5 * One Ring to rule them all, One Ring to find them
7 * [p.v of _The Lord of the Rings_, opening poem]
8 * [p.50 of _The Lord of the Rings_, I/iii: "The Shadow of the Past"]
9 * [p.254 of _The Lord of the Rings_, II/ii: "The Council of Elrond"]
12 /* This file contains functions for executing a regular expression. See
13 * also regcomp.c which funnily enough, contains functions for compiling
14 * a regular expression.
16 * This file is also copied at build time to ext/re/re_exec.c, where
17 * it's built with -DPERL_EXT_RE_BUILD -DPERL_EXT_RE_DEBUG -DPERL_EXT.
18 * This causes the main functions to be compiled under new names and with
19 * debugging support added, which makes "use re 'debug'" work.
22 /* NOTE: this is derived from Henry Spencer's regexp code, and should not
23 * confused with the original package (see point 3 below). Thanks, Henry!
26 /* Additional note: this code is very heavily munged from Henry's version
27 * in places. In some spots I've traded clarity for efficiency, so don't
28 * blame Henry for some of the lack of readability.
31 /* The names of the functions have been changed from regcomp and
32 * regexec to pregcomp and pregexec in order to avoid conflicts
33 * with the POSIX routines of the same names.
36 #ifdef PERL_EXT_RE_BUILD
41 * pregcomp and pregexec -- regsub and regerror are not used in perl
43 * Copyright (c) 1986 by University of Toronto.
44 * Written by Henry Spencer. Not derived from licensed software.
46 * Permission is granted to anyone to use this software for any
47 * purpose on any computer system, and to redistribute it freely,
48 * subject to the following restrictions:
50 * 1. The author is not responsible for the consequences of use of
51 * this software, no matter how awful, even if they arise
54 * 2. The origin of this software must not be misrepresented, either
55 * by explicit claim or by omission.
57 * 3. Altered versions must be plainly marked as such, and must not
58 * be misrepresented as being the original software.
60 **** Alterations to Henry's code are...
62 **** Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
63 **** 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008
64 **** by Larry Wall and others
66 **** You may distribute under the terms of either the GNU General Public
67 **** License or the Artistic License, as specified in the README file.
69 * Beware that some of this code is subtly aware of the way operator
70 * precedence is structured in regular expressions. Serious changes in
71 * regular-expression syntax might require a total rethink.
74 #define PERL_IN_REGEXEC_C
77 #ifdef PERL_IN_XSUB_RE
83 #define RF_tainted 1 /* tainted information used? */
84 #define RF_warned 2 /* warned about big count? */
86 #define RF_utf8 8 /* Pattern contains multibyte chars? */
88 #define UTF_PATTERN ((PL_reg_flags & RF_utf8) != 0)
90 #define RS_init 1 /* eval environment created */
91 #define RS_set 2 /* replsv value is set */
97 #define REGINCLASS(prog,p,c) (ANYOF_FLAGS(p) ? reginclass(prog,p,c,0,0) : ANYOF_BITMAP_TEST(p,*(c)))
103 #define CHR_SVLEN(sv) (utf8_target ? sv_len_utf8(sv) : SvCUR(sv))
104 #define CHR_DIST(a,b) (PL_reg_match_utf8 ? utf8_distance(a,b) : a - b)
106 #define HOPc(pos,off) \
107 (char *)(PL_reg_match_utf8 \
108 ? reghop3((U8*)pos, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr)) \
110 #define HOPBACKc(pos, off) \
111 (char*)(PL_reg_match_utf8\
112 ? reghopmaybe3((U8*)pos, -off, (U8*)PL_bostr) \
113 : (pos - off >= PL_bostr) \
117 #define HOP3(pos,off,lim) (PL_reg_match_utf8 ? reghop3((U8*)(pos), off, (U8*)(lim)) : (U8*)(pos + off))
118 #define HOP3c(pos,off,lim) ((char*)HOP3(pos,off,lim))
120 /* these are unrolled below in the CCC_TRY_XXX defined */
121 #define LOAD_UTF8_CHARCLASS(class,str) STMT_START { \
122 if (!CAT2(PL_utf8_,class)) { bool ok; ENTER; save_re_context(); ok=CAT2(is_utf8_,class)((const U8*)str); assert(ok); LEAVE; } } STMT_END
124 /* Doesn't do an assert to verify that is correct */
125 #define LOAD_UTF8_CHARCLASS_NO_CHECK(class) STMT_START { \
126 if (!CAT2(PL_utf8_,class)) { bool ok; ENTER; save_re_context(); ok=CAT2(is_utf8_,class)((const U8*)" "); LEAVE; } } STMT_END
128 #define LOAD_UTF8_CHARCLASS_ALNUM() LOAD_UTF8_CHARCLASS(alnum,"a")
129 #define LOAD_UTF8_CHARCLASS_DIGIT() LOAD_UTF8_CHARCLASS(digit,"0")
130 #define LOAD_UTF8_CHARCLASS_SPACE() LOAD_UTF8_CHARCLASS(space," ")
132 #define LOAD_UTF8_CHARCLASS_GCB() /* Grapheme cluster boundaries */ \
133 LOAD_UTF8_CHARCLASS(X_begin, " "); \
134 LOAD_UTF8_CHARCLASS(X_non_hangul, "A"); \
135 /* These are utf8 constants, and not utf-ebcdic constants, so the \
136 * assert should likely and hopefully fail on an EBCDIC machine */ \
137 LOAD_UTF8_CHARCLASS(X_extend, "\xcc\x80"); /* U+0300 */ \
139 /* No asserts are done for these, in case called on an early \
140 * Unicode version in which they map to nothing */ \
141 LOAD_UTF8_CHARCLASS_NO_CHECK(X_prepend);/* U+0E40 "\xe0\xb9\x80" */ \
142 LOAD_UTF8_CHARCLASS_NO_CHECK(X_L); /* U+1100 "\xe1\x84\x80" */ \
143 LOAD_UTF8_CHARCLASS_NO_CHECK(X_LV); /* U+AC00 "\xea\xb0\x80" */ \
144 LOAD_UTF8_CHARCLASS_NO_CHECK(X_LVT); /* U+AC01 "\xea\xb0\x81" */ \
145 LOAD_UTF8_CHARCLASS_NO_CHECK(X_LV_LVT_V);/* U+AC01 "\xea\xb0\x81" */\
146 LOAD_UTF8_CHARCLASS_NO_CHECK(X_T); /* U+11A8 "\xe1\x86\xa8" */ \
147 LOAD_UTF8_CHARCLASS_NO_CHECK(X_V) /* U+1160 "\xe1\x85\xa0" */
150 We dont use PERL_LEGACY_UNICODE_CHARCLASS_MAPPINGS as the direct test
151 so that it is possible to override the option here without having to
152 rebuild the entire core. as we are required to do if we change regcomp.h
153 which is where PERL_LEGACY_UNICODE_CHARCLASS_MAPPINGS is defined.
155 #if PERL_LEGACY_UNICODE_CHARCLASS_MAPPINGS
156 #define BROKEN_UNICODE_CHARCLASS_MAPPINGS
159 #ifdef BROKEN_UNICODE_CHARCLASS_MAPPINGS
160 #define LOAD_UTF8_CHARCLASS_PERL_WORD() LOAD_UTF8_CHARCLASS_ALNUM()
161 #define LOAD_UTF8_CHARCLASS_PERL_SPACE() LOAD_UTF8_CHARCLASS_SPACE()
162 #define LOAD_UTF8_CHARCLASS_POSIX_DIGIT() LOAD_UTF8_CHARCLASS_DIGIT()
163 #define RE_utf8_perl_word PL_utf8_alnum
164 #define RE_utf8_perl_space PL_utf8_space
165 #define RE_utf8_posix_digit PL_utf8_digit
166 #define perl_word alnum
167 #define perl_space space
168 #define posix_digit digit
170 #define LOAD_UTF8_CHARCLASS_PERL_WORD() LOAD_UTF8_CHARCLASS(perl_word,"a")
171 #define LOAD_UTF8_CHARCLASS_PERL_SPACE() LOAD_UTF8_CHARCLASS(perl_space," ")
172 #define LOAD_UTF8_CHARCLASS_POSIX_DIGIT() LOAD_UTF8_CHARCLASS(posix_digit,"0")
173 #define RE_utf8_perl_word PL_utf8_perl_word
174 #define RE_utf8_perl_space PL_utf8_perl_space
175 #define RE_utf8_posix_digit PL_utf8_posix_digit
179 #define CCC_TRY_AFF(NAME,NAMEL,CLASS,STR,LCFUNC_utf8,FUNC,LCFUNC) \
181 PL_reg_flags |= RF_tainted; \
186 if (utf8_target && UTF8_IS_CONTINUED(nextchr)) { \
187 if (!CAT2(PL_utf8_,CLASS)) { \
191 ok=CAT2(is_utf8_,CLASS)((const U8*)STR); \
195 if (!(OP(scan) == NAME \
196 ? cBOOL(swash_fetch(CAT2(PL_utf8_,CLASS), (U8*)locinput, utf8_target)) \
197 : LCFUNC_utf8((U8*)locinput))) \
201 locinput += PL_utf8skip[nextchr]; \
202 nextchr = UCHARAT(locinput); \
205 if (!(OP(scan) == NAME ? FUNC(nextchr) : LCFUNC(nextchr))) \
207 nextchr = UCHARAT(++locinput); \
210 #define CCC_TRY_NEG(NAME,NAMEL,CLASS,STR,LCFUNC_utf8,FUNC,LCFUNC) \
212 PL_reg_flags |= RF_tainted; \
215 if (!nextchr && locinput >= PL_regeol) \
217 if (utf8_target && UTF8_IS_CONTINUED(nextchr)) { \
218 if (!CAT2(PL_utf8_,CLASS)) { \
222 ok=CAT2(is_utf8_,CLASS)((const U8*)STR); \
226 if ((OP(scan) == NAME \
227 ? cBOOL(swash_fetch(CAT2(PL_utf8_,CLASS), (U8*)locinput, utf8_target)) \
228 : LCFUNC_utf8((U8*)locinput))) \
232 locinput += PL_utf8skip[nextchr]; \
233 nextchr = UCHARAT(locinput); \
236 if ((OP(scan) == NAME ? FUNC(nextchr) : LCFUNC(nextchr))) \
238 nextchr = UCHARAT(++locinput); \
245 /* TODO: Combine JUMPABLE and HAS_TEXT to cache OP(rn) */
247 /* for use after a quantifier and before an EXACT-like node -- japhy */
248 /* it would be nice to rework regcomp.sym to generate this stuff. sigh
250 * NOTE that *nothing* that affects backtracking should be in here, specifically
251 * VERBS must NOT be included. JUMPABLE is used to determine if we can ignore a
252 * node that is in between two EXACT like nodes when ascertaining what the required
253 * "follow" character is. This should probably be moved to regex compile time
254 * although it may be done at run time beause of the REF possibility - more
255 * investigation required. -- demerphq
257 #define JUMPABLE(rn) ( \
259 (OP(rn) == CLOSE && (!cur_eval || cur_eval->u.eval.close_paren != ARG(rn))) || \
261 OP(rn) == SUSPEND || OP(rn) == IFMATCH || \
262 OP(rn) == PLUS || OP(rn) == MINMOD || \
264 (PL_regkind[OP(rn)] == CURLY && ARG1(rn) > 0) \
266 #define IS_EXACT(rn) (PL_regkind[OP(rn)] == EXACT)
268 #define HAS_TEXT(rn) ( IS_EXACT(rn) || PL_regkind[OP(rn)] == REF )
271 /* Currently these are only used when PL_regkind[OP(rn)] == EXACT so
272 we don't need this definition. */
273 #define IS_TEXT(rn) ( OP(rn)==EXACT || OP(rn)==REF || OP(rn)==NREF )
274 #define IS_TEXTF(rn) ( OP(rn)==EXACTF || OP(rn)==REFF || OP(rn)==NREFF )
275 #define IS_TEXTFL(rn) ( OP(rn)==EXACTFL || OP(rn)==REFFL || OP(rn)==NREFFL )
278 /* ... so we use this as its faster. */
279 #define IS_TEXT(rn) ( OP(rn)==EXACT )
280 #define IS_TEXTF(rn) ( OP(rn)==EXACTF )
281 #define IS_TEXTFL(rn) ( OP(rn)==EXACTFL )
286 Search for mandatory following text node; for lookahead, the text must
287 follow but for lookbehind (rn->flags != 0) we skip to the next step.
289 #define FIND_NEXT_IMPT(rn) STMT_START { \
290 while (JUMPABLE(rn)) { \
291 const OPCODE type = OP(rn); \
292 if (type == SUSPEND || PL_regkind[type] == CURLY) \
293 rn = NEXTOPER(NEXTOPER(rn)); \
294 else if (type == PLUS) \
296 else if (type == IFMATCH) \
297 rn = (rn->flags == 0) ? NEXTOPER(NEXTOPER(rn)) : rn + ARG(rn); \
298 else rn += NEXT_OFF(rn); \
303 static void restore_pos(pTHX_ void *arg);
305 #define REGCP_PAREN_ELEMS 4
306 #define REGCP_OTHER_ELEMS 5
307 #define REGCP_FRAME_ELEMS 1
308 /* REGCP_FRAME_ELEMS are not part of the REGCP_OTHER_ELEMS and
309 * are needed for the regexp context stack bookkeeping. */
312 S_regcppush(pTHX_ I32 parenfloor)
315 const int retval = PL_savestack_ix;
316 const int paren_elems_to_push = (PL_regsize - parenfloor) * REGCP_PAREN_ELEMS;
317 const UV total_elems = paren_elems_to_push + REGCP_OTHER_ELEMS;
318 const UV elems_shifted = total_elems << SAVE_TIGHT_SHIFT;
320 GET_RE_DEBUG_FLAGS_DECL;
322 if (paren_elems_to_push < 0)
323 Perl_croak(aTHX_ "panic: paren_elems_to_push < 0");
325 if ((elems_shifted >> SAVE_TIGHT_SHIFT) != total_elems)
326 Perl_croak(aTHX_ "panic: paren_elems_to_push offset %"UVuf
327 " out of range (%lu-%ld)",
328 total_elems, (unsigned long)PL_regsize, (long)parenfloor);
330 SSGROW(total_elems + REGCP_FRAME_ELEMS);
332 for (p = PL_regsize; p > parenfloor; p--) {
333 /* REGCP_PARENS_ELEMS are pushed per pairs of parentheses. */
334 SSPUSHINT(PL_regoffs[p].end);
335 SSPUSHINT(PL_regoffs[p].start);
336 SSPUSHPTR(PL_reg_start_tmp[p]);
338 DEBUG_BUFFERS_r(PerlIO_printf(Perl_debug_log,
339 " saving \\%"UVuf" %"IVdf"(%"IVdf")..%"IVdf"\n",
340 (UV)p, (IV)PL_regoffs[p].start,
341 (IV)(PL_reg_start_tmp[p] - PL_bostr),
342 (IV)PL_regoffs[p].end
345 /* REGCP_OTHER_ELEMS are pushed in any case, parentheses or no. */
346 SSPUSHPTR(PL_regoffs);
347 SSPUSHINT(PL_regsize);
348 SSPUSHINT(*PL_reglastparen);
349 SSPUSHINT(*PL_reglastcloseparen);
350 SSPUSHPTR(PL_reginput);
351 SSPUSHUV(SAVEt_REGCONTEXT | elems_shifted); /* Magic cookie. */
356 /* These are needed since we do not localize EVAL nodes: */
357 #define REGCP_SET(cp) \
359 PerlIO_printf(Perl_debug_log, \
360 " Setting an EVAL scope, savestack=%"IVdf"\n", \
361 (IV)PL_savestack_ix)); \
364 #define REGCP_UNWIND(cp) \
366 if (cp != PL_savestack_ix) \
367 PerlIO_printf(Perl_debug_log, \
368 " Clearing an EVAL scope, savestack=%"IVdf"..%"IVdf"\n", \
369 (IV)(cp), (IV)PL_savestack_ix)); \
373 S_regcppop(pTHX_ const regexp *rex)
378 GET_RE_DEBUG_FLAGS_DECL;
380 PERL_ARGS_ASSERT_REGCPPOP;
382 /* Pop REGCP_OTHER_ELEMS before the parentheses loop starts. */
384 assert((i & SAVE_MASK) == SAVEt_REGCONTEXT); /* Check that the magic cookie is there. */
385 i >>= SAVE_TIGHT_SHIFT; /* Parentheses elements to pop. */
386 input = (char *) SSPOPPTR;
387 *PL_reglastcloseparen = SSPOPINT;
388 *PL_reglastparen = SSPOPINT;
389 PL_regsize = SSPOPINT;
390 PL_regoffs=(regexp_paren_pair *) SSPOPPTR;
392 i -= REGCP_OTHER_ELEMS;
393 /* Now restore the parentheses context. */
394 for ( ; i > 0; i -= REGCP_PAREN_ELEMS) {
396 U32 paren = (U32)SSPOPINT;
397 PL_reg_start_tmp[paren] = (char *) SSPOPPTR;
398 PL_regoffs[paren].start = SSPOPINT;
400 if (paren <= *PL_reglastparen)
401 PL_regoffs[paren].end = tmps;
403 PerlIO_printf(Perl_debug_log,
404 " restoring \\%"UVuf" to %"IVdf"(%"IVdf")..%"IVdf"%s\n",
405 (UV)paren, (IV)PL_regoffs[paren].start,
406 (IV)(PL_reg_start_tmp[paren] - PL_bostr),
407 (IV)PL_regoffs[paren].end,
408 (paren > *PL_reglastparen ? "(no)" : ""));
412 if (*PL_reglastparen + 1 <= rex->nparens) {
413 PerlIO_printf(Perl_debug_log,
414 " restoring \\%"IVdf"..\\%"IVdf" to undef\n",
415 (IV)(*PL_reglastparen + 1), (IV)rex->nparens);
419 /* It would seem that the similar code in regtry()
420 * already takes care of this, and in fact it is in
421 * a better location to since this code can #if 0-ed out
422 * but the code in regtry() is needed or otherwise tests
423 * requiring null fields (pat.t#187 and split.t#{13,14}
424 * (as of patchlevel 7877) will fail. Then again,
425 * this code seems to be necessary or otherwise
426 * this erroneously leaves $1 defined: "1" =~ /^(?:(\d)x)?\d$/
427 * --jhi updated by dapm */
428 for (i = *PL_reglastparen + 1; i <= rex->nparens; i++) {
430 PL_regoffs[i].start = -1;
431 PL_regoffs[i].end = -1;
437 #define regcpblow(cp) LEAVE_SCOPE(cp) /* Ignores regcppush()ed data. */
440 * pregexec and friends
443 #ifndef PERL_IN_XSUB_RE
445 - pregexec - match a regexp against a string
448 Perl_pregexec(pTHX_ REGEXP * const prog, char* stringarg, register char *strend,
449 char *strbeg, I32 minend, SV *screamer, U32 nosave)
450 /* strend: pointer to null at end of string */
451 /* strbeg: real beginning of string */
452 /* minend: end of match must be >=minend after stringarg. */
453 /* nosave: For optimizations. */
455 PERL_ARGS_ASSERT_PREGEXEC;
458 regexec_flags(prog, stringarg, strend, strbeg, minend, screamer, NULL,
459 nosave ? 0 : REXEC_COPY_STR);
464 * Need to implement the following flags for reg_anch:
466 * USE_INTUIT_NOML - Useful to call re_intuit_start() first
468 * INTUIT_AUTORITATIVE_NOML - Can trust a positive answer
469 * INTUIT_AUTORITATIVE_ML
470 * INTUIT_ONCE_NOML - Intuit can match in one location only.
473 * Another flag for this function: SECOND_TIME (so that float substrs
474 * with giant delta may be not rechecked).
477 /* Assumptions: if ANCH_GPOS, then strpos is anchored. XXXX Check GPOS logic */
479 /* If SCREAM, then SvPVX_const(sv) should be compatible with strpos and strend.
480 Otherwise, only SvCUR(sv) is used to get strbeg. */
482 /* XXXX We assume that strpos is strbeg unless sv. */
484 /* XXXX Some places assume that there is a fixed substring.
485 An update may be needed if optimizer marks as "INTUITable"
486 RExen without fixed substrings. Similarly, it is assumed that
487 lengths of all the strings are no more than minlen, thus they
488 cannot come from lookahead.
489 (Or minlen should take into account lookahead.)
490 NOTE: Some of this comment is not correct. minlen does now take account
491 of lookahead/behind. Further research is required. -- demerphq
495 /* A failure to find a constant substring means that there is no need to make
496 an expensive call to REx engine, thus we celebrate a failure. Similarly,
497 finding a substring too deep into the string means that less calls to
498 regtry() should be needed.
500 REx compiler's optimizer found 4 possible hints:
501 a) Anchored substring;
503 c) Whether we are anchored (beginning-of-line or \G);
504 d) First node (of those at offset 0) which may distingush positions;
505 We use a)b)d) and multiline-part of c), and try to find a position in the
506 string which does not contradict any of them.
509 /* Most of decisions we do here should have been done at compile time.
510 The nodes of the REx which we used for the search should have been
511 deleted from the finite automaton. */
514 Perl_re_intuit_start(pTHX_ REGEXP * const rx, SV *sv, char *strpos,
515 char *strend, const U32 flags, re_scream_pos_data *data)
518 struct regexp *const prog = (struct regexp *)SvANY(rx);
519 register I32 start_shift = 0;
520 /* Should be nonnegative! */
521 register I32 end_shift = 0;
526 const bool utf8_target = (sv && SvUTF8(sv)) ? 1 : 0; /* if no sv we have to assume bytes */
528 register char *other_last = NULL; /* other substr checked before this */
529 char *check_at = NULL; /* check substr found at this pos */
530 const I32 multiline = prog->extflags & RXf_PMf_MULTILINE;
531 RXi_GET_DECL(prog,progi);
533 const char * const i_strpos = strpos;
535 GET_RE_DEBUG_FLAGS_DECL;
537 PERL_ARGS_ASSERT_RE_INTUIT_START;
539 RX_MATCH_UTF8_set(rx,utf8_target);
542 PL_reg_flags |= RF_utf8;
545 debug_start_match(rx, utf8_target, strpos, strend,
546 sv ? "Guessing start of match in sv for"
547 : "Guessing start of match in string for");
550 /* CHR_DIST() would be more correct here but it makes things slow. */
551 if (prog->minlen > strend - strpos) {
552 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
553 "String too short... [re_intuit_start]\n"));
557 strbeg = (sv && SvPOK(sv)) ? strend - SvCUR(sv) : strpos;
560 if (!prog->check_utf8 && prog->check_substr)
561 to_utf8_substr(prog);
562 check = prog->check_utf8;
564 if (!prog->check_substr && prog->check_utf8)
565 to_byte_substr(prog);
566 check = prog->check_substr;
568 if (check == &PL_sv_undef) {
569 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
570 "Non-utf8 string cannot match utf8 check string\n"));
573 if (prog->extflags & RXf_ANCH) { /* Match at beg-of-str or after \n */
574 ml_anch = !( (prog->extflags & RXf_ANCH_SINGLE)
575 || ( (prog->extflags & RXf_ANCH_BOL)
576 && !multiline ) ); /* Check after \n? */
579 if ( !(prog->extflags & RXf_ANCH_GPOS) /* Checked by the caller */
580 && !(prog->intflags & PREGf_IMPLICIT) /* not a real BOL */
581 /* SvCUR is not set on references: SvRV and SvPVX_const overlap */
583 && (strpos != strbeg)) {
584 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Not at start...\n"));
587 if (prog->check_offset_min == prog->check_offset_max &&
588 !(prog->extflags & RXf_CANY_SEEN)) {
589 /* Substring at constant offset from beg-of-str... */
592 s = HOP3c(strpos, prog->check_offset_min, strend);
595 slen = SvCUR(check); /* >= 1 */
597 if ( strend - s > slen || strend - s < slen - 1
598 || (strend - s == slen && strend[-1] != '\n')) {
599 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "String too long...\n"));
602 /* Now should match s[0..slen-2] */
604 if (slen && (*SvPVX_const(check) != *s
606 && memNE(SvPVX_const(check), s, slen)))) {
608 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "String not equal...\n"));
612 else if (*SvPVX_const(check) != *s
613 || ((slen = SvCUR(check)) > 1
614 && memNE(SvPVX_const(check), s, slen)))
617 goto success_at_start;
620 /* Match is anchored, but substr is not anchored wrt beg-of-str. */
622 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
623 end_shift = prog->check_end_shift;
626 const I32 end = prog->check_offset_max + CHR_SVLEN(check)
627 - (SvTAIL(check) != 0);
628 const I32 eshift = CHR_DIST((U8*)strend, (U8*)s) - end;
630 if (end_shift < eshift)
634 else { /* Can match at random position */
637 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
638 end_shift = prog->check_end_shift;
640 /* end shift should be non negative here */
643 #ifdef QDEBUGGING /* 7/99: reports of failure (with the older version) */
645 Perl_croak(aTHX_ "panic: end_shift: %"IVdf" pattern:\n%s\n ",
646 (IV)end_shift, RX_PRECOMP(prog));
650 /* Find a possible match in the region s..strend by looking for
651 the "check" substring in the region corrected by start/end_shift. */
654 I32 srch_start_shift = start_shift;
655 I32 srch_end_shift = end_shift;
656 if (srch_start_shift < 0 && strbeg - s > srch_start_shift) {
657 srch_end_shift -= ((strbeg - s) - srch_start_shift);
658 srch_start_shift = strbeg - s;
660 DEBUG_OPTIMISE_MORE_r({
661 PerlIO_printf(Perl_debug_log, "Check offset min: %"IVdf" Start shift: %"IVdf" End shift %"IVdf" Real End Shift: %"IVdf"\n",
662 (IV)prog->check_offset_min,
663 (IV)srch_start_shift,
665 (IV)prog->check_end_shift);
668 if (flags & REXEC_SCREAM) {
669 I32 p = -1; /* Internal iterator of scream. */
670 I32 * const pp = data ? data->scream_pos : &p;
672 if (PL_screamfirst[BmRARE(check)] >= 0
673 || ( BmRARE(check) == '\n'
674 && (BmPREVIOUS(check) == SvCUR(check) - 1)
676 s = screaminstr(sv, check,
677 srch_start_shift + (s - strbeg), srch_end_shift, pp, 0);
680 /* we may be pointing at the wrong string */
681 if (s && RXp_MATCH_COPIED(prog))
682 s = strbeg + (s - SvPVX_const(sv));
684 *data->scream_olds = s;
689 if (prog->extflags & RXf_CANY_SEEN) {
690 start_point= (U8*)(s + srch_start_shift);
691 end_point= (U8*)(strend - srch_end_shift);
693 start_point= HOP3(s, srch_start_shift, srch_start_shift < 0 ? strbeg : strend);
694 end_point= HOP3(strend, -srch_end_shift, strbeg);
696 DEBUG_OPTIMISE_MORE_r({
697 PerlIO_printf(Perl_debug_log, "fbm_instr len=%d str=<%.*s>\n",
698 (int)(end_point - start_point),
699 (int)(end_point - start_point) > 20 ? 20 : (int)(end_point - start_point),
703 s = fbm_instr( start_point, end_point,
704 check, multiline ? FBMrf_MULTILINE : 0);
707 /* Update the count-of-usability, remove useless subpatterns,
711 RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0),
712 SvPVX_const(check), RE_SV_DUMPLEN(check), 30);
713 PerlIO_printf(Perl_debug_log, "%s %s substr %s%s%s",
714 (s ? "Found" : "Did not find"),
715 (check == (utf8_target ? prog->anchored_utf8 : prog->anchored_substr)
716 ? "anchored" : "floating"),
719 (s ? " at offset " : "...\n") );
724 /* Finish the diagnostic message */
725 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%ld...\n", (long)(s - i_strpos)) );
727 /* XXX dmq: first branch is for positive lookbehind...
728 Our check string is offset from the beginning of the pattern.
729 So we need to do any stclass tests offset forward from that
738 /* Got a candidate. Check MBOL anchoring, and the *other* substr.
739 Start with the other substr.
740 XXXX no SCREAM optimization yet - and a very coarse implementation
741 XXXX /ttx+/ results in anchored="ttx", floating="x". floating will
742 *always* match. Probably should be marked during compile...
743 Probably it is right to do no SCREAM here...
746 if (utf8_target ? (prog->float_utf8 && prog->anchored_utf8)
747 : (prog->float_substr && prog->anchored_substr))
749 /* Take into account the "other" substring. */
750 /* XXXX May be hopelessly wrong for UTF... */
753 if (check == (utf8_target ? prog->float_utf8 : prog->float_substr)) {
756 char * const last = HOP3c(s, -start_shift, strbeg);
758 char * const saved_s = s;
761 t = s - prog->check_offset_max;
762 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
764 || ((t = (char*)reghopmaybe3((U8*)s, -(prog->check_offset_max), (U8*)strpos))
769 t = HOP3c(t, prog->anchored_offset, strend);
770 if (t < other_last) /* These positions already checked */
772 last2 = last1 = HOP3c(strend, -prog->minlen, strbeg);
775 /* XXXX It is not documented what units *_offsets are in.
776 We assume bytes, but this is clearly wrong.
777 Meaning this code needs to be carefully reviewed for errors.
781 /* On end-of-str: see comment below. */
782 must = utf8_target ? prog->anchored_utf8 : prog->anchored_substr;
783 if (must == &PL_sv_undef) {
785 DEBUG_r(must = prog->anchored_utf8); /* for debug */
790 HOP3(HOP3(last1, prog->anchored_offset, strend)
791 + SvCUR(must), -(SvTAIL(must)!=0), strbeg),
793 multiline ? FBMrf_MULTILINE : 0
796 RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0),
797 SvPVX_const(must), RE_SV_DUMPLEN(must), 30);
798 PerlIO_printf(Perl_debug_log, "%s anchored substr %s%s",
799 (s ? "Found" : "Contradicts"),
800 quoted, RE_SV_TAIL(must));
805 if (last1 >= last2) {
806 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
807 ", giving up...\n"));
810 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
811 ", trying floating at offset %ld...\n",
812 (long)(HOP3c(saved_s, 1, strend) - i_strpos)));
813 other_last = HOP3c(last1, prog->anchored_offset+1, strend);
814 s = HOP3c(last, 1, strend);
818 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
819 (long)(s - i_strpos)));
820 t = HOP3c(s, -prog->anchored_offset, strbeg);
821 other_last = HOP3c(s, 1, strend);
829 else { /* Take into account the floating substring. */
831 char * const saved_s = s;
834 t = HOP3c(s, -start_shift, strbeg);
836 HOP3c(strend, -prog->minlen + prog->float_min_offset, strbeg);
837 if (CHR_DIST((U8*)last, (U8*)t) > prog->float_max_offset)
838 last = HOP3c(t, prog->float_max_offset, strend);
839 s = HOP3c(t, prog->float_min_offset, strend);
842 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
843 must = utf8_target ? prog->float_utf8 : prog->float_substr;
844 /* fbm_instr() takes into account exact value of end-of-str
845 if the check is SvTAIL(ed). Since false positives are OK,
846 and end-of-str is not later than strend we are OK. */
847 if (must == &PL_sv_undef) {
849 DEBUG_r(must = prog->float_utf8); /* for debug message */
852 s = fbm_instr((unsigned char*)s,
853 (unsigned char*)last + SvCUR(must)
855 must, multiline ? FBMrf_MULTILINE : 0);
857 RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0),
858 SvPVX_const(must), RE_SV_DUMPLEN(must), 30);
859 PerlIO_printf(Perl_debug_log, "%s floating substr %s%s",
860 (s ? "Found" : "Contradicts"),
861 quoted, RE_SV_TAIL(must));
865 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
866 ", giving up...\n"));
869 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
870 ", trying anchored starting at offset %ld...\n",
871 (long)(saved_s + 1 - i_strpos)));
873 s = HOP3c(t, 1, strend);
877 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
878 (long)(s - i_strpos)));
879 other_last = s; /* Fix this later. --Hugo */
889 t= (char*)HOP3( s, -prog->check_offset_max, (prog->check_offset_max<0) ? strend : strpos);
891 DEBUG_OPTIMISE_MORE_r(
892 PerlIO_printf(Perl_debug_log,
893 "Check offset min:%"IVdf" max:%"IVdf" S:%"IVdf" t:%"IVdf" D:%"IVdf" end:%"IVdf"\n",
894 (IV)prog->check_offset_min,
895 (IV)prog->check_offset_max,
903 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
905 || ((t = (char*)reghopmaybe3((U8*)s, -prog->check_offset_max, (U8*) ((prog->check_offset_max<0) ? strend : strpos)))
908 /* Fixed substring is found far enough so that the match
909 cannot start at strpos. */
911 if (ml_anch && t[-1] != '\n') {
912 /* Eventually fbm_*() should handle this, but often
913 anchored_offset is not 0, so this check will not be wasted. */
914 /* XXXX In the code below we prefer to look for "^" even in
915 presence of anchored substrings. And we search even
916 beyond the found float position. These pessimizations
917 are historical artefacts only. */
919 while (t < strend - prog->minlen) {
921 if (t < check_at - prog->check_offset_min) {
922 if (utf8_target ? prog->anchored_utf8 : prog->anchored_substr) {
923 /* Since we moved from the found position,
924 we definitely contradict the found anchored
925 substr. Due to the above check we do not
926 contradict "check" substr.
927 Thus we can arrive here only if check substr
928 is float. Redo checking for "other"=="fixed".
931 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld, rescanning for anchored from offset %ld...\n",
932 PL_colors[0], PL_colors[1], (long)(strpos - i_strpos), (long)(strpos - i_strpos + prog->anchored_offset)));
933 goto do_other_anchored;
935 /* We don't contradict the found floating substring. */
936 /* XXXX Why not check for STCLASS? */
938 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld...\n",
939 PL_colors[0], PL_colors[1], (long)(s - i_strpos)));
942 /* Position contradicts check-string */
943 /* XXXX probably better to look for check-string
944 than for "\n", so one should lower the limit for t? */
945 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m, restarting lookup for check-string at offset %ld...\n",
946 PL_colors[0], PL_colors[1], (long)(t + 1 - i_strpos)));
947 other_last = strpos = s = t + 1;
952 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Did not find /%s^%s/m...\n",
953 PL_colors[0], PL_colors[1]));
957 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Starting position does not contradict /%s^%s/m...\n",
958 PL_colors[0], PL_colors[1]));
962 ++BmUSEFUL(utf8_target ? prog->check_utf8 : prog->check_substr); /* hooray/5 */
965 /* The found string does not prohibit matching at strpos,
966 - no optimization of calling REx engine can be performed,
967 unless it was an MBOL and we are not after MBOL,
968 or a future STCLASS check will fail this. */
970 /* Even in this situation we may use MBOL flag if strpos is offset
971 wrt the start of the string. */
972 if (ml_anch && sv && !SvROK(sv) /* See prev comment on SvROK */
973 && (strpos != strbeg) && strpos[-1] != '\n'
974 /* May be due to an implicit anchor of m{.*foo} */
975 && !(prog->intflags & PREGf_IMPLICIT))
980 DEBUG_EXECUTE_r( if (ml_anch)
981 PerlIO_printf(Perl_debug_log, "Position at offset %ld does not contradict /%s^%s/m...\n",
982 (long)(strpos - i_strpos), PL_colors[0], PL_colors[1]);
985 if (!(prog->intflags & PREGf_NAUGHTY) /* XXXX If strpos moved? */
987 prog->check_utf8 /* Could be deleted already */
988 && --BmUSEFUL(prog->check_utf8) < 0
989 && (prog->check_utf8 == prog->float_utf8)
991 prog->check_substr /* Could be deleted already */
992 && --BmUSEFUL(prog->check_substr) < 0
993 && (prog->check_substr == prog->float_substr)
996 /* If flags & SOMETHING - do not do it many times on the same match */
997 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "... Disabling check substring...\n"));
998 /* XXX Does the destruction order has to change with utf8_target? */
999 SvREFCNT_dec(utf8_target ? prog->check_utf8 : prog->check_substr);
1000 SvREFCNT_dec(utf8_target ? prog->check_substr : prog->check_utf8);
1001 prog->check_substr = prog->check_utf8 = NULL; /* disable */
1002 prog->float_substr = prog->float_utf8 = NULL; /* clear */
1003 check = NULL; /* abort */
1005 /* XXXX If the check string was an implicit check MBOL, then we need to unset the relevent flag
1006 see http://bugs.activestate.com/show_bug.cgi?id=87173 */
1007 if (prog->intflags & PREGf_IMPLICIT)
1008 prog->extflags &= ~RXf_ANCH_MBOL;
1009 /* XXXX This is a remnant of the old implementation. It
1010 looks wasteful, since now INTUIT can use many
1011 other heuristics. */
1012 prog->extflags &= ~RXf_USE_INTUIT;
1013 /* XXXX What other flags might need to be cleared in this branch? */
1019 /* Last resort... */
1020 /* XXXX BmUSEFUL already changed, maybe multiple change is meaningful... */
1021 /* trie stclasses are too expensive to use here, we are better off to
1022 leave it to regmatch itself */
1023 if (progi->regstclass && PL_regkind[OP(progi->regstclass)]!=TRIE) {
1024 /* minlen == 0 is possible if regstclass is \b or \B,
1025 and the fixed substr is ''$.
1026 Since minlen is already taken into account, s+1 is before strend;
1027 accidentally, minlen >= 1 guaranties no false positives at s + 1
1028 even for \b or \B. But (minlen? 1 : 0) below assumes that
1029 regstclass does not come from lookahead... */
1030 /* If regstclass takes bytelength more than 1: If charlength==1, OK.
1031 This leaves EXACTF only, which is dealt with in find_byclass(). */
1032 const U8* const str = (U8*)STRING(progi->regstclass);
1033 const int cl_l = (PL_regkind[OP(progi->regstclass)] == EXACT
1034 ? CHR_DIST(str+STR_LEN(progi->regstclass), str)
1037 if (prog->anchored_substr || prog->anchored_utf8 || ml_anch)
1038 endpos= HOP3c(s, (prog->minlen ? cl_l : 0), strend);
1039 else if (prog->float_substr || prog->float_utf8)
1040 endpos= HOP3c(HOP3c(check_at, -start_shift, strbeg), cl_l, strend);
1044 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "start_shift: %"IVdf" check_at: %"IVdf" s: %"IVdf" endpos: %"IVdf"\n",
1045 (IV)start_shift, (IV)(check_at - strbeg), (IV)(s - strbeg), (IV)(endpos - strbeg)));
1048 s = find_byclass(prog, progi->regstclass, s, endpos, NULL);
1051 const char *what = NULL;
1053 if (endpos == strend) {
1054 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
1055 "Could not match STCLASS...\n") );
1058 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
1059 "This position contradicts STCLASS...\n") );
1060 if ((prog->extflags & RXf_ANCH) && !ml_anch)
1062 /* Contradict one of substrings */
1063 if (prog->anchored_substr || prog->anchored_utf8) {
1064 if ((utf8_target ? prog->anchored_utf8 : prog->anchored_substr) == check) {
1065 DEBUG_EXECUTE_r( what = "anchored" );
1067 s = HOP3c(t, 1, strend);
1068 if (s + start_shift + end_shift > strend) {
1069 /* XXXX Should be taken into account earlier? */
1070 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
1071 "Could not match STCLASS...\n") );
1076 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
1077 "Looking for %s substr starting at offset %ld...\n",
1078 what, (long)(s + start_shift - i_strpos)) );
1081 /* Have both, check_string is floating */
1082 if (t + start_shift >= check_at) /* Contradicts floating=check */
1083 goto retry_floating_check;
1084 /* Recheck anchored substring, but not floating... */
1088 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
1089 "Looking for anchored substr starting at offset %ld...\n",
1090 (long)(other_last - i_strpos)) );
1091 goto do_other_anchored;
1093 /* Another way we could have checked stclass at the
1094 current position only: */
1099 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
1100 "Looking for /%s^%s/m starting at offset %ld...\n",
1101 PL_colors[0], PL_colors[1], (long)(t - i_strpos)) );
1104 if (!(utf8_target ? prog->float_utf8 : prog->float_substr)) /* Could have been deleted */
1106 /* Check is floating subtring. */
1107 retry_floating_check:
1108 t = check_at - start_shift;
1109 DEBUG_EXECUTE_r( what = "floating" );
1110 goto hop_and_restart;
1113 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
1114 "By STCLASS: moving %ld --> %ld\n",
1115 (long)(t - i_strpos), (long)(s - i_strpos))
1119 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
1120 "Does not contradict STCLASS...\n");
1125 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%s%s:%s match at offset %ld\n",
1126 PL_colors[4], (check ? "Guessed" : "Giving up"),
1127 PL_colors[5], (long)(s - i_strpos)) );
1130 fail_finish: /* Substring not found */
1131 if (prog->check_substr || prog->check_utf8) /* could be removed already */
1132 BmUSEFUL(utf8_target ? prog->check_utf8 : prog->check_substr) += 5; /* hooray */
1134 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch rejected by optimizer%s\n",
1135 PL_colors[4], PL_colors[5]));
1139 #define DECL_TRIE_TYPE(scan) \
1140 const enum { trie_plain, trie_utf8, trie_utf8_fold, trie_latin_utf8_fold } \
1141 trie_type = (scan->flags != EXACT) \
1142 ? (utf8_target ? trie_utf8_fold : (UTF_PATTERN ? trie_latin_utf8_fold : trie_plain)) \
1143 : (utf8_target ? trie_utf8 : trie_plain)
1145 #define REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc, uscan, len, \
1146 uvc, charid, foldlen, foldbuf, uniflags) STMT_START { \
1147 switch (trie_type) { \
1148 case trie_utf8_fold: \
1149 if ( foldlen>0 ) { \
1150 uvc = utf8n_to_uvuni( uscan, UTF8_MAXLEN, &len, uniflags ); \
1155 uvc = utf8n_to_uvuni( (U8*)uc, UTF8_MAXLEN, &len, uniflags ); \
1156 uvc = to_uni_fold( uvc, foldbuf, &foldlen ); \
1157 foldlen -= UNISKIP( uvc ); \
1158 uscan = foldbuf + UNISKIP( uvc ); \
1161 case trie_latin_utf8_fold: \
1162 if ( foldlen>0 ) { \
1163 uvc = utf8n_to_uvuni( uscan, UTF8_MAXLEN, &len, uniflags ); \
1169 uvc = to_uni_fold( *(U8*)uc, foldbuf, &foldlen ); \
1170 foldlen -= UNISKIP( uvc ); \
1171 uscan = foldbuf + UNISKIP( uvc ); \
1175 uvc = utf8n_to_uvuni( (U8*)uc, UTF8_MAXLEN, &len, uniflags ); \
1182 charid = trie->charmap[ uvc ]; \
1186 if (widecharmap) { \
1187 SV** const svpp = hv_fetch(widecharmap, \
1188 (char*)&uvc, sizeof(UV), 0); \
1190 charid = (U16)SvIV(*svpp); \
1195 #define REXEC_FBC_EXACTISH_CHECK(CoNd) \
1197 char *my_strend= (char *)strend; \
1200 foldEQ_utf8(s, &my_strend, 0, utf8_target, \
1201 m, NULL, ln, cBOOL(UTF_PATTERN))) \
1202 && (!reginfo || regtry(reginfo, &s)) ) \
1205 U8 foldbuf[UTF8_MAXBYTES_CASE+1]; \
1206 uvchr_to_utf8(tmpbuf, c); \
1207 f = to_utf8_fold(tmpbuf, foldbuf, &foldlen); \
1209 && (f == c1 || f == c2) \
1211 foldEQ_utf8(s, &my_strend, 0, utf8_target,\
1212 m, NULL, ln, cBOOL(UTF_PATTERN)))\
1213 && (!reginfo || regtry(reginfo, &s)) ) \
1219 #define REXEC_FBC_EXACTISH_SCAN(CoNd) \
1223 && (ln == 1 || (OP(c) == EXACTF \
1224 ? foldEQ(s, m, ln) \
1225 : foldEQ_locale(s, m, ln))) \
1226 && (!reginfo || regtry(reginfo, &s)) ) \
1232 #define REXEC_FBC_UTF8_SCAN(CoDe) \
1234 while (s + (uskip = UTF8SKIP(s)) <= strend) { \
1240 #define REXEC_FBC_SCAN(CoDe) \
1242 while (s < strend) { \
1248 #define REXEC_FBC_UTF8_CLASS_SCAN(CoNd) \
1249 REXEC_FBC_UTF8_SCAN( \
1251 if (tmp && (!reginfo || regtry(reginfo, &s))) \
1260 #define REXEC_FBC_CLASS_SCAN(CoNd) \
1263 if (tmp && (!reginfo || regtry(reginfo, &s))) \
1272 #define REXEC_FBC_TRYIT \
1273 if ((!reginfo || regtry(reginfo, &s))) \
1276 #define REXEC_FBC_CSCAN(CoNdUtF8,CoNd) \
1277 if (utf8_target) { \
1278 REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \
1281 REXEC_FBC_CLASS_SCAN(CoNd); \
1285 #define REXEC_FBC_CSCAN_PRELOAD(UtFpReLoAd,CoNdUtF8,CoNd) \
1286 if (utf8_target) { \
1288 REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \
1291 REXEC_FBC_CLASS_SCAN(CoNd); \
1295 #define REXEC_FBC_CSCAN_TAINT(CoNdUtF8,CoNd) \
1296 PL_reg_flags |= RF_tainted; \
1297 if (utf8_target) { \
1298 REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \
1301 REXEC_FBC_CLASS_SCAN(CoNd); \
1305 #define DUMP_EXEC_POS(li,s,doutf8) \
1306 dump_exec_pos(li,s,(PL_regeol),(PL_bostr),(PL_reg_starttry),doutf8)
1308 /* We know what class REx starts with. Try to find this position... */
1309 /* if reginfo is NULL, its a dryrun */
1310 /* annoyingly all the vars in this routine have different names from their counterparts
1311 in regmatch. /grrr */
1314 S_find_byclass(pTHX_ regexp * prog, const regnode *c, char *s,
1315 const char *strend, regmatch_info *reginfo)
1318 const I32 doevery = (prog->intflags & PREGf_SKIP) == 0;
1322 register STRLEN uskip;
1326 register I32 tmp = 1; /* Scratch variable? */
1327 register const bool utf8_target = PL_reg_match_utf8;
1328 RXi_GET_DECL(prog,progi);
1330 PERL_ARGS_ASSERT_FIND_BYCLASS;
1332 /* We know what class it must start with. */
1336 REXEC_FBC_UTF8_CLASS_SCAN((ANYOF_FLAGS(c) & ANYOF_UNICODE) ||
1337 !UTF8_IS_INVARIANT((U8)s[0]) ?
1338 reginclass(prog, c, (U8*)s, 0, utf8_target) :
1339 REGINCLASS(prog, c, (U8*)s));
1342 while (s < strend) {
1345 if (REGINCLASS(prog, c, (U8*)s) ||
1346 (ANYOF_FOLD_SHARP_S(c, s, strend) &&
1347 /* The assignment of 2 is intentional:
1348 * for the folded sharp s, the skip is 2. */
1349 (skip = SHARP_S_SKIP))) {
1350 if (tmp && (!reginfo || regtry(reginfo, &s)))
1363 if (tmp && (!reginfo || regtry(reginfo, &s)))
1371 ln = STR_LEN(c); /* length to match in octets/bytes */
1372 lnc = (I32) ln; /* length to match in characters */
1374 STRLEN ulen1, ulen2;
1376 U8 tmpbuf1[UTF8_MAXBYTES_CASE+1];
1377 U8 tmpbuf2[UTF8_MAXBYTES_CASE+1];
1378 /* used by commented-out code below */
1379 /*const U32 uniflags = UTF8_ALLOW_DEFAULT;*/
1381 /* XXX: Since the node will be case folded at compile
1382 time this logic is a little odd, although im not
1383 sure that its actually wrong. --dmq */
1385 c1 = to_utf8_lower((U8*)m, tmpbuf1, &ulen1);
1386 c2 = to_utf8_upper((U8*)m, tmpbuf2, &ulen2);
1388 /* XXX: This is kinda strange. to_utf8_XYZ returns the
1389 codepoint of the first character in the converted
1390 form, yet originally we did the extra step.
1391 No tests fail by commenting this code out however
1392 so Ive left it out. -- dmq.
1394 c1 = utf8n_to_uvchr(tmpbuf1, UTF8_MAXBYTES_CASE,
1396 c2 = utf8n_to_uvchr(tmpbuf2, UTF8_MAXBYTES_CASE,
1401 while (sm < ((U8 *) m + ln)) {
1416 c2 = PL_fold_locale[c1];
1418 e = HOP3c(strend, -((I32)lnc), s);
1420 if (!reginfo && e < s)
1421 e = s; /* Due to minlen logic of intuit() */
1423 /* The idea in the EXACTF* cases is to first find the
1424 * first character of the EXACTF* node and then, if
1425 * necessary, case-insensitively compare the full
1426 * text of the node. The c1 and c2 are the first
1427 * characters (though in Unicode it gets a bit
1428 * more complicated because there are more cases
1429 * than just upper and lower: one needs to use
1430 * the so-called folding case for case-insensitive
1431 * matching (called "loose matching" in Unicode).
1432 * foldEQ_utf8() will do just that. */
1434 if (utf8_target || UTF_PATTERN) {
1436 U8 tmpbuf [UTF8_MAXBYTES+1];
1439 const U32 uniflags = UTF8_ALLOW_DEFAULT;
1441 /* Upper and lower of 1st char are equal -
1442 * probably not a "letter". */
1445 c = utf8n_to_uvchr((U8*)s, UTF8_MAXBYTES, &len,
1450 REXEC_FBC_EXACTISH_CHECK(c == c1);
1456 c = utf8n_to_uvchr((U8*)s, UTF8_MAXBYTES, &len,
1462 /* Handle some of the three Greek sigmas cases.
1463 * Note that not all the possible combinations
1464 * are handled here: some of them are handled
1465 * by the standard folding rules, and some of
1466 * them (the character class or ANYOF cases)
1467 * are handled during compiletime in
1468 * regexec.c:S_regclass(). */
1469 if (c == (UV)UNICODE_GREEK_CAPITAL_LETTER_SIGMA ||
1470 c == (UV)UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA)
1471 c = (UV)UNICODE_GREEK_SMALL_LETTER_SIGMA;
1473 REXEC_FBC_EXACTISH_CHECK(c == c1 || c == c2);
1478 /* Neither pattern nor string are UTF8 */
1480 REXEC_FBC_EXACTISH_SCAN(*(U8*)s == c1);
1482 REXEC_FBC_EXACTISH_SCAN(*(U8*)s == c1 || *(U8*)s == c2);
1486 PL_reg_flags |= RF_tainted;
1493 U8 * const r = reghop3((U8*)s, -1, (U8*)PL_bostr);
1494 tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, UTF8_ALLOW_DEFAULT);
1496 tmp = ((OP(c) == BOUND ?
1497 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1498 LOAD_UTF8_CHARCLASS_ALNUM();
1499 REXEC_FBC_UTF8_SCAN(
1500 if (tmp == !(OP(c) == BOUND ?
1501 cBOOL(swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target)) :
1502 isALNUM_LC_utf8((U8*)s)))
1510 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1511 tmp = ((OP(c) == BOUND ? isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1514 !(OP(c) == BOUND ? isALNUM(*s) : isALNUM_LC(*s))) {
1520 if ((!prog->minlen && tmp) && (!reginfo || regtry(reginfo, &s)))
1524 PL_reg_flags |= RF_tainted;
1531 U8 * const r = reghop3((U8*)s, -1, (U8*)PL_bostr);
1532 tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, UTF8_ALLOW_DEFAULT);
1534 tmp = ((OP(c) == NBOUND ?
1535 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1536 LOAD_UTF8_CHARCLASS_ALNUM();
1537 REXEC_FBC_UTF8_SCAN(
1538 if (tmp == !(OP(c) == NBOUND ?
1539 cBOOL(swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target)) :
1540 isALNUM_LC_utf8((U8*)s)))
1542 else REXEC_FBC_TRYIT;
1546 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1547 tmp = ((OP(c) == NBOUND ?
1548 isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1551 !(OP(c) == NBOUND ? isALNUM(*s) : isALNUM_LC(*s)))
1553 else REXEC_FBC_TRYIT;
1556 if ((!prog->minlen && !tmp) && (!reginfo || regtry(reginfo, &s)))
1560 REXEC_FBC_CSCAN_PRELOAD(
1561 LOAD_UTF8_CHARCLASS_PERL_WORD(),
1562 swash_fetch(RE_utf8_perl_word, (U8*)s, utf8_target),
1566 REXEC_FBC_CSCAN_TAINT(
1567 isALNUM_LC_utf8((U8*)s),
1571 REXEC_FBC_CSCAN_PRELOAD(
1572 LOAD_UTF8_CHARCLASS_PERL_WORD(),
1573 !swash_fetch(RE_utf8_perl_word, (U8*)s, utf8_target),
1577 REXEC_FBC_CSCAN_TAINT(
1578 !isALNUM_LC_utf8((U8*)s),
1582 REXEC_FBC_CSCAN_PRELOAD(
1583 LOAD_UTF8_CHARCLASS_PERL_SPACE(),
1584 *s == ' ' || swash_fetch(RE_utf8_perl_space,(U8*)s, utf8_target),
1588 REXEC_FBC_CSCAN_TAINT(
1589 *s == ' ' || isSPACE_LC_utf8((U8*)s),
1593 REXEC_FBC_CSCAN_PRELOAD(
1594 LOAD_UTF8_CHARCLASS_PERL_SPACE(),
1595 !(*s == ' ' || swash_fetch(RE_utf8_perl_space,(U8*)s, utf8_target)),
1599 REXEC_FBC_CSCAN_TAINT(
1600 !(*s == ' ' || isSPACE_LC_utf8((U8*)s)),
1604 REXEC_FBC_CSCAN_PRELOAD(
1605 LOAD_UTF8_CHARCLASS_POSIX_DIGIT(),
1606 swash_fetch(RE_utf8_posix_digit,(U8*)s, utf8_target),
1610 REXEC_FBC_CSCAN_TAINT(
1611 isDIGIT_LC_utf8((U8*)s),
1615 REXEC_FBC_CSCAN_PRELOAD(
1616 LOAD_UTF8_CHARCLASS_POSIX_DIGIT(),
1617 !swash_fetch(RE_utf8_posix_digit,(U8*)s, utf8_target),
1621 REXEC_FBC_CSCAN_TAINT(
1622 !isDIGIT_LC_utf8((U8*)s),
1628 is_LNBREAK_latin1(s)
1638 !is_VERTWS_latin1(s)
1643 is_HORIZWS_latin1(s)
1647 !is_HORIZWS_utf8(s),
1648 !is_HORIZWS_latin1(s)
1654 /* what trie are we using right now */
1656 = (reg_ac_data*)progi->data->data[ ARG( c ) ];
1658 = (reg_trie_data*)progi->data->data[ aho->trie ];
1659 HV *widecharmap = MUTABLE_HV(progi->data->data[ aho->trie + 1 ]);
1661 const char *last_start = strend - trie->minlen;
1663 const char *real_start = s;
1665 STRLEN maxlen = trie->maxlen;
1667 U8 **points; /* map of where we were in the input string
1668 when reading a given char. For ASCII this
1669 is unnecessary overhead as the relationship
1670 is always 1:1, but for Unicode, especially
1671 case folded Unicode this is not true. */
1672 U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ];
1676 GET_RE_DEBUG_FLAGS_DECL;
1678 /* We can't just allocate points here. We need to wrap it in
1679 * an SV so it gets freed properly if there is a croak while
1680 * running the match */
1683 sv_points=newSV(maxlen * sizeof(U8 *));
1684 SvCUR_set(sv_points,
1685 maxlen * sizeof(U8 *));
1686 SvPOK_on(sv_points);
1687 sv_2mortal(sv_points);
1688 points=(U8**)SvPV_nolen(sv_points );
1689 if ( trie_type != trie_utf8_fold
1690 && (trie->bitmap || OP(c)==AHOCORASICKC) )
1693 bitmap=(U8*)trie->bitmap;
1695 bitmap=(U8*)ANYOF_BITMAP(c);
1697 /* this is the Aho-Corasick algorithm modified a touch
1698 to include special handling for long "unknown char"
1699 sequences. The basic idea being that we use AC as long
1700 as we are dealing with a possible matching char, when
1701 we encounter an unknown char (and we have not encountered
1702 an accepting state) we scan forward until we find a legal
1704 AC matching is basically that of trie matching, except
1705 that when we encounter a failing transition, we fall back
1706 to the current states "fail state", and try the current char
1707 again, a process we repeat until we reach the root state,
1708 state 1, or a legal transition. If we fail on the root state
1709 then we can either terminate if we have reached an accepting
1710 state previously, or restart the entire process from the beginning
1714 while (s <= last_start) {
1715 const U32 uniflags = UTF8_ALLOW_DEFAULT;
1723 U8 *uscan = (U8*)NULL;
1724 U8 *leftmost = NULL;
1726 U32 accepted_word= 0;
1730 while ( state && uc <= (U8*)strend ) {
1732 U32 word = aho->states[ state ].wordnum;
1736 DEBUG_TRIE_EXECUTE_r(
1737 if ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) {
1738 dump_exec_pos( (char *)uc, c, strend, real_start,
1739 (char *)uc, utf8_target );
1740 PerlIO_printf( Perl_debug_log,
1741 " Scanning for legal start char...\n");
1744 while ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) {
1749 if (uc >(U8*)last_start) break;
1753 U8 *lpos= points[ (pointpos - trie->wordinfo[word].len) % maxlen ];
1754 if (!leftmost || lpos < leftmost) {
1755 DEBUG_r(accepted_word=word);
1761 points[pointpos++ % maxlen]= uc;
1762 REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc,
1763 uscan, len, uvc, charid, foldlen,
1765 DEBUG_TRIE_EXECUTE_r({
1766 dump_exec_pos( (char *)uc, c, strend, real_start,
1768 PerlIO_printf(Perl_debug_log,
1769 " Charid:%3u CP:%4"UVxf" ",
1775 word = aho->states[ state ].wordnum;
1777 base = aho->states[ state ].trans.base;
1779 DEBUG_TRIE_EXECUTE_r({
1781 dump_exec_pos( (char *)uc, c, strend, real_start,
1783 PerlIO_printf( Perl_debug_log,
1784 "%sState: %4"UVxf", word=%"UVxf,
1785 failed ? " Fail transition to " : "",
1786 (UV)state, (UV)word);
1792 ( ((offset = base + charid
1793 - 1 - trie->uniquecharcount)) >= 0)
1794 && ((U32)offset < trie->lasttrans)
1795 && trie->trans[offset].check == state
1796 && (tmp=trie->trans[offset].next))
1798 DEBUG_TRIE_EXECUTE_r(
1799 PerlIO_printf( Perl_debug_log," - legal\n"));
1804 DEBUG_TRIE_EXECUTE_r(
1805 PerlIO_printf( Perl_debug_log," - fail\n"));
1807 state = aho->fail[state];
1811 /* we must be accepting here */
1812 DEBUG_TRIE_EXECUTE_r(
1813 PerlIO_printf( Perl_debug_log," - accepting\n"));
1822 if (!state) state = 1;
1825 if ( aho->states[ state ].wordnum ) {
1826 U8 *lpos = points[ (pointpos - trie->wordinfo[aho->states[ state ].wordnum].len) % maxlen ];
1827 if (!leftmost || lpos < leftmost) {
1828 DEBUG_r(accepted_word=aho->states[ state ].wordnum);
1833 s = (char*)leftmost;
1834 DEBUG_TRIE_EXECUTE_r({
1836 Perl_debug_log,"Matches word #%"UVxf" at position %"IVdf". Trying full pattern...\n",
1837 (UV)accepted_word, (IV)(s - real_start)
1840 if (!reginfo || regtry(reginfo, &s)) {
1846 DEBUG_TRIE_EXECUTE_r({
1847 PerlIO_printf( Perl_debug_log,"Pattern failed. Looking for new start point...\n");
1850 DEBUG_TRIE_EXECUTE_r(
1851 PerlIO_printf( Perl_debug_log,"No match.\n"));
1860 Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c));
1870 - regexec_flags - match a regexp against a string
1873 Perl_regexec_flags(pTHX_ REGEXP * const rx, char *stringarg, register char *strend,
1874 char *strbeg, I32 minend, SV *sv, void *data, U32 flags)
1875 /* strend: pointer to null at end of string */
1876 /* strbeg: real beginning of string */
1877 /* minend: end of match must be >=minend after stringarg. */
1878 /* data: May be used for some additional optimizations.
1879 Currently its only used, with a U32 cast, for transmitting
1880 the ganch offset when doing a /g match. This will change */
1881 /* nosave: For optimizations. */
1884 struct regexp *const prog = (struct regexp *)SvANY(rx);
1885 /*register*/ char *s;
1886 register regnode *c;
1887 /*register*/ char *startpos = stringarg;
1888 I32 minlen; /* must match at least this many chars */
1889 I32 dontbother = 0; /* how many characters not to try at end */
1890 I32 end_shift = 0; /* Same for the end. */ /* CC */
1891 I32 scream_pos = -1; /* Internal iterator of scream. */
1892 char *scream_olds = NULL;
1893 const bool utf8_target = cBOOL(DO_UTF8(sv));
1895 RXi_GET_DECL(prog,progi);
1896 regmatch_info reginfo; /* create some info to pass to regtry etc */
1897 regexp_paren_pair *swap = NULL;
1898 GET_RE_DEBUG_FLAGS_DECL;
1900 PERL_ARGS_ASSERT_REGEXEC_FLAGS;
1901 PERL_UNUSED_ARG(data);
1903 /* Be paranoid... */
1904 if (prog == NULL || startpos == NULL) {
1905 Perl_croak(aTHX_ "NULL regexp parameter");
1909 multiline = prog->extflags & RXf_PMf_MULTILINE;
1910 reginfo.prog = rx; /* Yes, sorry that this is confusing. */
1912 RX_MATCH_UTF8_set(rx, utf8_target);
1914 debug_start_match(rx, utf8_target, startpos, strend,
1918 minlen = prog->minlen;
1920 if (strend - startpos < (minlen+(prog->check_offset_min<0?prog->check_offset_min:0))) {
1921 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
1922 "String too short [regexec_flags]...\n"));
1927 /* Check validity of program. */
1928 if (UCHARAT(progi->program) != REG_MAGIC) {
1929 Perl_croak(aTHX_ "corrupted regexp program");
1933 PL_reg_eval_set = 0;
1937 PL_reg_flags |= RF_utf8;
1939 /* Mark beginning of line for ^ and lookbehind. */
1940 reginfo.bol = startpos; /* XXX not used ??? */
1944 /* Mark end of line for $ (and such) */
1947 /* see how far we have to get to not match where we matched before */
1948 reginfo.till = startpos+minend;
1950 /* If there is a "must appear" string, look for it. */
1953 if (prog->extflags & RXf_GPOS_SEEN) { /* Need to set reginfo->ganch */
1955 if (flags & REXEC_IGNOREPOS){ /* Means: check only at start */
1956 reginfo.ganch = startpos + prog->gofs;
1957 DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log,
1958 "GPOS IGNOREPOS: reginfo.ganch = startpos + %"UVxf"\n",(UV)prog->gofs));
1959 } else if (sv && SvTYPE(sv) >= SVt_PVMG
1961 && (mg = mg_find(sv, PERL_MAGIC_regex_global))
1962 && mg->mg_len >= 0) {
1963 reginfo.ganch = strbeg + mg->mg_len; /* Defined pos() */
1964 DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log,
1965 "GPOS MAGIC: reginfo.ganch = strbeg + %"IVdf"\n",(IV)mg->mg_len));
1967 if (prog->extflags & RXf_ANCH_GPOS) {
1968 if (s > reginfo.ganch)
1970 s = reginfo.ganch - prog->gofs;
1971 DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log,
1972 "GPOS ANCH_GPOS: s = ganch - %"UVxf"\n",(UV)prog->gofs));
1978 reginfo.ganch = strbeg + PTR2UV(data);
1979 DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log,
1980 "GPOS DATA: reginfo.ganch= strbeg + %"UVxf"\n",PTR2UV(data)));
1982 } else { /* pos() not defined */
1983 reginfo.ganch = strbeg;
1984 DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log,
1985 "GPOS: reginfo.ganch = strbeg\n"));
1988 if (PL_curpm && (PM_GETRE(PL_curpm) == rx)) {
1989 /* We have to be careful. If the previous successful match
1990 was from this regex we don't want a subsequent partially
1991 successful match to clobber the old results.
1992 So when we detect this possibility we add a swap buffer
1993 to the re, and switch the buffer each match. If we fail
1994 we switch it back, otherwise we leave it swapped.
1997 /* do we need a save destructor here for eval dies? */
1998 Newxz(prog->offs, (prog->nparens + 1), regexp_paren_pair);
2000 if (!(flags & REXEC_CHECKED) && (prog->check_substr != NULL || prog->check_utf8 != NULL)) {
2001 re_scream_pos_data d;
2003 d.scream_olds = &scream_olds;
2004 d.scream_pos = &scream_pos;
2005 s = re_intuit_start(rx, sv, s, strend, flags, &d);
2007 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Not present...\n"));
2008 goto phooey; /* not present */
2014 /* Simplest case: anchored match need be tried only once. */
2015 /* [unless only anchor is BOL and multiline is set] */
2016 if (prog->extflags & (RXf_ANCH & ~RXf_ANCH_GPOS)) {
2017 if (s == startpos && regtry(®info, &startpos))
2019 else if (multiline || (prog->intflags & PREGf_IMPLICIT)
2020 || (prog->extflags & RXf_ANCH_MBOL)) /* XXXX SBOL? */
2025 dontbother = minlen - 1;
2026 end = HOP3c(strend, -dontbother, strbeg) - 1;
2027 /* for multiline we only have to try after newlines */
2028 if (prog->check_substr || prog->check_utf8) {
2029 /* because of the goto we can not easily reuse the macros for bifurcating the
2030 unicode/non-unicode match modes here like we do elsewhere - demerphq */
2033 goto after_try_utf8;
2035 if (regtry(®info, &s)) {
2042 if (prog->extflags & RXf_USE_INTUIT) {
2043 s = re_intuit_start(rx, sv, s + UTF8SKIP(s), strend, flags, NULL);
2052 } /* end search for check string in unicode */
2054 if (s == startpos) {
2055 goto after_try_latin;
2058 if (regtry(®info, &s)) {
2065 if (prog->extflags & RXf_USE_INTUIT) {
2066 s = re_intuit_start(rx, sv, s + 1, strend, flags, NULL);
2075 } /* end search for check string in latin*/
2076 } /* end search for check string */
2077 else { /* search for newline */
2079 /*XXX: The s-- is almost definitely wrong here under unicode - demeprhq*/
2082 /* We can use a more efficient search as newlines are the same in unicode as they are in latin */
2084 if (*s++ == '\n') { /* don't need PL_utf8skip here */
2085 if (regtry(®info, &s))
2089 } /* end search for newline */
2090 } /* end anchored/multiline check string search */
2092 } else if (RXf_GPOS_CHECK == (prog->extflags & RXf_GPOS_CHECK))
2094 /* the warning about reginfo.ganch being used without intialization
2095 is bogus -- we set it above, when prog->extflags & RXf_GPOS_SEEN
2096 and we only enter this block when the same bit is set. */
2097 char *tmp_s = reginfo.ganch - prog->gofs;
2099 if (tmp_s >= strbeg && regtry(®info, &tmp_s))
2104 /* Messy cases: unanchored match. */
2105 if ((prog->anchored_substr || prog->anchored_utf8) && prog->intflags & PREGf_SKIP) {
2106 /* we have /x+whatever/ */
2107 /* it must be a one character string (XXXX Except UTF_PATTERN?) */
2112 if (!(utf8_target ? prog->anchored_utf8 : prog->anchored_substr))
2113 utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog);
2114 ch = SvPVX_const(utf8_target ? prog->anchored_utf8 : prog->anchored_substr)[0];
2119 DEBUG_EXECUTE_r( did_match = 1 );
2120 if (regtry(®info, &s)) goto got_it;
2122 while (s < strend && *s == ch)
2130 DEBUG_EXECUTE_r( did_match = 1 );
2131 if (regtry(®info, &s)) goto got_it;
2133 while (s < strend && *s == ch)
2138 DEBUG_EXECUTE_r(if (!did_match)
2139 PerlIO_printf(Perl_debug_log,
2140 "Did not find anchored character...\n")
2143 else if (prog->anchored_substr != NULL
2144 || prog->anchored_utf8 != NULL
2145 || ((prog->float_substr != NULL || prog->float_utf8 != NULL)
2146 && prog->float_max_offset < strend - s)) {
2151 char *last1; /* Last position checked before */
2155 if (prog->anchored_substr || prog->anchored_utf8) {
2156 if (!(utf8_target ? prog->anchored_utf8 : prog->anchored_substr))
2157 utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog);
2158 must = utf8_target ? prog->anchored_utf8 : prog->anchored_substr;
2159 back_max = back_min = prog->anchored_offset;
2161 if (!(utf8_target ? prog->float_utf8 : prog->float_substr))
2162 utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog);
2163 must = utf8_target ? prog->float_utf8 : prog->float_substr;
2164 back_max = prog->float_max_offset;
2165 back_min = prog->float_min_offset;
2169 if (must == &PL_sv_undef)
2170 /* could not downgrade utf8 check substring, so must fail */
2176 last = HOP3c(strend, /* Cannot start after this */
2177 -(I32)(CHR_SVLEN(must)
2178 - (SvTAIL(must) != 0) + back_min), strbeg);
2181 last1 = HOPc(s, -1);
2183 last1 = s - 1; /* bogus */
2185 /* XXXX check_substr already used to find "s", can optimize if
2186 check_substr==must. */
2188 dontbother = end_shift;
2189 strend = HOPc(strend, -dontbother);
2190 while ( (s <= last) &&
2191 ((flags & REXEC_SCREAM)
2192 ? (s = screaminstr(sv, must, HOP3c(s, back_min, (back_min<0 ? strbeg : strend)) - strbeg,
2193 end_shift, &scream_pos, 0))
2194 : (s = fbm_instr((unsigned char*)HOP3(s, back_min, (back_min<0 ? strbeg : strend)),
2195 (unsigned char*)strend, must,
2196 multiline ? FBMrf_MULTILINE : 0))) ) {
2197 /* we may be pointing at the wrong string */
2198 if ((flags & REXEC_SCREAM) && RXp_MATCH_COPIED(prog))
2199 s = strbeg + (s - SvPVX_const(sv));
2200 DEBUG_EXECUTE_r( did_match = 1 );
2201 if (HOPc(s, -back_max) > last1) {
2202 last1 = HOPc(s, -back_min);
2203 s = HOPc(s, -back_max);
2206 char * const t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1;
2208 last1 = HOPc(s, -back_min);
2212 while (s <= last1) {
2213 if (regtry(®info, &s))
2219 while (s <= last1) {
2220 if (regtry(®info, &s))
2226 DEBUG_EXECUTE_r(if (!did_match) {
2227 RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0),
2228 SvPVX_const(must), RE_SV_DUMPLEN(must), 30);
2229 PerlIO_printf(Perl_debug_log, "Did not find %s substr %s%s...\n",
2230 ((must == prog->anchored_substr || must == prog->anchored_utf8)
2231 ? "anchored" : "floating"),
2232 quoted, RE_SV_TAIL(must));
2236 else if ( (c = progi->regstclass) ) {
2238 const OPCODE op = OP(progi->regstclass);
2239 /* don't bother with what can't match */
2240 if (PL_regkind[op] != EXACT && op != CANY && PL_regkind[op] != TRIE)
2241 strend = HOPc(strend, -(minlen - 1));
2244 SV * const prop = sv_newmortal();
2245 regprop(prog, prop, c);
2247 RE_PV_QUOTED_DECL(quoted,utf8_target,PERL_DEBUG_PAD_ZERO(1),
2249 PerlIO_printf(Perl_debug_log,
2250 "Matching stclass %.*s against %s (%d bytes)\n",
2251 (int)SvCUR(prop), SvPVX_const(prop),
2252 quoted, (int)(strend - s));
2255 if (find_byclass(prog, c, s, strend, ®info))
2257 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass... [regexec_flags]\n"));
2261 if (prog->float_substr != NULL || prog->float_utf8 != NULL) {
2266 if (!(utf8_target ? prog->float_utf8 : prog->float_substr))
2267 utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog);
2268 float_real = utf8_target ? prog->float_utf8 : prog->float_substr;
2270 if (flags & REXEC_SCREAM) {
2271 last = screaminstr(sv, float_real, s - strbeg,
2272 end_shift, &scream_pos, 1); /* last one */
2274 last = scream_olds; /* Only one occurrence. */
2275 /* we may be pointing at the wrong string */
2276 else if (RXp_MATCH_COPIED(prog))
2277 s = strbeg + (s - SvPVX_const(sv));
2281 const char * const little = SvPV_const(float_real, len);
2283 if (SvTAIL(float_real)) {
2284 if (memEQ(strend - len + 1, little, len - 1))
2285 last = strend - len + 1;
2286 else if (!multiline)
2287 last = memEQ(strend - len, little, len)
2288 ? strend - len : NULL;
2294 last = rninstr(s, strend, little, little + len);
2296 last = strend; /* matching "$" */
2301 PerlIO_printf(Perl_debug_log,
2302 "%sCan't trim the tail, match fails (should not happen)%s\n",
2303 PL_colors[4], PL_colors[5]));
2304 goto phooey; /* Should not happen! */
2306 dontbother = strend - last + prog->float_min_offset;
2308 if (minlen && (dontbother < minlen))
2309 dontbother = minlen - 1;
2310 strend -= dontbother; /* this one's always in bytes! */
2311 /* We don't know much -- general case. */
2314 if (regtry(®info, &s))
2323 if (regtry(®info, &s))
2325 } while (s++ < strend);
2334 RX_MATCH_TAINTED_set(rx, PL_reg_flags & RF_tainted);
2336 if (PL_reg_eval_set)
2337 restore_pos(aTHX_ prog);
2338 if (RXp_PAREN_NAMES(prog))
2339 (void)hv_iterinit(RXp_PAREN_NAMES(prog));
2341 /* make sure $`, $&, $', and $digit will work later */
2342 if ( !(flags & REXEC_NOT_FIRST) ) {
2343 RX_MATCH_COPY_FREE(rx);
2344 if (flags & REXEC_COPY_STR) {
2345 const I32 i = PL_regeol - startpos + (stringarg - strbeg);
2346 #ifdef PERL_OLD_COPY_ON_WRITE
2348 || (SvFLAGS(sv) & CAN_COW_MASK) == CAN_COW_FLAGS)) {
2350 PerlIO_printf(Perl_debug_log,
2351 "Copy on write: regexp capture, type %d\n",
2354 prog->saved_copy = sv_setsv_cow(prog->saved_copy, sv);
2355 prog->subbeg = (char *)SvPVX_const(prog->saved_copy);
2356 assert (SvPOKp(prog->saved_copy));
2360 RX_MATCH_COPIED_on(rx);
2361 s = savepvn(strbeg, i);
2367 prog->subbeg = strbeg;
2368 prog->sublen = PL_regeol - strbeg; /* strend may have been modified */
2375 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n",
2376 PL_colors[4], PL_colors[5]));
2377 if (PL_reg_eval_set)
2378 restore_pos(aTHX_ prog);
2380 /* we failed :-( roll it back */
2381 Safefree(prog->offs);
2390 - regtry - try match at specific point
2392 STATIC I32 /* 0 failure, 1 success */
2393 S_regtry(pTHX_ regmatch_info *reginfo, char **startpos)
2397 REGEXP *const rx = reginfo->prog;
2398 regexp *const prog = (struct regexp *)SvANY(rx);
2399 RXi_GET_DECL(prog,progi);
2400 GET_RE_DEBUG_FLAGS_DECL;
2402 PERL_ARGS_ASSERT_REGTRY;
2404 reginfo->cutpoint=NULL;
2406 if ((prog->extflags & RXf_EVAL_SEEN) && !PL_reg_eval_set) {
2409 PL_reg_eval_set = RS_init;
2410 DEBUG_EXECUTE_r(DEBUG_s(
2411 PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %"IVdf"\n",
2412 (IV)(PL_stack_sp - PL_stack_base));
2415 cxstack[cxstack_ix].blk_oldsp = PL_stack_sp - PL_stack_base;
2416 /* Otherwise OP_NEXTSTATE will free whatever on stack now. */
2418 /* Apparently this is not needed, judging by wantarray. */
2419 /* SAVEI8(cxstack[cxstack_ix].blk_gimme);
2420 cxstack[cxstack_ix].blk_gimme = G_SCALAR; */
2423 /* Make $_ available to executed code. */
2424 if (reginfo->sv != DEFSV) {
2426 DEFSV_set(reginfo->sv);
2429 if (!(SvTYPE(reginfo->sv) >= SVt_PVMG && SvMAGIC(reginfo->sv)
2430 && (mg = mg_find(reginfo->sv, PERL_MAGIC_regex_global)))) {
2431 /* prepare for quick setting of pos */
2432 #ifdef PERL_OLD_COPY_ON_WRITE
2433 if (SvIsCOW(reginfo->sv))
2434 sv_force_normal_flags(reginfo->sv, 0);
2436 mg = sv_magicext(reginfo->sv, NULL, PERL_MAGIC_regex_global,
2437 &PL_vtbl_mglob, NULL, 0);
2441 PL_reg_oldpos = mg->mg_len;
2442 SAVEDESTRUCTOR_X(restore_pos, prog);
2444 if (!PL_reg_curpm) {
2445 Newxz(PL_reg_curpm, 1, PMOP);
2448 SV* const repointer = &PL_sv_undef;
2449 /* this regexp is also owned by the new PL_reg_curpm, which
2450 will try to free it. */
2451 av_push(PL_regex_padav, repointer);
2452 PL_reg_curpm->op_pmoffset = av_len(PL_regex_padav);
2453 PL_regex_pad = AvARRAY(PL_regex_padav);
2458 /* It seems that non-ithreads works both with and without this code.
2459 So for efficiency reasons it seems best not to have the code
2460 compiled when it is not needed. */
2461 /* This is safe against NULLs: */
2462 ReREFCNT_dec(PM_GETRE(PL_reg_curpm));
2463 /* PM_reg_curpm owns a reference to this regexp. */
2466 PM_SETRE(PL_reg_curpm, rx);
2467 PL_reg_oldcurpm = PL_curpm;
2468 PL_curpm = PL_reg_curpm;
2469 if (RXp_MATCH_COPIED(prog)) {
2470 /* Here is a serious problem: we cannot rewrite subbeg,
2471 since it may be needed if this match fails. Thus
2472 $` inside (?{}) could fail... */
2473 PL_reg_oldsaved = prog->subbeg;
2474 PL_reg_oldsavedlen = prog->sublen;
2475 #ifdef PERL_OLD_COPY_ON_WRITE
2476 PL_nrs = prog->saved_copy;
2478 RXp_MATCH_COPIED_off(prog);
2481 PL_reg_oldsaved = NULL;
2482 prog->subbeg = PL_bostr;
2483 prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */
2485 DEBUG_EXECUTE_r(PL_reg_starttry = *startpos);
2486 prog->offs[0].start = *startpos - PL_bostr;
2487 PL_reginput = *startpos;
2488 PL_reglastparen = &prog->lastparen;
2489 PL_reglastcloseparen = &prog->lastcloseparen;
2490 prog->lastparen = 0;
2491 prog->lastcloseparen = 0;
2493 PL_regoffs = prog->offs;
2494 if (PL_reg_start_tmpl <= prog->nparens) {
2495 PL_reg_start_tmpl = prog->nparens*3/2 + 3;
2496 if(PL_reg_start_tmp)
2497 Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
2499 Newx(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
2502 /* XXXX What this code is doing here?!!! There should be no need
2503 to do this again and again, PL_reglastparen should take care of
2506 /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code.
2507 * Actually, the code in regcppop() (which Ilya may be meaning by
2508 * PL_reglastparen), is not needed at all by the test suite
2509 * (op/regexp, op/pat, op/split), but that code is needed otherwise
2510 * this erroneously leaves $1 defined: "1" =~ /^(?:(\d)x)?\d$/
2511 * Meanwhile, this code *is* needed for the
2512 * above-mentioned test suite tests to succeed. The common theme
2513 * on those tests seems to be returning null fields from matches.
2514 * --jhi updated by dapm */
2516 if (prog->nparens) {
2517 regexp_paren_pair *pp = PL_regoffs;
2519 for (i = prog->nparens; i > (I32)*PL_reglastparen; i--) {
2527 if (regmatch(reginfo, progi->program + 1)) {
2528 PL_regoffs[0].end = PL_reginput - PL_bostr;
2531 if (reginfo->cutpoint)
2532 *startpos= reginfo->cutpoint;
2533 REGCP_UNWIND(lastcp);
2538 #define sayYES goto yes
2539 #define sayNO goto no
2540 #define sayNO_SILENT goto no_silent
2542 /* we dont use STMT_START/END here because it leads to
2543 "unreachable code" warnings, which are bogus, but distracting. */
2544 #define CACHEsayNO \
2545 if (ST.cache_mask) \
2546 PL_reg_poscache[ST.cache_offset] |= ST.cache_mask; \
2549 /* this is used to determine how far from the left messages like
2550 'failed...' are printed. It should be set such that messages
2551 are inline with the regop output that created them.
2553 #define REPORT_CODE_OFF 32
2556 #define CHRTEST_UNINIT -1001 /* c1/c2 haven't been calculated yet */
2557 #define CHRTEST_VOID -1000 /* the c1/c2 "next char" test should be skipped */
2559 #define SLAB_FIRST(s) (&(s)->states[0])
2560 #define SLAB_LAST(s) (&(s)->states[PERL_REGMATCH_SLAB_SLOTS-1])
2562 /* grab a new slab and return the first slot in it */
2564 STATIC regmatch_state *
2567 #if PERL_VERSION < 9 && !defined(PERL_CORE)
2570 regmatch_slab *s = PL_regmatch_slab->next;
2572 Newx(s, 1, regmatch_slab);
2573 s->prev = PL_regmatch_slab;
2575 PL_regmatch_slab->next = s;
2577 PL_regmatch_slab = s;
2578 return SLAB_FIRST(s);
2582 /* push a new state then goto it */
2584 #define PUSH_STATE_GOTO(state, node) \
2586 st->resume_state = state; \
2589 /* push a new state with success backtracking, then goto it */
2591 #define PUSH_YES_STATE_GOTO(state, node) \
2593 st->resume_state = state; \
2594 goto push_yes_state;
2600 regmatch() - main matching routine
2602 This is basically one big switch statement in a loop. We execute an op,
2603 set 'next' to point the next op, and continue. If we come to a point which
2604 we may need to backtrack to on failure such as (A|B|C), we push a
2605 backtrack state onto the backtrack stack. On failure, we pop the top
2606 state, and re-enter the loop at the state indicated. If there are no more
2607 states to pop, we return failure.
2609 Sometimes we also need to backtrack on success; for example /A+/, where
2610 after successfully matching one A, we need to go back and try to
2611 match another one; similarly for lookahead assertions: if the assertion
2612 completes successfully, we backtrack to the state just before the assertion
2613 and then carry on. In these cases, the pushed state is marked as
2614 'backtrack on success too'. This marking is in fact done by a chain of
2615 pointers, each pointing to the previous 'yes' state. On success, we pop to
2616 the nearest yes state, discarding any intermediate failure-only states.
2617 Sometimes a yes state is pushed just to force some cleanup code to be
2618 called at the end of a successful match or submatch; e.g. (??{$re}) uses
2619 it to free the inner regex.
2621 Note that failure backtracking rewinds the cursor position, while
2622 success backtracking leaves it alone.
2624 A pattern is complete when the END op is executed, while a subpattern
2625 such as (?=foo) is complete when the SUCCESS op is executed. Both of these
2626 ops trigger the "pop to last yes state if any, otherwise return true"
2629 A common convention in this function is to use A and B to refer to the two
2630 subpatterns (or to the first nodes thereof) in patterns like /A*B/: so A is
2631 the subpattern to be matched possibly multiple times, while B is the entire
2632 rest of the pattern. Variable and state names reflect this convention.
2634 The states in the main switch are the union of ops and failure/success of
2635 substates associated with with that op. For example, IFMATCH is the op
2636 that does lookahead assertions /(?=A)B/ and so the IFMATCH state means
2637 'execute IFMATCH'; while IFMATCH_A is a state saying that we have just
2638 successfully matched A and IFMATCH_A_fail is a state saying that we have
2639 just failed to match A. Resume states always come in pairs. The backtrack
2640 state we push is marked as 'IFMATCH_A', but when that is popped, we resume
2641 at IFMATCH_A or IFMATCH_A_fail, depending on whether we are backtracking
2642 on success or failure.
2644 The struct that holds a backtracking state is actually a big union, with
2645 one variant for each major type of op. The variable st points to the
2646 top-most backtrack struct. To make the code clearer, within each
2647 block of code we #define ST to alias the relevant union.
2649 Here's a concrete example of a (vastly oversimplified) IFMATCH
2655 #define ST st->u.ifmatch
2657 case IFMATCH: // we are executing the IFMATCH op, (?=A)B
2658 ST.foo = ...; // some state we wish to save
2660 // push a yes backtrack state with a resume value of
2661 // IFMATCH_A/IFMATCH_A_fail, then continue execution at the
2663 PUSH_YES_STATE_GOTO(IFMATCH_A, A);
2666 case IFMATCH_A: // we have successfully executed A; now continue with B
2668 bar = ST.foo; // do something with the preserved value
2671 case IFMATCH_A_fail: // A failed, so the assertion failed
2672 ...; // do some housekeeping, then ...
2673 sayNO; // propagate the failure
2680 For any old-timers reading this who are familiar with the old recursive
2681 approach, the code above is equivalent to:
2683 case IFMATCH: // we are executing the IFMATCH op, (?=A)B
2692 ...; // do some housekeeping, then ...
2693 sayNO; // propagate the failure
2696 The topmost backtrack state, pointed to by st, is usually free. If you
2697 want to claim it, populate any ST.foo fields in it with values you wish to
2698 save, then do one of
2700 PUSH_STATE_GOTO(resume_state, node);
2701 PUSH_YES_STATE_GOTO(resume_state, node);
2703 which sets that backtrack state's resume value to 'resume_state', pushes a
2704 new free entry to the top of the backtrack stack, then goes to 'node'.
2705 On backtracking, the free slot is popped, and the saved state becomes the
2706 new free state. An ST.foo field in this new top state can be temporarily
2707 accessed to retrieve values, but once the main loop is re-entered, it
2708 becomes available for reuse.
2710 Note that the depth of the backtrack stack constantly increases during the
2711 left-to-right execution of the pattern, rather than going up and down with
2712 the pattern nesting. For example the stack is at its maximum at Z at the
2713 end of the pattern, rather than at X in the following:
2715 /(((X)+)+)+....(Y)+....Z/
2717 The only exceptions to this are lookahead/behind assertions and the cut,
2718 (?>A), which pop all the backtrack states associated with A before
2721 Bascktrack state structs are allocated in slabs of about 4K in size.
2722 PL_regmatch_state and st always point to the currently active state,
2723 and PL_regmatch_slab points to the slab currently containing
2724 PL_regmatch_state. The first time regmatch() is called, the first slab is
2725 allocated, and is never freed until interpreter destruction. When the slab
2726 is full, a new one is allocated and chained to the end. At exit from
2727 regmatch(), slabs allocated since entry are freed.
2732 #define DEBUG_STATE_pp(pp) \
2734 DUMP_EXEC_POS(locinput, scan, utf8_target); \
2735 PerlIO_printf(Perl_debug_log, \
2736 " %*s"pp" %s%s%s%s%s\n", \
2738 PL_reg_name[st->resume_state], \
2739 ((st==yes_state||st==mark_state) ? "[" : ""), \
2740 ((st==yes_state) ? "Y" : ""), \
2741 ((st==mark_state) ? "M" : ""), \
2742 ((st==yes_state||st==mark_state) ? "]" : "") \
2747 #define REG_NODE_NUM(x) ((x) ? (int)((x)-prog) : -1)
2752 S_debug_start_match(pTHX_ const REGEXP *prog, const bool utf8_target,
2753 const char *start, const char *end, const char *blurb)
2755 const bool utf8_pat = RX_UTF8(prog) ? 1 : 0;
2757 PERL_ARGS_ASSERT_DEBUG_START_MATCH;
2762 RE_PV_QUOTED_DECL(s0, utf8_pat, PERL_DEBUG_PAD_ZERO(0),
2763 RX_PRECOMP_const(prog), RX_PRELEN(prog), 60);
2765 RE_PV_QUOTED_DECL(s1, utf8_target, PERL_DEBUG_PAD_ZERO(1),
2766 start, end - start, 60);
2768 PerlIO_printf(Perl_debug_log,
2769 "%s%s REx%s %s against %s\n",
2770 PL_colors[4], blurb, PL_colors[5], s0, s1);
2772 if (utf8_target||utf8_pat)
2773 PerlIO_printf(Perl_debug_log, "UTF-8 %s%s%s...\n",
2774 utf8_pat ? "pattern" : "",
2775 utf8_pat && utf8_target ? " and " : "",
2776 utf8_target ? "string" : ""
2782 S_dump_exec_pos(pTHX_ const char *locinput,
2783 const regnode *scan,
2784 const char *loc_regeol,
2785 const char *loc_bostr,
2786 const char *loc_reg_starttry,
2787 const bool utf8_target)
2789 const int docolor = *PL_colors[0] || *PL_colors[2] || *PL_colors[4];
2790 const int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */
2791 int l = (loc_regeol - locinput) > taill ? taill : (loc_regeol - locinput);
2792 /* The part of the string before starttry has one color
2793 (pref0_len chars), between starttry and current
2794 position another one (pref_len - pref0_len chars),
2795 after the current position the third one.
2796 We assume that pref0_len <= pref_len, otherwise we
2797 decrease pref0_len. */
2798 int pref_len = (locinput - loc_bostr) > (5 + taill) - l
2799 ? (5 + taill) - l : locinput - loc_bostr;
2802 PERL_ARGS_ASSERT_DUMP_EXEC_POS;
2804 while (utf8_target && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len)))
2806 pref0_len = pref_len - (locinput - loc_reg_starttry);
2807 if (l + pref_len < (5 + taill) && l < loc_regeol - locinput)
2808 l = ( loc_regeol - locinput > (5 + taill) - pref_len
2809 ? (5 + taill) - pref_len : loc_regeol - locinput);
2810 while (utf8_target && UTF8_IS_CONTINUATION(*(U8*)(locinput + l)))
2814 if (pref0_len > pref_len)
2815 pref0_len = pref_len;
2817 const int is_uni = (utf8_target && OP(scan) != CANY) ? 1 : 0;
2819 RE_PV_COLOR_DECL(s0,len0,is_uni,PERL_DEBUG_PAD(0),
2820 (locinput - pref_len),pref0_len, 60, 4, 5);
2822 RE_PV_COLOR_DECL(s1,len1,is_uni,PERL_DEBUG_PAD(1),
2823 (locinput - pref_len + pref0_len),
2824 pref_len - pref0_len, 60, 2, 3);
2826 RE_PV_COLOR_DECL(s2,len2,is_uni,PERL_DEBUG_PAD(2),
2827 locinput, loc_regeol - locinput, 10, 0, 1);
2829 const STRLEN tlen=len0+len1+len2;
2830 PerlIO_printf(Perl_debug_log,
2831 "%4"IVdf" <%.*s%.*s%s%.*s>%*s|",
2832 (IV)(locinput - loc_bostr),
2835 (docolor ? "" : "> <"),
2837 (int)(tlen > 19 ? 0 : 19 - tlen),
2844 /* reg_check_named_buff_matched()
2845 * Checks to see if a named buffer has matched. The data array of
2846 * buffer numbers corresponding to the buffer is expected to reside
2847 * in the regexp->data->data array in the slot stored in the ARG() of
2848 * node involved. Note that this routine doesn't actually care about the
2849 * name, that information is not preserved from compilation to execution.
2850 * Returns the index of the leftmost defined buffer with the given name
2851 * or 0 if non of the buffers matched.
2854 S_reg_check_named_buff_matched(pTHX_ const regexp *rex, const regnode *scan)
2857 RXi_GET_DECL(rex,rexi);
2858 SV *sv_dat= MUTABLE_SV(rexi->data->data[ ARG( scan ) ]);
2859 I32 *nums=(I32*)SvPVX(sv_dat);
2861 PERL_ARGS_ASSERT_REG_CHECK_NAMED_BUFF_MATCHED;
2863 for ( n=0; n<SvIVX(sv_dat); n++ ) {
2864 if ((I32)*PL_reglastparen >= nums[n] &&
2865 PL_regoffs[nums[n]].end != -1)
2874 /* free all slabs above current one - called during LEAVE_SCOPE */
2877 S_clear_backtrack_stack(pTHX_ void *p)
2879 regmatch_slab *s = PL_regmatch_slab->next;
2884 PL_regmatch_slab->next = NULL;
2886 regmatch_slab * const osl = s;
2893 #define SETREX(Re1,Re2) \
2894 if (PL_reg_eval_set) PM_SETRE((PL_reg_curpm), (Re2)); \
2897 STATIC I32 /* 0 failure, 1 success */
2898 S_regmatch(pTHX_ regmatch_info *reginfo, regnode *prog)
2900 #if PERL_VERSION < 9 && !defined(PERL_CORE)
2904 register const bool utf8_target = PL_reg_match_utf8;
2905 const U32 uniflags = UTF8_ALLOW_DEFAULT;
2906 REGEXP *rex_sv = reginfo->prog;
2907 regexp *rex = (struct regexp *)SvANY(rex_sv);
2908 RXi_GET_DECL(rex,rexi);
2910 /* the current state. This is a cached copy of PL_regmatch_state */
2911 register regmatch_state *st;
2912 /* cache heavy used fields of st in registers */
2913 register regnode *scan;
2914 register regnode *next;
2915 register U32 n = 0; /* general value; init to avoid compiler warning */
2916 register I32 ln = 0; /* len or last; init to avoid compiler warning */
2917 register char *locinput = PL_reginput;
2918 register I32 nextchr; /* is always set to UCHARAT(locinput) */
2920 bool result = 0; /* return value of S_regmatch */
2921 int depth = 0; /* depth of backtrack stack */
2922 U32 nochange_depth = 0; /* depth of GOSUB recursion with nochange */
2923 const U32 max_nochange_depth =
2924 (3 * rex->nparens > MAX_RECURSE_EVAL_NOCHANGE_DEPTH) ?
2925 3 * rex->nparens : MAX_RECURSE_EVAL_NOCHANGE_DEPTH;
2926 regmatch_state *yes_state = NULL; /* state to pop to on success of
2928 /* mark_state piggy backs on the yes_state logic so that when we unwind
2929 the stack on success we can update the mark_state as we go */
2930 regmatch_state *mark_state = NULL; /* last mark state we have seen */
2931 regmatch_state *cur_eval = NULL; /* most recent EVAL_AB state */
2932 struct regmatch_state *cur_curlyx = NULL; /* most recent curlyx */
2934 bool no_final = 0; /* prevent failure from backtracking? */
2935 bool do_cutgroup = 0; /* no_final only until next branch/trie entry */
2936 char *startpoint = PL_reginput;
2937 SV *popmark = NULL; /* are we looking for a mark? */
2938 SV *sv_commit = NULL; /* last mark name seen in failure */
2939 SV *sv_yes_mark = NULL; /* last mark name we have seen
2940 during a successfull match */
2941 U32 lastopen = 0; /* last open we saw */
2942 bool has_cutgroup = RX_HAS_CUTGROUP(rex) ? 1 : 0;
2943 SV* const oreplsv = GvSV(PL_replgv);
2944 /* these three flags are set by various ops to signal information to
2945 * the very next op. They have a useful lifetime of exactly one loop
2946 * iteration, and are not preserved or restored by state pushes/pops
2948 bool sw = 0; /* the condition value in (?(cond)a|b) */
2949 bool minmod = 0; /* the next "{n,m}" is a "{n,m}?" */
2950 int logical = 0; /* the following EVAL is:
2954 or the following IFMATCH/UNLESSM is:
2955 false: plain (?=foo)
2956 true: used as a condition: (?(?=foo))
2959 GET_RE_DEBUG_FLAGS_DECL;
2962 PERL_ARGS_ASSERT_REGMATCH;
2964 DEBUG_OPTIMISE_r( DEBUG_EXECUTE_r({
2965 PerlIO_printf(Perl_debug_log,"regmatch start\n");
2967 /* on first ever call to regmatch, allocate first slab */
2968 if (!PL_regmatch_slab) {
2969 Newx(PL_regmatch_slab, 1, regmatch_slab);
2970 PL_regmatch_slab->prev = NULL;
2971 PL_regmatch_slab->next = NULL;
2972 PL_regmatch_state = SLAB_FIRST(PL_regmatch_slab);
2975 oldsave = PL_savestack_ix;
2976 SAVEDESTRUCTOR_X(S_clear_backtrack_stack, NULL);
2977 SAVEVPTR(PL_regmatch_slab);
2978 SAVEVPTR(PL_regmatch_state);
2980 /* grab next free state slot */
2981 st = ++PL_regmatch_state;
2982 if (st > SLAB_LAST(PL_regmatch_slab))
2983 st = PL_regmatch_state = S_push_slab(aTHX);
2985 /* Note that nextchr is a byte even in UTF */
2986 nextchr = UCHARAT(locinput);
2988 while (scan != NULL) {
2991 SV * const prop = sv_newmortal();
2992 regnode *rnext=regnext(scan);
2993 DUMP_EXEC_POS( locinput, scan, utf8_target );
2994 regprop(rex, prop, scan);
2996 PerlIO_printf(Perl_debug_log,
2997 "%3"IVdf":%*s%s(%"IVdf")\n",
2998 (IV)(scan - rexi->program), depth*2, "",
3000 (PL_regkind[OP(scan)] == END || !rnext) ?
3001 0 : (IV)(rnext - rexi->program));
3004 next = scan + NEXT_OFF(scan);
3007 state_num = OP(scan);
3011 assert(PL_reglastparen == &rex->lastparen);
3012 assert(PL_reglastcloseparen == &rex->lastcloseparen);
3013 assert(PL_regoffs == rex->offs);
3015 switch (state_num) {
3017 if (locinput == PL_bostr)
3019 /* reginfo->till = reginfo->bol; */
3024 if (locinput == PL_bostr ||
3025 ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n'))
3031 if (locinput == PL_bostr)
3035 if (locinput == reginfo->ganch)
3040 /* update the startpoint */
3041 st->u.keeper.val = PL_regoffs[0].start;
3042 PL_reginput = locinput;
3043 PL_regoffs[0].start = locinput - PL_bostr;
3044 PUSH_STATE_GOTO(KEEPS_next, next);
3046 case KEEPS_next_fail:
3047 /* rollback the start point change */
3048 PL_regoffs[0].start = st->u.keeper.val;
3054 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
3059 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
3061 if (PL_regeol - locinput > 1)
3065 if (PL_regeol != locinput)
3069 if (!nextchr && locinput >= PL_regeol)
3072 locinput += PL_utf8skip[nextchr];
3073 if (locinput > PL_regeol)
3075 nextchr = UCHARAT(locinput);
3078 nextchr = UCHARAT(++locinput);
3081 if (!nextchr && locinput >= PL_regeol)
3083 nextchr = UCHARAT(++locinput);
3086 if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n')
3089 locinput += PL_utf8skip[nextchr];
3090 if (locinput > PL_regeol)
3092 nextchr = UCHARAT(locinput);
3095 nextchr = UCHARAT(++locinput);
3099 #define ST st->u.trie
3101 /* In this case the charclass data is available inline so
3102 we can fail fast without a lot of extra overhead.
3104 if (scan->flags == EXACT || !utf8_target) {
3105 if(!ANYOF_BITMAP_TEST(scan, *locinput)) {
3107 PerlIO_printf(Perl_debug_log,
3108 "%*s %sfailed to match trie start class...%s\n",
3109 REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5])
3117 /* the basic plan of execution of the trie is:
3118 * At the beginning, run though all the states, and
3119 * find the longest-matching word. Also remember the position
3120 * of the shortest matching word. For example, this pattern:
3123 * when matched against the string "abcde", will generate
3124 * accept states for all words except 3, with the longest
3125 * matching word being 4, and the shortest being 1 (with
3126 * the position being after char 1 of the string).
3128 * Then for each matching word, in word order (i.e. 1,2,4,5),
3129 * we run the remainder of the pattern; on each try setting
3130 * the current position to the character following the word,
3131 * returning to try the next word on failure.
3133 * We avoid having to build a list of words at runtime by
3134 * using a compile-time structure, wordinfo[].prev, which
3135 * gives, for each word, the previous accepting word (if any).
3136 * In the case above it would contain the mappings 1->2, 2->0,
3137 * 3->0, 4->5, 5->1. We can use this table to generate, from
3138 * the longest word (4 above), a list of all words, by
3139 * following the list of prev pointers; this gives us the
3140 * unordered list 4,5,1,2. Then given the current word we have
3141 * just tried, we can go through the list and find the
3142 * next-biggest word to try (so if we just failed on word 2,
3143 * the next in the list is 4).
3145 * Since at runtime we don't record the matching position in
3146 * the string for each word, we have to work that out for
3147 * each word we're about to process. The wordinfo table holds
3148 * the character length of each word; given that we recorded
3149 * at the start: the position of the shortest word and its
3150 * length in chars, we just need to move the pointer the
3151 * difference between the two char lengths. Depending on
3152 * Unicode status and folding, that's cheap or expensive.
3154 * This algorithm is optimised for the case where are only a
3155 * small number of accept states, i.e. 0,1, or maybe 2.
3156 * With lots of accepts states, and having to try all of them,
3157 * it becomes quadratic on number of accept states to find all
3162 /* what type of TRIE am I? (utf8 makes this contextual) */
3163 DECL_TRIE_TYPE(scan);
3165 /* what trie are we using right now */
3166 reg_trie_data * const trie
3167 = (reg_trie_data*)rexi->data->data[ ARG( scan ) ];
3168 HV * widecharmap = MUTABLE_HV(rexi->data->data[ ARG( scan ) + 1 ]);
3169 U32 state = trie->startstate;
3171 if (trie->bitmap && trie_type != trie_utf8_fold &&
3172 !TRIE_BITMAP_TEST(trie,*locinput)
3174 if (trie->states[ state ].wordnum) {
3176 PerlIO_printf(Perl_debug_log,
3177 "%*s %smatched empty string...%s\n",
3178 REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5])
3183 PerlIO_printf(Perl_debug_log,
3184 "%*s %sfailed to match trie start class...%s\n",
3185 REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5])
3192 U8 *uc = ( U8* )locinput;
3196 U8 *uscan = (U8*)NULL;
3197 U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ];
3198 U32 charcount = 0; /* how many input chars we have matched */
3199 U32 accepted = 0; /* have we seen any accepting states? */
3202 ST.jump = trie->jump;
3205 ST.longfold = FALSE; /* char longer if folded => it's harder */
3208 /* fully traverse the TRIE; note the position of the
3209 shortest accept state and the wordnum of the longest
3212 while ( state && uc <= (U8*)PL_regeol ) {
3213 U32 base = trie->states[ state ].trans.base;
3217 wordnum = trie->states[ state ].wordnum;
3219 if (wordnum) { /* it's an accept state */
3222 /* record first match position */
3224 ST.firstpos = (U8*)locinput;
3229 ST.firstchars = charcount;
3232 if (!ST.nextword || wordnum < ST.nextword)
3233 ST.nextword = wordnum;
3234 ST.topword = wordnum;
3237 DEBUG_TRIE_EXECUTE_r({
3238 DUMP_EXEC_POS( (char *)uc, scan, utf8_target );
3239 PerlIO_printf( Perl_debug_log,
3240 "%*s %sState: %4"UVxf" Accepted: %c ",
3241 2+depth * 2, "", PL_colors[4],
3242 (UV)state, (accepted ? 'Y' : 'N'));
3245 /* read a char and goto next state */
3248 REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc,
3249 uscan, len, uvc, charid, foldlen,
3256 base + charid - 1 - trie->uniquecharcount)) >= 0)
3258 && ((U32)offset < trie->lasttrans)
3259 && trie->trans[offset].check == state)
3261 state = trie->trans[offset].next;
3272 DEBUG_TRIE_EXECUTE_r(
3273 PerlIO_printf( Perl_debug_log,
3274 "Charid:%3x CP:%4"UVxf" After State: %4"UVxf"%s\n",
3275 charid, uvc, (UV)state, PL_colors[5] );
3281 /* calculate total number of accept states */
3286 w = trie->wordinfo[w].prev;
3289 ST.accepted = accepted;
3293 PerlIO_printf( Perl_debug_log,
3294 "%*s %sgot %"IVdf" possible matches%s\n",
3295 REPORT_CODE_OFF + depth * 2, "",
3296 PL_colors[4], (IV)ST.accepted, PL_colors[5] );
3298 goto trie_first_try; /* jump into the fail handler */
3302 case TRIE_next_fail: /* we failed - try next alternative */
3304 REGCP_UNWIND(ST.cp);
3305 for (n = *PL_reglastparen; n > ST.lastparen; n--)
3306 PL_regoffs[n].end = -1;
3307 *PL_reglastparen = n;
3309 if (!--ST.accepted) {
3311 PerlIO_printf( Perl_debug_log,
3312 "%*s %sTRIE failed...%s\n",
3313 REPORT_CODE_OFF+depth*2, "",
3320 /* Find next-highest word to process. Note that this code
3321 * is O(N^2) per trie run (O(N) per branch), so keep tight */
3322 register U16 min = 0;
3324 register U16 const nextword = ST.nextword;
3325 register reg_trie_wordinfo * const wordinfo
3326 = ((reg_trie_data*)rexi->data->data[ARG(ST.me)])->wordinfo;
3327 for (word=ST.topword; word; word=wordinfo[word].prev) {
3328 if (word > nextword && (!min || word < min))
3341 ST.lastparen = *PL_reglastparen;
3345 /* find start char of end of current word */
3347 U32 chars; /* how many chars to skip */
3348 U8 *uc = ST.firstpos;
3349 reg_trie_data * const trie
3350 = (reg_trie_data*)rexi->data->data[ARG(ST.me)];
3352 assert((trie->wordinfo[ST.nextword].len - trie->prefixlen)
3354 chars = (trie->wordinfo[ST.nextword].len - trie->prefixlen)
3358 /* the hard option - fold each char in turn and find
3359 * its folded length (which may be different */
3360 U8 foldbuf[UTF8_MAXBYTES_CASE + 1];
3368 uvc = utf8n_to_uvuni((U8*)uc, UTF8_MAXLEN, &len,
3376 uvc = to_uni_fold(uvc, foldbuf, &foldlen);
3381 uvc = utf8n_to_uvuni(uscan, UTF8_MAXLEN, &len,
3395 PL_reginput = (char *)uc;
3398 scan = (ST.jump && ST.jump[ST.nextword])
3399 ? ST.me + ST.jump[ST.nextword]
3403 PerlIO_printf( Perl_debug_log,
3404 "%*s %sTRIE matched word #%d, continuing%s\n",
3405 REPORT_CODE_OFF+depth*2, "",
3412 if (ST.accepted > 1 || has_cutgroup) {
3413 PUSH_STATE_GOTO(TRIE_next, scan);
3416 /* only one choice left - just continue */
3418 AV *const trie_words
3419 = MUTABLE_AV(rexi->data->data[ARG(ST.me)+TRIE_WORDS_OFFSET]);
3420 SV ** const tmp = av_fetch( trie_words,
3422 SV *sv= tmp ? sv_newmortal() : NULL;
3424 PerlIO_printf( Perl_debug_log,
3425 "%*s %sonly one match left, short-circuiting: #%d <%s>%s\n",
3426 REPORT_CODE_OFF+depth*2, "", PL_colors[4],
3428 tmp ? pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), 0,
3429 PL_colors[0], PL_colors[1],
3430 (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0)
3432 : "not compiled under -Dr",
3436 locinput = PL_reginput;
3437 nextchr = UCHARAT(locinput);
3438 continue; /* execute rest of RE */
3443 char *s = STRING(scan);
3445 if (utf8_target != UTF_PATTERN) {
3446 /* The target and the pattern have differing utf8ness. */
3448 const char * const e = s + ln;
3451 /* The target is utf8, the pattern is not utf8. */
3456 if (NATIVE_TO_UNI(*(U8*)s) !=
3457 utf8n_to_uvuni((U8*)l, UTF8_MAXBYTES, &ulen,
3465 /* The target is not utf8, the pattern is utf8. */
3470 if (NATIVE_TO_UNI(*((U8*)l)) !=
3471 utf8n_to_uvuni((U8*)s, UTF8_MAXBYTES, &ulen,
3479 nextchr = UCHARAT(locinput);
3482 /* The target and the pattern have the same utf8ness. */
3483 /* Inline the first character, for speed. */
3484 if (UCHARAT(s) != nextchr)
3486 if (PL_regeol - locinput < ln)
3488 if (ln > 1 && memNE(s, locinput, ln))
3491 nextchr = UCHARAT(locinput);
3495 PL_reg_flags |= RF_tainted;
3498 char * const s = STRING(scan);
3501 if (utf8_target || UTF_PATTERN) {
3502 /* Either target or the pattern are utf8. */
3503 const char * const l = locinput;
3504 char *e = PL_regeol;
3506 if (! foldEQ_utf8(s, 0, ln, cBOOL(UTF_PATTERN),
3507 l, &e, 0, utf8_target)) {
3508 /* One more case for the sharp s:
3509 * pack("U0U*", 0xDF) =~ /ss/i,
3510 * the 0xC3 0x9F are the UTF-8
3511 * byte sequence for the U+00DF. */
3513 if (!(utf8_target &&
3514 toLOWER(s[0]) == 's' &&
3516 toLOWER(s[1]) == 's' &&
3523 nextchr = UCHARAT(locinput);
3527 /* Neither the target and the pattern are utf8. */
3529 /* Inline the first character, for speed. */
3530 if (UCHARAT(s) != nextchr &&
3531 UCHARAT(s) != ((OP(scan) == EXACTF)
3532 ? PL_fold : PL_fold_locale)[nextchr])
3534 if (PL_regeol - locinput < ln)
3536 if (ln > 1 && (OP(scan) == EXACTF
3537 ? ! foldEQ(s, locinput, ln)
3538 : ! foldEQ_locale(s, locinput, ln)))
3541 nextchr = UCHARAT(locinput);
3546 PL_reg_flags |= RF_tainted;
3550 /* was last char in word? */
3552 if (locinput == PL_bostr)
3555 const U8 * const r = reghop3((U8*)locinput, -1, (U8*)PL_bostr);
3557 ln = utf8n_to_uvchr(r, UTF8SKIP(r), 0, uniflags);
3559 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
3560 ln = isALNUM_uni(ln);
3561 LOAD_UTF8_CHARCLASS_ALNUM();
3562 n = swash_fetch(PL_utf8_alnum, (U8*)locinput, utf8_target);
3565 ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(ln));
3566 n = isALNUM_LC_utf8((U8*)locinput);
3570 ln = (locinput != PL_bostr) ?
3571 UCHARAT(locinput - 1) : '\n';
3572 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
3574 n = isALNUM(nextchr);
3577 ln = isALNUM_LC(ln);
3578 n = isALNUM_LC(nextchr);
3581 if (((!ln) == (!n)) == (OP(scan) == BOUND ||
3582 OP(scan) == BOUNDL))
3587 STRLEN inclasslen = PL_regeol - locinput;
3589 if (!reginclass(rex, scan, (U8*)locinput, &inclasslen, utf8_target))
3591 if (locinput >= PL_regeol)
3593 locinput += inclasslen ? inclasslen : UTF8SKIP(locinput);
3594 nextchr = UCHARAT(locinput);
3599 nextchr = UCHARAT(locinput);
3600 if (!REGINCLASS(rex, scan, (U8*)locinput))
3602 if (!nextchr && locinput >= PL_regeol)
3604 nextchr = UCHARAT(++locinput);
3608 /* If we might have the case of the German sharp s
3609 * in a casefolding Unicode character class. */
3611 if (ANYOF_FOLD_SHARP_S(scan, locinput, PL_regeol)) {
3612 locinput += SHARP_S_SKIP;
3613 nextchr = UCHARAT(locinput);
3618 /* Special char classes - The defines start on line 129 or so */
3619 CCC_TRY_AFF( ALNUM, ALNUML, perl_word, "a", isALNUM_LC_utf8, isALNUM, isALNUM_LC);
3620 CCC_TRY_NEG(NALNUM, NALNUML, perl_word, "a", isALNUM_LC_utf8, isALNUM, isALNUM_LC);
3622 CCC_TRY_AFF( SPACE, SPACEL, perl_space, " ", isSPACE_LC_utf8, isSPACE, isSPACE_LC);
3623 CCC_TRY_NEG(NSPACE, NSPACEL, perl_space, " ", isSPACE_LC_utf8, isSPACE, isSPACE_LC);
3625 CCC_TRY_AFF( DIGIT, DIGITL, posix_digit, "0", isDIGIT_LC_utf8, isDIGIT, isDIGIT_LC);
3626 CCC_TRY_NEG(NDIGIT, NDIGITL, posix_digit, "0", isDIGIT_LC_utf8, isDIGIT, isDIGIT_LC);
3628 case CLUMP: /* Match \X: logical Unicode character. This is defined as
3629 a Unicode extended Grapheme Cluster */
3630 /* From http://www.unicode.org/reports/tr29 (5.2 version). An
3631 extended Grapheme Cluster is:
3634 | Prepend* Begin Extend*
3637 Begin is (Hangul-syllable | ! Control)
3638 Extend is (Grapheme_Extend | Spacing_Mark)
3639 Control is [ GCB_Control CR LF ]
3641 The discussion below shows how the code for CLUMP is derived
3642 from this regex. Note that most of these concepts are from
3643 property values of the Grapheme Cluster Boundary (GCB) property.
3644 No code point can have multiple property values for a given
3645 property. Thus a code point in Prepend can't be in Control, but
3646 it must be in !Control. This is why Control above includes
3647 GCB_Control plus CR plus LF. The latter two are used in the GCB
3648 property separately, and so can't be in GCB_Control, even though
3649 they logically are controls. Control is not the same as gc=cc,
3650 but includes format and other characters as well.
3652 The Unicode definition of Hangul-syllable is:
3654 | (L* ( ( V | LV ) V* | LVT ) T*)
3657 Each of these is a value for the GCB property, and hence must be
3658 disjoint, so the order they are tested is immaterial, so the
3659 above can safely be changed to
3662 | (L* ( LVT | ( V | LV ) V*) T*)
3664 The last two terms can be combined like this:
3666 | (( LVT | ( V | LV ) V*) T*))
3668 And refactored into this:
3669 L* (L | LVT T* | V V* T* | LV V* T*)
3671 That means that if we have seen any L's at all we can quit
3672 there, but if the next character is a LVT, a V or and LV we
3675 There is a subtlety with Prepend* which showed up in testing.
3676 Note that the Begin, and only the Begin is required in:
3677 | Prepend* Begin Extend*
3678 Also, Begin contains '! Control'. A Prepend must be a '!
3679 Control', which means it must be a Begin. What it comes down to
3680 is that if we match Prepend* and then find no suitable Begin
3681 afterwards, that if we backtrack the last Prepend, that one will
3682 be a suitable Begin.
3685 if (locinput >= PL_regeol)
3687 if (! utf8_target) {
3689 /* Match either CR LF or '.', as all the other possibilities
3691 locinput++; /* Match the . or CR */
3693 && locinput < PL_regeol
3694 && UCHARAT(locinput) == '\n') locinput++;
3698 /* Utf8: See if is ( CR LF ); already know that locinput <
3699 * PL_regeol, so locinput+1 is in bounds */
3700 if (nextchr == '\r' && UCHARAT(locinput + 1) == '\n') {
3704 /* In case have to backtrack to beginning, then match '.' */
3705 char *starting = locinput;
3707 /* In case have to backtrack the last prepend */
3708 char *previous_prepend = 0;
3710 LOAD_UTF8_CHARCLASS_GCB();
3712 /* Match (prepend)* */
3713 while (locinput < PL_regeol
3714 && swash_fetch(PL_utf8_X_prepend,
3715 (U8*)locinput, utf8_target))
3717 previous_prepend = locinput;
3718 locinput += UTF8SKIP(locinput);
3721 /* As noted above, if we matched a prepend character, but
3722 * the next thing won't match, back off the last prepend we
3723 * matched, as it is guaranteed to match the begin */
3724 if (previous_prepend
3725 && (locinput >= PL_regeol
3726 || ! swash_fetch(PL_utf8_X_begin,
3727 (U8*)locinput, utf8_target)))
3729 locinput = previous_prepend;
3732 /* Note that here we know PL_regeol > locinput, as we
3733 * tested that upon input to this switch case, and if we
3734 * moved locinput forward, we tested the result just above
3735 * and it either passed, or we backed off so that it will
3737 if (! swash_fetch(PL_utf8_X_begin, (U8*)locinput, utf8_target)) {
3739 /* Here did not match the required 'Begin' in the
3740 * second term. So just match the very first
3741 * character, the '.' of the final term of the regex */
3742 locinput = starting + UTF8SKIP(starting);
3745 /* Here is the beginning of a character that can have
3746 * an extender. It is either a hangul syllable, or a
3748 if (swash_fetch(PL_utf8_X_non_hangul,
3749 (U8*)locinput, utf8_target))
3752 /* Here not a Hangul syllable, must be a
3753 * ('! * Control') */
3754 locinput += UTF8SKIP(locinput);
3757 /* Here is a Hangul syllable. It can be composed
3758 * of several individual characters. One
3759 * possibility is T+ */
3760 if (swash_fetch(PL_utf8_X_T,
3761 (U8*)locinput, utf8_target))
3763 while (locinput < PL_regeol
3764 && swash_fetch(PL_utf8_X_T,
3765 (U8*)locinput, utf8_target))
3767 locinput += UTF8SKIP(locinput);
3771 /* Here, not T+, but is a Hangul. That means
3772 * it is one of the others: L, LV, LVT or V,
3774 * L* (L | LVT T* | V V* T* | LV V* T*) */
3777 while (locinput < PL_regeol
3778 && swash_fetch(PL_utf8_X_L,
3779 (U8*)locinput, utf8_target))
3781 locinput += UTF8SKIP(locinput);
3784 /* Here, have exhausted L*. If the next
3785 * character is not an LV, LVT nor V, it means
3786 * we had to have at least one L, so matches L+
3787 * in the original equation, we have a complete
3788 * hangul syllable. Are done. */
3790 if (locinput < PL_regeol
3791 && swash_fetch(PL_utf8_X_LV_LVT_V,
3792 (U8*)locinput, utf8_target))
3795 /* Otherwise keep going. Must be LV, LVT
3796 * or V. See if LVT */
3797 if (swash_fetch(PL_utf8_X_LVT,
3798 (U8*)locinput, utf8_target))
3800 locinput += UTF8SKIP(locinput);
3803 /* Must be V or LV. Take it, then
3805 locinput += UTF8SKIP(locinput);
3806 while (locinput < PL_regeol
3807 && swash_fetch(PL_utf8_X_V,
3808 (U8*)locinput, utf8_target))
3810 locinput += UTF8SKIP(locinput);
3814 /* And any of LV, LVT, or V can be followed
3816 while (locinput < PL_regeol
3817 && swash_fetch(PL_utf8_X_T,
3821 locinput += UTF8SKIP(locinput);
3827 /* Match any extender */
3828 while (locinput < PL_regeol
3829 && swash_fetch(PL_utf8_X_extend,
3830 (U8*)locinput, utf8_target))
3832 locinput += UTF8SKIP(locinput);
3836 if (locinput > PL_regeol) sayNO;
3838 nextchr = UCHARAT(locinput);
3845 PL_reg_flags |= RF_tainted;
3850 n = reg_check_named_buff_matched(rex,scan);
3853 type = REF + ( type - NREF );
3860 PL_reg_flags |= RF_tainted;
3864 n = ARG(scan); /* which paren pair */
3867 ln = PL_regoffs[n].start;
3868 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
3869 if (*PL_reglastparen < n || ln == -1)
3870 sayNO; /* Do not match unless seen CLOSEn. */
3871 if (ln == PL_regoffs[n].end)
3875 if (utf8_target && type != REF) { /* REF can do byte comparison */
3877 const char *e = PL_bostr + PL_regoffs[n].end;
3879 * Note that we can't do the "other character" lookup trick as
3880 * in the 8-bit case (no pun intended) because in Unicode we
3881 * have to map both upper and title case to lower case.
3885 STRLEN ulen1, ulen2;
3886 U8 tmpbuf1[UTF8_MAXBYTES_CASE+1];
3887 U8 tmpbuf2[UTF8_MAXBYTES_CASE+1];
3891 toLOWER_utf8((U8*)s, tmpbuf1, &ulen1);
3892 toLOWER_utf8((U8*)l, tmpbuf2, &ulen2);
3893 if (ulen1 != ulen2 || memNE((char *)tmpbuf1, (char *)tmpbuf2, ulen1))
3900 nextchr = UCHARAT(locinput);
3904 /* Inline the first character, for speed. */
3905 if (UCHARAT(s) != nextchr &&
3907 (UCHARAT(s) != (type == REFF
3908 ? PL_fold : PL_fold_locale)[nextchr])))
3910 ln = PL_regoffs[n].end - ln;
3911 if (locinput + ln > PL_regeol)
3913 if (ln > 1 && (type == REF
3914 ? memNE(s, locinput, ln)
3916 ? ! foldEQ(s, locinput, ln)
3917 : ! foldEQ_locale(s, locinput, ln))))
3920 nextchr = UCHARAT(locinput);
3930 #define ST st->u.eval
3935 regexp_internal *rei;
3936 regnode *startpoint;
3939 case GOSUB: /* /(...(?1))/ /(...(?&foo))/ */
3940 if (cur_eval && cur_eval->locinput==locinput) {
3941 if (cur_eval->u.eval.close_paren == (U32)ARG(scan))
3942 Perl_croak(aTHX_ "Infinite recursion in regex");
3943 if ( ++nochange_depth > max_nochange_depth )
3945 "Pattern subroutine nesting without pos change"
3946 " exceeded limit in regex");
3953 (void)ReREFCNT_inc(rex_sv);
3954 if (OP(scan)==GOSUB) {
3955 startpoint = scan + ARG2L(scan);
3956 ST.close_paren = ARG(scan);
3958 startpoint = rei->program+1;
3961 goto eval_recurse_doit;
3963 case EVAL: /* /(?{A})B/ /(??{A})B/ and /(?(?{A})X|Y)B/ */
3964 if (cur_eval && cur_eval->locinput==locinput) {
3965 if ( ++nochange_depth > max_nochange_depth )
3966 Perl_croak(aTHX_ "EVAL without pos change exceeded limit in regex");
3971 /* execute the code in the {...} */
3973 SV ** const before = SP;
3974 OP_4tree * const oop = PL_op;
3975 COP * const ocurcop = PL_curcop;
3977 char *saved_regeol = PL_regeol;
3980 PL_op = (OP_4tree*)rexi->data->data[n];
3981 DEBUG_STATE_r( PerlIO_printf(Perl_debug_log,
3982 " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) );
3983 PAD_SAVE_LOCAL(old_comppad, (PAD*)rexi->data->data[n + 2]);
3984 PL_regoffs[0].end = PL_reg_magic->mg_len = locinput - PL_bostr;
3987 SV *sv_mrk = get_sv("REGMARK", 1);
3988 sv_setsv(sv_mrk, sv_yes_mark);
3991 CALLRUNOPS(aTHX); /* Scalar context. */
3994 ret = &PL_sv_undef; /* protect against empty (?{}) blocks. */
4001 PAD_RESTORE_LOCAL(old_comppad);
4002 PL_curcop = ocurcop;
4003 PL_regeol = saved_regeol;
4006 sv_setsv(save_scalar(PL_replgv), ret);
4010 if (logical == 2) { /* Postponed subexpression: /(??{...})/ */
4013 /* extract RE object from returned value; compiling if
4019 SV *const sv = SvRV(ret);
4021 if (SvTYPE(sv) == SVt_REGEXP) {
4023 } else if (SvSMAGICAL(sv)) {
4024 mg = mg_find(sv, PERL_MAGIC_qr);
4027 } else if (SvTYPE(ret) == SVt_REGEXP) {
4029 } else if (SvSMAGICAL(ret)) {
4030 if (SvGMAGICAL(ret)) {
4031 /* I don't believe that there is ever qr magic
4033 assert(!mg_find(ret, PERL_MAGIC_qr));
4034 sv_unmagic(ret, PERL_MAGIC_qr);
4037 mg = mg_find(ret, PERL_MAGIC_qr);
4038 /* testing suggests mg only ends up non-NULL for
4039 scalars who were upgraded and compiled in the
4040 else block below. In turn, this is only
4041 triggered in the "postponed utf8 string" tests
4047 rx = (REGEXP *) mg->mg_obj; /*XXX:dmq*/
4051 rx = reg_temp_copy(NULL, rx);
4055 const I32 osize = PL_regsize;
4058 assert (SvUTF8(ret));
4059 } else if (SvUTF8(ret)) {
4060 /* Not doing UTF-8, despite what the SV says. Is
4061 this only if we're trapped in use 'bytes'? */
4062 /* Make a copy of the octet sequence, but without
4063 the flag on, as the compiler now honours the
4064 SvUTF8 flag on ret. */
4066 const char *const p = SvPV(ret, len);
4067 ret = newSVpvn_flags(p, len, SVs_TEMP);
4069 rx = CALLREGCOMP(ret, pm_flags);
4071 & (SVs_TEMP | SVs_PADTMP | SVf_READONLY
4073 /* This isn't a first class regexp. Instead, it's
4074 caching a regexp onto an existing, Perl visible
4076 sv_magic(ret, MUTABLE_SV(rx), PERL_MAGIC_qr, 0, 0);
4081 re = (struct regexp *)SvANY(rx);
4083 RXp_MATCH_COPIED_off(re);
4084 re->subbeg = rex->subbeg;
4085 re->sublen = rex->sublen;
4088 debug_start_match(re_sv, utf8_target, locinput, PL_regeol,
4089 "Matching embedded");
4091 startpoint = rei->program + 1;
4092 ST.close_paren = 0; /* only used for GOSUB */
4093 /* borrowed from regtry */
4094 if (PL_reg_start_tmpl <= re->nparens) {
4095 PL_reg_start_tmpl = re->nparens*3/2 + 3;
4096 if(PL_reg_start_tmp)
4097 Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
4099 Newx(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
4102 eval_recurse_doit: /* Share code with GOSUB below this line */
4103 /* run the pattern returned from (??{...}) */
4104 ST.cp = regcppush(0); /* Save *all* the positions. */
4105 REGCP_SET(ST.lastcp);
4107 PL_regoffs = re->offs; /* essentially NOOP on GOSUB */
4109 /* see regtry, specifically PL_reglast(?:close)?paren is a pointer! (i dont know why) :dmq */
4110 PL_reglastparen = &re->lastparen;
4111 PL_reglastcloseparen = &re->lastcloseparen;
4113 re->lastcloseparen = 0;
4115 PL_reginput = locinput;
4118 /* XXXX This is too dramatic a measure... */
4121 ST.toggle_reg_flags = PL_reg_flags;
4123 PL_reg_flags |= RF_utf8;
4125 PL_reg_flags &= ~RF_utf8;
4126 ST.toggle_reg_flags ^= PL_reg_flags; /* diff of old and new */
4128 ST.prev_rex = rex_sv;
4129 ST.prev_curlyx = cur_curlyx;
4130 SETREX(rex_sv,re_sv);
4135 ST.prev_eval = cur_eval;
4137 /* now continue from first node in postoned RE */
4138 PUSH_YES_STATE_GOTO(EVAL_AB, startpoint);
4141 /* logical is 1, /(?(?{...})X|Y)/ */
4142 sw = cBOOL(SvTRUE(ret));
4147 case EVAL_AB: /* cleanup after a successful (??{A})B */
4148 /* note: this is called twice; first after popping B, then A */
4149 PL_reg_flags ^= ST.toggle_reg_flags;
4150 ReREFCNT_dec(rex_sv);
4151 SETREX(rex_sv,ST.prev_rex);
4152 rex = (struct regexp *)SvANY(rex_sv);
4153 rexi = RXi_GET(rex);
4155 cur_eval = ST.prev_eval;
4156 cur_curlyx = ST.prev_curlyx;
4158 /* rex was changed so update the pointer in PL_reglastparen and PL_reglastcloseparen */
4159 PL_reglastparen = &rex->lastparen;
4160 PL_reglastcloseparen = &rex->lastcloseparen;
4161 /* also update PL_regoffs */
4162 PL_regoffs = rex->offs;
4164 /* XXXX This is too dramatic a measure... */
4166 if ( nochange_depth )
4171 case EVAL_AB_fail: /* unsuccessfully ran A or B in (??{A})B */
4172 /* note: this is called twice; first after popping B, then A */
4173 PL_reg_flags ^= ST.toggle_reg_flags;
4174 ReREFCNT_dec(rex_sv);
4175 SETREX(rex_sv,ST.prev_rex);
4176 rex = (struct regexp *)SvANY(rex_sv);
4177 rexi = RXi_GET(rex);
4178 /* rex was changed so update the pointer in PL_reglastparen and PL_reglastcloseparen */
4179 PL_reglastparen = &rex->lastparen;
4180 PL_reglastcloseparen = &rex->lastcloseparen;
4182 PL_reginput = locinput;
4183 REGCP_UNWIND(ST.lastcp);
4185 cur_eval = ST.prev_eval;
4186 cur_curlyx = ST.prev_curlyx;
4187 /* XXXX This is too dramatic a measure... */
4189 if ( nochange_depth )
4195 n = ARG(scan); /* which paren pair */
4196 PL_reg_start_tmp[n] = locinput;
4202 n = ARG(scan); /* which paren pair */
4203 PL_regoffs[n].start = PL_reg_start_tmp[n] - PL_bostr;
4204 PL_regoffs[n].end = locinput - PL_bostr;
4205 /*if (n > PL_regsize)
4207 if (n > *PL_reglastparen)
4208 *PL_reglastparen = n;
4209 *PL_reglastcloseparen = n;
4210 if (cur_eval && cur_eval->u.eval.close_paren == n) {
4218 cursor && OP(cursor)!=END;
4219 cursor=regnext(cursor))
4221 if ( OP(cursor)==CLOSE ){
4223 if ( n <= lastopen ) {
4225 = PL_reg_start_tmp[n] - PL_bostr;
4226 PL_regoffs[n].end = locinput - PL_bostr;
4227 /*if (n > PL_regsize)
4229 if (n > *PL_reglastparen)
4230 *PL_reglastparen = n;
4231 *PL_reglastcloseparen = n;
4232 if ( n == ARG(scan) || (cur_eval &&
4233 cur_eval->u.eval.close_paren == n))
4242 n = ARG(scan); /* which paren pair */
4243 sw = cBOOL(*PL_reglastparen >= n && PL_regoffs[n].end != -1);
4246 /* reg_check_named_buff_matched returns 0 for no match */
4247 sw = cBOOL(0 < reg_check_named_buff_matched(rex,scan));
4251 sw = (cur_eval && (!n || cur_eval->u.eval.close_paren == n));
4257 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
4259 next = NEXTOPER(NEXTOPER(scan));
4261 next = scan + ARG(scan);
4262 if (OP(next) == IFTHEN) /* Fake one. */
4263 next = NEXTOPER(NEXTOPER(next));
4267 logical = scan->flags;
4270 /*******************************************************************
4272 The CURLYX/WHILEM pair of ops handle the most generic case of the /A*B/
4273 pattern, where A and B are subpatterns. (For simple A, CURLYM or
4274 STAR/PLUS/CURLY/CURLYN are used instead.)
4276 A*B is compiled as <CURLYX><A><WHILEM><B>
4278 On entry to the subpattern, CURLYX is called. This pushes a CURLYX
4279 state, which contains the current count, initialised to -1. It also sets
4280 cur_curlyx to point to this state, with any previous value saved in the
4283 CURLYX then jumps straight to the WHILEM op, rather than executing A,
4284 since the pattern may possibly match zero times (i.e. it's a while {} loop
4285 rather than a do {} while loop).
4287 Each entry to WHILEM represents a successful match of A. The count in the
4288 CURLYX block is incremented, another WHILEM state is pushed, and execution
4289 passes to A or B depending on greediness and the current count.
4291 For example, if matching against the string a1a2a3b (where the aN are
4292 substrings that match /A/), then the match progresses as follows: (the
4293 pushed states are interspersed with the bits of strings matched so far):
4296 <CURLYX cnt=0><WHILEM>
4297 <CURLYX cnt=1><WHILEM> a1 <WHILEM>
4298 <CURLYX cnt=2><WHILEM> a1 <WHILEM> a2 <WHILEM>
4299 <CURLYX cnt=3><WHILEM> a1 <WHILEM> a2 <WHILEM> a3 <WHILEM>
4300 <CURLYX cnt=3><WHILEM> a1 <WHILEM> a2 <WHILEM> a3 <WHILEM> b
4302 (Contrast this with something like CURLYM, which maintains only a single
4306 a1 <CURLYM cnt=1> a2
4307 a1 a2 <CURLYM cnt=2> a3
4308 a1 a2 a3 <CURLYM cnt=3> b
4311 Each WHILEM state block marks a point to backtrack to upon partial failure
4312 of A or B, and also contains some minor state data related to that
4313 iteration. The CURLYX block, pointed to by cur_curlyx, contains the
4314 overall state, such as the count, and pointers to the A and B ops.
4316 This is complicated slightly by nested CURLYX/WHILEM's. Since cur_curlyx
4317 must always point to the *current* CURLYX block, the rules are:
4319 When executing CURLYX, save the old cur_curlyx in the CURLYX state block,
4320 and set cur_curlyx to point the new block.
4322 When popping the CURLYX block after a successful or unsuccessful match,
4323 restore the previous cur_curlyx.
4325 When WHILEM is about to execute B, save the current cur_curlyx, and set it
4326 to the outer one saved in the CURLYX block.
4328 When popping the WHILEM block after a successful or unsuccessful B match,
4329 restore the previous cur_curlyx.
4331 Here's an example for the pattern (AI* BI)*BO
4332 I and O refer to inner and outer, C and W refer to CURLYX and WHILEM:
4335 curlyx backtrack stack
4336 ------ ---------------
4338 CO <CO prev=NULL> <WO>
4339 CI <CO prev=NULL> <WO> <CI prev=CO> <WI> ai
4340 CO <CO prev=NULL> <WO> <CI prev=CO> <WI> ai <WI prev=CI> bi
4341 NULL <CO prev=NULL> <WO> <CI prev=CO> <WI> ai <WI prev=CI> bi <WO prev=CO> bo
4343 At this point the pattern succeeds, and we work back down the stack to
4344 clean up, restoring as we go:
4346 CO <CO prev=NULL> <WO> <CI prev=CO> <WI> ai <WI prev=CI> bi
4347 CI <CO prev=NULL> <WO> <CI prev=CO> <WI> ai
4348 CO <CO prev=NULL> <WO>
4351 *******************************************************************/
4353 #define ST st->u.curlyx
4355 case CURLYX: /* start of /A*B/ (for complex A) */
4357 /* No need to save/restore up to this paren */
4358 I32 parenfloor = scan->flags;
4360 assert(next); /* keep Coverity happy */
4361 if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */
4364 /* XXXX Probably it is better to teach regpush to support
4365 parenfloor > PL_regsize... */
4366 if (parenfloor > (I32)*PL_reglastparen)
4367 parenfloor = *PL_reglastparen; /* Pessimization... */
4369 ST.prev_curlyx= cur_curlyx;
4371 ST.cp = PL_savestack_ix;
4373 /* these fields contain the state of the current curly.
4374 * they are accessed by subsequent WHILEMs */
4375 ST.parenfloor = parenfloor;
4380 ST.count = -1; /* this will be updated by WHILEM */
4381 ST.lastloc = NULL; /* this will be updated by WHILEM */
4383 PL_reginput = locinput;
4384 PUSH_YES_STATE_GOTO(CURLYX_end, PREVOPER(next));
4388 case CURLYX_end: /* just finished matching all of A*B */
4389 cur_curlyx = ST.prev_curlyx;
4393 case CURLYX_end_fail: /* just failed to match all of A*B */
4395 cur_curlyx = ST.prev_curlyx;
4401 #define ST st->u.whilem
4403 case WHILEM: /* just matched an A in /A*B/ (for complex A) */
4405 /* see the discussion above about CURLYX/WHILEM */
4407 int min = ARG1(cur_curlyx->u.curlyx.me);
4408 int max = ARG2(cur_curlyx->u.curlyx.me);
4409 regnode *A = NEXTOPER(cur_curlyx->u.curlyx.me) + EXTRA_STEP_2ARGS;
4411 assert(cur_curlyx); /* keep Coverity happy */
4412 n = ++cur_curlyx->u.curlyx.count; /* how many A's matched */
4413 ST.save_lastloc = cur_curlyx->u.curlyx.lastloc;
4414 ST.cache_offset = 0;
4417 PL_reginput = locinput;
4419 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
4420 "%*s whilem: matched %ld out of %d..%d\n",
4421 REPORT_CODE_OFF+depth*2, "", (long)n, min, max)
4424 /* First just match a string of min A's. */
4427 cur_curlyx->u.curlyx.lastloc = locinput;
4428 PUSH_STATE_GOTO(WHILEM_A_pre, A);
4432 /* If degenerate A matches "", assume A done. */
4434 if (locinput == cur_curlyx->u.curlyx.lastloc) {
4435 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
4436 "%*s whilem: empty match detected, trying continuation...\n",
4437 REPORT_CODE_OFF+depth*2, "")
4439 goto do_whilem_B_max;
4442 /* super-linear cache processing */
4446 if (!PL_reg_maxiter) {
4447 /* start the countdown: Postpone detection until we
4448 * know the match is not *that* much linear. */
4449 PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4);
4450 /* possible overflow for long strings and many CURLYX's */
4451 if (PL_reg_maxiter < 0)
4452 PL_reg_maxiter = I32_MAX;
4453 PL_reg_leftiter = PL_reg_maxiter;
4456 if (PL_reg_leftiter-- == 0) {
4457 /* initialise cache */
4458 const I32 size = (PL_reg_maxiter + 7)/8;
4459 if (PL_reg_poscache) {
4460 if ((I32)PL_reg_poscache_size < size) {
4461 Renew(PL_reg_poscache, size, char);
4462 PL_reg_poscache_size = size;
4464 Zero(PL_reg_poscache, size, char);
4467 PL_reg_poscache_size = size;
4468 Newxz(PL_reg_poscache, size, char);
4470 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
4471 "%swhilem: Detected a super-linear match, switching on caching%s...\n",
4472 PL_colors[4], PL_colors[5])
4476 if (PL_reg_leftiter < 0) {
4477 /* have we already failed at this position? */
4479 offset = (scan->flags & 0xf) - 1
4480 + (locinput - PL_bostr) * (scan->flags>>4);
4481 mask = 1 << (offset % 8);
4483 if (PL_reg_poscache[offset] & mask) {
4484 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
4485 "%*s whilem: (cache) already tried at this position...\n",
4486 REPORT_CODE_OFF+depth*2, "")
4488 sayNO; /* cache records failure */
4490 ST.cache_offset = offset;
4491 ST.cache_mask = mask;
4495 /* Prefer B over A for minimal matching. */
4497 if (cur_curlyx->u.curlyx.minmod) {
4498 ST.save_curlyx = cur_curlyx;
4499 cur_curlyx = cur_curlyx->u.curlyx.prev_curlyx;
4500 ST.cp = regcppush(ST.save_curlyx->u.curlyx.parenfloor);
4501 REGCP_SET(ST.lastcp);
4502 PUSH_YES_STATE_GOTO(WHILEM_B_min, ST.save_curlyx->u.curlyx.B);
4506 /* Prefer A over B for maximal matching. */
4508 if (n < max) { /* More greed allowed? */
4509 ST.cp = regcppush(cur_curlyx->u.curlyx.parenfloor);
4510 cur_curlyx->u.curlyx.lastloc = locinput;
4511 REGCP_SET(ST.lastcp);
4512 PUSH_STATE_GOTO(WHILEM_A_max, A);
4515 goto do_whilem_B_max;
4519 case WHILEM_B_min: /* just matched B in a minimal match */
4520 case WHILEM_B_max: /* just matched B in a maximal match */
4521 cur_curlyx = ST.save_curlyx;
4525 case WHILEM_B_max_fail: /* just failed to match B in a maximal match */
4526 cur_curlyx = ST.save_curlyx;
4527 cur_curlyx->u.curlyx.lastloc = ST.save_lastloc;
4528 cur_curlyx->u.curlyx.count--;
4532 case WHILEM_A_min_fail: /* just failed to match A in a minimal match */
4533 REGCP_UNWIND(ST.lastcp);
4536 case WHILEM_A_pre_fail: /* just failed to match even minimal A */
4537 cur_curlyx->u.curlyx.lastloc = ST.save_lastloc;
4538 cur_curlyx->u.curlyx.count--;
4542 case WHILEM_A_max_fail: /* just failed to match A in a maximal match */
4543 REGCP_UNWIND(ST.lastcp);
4544 regcppop(rex); /* Restore some previous $<digit>s? */
4545 PL_reginput = locinput;
4546 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
4547 "%*s whilem: failed, trying continuation...\n",
4548 REPORT_CODE_OFF+depth*2, "")
4551 if (cur_curlyx->u.curlyx.count >= REG_INFTY
4552 && ckWARN(WARN_REGEXP)
4553 && !(PL_reg_flags & RF_warned))
4555 PL_reg_flags |= RF_warned;
4556 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s limit (%d) exceeded",
4557 "Complex regular subexpression recursion",
4562 ST.save_curlyx = cur_curlyx;
4563 cur_curlyx = cur_curlyx->u.curlyx.prev_curlyx;
4564 PUSH_YES_STATE_GOTO(WHILEM_B_max, ST.save_curlyx->u.curlyx.B);
4567 case WHILEM_B_min_fail: /* just failed to match B in a minimal match */
4568 cur_curlyx = ST.save_curlyx;
4569 REGCP_UNWIND(ST.lastcp);
4572 if (cur_curlyx->u.curlyx.count >= /*max*/ARG2(cur_curlyx->u.curlyx.me)) {
4573 /* Maximum greed exceeded */
4574 if (cur_curlyx->u.curlyx.count >= REG_INFTY
4575 && ckWARN(WARN_REGEXP)
4576 && !(PL_reg_flags & RF_warned))
4578 PL_reg_flags |= RF_warned;
4579 Perl_warner(aTHX_ packWARN(WARN_REGEXP),
4580 "%s limit (%d) exceeded",
4581 "Complex regular subexpression recursion",
4584 cur_curlyx->u.curlyx.count--;
4588 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
4589 "%*s trying longer...\n", REPORT_CODE_OFF+depth*2, "")
4591 /* Try grabbing another A and see if it helps. */
4592 PL_reginput = locinput;
4593 cur_curlyx->u.curlyx.lastloc = locinput;
4594 ST.cp = regcppush(cur_curlyx->u.curlyx.parenfloor);
4595 REGCP_SET(ST.lastcp);
4596 PUSH_STATE_GOTO(WHILEM_A_min,
4597 /*A*/ NEXTOPER(ST.save_curlyx->u.curlyx.me) + EXTRA_STEP_2ARGS);
4601 #define ST st->u.branch
4603 case BRANCHJ: /* /(...|A|...)/ with long next pointer */
4604 next = scan + ARG(scan);
4607 scan = NEXTOPER(scan);
4610 case BRANCH: /* /(...|A|...)/ */
4611 scan = NEXTOPER(scan); /* scan now points to inner node */
4612 ST.lastparen = *PL_reglastparen;
4613 ST.next_branch = next;
4615 PL_reginput = locinput;
4617 /* Now go into the branch */
4619 PUSH_YES_STATE_GOTO(BRANCH_next, scan);
4621 PUSH_STATE_GOTO(BRANCH_next, scan);
4625 PL_reginput = locinput;
4626 sv_yes_mark = st->u.mark.mark_name = scan->flags ? NULL :
4627 MUTABLE_SV(rexi->data->data[ ARG( scan ) ]);
4628 PUSH_STATE_GOTO(CUTGROUP_next,next);
4630 case CUTGROUP_next_fail:
4633 if (st->u.mark.mark_name)
4634 sv_commit = st->u.mark.mark_name;
4640 case BRANCH_next_fail: /* that branch failed; try the next, if any */
4645 REGCP_UNWIND(ST.cp);
4646 for (n = *PL_reglastparen; n > ST.lastparen; n--)
4647 PL_regoffs[n].end = -1;
4648 *PL_reglastparen = n;
4649 /*dmq: *PL_reglastcloseparen = n; */
4650 scan = ST.next_branch;
4651 /* no more branches? */
4652 if (!scan || (OP(scan) != BRANCH && OP(scan) != BRANCHJ)) {
4654 PerlIO_printf( Perl_debug_log,
4655 "%*s %sBRANCH failed...%s\n",
4656 REPORT_CODE_OFF+depth*2, "",
4662 continue; /* execute next BRANCH[J] op */
4670 #define ST st->u.curlym
4672 case CURLYM: /* /A{m,n}B/ where A is fixed-length */
4674 /* This is an optimisation of CURLYX that enables us to push
4675 * only a single backtracking state, no matter how many matches
4676 * there are in {m,n}. It relies on the pattern being constant
4677 * length, with no parens to influence future backrefs
4681 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
4683 /* if paren positive, emulate an OPEN/CLOSE around A */
4685 U32 paren = ST.me->flags;
4686 if (paren > PL_regsize)
4688 if (paren > *PL_reglastparen)
4689 *PL_reglastparen = paren;
4690 scan += NEXT_OFF(scan); /* Skip former OPEN. */
4698 ST.c1 = CHRTEST_UNINIT;
4701 if (!(ST.minmod ? ARG1(ST.me) : ARG2(ST.me))) /* min/max */
4704 curlym_do_A: /* execute the A in /A{m,n}B/ */
4705 PL_reginput = locinput;
4706 PUSH_YES_STATE_GOTO(CURLYM_A, ST.A); /* match A */
4709 case CURLYM_A: /* we've just matched an A */
4710 locinput = st->locinput;
4711 nextchr = UCHARAT(locinput);
4714 /* after first match, determine A's length: u.curlym.alen */
4715 if (ST.count == 1) {
4716 if (PL_reg_match_utf8) {
4718 while (s < PL_reginput) {
4724 ST.alen = PL_reginput - locinput;
4727 ST.count = ST.minmod ? ARG1(ST.me) : ARG2(ST.me);
4730 PerlIO_printf(Perl_debug_log,
4731 "%*s CURLYM now matched %"IVdf" times, len=%"IVdf"...\n",
4732 (int)(REPORT_CODE_OFF+(depth*2)), "",
4733 (IV) ST.count, (IV)ST.alen)
4736 locinput = PL_reginput;
4738 if (cur_eval && cur_eval->u.eval.close_paren &&
4739 cur_eval->u.eval.close_paren == (U32)ST.me->flags)
4743 I32 max = (ST.minmod ? ARG1(ST.me) : ARG2(ST.me));
4744 if ( max == REG_INFTY || ST.count < max )
4745 goto curlym_do_A; /* try to match another A */
4747 goto curlym_do_B; /* try to match B */
4749 case CURLYM_A_fail: /* just failed to match an A */
4750 REGCP_UNWIND(ST.cp);
4752 if (ST.minmod || ST.count < ARG1(ST.me) /* min*/
4753 || (cur_eval && cur_eval->u.eval.close_paren &&
4754 cur_eval->u.eval.close_paren == (U32)ST.me->flags))
4757 curlym_do_B: /* execute the B in /A{m,n}B/ */
4758 PL_reginput = locinput;
4759 if (ST.c1 == CHRTEST_UNINIT) {
4760 /* calculate c1 and c2 for possible match of 1st char
4761 * following curly */
4762 ST.c1 = ST.c2 = CHRTEST_VOID;
4763 if (HAS_TEXT(ST.B) || JUMPABLE(ST.B)) {
4764 regnode *text_node = ST.B;
4765 if (! HAS_TEXT(text_node))
4766 FIND_NEXT_IMPT(text_node);
4769 (HAS_TEXT(text_node) && PL_regkind[OP(text_node)] == EXACT)
4771 But the former is redundant in light of the latter.
4773 if this changes back then the macro for
4774 IS_TEXT and friends need to change.
4776 if (PL_regkind[OP(text_node)] == EXACT)
4779 ST.c1 = (U8)*STRING(text_node);
4781 (IS_TEXTF(text_node))
4783 : (IS_TEXTFL(text_node))
4784 ? PL_fold_locale[ST.c1]
4791 PerlIO_printf(Perl_debug_log,
4792 "%*s CURLYM trying tail with matches=%"IVdf"...\n",
4793 (int)(REPORT_CODE_OFF+(depth*2)),
4796 if (ST.c1 != CHRTEST_VOID
4797 && UCHARAT(PL_reginput) != ST.c1
4798 && UCHARAT(PL_reginput) != ST.c2)
4800 /* simulate B failing */
4802 PerlIO_printf(Perl_debug_log,
4803 "%*s CURLYM Fast bail c1=%"IVdf" c2=%"IVdf"\n",
4804 (int)(REPORT_CODE_OFF+(depth*2)),"",
4807 state_num = CURLYM_B_fail;
4808 goto reenter_switch;
4812 /* mark current A as captured */
4813 I32 paren = ST.me->flags;
4815 PL_regoffs[paren].start
4816 = HOPc(PL_reginput, -ST.alen) - PL_bostr;
4817 PL_regoffs[paren].end = PL_reginput - PL_bostr;
4818 /*dmq: *PL_reglastcloseparen = paren; */
4821 PL_regoffs[paren].end = -1;
4822 if (cur_eval && cur_eval->u.eval.close_paren &&
4823 cur_eval->u.eval.close_paren == (U32)ST.me->flags)
4832 PUSH_STATE_GOTO(CURLYM_B, ST.B); /* match B */
4835 case CURLYM_B_fail: /* just failed to match a B */
4836 REGCP_UNWIND(ST.cp);
4838 I32 max = ARG2(ST.me);
4839 if (max != REG_INFTY && ST.count == max)
4841 goto curlym_do_A; /* try to match a further A */
4843 /* backtrack one A */
4844 if (ST.count == ARG1(ST.me) /* min */)
4847 locinput = HOPc(locinput, -ST.alen);
4848 goto curlym_do_B; /* try to match B */
4851 #define ST st->u.curly
4853 #define CURLY_SETPAREN(paren, success) \
4856 PL_regoffs[paren].start = HOPc(locinput, -1) - PL_bostr; \
4857 PL_regoffs[paren].end = locinput - PL_bostr; \
4858 *PL_reglastcloseparen = paren; \
4861 PL_regoffs[paren].end = -1; \
4864 case STAR: /* /A*B/ where A is width 1 */
4868 scan = NEXTOPER(scan);
4870 case PLUS: /* /A+B/ where A is width 1 */
4874 scan = NEXTOPER(scan);
4876 case CURLYN: /* /(A){m,n}B/ where A is width 1 */
4877 ST.paren = scan->flags; /* Which paren to set */
4878 if (ST.paren > PL_regsize)
4879 PL_regsize = ST.paren;
4880 if (ST.paren > *PL_reglastparen)
4881 *PL_reglastparen = ST.paren;
4882 ST.min = ARG1(scan); /* min to match */
4883 ST.max = ARG2(scan); /* max to match */
4884 if (cur_eval && cur_eval->u.eval.close_paren &&
4885 cur_eval->u.eval.close_paren == (U32)ST.paren) {
4889 scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE);
4891 case CURLY: /* /A{m,n}B/ where A is width 1 */
4893 ST.min = ARG1(scan); /* min to match */
4894 ST.max = ARG2(scan); /* max to match */
4895 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
4898 * Lookahead to avoid useless match attempts
4899 * when we know what character comes next.
4901 * Used to only do .*x and .*?x, but now it allows
4902 * for )'s, ('s and (?{ ... })'s to be in the way
4903 * of the quantifier and the EXACT-like node. -- japhy
4906 if (ST.min > ST.max) /* XXX make this a compile-time check? */
4908 if (HAS_TEXT(next) || JUMPABLE(next)) {
4910 regnode *text_node = next;
4912 if (! HAS_TEXT(text_node))
4913 FIND_NEXT_IMPT(text_node);
4915 if (! HAS_TEXT(text_node))
4916 ST.c1 = ST.c2 = CHRTEST_VOID;
4918 if ( PL_regkind[OP(text_node)] != EXACT ) {
4919 ST.c1 = ST.c2 = CHRTEST_VOID;
4920 goto assume_ok_easy;
4923 s = (U8*)STRING(text_node);
4925 /* Currently we only get here when
4927 PL_rekind[OP(text_node)] == EXACT
4929 if this changes back then the macro for IS_TEXT and
4930 friends need to change. */
4933 if (IS_TEXTF(text_node))
4934 ST.c2 = PL_fold[ST.c1];
4935 else if (IS_TEXTFL(text_node))
4936 ST.c2 = PL_fold_locale[ST.c1];
4938 else { /* UTF_PATTERN */
4939 if (IS_TEXTF(text_node)) {
4940 STRLEN ulen1, ulen2;
4941 U8 tmpbuf1[UTF8_MAXBYTES_CASE+1];
4942 U8 tmpbuf2[UTF8_MAXBYTES_CASE+1];
4944 to_utf8_lower((U8*)s, tmpbuf1, &ulen1);
4945 to_utf8_upper((U8*)s, tmpbuf2, &ulen2);
4947 ST.c1 = utf8n_to_uvchr(tmpbuf1, UTF8_MAXLEN, 0,
4949 0 : UTF8_ALLOW_ANY);
4950 ST.c2 = utf8n_to_uvchr(tmpbuf2, UTF8_MAXLEN, 0,
4952 0 : UTF8_ALLOW_ANY);
4954 ST.c1 = utf8n_to_uvuni(tmpbuf1, UTF8_MAXBYTES, 0,
4956 ST.c2 = utf8n_to_uvuni(tmpbuf2, UTF8_MAXBYTES, 0,
4961 ST.c2 = ST.c1 = utf8n_to_uvchr(s, UTF8_MAXBYTES, 0,
4968 ST.c1 = ST.c2 = CHRTEST_VOID;
4973 PL_reginput = locinput;
4976 if (ST.min && regrepeat(rex, ST.A, ST.min, depth) < ST.min)
4979 locinput = PL_reginput;
4981 if (ST.c1 == CHRTEST_VOID)
4982 goto curly_try_B_min;
4984 ST.oldloc = locinput;
4986 /* set ST.maxpos to the furthest point along the
4987 * string that could possibly match */
4988 if (ST.max == REG_INFTY) {
4989 ST.maxpos = PL_regeol - 1;
4991 while (UTF8_IS_CONTINUATION(*(U8*)ST.maxpos))
4994 else if (utf8_target) {
4995 int m = ST.max - ST.min;
4996 for (ST.maxpos = locinput;
4997 m >0 && ST.maxpos + UTF8SKIP(ST.maxpos) <= PL_regeol; m--)
4998 ST.maxpos += UTF8SKIP(ST.maxpos);
5001 ST.maxpos = locinput + ST.max - ST.min;
5002 if (ST.maxpos >= PL_regeol)
5003 ST.maxpos = PL_regeol - 1;
5005 goto curly_try_B_min_known;
5009 ST.count = regrepeat(rex, ST.A, ST.max, depth);
5010 locinput = PL_reginput;
5011 if (ST.count < ST.min)
5013 if ((ST.count > ST.min)
5014 && (PL_regkind[OP(ST.B)] == EOL) && (OP(ST.B) != MEOL))
5016 /* A{m,n} must come at the end of the string, there's
5017 * no point in backing off ... */
5019 /* ...except that $ and \Z can match before *and* after
5020 newline at the end. Consider "\n\n" =~ /\n+\Z\n/.
5021 We may back off by one in this case. */
5022 if (UCHARAT(PL_reginput - 1) == '\n' && OP(ST.B) != EOS)
5026 goto curly_try_B_max;
5031 case CURLY_B_min_known_fail:
5032 /* failed to find B in a non-greedy match where c1,c2 valid */
5033 if (ST.paren && ST.count)
5034 PL_regoffs[ST.paren].end = -1;
5036 PL_reginput = locinput; /* Could be reset... */
5037 REGCP_UNWIND(ST.cp);
5038 /* Couldn't or didn't -- move forward. */
5039 ST.oldloc = locinput;
5041 locinput += UTF8SKIP(locinput);
5045 curly_try_B_min_known:
5046 /* find the next place where 'B' could work, then call B */
5050 n = (ST.oldloc == locinput) ? 0 : 1;
5051 if (ST.c1 == ST.c2) {
5053 /* set n to utf8_distance(oldloc, locinput) */
5054 while (locinput <= ST.maxpos &&
5055 utf8n_to_uvchr((U8*)locinput,
5056 UTF8_MAXBYTES, &len,
5057 uniflags) != (UV)ST.c1) {
5063 /* set n to utf8_distance(oldloc, locinput) */
5064 while (locinput <= ST.maxpos) {
5066 const UV c = utf8n_to_uvchr((U8*)locinput,
5067 UTF8_MAXBYTES, &len,
5069 if (c == (UV)ST.c1 || c == (UV)ST.c2)
5077 if (ST.c1 == ST.c2) {
5078 while (locinput <= ST.maxpos &&
5079 UCHARAT(locinput) != ST.c1)
5083 while (locinput <= ST.maxpos
5084 && UCHARAT(locinput) != ST.c1
5085 && UCHARAT(locinput) != ST.c2)
5088 n = locinput - ST.oldloc;
5090 if (locinput > ST.maxpos)
5092 /* PL_reginput == oldloc now */
5095 if (regrepeat(rex, ST.A, n, depth) < n)
5098 PL_reginput = locinput;
5099 CURLY_SETPAREN(ST.paren, ST.count);
5100 if (cur_eval && cur_eval->u.eval.close_paren &&
5101 cur_eval->u.eval.close_paren == (U32)ST.paren) {
5104 PUSH_STATE_GOTO(CURLY_B_min_known, ST.B);
5109 case CURLY_B_min_fail:
5110 /* failed to find B in a non-greedy match where c1,c2 invalid */
5111 if (ST.paren && ST.count)
5112 PL_regoffs[ST.paren].end = -1;
5114 REGCP_UNWIND(ST.cp);
5115 /* failed -- move forward one */
5116 PL_reginput = locinput;
5117 if (regrepeat(rex, ST.A, 1, depth)) {
5119 locinput = PL_reginput;
5120 if (ST.count <= ST.max || (ST.max == REG_INFTY &&
5121 ST.count > 0)) /* count overflow ? */
5124 CURLY_SETPAREN(ST.paren, ST.count);
5125 if (cur_eval && cur_eval->u.eval.close_paren &&
5126 cur_eval->u.eval.close_paren == (U32)ST.paren) {
5129 PUSH_STATE_GOTO(CURLY_B_min, ST.B);
5137 /* a successful greedy match: now try to match B */
5138 if (cur_eval && cur_eval->u.eval.close_paren &&
5139 cur_eval->u.eval.close_paren == (U32)ST.paren) {
5144 if (ST.c1 != CHRTEST_VOID)
5145 c = utf8_target ? utf8n_to_uvchr((U8*)PL_reginput,
5146 UTF8_MAXBYTES, 0, uniflags)
5147 : (UV) UCHARAT(PL_reginput);
5148 /* If it could work, try it. */
5149 if (ST.c1 == CHRTEST_VOID || c == (UV)ST.c1 || c == (UV)ST.c2) {
5150 CURLY_SETPAREN(ST.paren, ST.count);
5151 PUSH_STATE_GOTO(CURLY_B_max, ST.B);
5156 case CURLY_B_max_fail:
5157 /* failed to find B in a greedy match */
5158 if (ST.paren && ST.count)
5159 PL_regoffs[ST.paren].end = -1;
5161 REGCP_UNWIND(ST.cp);
5163 if (--ST.count < ST.min)
5165 PL_reginput = locinput = HOPc(locinput, -1);
5166 goto curly_try_B_max;
5173 /* we've just finished A in /(??{A})B/; now continue with B */
5175 st->u.eval.toggle_reg_flags
5176 = cur_eval->u.eval.toggle_reg_flags;
5177 PL_reg_flags ^= st->u.eval.toggle_reg_flags;
5179 st->u.eval.prev_rex = rex_sv; /* inner */
5180 SETREX(rex_sv,cur_eval->u.eval.prev_rex);
5181 rex = (struct regexp *)SvANY(rex_sv);
5182 rexi = RXi_GET(rex);
5183 cur_curlyx = cur_eval->u.eval.prev_curlyx;
5184 ReREFCNT_inc(rex_sv);
5185 st->u.eval.cp = regcppush(0); /* Save *all* the positions. */
5187 /* rex was changed so update the pointer in PL_reglastparen and PL_reglastcloseparen */
5188 PL_reglastparen = &rex->lastparen;
5189 PL_reglastcloseparen = &rex->lastcloseparen;
5191 REGCP_SET(st->u.eval.lastcp);
5192 PL_reginput = locinput;
5194 /* Restore parens of the outer rex without popping the
5196 tmpix = PL_savestack_ix;
5197 PL_savestack_ix = cur_eval->u.eval.lastcp;
5199 PL_savestack_ix = tmpix;
5201 st->u.eval.prev_eval = cur_eval;
5202 cur_eval = cur_eval->u.eval.prev_eval;
5204 PerlIO_printf(Perl_debug_log, "%*s EVAL trying tail ... %"UVxf"\n",
5205 REPORT_CODE_OFF+depth*2, "",PTR2UV(cur_eval)););
5206 if ( nochange_depth )
5209 PUSH_YES_STATE_GOTO(EVAL_AB,
5210 st->u.eval.prev_eval->u.eval.B); /* match B */
5213 if (locinput < reginfo->till) {
5214 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
5215 "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n",
5217 (long)(locinput - PL_reg_starttry),
5218 (long)(reginfo->till - PL_reg_starttry),
5221 sayNO_SILENT; /* Cannot match: too short. */
5223 PL_reginput = locinput; /* put where regtry can find it */
5224 sayYES; /* Success! */
5226 case SUCCEED: /* successful SUSPEND/UNLESSM/IFMATCH/CURLYM */
5228 PerlIO_printf(Perl_debug_log,
5229 "%*s %ssubpattern success...%s\n",
5230 REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5]));
5231 PL_reginput = locinput; /* put where regtry can find it */
5232 sayYES; /* Success! */
5235 #define ST st->u.ifmatch
5237 case SUSPEND: /* (?>A) */
5239 PL_reginput = locinput;
5242 case UNLESSM: /* -ve lookaround: (?!A), or with flags, (?<!A) */
5244 goto ifmatch_trivial_fail_test;
5246 case IFMATCH: /* +ve lookaround: (?=A), or with flags, (?<=A) */
5248 ifmatch_trivial_fail_test:
5250 char * const s = HOPBACKc(locinput, scan->flags);
5255 sw = 1 - cBOOL(ST.wanted);
5259 next = scan + ARG(scan);
5267 PL_reginput = locinput;
5271 ST.logical = logical;
5272 logical = 0; /* XXX: reset state of logical once it has been saved into ST */
5274 /* execute body of (?...A) */
5275 PUSH_YES_STATE_GOTO(IFMATCH_A, NEXTOPER(NEXTOPER(scan)));
5278 case IFMATCH_A_fail: /* body of (?...A) failed */
5279 ST.wanted = !ST.wanted;
5282 case IFMATCH_A: /* body of (?...A) succeeded */
5284 sw = cBOOL(ST.wanted);
5286 else if (!ST.wanted)
5289 if (OP(ST.me) == SUSPEND)
5290 locinput = PL_reginput;
5292 locinput = PL_reginput = st->locinput;
5293 nextchr = UCHARAT(locinput);
5295 scan = ST.me + ARG(ST.me);
5298 continue; /* execute B */
5303 next = scan + ARG(scan);
5308 reginfo->cutpoint = PL_regeol;
5311 PL_reginput = locinput;
5313 sv_yes_mark = sv_commit = MUTABLE_SV(rexi->data->data[ ARG( scan ) ]);
5314 PUSH_STATE_GOTO(COMMIT_next,next);
5316 case COMMIT_next_fail:
5323 #define ST st->u.mark
5325 ST.prev_mark = mark_state;
5326 ST.mark_name = sv_commit = sv_yes_mark
5327 = MUTABLE_SV(rexi->data->data[ ARG( scan ) ]);
5329 ST.mark_loc = PL_reginput = locinput;
5330 PUSH_YES_STATE_GOTO(MARKPOINT_next,next);
5332 case MARKPOINT_next:
5333 mark_state = ST.prev_mark;
5336 case MARKPOINT_next_fail:
5337 if (popmark && sv_eq(ST.mark_name,popmark))
5339 if (ST.mark_loc > startpoint)
5340 reginfo->cutpoint = HOPBACKc(ST.mark_loc, 1);
5341 popmark = NULL; /* we found our mark */
5342 sv_commit = ST.mark_name;
5345 PerlIO_printf(Perl_debug_log,
5346 "%*s %ssetting cutpoint to mark:%"SVf"...%s\n",
5347 REPORT_CODE_OFF+depth*2, "",
5348 PL_colors[4], SVfARG(sv_commit), PL_colors[5]);
5351 mark_state = ST.prev_mark;
5352 sv_yes_mark = mark_state ?
5353 mark_state->u.mark.mark_name : NULL;
5357 PL_reginput = locinput;
5359 /* (*SKIP) : if we fail we cut here*/
5360 ST.mark_name = NULL;
5361 ST.mark_loc = locinput;
5362 PUSH_STATE_GOTO(SKIP_next,next);
5364 /* (*SKIP:NAME) : if there is a (*MARK:NAME) fail where it was,
5365 otherwise do nothing. Meaning we need to scan
5367 regmatch_state *cur = mark_state;
5368 SV *find = MUTABLE_SV(rexi->data->data[ ARG( scan ) ]);
5371 if ( sv_eq( cur->u.mark.mark_name,
5374 ST.mark_name = find;
5375 PUSH_STATE_GOTO( SKIP_next, next );
5377 cur = cur->u.mark.prev_mark;
5380 /* Didn't find our (*MARK:NAME) so ignore this (*SKIP:NAME) */
5382 case SKIP_next_fail:
5384 /* (*CUT:NAME) - Set up to search for the name as we
5385 collapse the stack*/
5386 popmark = ST.mark_name;
5388 /* (*CUT) - No name, we cut here.*/
5389 if (ST.mark_loc > startpoint)
5390 reginfo->cutpoint = HOPBACKc(ST.mark_loc, 1);
5391 /* but we set sv_commit to latest mark_name if there
5392 is one so they can test to see how things lead to this
5395 sv_commit=mark_state->u.mark.mark_name;
5403 if ( n == (U32)what_len_TRICKYFOLD(locinput,utf8_target,ln) ) {
5405 } else if ( 0xDF == n && !utf8_target && !UTF_PATTERN ) {
5408 U8 folded[UTF8_MAXBYTES_CASE+1];
5410 const char * const l = locinput;
5411 char *e = PL_regeol;
5412 to_uni_fold(n, folded, &foldlen);
5414 if (! foldEQ_utf8((const char*) folded, 0, foldlen, 1,
5415 l, &e, 0, utf8_target)) {
5420 nextchr = UCHARAT(locinput);
5423 if ((n=is_LNBREAK(locinput,utf8_target))) {
5425 nextchr = UCHARAT(locinput);
5430 #define CASE_CLASS(nAmE) \
5432 if ((n=is_##nAmE(locinput,utf8_target))) { \
5434 nextchr = UCHARAT(locinput); \
5439 if ((n=is_##nAmE(locinput,utf8_target))) { \
5442 locinput += UTF8SKIP(locinput); \
5443 nextchr = UCHARAT(locinput); \
5448 CASE_CLASS(HORIZWS);
5452 PerlIO_printf(Perl_error_log, "%"UVxf" %d\n",
5453 PTR2UV(scan), OP(scan));
5454 Perl_croak(aTHX_ "regexp memory corruption");
5458 /* switch break jumps here */
5459 scan = next; /* prepare to execute the next op and ... */
5460 continue; /* ... jump back to the top, reusing st */
5464 /* push a state that backtracks on success */
5465 st->u.yes.prev_yes_state = yes_state;
5469 /* push a new regex state, then continue at scan */
5471 regmatch_state *newst;
5474 regmatch_state *cur = st;
5475 regmatch_state *curyes = yes_state;
5477 regmatch_slab *slab = PL_regmatch_slab;
5478 for (;curd > -1;cur--,curd--) {
5479 if (cur < SLAB_FIRST(slab)) {
5481 cur = SLAB_LAST(slab);
5483 PerlIO_printf(Perl_error_log, "%*s#%-3d %-10s %s\n",
5484 REPORT_CODE_OFF + 2 + depth * 2,"",
5485 curd, PL_reg_name[cur->resume_state],
5486 (curyes == cur) ? "yes" : ""
5489 curyes = cur->u.yes.prev_yes_state;
5492 DEBUG_STATE_pp("push")
5495 st->locinput = locinput;
5497 if (newst > SLAB_LAST(PL_regmatch_slab))
5498 newst = S_push_slab(aTHX);
5499 PL_regmatch_state = newst;
5501 locinput = PL_reginput;
5502 nextchr = UCHARAT(locinput);
5510 * We get here only if there's trouble -- normally "case END" is
5511 * the terminating point.
5513 Perl_croak(aTHX_ "corrupted regexp pointers");
5519 /* we have successfully completed a subexpression, but we must now
5520 * pop to the state marked by yes_state and continue from there */
5521 assert(st != yes_state);
5523 while (st != yes_state) {
5525 if (st < SLAB_FIRST(PL_regmatch_slab)) {
5526 PL_regmatch_slab = PL_regmatch_slab->prev;
5527 st = SLAB_LAST(PL_regmatch_slab);
5531 DEBUG_STATE_pp("pop (no final)");
5533 DEBUG_STATE_pp("pop (yes)");
5539 while (yes_state < SLAB_FIRST(PL_regmatch_slab)
5540 || yes_state > SLAB_LAST(PL_regmatch_slab))
5542 /* not in this slab, pop slab */
5543 depth -= (st - SLAB_FIRST(PL_regmatch_slab) + 1);
5544 PL_regmatch_slab = PL_regmatch_slab->prev;
5545 st = SLAB_LAST(PL_regmatch_slab);
5547 depth -= (st - yes_state);
5550 yes_state = st->u.yes.prev_yes_state;
5551 PL_regmatch_state = st;
5554 locinput= st->locinput;
5555 nextchr = UCHARAT(locinput);
5557 state_num = st->resume_state + no_final;
5558 goto reenter_switch;
5561 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n",
5562 PL_colors[4], PL_colors[5]));
5564 if (PL_reg_eval_set) {
5565 /* each successfully executed (?{...}) block does the equivalent of
5566 * local $^R = do {...}
5567 * When popping the save stack, all these locals would be undone;
5568 * bypass this by setting the outermost saved $^R to the latest
5570 if (oreplsv != GvSV(PL_replgv))
5571 sv_setsv(oreplsv, GvSV(PL_replgv));
5578 PerlIO_printf(Perl_debug_log,
5579 "%*s %sfailed...%s\n",
5580 REPORT_CODE_OFF+depth*2, "",
5581 PL_colors[4], PL_colors[5])
5593 /* there's a previous state to backtrack to */
5595 if (st < SLAB_FIRST(PL_regmatch_slab)) {
5596 PL_regmatch_slab = PL_regmatch_slab->prev;
5597 st = SLAB_LAST(PL_regmatch_slab);
5599 PL_regmatch_state = st;
5600 locinput= st->locinput;
5601 nextchr = UCHARAT(locinput);
5603 DEBUG_STATE_pp("pop");
5605 if (yes_state == st)
5606 yes_state = st->u.yes.prev_yes_state;
5608 state_num = st->resume_state + 1; /* failure = success + 1 */
5609 goto reenter_switch;
5614 if (rex->intflags & PREGf_VERBARG_SEEN) {
5615 SV *sv_err = get_sv("REGERROR", 1);
5616 SV *sv_mrk = get_sv("REGMARK", 1);
5618 sv_commit = &PL_sv_no;
5620 sv_yes_mark = &PL_sv_yes;
5623 sv_commit = &PL_sv_yes;
5624 sv_yes_mark = &PL_sv_no;
5626 sv_setsv(sv_err, sv_commit);
5627 sv_setsv(sv_mrk, sv_yes_mark);
5630 /* clean up; in particular, free all slabs above current one */
5631 LEAVE_SCOPE(oldsave);
5637 - regrepeat - repeatedly match something simple, report how many
5640 * [This routine now assumes that it will only match on things of length 1.
5641 * That was true before, but now we assume scan - reginput is the count,
5642 * rather than incrementing count on every character. [Er, except utf8.]]
5645 S_regrepeat(pTHX_ const regexp *prog, const regnode *p, I32 max, int depth)
5648 register char *scan;
5650 register char *loceol = PL_regeol;
5651 register I32 hardcount = 0;
5652 register bool utf8_target = PL_reg_match_utf8;
5654 PERL_UNUSED_ARG(depth);
5657 PERL_ARGS_ASSERT_REGREPEAT;
5660 if (max == REG_INFTY)
5662 else if (max < loceol - scan)
5663 loceol = scan + max;
5668 while (scan < loceol && hardcount < max && *scan != '\n') {
5669 scan += UTF8SKIP(scan);
5673 while (scan < loceol && *scan != '\n')
5680 while (scan < loceol && hardcount < max) {
5681 scan += UTF8SKIP(scan);
5691 case EXACT: /* length of string is 1 */
5693 while (scan < loceol && UCHARAT(scan) == c)
5696 case EXACTF: /* length of string is 1 */
5698 while (scan < loceol &&
5699 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold[c]))
5702 case EXACTFL: /* length of string is 1 */
5703 PL_reg_flags |= RF_tainted;
5705 while (scan < loceol &&
5706 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold_locale[c]))
5712 while (hardcount < max && scan < loceol &&
5713 reginclass(prog, p, (U8*)scan, 0, utf8_target)) {
5714 scan += UTF8SKIP(scan);
5718 while (scan < loceol && REGINCLASS(prog, p, (U8*)scan))
5725 LOAD_UTF8_CHARCLASS_ALNUM();
5726 while (hardcount < max && scan < loceol &&
5727 swash_fetch(PL_utf8_alnum, (U8*)scan, utf8_target)) {
5728 scan += UTF8SKIP(scan);
5732 while (scan < loceol && isALNUM(*scan))
5737 PL_reg_flags |= RF_tainted;
5740 while (hardcount < max && scan < loceol &&
5741 isALNUM_LC_utf8((U8*)scan)) {
5742 scan += UTF8SKIP(scan);
5746 while (scan < loceol && isALNUM_LC(*scan))
5753 LOAD_UTF8_CHARCLASS_ALNUM();
5754 while (hardcount < max && scan < loceol &&
5755 !swash_fetch(PL_utf8_alnum, (U8*)scan, utf8_target)) {
5756 scan += UTF8SKIP(scan);
5760 while (scan < loceol && !isALNUM(*scan))
5765 PL_reg_flags |= RF_tainted;
5768 while (hardcount < max && scan < loceol &&
5769 !isALNUM_LC_utf8((U8*)scan)) {
5770 scan += UTF8SKIP(scan);
5774 while (scan < loceol && !isALNUM_LC(*scan))
5781 LOAD_UTF8_CHARCLASS_SPACE();
5782 while (hardcount < max && scan < loceol &&
5784 swash_fetch(PL_utf8_space,(U8*)scan, utf8_target))) {
5785 scan += UTF8SKIP(scan);
5789 while (scan < loceol && isSPACE(*scan))
5794 PL_reg_flags |= RF_tainted;
5797 while (hardcount < max && scan < loceol &&
5798 (*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
5799 scan += UTF8SKIP(scan);
5803 while (scan < loceol && isSPACE_LC(*scan))
5810 LOAD_UTF8_CHARCLASS_SPACE();
5811 while (hardcount < max && scan < loceol &&
5813 swash_fetch(PL_utf8_space,(U8*)scan, utf8_target))) {
5814 scan += UTF8SKIP(scan);
5818 while (scan < loceol && !isSPACE(*scan))
5823 PL_reg_flags |= RF_tainted;
5826 while (hardcount < max && scan < loceol &&
5827 !(*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
5828 scan += UTF8SKIP(scan);
5832 while (scan < loceol && !isSPACE_LC(*scan))
5839 LOAD_UTF8_CHARCLASS_DIGIT();
5840 while (hardcount < max && scan < loceol &&
5841 swash_fetch(PL_utf8_digit, (U8*)scan, utf8_target)) {
5842 scan += UTF8SKIP(scan);
5846 while (scan < loceol && isDIGIT(*scan))
5853 LOAD_UTF8_CHARCLASS_DIGIT();
5854 while (hardcount < max && scan < loceol &&
5855 !swash_fetch(PL_utf8_digit, (U8*)scan, utf8_target)) {
5856 scan += UTF8SKIP(scan);
5860 while (scan < loceol && !isDIGIT(*scan))
5866 while (hardcount < max && scan < loceol && (c=is_LNBREAK_utf8(scan))) {
5872 LNBREAK can match two latin chars, which is ok,
5873 because we have a null terminated string, but we
5874 have to use hardcount in this situation
5876 while (scan < loceol && (c=is_LNBREAK_latin1(scan))) {
5885 while (hardcount < max && scan < loceol && (c=is_HORIZWS_utf8(scan))) {
5890 while (scan < loceol && is_HORIZWS_latin1(scan))
5897 while (hardcount < max && scan < loceol && !is_HORIZWS_utf8(scan)) {
5898 scan += UTF8SKIP(scan);
5902 while (scan < loceol && !is_HORIZWS_latin1(scan))
5910 while (hardcount < max && scan < loceol && (c=is_VERTWS_utf8(scan))) {
5915 while (scan < loceol && is_VERTWS_latin1(scan))
5923 while (hardcount < max && scan < loceol && !is_VERTWS_utf8(scan)) {
5924 scan += UTF8SKIP(scan);
5928 while (scan < loceol && !is_VERTWS_latin1(scan))
5934 default: /* Called on something of 0 width. */
5935 break; /* So match right here or not at all. */
5941 c = scan - PL_reginput;
5945 GET_RE_DEBUG_FLAGS_DECL;
5947 SV * const prop = sv_newmortal();
5948 regprop(prog, prop, p);
5949 PerlIO_printf(Perl_debug_log,
5950 "%*s %s can match %"IVdf" times out of %"IVdf"...\n",
5951 REPORT_CODE_OFF + depth*2, "", SvPVX_const(prop),(IV)c,(IV)max);
5959 #if !defined(PERL_IN_XSUB_RE) || defined(PLUGGABLE_RE_EXTENSION)
5961 - regclass_swash - prepare the utf8 swash
5965 Perl_regclass_swash(pTHX_ const regexp *prog, register const regnode* node, bool doinit, SV** listsvp, SV **altsvp)
5971 RXi_GET_DECL(prog,progi);
5972 const struct reg_data * const data = prog ? progi->data : NULL;
5974 PERL_ARGS_ASSERT_REGCLASS_SWASH;
5976 if (data && data->count) {
5977 const U32 n = ARG(node);
5979 if (data->what[n] == 's') {
5980 SV * const rv = MUTABLE_SV(data->data[n]);
5981 AV * const av = MUTABLE_AV(SvRV(rv));
5982 SV **const ary = AvARRAY(av);
5985 /* See the end of regcomp.c:S_regclass() for
5986 * documentation of these array elements. */
5989 a = SvROK(ary[1]) ? &ary[1] : NULL;
5990 b = SvTYPE(ary[2]) == SVt_PVAV ? &ary[2] : NULL;
5994 else if (si && doinit) {
5995 sw = swash_init("utf8", "", si, 1, 0);
5996 (void)av_store(av, 1, sw);
6013 - reginclass - determine if a character falls into a character class
6015 The n is the ANYOF regnode, the p is the target string, lenp
6016 is pointer to the maximum length of how far to go in the p
6017 (if the lenp is zero, UTF8SKIP(p) is used),
6018 utf8_target tells whether the target string is in UTF-8.
6023 S_reginclass(pTHX_ const regexp *prog, register const regnode *n, register const U8* p, STRLEN* lenp, register bool utf8_target)
6026 const char flags = ANYOF_FLAGS(n);
6032 PERL_ARGS_ASSERT_REGINCLASS;
6034 if (utf8_target && !UTF8_IS_INVARIANT(c)) {
6035 c = utf8n_to_uvchr(p, UTF8_MAXBYTES, &len,
6036 (UTF8_ALLOW_DEFAULT & UTF8_ALLOW_ANYUV)
6037 | UTF8_ALLOW_FFFF | UTF8_CHECK_ONLY);
6038 /* see [perl #37836] for UTF8_ALLOW_ANYUV; [perl #38293] for
6039 * UTF8_ALLOW_FFFF */
6040 if (len == (STRLEN)-1)
6041 Perl_croak(aTHX_ "Malformed UTF-8 character (fatal)");
6044 plen = lenp ? *lenp : UNISKIP(NATIVE_TO_UNI(c));
6045 if (utf8_target || (flags & ANYOF_UNICODE)) {
6048 if (utf8_target && !ANYOF_RUNTIME(n)) {
6049 if (len != (STRLEN)-1 && c < 256 && ANYOF_BITMAP_TEST(n, c))
6052 if (!match && utf8_target && (flags & ANYOF_UNICODE_ALL) && c >= 256)
6056 SV * const sw = regclass_swash(prog, n, TRUE, 0, (SV**)&av);
6064 utf8_p = bytes_to_utf8(p, &len);
6066 if (swash_fetch(sw, utf8_p, 1))
6068 else if (flags & ANYOF_FOLD) {
6069 if (!match && lenp && av) {
6071 for (i = 0; i <= av_len(av); i++) {
6072 SV* const sv = *av_fetch(av, i, FALSE);
6074 const char * const s = SvPV_const(sv, len);
6075 if (len <= plen && memEQ(s, (char*)utf8_p, len)) {
6083 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
6086 to_utf8_fold(utf8_p, tmpbuf, &tmplen);
6087 if (swash_fetch(sw, tmpbuf, 1))
6092 /* If we allocated a string above, free it */
6093 if (! utf8_target) Safefree(utf8_p);
6096 if (match && lenp && *lenp == 0)
6097 *lenp = UNISKIP(NATIVE_TO_UNI(c));
6099 if (!match && c < 256) {
6100 if (ANYOF_BITMAP_TEST(n, c))
6102 else if (flags & ANYOF_FOLD) {
6105 if (flags & ANYOF_LOCALE) {
6106 PL_reg_flags |= RF_tainted;
6107 f = PL_fold_locale[c];
6111 if (f != c && ANYOF_BITMAP_TEST(n, f))
6115 if (!match && (flags & ANYOF_CLASS)) {
6116 PL_reg_flags |= RF_tainted;
6118 (ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) ||
6119 (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) ||
6120 (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) ||
6121 (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) ||
6122 (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) ||
6123 (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) ||
6124 (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) ||
6125 (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) ||
6126 (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) ||
6127 (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) ||
6128 (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII(c)) ||
6129 (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII(c)) ||
6130 (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) ||
6131 (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) ||
6132 (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) ||
6133 (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) ||
6134 (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) ||
6135 (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) ||
6136 (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) ||
6137 (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) ||
6138 (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) ||
6139 (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) ||
6140 (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) ||
6141 (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) ||
6142 (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) ||
6143 (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) ||
6144 (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) ||
6145 (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) ||
6146 (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK(c)) ||
6147 (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK(c))
6148 ) /* How's that for a conditional? */
6155 return (flags & ANYOF_INVERT) ? !match : match;
6159 S_reghop3(U8 *s, I32 off, const U8* lim)
6163 PERL_ARGS_ASSERT_REGHOP3;
6166 while (off-- && s < lim) {
6167 /* XXX could check well-formedness here */
6172 while (off++ && s > lim) {
6174 if (UTF8_IS_CONTINUED(*s)) {
6175 while (s > lim && UTF8_IS_CONTINUATION(*s))
6178 /* XXX could check well-formedness here */
6185 /* there are a bunch of places where we use two reghop3's that should
6186 be replaced with this routine. but since thats not done yet
6187 we ifdef it out - dmq
6190 S_reghop4(U8 *s, I32 off, const U8* llim, const U8* rlim)
6194 PERL_ARGS_ASSERT_REGHOP4;
6197 while (off-- && s < rlim) {
6198 /* XXX could check well-formedness here */
6203 while (off++ && s > llim) {
6205 if (UTF8_IS_CONTINUED(*s)) {
6206 while (s > llim && UTF8_IS_CONTINUATION(*s))
6209 /* XXX could check well-formedness here */
6217 S_reghopmaybe3(U8* s, I32 off, const U8* lim)
6221 PERL_ARGS_ASSERT_REGHOPMAYBE3;
6224 while (off-- && s < lim) {
6225 /* XXX could check well-formedness here */
6232 while (off++ && s > lim) {
6234 if (UTF8_IS_CONTINUED(*s)) {
6235 while (s > lim && UTF8_IS_CONTINUATION(*s))
6238 /* XXX could check well-formedness here */
6247 restore_pos(pTHX_ void *arg)
6250 regexp * const rex = (regexp *)arg;
6251 if (PL_reg_eval_set) {
6252 if (PL_reg_oldsaved) {
6253 rex->subbeg = PL_reg_oldsaved;
6254 rex->sublen = PL_reg_oldsavedlen;
6255 #ifdef PERL_OLD_COPY_ON_WRITE
6256 rex->saved_copy = PL_nrs;
6258 RXp_MATCH_COPIED_on(rex);
6260 PL_reg_magic->mg_len = PL_reg_oldpos;
6261 PL_reg_eval_set = 0;
6262 PL_curpm = PL_reg_oldcurpm;
6267 S_to_utf8_substr(pTHX_ register regexp *prog)
6271 PERL_ARGS_ASSERT_TO_UTF8_SUBSTR;
6274 if (prog->substrs->data[i].substr
6275 && !prog->substrs->data[i].utf8_substr) {
6276 SV* const sv = newSVsv(prog->substrs->data[i].substr);
6277 prog->substrs->data[i].utf8_substr = sv;
6278 sv_utf8_upgrade(sv);
6279 if (SvVALID(prog->substrs->data[i].substr)) {
6280 const U8 flags = BmFLAGS(prog->substrs->data[i].substr);
6281 if (flags & FBMcf_TAIL) {
6282 /* Trim the trailing \n that fbm_compile added last
6284 SvCUR_set(sv, SvCUR(sv) - 1);
6285 /* Whilst this makes the SV technically "invalid" (as its
6286 buffer is no longer followed by "\0") when fbm_compile()
6287 adds the "\n" back, a "\0" is restored. */
6289 fbm_compile(sv, flags);
6291 if (prog->substrs->data[i].substr == prog->check_substr)
6292 prog->check_utf8 = sv;
6298 S_to_byte_substr(pTHX_ register regexp *prog)
6303 PERL_ARGS_ASSERT_TO_BYTE_SUBSTR;
6306 if (prog->substrs->data[i].utf8_substr
6307 && !prog->substrs->data[i].substr) {
6308 SV* sv = newSVsv(prog->substrs->data[i].utf8_substr);
6309 if (sv_utf8_downgrade(sv, TRUE)) {
6310 if (SvVALID(prog->substrs->data[i].utf8_substr)) {
6312 = BmFLAGS(prog->substrs->data[i].utf8_substr);
6313 if (flags & FBMcf_TAIL) {
6314 /* Trim the trailing \n that fbm_compile added last
6316 SvCUR_set(sv, SvCUR(sv) - 1);
6318 fbm_compile(sv, flags);
6324 prog->substrs->data[i].substr = sv;
6325 if (prog->substrs->data[i].utf8_substr == prog->check_utf8)
6326 prog->check_substr = sv;
6333 * c-indentation-style: bsd
6335 * indent-tabs-mode: t
6338 * ex: set ts=8 sts=4 sw=4 noet: