3 * Copyright (C) 1993, 1994, 1996, 1997, 1999, 2000, 2001, 2003,
4 * 2005, 2006, 2007, 2008 by Larry Wall and others
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.
12 * Definitions etc. for regexp(3) routines.
14 * Caveat: this is V8 regexp(3) [actually, a reimplementation thereof],
15 * not the System V one.
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'*/
29 typedef struct regnode regnode;
31 struct reg_substr_data;
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 */
45 struct reg_substr_data {
46 U8 check_ix; /* index into data[] of check substr */
47 struct reg_substr_datum data[3]; /* Actual array */
51 #define SV_SAVED_COPY SV *saved_copy; /* If non-NULL, SV which is COW from original */
56 /* offsets within a string of a particular /(.)/ capture */
58 typedef struct regexp_paren_pair {
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
64 * "abc" =~ /(.(?{print "[$1]"}))+/
66 * This field is not part of the API. */
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)
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)
79 /* record the position of a (?{...}) within a pattern */
81 struct reg_code_block {
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.
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.
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. */ \
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/(?{..})/ */
140 typedef struct regexp {
145 #define RXp_PAREN_NAMES(rx) ((rx)->paren_names)
147 /* used for high speed searches */
148 typedef struct re_scream_pos_data_s
150 char **scream_olds; /* match pos */
151 SSize_t *scream_pos; /* Internal iterator of scream. */
152 } re_scream_pos_data;
154 /* regexp_engine structure. This is the dispatch table for regexes.
155 * Any regex engine implementation must be able to build one of these.
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_
165 const char * const strbeg,
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,
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,
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,
182 SV* (*qr_package)(pTHX_ REGEXP * const rx);
184 void* (*dupe) (pTHX_ REGEXP * const rx, CLONE_PARAMS *param);
186 REGEXP* (*op_comp) (pTHX_ SV ** const patternp, int pat_count,
187 OP *expr, const struct regexp_engine* eng,
189 bool *is_bare_re, U32 orig_rx_flags, U32 pm_flags);
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,
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 /* $& */
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.
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
221 /* Whether %+ or %- is being operated on */
222 #define RXapif_ONE 0x0100 /* %+ */
223 #define RXapif_ALL 0x0200 /* %- */
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
231 =head1 REGEXP Functions
233 =for apidoc Am|REGEXP *|SvRX|SV *sv
235 Convenience macro to get the REGEXP from a SV. This is approximately
236 equivalent to the following snippet:
241 sv = MUTABLE_SV(SvRV(sv));
242 if (SvTYPE(sv) == SVt_REGEXP)
245 NULL will be returned if a REGEXP* is not found.
247 =for apidoc Am|bool|SvRXOK|SV* sv
249 Returns a boolean indicating whether the SV (or the one it references)
252 If you want to do something with the REGEXP* later use SvRX instead
258 #define SvRX(sv) (Perl_get_re_arg(aTHX_ sv))
259 #define SvRXOK(sv) (Perl_get_re_arg(aTHX_ sv) ? TRUE : FALSE)
262 /* Flags stored in regexp->extflags
263 * These are used by code external to the regexp engine
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
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.
273 #include "op_reg_common.h"
275 #define RXf_PMf_STD_PMMOD (RXf_PMf_MULTILINE|RXf_PMf_SINGLELINE|RXf_PMf_FOLD|RXf_PMf_EXTENDED)
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
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)
287 /* chars and strings used as regex pattern modifiers
288 * Singular is a 'c'har, plural is a "string"
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.
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'
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"
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"
326 #define CHARSET_PAT_MODS ASCII_RESTRICT_PAT_MODS DEPENDS_PAT_MODS LOCALE_PAT_MODS UNICODE_PAT_MODS
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
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
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.
345 Set in Perl_pmruntime if op_flags & OPf_SPECIAL, i.e. split. Will
346 be used by regex engines to check whether they should set
349 #define RXf_SPLIT RXf_PMf_SPLIT
351 /* Currently the regex flags occupy a single 32-bit word. Not all bits are
352 * currently used. The lower bits are shared with their corresponding PMf flag
353 * bits, up to but not including _RXf_PMf_SHIFT_NEXT. The unused bits
354 * immediately follow; finally the used RXf-only (unshared) bits, so that the
355 * highest bit in the word is used. This gathers all the unused bits as a pool
356 * in the middle, like so: 11111111111111110000001111111111
357 * where the '1's represent used bits, and the '0's unused. This design allows
358 * us to allocate off one end of the pool if we need to add a shared bit, and
359 * off the other end if we need a non-shared bit, without disturbing the other
360 * bits. This maximizes the likelihood of being able to change things without
361 * breaking binary compatibility.
363 * To add shared bits, do so in op_reg_common.h. This should change
364 * _RXf_PMf_SHIFT_NEXT so that things won't compile. Then come to regexp.h and
365 * op.h and adjust the constant adders in the definitions of RXf_BASE_SHIFT and
366 * Pmf_BASE_SHIFT down by the number of shared bits you added. That's it.
367 * Things should be binary compatible. But if either of these gets to having
368 * to subtract rather than add, leave at 0 and instead adjust all the entries
369 * that are in terms of it. But if the first one of those is already
370 * RXf_BASE_SHIFT+0, there are no bits left, and a redesign is in order.
372 * To remove unshared bits, just delete its entry. If you're where breaking
373 * binary compatibility is ok to do, you might want to adjust things to move
374 * the newly opened space so that it gets absorbed into the common pool.
376 * To add unshared bits, first use up any gaps in the middle. Otherwise,
377 * allocate off the low end until you get to RXf_BASE_SHIFT+0. If that isn't
378 * enough, move RXf_BASE_SHIFT down (if possible) and add the new bit at the
379 * other end instead; this preserves binary compatibility.
381 * For the regexp bits, PL_reg_extflags_name[] in regnodes.h has a comment
382 * giving which bits are used/unused */
384 #define RXf_BASE_SHIFT (_RXf_PMf_SHIFT_NEXT + 5)
386 /* What we have seen */
387 #define RXf_NO_INPLACE_SUBST (1<<(RXf_BASE_SHIFT+2))
388 #define RXf_EVAL_SEEN (1<<(RXf_BASE_SHIFT+3))
391 #define RXf_UNBOUNDED_QUANTIFIER_SEEN (1<<(RXf_BASE_SHIFT+4))
392 #define RXf_CHECK_ALL (1<<(RXf_BASE_SHIFT+5))
395 #define RXf_MATCH_UTF8 (1<<(RXf_BASE_SHIFT+6)) /* $1 etc are utf8 */
398 #define RXf_USE_INTUIT_NOML (1<<(RXf_BASE_SHIFT+7))
399 #define RXf_USE_INTUIT_ML (1<<(RXf_BASE_SHIFT+8))
400 #define RXf_INTUIT_TAIL (1<<(RXf_BASE_SHIFT+9))
401 #define RXf_USE_INTUIT (RXf_USE_INTUIT_NOML|RXf_USE_INTUIT_ML)
403 /* Do we have some sort of anchor? */
404 #define RXf_IS_ANCHORED (1<<(RXf_BASE_SHIFT+10))
406 /* Copy and tainted info */
407 #define RXf_COPY_DONE (1<<(RXf_BASE_SHIFT+11))
409 /* post-execution: $1 et al are tainted */
410 #define RXf_TAINTED_SEEN (1<<(RXf_BASE_SHIFT+12))
411 /* this pattern was tainted during compilation */
412 #define RXf_TAINTED (1<<(RXf_BASE_SHIFT+13))
414 /* Flags indicating special patterns */
415 #define RXf_START_ONLY (1<<(RXf_BASE_SHIFT+14)) /* Pattern is /^/ */
416 #define RXf_SKIPWHITE (1<<(RXf_BASE_SHIFT+15)) /* Pattern is for a split " " */
417 #define RXf_WHITE (1<<(RXf_BASE_SHIFT+16)) /* Pattern is /\s+/ */
418 #define RXf_NULL (1U<<(RXf_BASE_SHIFT+17)) /* Pattern is // */
420 /* See comments at the beginning of these defines about adding bits. The
421 * highest bit position should be used, so that if RXf_BASE_SHIFT gets
422 * increased, the #error below will be triggered so that you will be reminded
423 * to adjust things at the other end to keep the bit positions unchanged */
424 #if RXf_BASE_SHIFT+17 > 31
425 # error Too many RXf_PMf bits used. See comments at beginning of these for what to do
429 * NOTE: if you modify any RXf flags you should run regen.pl or
430 * regen/regcomp.pl so that regnodes.h is updated with the changes.
434 #ifdef NO_TAINT_SUPPORT
435 # define RX_ISTAINTED(prog) 0
436 # define RX_TAINT_on(prog) NOOP
437 # define RXp_MATCH_TAINTED(prog) 0
438 # define RX_MATCH_TAINTED(prog) 0
439 # define RXp_MATCH_TAINTED_on(prog) NOOP
440 # define RX_MATCH_TAINTED_on(prog) NOOP
441 # define RX_MATCH_TAINTED_off(prog) NOOP
443 # define RX_ISTAINTED(prog) (RX_EXTFLAGS(prog) & RXf_TAINTED)
444 # define RX_TAINT_on(prog) (RX_EXTFLAGS(prog) |= RXf_TAINTED)
445 # define RXp_MATCH_TAINTED(prog) (RXp_EXTFLAGS(prog) & RXf_TAINTED_SEEN)
446 # define RX_MATCH_TAINTED(prog) (RX_EXTFLAGS(prog) & RXf_TAINTED_SEEN)
447 # define RXp_MATCH_TAINTED_on(prog) (RXp_EXTFLAGS(prog) |= RXf_TAINTED_SEEN)
448 # define RX_MATCH_TAINTED_on(prog) (RX_EXTFLAGS(prog) |= RXf_TAINTED_SEEN)
449 # define RX_MATCH_TAINTED_off(prog) (RX_EXTFLAGS(prog) &= ~RXf_TAINTED_SEEN)
452 #define RX_HAS_CUTGROUP(prog) ((prog)->intflags & PREGf_CUTGROUP_SEEN)
453 #define RX_MATCH_TAINTED_set(prog, t) ((t) \
454 ? RX_MATCH_TAINTED_on(prog) \
455 : RX_MATCH_TAINTED_off(prog))
457 #define RXp_MATCH_COPIED(prog) (RXp_EXTFLAGS(prog) & RXf_COPY_DONE)
458 #define RX_MATCH_COPIED(prog) (RX_EXTFLAGS(prog) & RXf_COPY_DONE)
459 #define RXp_MATCH_COPIED_on(prog) (RXp_EXTFLAGS(prog) |= RXf_COPY_DONE)
460 #define RX_MATCH_COPIED_on(prog) (RX_EXTFLAGS(prog) |= RXf_COPY_DONE)
461 #define RXp_MATCH_COPIED_off(prog) (RXp_EXTFLAGS(prog) &= ~RXf_COPY_DONE)
462 #define RX_MATCH_COPIED_off(prog) (RX_EXTFLAGS(prog) &= ~RXf_COPY_DONE)
463 #define RX_MATCH_COPIED_set(prog,t) ((t) \
464 ? RX_MATCH_COPIED_on(prog) \
465 : RX_MATCH_COPIED_off(prog))
467 #define RXp_EXTFLAGS(rx) ((rx)->extflags)
468 #define RXp_COMPFLAGS(rx) ((rx)->compflags)
470 /* For source compatibility. We used to store these explicitly. */
471 #define RX_PRECOMP(prog) (RX_WRAPPED(prog) + ReANY(prog)->pre_prefix)
472 #define RX_PRECOMP_const(prog) (RX_WRAPPED_const(prog) + ReANY(prog)->pre_prefix)
473 /* FIXME? Are we hardcoding too much here and constraining plugin extension
474 writers? Specifically, the value 1 assumes that the wrapped version always
475 has exactly one character at the end, a ')'. Will that always be true? */
476 #define RX_PRELEN(prog) (RX_WRAPLEN(prog) - ReANY(prog)->pre_prefix - 1)
477 #define RX_WRAPPED(prog) ReANY(prog)->xpv_len_u.xpvlenu_pv
478 #define RX_WRAPPED_const(prog) ((const char *)RX_WRAPPED(prog))
479 #define RX_WRAPLEN(prog) SvCUR(prog)
480 #define RX_CHECK_SUBSTR(prog) (ReANY(prog)->check_substr)
481 #define RX_REFCNT(prog) SvREFCNT(prog)
482 #define RX_EXTFLAGS(prog) RXp_EXTFLAGS(ReANY(prog))
483 #define RX_COMPFLAGS(prog) RXp_COMPFLAGS(ReANY(prog))
484 #define RX_ENGINE(prog) (ReANY(prog)->engine)
485 #define RX_SUBBEG(prog) (ReANY(prog)->subbeg)
486 #define RX_SUBOFFSET(prog) (ReANY(prog)->suboffset)
487 #define RX_SUBCOFFSET(prog) (ReANY(prog)->subcoffset)
488 #define RX_OFFS(prog) (ReANY(prog)->offs)
489 #define RX_NPARENS(prog) (ReANY(prog)->nparens)
490 #define RX_SUBLEN(prog) (ReANY(prog)->sublen)
491 #define RX_MINLEN(prog) (ReANY(prog)->minlen)
492 #define RX_MINLENRET(prog) (ReANY(prog)->minlenret)
493 #define RX_GOFS(prog) (ReANY(prog)->gofs)
494 #define RX_LASTPAREN(prog) (ReANY(prog)->lastparen)
495 #define RX_LASTCLOSEPAREN(prog) (ReANY(prog)->lastcloseparen)
496 #define RX_SAVED_COPY(prog) (ReANY(prog)->saved_copy)
497 /* last match was zero-length */
498 #define RX_ZERO_LEN(prog) \
499 (RX_OFFS(prog)[0].start + (SSize_t)RX_GOFS(prog) \
500 == RX_OFFS(prog)[0].end)
502 #endif /* PLUGGABLE_RE_EXTENSION */
504 /* Stuff that needs to be included in the pluggable extension goes below here */
507 #define RX_MATCH_COPY_FREE(rx) \
508 STMT_START {if (RX_SAVED_COPY(rx)) { \
509 SV_CHECK_THINKFIRST_COW_DROP(RX_SAVED_COPY(rx)); \
511 if (RX_MATCH_COPIED(rx)) { \
512 Safefree(RX_SUBBEG(rx)); \
513 RX_MATCH_COPIED_off(rx); \
516 #define RX_MATCH_COPY_FREE(rx) \
517 STMT_START {if (RX_MATCH_COPIED(rx)) { \
518 Safefree(RX_SUBBEG(rx)); \
519 RX_MATCH_COPIED_off(rx); \
523 #define RXp_MATCH_UTF8(prog) (RXp_EXTFLAGS(prog) & RXf_MATCH_UTF8)
524 #define RX_MATCH_UTF8(prog) (RX_EXTFLAGS(prog) & RXf_MATCH_UTF8)
525 #define RX_MATCH_UTF8_on(prog) (RX_EXTFLAGS(prog) |= RXf_MATCH_UTF8)
526 #define RX_MATCH_UTF8_off(prog) (RX_EXTFLAGS(prog) &= ~RXf_MATCH_UTF8)
527 #define RX_MATCH_UTF8_set(prog, t) ((t) \
528 ? RX_MATCH_UTF8_on(prog) \
529 : RX_MATCH_UTF8_off(prog))
531 /* Whether the pattern stored at RX_WRAPPED is in UTF-8 */
532 #define RX_UTF8(prog) SvUTF8(prog)
535 /* bits in flags arg of Perl_regexec_flags() */
537 #define REXEC_COPY_STR 0x01 /* Need to copy the string for captures. */
538 #define REXEC_CHECKED 0x02 /* re_intuit_start() already called. */
539 #define REXEC_SCREAM 0x04 /* currently unused. */
540 #define REXEC_IGNOREPOS 0x08 /* use stringarg, not pos(), for \G match */
541 #define REXEC_NOT_FIRST 0x10 /* This is another iteration of //g:
542 no need to copy string again */
544 /* under REXEC_COPY_STR, it's ok for the
545 engine (modulo PL_sawamperand etc)
546 to skip copying: ... */
547 #define REXEC_COPY_SKIP_PRE 0x20 /* ...the $` part of the string, or */
548 #define REXEC_COPY_SKIP_POST 0x40 /* ...the $' part of the string */
549 #define REXEC_FAIL_ON_UNDERFLOW 0x80 /* fail the match if $& would start before
550 the start pos (so s/.\G// would fail
551 on second iteration */
553 #if defined(__GNUC__) && !defined(PERL_GCC_BRACE_GROUPS_FORBIDDEN)
554 # define ReREFCNT_inc(re) \
556 /* This is here to generate a casting warning if incorrect. */ \
557 REGEXP *const _rerefcnt_inc = (re); \
558 assert(SvTYPE(_rerefcnt_inc) == SVt_REGEXP); \
559 SvREFCNT_inc(_rerefcnt_inc); \
562 # define ReREFCNT_dec(re) \
564 /* This is here to generate a casting warning if incorrect. */ \
565 REGEXP *const _rerefcnt_dec = (re); \
566 SvREFCNT_dec(_rerefcnt_dec); \
569 # define ReREFCNT_dec(re) SvREFCNT_dec(re)
570 # define ReREFCNT_inc(re) ((REGEXP *) SvREFCNT_inc(re))
572 #define ReANY(re) S_ReANY((const REGEXP *)(re))
574 /* FIXME for plugins. */
576 #define FBMcf_TAIL_DOLLAR 1
577 #define FBMcf_TAIL_DOLLARM 2
578 #define FBMcf_TAIL_Z 4
579 #define FBMcf_TAIL_z 8
580 #define FBMcf_TAIL (FBMcf_TAIL_DOLLAR|FBMcf_TAIL_DOLLARM|FBMcf_TAIL_Z|FBMcf_TAIL_z)
582 #define FBMrf_MULTILINE 1
584 struct regmatch_state;
585 struct regmatch_slab;
587 /* like regmatch_info_aux, but contains extra fields only needed if the
588 * pattern contains (?{}). If used, is snuck into the second slot in the
589 * regmatch_state stack at the start of execution */
593 PMOP *curpm; /* saved PL_curpm */
595 SV *saved_copy; /* saved saved_copy field from rex */
597 char *subbeg; /* saved subbeg field from rex */
598 STRLEN sublen; /* saved sublen field from rex */
599 STRLEN suboffset; /* saved suboffset field from rex */
600 STRLEN subcoffset; /* saved subcoffset field from rex */
601 MAGIC *pos_magic; /* pos() magic attached to $_ */
602 SSize_t pos; /* the original value of pos() in pos_magic */
603 U8 pos_flags; /* flags to be restored; currently only MGf_BYTES*/
604 } regmatch_info_aux_eval;
607 /* fields that logically live in regmatch_info, but which need cleaning
608 * up on croak(), and so are instead are snuck into the first slot in
609 * the regmatch_state stack at the start of execution */
612 regmatch_info_aux_eval *info_aux_eval;
613 struct regmatch_state *old_regmatch_state; /* saved PL_regmatch_state */
614 struct regmatch_slab *old_regmatch_slab; /* saved PL_regmatch_slab */
615 char *poscache; /* S-L cache of fail positions of WHILEMs */
619 /* some basic information about the current match that is created by
620 * Perl_regexec_flags and then passed to regtry(), regmatch() etc.
621 * It is allocated as a local var on the stack, so nothing should be
622 * stored in it that needs preserving or clearing up on croak().
623 * For that, see the aux_info and aux_info_eval members of the
624 * regmatch_state union. */
627 REGEXP *prog; /* the regex being executed */
628 const char * strbeg; /* real start of string */
629 char *strend; /* one byte beyond last char of match string */
630 char *till; /* matches shorter than this fail (see minlen arg) */
631 SV *sv; /* the SV string currently being matched */
632 char *ganch; /* position of \G anchor */
633 char *cutpoint; /* (*COMMIT) position (if any) */
634 regmatch_info_aux *info_aux; /* extra fields that need cleanup */
635 regmatch_info_aux_eval *info_aux_eval; /* extra saved state for (?{}) */
636 I32 poscache_maxiter; /* how many whilems todo before S-L cache kicks in */
637 I32 poscache_iter; /* current countdown from _maxiter to zero */
638 STRLEN poscache_size; /* size of regmatch_info_aux.poscache */
639 bool intuit; /* re_intuit_start() is the top-level caller */
640 bool is_utf8_pat; /* regex is utf8 */
641 bool is_utf8_target; /* string being matched is utf8 */
642 bool warned; /* we have issued a recursion warning; no need for more */
646 /* structures for holding and saving the state maintained by regmatch() */
648 #ifndef MAX_RECURSE_EVAL_NOCHANGE_DEPTH
649 #define MAX_RECURSE_EVAL_NOCHANGE_DEPTH 1000
652 typedef I32 CHECKPOINT;
654 typedef struct regmatch_state {
655 int resume_state; /* where to jump to on return */
656 char *locinput; /* where to backtrack in string on failure */
660 /* the 'info_aux' and 'info_aux_eval' union members are cuckoos in
661 * the nest. They aren't saved backtrack state; rather they
662 * represent one or two extra chunks of data that need allocating
663 * at the start of a match. These fields would logically live in
664 * the regmatch_info struct, except that is allocated on the
665 * C stack, and these fields are all things that require cleanup
666 * after a croak(), when the stack is lost.
667 * As a convenience, we just use the first 1 or 2 regmatch_state
668 * slots to store this info, as we will be allocating a slab of
669 * these anyway. Otherwise we'd have to malloc and then free them,
670 * or allocate them on the save stack (where they will get
671 * realloced if the save stack grows).
672 * info_aux contains the extra fields that are always needed;
673 * info_aux_eval contains extra fields that only needed if
674 * the pattern contains code blocks
675 * We split them into two separate structs to avoid increasing
676 * the size of the union.
679 regmatch_info_aux info_aux;
681 regmatch_info_aux_eval info_aux_eval;
683 /* this is a fake union member that matches the first element
684 * of each member that needs to store positive backtrack
687 struct regmatch_state *prev_yes_state;
690 /* branchlike members */
691 /* this is a fake union member that matches the first elements
692 * of each member that needs to behave like a branch */
694 /* this first element must match u.yes */
695 struct regmatch_state *prev_yes_state;
703 /* the first elements must match u.branchlike */
704 struct regmatch_state *prev_yes_state;
709 regnode *next_branch; /* next branch node */
713 /* the first elements must match u.branchlike */
714 struct regmatch_state *prev_yes_state;
719 U32 accepted; /* how many accepting states left */
720 bool longfold;/* saw a fold with a 1->n char mapping */
721 U16 *jump; /* positive offsets from me */
722 regnode *me; /* Which node am I - needed for jump tries*/
723 U8 *firstpos;/* pos in string of first trie match */
724 U32 firstchars;/* len in chars of firstpos from start */
725 U16 nextword;/* next word to try */
726 U16 topword; /* longest accepted word */
729 /* special types - these members are used to store state for special
730 regops like eval, if/then, lookaround and the markpoint state */
732 /* this first element must match u.yes */
733 struct regmatch_state *prev_yes_state;
734 struct regmatch_state *prev_eval;
735 struct regmatch_state *prev_curlyx;
737 CHECKPOINT cp; /* remember current savestack indexes */
739 U32 close_paren; /* which close bracket is our end */
740 regnode *B; /* the node following us */
744 /* this first element must match u.yes */
745 struct regmatch_state *prev_yes_state;
747 I32 logical; /* saved copy of 'logical' var */
748 regnode *me; /* the IFMATCH/SUSPEND/UNLESSM node */
749 } ifmatch; /* and SUSPEND/UNLESSM */
752 /* this first element must match u.yes */
753 struct regmatch_state *prev_yes_state;
754 struct regmatch_state *prev_mark;
763 /* quantifiers - these members are used for storing state for
764 for the regops used to implement quantifiers */
766 /* this first element must match u.yes */
767 struct regmatch_state *prev_yes_state;
768 struct regmatch_state *prev_curlyx; /* previous cur_curlyx */
769 regnode *me; /* the CURLYX node */
770 regnode *B; /* the B node in /A*B/ */
771 CHECKPOINT cp; /* remember current savestack index */
773 int parenfloor;/* how far back to strip paren data */
775 /* these two are modified by WHILEM */
776 int count; /* how many instances of A we've matched */
777 char *lastloc;/* where previous A matched (0-len detect) */
781 /* this first element must match u.yes */
782 struct regmatch_state *prev_yes_state;
783 struct regmatch_state *save_curlyx;
784 CHECKPOINT cp; /* remember current savestack indexes */
786 char *save_lastloc; /* previous curlyx.lastloc */
792 /* this first element must match u.yes */
793 struct regmatch_state *prev_yes_state;
794 int c1, c2; /* case fold search */
798 I32 alen; /* length of first-matched A string */
801 regnode *A, *B; /* the nodes corresponding to /A*B/ */
802 regnode *me; /* the curlym node */
803 U8 c1_utf8[UTF8_MAXBYTES+1]; /* */
804 U8 c2_utf8[UTF8_MAXBYTES+1];
812 int c1, c2; /* case fold search */
813 char *maxpos; /* highest possible point in string to match */
814 char *oldloc; /* the previous locinput */
816 int min, max; /* {m,n} */
817 regnode *A, *B; /* the nodes corresponding to /A*B/ */
818 U8 c1_utf8[UTF8_MAXBYTES+1]; /* */
819 U8 c2_utf8[UTF8_MAXBYTES+1];
820 } curly; /* and CURLYN/PLUS/STAR */
825 /* how many regmatch_state structs to allocate as a single slab.
826 * We do it in 4K blocks for efficiency. The "3" is 2 for the next/prev
827 * pointers, plus 1 for any mythical malloc overhead. */
829 #define PERL_REGMATCH_SLAB_SLOTS \
830 ((4096 - 3 * sizeof (void*)) / sizeof(regmatch_state))
832 typedef struct regmatch_slab {
833 regmatch_state states[PERL_REGMATCH_SLAB_SLOTS];
834 struct regmatch_slab *prev, *next;
841 * c-indentation-style: bsd
843 * indent-tabs-mode: nil
846 * ex: set ts=8 sts=4 sw=4 et: