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 /* Information about the match that the perl core uses to */ \ 106 /* manage things */ \ 107 U32 extflags; /* Flags used both externally and internally */ \ 108 SSize_t minlen; /* mininum possible number of chars in string to match */\ 109 SSize_t minlenret; /* mininum possible number of chars in $& */ \ 110 STRLEN gofs; /* chars left of pos that we search from */ \ 111 /* substring data about strings that must appear in the */ \ 112 /* final match, used for optimisations */ \ 113 struct reg_substr_data *substrs; \ 114 U32 nparens; /* number of capture buffers */ \ 115 /* private engine specific data */ \ 116 U32 intflags; /* Engine Specific Internal flags */ \ 117 void *pprivate; /* Data private to the regex engine which */ \ 118 /* created this object. */ \ 119 /* Data about the last/current match. These are modified */ \ 120 /* during matching */ \ 121 U32 lastparen; /* last open paren matched */ \ 122 U32 lastcloseparen; /* last close paren matched */ \ 123 /* Array of offsets for (@-) and (@+) */ \ 124 regexp_paren_pair *offs; \ 125 /* saved or original string so \digit works forever. */ \ 126 char *subbeg; \ 127 SV_SAVED_COPY /* If non-NULL, SV which is COW from original */\ 128 SSize_t sublen; /* Length of string pointed by subbeg */ \ 129 SSize_t suboffset; /* byte offset of subbeg from logical start of str */ \ 130 SSize_t subcoffset; /* suboffset equiv, but in chars (for @-/@+) */ \ 131 /* Information about the match that isn't often used */ \ 132 SSize_t maxlen; /* mininum possible number of chars in string to match */\ 133 /* offset from wrapped to the start of precomp */ \ 134 PERL_BITFIELD32 pre_prefix:4; \ 135 /* original flags used to compile the pattern, may differ */ \ 136 /* from extflags in various ways */ \ 137 PERL_BITFIELD32 compflags:9; \ 138 CV *qr_anoncv /* the anon sub wrapped round qr/(?{..})/ */ 139 140 typedef struct regexp { 141 _XPV_HEAD; 142 _REGEXP_COMMON; 143 } regexp; 144 145 #define RXp_PAREN_NAMES(rx) ((rx)->paren_names) 146 147 /* used for high speed searches */ 148 typedef struct re_scream_pos_data_s 149 { 150 char **scream_olds; /* match pos */ 151 SSize_t *scream_pos; /* Internal iterator of scream. */ 152 } re_scream_pos_data; 153 154 /* regexp_engine structure. This is the dispatch table for regexes. 155 * Any regex engine implementation must be able to build one of these. 156 */ 157 typedef struct regexp_engine { 158 REGEXP* (*comp) (pTHX_ SV * const pattern, U32 flags); 159 I32 (*exec) (pTHX_ REGEXP * const rx, char* stringarg, char* strend, 160 char* strbeg, SSize_t minend, SV* sv, 161 void* data, U32 flags); 162 char* (*intuit) (pTHX_ 163 REGEXP * const rx, 164 SV *sv, 165 const char * const strbeg, 166 char *strpos, 167 char *strend, 168 const U32 flags, 169 re_scream_pos_data *data); 170 SV* (*checkstr) (pTHX_ REGEXP * const rx); 171 void (*free) (pTHX_ REGEXP * const rx); 172 void (*numbered_buff_FETCH) (pTHX_ REGEXP * const rx, const I32 paren, 173 SV * const sv); 174 void (*numbered_buff_STORE) (pTHX_ REGEXP * const rx, const I32 paren, 175 SV const * const value); 176 I32 (*numbered_buff_LENGTH) (pTHX_ REGEXP * const rx, const SV * const sv, 177 const I32 paren); 178 SV* (*named_buff) (pTHX_ REGEXP * const rx, SV * const key, 179 SV * const value, const U32 flags); 180 SV* (*named_buff_iter) (pTHX_ REGEXP * const rx, const SV * const lastkey, 181 const U32 flags); 182 SV* (*qr_package)(pTHX_ REGEXP * const rx); 183 #ifdef USE_ITHREADS 184 void* (*dupe) (pTHX_ REGEXP * const rx, CLONE_PARAMS *param); 185 #endif 186 REGEXP* (*op_comp) (pTHX_ SV ** const patternp, int pat_count, 187 OP *expr, const struct regexp_engine* eng, 188 REGEXP *VOL old_re, 189 bool *is_bare_re, U32 orig_rx_flags, U32 pm_flags); 190 } regexp_engine; 191 192 /* 193 These are passed to the numbered capture variable callbacks as the 194 paren name. >= 1 is reserved for actual numbered captures, i.e. $1, 195 $2 etc. 196 */ 197 #define RX_BUFF_IDX_CARET_PREMATCH -5 /* ${^PREMATCH} */ 198 #define RX_BUFF_IDX_CARET_POSTMATCH -4 /* ${^POSTMATCH} */ 199 #define RX_BUFF_IDX_CARET_FULLMATCH -3 /* ${^MATCH} */ 200 #define RX_BUFF_IDX_PREMATCH -2 /* $` */ 201 #define RX_BUFF_IDX_POSTMATCH -1 /* $' */ 202 #define RX_BUFF_IDX_FULLMATCH 0 /* $& */ 203 204 /* 205 Flags that are passed to the named_buff and named_buff_iter 206 callbacks above. Those routines are called from universal.c via the 207 Tie::Hash::NamedCapture interface for %+ and %- and the re:: 208 functions in the same file. 209 */ 210 211 /* The Tie::Hash::NamedCapture operation this is part of, if any */ 212 #define RXapif_FETCH 0x0001 213 #define RXapif_STORE 0x0002 214 #define RXapif_DELETE 0x0004 215 #define RXapif_CLEAR 0x0008 216 #define RXapif_EXISTS 0x0010 217 #define RXapif_SCALAR 0x0020 218 #define RXapif_FIRSTKEY 0x0040 219 #define RXapif_NEXTKEY 0x0080 220 221 /* Whether %+ or %- is being operated on */ 222 #define RXapif_ONE 0x0100 /* %+ */ 223 #define RXapif_ALL 0x0200 /* %- */ 224 225 /* Whether this is being called from a re:: function */ 226 #define RXapif_REGNAME 0x0400 227 #define RXapif_REGNAMES 0x0800 228 #define RXapif_REGNAMES_COUNT 0x1000 229 230 /* 231 =head1 REGEXP Functions 232 233 =for apidoc Am|REGEXP *|SvRX|SV *sv 234 235 Convenience macro to get the REGEXP from a SV. This is approximately 236 equivalent to the following snippet: 237 238 if (SvMAGICAL(sv)) 239 mg_get(sv); 240 if (SvROK(sv)) 241 sv = MUTABLE_SV(SvRV(sv)); 242 if (SvTYPE(sv) == SVt_REGEXP) 243 return (REGEXP*) sv; 244 245 NULL will be returned if a REGEXP* is not found. 246 247 =for apidoc Am|bool|SvRXOK|SV* sv 248 249 Returns a boolean indicating whether the SV (or the one it references) 250 is a REGEXP. 251 252 If you want to do something with the REGEXP* later use SvRX instead 253 and check for NULL. 254 255 =cut 256 */ 257 258 #define SvRX(sv) (Perl_get_re_arg(aTHX_ sv)) 259 #define SvRXOK(sv) (Perl_get_re_arg(aTHX_ sv) ? TRUE : FALSE) 260 261 262 /* Flags stored in regexp->extflags 263 * These are used by code external to the regexp engine 264 * 265 * Note that the flags whose names start with RXf_PMf_ are defined in 266 * op_reg_common.h, being copied from the parallel flags of op_pmflags 267 * 268 * NOTE: if you modify any RXf flags you should run regen.pl or 269 * regen/regcomp.pl so that regnodes.h is updated with the changes. 270 * 271 */ 272 273 #include "op_reg_common.h" 274 275 #define RXf_PMf_STD_PMMOD (RXf_PMf_MULTILINE|RXf_PMf_SINGLELINE|RXf_PMf_FOLD|RXf_PMf_EXTENDED) 276 277 #define CASE_STD_PMMOD_FLAGS_PARSE_SET(pmfl) \ 278 case IGNORE_PAT_MOD: *(pmfl) |= RXf_PMf_FOLD; break; \ 279 case MULTILINE_PAT_MOD: *(pmfl) |= RXf_PMf_MULTILINE; break; \ 280 case SINGLE_PAT_MOD: *(pmfl) |= RXf_PMf_SINGLELINE; break; \ 281 case XTENDED_PAT_MOD: *(pmfl) |= RXf_PMf_EXTENDED; break 282 283 /* Note, includes charset ones, assumes 0 is the default for them */ 284 #define STD_PMMOD_FLAGS_CLEAR(pmfl) \ 285 *(pmfl) &= ~(RXf_PMf_FOLD|RXf_PMf_MULTILINE|RXf_PMf_SINGLELINE|RXf_PMf_EXTENDED|RXf_PMf_CHARSET) 286 287 /* chars and strings used as regex pattern modifiers 288 * Singular is a 'c'har, plural is a "string" 289 * 290 * NOTE, KEEPCOPY was originally 'k', but was changed to 'p' for preserve 291 * for compatibility reasons with Regexp::Common which highjacked (?k:...) 292 * for its own uses. So 'k' is out as well. 293 */ 294 #define DEFAULT_PAT_MOD '^' /* Short for all the default modifiers */ 295 #define EXEC_PAT_MOD 'e' 296 #define KEEPCOPY_PAT_MOD 'p' 297 #define ONCE_PAT_MOD 'o' 298 #define GLOBAL_PAT_MOD 'g' 299 #define CONTINUE_PAT_MOD 'c' 300 #define MULTILINE_PAT_MOD 'm' 301 #define SINGLE_PAT_MOD 's' 302 #define IGNORE_PAT_MOD 'i' 303 #define XTENDED_PAT_MOD 'x' 304 #define NONDESTRUCT_PAT_MOD 'r' 305 #define LOCALE_PAT_MOD 'l' 306 #define UNICODE_PAT_MOD 'u' 307 #define DEPENDS_PAT_MOD 'd' 308 #define ASCII_RESTRICT_PAT_MOD 'a' 309 310 #define ONCE_PAT_MODS "o" 311 #define KEEPCOPY_PAT_MODS "p" 312 #define EXEC_PAT_MODS "e" 313 #define LOOP_PAT_MODS "gc" 314 #define NONDESTRUCT_PAT_MODS "r" 315 #define LOCALE_PAT_MODS "l" 316 #define UNICODE_PAT_MODS "u" 317 #define DEPENDS_PAT_MODS "d" 318 #define ASCII_RESTRICT_PAT_MODS "a" 319 #define ASCII_MORE_RESTRICT_PAT_MODS "aa" 320 321 /* This string is expected by regcomp.c to be ordered so that the first 322 * character is the flag in bit RXf_PMf_STD_PMMOD_SHIFT of extflags; the next 323 * character is bit +1, etc. */ 324 #define STD_PAT_MODS "msix" 325 326 #define CHARSET_PAT_MODS ASCII_RESTRICT_PAT_MODS DEPENDS_PAT_MODS LOCALE_PAT_MODS UNICODE_PAT_MODS 327 328 /* This string is expected by XS_re_regexp_pattern() in universal.c to be ordered 329 * so that the first character is the flag in bit RXf_PMf_STD_PMMOD_SHIFT of 330 * extflags; the next character is in bit +1, etc. */ 331 #define INT_PAT_MODS STD_PAT_MODS KEEPCOPY_PAT_MODS 332 333 #define EXT_PAT_MODS ONCE_PAT_MODS KEEPCOPY_PAT_MODS 334 #define QR_PAT_MODS STD_PAT_MODS EXT_PAT_MODS CHARSET_PAT_MODS 335 #define M_PAT_MODS QR_PAT_MODS LOOP_PAT_MODS 336 #define S_PAT_MODS M_PAT_MODS EXEC_PAT_MODS NONDESTRUCT_PAT_MODS 337 338 /* 339 * NOTE: if you modify any RXf flags you should run regen.pl or 340 * regen/regcomp.pl so that regnodes.h is updated with the changes. 341 * 342 */ 343 344 /* Leave some space, so future bit allocations can go either in the shared or 345 * unshared area without affecting binary compatibility */ 346 #define RXf_BASE_SHIFT (_RXf_PMf_SHIFT_NEXT) 347 348 /* 349 Set in Perl_pmruntime if op_flags & OPf_SPECIAL, i.e. split. Will 350 be used by regex engines to check whether they should set 351 RXf_SKIPWHITE 352 */ 353 #define RXf_SPLIT (1<<(RXf_BASE_SHIFT-1)) 354 #if RXf_SPLIT != RXf_PMf_SPLIT 355 # error "RXf_SPLIT does not match RXf_PMf_SPLIT" 356 #endif 357 358 /* Manually decorate this function with gcc-style attributes just to 359 * avoid having to restructure the header files and their called order, 360 * as proto.h would have to be included before this file, and isn't */ 361 362 PERL_STATIC_INLINE const char * 363 get_regex_charset_name(const U32 flags, STRLEN* const lenp) 364 __attribute__warn_unused_result__; 365 366 #define MAX_CHARSET_NAME_LENGTH 2 367 368 PERL_STATIC_INLINE const char * 369 get_regex_charset_name(const U32 flags, STRLEN* const lenp) 370 { 371 /* Returns a string that corresponds to the name of the regex character set 372 * given by 'flags', and *lenp is set the length of that string, which 373 * cannot exceed MAX_CHARSET_NAME_LENGTH characters */ 374 375 *lenp = 1; 376 switch (get_regex_charset(flags)) { 377 case REGEX_DEPENDS_CHARSET: return DEPENDS_PAT_MODS; 378 case REGEX_LOCALE_CHARSET: return LOCALE_PAT_MODS; 379 case REGEX_UNICODE_CHARSET: return UNICODE_PAT_MODS; 380 case REGEX_ASCII_RESTRICTED_CHARSET: return ASCII_RESTRICT_PAT_MODS; 381 case REGEX_ASCII_MORE_RESTRICTED_CHARSET: 382 *lenp = 2; 383 return ASCII_MORE_RESTRICT_PAT_MODS; 384 default: 385 return "?"; /* Unknown */ 386 } 387 } 388 389 /* Do we have some sort of anchor? */ 390 #define RXf_IS_ANCHORED (1<<(RXf_BASE_SHIFT+0)) 391 #define RXf_UNUSED1 (1<<(RXf_BASE_SHIFT+1)) 392 #define RXf_UNUSED2 (1<<(RXf_BASE_SHIFT+2)) 393 #define RXf_UNUSED3 (1<<(RXf_BASE_SHIFT+3)) 394 #define RXf_UNUSED4 (1<<(RXf_BASE_SHIFT+4)) 395 #define RXf_UNUSED5 (1<<(RXf_BASE_SHIFT+5)) 396 397 /* What we have seen */ 398 #define RXf_NO_INPLACE_SUBST (1<<(RXf_BASE_SHIFT+6)) 399 #define RXf_EVAL_SEEN (1<<(RXf_BASE_SHIFT+7)) 400 #define RXf_UNUSED8 (1<<(RXf_BASE_SHIFT+8)) 401 402 /* Special */ 403 #define RXf_UNBOUNDED_QUANTIFIER_SEEN (1<<(RXf_BASE_SHIFT+9)) 404 #define RXf_CHECK_ALL (1<<(RXf_BASE_SHIFT+10)) 405 406 /* UTF8 related */ 407 #define RXf_MATCH_UTF8 (1<<(RXf_BASE_SHIFT+11)) /* $1 etc are utf8 */ 408 409 /* Intuit related */ 410 #define RXf_USE_INTUIT_NOML (1<<(RXf_BASE_SHIFT+12)) 411 #define RXf_USE_INTUIT_ML (1<<(RXf_BASE_SHIFT+13)) 412 #define RXf_INTUIT_TAIL (1<<(RXf_BASE_SHIFT+14)) 413 #define RXf_USE_INTUIT (RXf_USE_INTUIT_NOML|RXf_USE_INTUIT_ML) 414 415 /* Copy and tainted info */ 416 #define RXf_COPY_DONE (1<<(RXf_BASE_SHIFT+16)) 417 418 /* post-execution: $1 et al are tainted */ 419 #define RXf_TAINTED_SEEN (1<<(RXf_BASE_SHIFT+17)) 420 /* this pattern was tainted during compilation */ 421 #define RXf_TAINTED (1<<(RXf_BASE_SHIFT+18)) 422 423 /* Flags indicating special patterns */ 424 #define RXf_START_ONLY (1<<(RXf_BASE_SHIFT+19)) /* Pattern is /^/ */ 425 #define RXf_SKIPWHITE (1<<(RXf_BASE_SHIFT+20)) /* Pattern is for a split " " */ 426 #define RXf_WHITE (1<<(RXf_BASE_SHIFT+21)) /* Pattern is /\s+/ */ 427 #define RXf_NULL (1U<<(RXf_BASE_SHIFT+22)) /* Pattern is // */ 428 #if RXf_BASE_SHIFT+22 > 31 429 # error Too many RXf_PMf bits used. See regnodes.h for any spare in middle 430 #endif 431 432 /* 433 * NOTE: if you modify any RXf flags you should run regen.pl or 434 * regen/regcomp.pl so that regnodes.h is updated with the changes. 435 * 436 */ 437 438 #ifdef NO_TAINT_SUPPORT 439 # define RX_ISTAINTED(prog) 0 440 # define RX_TAINT_on(prog) NOOP 441 # define RXp_MATCH_TAINTED(prog) 0 442 # define RX_MATCH_TAINTED(prog) 0 443 # define RXp_MATCH_TAINTED_on(prog) NOOP 444 # define RX_MATCH_TAINTED_on(prog) NOOP 445 # define RX_MATCH_TAINTED_off(prog) NOOP 446 #else 447 # define RX_ISTAINTED(prog) (RX_EXTFLAGS(prog) & RXf_TAINTED) 448 # define RX_TAINT_on(prog) (RX_EXTFLAGS(prog) |= RXf_TAINTED) 449 # define RXp_MATCH_TAINTED(prog) (RXp_EXTFLAGS(prog) & RXf_TAINTED_SEEN) 450 # define RX_MATCH_TAINTED(prog) (RX_EXTFLAGS(prog) & RXf_TAINTED_SEEN) 451 # define RXp_MATCH_TAINTED_on(prog) (RXp_EXTFLAGS(prog) |= RXf_TAINTED_SEEN) 452 # define RX_MATCH_TAINTED_on(prog) (RX_EXTFLAGS(prog) |= RXf_TAINTED_SEEN) 453 # define RX_MATCH_TAINTED_off(prog) (RX_EXTFLAGS(prog) &= ~RXf_TAINTED_SEEN) 454 #endif 455 456 #define RX_HAS_CUTGROUP(prog) ((prog)->intflags & PREGf_CUTGROUP_SEEN) 457 #define RX_MATCH_TAINTED_set(prog, t) ((t) \ 458 ? RX_MATCH_TAINTED_on(prog) \ 459 : RX_MATCH_TAINTED_off(prog)) 460 461 #define RXp_MATCH_COPIED(prog) (RXp_EXTFLAGS(prog) & RXf_COPY_DONE) 462 #define RX_MATCH_COPIED(prog) (RX_EXTFLAGS(prog) & RXf_COPY_DONE) 463 #define RXp_MATCH_COPIED_on(prog) (RXp_EXTFLAGS(prog) |= RXf_COPY_DONE) 464 #define RX_MATCH_COPIED_on(prog) (RX_EXTFLAGS(prog) |= RXf_COPY_DONE) 465 #define RXp_MATCH_COPIED_off(prog) (RXp_EXTFLAGS(prog) &= ~RXf_COPY_DONE) 466 #define RX_MATCH_COPIED_off(prog) (RX_EXTFLAGS(prog) &= ~RXf_COPY_DONE) 467 #define RX_MATCH_COPIED_set(prog,t) ((t) \ 468 ? RX_MATCH_COPIED_on(prog) \ 469 : RX_MATCH_COPIED_off(prog)) 470 471 #define RXp_EXTFLAGS(rx) ((rx)->extflags) 472 #define RXp_COMPFLAGS(rx) ((rx)->compflags) 473 474 /* For source compatibility. We used to store these explicitly. */ 475 #define RX_PRECOMP(prog) (RX_WRAPPED(prog) + ReANY(prog)->pre_prefix) 476 #define RX_PRECOMP_const(prog) (RX_WRAPPED_const(prog) + ReANY(prog)->pre_prefix) 477 /* FIXME? Are we hardcoding too much here and constraining plugin extension 478 writers? Specifically, the value 1 assumes that the wrapped version always 479 has exactly one character at the end, a ')'. Will that always be true? */ 480 #define RX_PRELEN(prog) (RX_WRAPLEN(prog) - ReANY(prog)->pre_prefix - 1) 481 #define RX_WRAPPED(prog) ReANY(prog)->xpv_len_u.xpvlenu_pv 482 #define RX_WRAPPED_const(prog) ((const char *)RX_WRAPPED(prog)) 483 #define RX_WRAPLEN(prog) SvCUR(prog) 484 #define RX_CHECK_SUBSTR(prog) (ReANY(prog)->check_substr) 485 #define RX_REFCNT(prog) SvREFCNT(prog) 486 #define RX_EXTFLAGS(prog) RXp_EXTFLAGS(ReANY(prog)) 487 #define RX_COMPFLAGS(prog) RXp_COMPFLAGS(ReANY(prog)) 488 #define RX_ENGINE(prog) (ReANY(prog)->engine) 489 #define RX_SUBBEG(prog) (ReANY(prog)->subbeg) 490 #define RX_SUBOFFSET(prog) (ReANY(prog)->suboffset) 491 #define RX_SUBCOFFSET(prog) (ReANY(prog)->subcoffset) 492 #define RX_OFFS(prog) (ReANY(prog)->offs) 493 #define RX_NPARENS(prog) (ReANY(prog)->nparens) 494 #define RX_SUBLEN(prog) (ReANY(prog)->sublen) 495 #define RX_MINLEN(prog) (ReANY(prog)->minlen) 496 #define RX_MINLENRET(prog) (ReANY(prog)->minlenret) 497 #define RX_GOFS(prog) (ReANY(prog)->gofs) 498 #define RX_LASTPAREN(prog) (ReANY(prog)->lastparen) 499 #define RX_LASTCLOSEPAREN(prog) (ReANY(prog)->lastcloseparen) 500 #define RX_SAVED_COPY(prog) (ReANY(prog)->saved_copy) 501 /* last match was zero-length */ 502 #define RX_ZERO_LEN(prog) \ 503 (RX_OFFS(prog)[0].start + (SSize_t)RX_GOFS(prog) \ 504 == RX_OFFS(prog)[0].end) 505 506 #endif /* PLUGGABLE_RE_EXTENSION */ 507 508 /* Stuff that needs to be included in the pluggable extension goes below here */ 509 510 #ifdef PERL_ANY_COW 511 #define RX_MATCH_COPY_FREE(rx) \ 512 STMT_START {if (RX_SAVED_COPY(rx)) { \ 513 SV_CHECK_THINKFIRST_COW_DROP(RX_SAVED_COPY(rx)); \ 514 } \ 515 if (RX_MATCH_COPIED(rx)) { \ 516 Safefree(RX_SUBBEG(rx)); \ 517 RX_MATCH_COPIED_off(rx); \ 518 }} STMT_END 519 #else 520 #define RX_MATCH_COPY_FREE(rx) \ 521 STMT_START {if (RX_MATCH_COPIED(rx)) { \ 522 Safefree(RX_SUBBEG(rx)); \ 523 RX_MATCH_COPIED_off(rx); \ 524 }} STMT_END 525 #endif 526 527 #define RXp_MATCH_UTF8(prog) (RXp_EXTFLAGS(prog) & RXf_MATCH_UTF8) 528 #define RX_MATCH_UTF8(prog) (RX_EXTFLAGS(prog) & RXf_MATCH_UTF8) 529 #define RX_MATCH_UTF8_on(prog) (RX_EXTFLAGS(prog) |= RXf_MATCH_UTF8) 530 #define RX_MATCH_UTF8_off(prog) (RX_EXTFLAGS(prog) &= ~RXf_MATCH_UTF8) 531 #define RX_MATCH_UTF8_set(prog, t) ((t) \ 532 ? RX_MATCH_UTF8_on(prog) \ 533 : RX_MATCH_UTF8_off(prog)) 534 535 /* Whether the pattern stored at RX_WRAPPED is in UTF-8 */ 536 #define RX_UTF8(prog) SvUTF8(prog) 537 538 539 /* bits in flags arg of Perl_regexec_flags() */ 540 541 #define REXEC_COPY_STR 0x01 /* Need to copy the string for captures. */ 542 #define REXEC_CHECKED 0x02 /* re_intuit_start() already called. */ 543 #define REXEC_SCREAM 0x04 /* currently unused. */ 544 #define REXEC_IGNOREPOS 0x08 /* use stringarg, not pos(), for \G match */ 545 #define REXEC_NOT_FIRST 0x10 /* This is another iteration of //g: 546 no need to copy string again */ 547 548 /* under REXEC_COPY_STR, it's ok for the 549 engine (modulo PL_sawamperand etc) 550 to skip copying: ... */ 551 #define REXEC_COPY_SKIP_PRE 0x20 /* ...the $` part of the string, or */ 552 #define REXEC_COPY_SKIP_POST 0x40 /* ...the $' part of the string */ 553 #define REXEC_FAIL_ON_UNDERFLOW 0x80 /* fail the match if $& would start before 554 the start pos (so s/.\G// would fail 555 on second iteration */ 556 557 #if defined(__GNUC__) && !defined(PERL_GCC_BRACE_GROUPS_FORBIDDEN) 558 # define ReREFCNT_inc(re) \ 559 ({ \ 560 /* This is here to generate a casting warning if incorrect. */ \ 561 REGEXP *const _rerefcnt_inc = (re); \ 562 assert(SvTYPE(_rerefcnt_inc) == SVt_REGEXP); \ 563 SvREFCNT_inc(_rerefcnt_inc); \ 564 _rerefcnt_inc; \ 565 }) 566 # define ReREFCNT_dec(re) \ 567 ({ \ 568 /* This is here to generate a casting warning if incorrect. */ \ 569 REGEXP *const _rerefcnt_dec = (re); \ 570 SvREFCNT_dec(_rerefcnt_dec); \ 571 }) 572 #else 573 # define ReREFCNT_dec(re) SvREFCNT_dec(re) 574 # define ReREFCNT_inc(re) ((REGEXP *) SvREFCNT_inc(re)) 575 #endif 576 #define ReANY(re) S_ReANY((const REGEXP *)(re)) 577 578 /* FIXME for plugins. */ 579 580 #define FBMcf_TAIL_DOLLAR 1 581 #define FBMcf_TAIL_DOLLARM 2 582 #define FBMcf_TAIL_Z 4 583 #define FBMcf_TAIL_z 8 584 #define FBMcf_TAIL (FBMcf_TAIL_DOLLAR|FBMcf_TAIL_DOLLARM|FBMcf_TAIL_Z|FBMcf_TAIL_z) 585 586 #define FBMrf_MULTILINE 1 587 588 struct regmatch_state; 589 struct regmatch_slab; 590 591 /* like regmatch_info_aux, but contains extra fields only needed if the 592 * pattern contains (?{}). If used, is snuck into the second slot in the 593 * regmatch_state stack at the start of execution */ 594 595 typedef struct { 596 regexp *rex; 597 PMOP *curpm; /* saved PL_curpm */ 598 #ifdef PERL_ANY_COW 599 SV *saved_copy; /* saved saved_copy field from rex */ 600 #endif 601 char *subbeg; /* saved subbeg field from rex */ 602 STRLEN sublen; /* saved sublen field from rex */ 603 STRLEN suboffset; /* saved suboffset field from rex */ 604 STRLEN subcoffset; /* saved subcoffset field from rex */ 605 MAGIC *pos_magic; /* pos() magic attached to $_ */ 606 SSize_t pos; /* the original value of pos() in pos_magic */ 607 U8 pos_flags; /* flags to be restored; currently only MGf_BYTES*/ 608 } regmatch_info_aux_eval; 609 610 611 /* fields that logically live in regmatch_info, but which need cleaning 612 * up on croak(), and so are instead are snuck into the first slot in 613 * the regmatch_state stack at the start of execution */ 614 615 typedef struct { 616 regmatch_info_aux_eval *info_aux_eval; 617 struct regmatch_state *old_regmatch_state; /* saved PL_regmatch_state */ 618 struct regmatch_slab *old_regmatch_slab; /* saved PL_regmatch_slab */ 619 char *poscache; /* S-L cache of fail positions of WHILEMs */ 620 } regmatch_info_aux; 621 622 623 /* some basic information about the current match that is created by 624 * Perl_regexec_flags and then passed to regtry(), regmatch() etc. 625 * It is allocated as a local var on the stack, so nothing should be 626 * stored in it that needs preserving or clearing up on croak(). 627 * For that, see the aux_info and aux_info_eval members of the 628 * regmatch_state union. */ 629 630 typedef struct { 631 REGEXP *prog; /* the regex being executed */ 632 const char * strbeg; /* real start of string */ 633 char *strend; /* one byte beyond last char of match string */ 634 char *till; /* matches shorter than this fail (see minlen arg) */ 635 SV *sv; /* the SV string currently being matched */ 636 char *ganch; /* position of \G anchor */ 637 char *cutpoint; /* (*COMMIT) position (if any) */ 638 regmatch_info_aux *info_aux; /* extra fields that need cleanup */ 639 regmatch_info_aux_eval *info_aux_eval; /* extra saved state for (?{}) */ 640 I32 poscache_maxiter; /* how many whilems todo before S-L cache kicks in */ 641 I32 poscache_iter; /* current countdown from _maxiter to zero */ 642 STRLEN poscache_size; /* size of regmatch_info_aux.poscache */ 643 bool intuit; /* re_intuit_start() is the top-level caller */ 644 bool is_utf8_pat; /* regex is utf8 */ 645 bool is_utf8_target; /* string being matched is utf8 */ 646 bool warned; /* we have issued a recursion warning; no need for more */ 647 } regmatch_info; 648 649 650 /* structures for holding and saving the state maintained by regmatch() */ 651 652 #ifndef MAX_RECURSE_EVAL_NOCHANGE_DEPTH 653 #define MAX_RECURSE_EVAL_NOCHANGE_DEPTH 1000 654 #endif 655 656 typedef I32 CHECKPOINT; 657 658 typedef struct regmatch_state { 659 int resume_state; /* where to jump to on return */ 660 char *locinput; /* where to backtrack in string on failure */ 661 662 union { 663 664 /* the 'info_aux' and 'info_aux_eval' union members are cuckoos in 665 * the nest. They aren't saved backtrack state; rather they 666 * represent one or two extra chunks of data that need allocating 667 * at the start of a match. These fields would logically live in 668 * the regmatch_info struct, except that is allocated on the 669 * C stack, and these fields are all things that require cleanup 670 * after a croak(), when the stack is lost. 671 * As a convenience, we just use the first 1 or 2 regmatch_state 672 * slots to store this info, as we will be allocating a slab of 673 * these anyway. Otherwise we'd have to malloc and then free them, 674 * or allocate them on the save stack (where they will get 675 * realloced if the save stack grows). 676 * info_aux contains the extra fields that are always needed; 677 * info_aux_eval contains extra fields that only needed if 678 * the pattern contains code blocks 679 * We split them into two separate structs to avoid increasing 680 * the size of the union. 681 */ 682 683 regmatch_info_aux info_aux; 684 685 regmatch_info_aux_eval info_aux_eval; 686 687 /* this is a fake union member that matches the first element 688 * of each member that needs to store positive backtrack 689 * information */ 690 struct { 691 struct regmatch_state *prev_yes_state; 692 } yes; 693 694 /* branchlike members */ 695 /* this is a fake union member that matches the first elements 696 * of each member that needs to behave like a branch */ 697 struct { 698 /* this first element must match u.yes */ 699 struct regmatch_state *prev_yes_state; 700 U32 lastparen; 701 U32 lastcloseparen; 702 CHECKPOINT cp; 703 704 } branchlike; 705 706 struct { 707 /* the first elements must match u.branchlike */ 708 struct regmatch_state *prev_yes_state; 709 U32 lastparen; 710 U32 lastcloseparen; 711 CHECKPOINT cp; 712 713 regnode *next_branch; /* next branch node */ 714 } branch; 715 716 struct { 717 /* the first elements must match u.branchlike */ 718 struct regmatch_state *prev_yes_state; 719 U32 lastparen; 720 U32 lastcloseparen; 721 CHECKPOINT cp; 722 723 U32 accepted; /* how many accepting states left */ 724 bool longfold;/* saw a fold with a 1->n char mapping */ 725 U16 *jump; /* positive offsets from me */ 726 regnode *me; /* Which node am I - needed for jump tries*/ 727 U8 *firstpos;/* pos in string of first trie match */ 728 U32 firstchars;/* len in chars of firstpos from start */ 729 U16 nextword;/* next word to try */ 730 U16 topword; /* longest accepted word */ 731 } trie; 732 733 /* special types - these members are used to store state for special 734 regops like eval, if/then, lookaround and the markpoint state */ 735 struct { 736 /* this first element must match u.yes */ 737 struct regmatch_state *prev_yes_state; 738 struct regmatch_state *prev_eval; 739 struct regmatch_state *prev_curlyx; 740 REGEXP *prev_rex; 741 CHECKPOINT cp; /* remember current savestack indexes */ 742 CHECKPOINT lastcp; 743 U32 close_paren; /* which close bracket is our end */ 744 regnode *B; /* the node following us */ 745 } eval; 746 747 struct { 748 /* this first element must match u.yes */ 749 struct regmatch_state *prev_yes_state; 750 I32 wanted; 751 I32 logical; /* saved copy of 'logical' var */ 752 regnode *me; /* the IFMATCH/SUSPEND/UNLESSM node */ 753 } ifmatch; /* and SUSPEND/UNLESSM */ 754 755 struct { 756 /* this first element must match u.yes */ 757 struct regmatch_state *prev_yes_state; 758 struct regmatch_state *prev_mark; 759 SV* mark_name; 760 char *mark_loc; 761 } mark; 762 763 struct { 764 int val; 765 } keeper; 766 767 /* quantifiers - these members are used for storing state for 768 for the regops used to implement quantifiers */ 769 struct { 770 /* this first element must match u.yes */ 771 struct regmatch_state *prev_yes_state; 772 struct regmatch_state *prev_curlyx; /* previous cur_curlyx */ 773 regnode *me; /* the CURLYX node */ 774 regnode *B; /* the B node in /A*B/ */ 775 CHECKPOINT cp; /* remember current savestack index */ 776 bool minmod; 777 int parenfloor;/* how far back to strip paren data */ 778 779 /* these two are modified by WHILEM */ 780 int count; /* how many instances of A we've matched */ 781 char *lastloc;/* where previous A matched (0-len detect) */ 782 } curlyx; 783 784 struct { 785 /* this first element must match u.yes */ 786 struct regmatch_state *prev_yes_state; 787 struct regmatch_state *save_curlyx; 788 CHECKPOINT cp; /* remember current savestack indexes */ 789 CHECKPOINT lastcp; 790 char *save_lastloc; /* previous curlyx.lastloc */ 791 I32 cache_offset; 792 I32 cache_mask; 793 } whilem; 794 795 struct { 796 /* this first element must match u.yes */ 797 struct regmatch_state *prev_yes_state; 798 int c1, c2; /* case fold search */ 799 CHECKPOINT cp; 800 U32 lastparen; 801 U32 lastcloseparen; 802 I32 alen; /* length of first-matched A string */ 803 I32 count; 804 bool minmod; 805 regnode *A, *B; /* the nodes corresponding to /A*B/ */ 806 regnode *me; /* the curlym node */ 807 U8 c1_utf8[UTF8_MAXBYTES+1]; /* */ 808 U8 c2_utf8[UTF8_MAXBYTES+1]; 809 } curlym; 810 811 struct { 812 U32 paren; 813 CHECKPOINT cp; 814 U32 lastparen; 815 U32 lastcloseparen; 816 int c1, c2; /* case fold search */ 817 char *maxpos; /* highest possible point in string to match */ 818 char *oldloc; /* the previous locinput */ 819 int count; 820 int min, max; /* {m,n} */ 821 regnode *A, *B; /* the nodes corresponding to /A*B/ */ 822 U8 c1_utf8[UTF8_MAXBYTES+1]; /* */ 823 U8 c2_utf8[UTF8_MAXBYTES+1]; 824 } curly; /* and CURLYN/PLUS/STAR */ 825 826 } u; 827 } regmatch_state; 828 829 /* how many regmatch_state structs to allocate as a single slab. 830 * We do it in 4K blocks for efficiency. The "3" is 2 for the next/prev 831 * pointers, plus 1 for any mythical malloc overhead. */ 832 833 #define PERL_REGMATCH_SLAB_SLOTS \ 834 ((4096 - 3 * sizeof (void*)) / sizeof(regmatch_state)) 835 836 typedef struct regmatch_slab { 837 regmatch_state states[PERL_REGMATCH_SLAB_SLOTS]; 838 struct regmatch_slab *prev, *next; 839 } regmatch_slab; 840 841 842 843 /* 844 * Local variables: 845 * c-indentation-style: bsd 846 * c-basic-offset: 4 847 * indent-tabs-mode: nil 848 * End: 849 * 850 * ex: set ts=8 sts=4 sw=4 et: 851 */ 852