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 = for apidoc AyT||regexp 101 The regexp/REGEXP struct, see L<perlreapi> for further documentation 102 on the individual fields. The struct is ordered so that the most 103 commonly used fields are placed at the start. 104 105 Any patch that adds items to this struct will need to include 106 changes to F<sv.c> (C<Perl_re_dup()>) and F<regcomp.c> 107 (C<pregfree()>). This involves freeing or cloning items in the 108 regexp's data array based on the data item's type. 109 */ 110 111 typedef struct regexp { 112 _XPV_HEAD; 113 const struct regexp_engine* engine; /* what engine created this regexp? */ 114 REGEXP *mother_re; /* what re is this a lightweight copy of? */ 115 HV *paren_names; /* Optional hash of paren names */ 116 117 /*---------------------------------------------------------------------- 118 * Information about the match that the perl core uses to manage things 119 */ 120 121 U32 extflags; /* Flags used both externally and internally */ 122 U32 nparens; /* number of capture buffers */ 123 SSize_t minlen; /* minimum possible number of chars in string to match */ 124 SSize_t minlenret; /* mininum possible number of chars in $& */ 125 STRLEN gofs; /* chars left of pos that we search from */ 126 /* substring data about strings that must appear in 127 * the final match, used for optimisations */ 128 struct reg_substr_data *substrs; 129 130 /* private engine specific data */ 131 132 void *pprivate; /* Data private to the regex engine which 133 * created this object. */ 134 U32 intflags; /* Engine Specific Internal flags */ 135 136 /*---------------------------------------------------------------------- 137 * Data about the last/current match. These are modified during matching 138 */ 139 140 U32 lastparen; /* highest close paren matched ($+) */ 141 regexp_paren_pair *offs; /* Array of offsets for (@-) and (@+) */ 142 char **recurse_locinput; /* used to detect infinite recursion, XXX: move to internal */ 143 U32 lastcloseparen; /* last close paren matched ($^N) */ 144 145 /*---------------------------------------------------------------------- */ 146 147 /* offset from wrapped to the start of precomp */ 148 PERL_BITFIELD32 pre_prefix:4; 149 150 /* original flags used to compile the pattern, may differ from 151 * extflags in various ways */ 152 PERL_BITFIELD32 compflags:9; 153 154 /*---------------------------------------------------------------------- */ 155 156 char *subbeg; /* saved or original string so \digit works forever. */ 157 SV_SAVED_COPY /* If non-NULL, SV which is COW from original */ 158 SSize_t sublen; /* Length of string pointed by subbeg */ 159 SSize_t suboffset; /* byte offset of subbeg from logical start of str */ 160 SSize_t subcoffset; /* suboffset equiv, but in chars (for @-/@+) */ 161 SSize_t maxlen; /* minimum possible number of chars in string to match */ 162 163 /*---------------------------------------------------------------------- */ 164 165 166 CV *qr_anoncv; /* the anon sub wrapped round qr/(?{..})/ */ 167 } regexp; 168 169 170 # define RXp_PAREN_NAMES(rx) ((rx)->paren_names) 171 172 /* used for high speed searches */ 173 typedef struct re_scream_pos_data_s 174 { 175 char **scream_olds; /* match pos */ 176 SSize_t *scream_pos; /* Internal iterator of scream. */ 177 } re_scream_pos_data; 178 179 /* regexp_engine structure. This is the dispatch table for regexes. 180 * Any regex engine implementation must be able to build one of these. 181 */ 182 typedef struct regexp_engine { 183 REGEXP* (*comp) (pTHX_ SV * const pattern, U32 flags); 184 I32 (*exec) (pTHX_ REGEXP * const rx, char* stringarg, char* strend, 185 char* strbeg, SSize_t minend, SV* sv, 186 void* data, U32 flags); 187 char* (*intuit) (pTHX_ 188 REGEXP * const rx, 189 SV *sv, 190 const char * const strbeg, 191 char *strpos, 192 char *strend, 193 const U32 flags, 194 re_scream_pos_data *data); 195 SV* (*checkstr) (pTHX_ REGEXP * const rx); 196 void (*rxfree) (pTHX_ REGEXP * const rx); 197 void (*numbered_buff_FETCH) (pTHX_ REGEXP * const rx, const I32 paren, 198 SV * const sv); 199 void (*numbered_buff_STORE) (pTHX_ REGEXP * const rx, const I32 paren, 200 SV const * const value); 201 I32 (*numbered_buff_LENGTH) (pTHX_ REGEXP * const rx, const SV * const sv, 202 const I32 paren); 203 SV* (*named_buff) (pTHX_ REGEXP * const rx, SV * const key, 204 SV * const value, const U32 flags); 205 SV* (*named_buff_iter) (pTHX_ REGEXP * const rx, const SV * const lastkey, 206 const U32 flags); 207 SV* (*qr_package)(pTHX_ REGEXP * const rx); 208 # ifdef USE_ITHREADS 209 void* (*dupe) (pTHX_ REGEXP * const rx, CLONE_PARAMS *param); 210 # endif 211 REGEXP* (*op_comp) (pTHX_ SV ** const patternp, int pat_count, 212 OP *expr, const struct regexp_engine* eng, 213 REGEXP *old_re, 214 bool *is_bare_re, U32 orig_rx_flags, U32 pm_flags); 215 } regexp_engine; 216 217 /* 218 These are passed to the numbered capture variable callbacks as the 219 paren name. >= 1 is reserved for actual numbered captures, i.e. $1, 220 $2 etc. 221 */ 222 # define RX_BUFF_IDX_CARET_PREMATCH -5 /* ${^PREMATCH} */ 223 # define RX_BUFF_IDX_CARET_POSTMATCH -4 /* ${^POSTMATCH} */ 224 # define RX_BUFF_IDX_CARET_FULLMATCH -3 /* ${^MATCH} */ 225 # define RX_BUFF_IDX_PREMATCH -2 /* $` */ 226 # define RX_BUFF_IDX_POSTMATCH -1 /* $' */ 227 # define RX_BUFF_IDX_FULLMATCH 0 /* $& */ 228 229 /* 230 Flags that are passed to the named_buff and named_buff_iter 231 callbacks above. Those routines are called from universal.c via the 232 Tie::Hash::NamedCapture interface for %+ and %- and the re:: 233 functions in the same file. 234 */ 235 236 /* The Tie::Hash::NamedCapture operation this is part of, if any */ 237 # define RXapif_FETCH 0x0001 238 # define RXapif_STORE 0x0002 239 # define RXapif_DELETE 0x0004 240 # define RXapif_CLEAR 0x0008 241 # define RXapif_EXISTS 0x0010 242 # define RXapif_SCALAR 0x0020 243 # define RXapif_FIRSTKEY 0x0040 244 # define RXapif_NEXTKEY 0x0080 245 246 /* Whether %+ or %- is being operated on */ 247 # define RXapif_ONE 0x0100 /* %+ */ 248 # define RXapif_ALL 0x0200 /* %- */ 249 250 /* Whether this is being called from a re:: function */ 251 # define RXapif_REGNAME 0x0400 252 # define RXapif_REGNAMES 0x0800 253 # define RXapif_REGNAMES_COUNT 0x1000 254 255 /* 256 =for apidoc Am|REGEXP *|SvRX|SV *sv 257 258 Convenience macro to get the REGEXP from a SV. This is approximately 259 equivalent to the following snippet: 260 261 if (SvMAGICAL(sv)) 262 mg_get(sv); 263 if (SvROK(sv)) 264 sv = MUTABLE_SV(SvRV(sv)); 265 if (SvTYPE(sv) == SVt_REGEXP) 266 return (REGEXP*) sv; 267 268 C<NULL> will be returned if a REGEXP* is not found. 269 270 =for apidoc Am|bool|SvRXOK|SV* sv 271 272 Returns a boolean indicating whether the SV (or the one it references) 273 is a REGEXP. 274 275 If you want to do something with the REGEXP* later use SvRX instead 276 and check for NULL. 277 278 =cut 279 */ 280 281 # define SvRX(sv) (Perl_get_re_arg(aTHX_ sv)) 282 # define SvRXOK(sv) cBOOL(Perl_get_re_arg(aTHX_ sv)) 283 284 285 /* Flags stored in regexp->extflags 286 * These are used by code external to the regexp engine 287 * 288 * Note that the flags whose names start with RXf_PMf_ are defined in 289 * op_reg_common.h, being copied from the parallel flags of op_pmflags 290 * 291 * NOTE: if you modify any RXf flags you should run regen.pl or 292 * regen/regcomp.pl so that regnodes.h is updated with the changes. 293 * 294 */ 295 296 # include "op_reg_common.h" 297 298 # define RXf_PMf_STD_PMMOD (RXf_PMf_MULTILINE|RXf_PMf_SINGLELINE|RXf_PMf_FOLD|RXf_PMf_EXTENDED|RXf_PMf_EXTENDED_MORE|RXf_PMf_NOCAPTURE) 299 300 # define CASE_STD_PMMOD_FLAGS_PARSE_SET(pmfl, x_count) \ 301 case IGNORE_PAT_MOD: *(pmfl) |= RXf_PMf_FOLD; break; \ 302 case MULTILINE_PAT_MOD: *(pmfl) |= RXf_PMf_MULTILINE; break; \ 303 case SINGLE_PAT_MOD: *(pmfl) |= RXf_PMf_SINGLELINE; break; \ 304 case XTENDED_PAT_MOD: if (x_count == 0) { \ 305 *(pmfl) |= RXf_PMf_EXTENDED; \ 306 *(pmfl) &= ~RXf_PMf_EXTENDED_MORE; \ 307 } \ 308 else { \ 309 *(pmfl) |= RXf_PMf_EXTENDED \ 310 |RXf_PMf_EXTENDED_MORE; \ 311 } \ 312 (x_count)++; break; \ 313 case NOCAPTURE_PAT_MOD: *(pmfl) |= RXf_PMf_NOCAPTURE; break; 314 315 /* Note, includes charset ones, assumes 0 is the default for them */ 316 # define STD_PMMOD_FLAGS_CLEAR(pmfl) \ 317 *(pmfl) &= ~(RXf_PMf_FOLD|RXf_PMf_MULTILINE|RXf_PMf_SINGLELINE|RXf_PMf_EXTENDED|RXf_PMf_EXTENDED_MORE|RXf_PMf_CHARSET|RXf_PMf_NOCAPTURE) 318 319 /* chars and strings used as regex pattern modifiers 320 * Singular is a 'c'har, plural is a "string" 321 * 322 * NOTE, KEEPCOPY was originally 'k', but was changed to 'p' for preserve 323 * for compatibility reasons with Regexp::Common which highjacked (?k:...) 324 * for its own uses. So 'k' is out as well. 325 */ 326 # define DEFAULT_PAT_MOD '^' /* Short for all the default modifiers */ 327 # define EXEC_PAT_MOD 'e' 328 # define KEEPCOPY_PAT_MOD 'p' 329 # define NOCAPTURE_PAT_MOD 'n' 330 # define ONCE_PAT_MOD 'o' 331 # define GLOBAL_PAT_MOD 'g' 332 # define CONTINUE_PAT_MOD 'c' 333 # define MULTILINE_PAT_MOD 'm' 334 # define SINGLE_PAT_MOD 's' 335 # define IGNORE_PAT_MOD 'i' 336 # define XTENDED_PAT_MOD 'x' 337 # define NONDESTRUCT_PAT_MOD 'r' 338 # define LOCALE_PAT_MOD 'l' 339 # define UNICODE_PAT_MOD 'u' 340 # define DEPENDS_PAT_MOD 'd' 341 # define ASCII_RESTRICT_PAT_MOD 'a' 342 343 # define ONCE_PAT_MODS "o" 344 # define KEEPCOPY_PAT_MODS "p" 345 # define NOCAPTURE_PAT_MODS "n" 346 # define EXEC_PAT_MODS "e" 347 # define LOOP_PAT_MODS "gc" 348 # define NONDESTRUCT_PAT_MODS "r" 349 # define LOCALE_PAT_MODS "l" 350 # define UNICODE_PAT_MODS "u" 351 # define DEPENDS_PAT_MODS "d" 352 # define ASCII_RESTRICT_PAT_MODS "a" 353 # define ASCII_MORE_RESTRICT_PAT_MODS "aa" 354 355 /* This string is expected by regcomp.c to be ordered so that the first 356 * character is the flag in bit RXf_PMf_STD_PMMOD_SHIFT of extflags; the next 357 * character is bit +1, etc. */ 358 # define STD_PAT_MODS "msixxn" 359 360 # define CHARSET_PAT_MODS ASCII_RESTRICT_PAT_MODS DEPENDS_PAT_MODS LOCALE_PAT_MODS UNICODE_PAT_MODS 361 362 /* This string is expected by XS_re_regexp_pattern() in universal.c to be ordered 363 * so that the first character is the flag in bit RXf_PMf_STD_PMMOD_SHIFT of 364 * extflags; the next character is in bit +1, etc. */ 365 # define INT_PAT_MODS STD_PAT_MODS KEEPCOPY_PAT_MODS 366 367 # define EXT_PAT_MODS ONCE_PAT_MODS KEEPCOPY_PAT_MODS NOCAPTURE_PAT_MODS 368 # define QR_PAT_MODS STD_PAT_MODS EXT_PAT_MODS CHARSET_PAT_MODS 369 # define M_PAT_MODS QR_PAT_MODS LOOP_PAT_MODS 370 # define S_PAT_MODS M_PAT_MODS EXEC_PAT_MODS NONDESTRUCT_PAT_MODS 371 372 /* 373 * NOTE: if you modify any RXf flags you should run regen.pl or 374 * regen/regcomp.pl so that regnodes.h is updated with the changes. 375 * 376 */ 377 378 /* 379 Set in Perl_pmruntime for a split. Will be used by regex engines to 380 check whether they should set RXf_SKIPWHITE 381 */ 382 # define RXf_SPLIT RXf_PMf_SPLIT 383 384 /* Currently the regex flags occupy a single 32-bit word. Not all bits are 385 * currently used. The lower bits are shared with their corresponding PMf flag 386 * bits, up to but not including _RXf_PMf_SHIFT_NEXT. The unused bits 387 * immediately follow; finally the used RXf-only (unshared) bits, so that the 388 * highest bit in the word is used. This gathers all the unused bits as a pool 389 * in the middle, like so: 11111111111111110000001111111111 390 * where the '1's represent used bits, and the '0's unused. This design allows 391 * us to allocate off one end of the pool if we need to add a shared bit, and 392 * off the other end if we need a non-shared bit, without disturbing the other 393 * bits. This maximizes the likelihood of being able to change things without 394 * breaking binary compatibility. 395 * 396 * To add shared bits, do so in op_reg_common.h. This should change 397 * _RXf_PMf_SHIFT_NEXT so that things won't compile. Then come to regexp.h and 398 * op.h and adjust the constant adders in the definitions of RXf_BASE_SHIFT and 399 * Pmf_BASE_SHIFT down by the number of shared bits you added. That's it. 400 * Things should be binary compatible. But if either of these gets to having 401 * to subtract rather than add, leave at 0 and instead adjust all the entries 402 * that are in terms of it. But if the first one of those is already 403 * RXf_BASE_SHIFT+0, there are no bits left, and a redesign is in order. 404 * 405 * To remove unshared bits, just delete its entry. If you're where breaking 406 * binary compatibility is ok to do, you might want to adjust things to move 407 * the newly opened space so that it gets absorbed into the common pool. 408 * 409 * To add unshared bits, first use up any gaps in the middle. Otherwise, 410 * allocate off the low end until you get to RXf_BASE_SHIFT+0. If that isn't 411 * enough, move RXf_BASE_SHIFT down (if possible) and add the new bit at the 412 * other end instead; this preserves binary compatibility. 413 * 414 * For the regexp bits, PL_reg_extflags_name[] in regnodes.h has a comment 415 * giving which bits are used/unused */ 416 417 # define RXf_BASE_SHIFT (_RXf_PMf_SHIFT_NEXT + 2) 418 419 /* What we have seen */ 420 # define RXf_NO_INPLACE_SUBST (1U<<(RXf_BASE_SHIFT+2)) 421 # define RXf_EVAL_SEEN (1U<<(RXf_BASE_SHIFT+3)) 422 423 /* Special */ 424 # define RXf_UNBOUNDED_QUANTIFIER_SEEN (1U<<(RXf_BASE_SHIFT+4)) 425 # define RXf_CHECK_ALL (1U<<(RXf_BASE_SHIFT+5)) 426 427 /* UTF8 related */ 428 # define RXf_MATCH_UTF8 (1U<<(RXf_BASE_SHIFT+6)) /* $1 etc are utf8 */ 429 430 /* Intuit related */ 431 # define RXf_USE_INTUIT_NOML (1U<<(RXf_BASE_SHIFT+7)) 432 # define RXf_USE_INTUIT_ML (1U<<(RXf_BASE_SHIFT+8)) 433 # define RXf_INTUIT_TAIL (1U<<(RXf_BASE_SHIFT+9)) 434 # define RXf_USE_INTUIT (RXf_USE_INTUIT_NOML|RXf_USE_INTUIT_ML) 435 436 /* Do we have some sort of anchor? */ 437 # define RXf_IS_ANCHORED (1U<<(RXf_BASE_SHIFT+10)) 438 439 /* Copy and tainted info */ 440 # define RXf_COPY_DONE (1U<<(RXf_BASE_SHIFT+11)) 441 442 /* post-execution: $1 et al are tainted */ 443 # define RXf_TAINTED_SEEN (1U<<(RXf_BASE_SHIFT+12)) 444 /* this pattern was tainted during compilation */ 445 # define RXf_TAINTED (1U<<(RXf_BASE_SHIFT+13)) 446 447 /* Flags indicating special patterns */ 448 # define RXf_START_ONLY (1U<<(RXf_BASE_SHIFT+14)) /* Pattern is /^/ */ 449 # define RXf_SKIPWHITE (1U<<(RXf_BASE_SHIFT+15)) /* Pattern is for a */ 450 /* split " " */ 451 # define RXf_WHITE (1U<<(RXf_BASE_SHIFT+16)) /* Pattern is /\s+/ */ 452 # define RXf_NULL (1U<<(RXf_BASE_SHIFT+17)) /* Pattern is // */ 453 454 /* See comments at the beginning of these defines about adding bits. The 455 * highest bit position should be used, so that if RXf_BASE_SHIFT gets 456 * increased, the #error below will be triggered so that you will be reminded 457 * to adjust things at the other end to keep the bit positions unchanged */ 458 # if RXf_BASE_SHIFT+17 > 31 459 # error Too many RXf_PMf bits used. See comments at beginning of these for what to do 460 # endif 461 462 /* 463 * NOTE: if you modify any RXf flags you should run regen.pl or 464 * regen/regcomp.pl so that regnodes.h is updated with the changes. 465 * 466 */ 467 468 # ifdef NO_TAINT_SUPPORT 469 # define RX_ISTAINTED(rx_sv) 0 470 # define RXp_ISTAINTED(prog) 0 471 # define RX_TAINT_on(rx_sv) NOOP 472 # define RXp_MATCH_TAINTED(prog) 0 473 # define RX_MATCH_TAINTED(rx_sv) 0 474 # define RXp_MATCH_TAINTED_on(prog) NOOP 475 # define RX_MATCH_TAINTED_on(rx_sv) NOOP 476 # define RXp_MATCH_TAINTED_off(prog) NOOP 477 # define RX_MATCH_TAINTED_off(rx_sv) NOOP 478 # else 479 # define RX_ISTAINTED(rx_sv) (RX_EXTFLAGS(rx_sv) & RXf_TAINTED) 480 # define RXp_ISTAINTED(prog) (RXp_EXTFLAGS(prog) & RXf_TAINTED) 481 # define RX_TAINT_on(rx_sv) (RX_EXTFLAGS(rx_sv) |= RXf_TAINTED) 482 # define RXp_MATCH_TAINTED(prog) (RXp_EXTFLAGS(prog) & RXf_TAINTED_SEEN) 483 # define RX_MATCH_TAINTED(rx_sv) (RX_EXTFLAGS(rx_sv) & RXf_TAINTED_SEEN) 484 # define RXp_MATCH_TAINTED_on(prog) (RXp_EXTFLAGS(prog) |= RXf_TAINTED_SEEN) 485 # define RX_MATCH_TAINTED_on(rx_sv) (RX_EXTFLAGS(rx_sv) |= RXf_TAINTED_SEEN) 486 # define RXp_MATCH_TAINTED_off(prog) (RXp_EXTFLAGS(prog) &= ~RXf_TAINTED_SEEN) 487 # define RX_MATCH_TAINTED_off(rx_sv) (RX_EXTFLAGS(rx_sv) &= ~RXf_TAINTED_SEEN) 488 # endif 489 490 # define RXp_HAS_CUTGROUP(prog) ((prog)->intflags & PREGf_CUTGROUP_SEEN) 491 492 # define RX_MATCH_TAINTED_set(rx_sv, t) ((t) \ 493 ? RX_MATCH_TAINTED_on(rx_sv) \ 494 : RX_MATCH_TAINTED_off(rx_sv)) 495 496 # define RXp_MATCH_COPIED(prog) (RXp_EXTFLAGS(prog) & RXf_COPY_DONE) 497 # define RX_MATCH_COPIED(rx_sv) (RX_EXTFLAGS(rx_sv) & RXf_COPY_DONE) 498 # define RXp_MATCH_COPIED_on(prog) (RXp_EXTFLAGS(prog) |= RXf_COPY_DONE) 499 # define RX_MATCH_COPIED_on(rx_sv) (RX_EXTFLAGS(rx_sv) |= RXf_COPY_DONE) 500 # define RXp_MATCH_COPIED_off(prog) (RXp_EXTFLAGS(prog) &= ~RXf_COPY_DONE) 501 # define RX_MATCH_COPIED_off(rx_sv) (RX_EXTFLAGS(rx_sv) &= ~RXf_COPY_DONE) 502 # define RX_MATCH_COPIED_set(rx_sv,t) ((t) \ 503 ? RX_MATCH_COPIED_on(rx_sv) \ 504 : RX_MATCH_COPIED_off(rx_sv)) 505 506 # define RXp_EXTFLAGS(rx) ((rx)->extflags) 507 # define RXp_COMPFLAGS(rx) ((rx)->compflags) 508 509 /* For source compatibility. We used to store these explicitly. */ 510 # define RX_PRECOMP(rx_sv) (RX_WRAPPED(rx_sv) \ 511 + ReANY(rx_sv)->pre_prefix) 512 # define RX_PRECOMP_const(rx_sv) (RX_WRAPPED_const(rx_sv) \ 513 + ReANY(rx_sv)->pre_prefix) 514 /* FIXME? Are we hardcoding too much here and constraining plugin extension 515 writers? Specifically, the value 1 assumes that the wrapped version always 516 has exactly one character at the end, a ')'. Will that always be true? */ 517 # define RX_PRELEN(rx_sv) (RX_WRAPLEN(rx_sv) \ 518 - ReANY(rx_sv)->pre_prefix - 1) 519 520 # define RX_WRAPPED(rx_sv) SvPVX(rx_sv) 521 # define RX_WRAPPED_const(rx_sv) SvPVX_const(rx_sv) 522 # define RX_WRAPLEN(rx_sv) SvCUR(rx_sv) 523 # define RX_CHECK_SUBSTR(rx_sv) (ReANY(rx_sv)->check_substr) 524 # define RX_REFCNT(rx_sv) SvREFCNT(rx_sv) 525 # define RX_EXTFLAGS(rx_sv) RXp_EXTFLAGS(ReANY(rx_sv)) 526 # define RX_COMPFLAGS(rx_sv) RXp_COMPFLAGS(ReANY(rx_sv)) 527 # define RXp_ENGINE(prog) ((prog)->engine) 528 # define RX_ENGINE(rx_sv) (RXp_ENGINE(ReANY(rx_sv))) 529 # define RXp_SUBBEG(prog) (prog->subbeg) 530 # define RX_SUBBEG(rx_sv) (RXp_SUBBEG(ReANY(rx_sv))) 531 # define RXp_SUBOFFSET(prog) (prog->suboffset) 532 # define RX_SUBOFFSET(rx_sv) (RXp_SUBOFFSET(ReANY(rx_sv))) 533 # define RX_SUBCOFFSET(rx_sv) (ReANY(rx_sv)->subcoffset) 534 # define RXp_OFFS(prog) (prog->offs) 535 # define RX_OFFS(rx_sv) (RXp_OFFS(ReANY(rx_sv))) 536 # define RXp_NPARENS(prog) (prog->nparens) 537 # define RX_NPARENS(rx_sv) (RXp_NPARENS(ReANY(rx_sv))) 538 # define RX_SUBLEN(rx_sv) (ReANY(rx_sv)->sublen) 539 # define RXp_MINLEN(prog) (prog->minlen) 540 # define RX_MINLEN(rx_sv) (RXp_MINLEN(ReANY(rx_sv))) 541 # define RXp_MINLENRET(prog) (prog->minlenret) 542 # define RX_MINLENRET(rx_sv) (RXp_MINLENRET(ReANY(rx_sv))) 543 # define RXp_GOFS(prog) (prog->gofs) 544 # define RX_GOFS(rx_sv) (RXp_GOFS(ReANY(rx_sv))) 545 # define RX_LASTPAREN(rx_sv) (ReANY(rx_sv)->lastparen) 546 # define RX_LASTCLOSEPAREN(rx_sv) (ReANY(rx_sv)->lastcloseparen) 547 # define RXp_SAVED_COPY(prog) (prog->saved_copy) 548 # define RX_SAVED_COPY(rx_sv) (RXp_SAVED_COPY(ReANY(rx_sv))) 549 /* last match was zero-length */ 550 # define RXp_ZERO_LEN(prog) \ 551 (RXp_OFFS(prog)[0].start + (SSize_t)RXp_GOFS(prog) \ 552 == RXp_OFFS(prog)[0].end) 553 # define RX_ZERO_LEN(rx_sv) (RXp_ZERO_LEN(ReANY(rx_sv))) 554 555 #endif /* PLUGGABLE_RE_EXTENSION */ 556 557 /* Stuff that needs to be included in the pluggable extension goes below here */ 558 559 #ifdef PERL_ANY_COW 560 # define RXp_MATCH_COPY_FREE(prog) \ 561 STMT_START {if (RXp_SAVED_COPY(prog)) { \ 562 SV_CHECK_THINKFIRST_COW_DROP(RXp_SAVED_COPY(prog)); \ 563 } \ 564 if (RXp_MATCH_COPIED(prog)) { \ 565 Safefree(RXp_SUBBEG(prog)); \ 566 RXp_MATCH_COPIED_off(prog); \ 567 }} STMT_END 568 #else 569 # define RXp_MATCH_COPY_FREE(prog) \ 570 STMT_START {if (RXp_MATCH_COPIED(prog)) { \ 571 Safefree(RXp_SUBBEG(prog)); \ 572 RXp_MATCH_COPIED_off(prog); \ 573 }} STMT_END 574 #endif 575 #define RX_MATCH_COPY_FREE(rx_sv) RXp_MATCH_COPY_FREE(ReANY(rx_sv)) 576 577 #define RXp_MATCH_UTF8(prog) (RXp_EXTFLAGS(prog) & RXf_MATCH_UTF8) 578 #define RX_MATCH_UTF8(rx_sv) (RX_EXTFLAGS(rx_sv) & RXf_MATCH_UTF8) 579 #define RXp_MATCH_UTF8_on(prog) (RXp_EXTFLAGS(prog) |= RXf_MATCH_UTF8) 580 #define RX_MATCH_UTF8_on(rx_sv) (RXp_MATCH_UTF8_on(ReANY(rx_sv))) 581 #define RXp_MATCH_UTF8_off(prog) (RXp_EXTFLAGS(prog) &= ~RXf_MATCH_UTF8) 582 #define RX_MATCH_UTF8_off(rx_sv) (RXp_MATCH_UTF8_off(ReANY(rx_sv)) 583 #define RXp_MATCH_UTF8_set(prog, t) ((t) \ 584 ? RXp_MATCH_UTF8_on(prog) \ 585 : RXp_MATCH_UTF8_off(prog)) 586 #define RX_MATCH_UTF8_set(rx_sv, t) (RXp_MATCH_UTF8_set(ReANY(rx_sv), t)) 587 588 /* Whether the pattern stored at RX_WRAPPED is in UTF-8 */ 589 #define RX_UTF8(rx_sv) SvUTF8(rx_sv) 590 591 592 /* bits in flags arg of Perl_regexec_flags() */ 593 594 #define REXEC_COPY_STR 0x01 /* Need to copy the string for captures. */ 595 #define REXEC_CHECKED 0x02 /* re_intuit_start() already called. */ 596 #define REXEC_SCREAM 0x04 /* currently unused. */ 597 #define REXEC_IGNOREPOS 0x08 /* use stringarg, not pos(), for \G match */ 598 #define REXEC_NOT_FIRST 0x10 /* This is another iteration of //g: 599 no need to copy string again */ 600 601 /* under REXEC_COPY_STR, it's ok for the 602 engine (modulo PL_sawamperand etc) 603 to skip copying: ... */ 604 #define REXEC_COPY_SKIP_PRE 0x20 /* ...the $` part of the string, or */ 605 #define REXEC_COPY_SKIP_POST 0x40 /* ...the $' part of the string */ 606 #define REXEC_FAIL_ON_UNDERFLOW 0x80 /* fail the match if $& would start before 607 the start pos (so s/.\G// would fail 608 on second iteration */ 609 610 #if defined(PERL_USE_GCC_BRACE_GROUPS) 611 # define ReREFCNT_inc(re) \ 612 ({ \ 613 /* This is here to generate a casting warning if incorrect. */ \ 614 REGEXP *const _rerefcnt_inc = (re); \ 615 assert(SvTYPE(_rerefcnt_inc) == SVt_REGEXP); \ 616 SvREFCNT_inc(_rerefcnt_inc); \ 617 _rerefcnt_inc; \ 618 }) 619 # define ReREFCNT_dec(re) \ 620 ({ \ 621 /* This is here to generate a casting warning if incorrect. */ \ 622 REGEXP *const _rerefcnt_dec = (re); \ 623 SvREFCNT_dec(_rerefcnt_dec); \ 624 }) 625 #else 626 # define ReREFCNT_dec(re) SvREFCNT_dec(re) 627 # define ReREFCNT_inc(re) ((REGEXP *) SvREFCNT_inc(re)) 628 #endif 629 #define ReANY(re) Perl_ReANY((const REGEXP *)(re)) 630 631 /* FIXME for plugins. */ 632 633 #define FBMcf_TAIL_DOLLAR 1 634 #define FBMcf_TAIL_DOLLARM 2 635 #define FBMcf_TAIL_Z 4 636 #define FBMcf_TAIL_z 8 637 #define FBMcf_TAIL (FBMcf_TAIL_DOLLAR|FBMcf_TAIL_DOLLARM|FBMcf_TAIL_Z|FBMcf_TAIL_z) 638 639 #define FBMrf_MULTILINE 1 640 641 struct regmatch_state; 642 struct regmatch_slab; 643 644 /* like regmatch_info_aux, but contains extra fields only needed if the 645 * pattern contains (?{}). If used, is snuck into the second slot in the 646 * regmatch_state stack at the start of execution */ 647 648 typedef struct { 649 regexp *rex; 650 PMOP *curpm; /* saved PL_curpm */ 651 #ifdef PERL_ANY_COW 652 SV *saved_copy; /* saved saved_copy field from rex */ 653 #endif 654 char *subbeg; /* saved subbeg field from rex */ 655 STRLEN sublen; /* saved sublen field from rex */ 656 STRLEN suboffset; /* saved suboffset field from rex */ 657 STRLEN subcoffset; /* saved subcoffset field from rex */ 658 SV *sv; /* $_ during (?{}) */ 659 MAGIC *pos_magic; /* pos() magic attached to $_ */ 660 SSize_t pos; /* the original value of pos() in pos_magic */ 661 U8 pos_flags; /* flags to be restored; currently only MGf_BYTES*/ 662 } regmatch_info_aux_eval; 663 664 665 /* fields that logically live in regmatch_info, but which need cleaning 666 * up on croak(), and so are instead are snuck into the first slot in 667 * the regmatch_state stack at the start of execution */ 668 669 typedef struct { 670 regmatch_info_aux_eval *info_aux_eval; 671 struct regmatch_state *old_regmatch_state; /* saved PL_regmatch_state */ 672 struct regmatch_slab *old_regmatch_slab; /* saved PL_regmatch_slab */ 673 char *poscache; /* S-L cache of fail positions of WHILEMs */ 674 } regmatch_info_aux; 675 676 677 /* 678 =for apidoc Ay||regmatch_info 679 Some basic information about the current match that is created by 680 Perl_regexec_flags and then passed to regtry(), regmatch() etc. 681 It is allocated as a local var on the stack, so nothing should be 682 stored in it that needs preserving or clearing up on croak(). 683 For that, see the aux_info and aux_info_eval members of the 684 regmatch_state union. 685 686 =cut 687 */ 688 689 typedef struct { 690 REGEXP *prog; /* the regex being executed */ 691 const char * strbeg; /* real start of string */ 692 char *strend; /* one byte beyond last char of match string */ 693 char *till; /* matches shorter than this fail (see minlen arg) */ 694 SV *sv; /* the SV string currently being matched */ 695 char *ganch; /* position of \G anchor */ 696 char *cutpoint; /* (*COMMIT) position (if any) */ 697 regmatch_info_aux *info_aux; /* extra fields that need cleanup */ 698 regmatch_info_aux_eval *info_aux_eval; /* extra saved state for (?{}) */ 699 I32 poscache_maxiter; /* how many whilems todo before S-L cache kicks in */ 700 I32 poscache_iter; /* current countdown from _maxiter to zero */ 701 STRLEN poscache_size; /* size of regmatch_info_aux.poscache */ 702 bool intuit; /* re_intuit_start() is the top-level caller */ 703 bool is_utf8_pat; /* regex is utf8 */ 704 bool is_utf8_target; /* string being matched is utf8 */ 705 bool warned; /* we have issued a recursion warning; no need for more */ 706 } regmatch_info; 707 708 709 /* structures for holding and saving the state maintained by regmatch() */ 710 711 #ifndef MAX_RECURSE_EVAL_NOCHANGE_DEPTH 712 # define MAX_RECURSE_EVAL_NOCHANGE_DEPTH 10 713 #endif 714 715 /* The +1 is because everything matches itself, which isn't included in 716 * MAX_FOLD_FROMS; the +2 is based on the current Unicode standards needs, and 717 * is unlikely to change. An assertion should fail in regexec.c if it is too 718 * low. It is needed for certain edge cases involving multi-character folds 719 * when the first component also participates in a fold individually. */ 720 #define MAX_MATCHES (MAX_FOLD_FROMS + 1 + 2) 721 722 struct next_matchable_info { 723 U8 first_byte_mask; 724 U8 first_byte_anded; 725 U32 mask32; 726 U32 anded32; 727 PERL_INT_FAST8_T count; /* Negative means not initialized */ 728 PERL_UINT_FAST8_T min_length; 729 PERL_UINT_FAST8_T max_length; 730 PERL_UINT_FAST8_T initial_definitive; 731 PERL_UINT_FAST8_T initial_exact; 732 PERL_UINT_FAST8_T lengths[MAX_MATCHES]; 733 734 /* The size is from trial and error, and could change with new Unicode 735 * standards, in which case there is an assertion that should start 736 * failing. This size could be calculated in one of the regen scripts 737 * dealing with Unicode, but khw thinks the likelihood of it changing is 738 * low enough that it isn't worth the effort. */ 739 U8 matches[18]; 740 }; 741 742 typedef I32 CHECKPOINT; 743 744 typedef struct regmatch_state { 745 int resume_state; /* where to jump to on return */ 746 char *locinput; /* where to backtrack in string on failure */ 747 char *loceol; 748 U8 *sr0; /* position of start of script run, or NULL */ 749 750 union { 751 752 /* the 'info_aux' and 'info_aux_eval' union members are cuckoos in 753 * the nest. They aren't saved backtrack state; rather they 754 * represent one or two extra chunks of data that need allocating 755 * at the start of a match. These fields would logically live in 756 * the regmatch_info struct, except that is allocated on the 757 * C stack, and these fields are all things that require cleanup 758 * after a croak(), when the stack is lost. 759 * As a convenience, we just use the first 1 or 2 regmatch_state 760 * slots to store this info, as we will be allocating a slab of 761 * these anyway. Otherwise we'd have to malloc and then free them, 762 * or allocate them on the save stack (where they will get 763 * realloced if the save stack grows). 764 * info_aux contains the extra fields that are always needed; 765 * info_aux_eval contains extra fields that only needed if 766 * the pattern contains code blocks 767 * We split them into two separate structs to avoid increasing 768 * the size of the union. 769 */ 770 771 regmatch_info_aux info_aux; 772 773 regmatch_info_aux_eval info_aux_eval; 774 775 /* this is a fake union member that matches the first element 776 * of each member that needs to store positive backtrack 777 * information */ 778 struct { 779 struct regmatch_state *prev_yes_state; 780 } yes; 781 782 /* branchlike members */ 783 /* this is a fake union member that matches the first elements 784 * of each member that needs to behave like a branch */ 785 struct { 786 /* this first element must match u.yes */ 787 struct regmatch_state *prev_yes_state; 788 U32 lastparen; 789 U32 lastcloseparen; 790 CHECKPOINT cp; 791 792 } branchlike; 793 794 struct { 795 /* the first elements must match u.branchlike */ 796 struct regmatch_state *prev_yes_state; 797 U32 lastparen; 798 U32 lastcloseparen; 799 CHECKPOINT cp; 800 801 regnode *next_branch; /* next branch node */ 802 } branch; 803 804 struct { 805 /* the first elements must match u.branchlike */ 806 struct regmatch_state *prev_yes_state; 807 U32 lastparen; 808 U32 lastcloseparen; 809 CHECKPOINT cp; 810 811 U32 accepted; /* how many accepting states left */ 812 bool longfold;/* saw a fold with a 1->n char mapping */ 813 U16 *jump; /* positive offsets from me */ 814 regnode *me; /* Which node am I - needed for jump tries*/ 815 U8 *firstpos;/* pos in string of first trie match */ 816 U32 firstchars;/* len in chars of firstpos from start */ 817 U16 nextword;/* next word to try */ 818 U16 topword; /* longest accepted word */ 819 } trie; 820 821 /* special types - these members are used to store state for special 822 regops like eval, if/then, lookaround and the markpoint state */ 823 struct { 824 /* this first element must match u.yes */ 825 struct regmatch_state *prev_yes_state; 826 struct regmatch_state *prev_curlyx; 827 struct regmatch_state *prev_eval; 828 REGEXP *prev_rex; 829 CHECKPOINT cp; /* remember current savestack indexes */ 830 CHECKPOINT lastcp; 831 U32 close_paren; /* which close bracket is our end (+1) */ 832 regnode *B; /* the node following us */ 833 char *prev_recurse_locinput; 834 } eval; 835 836 struct { 837 /* this first element must match u.yes */ 838 struct regmatch_state *prev_yes_state; 839 I32 wanted; 840 I32 logical; /* saved copy of 'logical' var */ 841 U8 count; /* number of beginning positions */ 842 char *start; 843 char *end; 844 regnode *me; /* the IFMATCH/SUSPEND/UNLESSM node */ 845 } ifmatch; /* and SUSPEND/UNLESSM */ 846 847 struct { 848 /* this first element must match u.yes */ 849 struct regmatch_state *prev_yes_state; 850 struct regmatch_state *prev_mark; 851 SV* mark_name; 852 char *mark_loc; 853 } mark; 854 855 struct { 856 int val; 857 } keeper; 858 859 /* quantifiers - these members are used for storing state for 860 the regops used to implement quantifiers */ 861 struct { 862 /* this first element must match u.yes */ 863 struct regmatch_state *prev_yes_state; 864 struct regmatch_state *prev_curlyx; /* previous cur_curlyx */ 865 regnode *me; /* the CURLYX node */ 866 regnode *B; /* the B node in /A*B/ */ 867 CHECKPOINT cp; /* remember current savestack index */ 868 bool minmod; 869 int parenfloor;/* how far back to strip paren data */ 870 871 /* these two are modified by WHILEM */ 872 int count; /* how many instances of A we've matched */ 873 char *lastloc;/* where previous A matched (0-len detect) */ 874 } curlyx; 875 876 struct { 877 /* this first element must match u.yes */ 878 struct regmatch_state *prev_yes_state; 879 struct regmatch_state *save_curlyx; 880 CHECKPOINT cp; /* remember current savestack indexes */ 881 CHECKPOINT lastcp; 882 char *save_lastloc; /* previous curlyx.lastloc */ 883 I32 cache_offset; 884 I32 cache_mask; 885 } whilem; 886 887 struct { 888 /* this first element must match u.yes */ 889 struct regmatch_state *prev_yes_state; 890 CHECKPOINT cp; 891 U32 lastparen; 892 U32 lastcloseparen; 893 I32 alen; /* length of first-matched A string */ 894 I32 count; 895 bool minmod; 896 regnode *A, *B; /* the nodes corresponding to /A*B/ */ 897 regnode *me; /* the curlym node */ 898 struct next_matchable_info Binfo; 899 } curlym; 900 901 struct { 902 U32 paren; 903 CHECKPOINT cp; 904 U32 lastparen; 905 U32 lastcloseparen; 906 char *maxpos; /* highest possible point in string to match */ 907 char *oldloc; /* the previous locinput */ 908 int count; 909 int min, max; /* {m,n} */ 910 regnode *A, *B; /* the nodes corresponding to /A*B/ */ 911 struct next_matchable_info Binfo; 912 } curly; /* and CURLYN/PLUS/STAR */ 913 914 } u; 915 } regmatch_state; 916 917 918 919 /* how many regmatch_state structs to allocate as a single slab. 920 * We do it in 4K blocks for efficiency. The "3" is 2 for the next/prev 921 * pointers, plus 1 for any mythical malloc overhead. */ 922 923 #define PERL_REGMATCH_SLAB_SLOTS \ 924 ((4096 - 3 * sizeof (void*)) / sizeof(regmatch_state)) 925 926 typedef struct regmatch_slab { 927 regmatch_state states[PERL_REGMATCH_SLAB_SLOTS]; 928 struct regmatch_slab *prev, *next; 929 } regmatch_slab; 930 931 932 933 /* 934 * ex: set ts=8 sts=4 sw=4 et: 935 */ 936