RX_MATCH_TAINTED_off(prog)) 473 474 #define RXp_MATCH_COPIED(prog) (RXp_EXTFLAGS(prog) & RXf_COPY_DONE) 475 #define RX_MATCH_COPIED(prog) (RX_EXTFLAGS(prog) & RXf_COPY_DONE) 476 #define RXp_MATCH_COPIED_on(prog) (RXp_EXTFLAGS(prog) |= RXf_COPY_DONE) 477 #define RX_MATCH_COPIED_on(prog) (RX_EXTFLAGS(prog) |= RXf_COPY_DONE) 478 #define RXp_MATCH_COPIED_off(prog) (RXp_EXTFLAGS(prog) &= ~RXf_COPY_DONE) 479 #define RX_MATCH_COPIED_off(prog) (RX_EXTFLAGS(prog) &= ~RXf_COPY_DONE) 480 #define RX_MATCH_COPIED_set(prog,t) ((t) \ 481 ? RX_MATCH_COPIED_on(prog) \ 482 : RX_MATCH_COPIED_off(prog)) 483 484 #define RXp_EXTFLAGS(rx) ((rx)->extflags) 485 #define RXp_COMPFLAGS(rx) ((rx)->compflags) 486 487 /* For source compatibility. We used to store these explicitly. */ 488 #define RX_PRECOMP(prog) (RX_WRAPPED(prog) + ReANY(prog)->pre_prefix) 489 #define RX_PRECOMP_const(prog) (RX_WRAPPED_const(prog) + ReANY(prog)->pre_prefix) 490 /* FIXME? Are we hardcoding too much here and constraining plugin extension 491 writers? Specifically, the value 1 assumes that the wrapped version always 492 has exactly one character at the end, a ')'. Will that always be true? */ 493 #define RX_PRELEN(prog) (RX_WRAPLEN(prog) - ReANY(prog)->pre_prefix - 1) 494 #define RX_WRAPPED(prog) ReANY(prog)->xpv_len_u.xpvlenu_pv 495 #define RX_WRAPPED_const(prog) ((const char *)RX_WRAPPED(prog)) 496 #define RX_WRAPLEN(prog) SvCUR(prog) 497 #define RX_CHECK_SUBSTR(prog) (ReANY(prog)->check_substr) 498 #define RX_REFCNT(prog) SvREFCNT(prog) 499 #define RX_EXTFLAGS(prog) RXp_EXTFLAGS(ReANY(prog)) 500 #define RX_COMPFLAGS(prog) RXp_COMPFLAGS(ReANY(prog)) 501 #define RX_ENGINE(prog) (ReANY(prog)->engine) 502 #define RX_SUBBEG(prog) (ReANY(prog)->subbeg) 503 #define RX_SUBOFFSET(prog) (ReANY(prog)->suboffset) 504 #define RX_SUBCOFFSET(prog) (ReANY(prog)->subcoffset) 505 #define RX_OFFS(prog) (ReANY(prog)->offs) 506 #define RX_NPARENS(prog) (ReANY(prog)->nparens) 507 #define RX_SUBLEN(prog) (ReANY(prog)->sublen) 508 #define RX_MINLEN(prog) (ReANY(prog)->minlen) 509 #define RX_MINLENRET(prog) (ReANY(prog)->minlenret) 510 #define RX_GOFS(prog) (ReANY(prog)->gofs) 511 #define RX_LASTPAREN(prog) (ReANY(prog)->lastparen) 512 #define RX_LASTCLOSEPAREN(prog) (ReANY(prog)->lastcloseparen) 513 #define RX_SAVED_COPY(prog) (ReANY(prog)->saved_copy) 514 /* last match was zero-length */ 515 #define RX_ZERO_LEN(prog) \ 516 (RX_OFFS(prog)[0].start + (SSize_t)RX_GOFS(prog) \ 517 == RX_OFFS(prog)[0].end) 518 519 #endif /* PLUGGABLE_RE_EXTENSION */ 520 521 /* Stuff that needs to be included in the pluggable extension goes below here */ 522 523 #ifdef PERL_ANY_COW 524 #define RX_MATCH_COPY_FREE(rx) \ 525 STMT_START {if (RX_SAVED_COPY(rx)) { \ 526 SV_CHECK_THINKFIRST_COW_DROP(RX_SAVED_COPY(rx)); \ 527 } \ 528 if (RX_MATCH_COPIED(rx)) { \ 529 Safefree(RX_SUBBEG(rx)); \ 530 RX_MATCH_COPIED_off(rx); \ 531 }} STMT_END 532 #else 533 #define RX_MATCH_COPY_FREE(rx) \ 534 STMT_START {if (RX_MATCH_COPIED(rx)) { \ 535 Safefree(RX_SUBBEG(rx)); \ 536 RX_MATCH_COPIED_off(rx); \ 537 }} STMT_END 538 #endif 539 540 #define RXp_MATCH_UTF8(prog) (RXp_EXTFLAGS(prog) & RXf_MATCH_UTF8) 541 #define RX_MATCH_UTF8(prog) (RX_EXTFLAGS(prog) & RXf_MATCH_UTF8) 542 #define RX_MATCH_UTF8_on(prog) (RX_EXTFLAGS(prog) |= RXf_MATCH_UTF8) 543 #define RX_MATCH_UTF8_off(prog) (RX_EXTFLAGS(prog) &= ~RXf_MATCH_UTF8) 544 #define RX_MATCH_UTF8_set(prog, t) ((t) \ 545 ? RX_MATCH_UTF8_on(prog) \ 546 : RX_MATCH_UTF8_off(prog)) 547 548 /* Whether the pattern stored at RX_WRAPPED is in UTF-8 */ 549 #define RX_UTF8(prog) SvUTF8(prog) 550 551 552 /* bits in flags arg of Perl_regexec_flags() */ 553 554 #define REXEC_COPY_STR 0x01 /* Need to copy the string for captures. */ 555 #define REXEC_CHECKED 0x02 /* re_intuit_start() already called. */ 556 #define REXEC_SCREAM 0x04 /* currently unused. */ 557 #define REXEC_IGNOREPOS 0x08 /* use stringarg, not pos(), for \G match */ 558 #define REXEC_NOT_FIRST 0x10 /* This is another iteration of //g: 559 no need to copy string again */ 560 561 /* under REXEC_COPY_STR, it's ok for the 562 engine (modulo PL_sawamperand etc) 563 to skip copying: ... */ 564 #define REXEC_COPY_SKIP_PRE 0x20 /* ...the $` part of the string, or */ 565 #define REXEC_COPY_SKIP_POST 0x40 /* ...the $' part of the string */ 566 #define REXEC_FAIL_ON_UNDERFLOW 0x80 /* fail the match if $& would start before 567 the start pos (so s/.\G// would fail 568 on second iteration */ 569 570 #if defined(__GNUC__) && !defined(PERL_GCC_BRACE_GROUPS_FORBIDDEN) 571 # define ReREFCNT_inc(re) \ 572 ({ \ 573 /* This is here to generate a casting warning if incorrect. */ \ 574 REGEXP *const _rerefcnt_inc = (re); \ 575 assert(SvTYPE(_rerefcnt_inc) == SVt_REGEXP); \ 576 SvREFCNT_inc(_rerefcnt_inc); \ 577 _rerefcnt_inc; \ 578 }) 579 # define ReREFCNT_dec(re) \ 580 ({ \ 581 /* This is here to generate a casting warning if incorrect. */ \ 582 REGEXP *const _rerefcnt_dec = (re); \ 583 SvREFCNT_dec(_rerefcnt_dec); \ 584 }) 585 #else 586 # define ReREFCNT_dec(re) SvREFCNT_dec(re) 587 # define ReREFCNT_inc(re) ((REGEXP *) SvREFCNT_inc(re)) 588 #endif 589 #define ReANY(re) S_ReANY((const REGEXP *)(re)) 590 591 /* FIXME for plugins. */ 592 593 #define FBMcf_TAIL_DOLLAR 1 594 #define FBMcf_TAIL_DOLLARM 2 595 #define FBMcf_TAIL_Z 4 596 #define FBMcf_TAIL_z 8 597 #define FBMcf_TAIL (FBMcf_TAIL_DOLLAR|FBMcf_TAIL_DOLLARM|FBMcf_TAIL_Z|FBMcf_TAIL_z) 598 599 #define FBMrf_MULTILINE 1 600 601 struct regmatch_state; 602 struct regmatch_slab; 603 604 /* like regmatch_info_aux, but contains extra fields only needed if the 605 * pattern contains (?{}). If used, is snuck into the second slot in the 606 * regmatch_state stack at the start of execution */ 607 608 typedef struct { 609 regexp *rex; 610 PMOP *curpm; /* saved PL_curpm */ 611 #ifdef PERL_ANY_COW 612 SV *saved_copy; /* saved saved_copy field from rex */ 613 #endif 614 char *subbeg; /* saved subbeg field from rex */ 615 STRLEN sublen; /* saved sublen field from rex */ 616 STRLEN suboffset; /* saved suboffset field from rex */ 617 STRLEN subcoffset; /* saved subcoffset field from rex */ 618 MAGIC *pos_magic; /* pos() magic attached to $_ */ 619 SSize_t pos; /* the original value of pos() in pos_magic */ 620 U8 pos_flags; /* flags to be restored; currently only MGf_BYTES*/ 621 } regmatch_info_aux_eval; 622 623 624 /* fields that logically live in regmatch_info, but which need cleaning 625 * up on croak(), and so are instead are snuck into the first slot in 626 * the regmatch_state stack at the start of execution */ 627 628 typedef struct { 629 regmatch_info_aux_eval *info_aux_eval; 630 struct regmatch_state *old_regmatch_state; /* saved PL_regmatch_state */ 631 struct regmatch_slab *old_regmatch_slab; /* saved PL_regmatch_slab */ 632 char *poscache; /* S-L cache of fail positions of WHILEMs */ 633 } regmatch_info_aux; 634 635 636 /* some basic information about the current match that is created by 637 * Perl_regexec_flags and then passed to regtry(), regmatch() etc. 638 * It is allocated as a local var on the stack, so nothing should be 639 * stored in it that needs preserving or clearing up on croak(). 640 * For that, see the aux_info and aux_info_eval members of the 641 * regmatch_state union. */ 642 643 typedef struct { 644 REGEXP *prog; /* the regex being executed */ 645 const char * strbeg; /* real start of string */ 646 char *strend; /* one byte beyond last char of match string */ 647 char *till; /* matches shorter than this fail (see minlen arg) */ 648 SV *sv; /* the SV string currently being matched */ 649 char *ganch; /* position of \G anchor */ 650 char *cutpoint; /* (*COMMIT) position (if any) */ 651 regmatch_info_aux *info_aux; /* extra fields that need cleanup */ 652 regmatch_info_aux_eval *info_aux_eval; /* extra saved state for (?{}) */ 653 I32 poscache_maxiter; /* how many whilems todo before S-L cache kicks in */ 654 I32 poscache_iter; /* current countdown from _maxiter to zero */ 655 STRLEN poscache_size; /* size of regmatch_info_aux.poscache */ 656 bool intuit; /* re_intuit_start() is the top-level caller */ 657 bool is_utf8_pat; /* regex is utf8 */ 658 bool is_utf8_target; /* string being matched is utf8 */ 659 bool warned; /* we have issued a recursion warning; no need for more */ 660 } regmatch_info; 661 662 663 /* structures for holding and saving the state maintained by regmatch() */ 664 665 #ifndef MAX_RECURSE_EVAL_NOCHANGE_DEPTH 666 #define MAX_RECURSE_EVAL_NOCHANGE_DEPTH 10 667 #endif 668 669 typedef I32 CHECKPOINT; 670 671 typedef struct regmatch_state { 672 int resume_state; /* where to jump to on return */ 673 char *locinput; /* where to backtrack in string on failure */ 674 675 union { 676 677 /* the 'info_aux' and 'info_aux_eval' union members are cuckoos in 678 * the nest. They aren't saved backtrack state; rather they 679 * represent one or two extra chunks of data that need allocating 680 * at the start of a match. These fields would logically live in 681 * the regmatch_info struct, except that is allocated on the 682 * C stack, and these fields are all things that require cleanup 683 * after a croak(), when the stack is lost. 684 * As a convenience, we just use the first 1 or 2 regmatch_state 685 * slots to store this info, as we will be allocating a slab of 686 * these anyway. Otherwise we'd have to malloc and then free them, 687 * or allocate them on the save stack (where they will get 688 * realloced if the save stack grows). 689 * info_aux contains the extra fields that are always needed; 690 * info_aux_eval contains extra fields that only needed if 691 * the pattern contains code blocks 692 * We split them into two separate structs to avoid increasing 693 * the size of the union. 694 */ 695 696 regmatch_info_aux info_aux; 697 698 regmatch_info_aux_eval info_aux_eval; 699 700 /* this is a fake union member that matches the first element 701 * of each member that needs to store positive backtrack 702 * information */ 703 struct { 704 struct regmatch_state *prev_yes_state; 705 } yes; 706 707 /* branchlike members */ 708 /* this is a fake union member that matches the first elements 709 * of each member that needs to behave like a branch */ 710 struct { 711 /* this first element must match u.yes */ 712 struct regmatch_state *prev_yes_state; 713 U32 lastparen; 714 U32 lastcloseparen; 715 CHECKPOINT cp; 716 717 } branchlike; 718 719 struct { 720 /* the first elements must match u.branchlike */ 721 struct regmatch_state *prev_yes_state; 722 U32 lastparen; 723 U32 lastcloseparen; 724 CHECKPOINT cp; 725 726 regnode *next_branch; /* next branch node */ 727 } branch; 728 729 struct { 730 /* the first elements must match u.branchlike */ 731 struct regmatch_state *prev_yes_state; 732 U32 lastparen; 733 U32 lastcloseparen; 734 CHECKPOINT cp; 735 736 U32 accepted; /* how many accepting states left */ 737 bool longfold;/* saw a fold with a 1->n char mapping */ 738 U16 *jump; /* positive offsets from me */ 739 regnode *me; /* Which node am I - needed for jump tries*/ 740 U8 *firstpos;/* pos in string of first trie match */ 741 U32 firstchars;/* len in chars of firstpos from start */ 742 U16 nextword;/* next word to try */ 743 U16 topword; /* longest accepted word */ 744 } trie; 745 746 /* special types - these members are used to store state for special 747 regops like eval, if/then, lookaround and the markpoint state */ 748 struct { 749 /* this first element must match u.yes */ 750 struct regmatch_state *prev_yes_state; 751 struct regmatch_state *prev_curlyx; 752 struct regmatch_state *prev_eval; 753 REGEXP *prev_rex; 754 CHECKPOINT cp; /* remember current savestack indexes */ 755 CHECKPOINT lastcp; 756 U32 close_paren; /* which close bracket is our end (+1) */ 757 regnode *B; /* the node following us */ 758 char *prev_recurse_locinput; 759 } eval; 760 761 struct { 762 /* this first element must match u.yes */ 763 struct regmatch_state *prev_yes_state; 764 I32 wanted; 765 I32 logical; /* saved copy of 'logical' var */ 766 regnode *me; /* the IFMATCH/SUSPEND/UNLESSM node */ 767 } ifmatch; /* and SUSPEND/UNLESSM */ 768 769 struct { 770 /* this first element must match u.yes */ 771 struct regmatch_state *prev_yes_state; 772 struct regmatch_state *prev_mark; 773 SV* mark_name; 774 char *mark_loc; 775 } mark; 776 777 struct { 778 int val; 779 } keeper; 780 781 /* quantifiers - these members are used for storing state for 782 for the regops used to implement quantifiers */ 783 struct { 784 /* this first element must match u.yes */ 785 struct regmatch_state *prev_yes_state; 786 struct regmatch_state *prev_curlyx; /* previous cur_curlyx */ 787 regnode *me; /* the CURLYX node */ 788 regnode *B; /* the B node in /A*B/ */ 789 CHECKPOINT cp; /* remember current savestack index */ 790 bool minmod; 791 int parenfloor;/* how far back to strip paren data */ 792 793 /* these two are modified by WHILEM */ 794 int count; /* how many instances of A we've matched */ 795 char *lastloc;/* where previous A matched (0-len detect) */ 796 } curlyx; 797 798 struct { 799 /* this first element must match u.yes */ 800 struct regmatch_state *prev_yes_state; 801 struct regmatch_state *save_curlyx; 802 CHECKPOINT cp; /* remember current savestack indexes */ 803 CHECKPOINT lastcp; 804 char *save_lastloc; /* previous curlyx.lastloc */ 805 I32 cache_offset; 806 I32 cache_mask; 807 } whilem; 808 809 struct { 810 /* this first element must match u.yes */ 811 struct regmatch_state *prev_yes_state; 812 int c1, c2; /* case fold search */ 813 CHECKPOINT cp; 814 U32 lastparen; 815 U32 lastcloseparen; 816 I32 alen; /* length of first-matched A string */ 817 I32 count; 818 bool minmod; 819 regnode *A, *B; /* the nodes corresponding to /A*B/ */ 820 regnode *me; /* the curlym node */ 821 U8 c1_utf8[UTF8_MAXBYTES+1]; /* */ 822 U8 c2_utf8[UTF8_MAXBYTES+1]; 823 } curlym; 824 825 struct { 826 U32 paren; 827 CHECKPOINT cp; 828 U32 lastparen; 829 U32 lastcloseparen; 830 int c1, c2; /* case fold search */ 831 char *maxpos; /* highest possible point in string to match */ 832 char *oldloc; /* the previous locinput */ 833 int count; 834 int min, max; /* {m,n} */ 835 regnode *A, *B; /* the nodes corresponding to /A*B/ */ 836 U8 c1_utf8[UTF8_MAXBYTES+1]; /* */ 837 U8 c2_utf8[UTF8_MAXBYTES+1]; 838 } curly; /* and CURLYN/PLUS/STAR */ 839 840 } u; 841 } regmatch_state; 842 843 844 845 /* how many regmatch_state structs to allocate as a single slab. 846 * We do it in 4K blocks for efficiency. The "3" is 2 for the next/prev 847 * pointers, plus 1 for any mythical malloc overhead. */ 848 849 #define PERL_REGMATCH_SLAB_SLOTS \ 850 ((4096 - 3 * sizeof (void*)) / sizeof(regmatch_state)) 851 852 typedef struct regmatch_slab { 853 regmatch_state states[PERL_REGMATCH_SLAB_SLOTS]; 854 struct regmatch_slab *prev, *next; 855 } regmatch_slab; 856 857 858 859 /* 860 * ex: set ts=8 sts=4 sw=4 et: 861 */ 862