| 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 | /* Do we have some sort of anchor? */ |
| 359 | #define RXf_IS_ANCHORED (1<<(RXf_BASE_SHIFT+0)) |
| 360 | #define RXf_UNUSED1 (1<<(RXf_BASE_SHIFT+1)) |
| 361 | #define RXf_UNUSED2 (1<<(RXf_BASE_SHIFT+2)) |
| 362 | #define RXf_UNUSED3 (1<<(RXf_BASE_SHIFT+3)) |
| 363 | #define RXf_UNUSED4 (1<<(RXf_BASE_SHIFT+4)) |
| 364 | #define RXf_UNUSED5 (1<<(RXf_BASE_SHIFT+5)) |
| 365 | |
| 366 | /* What we have seen */ |
| 367 | #define RXf_NO_INPLACE_SUBST (1<<(RXf_BASE_SHIFT+6)) |
| 368 | #define RXf_EVAL_SEEN (1<<(RXf_BASE_SHIFT+7)) |
| 369 | #define RXf_UNUSED8 (1<<(RXf_BASE_SHIFT+8)) |
| 370 | |
| 371 | /* Special */ |
| 372 | #define RXf_UNBOUNDED_QUANTIFIER_SEEN (1<<(RXf_BASE_SHIFT+9)) |
| 373 | #define RXf_CHECK_ALL (1<<(RXf_BASE_SHIFT+10)) |
| 374 | |
| 375 | /* UTF8 related */ |
| 376 | #define RXf_MATCH_UTF8 (1<<(RXf_BASE_SHIFT+11)) /* $1 etc are utf8 */ |
| 377 | |
| 378 | /* Intuit related */ |
| 379 | #define RXf_USE_INTUIT_NOML (1<<(RXf_BASE_SHIFT+12)) |
| 380 | #define RXf_USE_INTUIT_ML (1<<(RXf_BASE_SHIFT+13)) |
| 381 | #define RXf_INTUIT_TAIL (1<<(RXf_BASE_SHIFT+14)) |
| 382 | #define RXf_USE_INTUIT (RXf_USE_INTUIT_NOML|RXf_USE_INTUIT_ML) |
| 383 | |
| 384 | /* Copy and tainted info */ |
| 385 | #define RXf_COPY_DONE (1<<(RXf_BASE_SHIFT+16)) |
| 386 | |
| 387 | /* post-execution: $1 et al are tainted */ |
| 388 | #define RXf_TAINTED_SEEN (1<<(RXf_BASE_SHIFT+17)) |
| 389 | /* this pattern was tainted during compilation */ |
| 390 | #define RXf_TAINTED (1<<(RXf_BASE_SHIFT+18)) |
| 391 | |
| 392 | /* Flags indicating special patterns */ |
| 393 | #define RXf_START_ONLY (1<<(RXf_BASE_SHIFT+19)) /* Pattern is /^/ */ |
| 394 | #define RXf_SKIPWHITE (1<<(RXf_BASE_SHIFT+20)) /* Pattern is for a split " " */ |
| 395 | #define RXf_WHITE (1<<(RXf_BASE_SHIFT+21)) /* Pattern is /\s+/ */ |
| 396 | #define RXf_NULL (1U<<(RXf_BASE_SHIFT+22)) /* Pattern is // */ |
| 397 | #if RXf_BASE_SHIFT+22 > 31 |
| 398 | # error Too many RXf_PMf bits used. See regnodes.h for any spare in middle |
| 399 | #endif |
| 400 | |
| 401 | /* |
| 402 | * NOTE: if you modify any RXf flags you should run regen.pl or |
| 403 | * regen/regcomp.pl so that regnodes.h is updated with the changes. |
| 404 | * |
| 405 | */ |
| 406 | |
| 407 | #ifdef NO_TAINT_SUPPORT |
| 408 | # define RX_ISTAINTED(prog) 0 |
| 409 | # define RX_TAINT_on(prog) NOOP |
| 410 | # define RXp_MATCH_TAINTED(prog) 0 |
| 411 | # define RX_MATCH_TAINTED(prog) 0 |
| 412 | # define RXp_MATCH_TAINTED_on(prog) NOOP |
| 413 | # define RX_MATCH_TAINTED_on(prog) NOOP |
| 414 | # define RX_MATCH_TAINTED_off(prog) NOOP |
| 415 | #else |
| 416 | # define RX_ISTAINTED(prog) (RX_EXTFLAGS(prog) & RXf_TAINTED) |
| 417 | # define RX_TAINT_on(prog) (RX_EXTFLAGS(prog) |= RXf_TAINTED) |
| 418 | # define RXp_MATCH_TAINTED(prog) (RXp_EXTFLAGS(prog) & RXf_TAINTED_SEEN) |
| 419 | # define RX_MATCH_TAINTED(prog) (RX_EXTFLAGS(prog) & RXf_TAINTED_SEEN) |
| 420 | # define RXp_MATCH_TAINTED_on(prog) (RXp_EXTFLAGS(prog) |= RXf_TAINTED_SEEN) |
| 421 | # define RX_MATCH_TAINTED_on(prog) (RX_EXTFLAGS(prog) |= RXf_TAINTED_SEEN) |
| 422 | # define RX_MATCH_TAINTED_off(prog) (RX_EXTFLAGS(prog) &= ~RXf_TAINTED_SEEN) |
| 423 | #endif |
| 424 | |
| 425 | #define RX_HAS_CUTGROUP(prog) ((prog)->intflags & PREGf_CUTGROUP_SEEN) |
| 426 | #define RX_MATCH_TAINTED_set(prog, t) ((t) \ |
| 427 | ? RX_MATCH_TAINTED_on(prog) \ |
| 428 | : RX_MATCH_TAINTED_off(prog)) |
| 429 | |
| 430 | #define RXp_MATCH_COPIED(prog) (RXp_EXTFLAGS(prog) & RXf_COPY_DONE) |
| 431 | #define RX_MATCH_COPIED(prog) (RX_EXTFLAGS(prog) & RXf_COPY_DONE) |
| 432 | #define RXp_MATCH_COPIED_on(prog) (RXp_EXTFLAGS(prog) |= RXf_COPY_DONE) |
| 433 | #define RX_MATCH_COPIED_on(prog) (RX_EXTFLAGS(prog) |= RXf_COPY_DONE) |
| 434 | #define RXp_MATCH_COPIED_off(prog) (RXp_EXTFLAGS(prog) &= ~RXf_COPY_DONE) |
| 435 | #define RX_MATCH_COPIED_off(prog) (RX_EXTFLAGS(prog) &= ~RXf_COPY_DONE) |
| 436 | #define RX_MATCH_COPIED_set(prog,t) ((t) \ |
| 437 | ? RX_MATCH_COPIED_on(prog) \ |
| 438 | : RX_MATCH_COPIED_off(prog)) |
| 439 | |
| 440 | #define RXp_EXTFLAGS(rx) ((rx)->extflags) |
| 441 | #define RXp_COMPFLAGS(rx) ((rx)->compflags) |
| 442 | |
| 443 | /* For source compatibility. We used to store these explicitly. */ |
| 444 | #define RX_PRECOMP(prog) (RX_WRAPPED(prog) + ReANY(prog)->pre_prefix) |
| 445 | #define RX_PRECOMP_const(prog) (RX_WRAPPED_const(prog) + ReANY(prog)->pre_prefix) |
| 446 | /* FIXME? Are we hardcoding too much here and constraining plugin extension |
| 447 | writers? Specifically, the value 1 assumes that the wrapped version always |
| 448 | has exactly one character at the end, a ')'. Will that always be true? */ |
| 449 | #define RX_PRELEN(prog) (RX_WRAPLEN(prog) - ReANY(prog)->pre_prefix - 1) |
| 450 | #define RX_WRAPPED(prog) ReANY(prog)->xpv_len_u.xpvlenu_pv |
| 451 | #define RX_WRAPPED_const(prog) ((const char *)RX_WRAPPED(prog)) |
| 452 | #define RX_WRAPLEN(prog) SvCUR(prog) |
| 453 | #define RX_CHECK_SUBSTR(prog) (ReANY(prog)->check_substr) |
| 454 | #define RX_REFCNT(prog) SvREFCNT(prog) |
| 455 | #define RX_EXTFLAGS(prog) RXp_EXTFLAGS(ReANY(prog)) |
| 456 | #define RX_COMPFLAGS(prog) RXp_COMPFLAGS(ReANY(prog)) |
| 457 | #define RX_ENGINE(prog) (ReANY(prog)->engine) |
| 458 | #define RX_SUBBEG(prog) (ReANY(prog)->subbeg) |
| 459 | #define RX_SUBOFFSET(prog) (ReANY(prog)->suboffset) |
| 460 | #define RX_SUBCOFFSET(prog) (ReANY(prog)->subcoffset) |
| 461 | #define RX_OFFS(prog) (ReANY(prog)->offs) |
| 462 | #define RX_NPARENS(prog) (ReANY(prog)->nparens) |
| 463 | #define RX_SUBLEN(prog) (ReANY(prog)->sublen) |
| 464 | #define RX_MINLEN(prog) (ReANY(prog)->minlen) |
| 465 | #define RX_MINLENRET(prog) (ReANY(prog)->minlenret) |
| 466 | #define RX_GOFS(prog) (ReANY(prog)->gofs) |
| 467 | #define RX_LASTPAREN(prog) (ReANY(prog)->lastparen) |
| 468 | #define RX_LASTCLOSEPAREN(prog) (ReANY(prog)->lastcloseparen) |
| 469 | #define RX_SAVED_COPY(prog) (ReANY(prog)->saved_copy) |
| 470 | /* last match was zero-length */ |
| 471 | #define RX_ZERO_LEN(prog) \ |
| 472 | (RX_OFFS(prog)[0].start + (SSize_t)RX_GOFS(prog) \ |
| 473 | == RX_OFFS(prog)[0].end) |
| 474 | |
| 475 | #endif /* PLUGGABLE_RE_EXTENSION */ |
| 476 | |
| 477 | /* Stuff that needs to be included in the pluggable extension goes below here */ |
| 478 | |
| 479 | #ifdef PERL_ANY_COW |
| 480 | #define RX_MATCH_COPY_FREE(rx) \ |
| 481 | STMT_START {if (RX_SAVED_COPY(rx)) { \ |
| 482 | SV_CHECK_THINKFIRST_COW_DROP(RX_SAVED_COPY(rx)); \ |
| 483 | } \ |
| 484 | if (RX_MATCH_COPIED(rx)) { \ |
| 485 | Safefree(RX_SUBBEG(rx)); \ |
| 486 | RX_MATCH_COPIED_off(rx); \ |
| 487 | }} STMT_END |
| 488 | #else |
| 489 | #define RX_MATCH_COPY_FREE(rx) \ |
| 490 | STMT_START {if (RX_MATCH_COPIED(rx)) { \ |
| 491 | Safefree(RX_SUBBEG(rx)); \ |
| 492 | RX_MATCH_COPIED_off(rx); \ |
| 493 | }} STMT_END |
| 494 | #endif |
| 495 | |
| 496 | #define RXp_MATCH_UTF8(prog) (RXp_EXTFLAGS(prog) & RXf_MATCH_UTF8) |
| 497 | #define RX_MATCH_UTF8(prog) (RX_EXTFLAGS(prog) & RXf_MATCH_UTF8) |
| 498 | #define RX_MATCH_UTF8_on(prog) (RX_EXTFLAGS(prog) |= RXf_MATCH_UTF8) |
| 499 | #define RX_MATCH_UTF8_off(prog) (RX_EXTFLAGS(prog) &= ~RXf_MATCH_UTF8) |
| 500 | #define RX_MATCH_UTF8_set(prog, t) ((t) \ |
| 501 | ? RX_MATCH_UTF8_on(prog) \ |
| 502 | : RX_MATCH_UTF8_off(prog)) |
| 503 | |
| 504 | /* Whether the pattern stored at RX_WRAPPED is in UTF-8 */ |
| 505 | #define RX_UTF8(prog) SvUTF8(prog) |
| 506 | |
| 507 | |
| 508 | /* bits in flags arg of Perl_regexec_flags() */ |
| 509 | |
| 510 | #define REXEC_COPY_STR 0x01 /* Need to copy the string for captures. */ |
| 511 | #define REXEC_CHECKED 0x02 /* re_intuit_start() already called. */ |
| 512 | #define REXEC_SCREAM 0x04 /* currently unused. */ |
| 513 | #define REXEC_IGNOREPOS 0x08 /* use stringarg, not pos(), for \G match */ |
| 514 | #define REXEC_NOT_FIRST 0x10 /* This is another iteration of //g: |
| 515 | no need to copy string again */ |
| 516 | |
| 517 | /* under REXEC_COPY_STR, it's ok for the |
| 518 | engine (modulo PL_sawamperand etc) |
| 519 | to skip copying: ... */ |
| 520 | #define REXEC_COPY_SKIP_PRE 0x20 /* ...the $` part of the string, or */ |
| 521 | #define REXEC_COPY_SKIP_POST 0x40 /* ...the $' part of the string */ |
| 522 | #define REXEC_FAIL_ON_UNDERFLOW 0x80 /* fail the match if $& would start before |
| 523 | the start pos (so s/.\G// would fail |
| 524 | on second iteration */ |
| 525 | |
| 526 | #if defined(__GNUC__) && !defined(PERL_GCC_BRACE_GROUPS_FORBIDDEN) |
| 527 | # define ReREFCNT_inc(re) \ |
| 528 | ({ \ |
| 529 | /* This is here to generate a casting warning if incorrect. */ \ |
| 530 | REGEXP *const _rerefcnt_inc = (re); \ |
| 531 | assert(SvTYPE(_rerefcnt_inc) == SVt_REGEXP); \ |
| 532 | SvREFCNT_inc(_rerefcnt_inc); \ |
| 533 | _rerefcnt_inc; \ |
| 534 | }) |
| 535 | # define ReREFCNT_dec(re) \ |
| 536 | ({ \ |
| 537 | /* This is here to generate a casting warning if incorrect. */ \ |
| 538 | REGEXP *const _rerefcnt_dec = (re); \ |
| 539 | SvREFCNT_dec(_rerefcnt_dec); \ |
| 540 | }) |
| 541 | #else |
| 542 | # define ReREFCNT_dec(re) SvREFCNT_dec(re) |
| 543 | # define ReREFCNT_inc(re) ((REGEXP *) SvREFCNT_inc(re)) |
| 544 | #endif |
| 545 | #define ReANY(re) S_ReANY((const REGEXP *)(re)) |
| 546 | |
| 547 | /* FIXME for plugins. */ |
| 548 | |
| 549 | #define FBMcf_TAIL_DOLLAR 1 |
| 550 | #define FBMcf_TAIL_DOLLARM 2 |
| 551 | #define FBMcf_TAIL_Z 4 |
| 552 | #define FBMcf_TAIL_z 8 |
| 553 | #define FBMcf_TAIL (FBMcf_TAIL_DOLLAR|FBMcf_TAIL_DOLLARM|FBMcf_TAIL_Z|FBMcf_TAIL_z) |
| 554 | |
| 555 | #define FBMrf_MULTILINE 1 |
| 556 | |
| 557 | struct regmatch_state; |
| 558 | struct regmatch_slab; |
| 559 | |
| 560 | /* like regmatch_info_aux, but contains extra fields only needed if the |
| 561 | * pattern contains (?{}). If used, is snuck into the second slot in the |
| 562 | * regmatch_state stack at the start of execution */ |
| 563 | |
| 564 | typedef struct { |
| 565 | regexp *rex; |
| 566 | PMOP *curpm; /* saved PL_curpm */ |
| 567 | #ifdef PERL_ANY_COW |
| 568 | SV *saved_copy; /* saved saved_copy field from rex */ |
| 569 | #endif |
| 570 | char *subbeg; /* saved subbeg field from rex */ |
| 571 | STRLEN sublen; /* saved sublen field from rex */ |
| 572 | STRLEN suboffset; /* saved suboffset field from rex */ |
| 573 | STRLEN subcoffset; /* saved subcoffset field from rex */ |
| 574 | MAGIC *pos_magic; /* pos() magic attached to $_ */ |
| 575 | SSize_t pos; /* the original value of pos() in pos_magic */ |
| 576 | U8 pos_flags; /* flags to be restored; currently only MGf_BYTES*/ |
| 577 | } regmatch_info_aux_eval; |
| 578 | |
| 579 | |
| 580 | /* fields that logically live in regmatch_info, but which need cleaning |
| 581 | * up on croak(), and so are instead are snuck into the first slot in |
| 582 | * the regmatch_state stack at the start of execution */ |
| 583 | |
| 584 | typedef struct { |
| 585 | regmatch_info_aux_eval *info_aux_eval; |
| 586 | struct regmatch_state *old_regmatch_state; /* saved PL_regmatch_state */ |
| 587 | struct regmatch_slab *old_regmatch_slab; /* saved PL_regmatch_slab */ |
| 588 | char *poscache; /* S-L cache of fail positions of WHILEMs */ |
| 589 | } regmatch_info_aux; |
| 590 | |
| 591 | |
| 592 | /* some basic information about the current match that is created by |
| 593 | * Perl_regexec_flags and then passed to regtry(), regmatch() etc. |
| 594 | * It is allocated as a local var on the stack, so nothing should be |
| 595 | * stored in it that needs preserving or clearing up on croak(). |
| 596 | * For that, see the aux_info and aux_info_eval members of the |
| 597 | * regmatch_state union. */ |
| 598 | |
| 599 | typedef struct { |
| 600 | REGEXP *prog; /* the regex being executed */ |
| 601 | const char * strbeg; /* real start of string */ |
| 602 | char *strend; /* one byte beyond last char of match string */ |
| 603 | char *till; /* matches shorter than this fail (see minlen arg) */ |
| 604 | SV *sv; /* the SV string currently being matched */ |
| 605 | char *ganch; /* position of \G anchor */ |
| 606 | char *cutpoint; /* (*COMMIT) position (if any) */ |
| 607 | regmatch_info_aux *info_aux; /* extra fields that need cleanup */ |
| 608 | regmatch_info_aux_eval *info_aux_eval; /* extra saved state for (?{}) */ |
| 609 | I32 poscache_maxiter; /* how many whilems todo before S-L cache kicks in */ |
| 610 | I32 poscache_iter; /* current countdown from _maxiter to zero */ |
| 611 | STRLEN poscache_size; /* size of regmatch_info_aux.poscache */ |
| 612 | bool intuit; /* re_intuit_start() is the top-level caller */ |
| 613 | bool is_utf8_pat; /* regex is utf8 */ |
| 614 | bool is_utf8_target; /* string being matched is utf8 */ |
| 615 | bool warned; /* we have issued a recursion warning; no need for more */ |
| 616 | } regmatch_info; |
| 617 | |
| 618 | |
| 619 | /* structures for holding and saving the state maintained by regmatch() */ |
| 620 | |
| 621 | #ifndef MAX_RECURSE_EVAL_NOCHANGE_DEPTH |
| 622 | #define MAX_RECURSE_EVAL_NOCHANGE_DEPTH 1000 |
| 623 | #endif |
| 624 | |
| 625 | typedef I32 CHECKPOINT; |
| 626 | |
| 627 | typedef struct regmatch_state { |
| 628 | int resume_state; /* where to jump to on return */ |
| 629 | char *locinput; /* where to backtrack in string on failure */ |
| 630 | |
| 631 | union { |
| 632 | |
| 633 | /* the 'info_aux' and 'info_aux_eval' union members are cuckoos in |
| 634 | * the nest. They aren't saved backtrack state; rather they |
| 635 | * represent one or two extra chunks of data that need allocating |
| 636 | * at the start of a match. These fields would logically live in |
| 637 | * the regmatch_info struct, except that is allocated on the |
| 638 | * C stack, and these fields are all things that require cleanup |
| 639 | * after a croak(), when the stack is lost. |
| 640 | * As a convenience, we just use the first 1 or 2 regmatch_state |
| 641 | * slots to store this info, as we will be allocating a slab of |
| 642 | * these anyway. Otherwise we'd have to malloc and then free them, |
| 643 | * or allocate them on the save stack (where they will get |
| 644 | * realloced if the save stack grows). |
| 645 | * info_aux contains the extra fields that are always needed; |
| 646 | * info_aux_eval contains extra fields that only needed if |
| 647 | * the pattern contains code blocks |
| 648 | * We split them into two separate structs to avoid increasing |
| 649 | * the size of the union. |
| 650 | */ |
| 651 | |
| 652 | regmatch_info_aux info_aux; |
| 653 | |
| 654 | regmatch_info_aux_eval info_aux_eval; |
| 655 | |
| 656 | /* this is a fake union member that matches the first element |
| 657 | * of each member that needs to store positive backtrack |
| 658 | * information */ |
| 659 | struct { |
| 660 | struct regmatch_state *prev_yes_state; |
| 661 | } yes; |
| 662 | |
| 663 | /* branchlike members */ |
| 664 | /* this is a fake union member that matches the first elements |
| 665 | * of each member that needs to behave like a branch */ |
| 666 | struct { |
| 667 | /* this first element must match u.yes */ |
| 668 | struct regmatch_state *prev_yes_state; |
| 669 | U32 lastparen; |
| 670 | U32 lastcloseparen; |
| 671 | CHECKPOINT cp; |
| 672 | |
| 673 | } branchlike; |
| 674 | |
| 675 | struct { |
| 676 | /* the first elements must match u.branchlike */ |
| 677 | struct regmatch_state *prev_yes_state; |
| 678 | U32 lastparen; |
| 679 | U32 lastcloseparen; |
| 680 | CHECKPOINT cp; |
| 681 | |
| 682 | regnode *next_branch; /* next branch node */ |
| 683 | } branch; |
| 684 | |
| 685 | struct { |
| 686 | /* the first elements must match u.branchlike */ |
| 687 | struct regmatch_state *prev_yes_state; |
| 688 | U32 lastparen; |
| 689 | U32 lastcloseparen; |
| 690 | CHECKPOINT cp; |
| 691 | |
| 692 | U32 accepted; /* how many accepting states left */ |
| 693 | bool longfold;/* saw a fold with a 1->n char mapping */ |
| 694 | U16 *jump; /* positive offsets from me */ |
| 695 | regnode *me; /* Which node am I - needed for jump tries*/ |
| 696 | U8 *firstpos;/* pos in string of first trie match */ |
| 697 | U32 firstchars;/* len in chars of firstpos from start */ |
| 698 | U16 nextword;/* next word to try */ |
| 699 | U16 topword; /* longest accepted word */ |
| 700 | } trie; |
| 701 | |
| 702 | /* special types - these members are used to store state for special |
| 703 | regops like eval, if/then, lookaround and the markpoint state */ |
| 704 | struct { |
| 705 | /* this first element must match u.yes */ |
| 706 | struct regmatch_state *prev_yes_state; |
| 707 | struct regmatch_state *prev_eval; |
| 708 | struct regmatch_state *prev_curlyx; |
| 709 | REGEXP *prev_rex; |
| 710 | CHECKPOINT cp; /* remember current savestack indexes */ |
| 711 | CHECKPOINT lastcp; |
| 712 | U32 close_paren; /* which close bracket is our end */ |
| 713 | regnode *B; /* the node following us */ |
| 714 | } eval; |
| 715 | |
| 716 | struct { |
| 717 | /* this first element must match u.yes */ |
| 718 | struct regmatch_state *prev_yes_state; |
| 719 | I32 wanted; |
| 720 | I32 logical; /* saved copy of 'logical' var */ |
| 721 | regnode *me; /* the IFMATCH/SUSPEND/UNLESSM node */ |
| 722 | } ifmatch; /* and SUSPEND/UNLESSM */ |
| 723 | |
| 724 | struct { |
| 725 | /* this first element must match u.yes */ |
| 726 | struct regmatch_state *prev_yes_state; |
| 727 | struct regmatch_state *prev_mark; |
| 728 | SV* mark_name; |
| 729 | char *mark_loc; |
| 730 | } mark; |
| 731 | |
| 732 | struct { |
| 733 | int val; |
| 734 | } keeper; |
| 735 | |
| 736 | /* quantifiers - these members are used for storing state for |
| 737 | for the regops used to implement quantifiers */ |
| 738 | struct { |
| 739 | /* this first element must match u.yes */ |
| 740 | struct regmatch_state *prev_yes_state; |
| 741 | struct regmatch_state *prev_curlyx; /* previous cur_curlyx */ |
| 742 | regnode *me; /* the CURLYX node */ |
| 743 | regnode *B; /* the B node in /A*B/ */ |
| 744 | CHECKPOINT cp; /* remember current savestack index */ |
| 745 | bool minmod; |
| 746 | int parenfloor;/* how far back to strip paren data */ |
| 747 | |
| 748 | /* these two are modified by WHILEM */ |
| 749 | int count; /* how many instances of A we've matched */ |
| 750 | char *lastloc;/* where previous A matched (0-len detect) */ |
| 751 | } curlyx; |
| 752 | |
| 753 | struct { |
| 754 | /* this first element must match u.yes */ |
| 755 | struct regmatch_state *prev_yes_state; |
| 756 | struct regmatch_state *save_curlyx; |
| 757 | CHECKPOINT cp; /* remember current savestack indexes */ |
| 758 | CHECKPOINT lastcp; |
| 759 | char *save_lastloc; /* previous curlyx.lastloc */ |
| 760 | I32 cache_offset; |
| 761 | I32 cache_mask; |
| 762 | } whilem; |
| 763 | |
| 764 | struct { |
| 765 | /* this first element must match u.yes */ |
| 766 | struct regmatch_state *prev_yes_state; |
| 767 | int c1, c2; /* case fold search */ |
| 768 | CHECKPOINT cp; |
| 769 | U32 lastparen; |
| 770 | U32 lastcloseparen; |
| 771 | I32 alen; /* length of first-matched A string */ |
| 772 | I32 count; |
| 773 | bool minmod; |
| 774 | regnode *A, *B; /* the nodes corresponding to /A*B/ */ |
| 775 | regnode *me; /* the curlym node */ |
| 776 | U8 c1_utf8[UTF8_MAXBYTES+1]; /* */ |
| 777 | U8 c2_utf8[UTF8_MAXBYTES+1]; |
| 778 | } curlym; |
| 779 | |
| 780 | struct { |
| 781 | U32 paren; |
| 782 | CHECKPOINT cp; |
| 783 | U32 lastparen; |
| 784 | U32 lastcloseparen; |
| 785 | int c1, c2; /* case fold search */ |
| 786 | char *maxpos; /* highest possible point in string to match */ |
| 787 | char *oldloc; /* the previous locinput */ |
| 788 | int count; |
| 789 | int min, max; /* {m,n} */ |
| 790 | regnode *A, *B; /* the nodes corresponding to /A*B/ */ |
| 791 | U8 c1_utf8[UTF8_MAXBYTES+1]; /* */ |
| 792 | U8 c2_utf8[UTF8_MAXBYTES+1]; |
| 793 | } curly; /* and CURLYN/PLUS/STAR */ |
| 794 | |
| 795 | } u; |
| 796 | } regmatch_state; |
| 797 | |
| 798 | /* how many regmatch_state structs to allocate as a single slab. |
| 799 | * We do it in 4K blocks for efficiency. The "3" is 2 for the next/prev |
| 800 | * pointers, plus 1 for any mythical malloc overhead. */ |
| 801 | |
| 802 | #define PERL_REGMATCH_SLAB_SLOTS \ |
| 803 | ((4096 - 3 * sizeof (void*)) / sizeof(regmatch_state)) |
| 804 | |
| 805 | typedef struct regmatch_slab { |
| 806 | regmatch_state states[PERL_REGMATCH_SLAB_SLOTS]; |
| 807 | struct regmatch_slab *prev, *next; |
| 808 | } regmatch_slab; |
| 809 | |
| 810 | |
| 811 | |
| 812 | /* |
| 813 | * Local variables: |
| 814 | * c-indentation-style: bsd |
| 815 | * c-basic-offset: 4 |
| 816 | * indent-tabs-mode: nil |
| 817 | * End: |
| 818 | * |
| 819 | * ex: set ts=8 sts=4 sw=4 et: |
| 820 | */ |