| 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 | struct regnode { |
| 22 | U8 flags; |
| 23 | U8 type; |
| 24 | U16 next_off; |
| 25 | }; |
| 26 | |
| 27 | typedef struct regnode regnode; |
| 28 | |
| 29 | struct reg_substr_data; |
| 30 | |
| 31 | struct reg_data; |
| 32 | |
| 33 | struct regexp_engine; |
| 34 | struct regexp; |
| 35 | |
| 36 | struct reg_substr_datum { |
| 37 | I32 min_offset; |
| 38 | I32 max_offset; |
| 39 | SV *substr; /* non-utf8 variant */ |
| 40 | SV *utf8_substr; /* utf8 variant */ |
| 41 | I32 end_shift; |
| 42 | }; |
| 43 | struct reg_substr_data { |
| 44 | struct reg_substr_datum data[3]; /* Actual array */ |
| 45 | }; |
| 46 | |
| 47 | #ifdef PERL_OLD_COPY_ON_WRITE |
| 48 | #define SV_SAVED_COPY SV *saved_copy; /* If non-NULL, SV which is COW from original */ |
| 49 | #else |
| 50 | #define SV_SAVED_COPY |
| 51 | #endif |
| 52 | |
| 53 | typedef struct regexp_paren_pair { |
| 54 | I32 start; |
| 55 | I32 end; |
| 56 | } regexp_paren_pair; |
| 57 | |
| 58 | /* |
| 59 | The regexp/REGEXP struct, see L<perlreapi> for further documentation |
| 60 | on the individual fields. The struct is ordered so that the most |
| 61 | commonly used fields are placed at the start. |
| 62 | |
| 63 | Any patch that adds items to this struct will need to include |
| 64 | changes to F<sv.c> (C<Perl_re_dup()>) and F<regcomp.c> |
| 65 | (C<pregfree()>). This involves freeing or cloning items in the |
| 66 | regexp's data array based on the data item's type. |
| 67 | */ |
| 68 | |
| 69 | #define _REGEXP_COMMON \ |
| 70 | /* what engine created this regexp? */ \ |
| 71 | const struct regexp_engine* engine; \ |
| 72 | REGEXP *mother_re; /* what re is this a lightweight copy of? */ \ |
| 73 | HV *paren_names; /* Optional hash of paren names */ \ |
| 74 | /* Information about the match that the perl core uses to */ \ |
| 75 | /* manage things */ \ |
| 76 | U32 extflags; /* Flags used both externally and internally */ \ |
| 77 | I32 minlen; /* mininum possible length of string to match */\ |
| 78 | I32 minlenret; /* mininum possible length of $& */ \ |
| 79 | U32 gofs; /* chars left of pos that we search from */ \ |
| 80 | /* substring data about strings that must appear in the */ \ |
| 81 | /* final match, used for optimisations */ \ |
| 82 | struct reg_substr_data *substrs; \ |
| 83 | U32 nparens; /* number of capture buffers */ \ |
| 84 | /* private engine specific data */ \ |
| 85 | U32 intflags; /* Engine Specific Internal flags */ \ |
| 86 | void *pprivate; /* Data private to the regex engine which */ \ |
| 87 | /* created this object. */ \ |
| 88 | /* Data about the last/current match. These are modified */ \ |
| 89 | /* during matching */ \ |
| 90 | U32 lastparen; /* last open paren matched */ \ |
| 91 | U32 lastcloseparen; /* last close paren matched */ \ |
| 92 | regexp_paren_pair *swap; /* Unused: 5.10.1 and later */ \ |
| 93 | /* Array of offsets for (@-) and (@+) */ \ |
| 94 | regexp_paren_pair *offs; \ |
| 95 | /* saved or original string so \digit works forever. */ \ |
| 96 | char *subbeg; \ |
| 97 | SV_SAVED_COPY /* If non-NULL, SV which is COW from original */\ |
| 98 | I32 sublen; /* Length of string pointed by subbeg */ \ |
| 99 | /* Information about the match that isn't often used */ \ |
| 100 | /* offset from wrapped to the start of precomp */ \ |
| 101 | PERL_BITFIELD32 pre_prefix:4; \ |
| 102 | /* number of eval groups in the pattern - for security checks */\ |
| 103 | PERL_BITFIELD32 seen_evals:28 |
| 104 | |
| 105 | typedef struct regexp { |
| 106 | _XPV_HEAD; |
| 107 | _REGEXP_COMMON; |
| 108 | } regexp; |
| 109 | |
| 110 | #define RXp_PAREN_NAMES(rx) ((rx)->paren_names) |
| 111 | |
| 112 | /* used for high speed searches */ |
| 113 | typedef struct re_scream_pos_data_s |
| 114 | { |
| 115 | char **scream_olds; /* match pos */ |
| 116 | I32 *scream_pos; /* Internal iterator of scream. */ |
| 117 | } re_scream_pos_data; |
| 118 | |
| 119 | /* regexp_engine structure. This is the dispatch table for regexes. |
| 120 | * Any regex engine implementation must be able to build one of these. |
| 121 | */ |
| 122 | typedef struct regexp_engine { |
| 123 | REGEXP* (*comp) (pTHX_ SV * const pattern, U32 flags); |
| 124 | I32 (*exec) (pTHX_ REGEXP * const rx, char* stringarg, char* strend, |
| 125 | char* strbeg, I32 minend, SV* screamer, |
| 126 | void* data, U32 flags); |
| 127 | char* (*intuit) (pTHX_ REGEXP * const rx, SV *sv, char *strpos, |
| 128 | char *strend, const U32 flags, |
| 129 | re_scream_pos_data *data); |
| 130 | SV* (*checkstr) (pTHX_ REGEXP * const rx); |
| 131 | void (*free) (pTHX_ REGEXP * const rx); |
| 132 | void (*numbered_buff_FETCH) (pTHX_ REGEXP * const rx, const I32 paren, |
| 133 | SV * const sv); |
| 134 | void (*numbered_buff_STORE) (pTHX_ REGEXP * const rx, const I32 paren, |
| 135 | SV const * const value); |
| 136 | I32 (*numbered_buff_LENGTH) (pTHX_ REGEXP * const rx, const SV * const sv, |
| 137 | const I32 paren); |
| 138 | SV* (*named_buff) (pTHX_ REGEXP * const rx, SV * const key, |
| 139 | SV * const value, const U32 flags); |
| 140 | SV* (*named_buff_iter) (pTHX_ REGEXP * const rx, const SV * const lastkey, |
| 141 | const U32 flags); |
| 142 | SV* (*qr_package)(pTHX_ REGEXP * const rx); |
| 143 | #ifdef USE_ITHREADS |
| 144 | void* (*dupe) (pTHX_ REGEXP * const rx, CLONE_PARAMS *param); |
| 145 | #endif |
| 146 | } regexp_engine; |
| 147 | |
| 148 | /* |
| 149 | These are passed to the numbered capture variable callbacks as the |
| 150 | paren name. >= 1 is reserved for actual numbered captures, i.e. $1, |
| 151 | $2 etc. |
| 152 | */ |
| 153 | #define RX_BUFF_IDX_PREMATCH -2 /* $` / ${^PREMATCH} */ |
| 154 | #define RX_BUFF_IDX_POSTMATCH -1 /* $' / ${^POSTMATCH} */ |
| 155 | #define RX_BUFF_IDX_FULLMATCH 0 /* $& / ${^MATCH} */ |
| 156 | |
| 157 | /* |
| 158 | Flags that are passed to the named_buff and named_buff_iter |
| 159 | callbacks above. Those routines are called from universal.c via the |
| 160 | Tie::Hash::NamedCapture interface for %+ and %- and the re:: |
| 161 | functions in the same file. |
| 162 | */ |
| 163 | |
| 164 | /* The Tie::Hash::NamedCapture operation this is part of, if any */ |
| 165 | #define RXapif_FETCH 0x0001 |
| 166 | #define RXapif_STORE 0x0002 |
| 167 | #define RXapif_DELETE 0x0004 |
| 168 | #define RXapif_CLEAR 0x0008 |
| 169 | #define RXapif_EXISTS 0x0010 |
| 170 | #define RXapif_SCALAR 0x0020 |
| 171 | #define RXapif_FIRSTKEY 0x0040 |
| 172 | #define RXapif_NEXTKEY 0x0080 |
| 173 | |
| 174 | /* Whether %+ or %- is being operated on */ |
| 175 | #define RXapif_ONE 0x0100 /* %+ */ |
| 176 | #define RXapif_ALL 0x0200 /* %- */ |
| 177 | |
| 178 | /* Whether this is being called from a re:: function */ |
| 179 | #define RXapif_REGNAME 0x0400 |
| 180 | #define RXapif_REGNAMES 0x0800 |
| 181 | #define RXapif_REGNAMES_COUNT 0x1000 |
| 182 | |
| 183 | /* |
| 184 | =head1 REGEXP Functions |
| 185 | |
| 186 | =for apidoc Am|REGEXP *|SvRX|SV *sv |
| 187 | |
| 188 | Convenience macro to get the REGEXP from a SV. This is approximately |
| 189 | equivalent to the following snippet: |
| 190 | |
| 191 | if (SvMAGICAL(sv)) |
| 192 | mg_get(sv); |
| 193 | if (SvROK(sv)) |
| 194 | sv = MUTABLE_SV(SvRV(sv)); |
| 195 | if (SvTYPE(sv) == SVt_REGEXP) |
| 196 | return (REGEXP*) sv; |
| 197 | |
| 198 | NULL will be returned if a REGEXP* is not found. |
| 199 | |
| 200 | =for apidoc Am|bool|SvRXOK|SV* sv |
| 201 | |
| 202 | Returns a boolean indicating whether the SV (or the one it references) |
| 203 | is a REGEXP. |
| 204 | |
| 205 | If you want to do something with the REGEXP* later use SvRX instead |
| 206 | and check for NULL. |
| 207 | |
| 208 | =cut |
| 209 | */ |
| 210 | |
| 211 | #define SvRX(sv) (Perl_get_re_arg(aTHX_ sv)) |
| 212 | #define SvRXOK(sv) (Perl_get_re_arg(aTHX_ sv) ? TRUE : FALSE) |
| 213 | |
| 214 | |
| 215 | /* Flags stored in regexp->extflags |
| 216 | * These are used by code external to the regexp engine |
| 217 | * |
| 218 | * Note that the flags whose names start with RXf_PMf_ are defined in |
| 219 | * op_reg_common.h, being copied from the parallel flags of op_pmflags |
| 220 | * |
| 221 | * NOTE: if you modify any RXf flags you should run regen.pl or |
| 222 | * regen/regcomp.pl so that regnodes.h is updated with the changes. |
| 223 | * |
| 224 | */ |
| 225 | |
| 226 | #include "op_reg_common.h" |
| 227 | |
| 228 | #define RXf_PMf_STD_PMMOD (RXf_PMf_MULTILINE|RXf_PMf_SINGLELINE|RXf_PMf_FOLD|RXf_PMf_EXTENDED) |
| 229 | |
| 230 | #define CASE_STD_PMMOD_FLAGS_PARSE_SET(pmfl) \ |
| 231 | case IGNORE_PAT_MOD: *(pmfl) |= RXf_PMf_FOLD; break; \ |
| 232 | case MULTILINE_PAT_MOD: *(pmfl) |= RXf_PMf_MULTILINE; break; \ |
| 233 | case SINGLE_PAT_MOD: *(pmfl) |= RXf_PMf_SINGLELINE; break; \ |
| 234 | case XTENDED_PAT_MOD: *(pmfl) |= RXf_PMf_EXTENDED; break |
| 235 | |
| 236 | /* Note, includes charset ones, assumes 0 is the default for them */ |
| 237 | #define STD_PMMOD_FLAGS_CLEAR(pmfl) \ |
| 238 | *(pmfl) &= ~(RXf_PMf_FOLD|RXf_PMf_MULTILINE|RXf_PMf_SINGLELINE|RXf_PMf_EXTENDED|RXf_PMf_CHARSET) |
| 239 | |
| 240 | /* chars and strings used as regex pattern modifiers |
| 241 | * Singular is a 'c'har, plural is a "string" |
| 242 | * |
| 243 | * NOTE, KEEPCOPY was originally 'k', but was changed to 'p' for preserve |
| 244 | * for compatibility reasons with Regexp::Common which highjacked (?k:...) |
| 245 | * for its own uses. So 'k' is out as well. |
| 246 | */ |
| 247 | #define DEFAULT_PAT_MOD '^' /* Short for all the default modifiers */ |
| 248 | #define EXEC_PAT_MOD 'e' |
| 249 | #define KEEPCOPY_PAT_MOD 'p' |
| 250 | #define ONCE_PAT_MOD 'o' |
| 251 | #define GLOBAL_PAT_MOD 'g' |
| 252 | #define CONTINUE_PAT_MOD 'c' |
| 253 | #define MULTILINE_PAT_MOD 'm' |
| 254 | #define SINGLE_PAT_MOD 's' |
| 255 | #define IGNORE_PAT_MOD 'i' |
| 256 | #define XTENDED_PAT_MOD 'x' |
| 257 | #define NONDESTRUCT_PAT_MOD 'r' |
| 258 | #define LOCALE_PAT_MOD 'l' |
| 259 | #define UNICODE_PAT_MOD 'u' |
| 260 | #define DEPENDS_PAT_MOD 'd' |
| 261 | #define ASCII_RESTRICT_PAT_MOD 'a' |
| 262 | |
| 263 | #define ONCE_PAT_MODS "o" |
| 264 | #define KEEPCOPY_PAT_MODS "p" |
| 265 | #define EXEC_PAT_MODS "e" |
| 266 | #define LOOP_PAT_MODS "gc" |
| 267 | #define NONDESTRUCT_PAT_MODS "r" |
| 268 | #define LOCALE_PAT_MODS "l" |
| 269 | #define UNICODE_PAT_MODS "u" |
| 270 | #define DEPENDS_PAT_MODS "d" |
| 271 | #define ASCII_RESTRICT_PAT_MODS "a" |
| 272 | #define ASCII_MORE_RESTRICT_PAT_MODS "aa" |
| 273 | |
| 274 | /* This string is expected by regcomp.c to be ordered so that the first |
| 275 | * character is the flag in bit RXf_PMf_STD_PMMOD_SHIFT of extflags; the next |
| 276 | * character is bit +1, etc. */ |
| 277 | #define STD_PAT_MODS "msix" |
| 278 | |
| 279 | #define CHARSET_PAT_MODS ASCII_RESTRICT_PAT_MODS DEPENDS_PAT_MODS LOCALE_PAT_MODS UNICODE_PAT_MODS |
| 280 | |
| 281 | /* This string is expected by XS_re_regexp_pattern() in universal.c to be ordered |
| 282 | * so that the first character is the flag in bit RXf_PMf_STD_PMMOD_SHIFT of |
| 283 | * extflags; the next character is in bit +1, etc. */ |
| 284 | #define INT_PAT_MODS STD_PAT_MODS KEEPCOPY_PAT_MODS |
| 285 | |
| 286 | #define EXT_PAT_MODS ONCE_PAT_MODS KEEPCOPY_PAT_MODS |
| 287 | #define QR_PAT_MODS STD_PAT_MODS EXT_PAT_MODS CHARSET_PAT_MODS |
| 288 | #define M_PAT_MODS QR_PAT_MODS LOOP_PAT_MODS |
| 289 | #define S_PAT_MODS M_PAT_MODS EXEC_PAT_MODS NONDESTRUCT_PAT_MODS |
| 290 | |
| 291 | /* |
| 292 | * NOTE: if you modify any RXf flags you should run regen.pl or |
| 293 | * regen/regcomp.pl so that regnodes.h is updated with the changes. |
| 294 | * |
| 295 | */ |
| 296 | |
| 297 | /* Leave some space, so future bit allocations can go either in the shared or |
| 298 | * unshared area without affecting binary compatibility */ |
| 299 | #define RXf_BASE_SHIFT (_RXf_PMf_SHIFT_NEXT+1) |
| 300 | |
| 301 | /* embed.pl doesn't yet know how to handle static inline functions, so |
| 302 | manually decorate them here with gcc-style attributes. |
| 303 | */ |
| 304 | PERL_STATIC_INLINE const char * |
| 305 | get_regex_charset_name(const U32 flags, STRLEN* const lenp) |
| 306 | __attribute__warn_unused_result__; |
| 307 | |
| 308 | #define MAX_CHARSET_NAME_LENGTH 2 |
| 309 | |
| 310 | PERL_STATIC_INLINE const char * |
| 311 | get_regex_charset_name(const U32 flags, STRLEN* const lenp) |
| 312 | { |
| 313 | /* Returns a string that corresponds to the name of the regex character set |
| 314 | * given by 'flags', and *lenp is set the length of that string, which |
| 315 | * cannot exceed MAX_CHARSET_NAME_LENGTH characters */ |
| 316 | |
| 317 | *lenp = 1; |
| 318 | switch (get_regex_charset(flags)) { |
| 319 | case REGEX_DEPENDS_CHARSET: return DEPENDS_PAT_MODS; |
| 320 | case REGEX_LOCALE_CHARSET: return LOCALE_PAT_MODS; |
| 321 | case REGEX_UNICODE_CHARSET: return UNICODE_PAT_MODS; |
| 322 | case REGEX_ASCII_RESTRICTED_CHARSET: return ASCII_RESTRICT_PAT_MODS; |
| 323 | case REGEX_ASCII_MORE_RESTRICTED_CHARSET: |
| 324 | *lenp = 2; |
| 325 | return ASCII_MORE_RESTRICT_PAT_MODS; |
| 326 | default: |
| 327 | return "?"; /* Unknown */ |
| 328 | } |
| 329 | } |
| 330 | |
| 331 | /* Anchor and GPOS related stuff */ |
| 332 | #define RXf_ANCH_BOL (1<<(RXf_BASE_SHIFT+0)) |
| 333 | #define RXf_ANCH_MBOL (1<<(RXf_BASE_SHIFT+1)) |
| 334 | #define RXf_ANCH_SBOL (1<<(RXf_BASE_SHIFT+2)) |
| 335 | #define RXf_ANCH_GPOS (1<<(RXf_BASE_SHIFT+3)) |
| 336 | #define RXf_GPOS_SEEN (1<<(RXf_BASE_SHIFT+4)) |
| 337 | #define RXf_GPOS_FLOAT (1<<(RXf_BASE_SHIFT+5)) |
| 338 | /* two bits here */ |
| 339 | #define RXf_ANCH (RXf_ANCH_BOL|RXf_ANCH_MBOL|RXf_ANCH_GPOS|RXf_ANCH_SBOL) |
| 340 | #define RXf_GPOS_CHECK (RXf_GPOS_SEEN|RXf_ANCH_GPOS) |
| 341 | #define RXf_ANCH_SINGLE (RXf_ANCH_SBOL|RXf_ANCH_GPOS) |
| 342 | |
| 343 | /* What we have seen */ |
| 344 | #define RXf_LOOKBEHIND_SEEN (1<<(RXf_BASE_SHIFT+6)) |
| 345 | #define RXf_EVAL_SEEN (1<<(RXf_BASE_SHIFT+7)) |
| 346 | #define RXf_CANY_SEEN (1<<(RXf_BASE_SHIFT+8)) |
| 347 | |
| 348 | /* Special */ |
| 349 | #define RXf_NOSCAN (1<<(RXf_BASE_SHIFT+9)) |
| 350 | #define RXf_CHECK_ALL (1<<(RXf_BASE_SHIFT+10)) |
| 351 | |
| 352 | /* UTF8 related */ |
| 353 | #define RXf_MATCH_UTF8 (1<<(RXf_BASE_SHIFT+11)) |
| 354 | |
| 355 | /* Intuit related */ |
| 356 | #define RXf_USE_INTUIT_NOML (1<<(RXf_BASE_SHIFT+12)) |
| 357 | #define RXf_USE_INTUIT_ML (1<<(RXf_BASE_SHIFT+13)) |
| 358 | #define RXf_INTUIT_TAIL (1<<(RXf_BASE_SHIFT+14)) |
| 359 | |
| 360 | /* |
| 361 | Set in Perl_pmruntime if op_flags & OPf_SPECIAL, i.e. split. Will |
| 362 | be used by regex engines to check whether they should set |
| 363 | RXf_SKIPWHITE |
| 364 | */ |
| 365 | #define RXf_SPLIT (1<<(RXf_BASE_SHIFT+15)) |
| 366 | |
| 367 | #define RXf_USE_INTUIT (RXf_USE_INTUIT_NOML|RXf_USE_INTUIT_ML) |
| 368 | |
| 369 | /* Copy and tainted info */ |
| 370 | #define RXf_COPY_DONE (1<<(RXf_BASE_SHIFT+16)) |
| 371 | |
| 372 | /* during execution: pattern temporarily tainted by executing locale ops; |
| 373 | * post-execution: $1 et al are tainted */ |
| 374 | #define RXf_TAINTED_SEEN (1<<(RXf_BASE_SHIFT+17)) |
| 375 | /* this pattern was tainted during compilation */ |
| 376 | #define RXf_TAINTED (1<<(RXf_BASE_SHIFT+18)) |
| 377 | |
| 378 | /* Flags indicating special patterns */ |
| 379 | #define RXf_START_ONLY (1<<(RXf_BASE_SHIFT+19)) /* Pattern is /^/ */ |
| 380 | #define RXf_SKIPWHITE (1<<(RXf_BASE_SHIFT+20)) /* Pattern is for a split / / */ |
| 381 | #define RXf_WHITE (1<<(RXf_BASE_SHIFT+21)) /* Pattern is /\s+/ */ |
| 382 | #define RXf_NULL (1<<(RXf_BASE_SHIFT+22)) /* Pattern is // */ |
| 383 | #if RXf_BASE_SHIFT+22 > 31 |
| 384 | # error Too many RXf_PMf bits used. See regnodes.h for any spare in middle |
| 385 | #endif |
| 386 | |
| 387 | /* |
| 388 | * NOTE: if you modify any RXf flags you should run regen.pl or |
| 389 | * regen/regcomp.pl so that regnodes.h is updated with the changes. |
| 390 | * |
| 391 | */ |
| 392 | |
| 393 | #define RX_HAS_CUTGROUP(prog) ((prog)->intflags & PREGf_CUTGROUP_SEEN) |
| 394 | #define RXp_MATCH_TAINTED(prog) (RXp_EXTFLAGS(prog) & RXf_TAINTED_SEEN) |
| 395 | #define RX_MATCH_TAINTED(prog) (RX_EXTFLAGS(prog) & RXf_TAINTED_SEEN) |
| 396 | #define RX_MATCH_TAINTED_on(prog) (RX_EXTFLAGS(prog) |= RXf_TAINTED_SEEN) |
| 397 | #define RX_MATCH_TAINTED_off(prog) (RX_EXTFLAGS(prog) &= ~RXf_TAINTED_SEEN) |
| 398 | #define RX_MATCH_TAINTED_set(prog, t) ((t) \ |
| 399 | ? RX_MATCH_TAINTED_on(prog) \ |
| 400 | : RX_MATCH_TAINTED_off(prog)) |
| 401 | |
| 402 | #define RXp_MATCH_COPIED(prog) (RXp_EXTFLAGS(prog) & RXf_COPY_DONE) |
| 403 | #define RX_MATCH_COPIED(prog) (RX_EXTFLAGS(prog) & RXf_COPY_DONE) |
| 404 | #define RXp_MATCH_COPIED_on(prog) (RXp_EXTFLAGS(prog) |= RXf_COPY_DONE) |
| 405 | #define RX_MATCH_COPIED_on(prog) (RX_EXTFLAGS(prog) |= RXf_COPY_DONE) |
| 406 | #define RXp_MATCH_COPIED_off(prog) (RXp_EXTFLAGS(prog) &= ~RXf_COPY_DONE) |
| 407 | #define RX_MATCH_COPIED_off(prog) (RX_EXTFLAGS(prog) &= ~RXf_COPY_DONE) |
| 408 | #define RX_MATCH_COPIED_set(prog,t) ((t) \ |
| 409 | ? RX_MATCH_COPIED_on(prog) \ |
| 410 | : RX_MATCH_COPIED_off(prog)) |
| 411 | |
| 412 | #define RXp_EXTFLAGS(rx) ((rx)->extflags) |
| 413 | |
| 414 | /* For source compatibility. We used to store these explicitly. */ |
| 415 | #define RX_PRECOMP(prog) (RX_WRAPPED(prog) + ((struct regexp *)SvANY(prog))->pre_prefix) |
| 416 | #define RX_PRECOMP_const(prog) (RX_WRAPPED_const(prog) + ((struct regexp *)SvANY(prog))->pre_prefix) |
| 417 | /* FIXME? Are we hardcoding too much here and constraining plugin extension |
| 418 | writers? Specifically, the value 1 assumes that the wrapped version always |
| 419 | has exactly one character at the end, a ')'. Will that always be true? */ |
| 420 | #define RX_PRELEN(prog) (RX_WRAPLEN(prog) - ((struct regexp *)SvANY(prog))->pre_prefix - 1) |
| 421 | #define RX_WRAPPED(prog) SvPVX(prog) |
| 422 | #define RX_WRAPPED_const(prog) SvPVX_const(prog) |
| 423 | #define RX_WRAPLEN(prog) SvCUR(prog) |
| 424 | #define RX_CHECK_SUBSTR(prog) (((struct regexp *)SvANY(prog))->check_substr) |
| 425 | #define RX_REFCNT(prog) SvREFCNT(prog) |
| 426 | #if defined(__GNUC__) && !defined(PERL_GCC_BRACE_GROUPS_FORBIDDEN) |
| 427 | # define RX_EXTFLAGS(prog) \ |
| 428 | (*({ \ |
| 429 | const REGEXP *const _rx_extflags = (prog); \ |
| 430 | assert(SvTYPE(_rx_extflags) == SVt_REGEXP); \ |
| 431 | &RXp_EXTFLAGS(SvANY(_rx_extflags)); \ |
| 432 | })) |
| 433 | # define RX_ENGINE(prog) \ |
| 434 | (*({ \ |
| 435 | const REGEXP *const _rx_engine = (prog); \ |
| 436 | assert(SvTYPE(_rx_engine) == SVt_REGEXP); \ |
| 437 | &SvANY(_rx_engine)->engine; \ |
| 438 | })) |
| 439 | # define RX_SUBBEG(prog) \ |
| 440 | (*({ \ |
| 441 | const REGEXP *const _rx_subbeg = (prog); \ |
| 442 | assert(SvTYPE(_rx_subbeg) == SVt_REGEXP); \ |
| 443 | &SvANY(_rx_subbeg)->subbeg; \ |
| 444 | })) |
| 445 | # define RX_OFFS(prog) \ |
| 446 | (*({ \ |
| 447 | const REGEXP *const _rx_offs = (prog); \ |
| 448 | assert(SvTYPE(_rx_offs) == SVt_REGEXP); \ |
| 449 | &SvANY(_rx_offs)->offs; \ |
| 450 | })) |
| 451 | # define RX_NPARENS(prog) \ |
| 452 | (*({ \ |
| 453 | const REGEXP *const _rx_nparens = (prog); \ |
| 454 | assert(SvTYPE(_rx_nparens) == SVt_REGEXP); \ |
| 455 | &SvANY(_rx_nparens)->nparens; \ |
| 456 | })) |
| 457 | #else |
| 458 | # define RX_EXTFLAGS(prog) RXp_EXTFLAGS((struct regexp *)SvANY(prog)) |
| 459 | # define RX_ENGINE(prog) (((struct regexp *)SvANY(prog))->engine) |
| 460 | # define RX_SUBBEG(prog) (((struct regexp *)SvANY(prog))->subbeg) |
| 461 | # define RX_OFFS(prog) (((struct regexp *)SvANY(prog))->offs) |
| 462 | # define RX_NPARENS(prog) (((struct regexp *)SvANY(prog))->nparens) |
| 463 | #endif |
| 464 | #define RX_SUBLEN(prog) (((struct regexp *)SvANY(prog))->sublen) |
| 465 | #define RX_MINLEN(prog) (((struct regexp *)SvANY(prog))->minlen) |
| 466 | #define RX_MINLENRET(prog) (((struct regexp *)SvANY(prog))->minlenret) |
| 467 | #define RX_GOFS(prog) (((struct regexp *)SvANY(prog))->gofs) |
| 468 | #define RX_LASTPAREN(prog) (((struct regexp *)SvANY(prog))->lastparen) |
| 469 | #define RX_LASTCLOSEPAREN(prog) (((struct regexp *)SvANY(prog))->lastcloseparen) |
| 470 | #define RX_SEEN_EVALS(prog) (((struct regexp *)SvANY(prog))->seen_evals) |
| 471 | #define RX_SAVED_COPY(prog) (((struct regexp *)SvANY(prog))->saved_copy) |
| 472 | |
| 473 | #endif /* PLUGGABLE_RE_EXTENSION */ |
| 474 | |
| 475 | /* Stuff that needs to be included in the pluggable extension goes below here */ |
| 476 | |
| 477 | #ifdef PERL_OLD_COPY_ON_WRITE |
| 478 | #define RX_MATCH_COPY_FREE(rx) \ |
| 479 | STMT_START {if (RX_SAVED_COPY(rx)) { \ |
| 480 | SV_CHECK_THINKFIRST_COW_DROP(RX_SAVED_COPY(rx)); \ |
| 481 | } \ |
| 482 | if (RX_MATCH_COPIED(rx)) { \ |
| 483 | Safefree(RX_SUBBEG(rx)); \ |
| 484 | RX_MATCH_COPIED_off(rx); \ |
| 485 | }} STMT_END |
| 486 | #else |
| 487 | #define RX_MATCH_COPY_FREE(rx) \ |
| 488 | STMT_START {if (RX_MATCH_COPIED(rx)) { \ |
| 489 | Safefree(RX_SUBBEG(rx)); \ |
| 490 | RX_MATCH_COPIED_off(rx); \ |
| 491 | }} STMT_END |
| 492 | #endif |
| 493 | |
| 494 | #define RXp_MATCH_UTF8(prog) (RXp_EXTFLAGS(prog) & RXf_MATCH_UTF8) |
| 495 | #define RX_MATCH_UTF8(prog) (RX_EXTFLAGS(prog) & RXf_MATCH_UTF8) |
| 496 | #define RX_MATCH_UTF8_on(prog) (RX_EXTFLAGS(prog) |= RXf_MATCH_UTF8) |
| 497 | #define RX_MATCH_UTF8_off(prog) (RX_EXTFLAGS(prog) &= ~RXf_MATCH_UTF8) |
| 498 | #define RX_MATCH_UTF8_set(prog, t) ((t) \ |
| 499 | ? (RX_MATCH_UTF8_on(prog), (PL_reg_match_utf8 = 1)) \ |
| 500 | : (RX_MATCH_UTF8_off(prog), (PL_reg_match_utf8 = 0))) |
| 501 | |
| 502 | /* Whether the pattern stored at RX_WRAPPED is in UTF-8 */ |
| 503 | #define RX_UTF8(prog) SvUTF8(prog) |
| 504 | |
| 505 | #define REXEC_COPY_STR 0x01 /* Need to copy the string. */ |
| 506 | #define REXEC_CHECKED 0x02 /* check_substr already checked. */ |
| 507 | #define REXEC_SCREAM 0x04 /* use scream table. */ |
| 508 | #define REXEC_IGNOREPOS 0x08 /* \G matches at start. */ |
| 509 | #define REXEC_NOT_FIRST 0x10 /* This is another iteration of //g. */ |
| 510 | |
| 511 | #if defined(__GNUC__) && !defined(PERL_GCC_BRACE_GROUPS_FORBIDDEN) |
| 512 | # define ReREFCNT_inc(re) \ |
| 513 | ({ \ |
| 514 | /* This is here to generate a casting warning if incorrect. */ \ |
| 515 | REGEXP *const _rerefcnt_inc = (re); \ |
| 516 | assert(SvTYPE(_rerefcnt_inc) == SVt_REGEXP); \ |
| 517 | SvREFCNT_inc(_rerefcnt_inc); \ |
| 518 | _rerefcnt_inc; \ |
| 519 | }) |
| 520 | # define ReREFCNT_dec(re) \ |
| 521 | ({ \ |
| 522 | /* This is here to generate a casting warning if incorrect. */ \ |
| 523 | REGEXP *const _rerefcnt_dec = (re); \ |
| 524 | SvREFCNT_dec(_rerefcnt_dec); \ |
| 525 | }) |
| 526 | #else |
| 527 | # define ReREFCNT_dec(re) SvREFCNT_dec(re) |
| 528 | # define ReREFCNT_inc(re) ((REGEXP *) SvREFCNT_inc(re)) |
| 529 | #endif |
| 530 | |
| 531 | /* FIXME for plugins. */ |
| 532 | |
| 533 | #define FBMcf_TAIL_DOLLAR 1 |
| 534 | #define FBMcf_TAIL_DOLLARM 2 |
| 535 | #define FBMcf_TAIL_Z 4 |
| 536 | #define FBMcf_TAIL_z 8 |
| 537 | #define FBMcf_TAIL (FBMcf_TAIL_DOLLAR|FBMcf_TAIL_DOLLARM|FBMcf_TAIL_Z|FBMcf_TAIL_z) |
| 538 | |
| 539 | #define FBMrf_MULTILINE 1 |
| 540 | |
| 541 | /* some basic information about the current match that is created by |
| 542 | * Perl_regexec_flags and then passed to regtry(), regmatch() etc */ |
| 543 | |
| 544 | typedef struct { |
| 545 | REGEXP *prog; |
| 546 | char *bol; |
| 547 | char *till; |
| 548 | SV *sv; |
| 549 | char *ganch; |
| 550 | char *cutpoint; |
| 551 | } regmatch_info; |
| 552 | |
| 553 | |
| 554 | /* structures for holding and saving the state maintained by regmatch() */ |
| 555 | |
| 556 | #ifndef MAX_RECURSE_EVAL_NOCHANGE_DEPTH |
| 557 | #define MAX_RECURSE_EVAL_NOCHANGE_DEPTH 1000 |
| 558 | #endif |
| 559 | |
| 560 | typedef I32 CHECKPOINT; |
| 561 | |
| 562 | typedef struct regmatch_state { |
| 563 | int resume_state; /* where to jump to on return */ |
| 564 | char *locinput; /* where to backtrack in string on failure */ |
| 565 | |
| 566 | union { |
| 567 | |
| 568 | /* this is a fake union member that matches the first element |
| 569 | * of each member that needs to store positive backtrack |
| 570 | * information */ |
| 571 | struct { |
| 572 | struct regmatch_state *prev_yes_state; |
| 573 | } yes; |
| 574 | |
| 575 | /* branchlike members */ |
| 576 | /* this is a fake union member that matches the first elements |
| 577 | * of each member that needs to behave like a branch */ |
| 578 | struct { |
| 579 | /* this first element must match u.yes */ |
| 580 | struct regmatch_state *prev_yes_state; |
| 581 | U32 lastparen; |
| 582 | CHECKPOINT cp; |
| 583 | |
| 584 | } branchlike; |
| 585 | |
| 586 | struct { |
| 587 | /* the first elements must match u.branchlike */ |
| 588 | struct regmatch_state *prev_yes_state; |
| 589 | U32 lastparen; |
| 590 | CHECKPOINT cp; |
| 591 | |
| 592 | regnode *next_branch; /* next branch node */ |
| 593 | } branch; |
| 594 | |
| 595 | struct { |
| 596 | /* the first elements must match u.branchlike */ |
| 597 | struct regmatch_state *prev_yes_state; |
| 598 | U32 lastparen; |
| 599 | CHECKPOINT cp; |
| 600 | |
| 601 | U32 accepted; /* how many accepting states left */ |
| 602 | U16 *jump; /* positive offsets from me */ |
| 603 | regnode *B; /* node following the trie */ |
| 604 | regnode *me; /* Which node am I - needed for jump tries*/ |
| 605 | U8 *firstpos;/* pos in string of first trie match */ |
| 606 | U32 firstchars;/* len in chars of firstpos from start */ |
| 607 | U16 nextword;/* next word to try */ |
| 608 | U16 topword; /* longest accepted word */ |
| 609 | bool longfold;/* saw a fold with a 1->n char mapping */ |
| 610 | } trie; |
| 611 | |
| 612 | /* special types - these members are used to store state for special |
| 613 | regops like eval, if/then, lookaround and the markpoint state */ |
| 614 | struct { |
| 615 | /* this first element must match u.yes */ |
| 616 | struct regmatch_state *prev_yes_state; |
| 617 | struct regmatch_state *prev_eval; |
| 618 | struct regmatch_state *prev_curlyx; |
| 619 | REGEXP *prev_rex; |
| 620 | U32 toggle_reg_flags; /* what bits in PL_reg_flags to |
| 621 | flip when transitioning between |
| 622 | inner and outer rexen */ |
| 623 | CHECKPOINT cp; /* remember current savestack indexes */ |
| 624 | CHECKPOINT lastcp; |
| 625 | U32 close_paren; /* which close bracket is our end */ |
| 626 | regnode *B; /* the node following us */ |
| 627 | } eval; |
| 628 | |
| 629 | struct { |
| 630 | /* this first element must match u.yes */ |
| 631 | struct regmatch_state *prev_yes_state; |
| 632 | I32 wanted; |
| 633 | I32 logical; /* saved copy of 'logical' var */ |
| 634 | regnode *me; /* the IFMATCH/SUSPEND/UNLESSM node */ |
| 635 | } ifmatch; /* and SUSPEND/UNLESSM */ |
| 636 | |
| 637 | struct { |
| 638 | /* this first element must match u.yes */ |
| 639 | struct regmatch_state *prev_yes_state; |
| 640 | struct regmatch_state *prev_mark; |
| 641 | SV* mark_name; |
| 642 | char *mark_loc; |
| 643 | } mark; |
| 644 | |
| 645 | struct { |
| 646 | int val; |
| 647 | } keeper; |
| 648 | |
| 649 | /* quantifiers - these members are used for storing state for |
| 650 | for the regops used to implement quantifiers */ |
| 651 | struct { |
| 652 | /* this first element must match u.yes */ |
| 653 | struct regmatch_state *prev_yes_state; |
| 654 | struct regmatch_state *prev_curlyx; /* previous cur_curlyx */ |
| 655 | regnode *me; /* the CURLYX node */ |
| 656 | regnode *B; /* the B node in /A*B/ */ |
| 657 | CHECKPOINT cp; /* remember current savestack index */ |
| 658 | bool minmod; |
| 659 | int parenfloor;/* how far back to strip paren data */ |
| 660 | |
| 661 | /* these two are modified by WHILEM */ |
| 662 | int count; /* how many instances of A we've matched */ |
| 663 | char *lastloc;/* where previous A matched (0-len detect) */ |
| 664 | } curlyx; |
| 665 | |
| 666 | struct { |
| 667 | /* this first element must match u.yes */ |
| 668 | struct regmatch_state *prev_yes_state; |
| 669 | struct regmatch_state *save_curlyx; |
| 670 | CHECKPOINT cp; /* remember current savestack indexes */ |
| 671 | CHECKPOINT lastcp; |
| 672 | char *save_lastloc; /* previous curlyx.lastloc */ |
| 673 | I32 cache_offset; |
| 674 | I32 cache_mask; |
| 675 | } whilem; |
| 676 | |
| 677 | struct { |
| 678 | /* this first element must match u.yes */ |
| 679 | struct regmatch_state *prev_yes_state; |
| 680 | I32 c1, c2; /* case fold search */ |
| 681 | CHECKPOINT cp; |
| 682 | I32 alen; /* length of first-matched A string */ |
| 683 | I32 count; |
| 684 | bool minmod; |
| 685 | regnode *A, *B; /* the nodes corresponding to /A*B/ */ |
| 686 | regnode *me; /* the curlym node */ |
| 687 | } curlym; |
| 688 | |
| 689 | struct { |
| 690 | U32 paren; |
| 691 | CHECKPOINT cp; |
| 692 | I32 c1, c2; /* case fold search */ |
| 693 | char *maxpos; /* highest possible point in string to match */ |
| 694 | char *oldloc; /* the previous locinput */ |
| 695 | int count; |
| 696 | int min, max; /* {m,n} */ |
| 697 | regnode *A, *B; /* the nodes corresponding to /A*B/ */ |
| 698 | } curly; /* and CURLYN/PLUS/STAR */ |
| 699 | |
| 700 | } u; |
| 701 | } regmatch_state; |
| 702 | |
| 703 | /* how many regmatch_state structs to allocate as a single slab. |
| 704 | * We do it in 4K blocks for efficiency. The "3" is 2 for the next/prev |
| 705 | * pointers, plus 1 for any mythical malloc overhead. */ |
| 706 | |
| 707 | #define PERL_REGMATCH_SLAB_SLOTS \ |
| 708 | ((4096 - 3 * sizeof (void*)) / sizeof(regmatch_state)) |
| 709 | |
| 710 | typedef struct regmatch_slab { |
| 711 | regmatch_state states[PERL_REGMATCH_SLAB_SLOTS]; |
| 712 | struct regmatch_slab *prev, *next; |
| 713 | } regmatch_slab; |
| 714 | |
| 715 | #define PL_reg_flags PL_reg_state.re_state_reg_flags |
| 716 | #define PL_bostr PL_reg_state.re_state_bostr |
| 717 | #define PL_reginput PL_reg_state.re_state_reginput |
| 718 | #define PL_regeol PL_reg_state.re_state_regeol |
| 719 | #define PL_regoffs PL_reg_state.re_state_regoffs |
| 720 | #define PL_reglastparen PL_reg_state.re_state_reglastparen |
| 721 | #define PL_reglastcloseparen PL_reg_state.re_state_reglastcloseparen |
| 722 | #define PL_reg_start_tmp PL_reg_state.re_state_reg_start_tmp |
| 723 | #define PL_reg_start_tmpl PL_reg_state.re_state_reg_start_tmpl |
| 724 | #define PL_reg_eval_set PL_reg_state.re_state_reg_eval_set |
| 725 | #define PL_reg_match_utf8 PL_reg_state.re_state_reg_match_utf8 |
| 726 | #define PL_reg_magic PL_reg_state.re_state_reg_magic |
| 727 | #define PL_reg_oldpos PL_reg_state.re_state_reg_oldpos |
| 728 | #define PL_reg_oldcurpm PL_reg_state.re_state_reg_oldcurpm |
| 729 | #define PL_reg_curpm PL_reg_state.re_state_reg_curpm |
| 730 | #define PL_reg_oldsaved PL_reg_state.re_state_reg_oldsaved |
| 731 | #define PL_reg_oldsavedlen PL_reg_state.re_state_reg_oldsavedlen |
| 732 | #define PL_reg_maxiter PL_reg_state.re_state_reg_maxiter |
| 733 | #define PL_reg_leftiter PL_reg_state.re_state_reg_leftiter |
| 734 | #define PL_reg_poscache PL_reg_state.re_state_reg_poscache |
| 735 | #define PL_reg_poscache_size PL_reg_state.re_state_reg_poscache_size |
| 736 | #define PL_regsize PL_reg_state.re_state_regsize |
| 737 | #define PL_reg_starttry PL_reg_state.re_state_reg_starttry |
| 738 | #define PL_nrs PL_reg_state.re_state_nrs |
| 739 | |
| 740 | struct re_save_state { |
| 741 | U32 re_state_reg_flags; /* from regexec.c */ |
| 742 | U32 re_state_reg_start_tmpl; /* from regexec.c */ |
| 743 | I32 re_state_reg_eval_set; /* from regexec.c */ |
| 744 | bool re_state_reg_match_utf8; /* from regexec.c */ |
| 745 | char *re_state_bostr; |
| 746 | char *re_state_reginput; /* String-input pointer. */ |
| 747 | char *re_state_regeol; /* End of input, for $ check. */ |
| 748 | regexp_paren_pair *re_state_regoffs; /* Pointer to start/end pairs */ |
| 749 | U32 *re_state_reglastparen; /* Similarly for lastparen. */ |
| 750 | U32 *re_state_reglastcloseparen; /* Similarly for lastcloseparen. */ |
| 751 | char **re_state_reg_start_tmp; /* from regexec.c */ |
| 752 | MAGIC *re_state_reg_magic; /* from regexec.c */ |
| 753 | PMOP *re_state_reg_oldcurpm; /* from regexec.c */ |
| 754 | PMOP *re_state_reg_curpm; /* from regexec.c */ |
| 755 | char *re_state_reg_oldsaved; /* old saved substr during match */ |
| 756 | STRLEN re_state_reg_oldsavedlen; /* old length of saved substr during match */ |
| 757 | STRLEN re_state_reg_poscache_size; /* size of pos cache of WHILEM */ |
| 758 | I32 re_state_reg_oldpos; /* from regexec.c */ |
| 759 | I32 re_state_reg_maxiter; /* max wait until caching pos */ |
| 760 | I32 re_state_reg_leftiter; /* wait until caching pos */ |
| 761 | U32 re_state_regsize; /* from regexec.c */ |
| 762 | char *re_state_reg_poscache; /* cache of pos of WHILEM */ |
| 763 | char *re_state_reg_starttry; /* from regexec.c */ |
| 764 | #ifdef PERL_OLD_COPY_ON_WRITE |
| 765 | SV *re_state_nrs; /* was placeholder: unused since 5.8.0 (5.7.2 patch #12027 for bug ID 20010815.012). Used to save rx->saved_copy */ |
| 766 | #endif |
| 767 | }; |
| 768 | |
| 769 | #define SAVESTACK_ALLOC_FOR_RE_SAVE_STATE \ |
| 770 | (1 + ((sizeof(struct re_save_state) - 1) / sizeof(*PL_savestack))) |
| 771 | |
| 772 | /* |
| 773 | * Local variables: |
| 774 | * c-indentation-style: bsd |
| 775 | * c-basic-offset: 4 |
| 776 | * indent-tabs-mode: t |
| 777 | * End: |
| 778 | * |
| 779 | * ex: set ts=8 sts=4 sw=4 noet: |
| 780 | */ |