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 /* Valid for non-utf8 strings only: avoids the reginclass call if there are no
98 * complications: i.e., if everything matchable is straight forward in the
100 #define REGINCLASS(prog,p,c) (ANYOF_FLAGS(p) ? reginclass(prog,p,c,0,0) \
101 : ANYOF_BITMAP_TEST(p,*(c)))
107 #define CHR_SVLEN(sv) (utf8_target ? sv_len_utf8(sv) : SvCUR(sv))
108 #define CHR_DIST(a,b) (PL_reg_match_utf8 ? utf8_distance(a,b) : a - b)
110 #define HOPc(pos,off) \
111 (char *)(PL_reg_match_utf8 \
112 ? reghop3((U8*)pos, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr)) \
114 #define HOPBACKc(pos, off) \
115 (char*)(PL_reg_match_utf8\
116 ? reghopmaybe3((U8*)pos, -off, (U8*)PL_bostr) \
117 : (pos - off >= PL_bostr) \
121 #define HOP3(pos,off,lim) (PL_reg_match_utf8 ? reghop3((U8*)(pos), off, (U8*)(lim)) : (U8*)(pos + off))
122 #define HOP3c(pos,off,lim) ((char*)HOP3(pos,off,lim))
124 /* these are unrolled below in the CCC_TRY_XXX defined */
125 #define LOAD_UTF8_CHARCLASS(class,str) STMT_START { \
126 if (!CAT2(PL_utf8_,class)) { bool ok; ENTER; save_re_context(); ok=CAT2(is_utf8_,class)((const U8*)str); assert(ok); LEAVE; } } STMT_END
128 /* Doesn't do an assert to verify that is correct */
129 #define LOAD_UTF8_CHARCLASS_NO_CHECK(class) STMT_START { \
130 if (!CAT2(PL_utf8_,class)) { bool ok; ENTER; save_re_context(); ok=CAT2(is_utf8_,class)((const U8*)" "); LEAVE; } } STMT_END
132 #define LOAD_UTF8_CHARCLASS_ALNUM() LOAD_UTF8_CHARCLASS(alnum,"a")
133 #define LOAD_UTF8_CHARCLASS_DIGIT() LOAD_UTF8_CHARCLASS(digit,"0")
134 #define LOAD_UTF8_CHARCLASS_SPACE() LOAD_UTF8_CHARCLASS(space," ")
136 #define LOAD_UTF8_CHARCLASS_GCB() /* Grapheme cluster boundaries */ \
137 LOAD_UTF8_CHARCLASS(X_begin, " "); \
138 LOAD_UTF8_CHARCLASS(X_non_hangul, "A"); \
139 /* These are utf8 constants, and not utf-ebcdic constants, so the \
140 * assert should likely and hopefully fail on an EBCDIC machine */ \
141 LOAD_UTF8_CHARCLASS(X_extend, "\xcc\x80"); /* U+0300 */ \
143 /* No asserts are done for these, in case called on an early \
144 * Unicode version in which they map to nothing */ \
145 LOAD_UTF8_CHARCLASS_NO_CHECK(X_prepend);/* U+0E40 "\xe0\xb9\x80" */ \
146 LOAD_UTF8_CHARCLASS_NO_CHECK(X_L); /* U+1100 "\xe1\x84\x80" */ \
147 LOAD_UTF8_CHARCLASS_NO_CHECK(X_LV); /* U+AC00 "\xea\xb0\x80" */ \
148 LOAD_UTF8_CHARCLASS_NO_CHECK(X_LVT); /* U+AC01 "\xea\xb0\x81" */ \
149 LOAD_UTF8_CHARCLASS_NO_CHECK(X_LV_LVT_V);/* U+AC01 "\xea\xb0\x81" */\
150 LOAD_UTF8_CHARCLASS_NO_CHECK(X_T); /* U+11A8 "\xe1\x86\xa8" */ \
151 LOAD_UTF8_CHARCLASS_NO_CHECK(X_V) /* U+1160 "\xe1\x85\xa0" */
154 We dont use PERL_LEGACY_UNICODE_CHARCLASS_MAPPINGS as the direct test
155 so that it is possible to override the option here without having to
156 rebuild the entire core. as we are required to do if we change regcomp.h
157 which is where PERL_LEGACY_UNICODE_CHARCLASS_MAPPINGS is defined.
159 #if PERL_LEGACY_UNICODE_CHARCLASS_MAPPINGS
160 #define BROKEN_UNICODE_CHARCLASS_MAPPINGS
163 #ifdef BROKEN_UNICODE_CHARCLASS_MAPPINGS
164 #define LOAD_UTF8_CHARCLASS_PERL_WORD() LOAD_UTF8_CHARCLASS_ALNUM()
165 #define LOAD_UTF8_CHARCLASS_PERL_SPACE() LOAD_UTF8_CHARCLASS_SPACE()
166 #define LOAD_UTF8_CHARCLASS_POSIX_DIGIT() LOAD_UTF8_CHARCLASS_DIGIT()
167 #define RE_utf8_perl_word PL_utf8_alnum
168 #define RE_utf8_perl_space PL_utf8_space
169 #define RE_utf8_posix_digit PL_utf8_digit
170 #define perl_word alnum
171 #define perl_space space
172 #define posix_digit digit
174 #define LOAD_UTF8_CHARCLASS_PERL_WORD() LOAD_UTF8_CHARCLASS(perl_word,"a")
175 #define LOAD_UTF8_CHARCLASS_PERL_SPACE() LOAD_UTF8_CHARCLASS(perl_space," ")
176 #define LOAD_UTF8_CHARCLASS_POSIX_DIGIT() LOAD_UTF8_CHARCLASS(posix_digit,"0")
177 #define RE_utf8_perl_word PL_utf8_perl_word
178 #define RE_utf8_perl_space PL_utf8_perl_space
179 #define RE_utf8_posix_digit PL_utf8_posix_digit
183 #define _CCC_TRY_AFF_COMMON(NAME,NAMEL,CLASS,STR,LCFUNC_utf8,FUNC) \
185 PL_reg_flags |= RF_tainted; \
190 if (utf8_target && UTF8_IS_CONTINUED(nextchr)) { \
191 if (!CAT2(PL_utf8_,CLASS)) { \
195 ok=CAT2(is_utf8_,CLASS)((const U8*)STR); \
199 if (!(OP(scan) == NAME \
200 ? cBOOL(swash_fetch(CAT2(PL_utf8_,CLASS), (U8*)locinput, utf8_target)) \
201 : LCFUNC_utf8((U8*)locinput))) \
205 locinput += PL_utf8skip[nextchr]; \
206 nextchr = UCHARAT(locinput); \
209 /* Drops through to the macro that calls this one */
211 #define CCC_TRY_AFF(NAME,NAMEL,CLASS,STR,LCFUNC_utf8,FUNC,LCFUNC) \
212 _CCC_TRY_AFF_COMMON(NAME,NAMEL,CLASS,STR,LCFUNC_utf8,FUNC) \
213 if (!(OP(scan) == NAME ? FUNC(nextchr) : LCFUNC(nextchr))) \
215 nextchr = UCHARAT(++locinput); \
218 /* Almost identical to the above, but has a case for a node that matches chars
219 * between 128 and 255 using Unicode (latin1) semantics. */
220 #define CCC_TRY_AFF_U(NAME,NAMEL,CLASS,STR,LCFUNC_utf8,FUNCU,LCFUNC) \
221 _CCC_TRY_AFF_COMMON(NAME,NAMEL,CLASS,STR,LCFUNC_utf8,FUNC) \
222 if (!(OP(scan) == NAMEL ? LCFUNC(nextchr) : (FUNCU(nextchr) && (isASCII(nextchr) || (FLAGS(scan) & USE_UNI))))) \
224 nextchr = UCHARAT(++locinput); \
227 #define _CCC_TRY_NEG_COMMON(NAME,NAMEL,CLASS,STR,LCFUNC_utf8,FUNC) \
229 PL_reg_flags |= RF_tainted; \
232 if (!nextchr && locinput >= PL_regeol) \
234 if (utf8_target && UTF8_IS_CONTINUED(nextchr)) { \
235 if (!CAT2(PL_utf8_,CLASS)) { \
239 ok=CAT2(is_utf8_,CLASS)((const U8*)STR); \
243 if ((OP(scan) == NAME \
244 ? cBOOL(swash_fetch(CAT2(PL_utf8_,CLASS), (U8*)locinput, utf8_target)) \
245 : LCFUNC_utf8((U8*)locinput))) \
249 locinput += PL_utf8skip[nextchr]; \
250 nextchr = UCHARAT(locinput); \
254 #define CCC_TRY_NEG(NAME,NAMEL,CLASS,STR,LCFUNC_utf8,FUNC,LCFUNC) \
255 _CCC_TRY_NEG_COMMON(NAME,NAMEL,CLASS,STR,LCFUNC_utf8,FUNC) \
256 if ((OP(scan) == NAME ? FUNC(nextchr) : LCFUNC(nextchr))) \
258 nextchr = UCHARAT(++locinput); \
262 #define CCC_TRY_NEG_U(NAME,NAMEL,CLASS,STR,LCFUNC_utf8,FUNCU,LCFUNC) \
263 _CCC_TRY_NEG_COMMON(NAME,NAMEL,CLASS,STR,LCFUNC_utf8,FUNCU) \
264 if ((OP(scan) == NAMEL ? LCFUNC(nextchr) : (FUNCU(nextchr) && (isASCII(nextchr) || (FLAGS(scan) & USE_UNI))))) \
266 nextchr = UCHARAT(++locinput); \
271 /* TODO: Combine JUMPABLE and HAS_TEXT to cache OP(rn) */
273 /* for use after a quantifier and before an EXACT-like node -- japhy */
274 /* it would be nice to rework regcomp.sym to generate this stuff. sigh
276 * NOTE that *nothing* that affects backtracking should be in here, specifically
277 * VERBS must NOT be included. JUMPABLE is used to determine if we can ignore a
278 * node that is in between two EXACT like nodes when ascertaining what the required
279 * "follow" character is. This should probably be moved to regex compile time
280 * although it may be done at run time beause of the REF possibility - more
281 * investigation required. -- demerphq
283 #define JUMPABLE(rn) ( \
285 (OP(rn) == CLOSE && (!cur_eval || cur_eval->u.eval.close_paren != ARG(rn))) || \
287 OP(rn) == SUSPEND || OP(rn) == IFMATCH || \
288 OP(rn) == PLUS || OP(rn) == MINMOD || \
290 (PL_regkind[OP(rn)] == CURLY && ARG1(rn) > 0) \
292 #define IS_EXACT(rn) (PL_regkind[OP(rn)] == EXACT)
294 #define HAS_TEXT(rn) ( IS_EXACT(rn) || PL_regkind[OP(rn)] == REF )
297 /* Currently these are only used when PL_regkind[OP(rn)] == EXACT so
298 we don't need this definition. */
299 #define IS_TEXT(rn) ( OP(rn)==EXACT || OP(rn)==REF || OP(rn)==NREF )
300 #define IS_TEXTF(rn) ( OP(rn)==EXACTF || OP(rn)==REFF || OP(rn)==NREFF )
301 #define IS_TEXTFL(rn) ( OP(rn)==EXACTFL || OP(rn)==REFFL || OP(rn)==NREFFL )
304 /* ... so we use this as its faster. */
305 #define IS_TEXT(rn) ( OP(rn)==EXACT )
306 #define IS_TEXTF(rn) ( OP(rn)==EXACTF )
307 #define IS_TEXTFL(rn) ( OP(rn)==EXACTFL )
312 Search for mandatory following text node; for lookahead, the text must
313 follow but for lookbehind (rn->flags != 0) we skip to the next step.
315 #define FIND_NEXT_IMPT(rn) STMT_START { \
316 while (JUMPABLE(rn)) { \
317 const OPCODE type = OP(rn); \
318 if (type == SUSPEND || PL_regkind[type] == CURLY) \
319 rn = NEXTOPER(NEXTOPER(rn)); \
320 else if (type == PLUS) \
322 else if (type == IFMATCH) \
323 rn = (rn->flags == 0) ? NEXTOPER(NEXTOPER(rn)) : rn + ARG(rn); \
324 else rn += NEXT_OFF(rn); \
329 static void restore_pos(pTHX_ void *arg);
331 #define REGCP_PAREN_ELEMS 4
332 #define REGCP_OTHER_ELEMS 5
333 #define REGCP_FRAME_ELEMS 1
334 /* REGCP_FRAME_ELEMS are not part of the REGCP_OTHER_ELEMS and
335 * are needed for the regexp context stack bookkeeping. */
338 S_regcppush(pTHX_ I32 parenfloor)
341 const int retval = PL_savestack_ix;
342 const int paren_elems_to_push = (PL_regsize - parenfloor) * REGCP_PAREN_ELEMS;
343 const UV total_elems = paren_elems_to_push + REGCP_OTHER_ELEMS;
344 const UV elems_shifted = total_elems << SAVE_TIGHT_SHIFT;
346 GET_RE_DEBUG_FLAGS_DECL;
348 if (paren_elems_to_push < 0)
349 Perl_croak(aTHX_ "panic: paren_elems_to_push < 0");
351 if ((elems_shifted >> SAVE_TIGHT_SHIFT) != total_elems)
352 Perl_croak(aTHX_ "panic: paren_elems_to_push offset %"UVuf
353 " out of range (%lu-%ld)",
354 total_elems, (unsigned long)PL_regsize, (long)parenfloor);
356 SSGROW(total_elems + REGCP_FRAME_ELEMS);
358 for (p = PL_regsize; p > parenfloor; p--) {
359 /* REGCP_PARENS_ELEMS are pushed per pairs of parentheses. */
360 SSPUSHINT(PL_regoffs[p].end);
361 SSPUSHINT(PL_regoffs[p].start);
362 SSPUSHPTR(PL_reg_start_tmp[p]);
364 DEBUG_BUFFERS_r(PerlIO_printf(Perl_debug_log,
365 " saving \\%"UVuf" %"IVdf"(%"IVdf")..%"IVdf"\n",
366 (UV)p, (IV)PL_regoffs[p].start,
367 (IV)(PL_reg_start_tmp[p] - PL_bostr),
368 (IV)PL_regoffs[p].end
371 /* REGCP_OTHER_ELEMS are pushed in any case, parentheses or no. */
372 SSPUSHPTR(PL_regoffs);
373 SSPUSHINT(PL_regsize);
374 SSPUSHINT(*PL_reglastparen);
375 SSPUSHINT(*PL_reglastcloseparen);
376 SSPUSHPTR(PL_reginput);
377 SSPUSHUV(SAVEt_REGCONTEXT | elems_shifted); /* Magic cookie. */
382 /* These are needed since we do not localize EVAL nodes: */
383 #define REGCP_SET(cp) \
385 PerlIO_printf(Perl_debug_log, \
386 " Setting an EVAL scope, savestack=%"IVdf"\n", \
387 (IV)PL_savestack_ix)); \
390 #define REGCP_UNWIND(cp) \
392 if (cp != PL_savestack_ix) \
393 PerlIO_printf(Perl_debug_log, \
394 " Clearing an EVAL scope, savestack=%"IVdf"..%"IVdf"\n", \
395 (IV)(cp), (IV)PL_savestack_ix)); \
399 S_regcppop(pTHX_ const regexp *rex)
404 GET_RE_DEBUG_FLAGS_DECL;
406 PERL_ARGS_ASSERT_REGCPPOP;
408 /* Pop REGCP_OTHER_ELEMS before the parentheses loop starts. */
410 assert((i & SAVE_MASK) == SAVEt_REGCONTEXT); /* Check that the magic cookie is there. */
411 i >>= SAVE_TIGHT_SHIFT; /* Parentheses elements to pop. */
412 input = (char *) SSPOPPTR;
413 *PL_reglastcloseparen = SSPOPINT;
414 *PL_reglastparen = SSPOPINT;
415 PL_regsize = SSPOPINT;
416 PL_regoffs=(regexp_paren_pair *) SSPOPPTR;
418 i -= REGCP_OTHER_ELEMS;
419 /* Now restore the parentheses context. */
420 for ( ; i > 0; i -= REGCP_PAREN_ELEMS) {
422 U32 paren = (U32)SSPOPINT;
423 PL_reg_start_tmp[paren] = (char *) SSPOPPTR;
424 PL_regoffs[paren].start = SSPOPINT;
426 if (paren <= *PL_reglastparen)
427 PL_regoffs[paren].end = tmps;
429 PerlIO_printf(Perl_debug_log,
430 " restoring \\%"UVuf" to %"IVdf"(%"IVdf")..%"IVdf"%s\n",
431 (UV)paren, (IV)PL_regoffs[paren].start,
432 (IV)(PL_reg_start_tmp[paren] - PL_bostr),
433 (IV)PL_regoffs[paren].end,
434 (paren > *PL_reglastparen ? "(no)" : ""));
438 if (*PL_reglastparen + 1 <= rex->nparens) {
439 PerlIO_printf(Perl_debug_log,
440 " restoring \\%"IVdf"..\\%"IVdf" to undef\n",
441 (IV)(*PL_reglastparen + 1), (IV)rex->nparens);
445 /* It would seem that the similar code in regtry()
446 * already takes care of this, and in fact it is in
447 * a better location to since this code can #if 0-ed out
448 * but the code in regtry() is needed or otherwise tests
449 * requiring null fields (pat.t#187 and split.t#{13,14}
450 * (as of patchlevel 7877) will fail. Then again,
451 * this code seems to be necessary or otherwise
452 * this erroneously leaves $1 defined: "1" =~ /^(?:(\d)x)?\d$/
453 * --jhi updated by dapm */
454 for (i = *PL_reglastparen + 1; i <= rex->nparens; i++) {
456 PL_regoffs[i].start = -1;
457 PL_regoffs[i].end = -1;
463 #define regcpblow(cp) LEAVE_SCOPE(cp) /* Ignores regcppush()ed data. */
466 * pregexec and friends
469 #ifndef PERL_IN_XSUB_RE
471 - pregexec - match a regexp against a string
474 Perl_pregexec(pTHX_ REGEXP * const prog, char* stringarg, register char *strend,
475 char *strbeg, I32 minend, SV *screamer, U32 nosave)
476 /* strend: pointer to null at end of string */
477 /* strbeg: real beginning of string */
478 /* minend: end of match must be >=minend after stringarg. */
479 /* nosave: For optimizations. */
481 PERL_ARGS_ASSERT_PREGEXEC;
484 regexec_flags(prog, stringarg, strend, strbeg, minend, screamer, NULL,
485 nosave ? 0 : REXEC_COPY_STR);
490 * Need to implement the following flags for reg_anch:
492 * USE_INTUIT_NOML - Useful to call re_intuit_start() first
494 * INTUIT_AUTORITATIVE_NOML - Can trust a positive answer
495 * INTUIT_AUTORITATIVE_ML
496 * INTUIT_ONCE_NOML - Intuit can match in one location only.
499 * Another flag for this function: SECOND_TIME (so that float substrs
500 * with giant delta may be not rechecked).
503 /* Assumptions: if ANCH_GPOS, then strpos is anchored. XXXX Check GPOS logic */
505 /* If SCREAM, then SvPVX_const(sv) should be compatible with strpos and strend.
506 Otherwise, only SvCUR(sv) is used to get strbeg. */
508 /* XXXX We assume that strpos is strbeg unless sv. */
510 /* XXXX Some places assume that there is a fixed substring.
511 An update may be needed if optimizer marks as "INTUITable"
512 RExen without fixed substrings. Similarly, it is assumed that
513 lengths of all the strings are no more than minlen, thus they
514 cannot come from lookahead.
515 (Or minlen should take into account lookahead.)
516 NOTE: Some of this comment is not correct. minlen does now take account
517 of lookahead/behind. Further research is required. -- demerphq
521 /* A failure to find a constant substring means that there is no need to make
522 an expensive call to REx engine, thus we celebrate a failure. Similarly,
523 finding a substring too deep into the string means that less calls to
524 regtry() should be needed.
526 REx compiler's optimizer found 4 possible hints:
527 a) Anchored substring;
529 c) Whether we are anchored (beginning-of-line or \G);
530 d) First node (of those at offset 0) which may distingush positions;
531 We use a)b)d) and multiline-part of c), and try to find a position in the
532 string which does not contradict any of them.
535 /* Most of decisions we do here should have been done at compile time.
536 The nodes of the REx which we used for the search should have been
537 deleted from the finite automaton. */
540 Perl_re_intuit_start(pTHX_ REGEXP * const rx, SV *sv, char *strpos,
541 char *strend, const U32 flags, re_scream_pos_data *data)
544 struct regexp *const prog = (struct regexp *)SvANY(rx);
545 register I32 start_shift = 0;
546 /* Should be nonnegative! */
547 register I32 end_shift = 0;
552 const bool utf8_target = (sv && SvUTF8(sv)) ? 1 : 0; /* if no sv we have to assume bytes */
554 register char *other_last = NULL; /* other substr checked before this */
555 char *check_at = NULL; /* check substr found at this pos */
556 const I32 multiline = prog->extflags & RXf_PMf_MULTILINE;
557 RXi_GET_DECL(prog,progi);
559 const char * const i_strpos = strpos;
561 GET_RE_DEBUG_FLAGS_DECL;
563 PERL_ARGS_ASSERT_RE_INTUIT_START;
565 RX_MATCH_UTF8_set(rx,utf8_target);
568 PL_reg_flags |= RF_utf8;
571 debug_start_match(rx, utf8_target, strpos, strend,
572 sv ? "Guessing start of match in sv for"
573 : "Guessing start of match in string for");
576 /* CHR_DIST() would be more correct here but it makes things slow. */
577 if (prog->minlen > strend - strpos) {
578 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
579 "String too short... [re_intuit_start]\n"));
583 strbeg = (sv && SvPOK(sv)) ? strend - SvCUR(sv) : strpos;
586 if (!prog->check_utf8 && prog->check_substr)
587 to_utf8_substr(prog);
588 check = prog->check_utf8;
590 if (!prog->check_substr && prog->check_utf8)
591 to_byte_substr(prog);
592 check = prog->check_substr;
594 if (check == &PL_sv_undef) {
595 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
596 "Non-utf8 string cannot match utf8 check string\n"));
599 if (prog->extflags & RXf_ANCH) { /* Match at beg-of-str or after \n */
600 ml_anch = !( (prog->extflags & RXf_ANCH_SINGLE)
601 || ( (prog->extflags & RXf_ANCH_BOL)
602 && !multiline ) ); /* Check after \n? */
605 if ( !(prog->extflags & RXf_ANCH_GPOS) /* Checked by the caller */
606 && !(prog->intflags & PREGf_IMPLICIT) /* not a real BOL */
607 /* SvCUR is not set on references: SvRV and SvPVX_const overlap */
609 && (strpos != strbeg)) {
610 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Not at start...\n"));
613 if (prog->check_offset_min == prog->check_offset_max &&
614 !(prog->extflags & RXf_CANY_SEEN)) {
615 /* Substring at constant offset from beg-of-str... */
618 s = HOP3c(strpos, prog->check_offset_min, strend);
621 slen = SvCUR(check); /* >= 1 */
623 if ( strend - s > slen || strend - s < slen - 1
624 || (strend - s == slen && strend[-1] != '\n')) {
625 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "String too long...\n"));
628 /* Now should match s[0..slen-2] */
630 if (slen && (*SvPVX_const(check) != *s
632 && memNE(SvPVX_const(check), s, slen)))) {
634 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "String not equal...\n"));
638 else if (*SvPVX_const(check) != *s
639 || ((slen = SvCUR(check)) > 1
640 && memNE(SvPVX_const(check), s, slen)))
643 goto success_at_start;
646 /* Match is anchored, but substr is not anchored wrt beg-of-str. */
648 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
649 end_shift = prog->check_end_shift;
652 const I32 end = prog->check_offset_max + CHR_SVLEN(check)
653 - (SvTAIL(check) != 0);
654 const I32 eshift = CHR_DIST((U8*)strend, (U8*)s) - end;
656 if (end_shift < eshift)
660 else { /* Can match at random position */
663 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
664 end_shift = prog->check_end_shift;
666 /* end shift should be non negative here */
669 #ifdef QDEBUGGING /* 7/99: reports of failure (with the older version) */
671 Perl_croak(aTHX_ "panic: end_shift: %"IVdf" pattern:\n%s\n ",
672 (IV)end_shift, RX_PRECOMP(prog));
676 /* Find a possible match in the region s..strend by looking for
677 the "check" substring in the region corrected by start/end_shift. */
680 I32 srch_start_shift = start_shift;
681 I32 srch_end_shift = end_shift;
682 if (srch_start_shift < 0 && strbeg - s > srch_start_shift) {
683 srch_end_shift -= ((strbeg - s) - srch_start_shift);
684 srch_start_shift = strbeg - s;
686 DEBUG_OPTIMISE_MORE_r({
687 PerlIO_printf(Perl_debug_log, "Check offset min: %"IVdf" Start shift: %"IVdf" End shift %"IVdf" Real End Shift: %"IVdf"\n",
688 (IV)prog->check_offset_min,
689 (IV)srch_start_shift,
691 (IV)prog->check_end_shift);
694 if (flags & REXEC_SCREAM) {
695 I32 p = -1; /* Internal iterator of scream. */
696 I32 * const pp = data ? data->scream_pos : &p;
698 if (PL_screamfirst[BmRARE(check)] >= 0
699 || ( BmRARE(check) == '\n'
700 && (BmPREVIOUS(check) == SvCUR(check) - 1)
702 s = screaminstr(sv, check,
703 srch_start_shift + (s - strbeg), srch_end_shift, pp, 0);
706 /* we may be pointing at the wrong string */
707 if (s && RXp_MATCH_COPIED(prog))
708 s = strbeg + (s - SvPVX_const(sv));
710 *data->scream_olds = s;
715 if (prog->extflags & RXf_CANY_SEEN) {
716 start_point= (U8*)(s + srch_start_shift);
717 end_point= (U8*)(strend - srch_end_shift);
719 start_point= HOP3(s, srch_start_shift, srch_start_shift < 0 ? strbeg : strend);
720 end_point= HOP3(strend, -srch_end_shift, strbeg);
722 DEBUG_OPTIMISE_MORE_r({
723 PerlIO_printf(Perl_debug_log, "fbm_instr len=%d str=<%.*s>\n",
724 (int)(end_point - start_point),
725 (int)(end_point - start_point) > 20 ? 20 : (int)(end_point - start_point),
729 s = fbm_instr( start_point, end_point,
730 check, multiline ? FBMrf_MULTILINE : 0);
733 /* Update the count-of-usability, remove useless subpatterns,
737 RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0),
738 SvPVX_const(check), RE_SV_DUMPLEN(check), 30);
739 PerlIO_printf(Perl_debug_log, "%s %s substr %s%s%s",
740 (s ? "Found" : "Did not find"),
741 (check == (utf8_target ? prog->anchored_utf8 : prog->anchored_substr)
742 ? "anchored" : "floating"),
745 (s ? " at offset " : "...\n") );
750 /* Finish the diagnostic message */
751 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%ld...\n", (long)(s - i_strpos)) );
753 /* XXX dmq: first branch is for positive lookbehind...
754 Our check string is offset from the beginning of the pattern.
755 So we need to do any stclass tests offset forward from that
764 /* Got a candidate. Check MBOL anchoring, and the *other* substr.
765 Start with the other substr.
766 XXXX no SCREAM optimization yet - and a very coarse implementation
767 XXXX /ttx+/ results in anchored="ttx", floating="x". floating will
768 *always* match. Probably should be marked during compile...
769 Probably it is right to do no SCREAM here...
772 if (utf8_target ? (prog->float_utf8 && prog->anchored_utf8)
773 : (prog->float_substr && prog->anchored_substr))
775 /* Take into account the "other" substring. */
776 /* XXXX May be hopelessly wrong for UTF... */
779 if (check == (utf8_target ? prog->float_utf8 : prog->float_substr)) {
782 char * const last = HOP3c(s, -start_shift, strbeg);
784 char * const saved_s = s;
787 t = s - prog->check_offset_max;
788 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
790 || ((t = (char*)reghopmaybe3((U8*)s, -(prog->check_offset_max), (U8*)strpos))
795 t = HOP3c(t, prog->anchored_offset, strend);
796 if (t < other_last) /* These positions already checked */
798 last2 = last1 = HOP3c(strend, -prog->minlen, strbeg);
801 /* XXXX It is not documented what units *_offsets are in.
802 We assume bytes, but this is clearly wrong.
803 Meaning this code needs to be carefully reviewed for errors.
807 /* On end-of-str: see comment below. */
808 must = utf8_target ? prog->anchored_utf8 : prog->anchored_substr;
809 if (must == &PL_sv_undef) {
811 DEBUG_r(must = prog->anchored_utf8); /* for debug */
816 HOP3(HOP3(last1, prog->anchored_offset, strend)
817 + SvCUR(must), -(SvTAIL(must)!=0), strbeg),
819 multiline ? FBMrf_MULTILINE : 0
822 RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0),
823 SvPVX_const(must), RE_SV_DUMPLEN(must), 30);
824 PerlIO_printf(Perl_debug_log, "%s anchored substr %s%s",
825 (s ? "Found" : "Contradicts"),
826 quoted, RE_SV_TAIL(must));
831 if (last1 >= last2) {
832 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
833 ", giving up...\n"));
836 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
837 ", trying floating at offset %ld...\n",
838 (long)(HOP3c(saved_s, 1, strend) - i_strpos)));
839 other_last = HOP3c(last1, prog->anchored_offset+1, strend);
840 s = HOP3c(last, 1, strend);
844 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
845 (long)(s - i_strpos)));
846 t = HOP3c(s, -prog->anchored_offset, strbeg);
847 other_last = HOP3c(s, 1, strend);
855 else { /* Take into account the floating substring. */
857 char * const saved_s = s;
860 t = HOP3c(s, -start_shift, strbeg);
862 HOP3c(strend, -prog->minlen + prog->float_min_offset, strbeg);
863 if (CHR_DIST((U8*)last, (U8*)t) > prog->float_max_offset)
864 last = HOP3c(t, prog->float_max_offset, strend);
865 s = HOP3c(t, prog->float_min_offset, strend);
868 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
869 must = utf8_target ? prog->float_utf8 : prog->float_substr;
870 /* fbm_instr() takes into account exact value of end-of-str
871 if the check is SvTAIL(ed). Since false positives are OK,
872 and end-of-str is not later than strend we are OK. */
873 if (must == &PL_sv_undef) {
875 DEBUG_r(must = prog->float_utf8); /* for debug message */
878 s = fbm_instr((unsigned char*)s,
879 (unsigned char*)last + SvCUR(must)
881 must, multiline ? FBMrf_MULTILINE : 0);
883 RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0),
884 SvPVX_const(must), RE_SV_DUMPLEN(must), 30);
885 PerlIO_printf(Perl_debug_log, "%s floating substr %s%s",
886 (s ? "Found" : "Contradicts"),
887 quoted, RE_SV_TAIL(must));
891 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
892 ", giving up...\n"));
895 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
896 ", trying anchored starting at offset %ld...\n",
897 (long)(saved_s + 1 - i_strpos)));
899 s = HOP3c(t, 1, strend);
903 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
904 (long)(s - i_strpos)));
905 other_last = s; /* Fix this later. --Hugo */
915 t= (char*)HOP3( s, -prog->check_offset_max, (prog->check_offset_max<0) ? strend : strpos);
917 DEBUG_OPTIMISE_MORE_r(
918 PerlIO_printf(Perl_debug_log,
919 "Check offset min:%"IVdf" max:%"IVdf" S:%"IVdf" t:%"IVdf" D:%"IVdf" end:%"IVdf"\n",
920 (IV)prog->check_offset_min,
921 (IV)prog->check_offset_max,
929 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
931 || ((t = (char*)reghopmaybe3((U8*)s, -prog->check_offset_max, (U8*) ((prog->check_offset_max<0) ? strend : strpos)))
934 /* Fixed substring is found far enough so that the match
935 cannot start at strpos. */
937 if (ml_anch && t[-1] != '\n') {
938 /* Eventually fbm_*() should handle this, but often
939 anchored_offset is not 0, so this check will not be wasted. */
940 /* XXXX In the code below we prefer to look for "^" even in
941 presence of anchored substrings. And we search even
942 beyond the found float position. These pessimizations
943 are historical artefacts only. */
945 while (t < strend - prog->minlen) {
947 if (t < check_at - prog->check_offset_min) {
948 if (utf8_target ? prog->anchored_utf8 : prog->anchored_substr) {
949 /* Since we moved from the found position,
950 we definitely contradict the found anchored
951 substr. Due to the above check we do not
952 contradict "check" substr.
953 Thus we can arrive here only if check substr
954 is float. Redo checking for "other"=="fixed".
957 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld, rescanning for anchored from offset %ld...\n",
958 PL_colors[0], PL_colors[1], (long)(strpos - i_strpos), (long)(strpos - i_strpos + prog->anchored_offset)));
959 goto do_other_anchored;
961 /* We don't contradict the found floating substring. */
962 /* XXXX Why not check for STCLASS? */
964 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld...\n",
965 PL_colors[0], PL_colors[1], (long)(s - i_strpos)));
968 /* Position contradicts check-string */
969 /* XXXX probably better to look for check-string
970 than for "\n", so one should lower the limit for t? */
971 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m, restarting lookup for check-string at offset %ld...\n",
972 PL_colors[0], PL_colors[1], (long)(t + 1 - i_strpos)));
973 other_last = strpos = s = t + 1;
978 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Did not find /%s^%s/m...\n",
979 PL_colors[0], PL_colors[1]));
983 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Starting position does not contradict /%s^%s/m...\n",
984 PL_colors[0], PL_colors[1]));
988 ++BmUSEFUL(utf8_target ? prog->check_utf8 : prog->check_substr); /* hooray/5 */
991 /* The found string does not prohibit matching at strpos,
992 - no optimization of calling REx engine can be performed,
993 unless it was an MBOL and we are not after MBOL,
994 or a future STCLASS check will fail this. */
996 /* Even in this situation we may use MBOL flag if strpos is offset
997 wrt the start of the string. */
998 if (ml_anch && sv && !SvROK(sv) /* See prev comment on SvROK */
999 && (strpos != strbeg) && strpos[-1] != '\n'
1000 /* May be due to an implicit anchor of m{.*foo} */
1001 && !(prog->intflags & PREGf_IMPLICIT))
1006 DEBUG_EXECUTE_r( if (ml_anch)
1007 PerlIO_printf(Perl_debug_log, "Position at offset %ld does not contradict /%s^%s/m...\n",
1008 (long)(strpos - i_strpos), PL_colors[0], PL_colors[1]);
1011 if (!(prog->intflags & PREGf_NAUGHTY) /* XXXX If strpos moved? */
1013 prog->check_utf8 /* Could be deleted already */
1014 && --BmUSEFUL(prog->check_utf8) < 0
1015 && (prog->check_utf8 == prog->float_utf8)
1017 prog->check_substr /* Could be deleted already */
1018 && --BmUSEFUL(prog->check_substr) < 0
1019 && (prog->check_substr == prog->float_substr)
1022 /* If flags & SOMETHING - do not do it many times on the same match */
1023 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "... Disabling check substring...\n"));
1024 /* XXX Does the destruction order has to change with utf8_target? */
1025 SvREFCNT_dec(utf8_target ? prog->check_utf8 : prog->check_substr);
1026 SvREFCNT_dec(utf8_target ? prog->check_substr : prog->check_utf8);
1027 prog->check_substr = prog->check_utf8 = NULL; /* disable */
1028 prog->float_substr = prog->float_utf8 = NULL; /* clear */
1029 check = NULL; /* abort */
1031 /* XXXX If the check string was an implicit check MBOL, then we need to unset the relevent flag
1032 see http://bugs.activestate.com/show_bug.cgi?id=87173 */
1033 if (prog->intflags & PREGf_IMPLICIT)
1034 prog->extflags &= ~RXf_ANCH_MBOL;
1035 /* XXXX This is a remnant of the old implementation. It
1036 looks wasteful, since now INTUIT can use many
1037 other heuristics. */
1038 prog->extflags &= ~RXf_USE_INTUIT;
1039 /* XXXX What other flags might need to be cleared in this branch? */
1045 /* Last resort... */
1046 /* XXXX BmUSEFUL already changed, maybe multiple change is meaningful... */
1047 /* trie stclasses are too expensive to use here, we are better off to
1048 leave it to regmatch itself */
1049 if (progi->regstclass && PL_regkind[OP(progi->regstclass)]!=TRIE) {
1050 /* minlen == 0 is possible if regstclass is \b or \B,
1051 and the fixed substr is ''$.
1052 Since minlen is already taken into account, s+1 is before strend;
1053 accidentally, minlen >= 1 guaranties no false positives at s + 1
1054 even for \b or \B. But (minlen? 1 : 0) below assumes that
1055 regstclass does not come from lookahead... */
1056 /* If regstclass takes bytelength more than 1: If charlength==1, OK.
1057 This leaves EXACTF only, which is dealt with in find_byclass(). */
1058 const U8* const str = (U8*)STRING(progi->regstclass);
1059 const int cl_l = (PL_regkind[OP(progi->regstclass)] == EXACT
1060 ? CHR_DIST(str+STR_LEN(progi->regstclass), str)
1063 if (prog->anchored_substr || prog->anchored_utf8 || ml_anch)
1064 endpos= HOP3c(s, (prog->minlen ? cl_l : 0), strend);
1065 else if (prog->float_substr || prog->float_utf8)
1066 endpos= HOP3c(HOP3c(check_at, -start_shift, strbeg), cl_l, strend);
1070 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "start_shift: %"IVdf" check_at: %"IVdf" s: %"IVdf" endpos: %"IVdf"\n",
1071 (IV)start_shift, (IV)(check_at - strbeg), (IV)(s - strbeg), (IV)(endpos - strbeg)));
1074 s = find_byclass(prog, progi->regstclass, s, endpos, NULL);
1077 const char *what = NULL;
1079 if (endpos == strend) {
1080 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
1081 "Could not match STCLASS...\n") );
1084 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
1085 "This position contradicts STCLASS...\n") );
1086 if ((prog->extflags & RXf_ANCH) && !ml_anch)
1088 /* Contradict one of substrings */
1089 if (prog->anchored_substr || prog->anchored_utf8) {
1090 if ((utf8_target ? prog->anchored_utf8 : prog->anchored_substr) == check) {
1091 DEBUG_EXECUTE_r( what = "anchored" );
1093 s = HOP3c(t, 1, strend);
1094 if (s + start_shift + end_shift > strend) {
1095 /* XXXX Should be taken into account earlier? */
1096 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
1097 "Could not match STCLASS...\n") );
1102 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
1103 "Looking for %s substr starting at offset %ld...\n",
1104 what, (long)(s + start_shift - i_strpos)) );
1107 /* Have both, check_string is floating */
1108 if (t + start_shift >= check_at) /* Contradicts floating=check */
1109 goto retry_floating_check;
1110 /* Recheck anchored substring, but not floating... */
1114 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
1115 "Looking for anchored substr starting at offset %ld...\n",
1116 (long)(other_last - i_strpos)) );
1117 goto do_other_anchored;
1119 /* Another way we could have checked stclass at the
1120 current position only: */
1125 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
1126 "Looking for /%s^%s/m starting at offset %ld...\n",
1127 PL_colors[0], PL_colors[1], (long)(t - i_strpos)) );
1130 if (!(utf8_target ? prog->float_utf8 : prog->float_substr)) /* Could have been deleted */
1132 /* Check is floating subtring. */
1133 retry_floating_check:
1134 t = check_at - start_shift;
1135 DEBUG_EXECUTE_r( what = "floating" );
1136 goto hop_and_restart;
1139 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
1140 "By STCLASS: moving %ld --> %ld\n",
1141 (long)(t - i_strpos), (long)(s - i_strpos))
1145 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
1146 "Does not contradict STCLASS...\n");
1151 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%s%s:%s match at offset %ld\n",
1152 PL_colors[4], (check ? "Guessed" : "Giving up"),
1153 PL_colors[5], (long)(s - i_strpos)) );
1156 fail_finish: /* Substring not found */
1157 if (prog->check_substr || prog->check_utf8) /* could be removed already */
1158 BmUSEFUL(utf8_target ? prog->check_utf8 : prog->check_substr) += 5; /* hooray */
1160 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch rejected by optimizer%s\n",
1161 PL_colors[4], PL_colors[5]));
1165 #define DECL_TRIE_TYPE(scan) \
1166 const enum { trie_plain, trie_utf8, trie_utf8_fold, trie_latin_utf8_fold } \
1167 trie_type = (scan->flags != EXACT) \
1168 ? (utf8_target ? trie_utf8_fold : (UTF_PATTERN ? trie_latin_utf8_fold : trie_plain)) \
1169 : (utf8_target ? trie_utf8 : trie_plain)
1171 #define REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc, uscan, len, \
1172 uvc, charid, foldlen, foldbuf, uniflags) STMT_START { \
1173 switch (trie_type) { \
1174 case trie_utf8_fold: \
1175 if ( foldlen>0 ) { \
1176 uvc = utf8n_to_uvuni( uscan, UTF8_MAXLEN, &len, uniflags ); \
1181 uvc = utf8n_to_uvuni( (U8*)uc, UTF8_MAXLEN, &len, uniflags ); \
1182 uvc = to_uni_fold( uvc, foldbuf, &foldlen ); \
1183 foldlen -= UNISKIP( uvc ); \
1184 uscan = foldbuf + UNISKIP( uvc ); \
1187 case trie_latin_utf8_fold: \
1188 if ( foldlen>0 ) { \
1189 uvc = utf8n_to_uvuni( uscan, UTF8_MAXLEN, &len, uniflags ); \
1195 uvc = to_uni_fold( *(U8*)uc, foldbuf, &foldlen ); \
1196 foldlen -= UNISKIP( uvc ); \
1197 uscan = foldbuf + UNISKIP( uvc ); \
1201 uvc = utf8n_to_uvuni( (U8*)uc, UTF8_MAXLEN, &len, uniflags ); \
1208 charid = trie->charmap[ uvc ]; \
1212 if (widecharmap) { \
1213 SV** const svpp = hv_fetch(widecharmap, \
1214 (char*)&uvc, sizeof(UV), 0); \
1216 charid = (U16)SvIV(*svpp); \
1221 #define REXEC_FBC_EXACTISH_CHECK(CoNd) \
1223 char *my_strend= (char *)strend; \
1226 foldEQ_utf8(s, &my_strend, 0, utf8_target, \
1227 m, NULL, ln, cBOOL(UTF_PATTERN))) \
1228 && (!reginfo || regtry(reginfo, &s)) ) \
1231 U8 foldbuf[UTF8_MAXBYTES_CASE+1]; \
1232 uvchr_to_utf8(tmpbuf, c); \
1233 f = to_utf8_fold(tmpbuf, foldbuf, &foldlen); \
1235 && (f == c1 || f == c2) \
1237 foldEQ_utf8(s, &my_strend, 0, utf8_target,\
1238 m, NULL, ln, cBOOL(UTF_PATTERN)))\
1239 && (!reginfo || regtry(reginfo, &s)) ) \
1245 #define REXEC_FBC_EXACTISH_SCAN(CoNd) \
1249 && (ln == 1 || (OP(c) == EXACTF \
1250 ? foldEQ(s, m, ln) \
1251 : foldEQ_locale(s, m, ln))) \
1252 && (!reginfo || regtry(reginfo, &s)) ) \
1258 #define REXEC_FBC_UTF8_SCAN(CoDe) \
1260 while (s + (uskip = UTF8SKIP(s)) <= strend) { \
1266 #define REXEC_FBC_SCAN(CoDe) \
1268 while (s < strend) { \
1274 #define REXEC_FBC_UTF8_CLASS_SCAN(CoNd) \
1275 REXEC_FBC_UTF8_SCAN( \
1277 if (tmp && (!reginfo || regtry(reginfo, &s))) \
1286 #define REXEC_FBC_CLASS_SCAN(CoNd) \
1289 if (tmp && (!reginfo || regtry(reginfo, &s))) \
1298 #define REXEC_FBC_TRYIT \
1299 if ((!reginfo || regtry(reginfo, &s))) \
1302 #define REXEC_FBC_CSCAN(CoNdUtF8,CoNd) \
1303 if (utf8_target) { \
1304 REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \
1307 REXEC_FBC_CLASS_SCAN(CoNd); \
1311 #define REXEC_FBC_CSCAN_PRELOAD(UtFpReLoAd,CoNdUtF8,CoNd) \
1312 if (utf8_target) { \
1314 REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \
1317 REXEC_FBC_CLASS_SCAN(CoNd); \
1321 #define REXEC_FBC_CSCAN_TAINT(CoNdUtF8,CoNd) \
1322 PL_reg_flags |= RF_tainted; \
1323 if (utf8_target) { \
1324 REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \
1327 REXEC_FBC_CLASS_SCAN(CoNd); \
1331 #define DUMP_EXEC_POS(li,s,doutf8) \
1332 dump_exec_pos(li,s,(PL_regeol),(PL_bostr),(PL_reg_starttry),doutf8)
1334 /* We know what class REx starts with. Try to find this position... */
1335 /* if reginfo is NULL, its a dryrun */
1336 /* annoyingly all the vars in this routine have different names from their counterparts
1337 in regmatch. /grrr */
1340 S_find_byclass(pTHX_ regexp * prog, const regnode *c, char *s,
1341 const char *strend, regmatch_info *reginfo)
1344 const I32 doevery = (prog->intflags & PREGf_SKIP) == 0;
1348 register STRLEN uskip;
1352 register I32 tmp = 1; /* Scratch variable? */
1353 register const bool utf8_target = PL_reg_match_utf8;
1354 RXi_GET_DECL(prog,progi);
1356 PERL_ARGS_ASSERT_FIND_BYCLASS;
1358 /* We know what class it must start with. */
1362 REXEC_FBC_UTF8_CLASS_SCAN((ANYOF_FLAGS(c) & ANYOF_UNICODE) ||
1363 !UTF8_IS_INVARIANT((U8)s[0]) ?
1364 reginclass(prog, c, (U8*)s, 0, utf8_target) :
1365 REGINCLASS(prog, c, (U8*)s));
1368 while (s < strend) {
1371 if (REGINCLASS(prog, c, (U8*)s) ||
1372 (ANYOF_FOLD_SHARP_S(c, s, strend) &&
1373 /* The assignment of 2 is intentional:
1374 * for the folded sharp s, the skip is 2. */
1375 (skip = SHARP_S_SKIP))) {
1376 if (tmp && (!reginfo || regtry(reginfo, &s)))
1389 if (tmp && (!reginfo || regtry(reginfo, &s)))
1397 ln = STR_LEN(c); /* length to match in octets/bytes */
1398 lnc = (I32) ln; /* length to match in characters */
1400 STRLEN ulen1, ulen2;
1402 U8 tmpbuf1[UTF8_MAXBYTES_CASE+1];
1403 U8 tmpbuf2[UTF8_MAXBYTES_CASE+1];
1404 /* used by commented-out code below */
1405 /*const U32 uniflags = UTF8_ALLOW_DEFAULT;*/
1407 /* XXX: Since the node will be case folded at compile
1408 time this logic is a little odd, although im not
1409 sure that its actually wrong. --dmq */
1411 c1 = to_utf8_lower((U8*)m, tmpbuf1, &ulen1);
1412 c2 = to_utf8_upper((U8*)m, tmpbuf2, &ulen2);
1414 /* XXX: This is kinda strange. to_utf8_XYZ returns the
1415 codepoint of the first character in the converted
1416 form, yet originally we did the extra step.
1417 No tests fail by commenting this code out however
1418 so Ive left it out. -- dmq.
1420 c1 = utf8n_to_uvchr(tmpbuf1, UTF8_MAXBYTES_CASE,
1422 c2 = utf8n_to_uvchr(tmpbuf2, UTF8_MAXBYTES_CASE,
1427 while (sm < ((U8 *) m + ln)) {
1442 c2 = PL_fold_locale[c1];
1444 e = HOP3c(strend, -((I32)lnc), s);
1446 if (!reginfo && e < s)
1447 e = s; /* Due to minlen logic of intuit() */
1449 /* The idea in the EXACTF* cases is to first find the
1450 * first character of the EXACTF* node and then, if
1451 * necessary, case-insensitively compare the full
1452 * text of the node. The c1 and c2 are the first
1453 * characters (though in Unicode it gets a bit
1454 * more complicated because there are more cases
1455 * than just upper and lower: one needs to use
1456 * the so-called folding case for case-insensitive
1457 * matching (called "loose matching" in Unicode).
1458 * foldEQ_utf8() will do just that. */
1460 if (utf8_target || UTF_PATTERN) {
1462 U8 tmpbuf [UTF8_MAXBYTES+1];
1465 const U32 uniflags = UTF8_ALLOW_DEFAULT;
1467 /* Upper and lower of 1st char are equal -
1468 * probably not a "letter". */
1471 c = utf8n_to_uvchr((U8*)s, UTF8_MAXBYTES, &len,
1476 REXEC_FBC_EXACTISH_CHECK(c == c1);
1482 c = utf8n_to_uvchr((U8*)s, UTF8_MAXBYTES, &len,
1488 /* Handle some of the three Greek sigmas cases.
1489 * Note that not all the possible combinations
1490 * are handled here: some of them are handled
1491 * by the standard folding rules, and some of
1492 * them (the character class or ANYOF cases)
1493 * are handled during compiletime in
1494 * regexec.c:S_regclass(). */
1495 if (c == (UV)UNICODE_GREEK_CAPITAL_LETTER_SIGMA ||
1496 c == (UV)UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA)
1497 c = (UV)UNICODE_GREEK_SMALL_LETTER_SIGMA;
1499 REXEC_FBC_EXACTISH_CHECK(c == c1 || c == c2);
1504 /* Neither pattern nor string are UTF8 */
1506 REXEC_FBC_EXACTISH_SCAN(*(U8*)s == c1);
1508 REXEC_FBC_EXACTISH_SCAN(*(U8*)s == c1 || *(U8*)s == c2);
1512 PL_reg_flags |= RF_tainted;
1519 U8 * const r = reghop3((U8*)s, -1, (U8*)PL_bostr);
1520 tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, UTF8_ALLOW_DEFAULT);
1522 tmp = ((OP(c) == BOUND ?
1523 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1524 LOAD_UTF8_CHARCLASS_ALNUM();
1525 REXEC_FBC_UTF8_SCAN(
1526 if (tmp == !(OP(c) == BOUND ?
1527 cBOOL(swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target)) :
1528 isALNUM_LC_utf8((U8*)s)))
1535 else { /* Not utf8 */
1536 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1537 tmp = cBOOL((OP(c) == BOUNDL)
1539 : (isWORDCHAR_L1(tmp)
1540 && (isASCII(tmp) || (FLAGS(c) & USE_UNI))));
1545 : (isWORDCHAR_L1((U8) *s)
1546 && (isASCII((U8) *s) || (FLAGS(c) & USE_UNI)))))
1553 if ((!prog->minlen && tmp) && (!reginfo || regtry(reginfo, &s)))
1557 PL_reg_flags |= RF_tainted;
1564 U8 * const r = reghop3((U8*)s, -1, (U8*)PL_bostr);
1565 tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, UTF8_ALLOW_DEFAULT);
1567 tmp = ((OP(c) == NBOUND ?
1568 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1569 LOAD_UTF8_CHARCLASS_ALNUM();
1570 REXEC_FBC_UTF8_SCAN(
1571 if (tmp == !(OP(c) == NBOUND ?
1572 cBOOL(swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target)) :
1573 isALNUM_LC_utf8((U8*)s)))
1575 else REXEC_FBC_TRYIT;
1579 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1580 tmp = cBOOL((OP(c) == NBOUNDL)
1582 : (isWORDCHAR_L1(tmp)
1583 && (isASCII(tmp) || (FLAGS(c) & USE_UNI))));
1588 : (isWORDCHAR_L1((U8) *s)
1589 && (isASCII((U8) *s) || (FLAGS(c) & USE_UNI)))))
1593 else REXEC_FBC_TRYIT;
1596 if ((!prog->minlen && !tmp) && (!reginfo || regtry(reginfo, &s)))
1600 REXEC_FBC_CSCAN_PRELOAD(
1601 LOAD_UTF8_CHARCLASS_PERL_WORD(),
1602 swash_fetch(RE_utf8_perl_word, (U8*)s, utf8_target),
1603 (FLAGS(c) & USE_UNI) ? isWORDCHAR_L1((U8) *s) : isALNUM(*s)
1606 REXEC_FBC_CSCAN_TAINT(
1607 isALNUM_LC_utf8((U8*)s),
1611 REXEC_FBC_CSCAN_PRELOAD(
1612 LOAD_UTF8_CHARCLASS_PERL_WORD(),
1613 !swash_fetch(RE_utf8_perl_word, (U8*)s, utf8_target),
1614 ! ((FLAGS(c) & USE_UNI) ? isWORDCHAR_L1((U8) *s) : isALNUM(*s))
1617 REXEC_FBC_CSCAN_TAINT(
1618 !isALNUM_LC_utf8((U8*)s),
1622 REXEC_FBC_CSCAN_PRELOAD(
1623 LOAD_UTF8_CHARCLASS_PERL_SPACE(),
1624 *s == ' ' || swash_fetch(RE_utf8_perl_space,(U8*)s, utf8_target),
1625 isSPACE_L1((U8) *s) && (isASCII((U8) *s) || (FLAGS(c) & USE_UNI))
1628 REXEC_FBC_CSCAN_TAINT(
1629 *s == ' ' || isSPACE_LC_utf8((U8*)s),
1633 REXEC_FBC_CSCAN_PRELOAD(
1634 LOAD_UTF8_CHARCLASS_PERL_SPACE(),
1635 !(*s == ' ' || swash_fetch(RE_utf8_perl_space,(U8*)s, utf8_target)),
1636 !(isSPACE_L1((U8) *s) && (isASCII((U8) *s) || (FLAGS(c) & USE_UNI)))
1639 REXEC_FBC_CSCAN_TAINT(
1640 !(*s == ' ' || isSPACE_LC_utf8((U8*)s)),
1644 REXEC_FBC_CSCAN_PRELOAD(
1645 LOAD_UTF8_CHARCLASS_POSIX_DIGIT(),
1646 swash_fetch(RE_utf8_posix_digit,(U8*)s, utf8_target),
1650 REXEC_FBC_CSCAN_TAINT(
1651 isDIGIT_LC_utf8((U8*)s),
1655 REXEC_FBC_CSCAN_PRELOAD(
1656 LOAD_UTF8_CHARCLASS_POSIX_DIGIT(),
1657 !swash_fetch(RE_utf8_posix_digit,(U8*)s, utf8_target),
1661 REXEC_FBC_CSCAN_TAINT(
1662 !isDIGIT_LC_utf8((U8*)s),
1668 is_LNBREAK_latin1(s)
1678 !is_VERTWS_latin1(s)
1683 is_HORIZWS_latin1(s)
1687 !is_HORIZWS_utf8(s),
1688 !is_HORIZWS_latin1(s)
1694 /* what trie are we using right now */
1696 = (reg_ac_data*)progi->data->data[ ARG( c ) ];
1698 = (reg_trie_data*)progi->data->data[ aho->trie ];
1699 HV *widecharmap = MUTABLE_HV(progi->data->data[ aho->trie + 1 ]);
1701 const char *last_start = strend - trie->minlen;
1703 const char *real_start = s;
1705 STRLEN maxlen = trie->maxlen;
1707 U8 **points; /* map of where we were in the input string
1708 when reading a given char. For ASCII this
1709 is unnecessary overhead as the relationship
1710 is always 1:1, but for Unicode, especially
1711 case folded Unicode this is not true. */
1712 U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ];
1716 GET_RE_DEBUG_FLAGS_DECL;
1718 /* We can't just allocate points here. We need to wrap it in
1719 * an SV so it gets freed properly if there is a croak while
1720 * running the match */
1723 sv_points=newSV(maxlen * sizeof(U8 *));
1724 SvCUR_set(sv_points,
1725 maxlen * sizeof(U8 *));
1726 SvPOK_on(sv_points);
1727 sv_2mortal(sv_points);
1728 points=(U8**)SvPV_nolen(sv_points );
1729 if ( trie_type != trie_utf8_fold
1730 && (trie->bitmap || OP(c)==AHOCORASICKC) )
1733 bitmap=(U8*)trie->bitmap;
1735 bitmap=(U8*)ANYOF_BITMAP(c);
1737 /* this is the Aho-Corasick algorithm modified a touch
1738 to include special handling for long "unknown char"
1739 sequences. The basic idea being that we use AC as long
1740 as we are dealing with a possible matching char, when
1741 we encounter an unknown char (and we have not encountered
1742 an accepting state) we scan forward until we find a legal
1744 AC matching is basically that of trie matching, except
1745 that when we encounter a failing transition, we fall back
1746 to the current states "fail state", and try the current char
1747 again, a process we repeat until we reach the root state,
1748 state 1, or a legal transition. If we fail on the root state
1749 then we can either terminate if we have reached an accepting
1750 state previously, or restart the entire process from the beginning
1754 while (s <= last_start) {
1755 const U32 uniflags = UTF8_ALLOW_DEFAULT;
1763 U8 *uscan = (U8*)NULL;
1764 U8 *leftmost = NULL;
1766 U32 accepted_word= 0;
1770 while ( state && uc <= (U8*)strend ) {
1772 U32 word = aho->states[ state ].wordnum;
1776 DEBUG_TRIE_EXECUTE_r(
1777 if ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) {
1778 dump_exec_pos( (char *)uc, c, strend, real_start,
1779 (char *)uc, utf8_target );
1780 PerlIO_printf( Perl_debug_log,
1781 " Scanning for legal start char...\n");
1785 while ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) {
1789 while ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) {
1795 if (uc >(U8*)last_start) break;
1799 U8 *lpos= points[ (pointpos - trie->wordinfo[word].len) % maxlen ];
1800 if (!leftmost || lpos < leftmost) {
1801 DEBUG_r(accepted_word=word);
1807 points[pointpos++ % maxlen]= uc;
1808 REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc,
1809 uscan, len, uvc, charid, foldlen,
1811 DEBUG_TRIE_EXECUTE_r({
1812 dump_exec_pos( (char *)uc, c, strend, real_start,
1814 PerlIO_printf(Perl_debug_log,
1815 " Charid:%3u CP:%4"UVxf" ",
1821 word = aho->states[ state ].wordnum;
1823 base = aho->states[ state ].trans.base;
1825 DEBUG_TRIE_EXECUTE_r({
1827 dump_exec_pos( (char *)uc, c, strend, real_start,
1829 PerlIO_printf( Perl_debug_log,
1830 "%sState: %4"UVxf", word=%"UVxf,
1831 failed ? " Fail transition to " : "",
1832 (UV)state, (UV)word);
1838 ( ((offset = base + charid
1839 - 1 - trie->uniquecharcount)) >= 0)
1840 && ((U32)offset < trie->lasttrans)
1841 && trie->trans[offset].check == state
1842 && (tmp=trie->trans[offset].next))
1844 DEBUG_TRIE_EXECUTE_r(
1845 PerlIO_printf( Perl_debug_log," - legal\n"));
1850 DEBUG_TRIE_EXECUTE_r(
1851 PerlIO_printf( Perl_debug_log," - fail\n"));
1853 state = aho->fail[state];
1857 /* we must be accepting here */
1858 DEBUG_TRIE_EXECUTE_r(
1859 PerlIO_printf( Perl_debug_log," - accepting\n"));
1868 if (!state) state = 1;
1871 if ( aho->states[ state ].wordnum ) {
1872 U8 *lpos = points[ (pointpos - trie->wordinfo[aho->states[ state ].wordnum].len) % maxlen ];
1873 if (!leftmost || lpos < leftmost) {
1874 DEBUG_r(accepted_word=aho->states[ state ].wordnum);
1879 s = (char*)leftmost;
1880 DEBUG_TRIE_EXECUTE_r({
1882 Perl_debug_log,"Matches word #%"UVxf" at position %"IVdf". Trying full pattern...\n",
1883 (UV)accepted_word, (IV)(s - real_start)
1886 if (!reginfo || regtry(reginfo, &s)) {
1892 DEBUG_TRIE_EXECUTE_r({
1893 PerlIO_printf( Perl_debug_log,"Pattern failed. Looking for new start point...\n");
1896 DEBUG_TRIE_EXECUTE_r(
1897 PerlIO_printf( Perl_debug_log,"No match.\n"));
1906 Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c));
1916 - regexec_flags - match a regexp against a string
1919 Perl_regexec_flags(pTHX_ REGEXP * const rx, char *stringarg, register char *strend,
1920 char *strbeg, I32 minend, SV *sv, void *data, U32 flags)
1921 /* strend: pointer to null at end of string */
1922 /* strbeg: real beginning of string */
1923 /* minend: end of match must be >=minend after stringarg. */
1924 /* data: May be used for some additional optimizations.
1925 Currently its only used, with a U32 cast, for transmitting
1926 the ganch offset when doing a /g match. This will change */
1927 /* nosave: For optimizations. */
1930 struct regexp *const prog = (struct regexp *)SvANY(rx);
1931 /*register*/ char *s;
1932 register regnode *c;
1933 /*register*/ char *startpos = stringarg;
1934 I32 minlen; /* must match at least this many chars */
1935 I32 dontbother = 0; /* how many characters not to try at end */
1936 I32 end_shift = 0; /* Same for the end. */ /* CC */
1937 I32 scream_pos = -1; /* Internal iterator of scream. */
1938 char *scream_olds = NULL;
1939 const bool utf8_target = cBOOL(DO_UTF8(sv));
1941 RXi_GET_DECL(prog,progi);
1942 regmatch_info reginfo; /* create some info to pass to regtry etc */
1943 regexp_paren_pair *swap = NULL;
1944 GET_RE_DEBUG_FLAGS_DECL;
1946 PERL_ARGS_ASSERT_REGEXEC_FLAGS;
1947 PERL_UNUSED_ARG(data);
1949 /* Be paranoid... */
1950 if (prog == NULL || startpos == NULL) {
1951 Perl_croak(aTHX_ "NULL regexp parameter");
1955 multiline = prog->extflags & RXf_PMf_MULTILINE;
1956 reginfo.prog = rx; /* Yes, sorry that this is confusing. */
1958 RX_MATCH_UTF8_set(rx, utf8_target);
1960 debug_start_match(rx, utf8_target, startpos, strend,
1964 minlen = prog->minlen;
1966 if (strend - startpos < (minlen+(prog->check_offset_min<0?prog->check_offset_min:0))) {
1967 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
1968 "String too short [regexec_flags]...\n"));
1973 /* Check validity of program. */
1974 if (UCHARAT(progi->program) != REG_MAGIC) {
1975 Perl_croak(aTHX_ "corrupted regexp program");
1979 PL_reg_eval_set = 0;
1983 PL_reg_flags |= RF_utf8;
1985 /* Mark beginning of line for ^ and lookbehind. */
1986 reginfo.bol = startpos; /* XXX not used ??? */
1990 /* Mark end of line for $ (and such) */
1993 /* see how far we have to get to not match where we matched before */
1994 reginfo.till = startpos+minend;
1996 /* If there is a "must appear" string, look for it. */
1999 if (prog->extflags & RXf_GPOS_SEEN) { /* Need to set reginfo->ganch */
2001 if (flags & REXEC_IGNOREPOS){ /* Means: check only at start */
2002 reginfo.ganch = startpos + prog->gofs;
2003 DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log,
2004 "GPOS IGNOREPOS: reginfo.ganch = startpos + %"UVxf"\n",(UV)prog->gofs));
2005 } else if (sv && SvTYPE(sv) >= SVt_PVMG
2007 && (mg = mg_find(sv, PERL_MAGIC_regex_global))
2008 && mg->mg_len >= 0) {
2009 reginfo.ganch = strbeg + mg->mg_len; /* Defined pos() */
2010 DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log,
2011 "GPOS MAGIC: reginfo.ganch = strbeg + %"IVdf"\n",(IV)mg->mg_len));
2013 if (prog->extflags & RXf_ANCH_GPOS) {
2014 if (s > reginfo.ganch)
2016 s = reginfo.ganch - prog->gofs;
2017 DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log,
2018 "GPOS ANCH_GPOS: s = ganch - %"UVxf"\n",(UV)prog->gofs));
2024 reginfo.ganch = strbeg + PTR2UV(data);
2025 DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log,
2026 "GPOS DATA: reginfo.ganch= strbeg + %"UVxf"\n",PTR2UV(data)));
2028 } else { /* pos() not defined */
2029 reginfo.ganch = strbeg;
2030 DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log,
2031 "GPOS: reginfo.ganch = strbeg\n"));
2034 if (PL_curpm && (PM_GETRE(PL_curpm) == rx)) {
2035 /* We have to be careful. If the previous successful match
2036 was from this regex we don't want a subsequent partially
2037 successful match to clobber the old results.
2038 So when we detect this possibility we add a swap buffer
2039 to the re, and switch the buffer each match. If we fail
2040 we switch it back, otherwise we leave it swapped.
2043 /* do we need a save destructor here for eval dies? */
2044 Newxz(prog->offs, (prog->nparens + 1), regexp_paren_pair);
2046 if (!(flags & REXEC_CHECKED) && (prog->check_substr != NULL || prog->check_utf8 != NULL)) {
2047 re_scream_pos_data d;
2049 d.scream_olds = &scream_olds;
2050 d.scream_pos = &scream_pos;
2051 s = re_intuit_start(rx, sv, s, strend, flags, &d);
2053 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Not present...\n"));
2054 goto phooey; /* not present */
2060 /* Simplest case: anchored match need be tried only once. */
2061 /* [unless only anchor is BOL and multiline is set] */
2062 if (prog->extflags & (RXf_ANCH & ~RXf_ANCH_GPOS)) {
2063 if (s == startpos && regtry(®info, &startpos))
2065 else if (multiline || (prog->intflags & PREGf_IMPLICIT)
2066 || (prog->extflags & RXf_ANCH_MBOL)) /* XXXX SBOL? */
2071 dontbother = minlen - 1;
2072 end = HOP3c(strend, -dontbother, strbeg) - 1;
2073 /* for multiline we only have to try after newlines */
2074 if (prog->check_substr || prog->check_utf8) {
2075 /* because of the goto we can not easily reuse the macros for bifurcating the
2076 unicode/non-unicode match modes here like we do elsewhere - demerphq */
2079 goto after_try_utf8;
2081 if (regtry(®info, &s)) {
2088 if (prog->extflags & RXf_USE_INTUIT) {
2089 s = re_intuit_start(rx, sv, s + UTF8SKIP(s), strend, flags, NULL);
2098 } /* end search for check string in unicode */
2100 if (s == startpos) {
2101 goto after_try_latin;
2104 if (regtry(®info, &s)) {
2111 if (prog->extflags & RXf_USE_INTUIT) {
2112 s = re_intuit_start(rx, sv, s + 1, strend, flags, NULL);
2121 } /* end search for check string in latin*/
2122 } /* end search for check string */
2123 else { /* search for newline */
2125 /*XXX: The s-- is almost definitely wrong here under unicode - demeprhq*/
2128 /* We can use a more efficient search as newlines are the same in unicode as they are in latin */
2130 if (*s++ == '\n') { /* don't need PL_utf8skip here */
2131 if (regtry(®info, &s))
2135 } /* end search for newline */
2136 } /* end anchored/multiline check string search */
2138 } else if (RXf_GPOS_CHECK == (prog->extflags & RXf_GPOS_CHECK))
2140 /* the warning about reginfo.ganch being used without intialization
2141 is bogus -- we set it above, when prog->extflags & RXf_GPOS_SEEN
2142 and we only enter this block when the same bit is set. */
2143 char *tmp_s = reginfo.ganch - prog->gofs;
2145 if (tmp_s >= strbeg && regtry(®info, &tmp_s))
2150 /* Messy cases: unanchored match. */
2151 if ((prog->anchored_substr || prog->anchored_utf8) && prog->intflags & PREGf_SKIP) {
2152 /* we have /x+whatever/ */
2153 /* it must be a one character string (XXXX Except UTF_PATTERN?) */
2158 if (!(utf8_target ? prog->anchored_utf8 : prog->anchored_substr))
2159 utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog);
2160 ch = SvPVX_const(utf8_target ? prog->anchored_utf8 : prog->anchored_substr)[0];
2165 DEBUG_EXECUTE_r( did_match = 1 );
2166 if (regtry(®info, &s)) goto got_it;
2168 while (s < strend && *s == ch)
2176 DEBUG_EXECUTE_r( did_match = 1 );
2177 if (regtry(®info, &s)) goto got_it;
2179 while (s < strend && *s == ch)
2184 DEBUG_EXECUTE_r(if (!did_match)
2185 PerlIO_printf(Perl_debug_log,
2186 "Did not find anchored character...\n")
2189 else if (prog->anchored_substr != NULL
2190 || prog->anchored_utf8 != NULL
2191 || ((prog->float_substr != NULL || prog->float_utf8 != NULL)
2192 && prog->float_max_offset < strend - s)) {
2197 char *last1; /* Last position checked before */
2201 if (prog->anchored_substr || prog->anchored_utf8) {
2202 if (!(utf8_target ? prog->anchored_utf8 : prog->anchored_substr))
2203 utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog);
2204 must = utf8_target ? prog->anchored_utf8 : prog->anchored_substr;
2205 back_max = back_min = prog->anchored_offset;
2207 if (!(utf8_target ? prog->float_utf8 : prog->float_substr))
2208 utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog);
2209 must = utf8_target ? prog->float_utf8 : prog->float_substr;
2210 back_max = prog->float_max_offset;
2211 back_min = prog->float_min_offset;
2215 if (must == &PL_sv_undef)
2216 /* could not downgrade utf8 check substring, so must fail */
2222 last = HOP3c(strend, /* Cannot start after this */
2223 -(I32)(CHR_SVLEN(must)
2224 - (SvTAIL(must) != 0) + back_min), strbeg);
2227 last1 = HOPc(s, -1);
2229 last1 = s - 1; /* bogus */
2231 /* XXXX check_substr already used to find "s", can optimize if
2232 check_substr==must. */
2234 dontbother = end_shift;
2235 strend = HOPc(strend, -dontbother);
2236 while ( (s <= last) &&
2237 ((flags & REXEC_SCREAM)
2238 ? (s = screaminstr(sv, must, HOP3c(s, back_min, (back_min<0 ? strbeg : strend)) - strbeg,
2239 end_shift, &scream_pos, 0))
2240 : (s = fbm_instr((unsigned char*)HOP3(s, back_min, (back_min<0 ? strbeg : strend)),
2241 (unsigned char*)strend, must,
2242 multiline ? FBMrf_MULTILINE : 0))) ) {
2243 /* we may be pointing at the wrong string */
2244 if ((flags & REXEC_SCREAM) && RXp_MATCH_COPIED(prog))
2245 s = strbeg + (s - SvPVX_const(sv));
2246 DEBUG_EXECUTE_r( did_match = 1 );
2247 if (HOPc(s, -back_max) > last1) {
2248 last1 = HOPc(s, -back_min);
2249 s = HOPc(s, -back_max);
2252 char * const t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1;
2254 last1 = HOPc(s, -back_min);
2258 while (s <= last1) {
2259 if (regtry(®info, &s))
2265 while (s <= last1) {
2266 if (regtry(®info, &s))
2272 DEBUG_EXECUTE_r(if (!did_match) {
2273 RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0),
2274 SvPVX_const(must), RE_SV_DUMPLEN(must), 30);
2275 PerlIO_printf(Perl_debug_log, "Did not find %s substr %s%s...\n",
2276 ((must == prog->anchored_substr || must == prog->anchored_utf8)
2277 ? "anchored" : "floating"),
2278 quoted, RE_SV_TAIL(must));
2282 else if ( (c = progi->regstclass) ) {
2284 const OPCODE op = OP(progi->regstclass);
2285 /* don't bother with what can't match */
2286 if (PL_regkind[op] != EXACT && op != CANY && PL_regkind[op] != TRIE)
2287 strend = HOPc(strend, -(minlen - 1));
2290 SV * const prop = sv_newmortal();
2291 regprop(prog, prop, c);
2293 RE_PV_QUOTED_DECL(quoted,utf8_target,PERL_DEBUG_PAD_ZERO(1),
2295 PerlIO_printf(Perl_debug_log,
2296 "Matching stclass %.*s against %s (%d bytes)\n",
2297 (int)SvCUR(prop), SvPVX_const(prop),
2298 quoted, (int)(strend - s));
2301 if (find_byclass(prog, c, s, strend, ®info))
2303 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass... [regexec_flags]\n"));
2307 if (prog->float_substr != NULL || prog->float_utf8 != NULL) {
2312 if (!(utf8_target ? prog->float_utf8 : prog->float_substr))
2313 utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog);
2314 float_real = utf8_target ? prog->float_utf8 : prog->float_substr;
2316 if (flags & REXEC_SCREAM) {
2317 last = screaminstr(sv, float_real, s - strbeg,
2318 end_shift, &scream_pos, 1); /* last one */
2320 last = scream_olds; /* Only one occurrence. */
2321 /* we may be pointing at the wrong string */
2322 else if (RXp_MATCH_COPIED(prog))
2323 s = strbeg + (s - SvPVX_const(sv));
2327 const char * const little = SvPV_const(float_real, len);
2329 if (SvTAIL(float_real)) {
2330 if (memEQ(strend - len + 1, little, len - 1))
2331 last = strend - len + 1;
2332 else if (!multiline)
2333 last = memEQ(strend - len, little, len)
2334 ? strend - len : NULL;
2340 last = rninstr(s, strend, little, little + len);
2342 last = strend; /* matching "$" */
2347 PerlIO_printf(Perl_debug_log,
2348 "%sCan't trim the tail, match fails (should not happen)%s\n",
2349 PL_colors[4], PL_colors[5]));
2350 goto phooey; /* Should not happen! */
2352 dontbother = strend - last + prog->float_min_offset;
2354 if (minlen && (dontbother < minlen))
2355 dontbother = minlen - 1;
2356 strend -= dontbother; /* this one's always in bytes! */
2357 /* We don't know much -- general case. */
2360 if (regtry(®info, &s))
2369 if (regtry(®info, &s))
2371 } while (s++ < strend);
2380 RX_MATCH_TAINTED_set(rx, PL_reg_flags & RF_tainted);
2382 if (PL_reg_eval_set)
2383 restore_pos(aTHX_ prog);
2384 if (RXp_PAREN_NAMES(prog))
2385 (void)hv_iterinit(RXp_PAREN_NAMES(prog));
2387 /* make sure $`, $&, $', and $digit will work later */
2388 if ( !(flags & REXEC_NOT_FIRST) ) {
2389 RX_MATCH_COPY_FREE(rx);
2390 if (flags & REXEC_COPY_STR) {
2391 const I32 i = PL_regeol - startpos + (stringarg - strbeg);
2392 #ifdef PERL_OLD_COPY_ON_WRITE
2394 || (SvFLAGS(sv) & CAN_COW_MASK) == CAN_COW_FLAGS)) {
2396 PerlIO_printf(Perl_debug_log,
2397 "Copy on write: regexp capture, type %d\n",
2400 prog->saved_copy = sv_setsv_cow(prog->saved_copy, sv);
2401 prog->subbeg = (char *)SvPVX_const(prog->saved_copy);
2402 assert (SvPOKp(prog->saved_copy));
2406 RX_MATCH_COPIED_on(rx);
2407 s = savepvn(strbeg, i);
2413 prog->subbeg = strbeg;
2414 prog->sublen = PL_regeol - strbeg; /* strend may have been modified */
2421 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n",
2422 PL_colors[4], PL_colors[5]));
2423 if (PL_reg_eval_set)
2424 restore_pos(aTHX_ prog);
2426 /* we failed :-( roll it back */
2427 Safefree(prog->offs);
2436 - regtry - try match at specific point
2438 STATIC I32 /* 0 failure, 1 success */
2439 S_regtry(pTHX_ regmatch_info *reginfo, char **startpos)
2443 REGEXP *const rx = reginfo->prog;
2444 regexp *const prog = (struct regexp *)SvANY(rx);
2445 RXi_GET_DECL(prog,progi);
2446 GET_RE_DEBUG_FLAGS_DECL;
2448 PERL_ARGS_ASSERT_REGTRY;
2450 reginfo->cutpoint=NULL;
2452 if ((prog->extflags & RXf_EVAL_SEEN) && !PL_reg_eval_set) {
2455 PL_reg_eval_set = RS_init;
2456 DEBUG_EXECUTE_r(DEBUG_s(
2457 PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %"IVdf"\n",
2458 (IV)(PL_stack_sp - PL_stack_base));
2461 cxstack[cxstack_ix].blk_oldsp = PL_stack_sp - PL_stack_base;
2462 /* Otherwise OP_NEXTSTATE will free whatever on stack now. */
2464 /* Apparently this is not needed, judging by wantarray. */
2465 /* SAVEI8(cxstack[cxstack_ix].blk_gimme);
2466 cxstack[cxstack_ix].blk_gimme = G_SCALAR; */
2469 /* Make $_ available to executed code. */
2470 if (reginfo->sv != DEFSV) {
2472 DEFSV_set(reginfo->sv);
2475 if (!(SvTYPE(reginfo->sv) >= SVt_PVMG && SvMAGIC(reginfo->sv)
2476 && (mg = mg_find(reginfo->sv, PERL_MAGIC_regex_global)))) {
2477 /* prepare for quick setting of pos */
2478 #ifdef PERL_OLD_COPY_ON_WRITE
2479 if (SvIsCOW(reginfo->sv))
2480 sv_force_normal_flags(reginfo->sv, 0);
2482 mg = sv_magicext(reginfo->sv, NULL, PERL_MAGIC_regex_global,
2483 &PL_vtbl_mglob, NULL, 0);
2487 PL_reg_oldpos = mg->mg_len;
2488 SAVEDESTRUCTOR_X(restore_pos, prog);
2490 if (!PL_reg_curpm) {
2491 Newxz(PL_reg_curpm, 1, PMOP);
2494 SV* const repointer = &PL_sv_undef;
2495 /* this regexp is also owned by the new PL_reg_curpm, which
2496 will try to free it. */
2497 av_push(PL_regex_padav, repointer);
2498 PL_reg_curpm->op_pmoffset = av_len(PL_regex_padav);
2499 PL_regex_pad = AvARRAY(PL_regex_padav);
2504 /* It seems that non-ithreads works both with and without this code.
2505 So for efficiency reasons it seems best not to have the code
2506 compiled when it is not needed. */
2507 /* This is safe against NULLs: */
2508 ReREFCNT_dec(PM_GETRE(PL_reg_curpm));
2509 /* PM_reg_curpm owns a reference to this regexp. */
2512 PM_SETRE(PL_reg_curpm, rx);
2513 PL_reg_oldcurpm = PL_curpm;
2514 PL_curpm = PL_reg_curpm;
2515 if (RXp_MATCH_COPIED(prog)) {
2516 /* Here is a serious problem: we cannot rewrite subbeg,
2517 since it may be needed if this match fails. Thus
2518 $` inside (?{}) could fail... */
2519 PL_reg_oldsaved = prog->subbeg;
2520 PL_reg_oldsavedlen = prog->sublen;
2521 #ifdef PERL_OLD_COPY_ON_WRITE
2522 PL_nrs = prog->saved_copy;
2524 RXp_MATCH_COPIED_off(prog);
2527 PL_reg_oldsaved = NULL;
2528 prog->subbeg = PL_bostr;
2529 prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */
2531 DEBUG_EXECUTE_r(PL_reg_starttry = *startpos);
2532 prog->offs[0].start = *startpos - PL_bostr;
2533 PL_reginput = *startpos;
2534 PL_reglastparen = &prog->lastparen;
2535 PL_reglastcloseparen = &prog->lastcloseparen;
2536 prog->lastparen = 0;
2537 prog->lastcloseparen = 0;
2539 PL_regoffs = prog->offs;
2540 if (PL_reg_start_tmpl <= prog->nparens) {
2541 PL_reg_start_tmpl = prog->nparens*3/2 + 3;
2542 if(PL_reg_start_tmp)
2543 Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
2545 Newx(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
2548 /* XXXX What this code is doing here?!!! There should be no need
2549 to do this again and again, PL_reglastparen should take care of
2552 /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code.
2553 * Actually, the code in regcppop() (which Ilya may be meaning by
2554 * PL_reglastparen), is not needed at all by the test suite
2555 * (op/regexp, op/pat, op/split), but that code is needed otherwise
2556 * this erroneously leaves $1 defined: "1" =~ /^(?:(\d)x)?\d$/
2557 * Meanwhile, this code *is* needed for the
2558 * above-mentioned test suite tests to succeed. The common theme
2559 * on those tests seems to be returning null fields from matches.
2560 * --jhi updated by dapm */
2562 if (prog->nparens) {
2563 regexp_paren_pair *pp = PL_regoffs;
2565 for (i = prog->nparens; i > (I32)*PL_reglastparen; i--) {
2573 if (regmatch(reginfo, progi->program + 1)) {
2574 PL_regoffs[0].end = PL_reginput - PL_bostr;
2577 if (reginfo->cutpoint)
2578 *startpos= reginfo->cutpoint;
2579 REGCP_UNWIND(lastcp);
2584 #define sayYES goto yes
2585 #define sayNO goto no
2586 #define sayNO_SILENT goto no_silent
2588 /* we dont use STMT_START/END here because it leads to
2589 "unreachable code" warnings, which are bogus, but distracting. */
2590 #define CACHEsayNO \
2591 if (ST.cache_mask) \
2592 PL_reg_poscache[ST.cache_offset] |= ST.cache_mask; \
2595 /* this is used to determine how far from the left messages like
2596 'failed...' are printed. It should be set such that messages
2597 are inline with the regop output that created them.
2599 #define REPORT_CODE_OFF 32
2602 #define CHRTEST_UNINIT -1001 /* c1/c2 haven't been calculated yet */
2603 #define CHRTEST_VOID -1000 /* the c1/c2 "next char" test should be skipped */
2605 #define SLAB_FIRST(s) (&(s)->states[0])
2606 #define SLAB_LAST(s) (&(s)->states[PERL_REGMATCH_SLAB_SLOTS-1])
2608 /* grab a new slab and return the first slot in it */
2610 STATIC regmatch_state *
2613 #if PERL_VERSION < 9 && !defined(PERL_CORE)
2616 regmatch_slab *s = PL_regmatch_slab->next;
2618 Newx(s, 1, regmatch_slab);
2619 s->prev = PL_regmatch_slab;
2621 PL_regmatch_slab->next = s;
2623 PL_regmatch_slab = s;
2624 return SLAB_FIRST(s);
2628 /* push a new state then goto it */
2630 #define PUSH_STATE_GOTO(state, node) \
2632 st->resume_state = state; \
2635 /* push a new state with success backtracking, then goto it */
2637 #define PUSH_YES_STATE_GOTO(state, node) \
2639 st->resume_state = state; \
2640 goto push_yes_state;
2646 regmatch() - main matching routine
2648 This is basically one big switch statement in a loop. We execute an op,
2649 set 'next' to point the next op, and continue. If we come to a point which
2650 we may need to backtrack to on failure such as (A|B|C), we push a
2651 backtrack state onto the backtrack stack. On failure, we pop the top
2652 state, and re-enter the loop at the state indicated. If there are no more
2653 states to pop, we return failure.
2655 Sometimes we also need to backtrack on success; for example /A+/, where
2656 after successfully matching one A, we need to go back and try to
2657 match another one; similarly for lookahead assertions: if the assertion
2658 completes successfully, we backtrack to the state just before the assertion
2659 and then carry on. In these cases, the pushed state is marked as
2660 'backtrack on success too'. This marking is in fact done by a chain of
2661 pointers, each pointing to the previous 'yes' state. On success, we pop to
2662 the nearest yes state, discarding any intermediate failure-only states.
2663 Sometimes a yes state is pushed just to force some cleanup code to be
2664 called at the end of a successful match or submatch; e.g. (??{$re}) uses
2665 it to free the inner regex.
2667 Note that failure backtracking rewinds the cursor position, while
2668 success backtracking leaves it alone.
2670 A pattern is complete when the END op is executed, while a subpattern
2671 such as (?=foo) is complete when the SUCCESS op is executed. Both of these
2672 ops trigger the "pop to last yes state if any, otherwise return true"
2675 A common convention in this function is to use A and B to refer to the two
2676 subpatterns (or to the first nodes thereof) in patterns like /A*B/: so A is
2677 the subpattern to be matched possibly multiple times, while B is the entire
2678 rest of the pattern. Variable and state names reflect this convention.
2680 The states in the main switch are the union of ops and failure/success of
2681 substates associated with with that op. For example, IFMATCH is the op
2682 that does lookahead assertions /(?=A)B/ and so the IFMATCH state means
2683 'execute IFMATCH'; while IFMATCH_A is a state saying that we have just
2684 successfully matched A and IFMATCH_A_fail is a state saying that we have
2685 just failed to match A. Resume states always come in pairs. The backtrack
2686 state we push is marked as 'IFMATCH_A', but when that is popped, we resume
2687 at IFMATCH_A or IFMATCH_A_fail, depending on whether we are backtracking
2688 on success or failure.
2690 The struct that holds a backtracking state is actually a big union, with
2691 one variant for each major type of op. The variable st points to the
2692 top-most backtrack struct. To make the code clearer, within each
2693 block of code we #define ST to alias the relevant union.
2695 Here's a concrete example of a (vastly oversimplified) IFMATCH
2701 #define ST st->u.ifmatch
2703 case IFMATCH: // we are executing the IFMATCH op, (?=A)B
2704 ST.foo = ...; // some state we wish to save
2706 // push a yes backtrack state with a resume value of
2707 // IFMATCH_A/IFMATCH_A_fail, then continue execution at the
2709 PUSH_YES_STATE_GOTO(IFMATCH_A, A);
2712 case IFMATCH_A: // we have successfully executed A; now continue with B
2714 bar = ST.foo; // do something with the preserved value
2717 case IFMATCH_A_fail: // A failed, so the assertion failed
2718 ...; // do some housekeeping, then ...
2719 sayNO; // propagate the failure
2726 For any old-timers reading this who are familiar with the old recursive
2727 approach, the code above is equivalent to:
2729 case IFMATCH: // we are executing the IFMATCH op, (?=A)B
2738 ...; // do some housekeeping, then ...
2739 sayNO; // propagate the failure
2742 The topmost backtrack state, pointed to by st, is usually free. If you
2743 want to claim it, populate any ST.foo fields in it with values you wish to
2744 save, then do one of
2746 PUSH_STATE_GOTO(resume_state, node);
2747 PUSH_YES_STATE_GOTO(resume_state, node);
2749 which sets that backtrack state's resume value to 'resume_state', pushes a
2750 new free entry to the top of the backtrack stack, then goes to 'node'.
2751 On backtracking, the free slot is popped, and the saved state becomes the
2752 new free state. An ST.foo field in this new top state can be temporarily
2753 accessed to retrieve values, but once the main loop is re-entered, it
2754 becomes available for reuse.
2756 Note that the depth of the backtrack stack constantly increases during the
2757 left-to-right execution of the pattern, rather than going up and down with
2758 the pattern nesting. For example the stack is at its maximum at Z at the
2759 end of the pattern, rather than at X in the following:
2761 /(((X)+)+)+....(Y)+....Z/
2763 The only exceptions to this are lookahead/behind assertions and the cut,
2764 (?>A), which pop all the backtrack states associated with A before
2767 Bascktrack state structs are allocated in slabs of about 4K in size.
2768 PL_regmatch_state and st always point to the currently active state,
2769 and PL_regmatch_slab points to the slab currently containing
2770 PL_regmatch_state. The first time regmatch() is called, the first slab is
2771 allocated, and is never freed until interpreter destruction. When the slab
2772 is full, a new one is allocated and chained to the end. At exit from
2773 regmatch(), slabs allocated since entry are freed.
2778 #define DEBUG_STATE_pp(pp) \
2780 DUMP_EXEC_POS(locinput, scan, utf8_target); \
2781 PerlIO_printf(Perl_debug_log, \
2782 " %*s"pp" %s%s%s%s%s\n", \
2784 PL_reg_name[st->resume_state], \
2785 ((st==yes_state||st==mark_state) ? "[" : ""), \
2786 ((st==yes_state) ? "Y" : ""), \
2787 ((st==mark_state) ? "M" : ""), \
2788 ((st==yes_state||st==mark_state) ? "]" : "") \
2793 #define REG_NODE_NUM(x) ((x) ? (int)((x)-prog) : -1)
2798 S_debug_start_match(pTHX_ const REGEXP *prog, const bool utf8_target,
2799 const char *start, const char *end, const char *blurb)
2801 const bool utf8_pat = RX_UTF8(prog) ? 1 : 0;
2803 PERL_ARGS_ASSERT_DEBUG_START_MATCH;
2808 RE_PV_QUOTED_DECL(s0, utf8_pat, PERL_DEBUG_PAD_ZERO(0),
2809 RX_PRECOMP_const(prog), RX_PRELEN(prog), 60);
2811 RE_PV_QUOTED_DECL(s1, utf8_target, PERL_DEBUG_PAD_ZERO(1),
2812 start, end - start, 60);
2814 PerlIO_printf(Perl_debug_log,
2815 "%s%s REx%s %s against %s\n",
2816 PL_colors[4], blurb, PL_colors[5], s0, s1);
2818 if (utf8_target||utf8_pat)
2819 PerlIO_printf(Perl_debug_log, "UTF-8 %s%s%s...\n",
2820 utf8_pat ? "pattern" : "",
2821 utf8_pat && utf8_target ? " and " : "",
2822 utf8_target ? "string" : ""
2828 S_dump_exec_pos(pTHX_ const char *locinput,
2829 const regnode *scan,
2830 const char *loc_regeol,
2831 const char *loc_bostr,
2832 const char *loc_reg_starttry,
2833 const bool utf8_target)
2835 const int docolor = *PL_colors[0] || *PL_colors[2] || *PL_colors[4];
2836 const int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */
2837 int l = (loc_regeol - locinput) > taill ? taill : (loc_regeol - locinput);
2838 /* The part of the string before starttry has one color
2839 (pref0_len chars), between starttry and current
2840 position another one (pref_len - pref0_len chars),
2841 after the current position the third one.
2842 We assume that pref0_len <= pref_len, otherwise we
2843 decrease pref0_len. */
2844 int pref_len = (locinput - loc_bostr) > (5 + taill) - l
2845 ? (5 + taill) - l : locinput - loc_bostr;
2848 PERL_ARGS_ASSERT_DUMP_EXEC_POS;
2850 while (utf8_target && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len)))
2852 pref0_len = pref_len - (locinput - loc_reg_starttry);
2853 if (l + pref_len < (5 + taill) && l < loc_regeol - locinput)
2854 l = ( loc_regeol - locinput > (5 + taill) - pref_len
2855 ? (5 + taill) - pref_len : loc_regeol - locinput);
2856 while (utf8_target && UTF8_IS_CONTINUATION(*(U8*)(locinput + l)))
2860 if (pref0_len > pref_len)
2861 pref0_len = pref_len;
2863 const int is_uni = (utf8_target && OP(scan) != CANY) ? 1 : 0;
2865 RE_PV_COLOR_DECL(s0,len0,is_uni,PERL_DEBUG_PAD(0),
2866 (locinput - pref_len),pref0_len, 60, 4, 5);
2868 RE_PV_COLOR_DECL(s1,len1,is_uni,PERL_DEBUG_PAD(1),
2869 (locinput - pref_len + pref0_len),
2870 pref_len - pref0_len, 60, 2, 3);
2872 RE_PV_COLOR_DECL(s2,len2,is_uni,PERL_DEBUG_PAD(2),
2873 locinput, loc_regeol - locinput, 10, 0, 1);
2875 const STRLEN tlen=len0+len1+len2;
2876 PerlIO_printf(Perl_debug_log,
2877 "%4"IVdf" <%.*s%.*s%s%.*s>%*s|",
2878 (IV)(locinput - loc_bostr),
2881 (docolor ? "" : "> <"),
2883 (int)(tlen > 19 ? 0 : 19 - tlen),
2890 /* reg_check_named_buff_matched()
2891 * Checks to see if a named buffer has matched. The data array of
2892 * buffer numbers corresponding to the buffer is expected to reside
2893 * in the regexp->data->data array in the slot stored in the ARG() of
2894 * node involved. Note that this routine doesn't actually care about the
2895 * name, that information is not preserved from compilation to execution.
2896 * Returns the index of the leftmost defined buffer with the given name
2897 * or 0 if non of the buffers matched.
2900 S_reg_check_named_buff_matched(pTHX_ const regexp *rex, const regnode *scan)
2903 RXi_GET_DECL(rex,rexi);
2904 SV *sv_dat= MUTABLE_SV(rexi->data->data[ ARG( scan ) ]);
2905 I32 *nums=(I32*)SvPVX(sv_dat);
2907 PERL_ARGS_ASSERT_REG_CHECK_NAMED_BUFF_MATCHED;
2909 for ( n=0; n<SvIVX(sv_dat); n++ ) {
2910 if ((I32)*PL_reglastparen >= nums[n] &&
2911 PL_regoffs[nums[n]].end != -1)
2920 /* free all slabs above current one - called during LEAVE_SCOPE */
2923 S_clear_backtrack_stack(pTHX_ void *p)
2925 regmatch_slab *s = PL_regmatch_slab->next;
2930 PL_regmatch_slab->next = NULL;
2932 regmatch_slab * const osl = s;
2939 #define SETREX(Re1,Re2) \
2940 if (PL_reg_eval_set) PM_SETRE((PL_reg_curpm), (Re2)); \
2943 STATIC I32 /* 0 failure, 1 success */
2944 S_regmatch(pTHX_ regmatch_info *reginfo, regnode *prog)
2946 #if PERL_VERSION < 9 && !defined(PERL_CORE)
2950 register const bool utf8_target = PL_reg_match_utf8;
2951 const U32 uniflags = UTF8_ALLOW_DEFAULT;
2952 REGEXP *rex_sv = reginfo->prog;
2953 regexp *rex = (struct regexp *)SvANY(rex_sv);
2954 RXi_GET_DECL(rex,rexi);
2956 /* the current state. This is a cached copy of PL_regmatch_state */
2957 register regmatch_state *st;
2958 /* cache heavy used fields of st in registers */
2959 register regnode *scan;
2960 register regnode *next;
2961 register U32 n = 0; /* general value; init to avoid compiler warning */
2962 register I32 ln = 0; /* len or last; init to avoid compiler warning */
2963 register char *locinput = PL_reginput;
2964 register I32 nextchr; /* is always set to UCHARAT(locinput) */
2966 bool result = 0; /* return value of S_regmatch */
2967 int depth = 0; /* depth of backtrack stack */
2968 U32 nochange_depth = 0; /* depth of GOSUB recursion with nochange */
2969 const U32 max_nochange_depth =
2970 (3 * rex->nparens > MAX_RECURSE_EVAL_NOCHANGE_DEPTH) ?
2971 3 * rex->nparens : MAX_RECURSE_EVAL_NOCHANGE_DEPTH;
2972 regmatch_state *yes_state = NULL; /* state to pop to on success of
2974 /* mark_state piggy backs on the yes_state logic so that when we unwind
2975 the stack on success we can update the mark_state as we go */
2976 regmatch_state *mark_state = NULL; /* last mark state we have seen */
2977 regmatch_state *cur_eval = NULL; /* most recent EVAL_AB state */
2978 struct regmatch_state *cur_curlyx = NULL; /* most recent curlyx */
2980 bool no_final = 0; /* prevent failure from backtracking? */
2981 bool do_cutgroup = 0; /* no_final only until next branch/trie entry */
2982 char *startpoint = PL_reginput;
2983 SV *popmark = NULL; /* are we looking for a mark? */
2984 SV *sv_commit = NULL; /* last mark name seen in failure */
2985 SV *sv_yes_mark = NULL; /* last mark name we have seen
2986 during a successfull match */
2987 U32 lastopen = 0; /* last open we saw */
2988 bool has_cutgroup = RX_HAS_CUTGROUP(rex) ? 1 : 0;
2989 SV* const oreplsv = GvSV(PL_replgv);
2990 /* these three flags are set by various ops to signal information to
2991 * the very next op. They have a useful lifetime of exactly one loop
2992 * iteration, and are not preserved or restored by state pushes/pops
2994 bool sw = 0; /* the condition value in (?(cond)a|b) */
2995 bool minmod = 0; /* the next "{n,m}" is a "{n,m}?" */
2996 int logical = 0; /* the following EVAL is:
3000 or the following IFMATCH/UNLESSM is:
3001 false: plain (?=foo)
3002 true: used as a condition: (?(?=foo))
3005 GET_RE_DEBUG_FLAGS_DECL;
3008 PERL_ARGS_ASSERT_REGMATCH;
3010 DEBUG_OPTIMISE_r( DEBUG_EXECUTE_r({
3011 PerlIO_printf(Perl_debug_log,"regmatch start\n");
3013 /* on first ever call to regmatch, allocate first slab */
3014 if (!PL_regmatch_slab) {
3015 Newx(PL_regmatch_slab, 1, regmatch_slab);
3016 PL_regmatch_slab->prev = NULL;
3017 PL_regmatch_slab->next = NULL;
3018 PL_regmatch_state = SLAB_FIRST(PL_regmatch_slab);
3021 oldsave = PL_savestack_ix;
3022 SAVEDESTRUCTOR_X(S_clear_backtrack_stack, NULL);
3023 SAVEVPTR(PL_regmatch_slab);
3024 SAVEVPTR(PL_regmatch_state);
3026 /* grab next free state slot */
3027 st = ++PL_regmatch_state;
3028 if (st > SLAB_LAST(PL_regmatch_slab))
3029 st = PL_regmatch_state = S_push_slab(aTHX);
3031 /* Note that nextchr is a byte even in UTF */
3032 nextchr = UCHARAT(locinput);
3034 while (scan != NULL) {
3037 SV * const prop = sv_newmortal();
3038 regnode *rnext=regnext(scan);
3039 DUMP_EXEC_POS( locinput, scan, utf8_target );
3040 regprop(rex, prop, scan);
3042 PerlIO_printf(Perl_debug_log,
3043 "%3"IVdf":%*s%s(%"IVdf")\n",
3044 (IV)(scan - rexi->program), depth*2, "",
3046 (PL_regkind[OP(scan)] == END || !rnext) ?
3047 0 : (IV)(rnext - rexi->program));
3050 next = scan + NEXT_OFF(scan);
3053 state_num = OP(scan);
3057 assert(PL_reglastparen == &rex->lastparen);
3058 assert(PL_reglastcloseparen == &rex->lastcloseparen);
3059 assert(PL_regoffs == rex->offs);
3061 switch (state_num) {
3063 if (locinput == PL_bostr)
3065 /* reginfo->till = reginfo->bol; */
3070 if (locinput == PL_bostr ||
3071 ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n'))
3077 if (locinput == PL_bostr)
3081 if (locinput == reginfo->ganch)
3086 /* update the startpoint */
3087 st->u.keeper.val = PL_regoffs[0].start;
3088 PL_reginput = locinput;
3089 PL_regoffs[0].start = locinput - PL_bostr;
3090 PUSH_STATE_GOTO(KEEPS_next, next);
3092 case KEEPS_next_fail:
3093 /* rollback the start point change */
3094 PL_regoffs[0].start = st->u.keeper.val;
3100 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
3105 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
3107 if (PL_regeol - locinput > 1)
3111 if (PL_regeol != locinput)
3115 if (!nextchr && locinput >= PL_regeol)
3118 locinput += PL_utf8skip[nextchr];
3119 if (locinput > PL_regeol)
3121 nextchr = UCHARAT(locinput);
3124 nextchr = UCHARAT(++locinput);
3127 if (!nextchr && locinput >= PL_regeol)
3129 nextchr = UCHARAT(++locinput);
3132 if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n')
3135 locinput += PL_utf8skip[nextchr];
3136 if (locinput > PL_regeol)
3138 nextchr = UCHARAT(locinput);
3141 nextchr = UCHARAT(++locinput);
3145 #define ST st->u.trie
3147 /* In this case the charclass data is available inline so
3148 we can fail fast without a lot of extra overhead.
3150 if (scan->flags == EXACT || !utf8_target) {
3151 if(!ANYOF_BITMAP_TEST(scan, *locinput)) {
3153 PerlIO_printf(Perl_debug_log,
3154 "%*s %sfailed to match trie start class...%s\n",
3155 REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5])
3163 /* the basic plan of execution of the trie is:
3164 * At the beginning, run though all the states, and
3165 * find the longest-matching word. Also remember the position
3166 * of the shortest matching word. For example, this pattern:
3169 * when matched against the string "abcde", will generate
3170 * accept states for all words except 3, with the longest
3171 * matching word being 4, and the shortest being 1 (with
3172 * the position being after char 1 of the string).
3174 * Then for each matching word, in word order (i.e. 1,2,4,5),
3175 * we run the remainder of the pattern; on each try setting
3176 * the current position to the character following the word,
3177 * returning to try the next word on failure.
3179 * We avoid having to build a list of words at runtime by
3180 * using a compile-time structure, wordinfo[].prev, which
3181 * gives, for each word, the previous accepting word (if any).
3182 * In the case above it would contain the mappings 1->2, 2->0,
3183 * 3->0, 4->5, 5->1. We can use this table to generate, from
3184 * the longest word (4 above), a list of all words, by
3185 * following the list of prev pointers; this gives us the
3186 * unordered list 4,5,1,2. Then given the current word we have
3187 * just tried, we can go through the list and find the
3188 * next-biggest word to try (so if we just failed on word 2,
3189 * the next in the list is 4).
3191 * Since at runtime we don't record the matching position in
3192 * the string for each word, we have to work that out for
3193 * each word we're about to process. The wordinfo table holds
3194 * the character length of each word; given that we recorded
3195 * at the start: the position of the shortest word and its
3196 * length in chars, we just need to move the pointer the
3197 * difference between the two char lengths. Depending on
3198 * Unicode status and folding, that's cheap or expensive.
3200 * This algorithm is optimised for the case where are only a
3201 * small number of accept states, i.e. 0,1, or maybe 2.
3202 * With lots of accepts states, and having to try all of them,
3203 * it becomes quadratic on number of accept states to find all
3208 /* what type of TRIE am I? (utf8 makes this contextual) */
3209 DECL_TRIE_TYPE(scan);
3211 /* what trie are we using right now */
3212 reg_trie_data * const trie
3213 = (reg_trie_data*)rexi->data->data[ ARG( scan ) ];
3214 HV * widecharmap = MUTABLE_HV(rexi->data->data[ ARG( scan ) + 1 ]);
3215 U32 state = trie->startstate;
3217 if (trie->bitmap && trie_type != trie_utf8_fold &&
3218 !TRIE_BITMAP_TEST(trie,*locinput)
3220 if (trie->states[ state ].wordnum) {
3222 PerlIO_printf(Perl_debug_log,
3223 "%*s %smatched empty string...%s\n",
3224 REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5])
3230 PerlIO_printf(Perl_debug_log,
3231 "%*s %sfailed to match trie start class...%s\n",
3232 REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5])
3239 U8 *uc = ( U8* )locinput;
3243 U8 *uscan = (U8*)NULL;
3244 U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ];
3245 U32 charcount = 0; /* how many input chars we have matched */
3246 U32 accepted = 0; /* have we seen any accepting states? */
3249 ST.jump = trie->jump;
3252 ST.longfold = FALSE; /* char longer if folded => it's harder */
3255 /* fully traverse the TRIE; note the position of the
3256 shortest accept state and the wordnum of the longest
3259 while ( state && uc <= (U8*)PL_regeol ) {
3260 U32 base = trie->states[ state ].trans.base;
3264 wordnum = trie->states[ state ].wordnum;
3266 if (wordnum) { /* it's an accept state */
3269 /* record first match position */
3271 ST.firstpos = (U8*)locinput;
3276 ST.firstchars = charcount;
3279 if (!ST.nextword || wordnum < ST.nextword)
3280 ST.nextword = wordnum;
3281 ST.topword = wordnum;
3284 DEBUG_TRIE_EXECUTE_r({
3285 DUMP_EXEC_POS( (char *)uc, scan, utf8_target );
3286 PerlIO_printf( Perl_debug_log,
3287 "%*s %sState: %4"UVxf" Accepted: %c ",
3288 2+depth * 2, "", PL_colors[4],
3289 (UV)state, (accepted ? 'Y' : 'N'));
3292 /* read a char and goto next state */
3295 REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc,
3296 uscan, len, uvc, charid, foldlen,
3303 base + charid - 1 - trie->uniquecharcount)) >= 0)
3305 && ((U32)offset < trie->lasttrans)
3306 && trie->trans[offset].check == state)
3308 state = trie->trans[offset].next;
3319 DEBUG_TRIE_EXECUTE_r(
3320 PerlIO_printf( Perl_debug_log,
3321 "Charid:%3x CP:%4"UVxf" After State: %4"UVxf"%s\n",
3322 charid, uvc, (UV)state, PL_colors[5] );
3328 /* calculate total number of accept states */
3333 w = trie->wordinfo[w].prev;
3336 ST.accepted = accepted;
3340 PerlIO_printf( Perl_debug_log,
3341 "%*s %sgot %"IVdf" possible matches%s\n",
3342 REPORT_CODE_OFF + depth * 2, "",
3343 PL_colors[4], (IV)ST.accepted, PL_colors[5] );
3345 goto trie_first_try; /* jump into the fail handler */
3349 case TRIE_next_fail: /* we failed - try next alternative */
3351 REGCP_UNWIND(ST.cp);
3352 for (n = *PL_reglastparen; n > ST.lastparen; n--)
3353 PL_regoffs[n].end = -1;
3354 *PL_reglastparen = n;
3356 if (!--ST.accepted) {
3358 PerlIO_printf( Perl_debug_log,
3359 "%*s %sTRIE failed...%s\n",
3360 REPORT_CODE_OFF+depth*2, "",
3367 /* Find next-highest word to process. Note that this code
3368 * is O(N^2) per trie run (O(N) per branch), so keep tight */
3369 register U16 min = 0;
3371 register U16 const nextword = ST.nextword;
3372 register reg_trie_wordinfo * const wordinfo
3373 = ((reg_trie_data*)rexi->data->data[ARG(ST.me)])->wordinfo;
3374 for (word=ST.topword; word; word=wordinfo[word].prev) {
3375 if (word > nextword && (!min || word < min))
3388 ST.lastparen = *PL_reglastparen;
3392 /* find start char of end of current word */
3394 U32 chars; /* how many chars to skip */
3395 U8 *uc = ST.firstpos;
3396 reg_trie_data * const trie
3397 = (reg_trie_data*)rexi->data->data[ARG(ST.me)];
3399 assert((trie->wordinfo[ST.nextword].len - trie->prefixlen)
3401 chars = (trie->wordinfo[ST.nextword].len - trie->prefixlen)
3405 /* the hard option - fold each char in turn and find
3406 * its folded length (which may be different */
3407 U8 foldbuf[UTF8_MAXBYTES_CASE + 1];
3415 uvc = utf8n_to_uvuni((U8*)uc, UTF8_MAXLEN, &len,
3423 uvc = to_uni_fold(uvc, foldbuf, &foldlen);
3428 uvc = utf8n_to_uvuni(uscan, UTF8_MAXLEN, &len,
3442 PL_reginput = (char *)uc;
3445 scan = (ST.jump && ST.jump[ST.nextword])
3446 ? ST.me + ST.jump[ST.nextword]
3450 PerlIO_printf( Perl_debug_log,
3451 "%*s %sTRIE matched word #%d, continuing%s\n",
3452 REPORT_CODE_OFF+depth*2, "",
3459 if (ST.accepted > 1 || has_cutgroup) {
3460 PUSH_STATE_GOTO(TRIE_next, scan);
3463 /* only one choice left - just continue */
3465 AV *const trie_words
3466 = MUTABLE_AV(rexi->data->data[ARG(ST.me)+TRIE_WORDS_OFFSET]);
3467 SV ** const tmp = av_fetch( trie_words,
3469 SV *sv= tmp ? sv_newmortal() : NULL;
3471 PerlIO_printf( Perl_debug_log,
3472 "%*s %sonly one match left, short-circuiting: #%d <%s>%s\n",
3473 REPORT_CODE_OFF+depth*2, "", PL_colors[4],
3475 tmp ? pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), 0,
3476 PL_colors[0], PL_colors[1],
3477 (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0)
3479 : "not compiled under -Dr",
3483 locinput = PL_reginput;
3484 nextchr = UCHARAT(locinput);
3485 continue; /* execute rest of RE */
3490 char *s = STRING(scan);
3492 if (utf8_target != UTF_PATTERN) {
3493 /* The target and the pattern have differing utf8ness. */
3495 const char * const e = s + ln;
3498 /* The target is utf8, the pattern is not utf8. */
3503 if (NATIVE_TO_UNI(*(U8*)s) !=
3504 utf8n_to_uvuni((U8*)l, UTF8_MAXBYTES, &ulen,
3512 /* The target is not utf8, the pattern is utf8. */
3517 if (NATIVE_TO_UNI(*((U8*)l)) !=
3518 utf8n_to_uvuni((U8*)s, UTF8_MAXBYTES, &ulen,
3526 nextchr = UCHARAT(locinput);
3529 /* The target and the pattern have the same utf8ness. */
3530 /* Inline the first character, for speed. */
3531 if (UCHARAT(s) != nextchr)
3533 if (PL_regeol - locinput < ln)
3535 if (ln > 1 && memNE(s, locinput, ln))
3538 nextchr = UCHARAT(locinput);
3542 PL_reg_flags |= RF_tainted;
3545 char * const s = STRING(scan);
3548 if (utf8_target || UTF_PATTERN) {
3549 /* Either target or the pattern are utf8. */
3550 const char * const l = locinput;
3551 char *e = PL_regeol;
3553 if (! foldEQ_utf8(s, 0, ln, cBOOL(UTF_PATTERN),
3554 l, &e, 0, utf8_target)) {
3555 /* One more case for the sharp s:
3556 * pack("U0U*", 0xDF) =~ /ss/i,
3557 * the 0xC3 0x9F are the UTF-8
3558 * byte sequence for the U+00DF. */
3560 if (!(utf8_target &&
3561 toLOWER(s[0]) == 's' &&
3563 toLOWER(s[1]) == 's' &&
3570 nextchr = UCHARAT(locinput);
3574 /* Neither the target and the pattern are utf8. */
3576 /* Inline the first character, for speed. */
3577 if (UCHARAT(s) != nextchr &&
3578 UCHARAT(s) != ((OP(scan) == EXACTF)
3579 ? PL_fold : PL_fold_locale)[nextchr])
3581 if (PL_regeol - locinput < ln)
3583 if (ln > 1 && (OP(scan) == EXACTF
3584 ? ! foldEQ(s, locinput, ln)
3585 : ! foldEQ_locale(s, locinput, ln)))
3588 nextchr = UCHARAT(locinput);
3593 PL_reg_flags |= RF_tainted;
3597 /* was last char in word? */
3599 if (locinput == PL_bostr)
3602 const U8 * const r = reghop3((U8*)locinput, -1, (U8*)PL_bostr);
3604 ln = utf8n_to_uvchr(r, UTF8SKIP(r), 0, uniflags);
3606 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
3607 ln = isALNUM_uni(ln);
3608 LOAD_UTF8_CHARCLASS_ALNUM();
3609 n = swash_fetch(PL_utf8_alnum, (U8*)locinput, utf8_target);
3612 ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(ln));
3613 n = isALNUM_LC_utf8((U8*)locinput);
3617 ln = (locinput != PL_bostr) ?
3618 UCHARAT(locinput - 1) : '\n';
3619 if (FLAGS(scan) & USE_UNI) {
3621 /* Here, can't be BOUNDL or NBOUNDL because they never set
3622 * the flags to USE_UNI */
3623 ln = isWORDCHAR_L1(ln);
3624 n = isWORDCHAR_L1(nextchr);
3626 else if (OP(scan) == BOUND || OP(scan) == NBOUND) {
3628 n = isALNUM(nextchr);
3631 ln = isALNUM_LC(ln);
3632 n = isALNUM_LC(nextchr);
3635 if (((!ln) == (!n)) == (OP(scan) == BOUND ||
3636 OP(scan) == BOUNDL))
3641 STRLEN inclasslen = PL_regeol - locinput;
3642 if (locinput >= PL_regeol)
3645 if (!reginclass(rex, scan, (U8*)locinput, &inclasslen, utf8_target))
3647 locinput += inclasslen;
3648 nextchr = UCHARAT(locinput);
3653 nextchr = UCHARAT(locinput);
3654 if (!nextchr && locinput >= PL_regeol)
3656 if (!REGINCLASS(rex, scan, (U8*)locinput))
3658 nextchr = UCHARAT(++locinput);
3662 /* If we might have the case of the German sharp s
3663 * in a casefolding Unicode character class. */
3665 if (ANYOF_FOLD_SHARP_S(scan, locinput, PL_regeol)) {
3666 locinput += SHARP_S_SKIP;
3667 nextchr = UCHARAT(locinput);
3672 /* Special char classes - The defines start on line 129 or so */
3673 CCC_TRY_AFF_U( ALNUM, ALNUML, perl_word, "a", isALNUM_LC_utf8, isWORDCHAR_L1, isALNUM_LC);
3674 CCC_TRY_NEG_U(NALNUM, NALNUML, perl_word, "a", isALNUM_LC_utf8, isWORDCHAR_L1, isALNUM_LC);
3676 CCC_TRY_AFF_U( SPACE, SPACEL, perl_space, " ", isSPACE_LC_utf8, isSPACE_L1, isSPACE_LC);
3677 CCC_TRY_NEG_U(NSPACE, NSPACEL, perl_space, " ", isSPACE_LC_utf8, isSPACE_L1, isSPACE_LC);
3679 CCC_TRY_AFF( DIGIT, DIGITL, posix_digit, "0", isDIGIT_LC_utf8, isDIGIT, isDIGIT_LC);
3680 CCC_TRY_NEG(NDIGIT, NDIGITL, posix_digit, "0", isDIGIT_LC_utf8, isDIGIT, isDIGIT_LC);
3682 case CLUMP: /* Match \X: logical Unicode character. This is defined as
3683 a Unicode extended Grapheme Cluster */
3684 /* From http://www.unicode.org/reports/tr29 (5.2 version). An
3685 extended Grapheme Cluster is:
3688 | Prepend* Begin Extend*
3691 Begin is (Hangul-syllable | ! Control)
3692 Extend is (Grapheme_Extend | Spacing_Mark)
3693 Control is [ GCB_Control CR LF ]
3695 The discussion below shows how the code for CLUMP is derived
3696 from this regex. Note that most of these concepts are from
3697 property values of the Grapheme Cluster Boundary (GCB) property.
3698 No code point can have multiple property values for a given
3699 property. Thus a code point in Prepend can't be in Control, but
3700 it must be in !Control. This is why Control above includes
3701 GCB_Control plus CR plus LF. The latter two are used in the GCB
3702 property separately, and so can't be in GCB_Control, even though
3703 they logically are controls. Control is not the same as gc=cc,
3704 but includes format and other characters as well.
3706 The Unicode definition of Hangul-syllable is:
3708 | (L* ( ( V | LV ) V* | LVT ) T*)
3711 Each of these is a value for the GCB property, and hence must be
3712 disjoint, so the order they are tested is immaterial, so the
3713 above can safely be changed to
3716 | (L* ( LVT | ( V | LV ) V*) T*)
3718 The last two terms can be combined like this:
3720 | (( LVT | ( V | LV ) V*) T*))
3722 And refactored into this:
3723 L* (L | LVT T* | V V* T* | LV V* T*)
3725 That means that if we have seen any L's at all we can quit
3726 there, but if the next character is a LVT, a V or and LV we
3729 There is a subtlety with Prepend* which showed up in testing.
3730 Note that the Begin, and only the Begin is required in:
3731 | Prepend* Begin Extend*
3732 Also, Begin contains '! Control'. A Prepend must be a '!
3733 Control', which means it must be a Begin. What it comes down to
3734 is that if we match Prepend* and then find no suitable Begin
3735 afterwards, that if we backtrack the last Prepend, that one will
3736 be a suitable Begin.
3739 if (locinput >= PL_regeol)
3741 if (! utf8_target) {
3743 /* Match either CR LF or '.', as all the other possibilities
3745 locinput++; /* Match the . or CR */
3747 && locinput < PL_regeol
3748 && UCHARAT(locinput) == '\n') locinput++;
3752 /* Utf8: See if is ( CR LF ); already know that locinput <
3753 * PL_regeol, so locinput+1 is in bounds */
3754 if (nextchr == '\r' && UCHARAT(locinput + 1) == '\n') {
3758 /* In case have to backtrack to beginning, then match '.' */
3759 char *starting = locinput;
3761 /* In case have to backtrack the last prepend */
3762 char *previous_prepend = 0;
3764 LOAD_UTF8_CHARCLASS_GCB();
3766 /* Match (prepend)* */
3767 while (locinput < PL_regeol
3768 && swash_fetch(PL_utf8_X_prepend,
3769 (U8*)locinput, utf8_target))
3771 previous_prepend = locinput;
3772 locinput += UTF8SKIP(locinput);
3775 /* As noted above, if we matched a prepend character, but
3776 * the next thing won't match, back off the last prepend we
3777 * matched, as it is guaranteed to match the begin */
3778 if (previous_prepend
3779 && (locinput >= PL_regeol
3780 || ! swash_fetch(PL_utf8_X_begin,
3781 (U8*)locinput, utf8_target)))
3783 locinput = previous_prepend;
3786 /* Note that here we know PL_regeol > locinput, as we
3787 * tested that upon input to this switch case, and if we
3788 * moved locinput forward, we tested the result just above
3789 * and it either passed, or we backed off so that it will
3791 if (! swash_fetch(PL_utf8_X_begin, (U8*)locinput, utf8_target)) {
3793 /* Here did not match the required 'Begin' in the
3794 * second term. So just match the very first
3795 * character, the '.' of the final term of the regex */
3796 locinput = starting + UTF8SKIP(starting);
3799 /* Here is the beginning of a character that can have
3800 * an extender. It is either a hangul syllable, or a
3802 if (swash_fetch(PL_utf8_X_non_hangul,
3803 (U8*)locinput, utf8_target))
3806 /* Here not a Hangul syllable, must be a
3807 * ('! * Control') */
3808 locinput += UTF8SKIP(locinput);
3811 /* Here is a Hangul syllable. It can be composed
3812 * of several individual characters. One
3813 * possibility is T+ */
3814 if (swash_fetch(PL_utf8_X_T,
3815 (U8*)locinput, utf8_target))
3817 while (locinput < PL_regeol
3818 && swash_fetch(PL_utf8_X_T,
3819 (U8*)locinput, utf8_target))
3821 locinput += UTF8SKIP(locinput);
3825 /* Here, not T+, but is a Hangul. That means
3826 * it is one of the others: L, LV, LVT or V,
3828 * L* (L | LVT T* | V V* T* | LV V* T*) */
3831 while (locinput < PL_regeol
3832 && swash_fetch(PL_utf8_X_L,
3833 (U8*)locinput, utf8_target))
3835 locinput += UTF8SKIP(locinput);
3838 /* Here, have exhausted L*. If the next
3839 * character is not an LV, LVT nor V, it means
3840 * we had to have at least one L, so matches L+
3841 * in the original equation, we have a complete
3842 * hangul syllable. Are done. */
3844 if (locinput < PL_regeol
3845 && swash_fetch(PL_utf8_X_LV_LVT_V,
3846 (U8*)locinput, utf8_target))
3849 /* Otherwise keep going. Must be LV, LVT
3850 * or V. See if LVT */
3851 if (swash_fetch(PL_utf8_X_LVT,
3852 (U8*)locinput, utf8_target))
3854 locinput += UTF8SKIP(locinput);
3857 /* Must be V or LV. Take it, then
3859 locinput += UTF8SKIP(locinput);
3860 while (locinput < PL_regeol
3861 && swash_fetch(PL_utf8_X_V,
3862 (U8*)locinput, utf8_target))
3864 locinput += UTF8SKIP(locinput);
3868 /* And any of LV, LVT, or V can be followed
3870 while (locinput < PL_regeol
3871 && swash_fetch(PL_utf8_X_T,
3875 locinput += UTF8SKIP(locinput);
3881 /* Match any extender */
3882 while (locinput < PL_regeol
3883 && swash_fetch(PL_utf8_X_extend,
3884 (U8*)locinput, utf8_target))
3886 locinput += UTF8SKIP(locinput);
3890 if (locinput > PL_regeol) sayNO;
3892 nextchr = UCHARAT(locinput);
3899 PL_reg_flags |= RF_tainted;
3904 n = reg_check_named_buff_matched(rex,scan);
3907 type = REF + ( type - NREF );
3914 PL_reg_flags |= RF_tainted;
3918 n = ARG(scan); /* which paren pair */
3921 ln = PL_regoffs[n].start;
3922 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
3923 if (*PL_reglastparen < n || ln == -1)
3924 sayNO; /* Do not match unless seen CLOSEn. */
3925 if (ln == PL_regoffs[n].end)
3929 if (utf8_target && type != REF) { /* REF can do byte comparison */
3931 const char *e = PL_bostr + PL_regoffs[n].end;
3933 * Note that we can't do the "other character" lookup trick as
3934 * in the 8-bit case (no pun intended) because in Unicode we
3935 * have to map both upper and title case to lower case.
3939 STRLEN ulen1, ulen2;
3940 U8 tmpbuf1[UTF8_MAXBYTES_CASE+1];
3941 U8 tmpbuf2[UTF8_MAXBYTES_CASE+1];
3945 toLOWER_utf8((U8*)s, tmpbuf1, &ulen1);
3946 toLOWER_utf8((U8*)l, tmpbuf2, &ulen2);
3947 if (ulen1 != ulen2 || memNE((char *)tmpbuf1, (char *)tmpbuf2, ulen1))
3954 nextchr = UCHARAT(locinput);
3958 /* Inline the first character, for speed. */
3959 if (UCHARAT(s) != nextchr &&
3961 (UCHARAT(s) != (type == REFF
3962 ? PL_fold : PL_fold_locale)[nextchr])))
3964 ln = PL_regoffs[n].end - ln;
3965 if (locinput + ln > PL_regeol)
3967 if (ln > 1 && (type == REF
3968 ? memNE(s, locinput, ln)
3970 ? ! foldEQ(s, locinput, ln)
3971 : ! foldEQ_locale(s, locinput, ln))))
3974 nextchr = UCHARAT(locinput);
3984 #define ST st->u.eval
3989 regexp_internal *rei;
3990 regnode *startpoint;
3993 case GOSUB: /* /(...(?1))/ /(...(?&foo))/ */
3994 if (cur_eval && cur_eval->locinput==locinput) {
3995 if (cur_eval->u.eval.close_paren == (U32)ARG(scan))
3996 Perl_croak(aTHX_ "Infinite recursion in regex");
3997 if ( ++nochange_depth > max_nochange_depth )
3999 "Pattern subroutine nesting without pos change"
4000 " exceeded limit in regex");
4007 (void)ReREFCNT_inc(rex_sv);
4008 if (OP(scan)==GOSUB) {
4009 startpoint = scan + ARG2L(scan);
4010 ST.close_paren = ARG(scan);
4012 startpoint = rei->program+1;
4015 goto eval_recurse_doit;
4017 case EVAL: /* /(?{A})B/ /(??{A})B/ and /(?(?{A})X|Y)B/ */
4018 if (cur_eval && cur_eval->locinput==locinput) {
4019 if ( ++nochange_depth > max_nochange_depth )
4020 Perl_croak(aTHX_ "EVAL without pos change exceeded limit in regex");
4025 /* execute the code in the {...} */
4027 SV ** const before = SP;
4028 OP_4tree * const oop = PL_op;
4029 COP * const ocurcop = PL_curcop;
4031 char *saved_regeol = PL_regeol;
4032 struct re_save_state saved_state;
4034 /* To not corrupt the existing regex state while executing the
4035 * eval we would normally put it on the save stack, like with
4036 * save_re_context. However, re-evals have a weird scoping so we
4037 * can't just add ENTER/LEAVE here. With that, things like
4039 * (?{$a=2})(a(?{local$a=$a+1}))*aak*c(?{$b=$a})
4041 * would break, as they expect the localisation to be unwound
4042 * only when the re-engine backtracks through the bit that
4045 * What we do instead is just saving the state in a local c
4048 Copy(&PL_reg_state, &saved_state, 1, struct re_save_state);
4051 PL_op = (OP_4tree*)rexi->data->data[n];
4052 DEBUG_STATE_r( PerlIO_printf(Perl_debug_log,
4053 " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) );
4054 PAD_SAVE_LOCAL(old_comppad, (PAD*)rexi->data->data[n + 2]);
4055 PL_regoffs[0].end = PL_reg_magic->mg_len = locinput - PL_bostr;
4058 SV *sv_mrk = get_sv("REGMARK", 1);
4059 sv_setsv(sv_mrk, sv_yes_mark);
4062 CALLRUNOPS(aTHX); /* Scalar context. */
4065 ret = &PL_sv_undef; /* protect against empty (?{}) blocks. */
4071 Copy(&saved_state, &PL_reg_state, 1, struct re_save_state);
4074 PAD_RESTORE_LOCAL(old_comppad);
4075 PL_curcop = ocurcop;
4076 PL_regeol = saved_regeol;
4079 sv_setsv(save_scalar(PL_replgv), ret);
4083 if (logical == 2) { /* Postponed subexpression: /(??{...})/ */
4086 /* extract RE object from returned value; compiling if
4092 SV *const sv = SvRV(ret);
4094 if (SvTYPE(sv) == SVt_REGEXP) {
4096 } else if (SvSMAGICAL(sv)) {
4097 mg = mg_find(sv, PERL_MAGIC_qr);
4100 } else if (SvTYPE(ret) == SVt_REGEXP) {
4102 } else if (SvSMAGICAL(ret)) {
4103 if (SvGMAGICAL(ret)) {
4104 /* I don't believe that there is ever qr magic
4106 assert(!mg_find(ret, PERL_MAGIC_qr));
4107 sv_unmagic(ret, PERL_MAGIC_qr);
4110 mg = mg_find(ret, PERL_MAGIC_qr);
4111 /* testing suggests mg only ends up non-NULL for
4112 scalars who were upgraded and compiled in the
4113 else block below. In turn, this is only
4114 triggered in the "postponed utf8 string" tests
4120 rx = (REGEXP *) mg->mg_obj; /*XXX:dmq*/
4124 rx = reg_temp_copy(NULL, rx);
4128 const I32 osize = PL_regsize;
4131 assert (SvUTF8(ret));
4132 } else if (SvUTF8(ret)) {
4133 /* Not doing UTF-8, despite what the SV says. Is
4134 this only if we're trapped in use 'bytes'? */
4135 /* Make a copy of the octet sequence, but without
4136 the flag on, as the compiler now honours the
4137 SvUTF8 flag on ret. */
4139 const char *const p = SvPV(ret, len);
4140 ret = newSVpvn_flags(p, len, SVs_TEMP);
4142 rx = CALLREGCOMP(ret, pm_flags);
4144 & (SVs_TEMP | SVs_PADTMP | SVf_READONLY
4146 /* This isn't a first class regexp. Instead, it's
4147 caching a regexp onto an existing, Perl visible
4149 sv_magic(ret, MUTABLE_SV(rx), PERL_MAGIC_qr, 0, 0);
4154 re = (struct regexp *)SvANY(rx);
4156 RXp_MATCH_COPIED_off(re);
4157 re->subbeg = rex->subbeg;
4158 re->sublen = rex->sublen;
4161 debug_start_match(re_sv, utf8_target, locinput, PL_regeol,
4162 "Matching embedded");
4164 startpoint = rei->program + 1;
4165 ST.close_paren = 0; /* only used for GOSUB */
4166 /* borrowed from regtry */
4167 if (PL_reg_start_tmpl <= re->nparens) {
4168 PL_reg_start_tmpl = re->nparens*3/2 + 3;
4169 if(PL_reg_start_tmp)
4170 Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
4172 Newx(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
4175 eval_recurse_doit: /* Share code with GOSUB below this line */
4176 /* run the pattern returned from (??{...}) */
4177 ST.cp = regcppush(0); /* Save *all* the positions. */
4178 REGCP_SET(ST.lastcp);
4180 PL_regoffs = re->offs; /* essentially NOOP on GOSUB */
4182 /* see regtry, specifically PL_reglast(?:close)?paren is a pointer! (i dont know why) :dmq */
4183 PL_reglastparen = &re->lastparen;
4184 PL_reglastcloseparen = &re->lastcloseparen;
4186 re->lastcloseparen = 0;
4188 PL_reginput = locinput;
4191 /* XXXX This is too dramatic a measure... */
4194 ST.toggle_reg_flags = PL_reg_flags;
4196 PL_reg_flags |= RF_utf8;
4198 PL_reg_flags &= ~RF_utf8;
4199 ST.toggle_reg_flags ^= PL_reg_flags; /* diff of old and new */
4201 ST.prev_rex = rex_sv;
4202 ST.prev_curlyx = cur_curlyx;
4203 SETREX(rex_sv,re_sv);
4208 ST.prev_eval = cur_eval;
4210 /* now continue from first node in postoned RE */
4211 PUSH_YES_STATE_GOTO(EVAL_AB, startpoint);
4214 /* logical is 1, /(?(?{...})X|Y)/ */
4215 sw = cBOOL(SvTRUE(ret));
4220 case EVAL_AB: /* cleanup after a successful (??{A})B */
4221 /* note: this is called twice; first after popping B, then A */
4222 PL_reg_flags ^= ST.toggle_reg_flags;
4223 ReREFCNT_dec(rex_sv);
4224 SETREX(rex_sv,ST.prev_rex);
4225 rex = (struct regexp *)SvANY(rex_sv);
4226 rexi = RXi_GET(rex);
4228 cur_eval = ST.prev_eval;
4229 cur_curlyx = ST.prev_curlyx;
4231 /* rex was changed so update the pointer in PL_reglastparen and PL_reglastcloseparen */
4232 PL_reglastparen = &rex->lastparen;
4233 PL_reglastcloseparen = &rex->lastcloseparen;
4234 /* also update PL_regoffs */
4235 PL_regoffs = rex->offs;
4237 /* XXXX This is too dramatic a measure... */
4239 if ( nochange_depth )
4244 case EVAL_AB_fail: /* unsuccessfully ran A or B in (??{A})B */
4245 /* note: this is called twice; first after popping B, then A */
4246 PL_reg_flags ^= ST.toggle_reg_flags;
4247 ReREFCNT_dec(rex_sv);
4248 SETREX(rex_sv,ST.prev_rex);
4249 rex = (struct regexp *)SvANY(rex_sv);
4250 rexi = RXi_GET(rex);
4251 /* rex was changed so update the pointer in PL_reglastparen and PL_reglastcloseparen */
4252 PL_reglastparen = &rex->lastparen;
4253 PL_reglastcloseparen = &rex->lastcloseparen;
4255 PL_reginput = locinput;
4256 REGCP_UNWIND(ST.lastcp);
4258 cur_eval = ST.prev_eval;
4259 cur_curlyx = ST.prev_curlyx;
4260 /* XXXX This is too dramatic a measure... */
4262 if ( nochange_depth )
4268 n = ARG(scan); /* which paren pair */
4269 PL_reg_start_tmp[n] = locinput;
4275 n = ARG(scan); /* which paren pair */
4276 PL_regoffs[n].start = PL_reg_start_tmp[n] - PL_bostr;
4277 PL_regoffs[n].end = locinput - PL_bostr;
4278 /*if (n > PL_regsize)
4280 if (n > *PL_reglastparen)
4281 *PL_reglastparen = n;
4282 *PL_reglastcloseparen = n;
4283 if (cur_eval && cur_eval->u.eval.close_paren == n) {
4291 cursor && OP(cursor)!=END;
4292 cursor=regnext(cursor))
4294 if ( OP(cursor)==CLOSE ){
4296 if ( n <= lastopen ) {
4298 = PL_reg_start_tmp[n] - PL_bostr;
4299 PL_regoffs[n].end = locinput - PL_bostr;
4300 /*if (n > PL_regsize)
4302 if (n > *PL_reglastparen)
4303 *PL_reglastparen = n;
4304 *PL_reglastcloseparen = n;
4305 if ( n == ARG(scan) || (cur_eval &&
4306 cur_eval->u.eval.close_paren == n))
4315 n = ARG(scan); /* which paren pair */
4316 sw = cBOOL(*PL_reglastparen >= n && PL_regoffs[n].end != -1);
4319 /* reg_check_named_buff_matched returns 0 for no match */
4320 sw = cBOOL(0 < reg_check_named_buff_matched(rex,scan));
4324 sw = (cur_eval && (!n || cur_eval->u.eval.close_paren == n));
4330 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
4332 next = NEXTOPER(NEXTOPER(scan));
4334 next = scan + ARG(scan);
4335 if (OP(next) == IFTHEN) /* Fake one. */
4336 next = NEXTOPER(NEXTOPER(next));
4340 logical = scan->flags;
4343 /*******************************************************************
4345 The CURLYX/WHILEM pair of ops handle the most generic case of the /A*B/
4346 pattern, where A and B are subpatterns. (For simple A, CURLYM or
4347 STAR/PLUS/CURLY/CURLYN are used instead.)
4349 A*B is compiled as <CURLYX><A><WHILEM><B>
4351 On entry to the subpattern, CURLYX is called. This pushes a CURLYX
4352 state, which contains the current count, initialised to -1. It also sets
4353 cur_curlyx to point to this state, with any previous value saved in the
4356 CURLYX then jumps straight to the WHILEM op, rather than executing A,
4357 since the pattern may possibly match zero times (i.e. it's a while {} loop
4358 rather than a do {} while loop).
4360 Each entry to WHILEM represents a successful match of A. The count in the
4361 CURLYX block is incremented, another WHILEM state is pushed, and execution
4362 passes to A or B depending on greediness and the current count.
4364 For example, if matching against the string a1a2a3b (where the aN are
4365 substrings that match /A/), then the match progresses as follows: (the
4366 pushed states are interspersed with the bits of strings matched so far):
4369 <CURLYX cnt=0><WHILEM>
4370 <CURLYX cnt=1><WHILEM> a1 <WHILEM>
4371 <CURLYX cnt=2><WHILEM> a1 <WHILEM> a2 <WHILEM>
4372 <CURLYX cnt=3><WHILEM> a1 <WHILEM> a2 <WHILEM> a3 <WHILEM>
4373 <CURLYX cnt=3><WHILEM> a1 <WHILEM> a2 <WHILEM> a3 <WHILEM> b
4375 (Contrast this with something like CURLYM, which maintains only a single
4379 a1 <CURLYM cnt=1> a2
4380 a1 a2 <CURLYM cnt=2> a3
4381 a1 a2 a3 <CURLYM cnt=3> b
4384 Each WHILEM state block marks a point to backtrack to upon partial failure
4385 of A or B, and also contains some minor state data related to that
4386 iteration. The CURLYX block, pointed to by cur_curlyx, contains the
4387 overall state, such as the count, and pointers to the A and B ops.
4389 This is complicated slightly by nested CURLYX/WHILEM's. Since cur_curlyx
4390 must always point to the *current* CURLYX block, the rules are:
4392 When executing CURLYX, save the old cur_curlyx in the CURLYX state block,
4393 and set cur_curlyx to point the new block.
4395 When popping the CURLYX block after a successful or unsuccessful match,
4396 restore the previous cur_curlyx.
4398 When WHILEM is about to execute B, save the current cur_curlyx, and set it
4399 to the outer one saved in the CURLYX block.
4401 When popping the WHILEM block after a successful or unsuccessful B match,
4402 restore the previous cur_curlyx.
4404 Here's an example for the pattern (AI* BI)*BO
4405 I and O refer to inner and outer, C and W refer to CURLYX and WHILEM:
4408 curlyx backtrack stack
4409 ------ ---------------
4411 CO <CO prev=NULL> <WO>
4412 CI <CO prev=NULL> <WO> <CI prev=CO> <WI> ai
4413 CO <CO prev=NULL> <WO> <CI prev=CO> <WI> ai <WI prev=CI> bi
4414 NULL <CO prev=NULL> <WO> <CI prev=CO> <WI> ai <WI prev=CI> bi <WO prev=CO> bo
4416 At this point the pattern succeeds, and we work back down the stack to
4417 clean up, restoring as we go:
4419 CO <CO prev=NULL> <WO> <CI prev=CO> <WI> ai <WI prev=CI> bi
4420 CI <CO prev=NULL> <WO> <CI prev=CO> <WI> ai
4421 CO <CO prev=NULL> <WO>
4424 *******************************************************************/
4426 #define ST st->u.curlyx
4428 case CURLYX: /* start of /A*B/ (for complex A) */
4430 /* No need to save/restore up to this paren */
4431 I32 parenfloor = scan->flags;
4433 assert(next); /* keep Coverity happy */
4434 if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */
4437 /* XXXX Probably it is better to teach regpush to support
4438 parenfloor > PL_regsize... */
4439 if (parenfloor > (I32)*PL_reglastparen)
4440 parenfloor = *PL_reglastparen; /* Pessimization... */
4442 ST.prev_curlyx= cur_curlyx;
4444 ST.cp = PL_savestack_ix;
4446 /* these fields contain the state of the current curly.
4447 * they are accessed by subsequent WHILEMs */
4448 ST.parenfloor = parenfloor;
4453 ST.count = -1; /* this will be updated by WHILEM */
4454 ST.lastloc = NULL; /* this will be updated by WHILEM */
4456 PL_reginput = locinput;
4457 PUSH_YES_STATE_GOTO(CURLYX_end, PREVOPER(next));
4461 case CURLYX_end: /* just finished matching all of A*B */
4462 cur_curlyx = ST.prev_curlyx;
4466 case CURLYX_end_fail: /* just failed to match all of A*B */
4468 cur_curlyx = ST.prev_curlyx;
4474 #define ST st->u.whilem
4476 case WHILEM: /* just matched an A in /A*B/ (for complex A) */
4478 /* see the discussion above about CURLYX/WHILEM */
4480 int min = ARG1(cur_curlyx->u.curlyx.me);
4481 int max = ARG2(cur_curlyx->u.curlyx.me);
4482 regnode *A = NEXTOPER(cur_curlyx->u.curlyx.me) + EXTRA_STEP_2ARGS;
4484 assert(cur_curlyx); /* keep Coverity happy */
4485 n = ++cur_curlyx->u.curlyx.count; /* how many A's matched */
4486 ST.save_lastloc = cur_curlyx->u.curlyx.lastloc;
4487 ST.cache_offset = 0;
4490 PL_reginput = locinput;
4492 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
4493 "%*s whilem: matched %ld out of %d..%d\n",
4494 REPORT_CODE_OFF+depth*2, "", (long)n, min, max)
4497 /* First just match a string of min A's. */
4500 cur_curlyx->u.curlyx.lastloc = locinput;
4501 PUSH_STATE_GOTO(WHILEM_A_pre, A);
4505 /* If degenerate A matches "", assume A done. */
4507 if (locinput == cur_curlyx->u.curlyx.lastloc) {
4508 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
4509 "%*s whilem: empty match detected, trying continuation...\n",
4510 REPORT_CODE_OFF+depth*2, "")
4512 goto do_whilem_B_max;
4515 /* super-linear cache processing */
4519 if (!PL_reg_maxiter) {
4520 /* start the countdown: Postpone detection until we
4521 * know the match is not *that* much linear. */
4522 PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4);
4523 /* possible overflow for long strings and many CURLYX's */
4524 if (PL_reg_maxiter < 0)
4525 PL_reg_maxiter = I32_MAX;
4526 PL_reg_leftiter = PL_reg_maxiter;
4529 if (PL_reg_leftiter-- == 0) {
4530 /* initialise cache */
4531 const I32 size = (PL_reg_maxiter + 7)/8;
4532 if (PL_reg_poscache) {
4533 if ((I32)PL_reg_poscache_size < size) {
4534 Renew(PL_reg_poscache, size, char);
4535 PL_reg_poscache_size = size;
4537 Zero(PL_reg_poscache, size, char);
4540 PL_reg_poscache_size = size;
4541 Newxz(PL_reg_poscache, size, char);
4543 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
4544 "%swhilem: Detected a super-linear match, switching on caching%s...\n",
4545 PL_colors[4], PL_colors[5])
4549 if (PL_reg_leftiter < 0) {
4550 /* have we already failed at this position? */
4552 offset = (scan->flags & 0xf) - 1
4553 + (locinput - PL_bostr) * (scan->flags>>4);
4554 mask = 1 << (offset % 8);
4556 if (PL_reg_poscache[offset] & mask) {
4557 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
4558 "%*s whilem: (cache) already tried at this position...\n",
4559 REPORT_CODE_OFF+depth*2, "")
4561 sayNO; /* cache records failure */
4563 ST.cache_offset = offset;
4564 ST.cache_mask = mask;
4568 /* Prefer B over A for minimal matching. */
4570 if (cur_curlyx->u.curlyx.minmod) {
4571 ST.save_curlyx = cur_curlyx;
4572 cur_curlyx = cur_curlyx->u.curlyx.prev_curlyx;
4573 ST.cp = regcppush(ST.save_curlyx->u.curlyx.parenfloor);
4574 REGCP_SET(ST.lastcp);
4575 PUSH_YES_STATE_GOTO(WHILEM_B_min, ST.save_curlyx->u.curlyx.B);
4579 /* Prefer A over B for maximal matching. */
4581 if (n < max) { /* More greed allowed? */
4582 ST.cp = regcppush(cur_curlyx->u.curlyx.parenfloor);
4583 cur_curlyx->u.curlyx.lastloc = locinput;
4584 REGCP_SET(ST.lastcp);
4585 PUSH_STATE_GOTO(WHILEM_A_max, A);
4588 goto do_whilem_B_max;
4592 case WHILEM_B_min: /* just matched B in a minimal match */
4593 case WHILEM_B_max: /* just matched B in a maximal match */
4594 cur_curlyx = ST.save_curlyx;
4598 case WHILEM_B_max_fail: /* just failed to match B in a maximal match */
4599 cur_curlyx = ST.save_curlyx;
4600 cur_curlyx->u.curlyx.lastloc = ST.save_lastloc;
4601 cur_curlyx->u.curlyx.count--;
4605 case WHILEM_A_min_fail: /* just failed to match A in a minimal match */
4606 REGCP_UNWIND(ST.lastcp);
4609 case WHILEM_A_pre_fail: /* just failed to match even minimal A */
4610 cur_curlyx->u.curlyx.lastloc = ST.save_lastloc;
4611 cur_curlyx->u.curlyx.count--;
4615 case WHILEM_A_max_fail: /* just failed to match A in a maximal match */
4616 REGCP_UNWIND(ST.lastcp);
4617 regcppop(rex); /* Restore some previous $<digit>s? */
4618 PL_reginput = locinput;
4619 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
4620 "%*s whilem: failed, trying continuation...\n",
4621 REPORT_CODE_OFF+depth*2, "")
4624 if (cur_curlyx->u.curlyx.count >= REG_INFTY
4625 && ckWARN(WARN_REGEXP)
4626 && !(PL_reg_flags & RF_warned))
4628 PL_reg_flags |= RF_warned;
4629 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s limit (%d) exceeded",
4630 "Complex regular subexpression recursion",
4635 ST.save_curlyx = cur_curlyx;
4636 cur_curlyx = cur_curlyx->u.curlyx.prev_curlyx;
4637 PUSH_YES_STATE_GOTO(WHILEM_B_max, ST.save_curlyx->u.curlyx.B);
4640 case WHILEM_B_min_fail: /* just failed to match B in a minimal match */
4641 cur_curlyx = ST.save_curlyx;
4642 REGCP_UNWIND(ST.lastcp);
4645 if (cur_curlyx->u.curlyx.count >= /*max*/ARG2(cur_curlyx->u.curlyx.me)) {
4646 /* Maximum greed exceeded */
4647 if (cur_curlyx->u.curlyx.count >= REG_INFTY
4648 && ckWARN(WARN_REGEXP)
4649 && !(PL_reg_flags & RF_warned))
4651 PL_reg_flags |= RF_warned;
4652 Perl_warner(aTHX_ packWARN(WARN_REGEXP),
4653 "%s limit (%d) exceeded",
4654 "Complex regular subexpression recursion",
4657 cur_curlyx->u.curlyx.count--;
4661 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
4662 "%*s trying longer...\n", REPORT_CODE_OFF+depth*2, "")
4664 /* Try grabbing another A and see if it helps. */
4665 PL_reginput = locinput;
4666 cur_curlyx->u.curlyx.lastloc = locinput;
4667 ST.cp = regcppush(cur_curlyx->u.curlyx.parenfloor);
4668 REGCP_SET(ST.lastcp);
4669 PUSH_STATE_GOTO(WHILEM_A_min,
4670 /*A*/ NEXTOPER(ST.save_curlyx->u.curlyx.me) + EXTRA_STEP_2ARGS);
4674 #define ST st->u.branch
4676 case BRANCHJ: /* /(...|A|...)/ with long next pointer */
4677 next = scan + ARG(scan);
4680 scan = NEXTOPER(scan);
4683 case BRANCH: /* /(...|A|...)/ */
4684 scan = NEXTOPER(scan); /* scan now points to inner node */
4685 ST.lastparen = *PL_reglastparen;
4686 ST.next_branch = next;
4688 PL_reginput = locinput;
4690 /* Now go into the branch */
4692 PUSH_YES_STATE_GOTO(BRANCH_next, scan);
4694 PUSH_STATE_GOTO(BRANCH_next, scan);
4698 PL_reginput = locinput;
4699 sv_yes_mark = st->u.mark.mark_name = scan->flags ? NULL :
4700 MUTABLE_SV(rexi->data->data[ ARG( scan ) ]);
4701 PUSH_STATE_GOTO(CUTGROUP_next,next);
4703 case CUTGROUP_next_fail:
4706 if (st->u.mark.mark_name)
4707 sv_commit = st->u.mark.mark_name;
4713 case BRANCH_next_fail: /* that branch failed; try the next, if any */
4718 REGCP_UNWIND(ST.cp);
4719 for (n = *PL_reglastparen; n > ST.lastparen; n--)
4720 PL_regoffs[n].end = -1;
4721 *PL_reglastparen = n;
4722 /*dmq: *PL_reglastcloseparen = n; */
4723 scan = ST.next_branch;
4724 /* no more branches? */
4725 if (!scan || (OP(scan) != BRANCH && OP(scan) != BRANCHJ)) {
4727 PerlIO_printf( Perl_debug_log,
4728 "%*s %sBRANCH failed...%s\n",
4729 REPORT_CODE_OFF+depth*2, "",
4735 continue; /* execute next BRANCH[J] op */
4743 #define ST st->u.curlym
4745 case CURLYM: /* /A{m,n}B/ where A is fixed-length */
4747 /* This is an optimisation of CURLYX that enables us to push
4748 * only a single backtracking state, no matter how many matches
4749 * there are in {m,n}. It relies on the pattern being constant
4750 * length, with no parens to influence future backrefs
4754 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
4756 /* if paren positive, emulate an OPEN/CLOSE around A */
4758 U32 paren = ST.me->flags;
4759 if (paren > PL_regsize)
4761 if (paren > *PL_reglastparen)
4762 *PL_reglastparen = paren;
4763 scan += NEXT_OFF(scan); /* Skip former OPEN. */
4771 ST.c1 = CHRTEST_UNINIT;
4774 if (!(ST.minmod ? ARG1(ST.me) : ARG2(ST.me))) /* min/max */
4777 curlym_do_A: /* execute the A in /A{m,n}B/ */
4778 PL_reginput = locinput;
4779 PUSH_YES_STATE_GOTO(CURLYM_A, ST.A); /* match A */
4782 case CURLYM_A: /* we've just matched an A */
4783 locinput = st->locinput;
4784 nextchr = UCHARAT(locinput);
4787 /* after first match, determine A's length: u.curlym.alen */
4788 if (ST.count == 1) {
4789 if (PL_reg_match_utf8) {
4791 while (s < PL_reginput) {
4797 ST.alen = PL_reginput - locinput;
4800 ST.count = ST.minmod ? ARG1(ST.me) : ARG2(ST.me);
4803 PerlIO_printf(Perl_debug_log,
4804 "%*s CURLYM now matched %"IVdf" times, len=%"IVdf"...\n",
4805 (int)(REPORT_CODE_OFF+(depth*2)), "",
4806 (IV) ST.count, (IV)ST.alen)
4809 locinput = PL_reginput;
4811 if (cur_eval && cur_eval->u.eval.close_paren &&
4812 cur_eval->u.eval.close_paren == (U32)ST.me->flags)
4816 I32 max = (ST.minmod ? ARG1(ST.me) : ARG2(ST.me));
4817 if ( max == REG_INFTY || ST.count < max )
4818 goto curlym_do_A; /* try to match another A */
4820 goto curlym_do_B; /* try to match B */
4822 case CURLYM_A_fail: /* just failed to match an A */
4823 REGCP_UNWIND(ST.cp);
4825 if (ST.minmod || ST.count < ARG1(ST.me) /* min*/
4826 || (cur_eval && cur_eval->u.eval.close_paren &&
4827 cur_eval->u.eval.close_paren == (U32)ST.me->flags))
4830 curlym_do_B: /* execute the B in /A{m,n}B/ */
4831 PL_reginput = locinput;
4832 if (ST.c1 == CHRTEST_UNINIT) {
4833 /* calculate c1 and c2 for possible match of 1st char
4834 * following curly */
4835 ST.c1 = ST.c2 = CHRTEST_VOID;
4836 if (HAS_TEXT(ST.B) || JUMPABLE(ST.B)) {
4837 regnode *text_node = ST.B;
4838 if (! HAS_TEXT(text_node))
4839 FIND_NEXT_IMPT(text_node);
4842 (HAS_TEXT(text_node) && PL_regkind[OP(text_node)] == EXACT)
4844 But the former is redundant in light of the latter.
4846 if this changes back then the macro for
4847 IS_TEXT and friends need to change.
4849 if (PL_regkind[OP(text_node)] == EXACT)
4852 ST.c1 = (U8)*STRING(text_node);
4854 (IS_TEXTF(text_node))
4856 : (IS_TEXTFL(text_node))
4857 ? PL_fold_locale[ST.c1]
4864 PerlIO_printf(Perl_debug_log,
4865 "%*s CURLYM trying tail with matches=%"IVdf"...\n",
4866 (int)(REPORT_CODE_OFF+(depth*2)),
4869 if (ST.c1 != CHRTEST_VOID
4870 && UCHARAT(PL_reginput) != ST.c1
4871 && UCHARAT(PL_reginput) != ST.c2)
4873 /* simulate B failing */
4875 PerlIO_printf(Perl_debug_log,
4876 "%*s CURLYM Fast bail c1=%"IVdf" c2=%"IVdf"\n",
4877 (int)(REPORT_CODE_OFF+(depth*2)),"",
4880 state_num = CURLYM_B_fail;
4881 goto reenter_switch;
4885 /* mark current A as captured */
4886 I32 paren = ST.me->flags;
4888 PL_regoffs[paren].start
4889 = HOPc(PL_reginput, -ST.alen) - PL_bostr;
4890 PL_regoffs[paren].end = PL_reginput - PL_bostr;
4891 /*dmq: *PL_reglastcloseparen = paren; */
4894 PL_regoffs[paren].end = -1;
4895 if (cur_eval && cur_eval->u.eval.close_paren &&
4896 cur_eval->u.eval.close_paren == (U32)ST.me->flags)
4905 PUSH_STATE_GOTO(CURLYM_B, ST.B); /* match B */
4908 case CURLYM_B_fail: /* just failed to match a B */
4909 REGCP_UNWIND(ST.cp);
4911 I32 max = ARG2(ST.me);
4912 if (max != REG_INFTY && ST.count == max)
4914 goto curlym_do_A; /* try to match a further A */
4916 /* backtrack one A */
4917 if (ST.count == ARG1(ST.me) /* min */)
4920 locinput = HOPc(locinput, -ST.alen);
4921 goto curlym_do_B; /* try to match B */
4924 #define ST st->u.curly
4926 #define CURLY_SETPAREN(paren, success) \
4929 PL_regoffs[paren].start = HOPc(locinput, -1) - PL_bostr; \
4930 PL_regoffs[paren].end = locinput - PL_bostr; \
4931 *PL_reglastcloseparen = paren; \
4934 PL_regoffs[paren].end = -1; \
4937 case STAR: /* /A*B/ where A is width 1 */
4941 scan = NEXTOPER(scan);
4943 case PLUS: /* /A+B/ where A is width 1 */
4947 scan = NEXTOPER(scan);
4949 case CURLYN: /* /(A){m,n}B/ where A is width 1 */
4950 ST.paren = scan->flags; /* Which paren to set */
4951 if (ST.paren > PL_regsize)
4952 PL_regsize = ST.paren;
4953 if (ST.paren > *PL_reglastparen)
4954 *PL_reglastparen = ST.paren;
4955 ST.min = ARG1(scan); /* min to match */
4956 ST.max = ARG2(scan); /* max to match */
4957 if (cur_eval && cur_eval->u.eval.close_paren &&
4958 cur_eval->u.eval.close_paren == (U32)ST.paren) {
4962 scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE);
4964 case CURLY: /* /A{m,n}B/ where A is width 1 */
4966 ST.min = ARG1(scan); /* min to match */
4967 ST.max = ARG2(scan); /* max to match */
4968 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
4971 * Lookahead to avoid useless match attempts
4972 * when we know what character comes next.
4974 * Used to only do .*x and .*?x, but now it allows
4975 * for )'s, ('s and (?{ ... })'s to be in the way
4976 * of the quantifier and the EXACT-like node. -- japhy
4979 if (ST.min > ST.max) /* XXX make this a compile-time check? */
4981 if (HAS_TEXT(next) || JUMPABLE(next)) {
4983 regnode *text_node = next;
4985 if (! HAS_TEXT(text_node))
4986 FIND_NEXT_IMPT(text_node);
4988 if (! HAS_TEXT(text_node))
4989 ST.c1 = ST.c2 = CHRTEST_VOID;
4991 if ( PL_regkind[OP(text_node)] != EXACT ) {
4992 ST.c1 = ST.c2 = CHRTEST_VOID;
4993 goto assume_ok_easy;
4996 s = (U8*)STRING(text_node);
4998 /* Currently we only get here when
5000 PL_rekind[OP(text_node)] == EXACT
5002 if this changes back then the macro for IS_TEXT and
5003 friends need to change. */
5006 if (IS_TEXTF(text_node))
5007 ST.c2 = PL_fold[ST.c1];
5008 else if (IS_TEXTFL(text_node))
5009 ST.c2 = PL_fold_locale[ST.c1];
5011 else { /* UTF_PATTERN */
5012 if (IS_TEXTF(text_node)) {
5013 STRLEN ulen1, ulen2;
5014 U8 tmpbuf1[UTF8_MAXBYTES_CASE+1];
5015 U8 tmpbuf2[UTF8_MAXBYTES_CASE+1];
5017 to_utf8_lower((U8*)s, tmpbuf1, &ulen1);
5018 to_utf8_upper((U8*)s, tmpbuf2, &ulen2);
5020 ST.c1 = utf8n_to_uvchr(tmpbuf1, UTF8_MAXLEN, 0,
5022 0 : UTF8_ALLOW_ANY);
5023 ST.c2 = utf8n_to_uvchr(tmpbuf2, UTF8_MAXLEN, 0,
5025 0 : UTF8_ALLOW_ANY);
5027 ST.c1 = utf8n_to_uvuni(tmpbuf1, UTF8_MAXBYTES, 0,
5029 ST.c2 = utf8n_to_uvuni(tmpbuf2, UTF8_MAXBYTES, 0,
5034 ST.c2 = ST.c1 = utf8n_to_uvchr(s, UTF8_MAXBYTES, 0,
5041 ST.c1 = ST.c2 = CHRTEST_VOID;
5046 PL_reginput = locinput;
5049 if (ST.min && regrepeat(rex, ST.A, ST.min, depth) < ST.min)
5052 locinput = PL_reginput;
5054 if (ST.c1 == CHRTEST_VOID)
5055 goto curly_try_B_min;
5057 ST.oldloc = locinput;
5059 /* set ST.maxpos to the furthest point along the
5060 * string that could possibly match */
5061 if (ST.max == REG_INFTY) {
5062 ST.maxpos = PL_regeol - 1;
5064 while (UTF8_IS_CONTINUATION(*(U8*)ST.maxpos))
5067 else if (utf8_target) {
5068 int m = ST.max - ST.min;
5069 for (ST.maxpos = locinput;
5070 m >0 && ST.maxpos + UTF8SKIP(ST.maxpos) <= PL_regeol; m--)
5071 ST.maxpos += UTF8SKIP(ST.maxpos);
5074 ST.maxpos = locinput + ST.max - ST.min;
5075 if (ST.maxpos >= PL_regeol)
5076 ST.maxpos = PL_regeol - 1;
5078 goto curly_try_B_min_known;
5082 ST.count = regrepeat(rex, ST.A, ST.max, depth);
5083 locinput = PL_reginput;
5084 if (ST.count < ST.min)
5086 if ((ST.count > ST.min)
5087 && (PL_regkind[OP(ST.B)] == EOL) && (OP(ST.B) != MEOL))
5089 /* A{m,n} must come at the end of the string, there's
5090 * no point in backing off ... */
5092 /* ...except that $ and \Z can match before *and* after
5093 newline at the end. Consider "\n\n" =~ /\n+\Z\n/.
5094 We may back off by one in this case. */
5095 if (UCHARAT(PL_reginput - 1) == '\n' && OP(ST.B) != EOS)
5099 goto curly_try_B_max;
5104 case CURLY_B_min_known_fail:
5105 /* failed to find B in a non-greedy match where c1,c2 valid */
5106 if (ST.paren && ST.count)
5107 PL_regoffs[ST.paren].end = -1;
5109 PL_reginput = locinput; /* Could be reset... */
5110 REGCP_UNWIND(ST.cp);
5111 /* Couldn't or didn't -- move forward. */
5112 ST.oldloc = locinput;
5114 locinput += UTF8SKIP(locinput);
5118 curly_try_B_min_known:
5119 /* find the next place where 'B' could work, then call B */
5123 n = (ST.oldloc == locinput) ? 0 : 1;
5124 if (ST.c1 == ST.c2) {
5126 /* set n to utf8_distance(oldloc, locinput) */
5127 while (locinput <= ST.maxpos &&
5128 utf8n_to_uvchr((U8*)locinput,
5129 UTF8_MAXBYTES, &len,
5130 uniflags) != (UV)ST.c1) {
5136 /* set n to utf8_distance(oldloc, locinput) */
5137 while (locinput <= ST.maxpos) {
5139 const UV c = utf8n_to_uvchr((U8*)locinput,
5140 UTF8_MAXBYTES, &len,
5142 if (c == (UV)ST.c1 || c == (UV)ST.c2)
5150 if (ST.c1 == ST.c2) {
5151 while (locinput <= ST.maxpos &&
5152 UCHARAT(locinput) != ST.c1)
5156 while (locinput <= ST.maxpos
5157 && UCHARAT(locinput) != ST.c1
5158 && UCHARAT(locinput) != ST.c2)
5161 n = locinput - ST.oldloc;
5163 if (locinput > ST.maxpos)
5165 /* PL_reginput == oldloc now */
5168 if (regrepeat(rex, ST.A, n, depth) < n)
5171 PL_reginput = locinput;
5172 CURLY_SETPAREN(ST.paren, ST.count);
5173 if (cur_eval && cur_eval->u.eval.close_paren &&
5174 cur_eval->u.eval.close_paren == (U32)ST.paren) {
5177 PUSH_STATE_GOTO(CURLY_B_min_known, ST.B);
5182 case CURLY_B_min_fail:
5183 /* failed to find B in a non-greedy match where c1,c2 invalid */
5184 if (ST.paren && ST.count)
5185 PL_regoffs[ST.paren].end = -1;
5187 REGCP_UNWIND(ST.cp);
5188 /* failed -- move forward one */
5189 PL_reginput = locinput;
5190 if (regrepeat(rex, ST.A, 1, depth)) {
5192 locinput = PL_reginput;
5193 if (ST.count <= ST.max || (ST.max == REG_INFTY &&
5194 ST.count > 0)) /* count overflow ? */
5197 CURLY_SETPAREN(ST.paren, ST.count);
5198 if (cur_eval && cur_eval->u.eval.close_paren &&
5199 cur_eval->u.eval.close_paren == (U32)ST.paren) {
5202 PUSH_STATE_GOTO(CURLY_B_min, ST.B);
5210 /* a successful greedy match: now try to match B */
5211 if (cur_eval && cur_eval->u.eval.close_paren &&
5212 cur_eval->u.eval.close_paren == (U32)ST.paren) {
5217 if (ST.c1 != CHRTEST_VOID)
5218 c = utf8_target ? utf8n_to_uvchr((U8*)PL_reginput,
5219 UTF8_MAXBYTES, 0, uniflags)
5220 : (UV) UCHARAT(PL_reginput);
5221 /* If it could work, try it. */
5222 if (ST.c1 == CHRTEST_VOID || c == (UV)ST.c1 || c == (UV)ST.c2) {
5223 CURLY_SETPAREN(ST.paren, ST.count);
5224 PUSH_STATE_GOTO(CURLY_B_max, ST.B);
5229 case CURLY_B_max_fail:
5230 /* failed to find B in a greedy match */
5231 if (ST.paren && ST.count)
5232 PL_regoffs[ST.paren].end = -1;
5234 REGCP_UNWIND(ST.cp);
5236 if (--ST.count < ST.min)
5238 PL_reginput = locinput = HOPc(locinput, -1);
5239 goto curly_try_B_max;
5246 /* we've just finished A in /(??{A})B/; now continue with B */
5248 st->u.eval.toggle_reg_flags
5249 = cur_eval->u.eval.toggle_reg_flags;
5250 PL_reg_flags ^= st->u.eval.toggle_reg_flags;
5252 st->u.eval.prev_rex = rex_sv; /* inner */
5253 SETREX(rex_sv,cur_eval->u.eval.prev_rex);
5254 rex = (struct regexp *)SvANY(rex_sv);
5255 rexi = RXi_GET(rex);
5256 cur_curlyx = cur_eval->u.eval.prev_curlyx;
5257 ReREFCNT_inc(rex_sv);
5258 st->u.eval.cp = regcppush(0); /* Save *all* the positions. */
5260 /* rex was changed so update the pointer in PL_reglastparen and PL_reglastcloseparen */
5261 PL_reglastparen = &rex->lastparen;
5262 PL_reglastcloseparen = &rex->lastcloseparen;
5264 REGCP_SET(st->u.eval.lastcp);
5265 PL_reginput = locinput;
5267 /* Restore parens of the outer rex without popping the
5269 tmpix = PL_savestack_ix;
5270 PL_savestack_ix = cur_eval->u.eval.lastcp;
5272 PL_savestack_ix = tmpix;
5274 st->u.eval.prev_eval = cur_eval;
5275 cur_eval = cur_eval->u.eval.prev_eval;
5277 PerlIO_printf(Perl_debug_log, "%*s EVAL trying tail ... %"UVxf"\n",
5278 REPORT_CODE_OFF+depth*2, "",PTR2UV(cur_eval)););
5279 if ( nochange_depth )
5282 PUSH_YES_STATE_GOTO(EVAL_AB,
5283 st->u.eval.prev_eval->u.eval.B); /* match B */
5286 if (locinput < reginfo->till) {
5287 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
5288 "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n",
5290 (long)(locinput - PL_reg_starttry),
5291 (long)(reginfo->till - PL_reg_starttry),
5294 sayNO_SILENT; /* Cannot match: too short. */
5296 PL_reginput = locinput; /* put where regtry can find it */
5297 sayYES; /* Success! */
5299 case SUCCEED: /* successful SUSPEND/UNLESSM/IFMATCH/CURLYM */
5301 PerlIO_printf(Perl_debug_log,
5302 "%*s %ssubpattern success...%s\n",
5303 REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5]));
5304 PL_reginput = locinput; /* put where regtry can find it */
5305 sayYES; /* Success! */
5308 #define ST st->u.ifmatch
5310 case SUSPEND: /* (?>A) */
5312 PL_reginput = locinput;
5315 case UNLESSM: /* -ve lookaround: (?!A), or with flags, (?<!A) */
5317 goto ifmatch_trivial_fail_test;
5319 case IFMATCH: /* +ve lookaround: (?=A), or with flags, (?<=A) */
5321 ifmatch_trivial_fail_test:
5323 char * const s = HOPBACKc(locinput, scan->flags);
5328 sw = 1 - cBOOL(ST.wanted);
5332 next = scan + ARG(scan);
5340 PL_reginput = locinput;
5344 ST.logical = logical;
5345 logical = 0; /* XXX: reset state of logical once it has been saved into ST */
5347 /* execute body of (?...A) */
5348 PUSH_YES_STATE_GOTO(IFMATCH_A, NEXTOPER(NEXTOPER(scan)));
5351 case IFMATCH_A_fail: /* body of (?...A) failed */
5352 ST.wanted = !ST.wanted;
5355 case IFMATCH_A: /* body of (?...A) succeeded */
5357 sw = cBOOL(ST.wanted);
5359 else if (!ST.wanted)
5362 if (OP(ST.me) == SUSPEND)
5363 locinput = PL_reginput;
5365 locinput = PL_reginput = st->locinput;
5366 nextchr = UCHARAT(locinput);
5368 scan = ST.me + ARG(ST.me);
5371 continue; /* execute B */
5376 next = scan + ARG(scan);
5381 reginfo->cutpoint = PL_regeol;
5384 PL_reginput = locinput;
5386 sv_yes_mark = sv_commit = MUTABLE_SV(rexi->data->data[ ARG( scan ) ]);
5387 PUSH_STATE_GOTO(COMMIT_next,next);
5389 case COMMIT_next_fail:
5396 #define ST st->u.mark
5398 ST.prev_mark = mark_state;
5399 ST.mark_name = sv_commit = sv_yes_mark
5400 = MUTABLE_SV(rexi->data->data[ ARG( scan ) ]);
5402 ST.mark_loc = PL_reginput = locinput;
5403 PUSH_YES_STATE_GOTO(MARKPOINT_next,next);
5405 case MARKPOINT_next:
5406 mark_state = ST.prev_mark;
5409 case MARKPOINT_next_fail:
5410 if (popmark && sv_eq(ST.mark_name,popmark))
5412 if (ST.mark_loc > startpoint)
5413 reginfo->cutpoint = HOPBACKc(ST.mark_loc, 1);
5414 popmark = NULL; /* we found our mark */
5415 sv_commit = ST.mark_name;
5418 PerlIO_printf(Perl_debug_log,
5419 "%*s %ssetting cutpoint to mark:%"SVf"...%s\n",
5420 REPORT_CODE_OFF+depth*2, "",
5421 PL_colors[4], SVfARG(sv_commit), PL_colors[5]);
5424 mark_state = ST.prev_mark;
5425 sv_yes_mark = mark_state ?
5426 mark_state->u.mark.mark_name : NULL;
5430 PL_reginput = locinput;
5432 /* (*SKIP) : if we fail we cut here*/
5433 ST.mark_name = NULL;
5434 ST.mark_loc = locinput;
5435 PUSH_STATE_GOTO(SKIP_next,next);
5437 /* (*SKIP:NAME) : if there is a (*MARK:NAME) fail where it was,
5438 otherwise do nothing. Meaning we need to scan
5440 regmatch_state *cur = mark_state;
5441 SV *find = MUTABLE_SV(rexi->data->data[ ARG( scan ) ]);
5444 if ( sv_eq( cur->u.mark.mark_name,
5447 ST.mark_name = find;
5448 PUSH_STATE_GOTO( SKIP_next, next );
5450 cur = cur->u.mark.prev_mark;
5453 /* Didn't find our (*MARK:NAME) so ignore this (*SKIP:NAME) */
5455 case SKIP_next_fail:
5457 /* (*CUT:NAME) - Set up to search for the name as we
5458 collapse the stack*/
5459 popmark = ST.mark_name;
5461 /* (*CUT) - No name, we cut here.*/
5462 if (ST.mark_loc > startpoint)
5463 reginfo->cutpoint = HOPBACKc(ST.mark_loc, 1);
5464 /* but we set sv_commit to latest mark_name if there
5465 is one so they can test to see how things lead to this
5468 sv_commit=mark_state->u.mark.mark_name;
5476 if ( n == (U32)what_len_TRICKYFOLD(locinput,utf8_target,ln) ) {
5478 } else if ( 0xDF == n && !utf8_target && !UTF_PATTERN ) {
5481 U8 folded[UTF8_MAXBYTES_CASE+1];
5483 const char * const l = locinput;
5484 char *e = PL_regeol;
5485 to_uni_fold(n, folded, &foldlen);
5487 if (! foldEQ_utf8((const char*) folded, 0, foldlen, 1,
5488 l, &e, 0, utf8_target)) {
5493 nextchr = UCHARAT(locinput);
5496 if ((n=is_LNBREAK(locinput,utf8_target))) {
5498 nextchr = UCHARAT(locinput);
5503 #define CASE_CLASS(nAmE) \
5505 if ((n=is_##nAmE(locinput,utf8_target))) { \
5507 nextchr = UCHARAT(locinput); \
5512 if ((n=is_##nAmE(locinput,utf8_target))) { \
5515 locinput += UTF8SKIP(locinput); \
5516 nextchr = UCHARAT(locinput); \
5521 CASE_CLASS(HORIZWS);
5525 PerlIO_printf(Perl_error_log, "%"UVxf" %d\n",
5526 PTR2UV(scan), OP(scan));
5527 Perl_croak(aTHX_ "regexp memory corruption");
5531 /* switch break jumps here */
5532 scan = next; /* prepare to execute the next op and ... */
5533 continue; /* ... jump back to the top, reusing st */
5537 /* push a state that backtracks on success */
5538 st->u.yes.prev_yes_state = yes_state;
5542 /* push a new regex state, then continue at scan */
5544 regmatch_state *newst;
5547 regmatch_state *cur = st;
5548 regmatch_state *curyes = yes_state;
5550 regmatch_slab *slab = PL_regmatch_slab;
5551 for (;curd > -1;cur--,curd--) {
5552 if (cur < SLAB_FIRST(slab)) {
5554 cur = SLAB_LAST(slab);
5556 PerlIO_printf(Perl_error_log, "%*s#%-3d %-10s %s\n",
5557 REPORT_CODE_OFF + 2 + depth * 2,"",
5558 curd, PL_reg_name[cur->resume_state],
5559 (curyes == cur) ? "yes" : ""
5562 curyes = cur->u.yes.prev_yes_state;
5565 DEBUG_STATE_pp("push")
5568 st->locinput = locinput;
5570 if (newst > SLAB_LAST(PL_regmatch_slab))
5571 newst = S_push_slab(aTHX);
5572 PL_regmatch_state = newst;
5574 locinput = PL_reginput;
5575 nextchr = UCHARAT(locinput);
5583 * We get here only if there's trouble -- normally "case END" is
5584 * the terminating point.
5586 Perl_croak(aTHX_ "corrupted regexp pointers");
5592 /* we have successfully completed a subexpression, but we must now
5593 * pop to the state marked by yes_state and continue from there */
5594 assert(st != yes_state);
5596 while (st != yes_state) {
5598 if (st < SLAB_FIRST(PL_regmatch_slab)) {
5599 PL_regmatch_slab = PL_regmatch_slab->prev;
5600 st = SLAB_LAST(PL_regmatch_slab);
5604 DEBUG_STATE_pp("pop (no final)");
5606 DEBUG_STATE_pp("pop (yes)");
5612 while (yes_state < SLAB_FIRST(PL_regmatch_slab)
5613 || yes_state > SLAB_LAST(PL_regmatch_slab))
5615 /* not in this slab, pop slab */
5616 depth -= (st - SLAB_FIRST(PL_regmatch_slab) + 1);
5617 PL_regmatch_slab = PL_regmatch_slab->prev;
5618 st = SLAB_LAST(PL_regmatch_slab);
5620 depth -= (st - yes_state);
5623 yes_state = st->u.yes.prev_yes_state;
5624 PL_regmatch_state = st;
5627 locinput= st->locinput;
5628 nextchr = UCHARAT(locinput);
5630 state_num = st->resume_state + no_final;
5631 goto reenter_switch;
5634 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n",
5635 PL_colors[4], PL_colors[5]));
5637 if (PL_reg_eval_set) {
5638 /* each successfully executed (?{...}) block does the equivalent of
5639 * local $^R = do {...}
5640 * When popping the save stack, all these locals would be undone;
5641 * bypass this by setting the outermost saved $^R to the latest
5643 if (oreplsv != GvSV(PL_replgv))
5644 sv_setsv(oreplsv, GvSV(PL_replgv));
5651 PerlIO_printf(Perl_debug_log,
5652 "%*s %sfailed...%s\n",
5653 REPORT_CODE_OFF+depth*2, "",
5654 PL_colors[4], PL_colors[5])
5666 /* there's a previous state to backtrack to */
5668 if (st < SLAB_FIRST(PL_regmatch_slab)) {
5669 PL_regmatch_slab = PL_regmatch_slab->prev;
5670 st = SLAB_LAST(PL_regmatch_slab);
5672 PL_regmatch_state = st;
5673 locinput= st->locinput;
5674 nextchr = UCHARAT(locinput);
5676 DEBUG_STATE_pp("pop");
5678 if (yes_state == st)
5679 yes_state = st->u.yes.prev_yes_state;
5681 state_num = st->resume_state + 1; /* failure = success + 1 */
5682 goto reenter_switch;
5687 if (rex->intflags & PREGf_VERBARG_SEEN) {
5688 SV *sv_err = get_sv("REGERROR", 1);
5689 SV *sv_mrk = get_sv("REGMARK", 1);
5691 sv_commit = &PL_sv_no;
5693 sv_yes_mark = &PL_sv_yes;
5696 sv_commit = &PL_sv_yes;
5697 sv_yes_mark = &PL_sv_no;
5699 sv_setsv(sv_err, sv_commit);
5700 sv_setsv(sv_mrk, sv_yes_mark);
5703 /* clean up; in particular, free all slabs above current one */
5704 LEAVE_SCOPE(oldsave);
5710 - regrepeat - repeatedly match something simple, report how many
5713 * [This routine now assumes that it will only match on things of length 1.
5714 * That was true before, but now we assume scan - reginput is the count,
5715 * rather than incrementing count on every character. [Er, except utf8.]]
5718 S_regrepeat(pTHX_ const regexp *prog, const regnode *p, I32 max, int depth)
5721 register char *scan;
5723 register char *loceol = PL_regeol;
5724 register I32 hardcount = 0;
5725 register bool utf8_target = PL_reg_match_utf8;
5727 PERL_UNUSED_ARG(depth);
5730 PERL_ARGS_ASSERT_REGREPEAT;
5733 if (max == REG_INFTY)
5735 else if (max < loceol - scan)
5736 loceol = scan + max;
5741 while (scan < loceol && hardcount < max && *scan != '\n') {
5742 scan += UTF8SKIP(scan);
5746 while (scan < loceol && *scan != '\n')
5753 while (scan < loceol && hardcount < max) {
5754 scan += UTF8SKIP(scan);
5765 PL_reg_flags |= RF_tainted;
5769 /* To get here, EXACTish nodes must have *byte* length == 1. That means
5770 * they match only characters in the string that can be expressed as a
5771 * single byte. For non-utf8 strings, that means a simple match. For
5772 * utf8 strings, the character matched must be an invariant, or
5773 * downgradable to a single byte. The pattern's utf8ness is
5774 * irrelevant, as it must be a single byte, so either it isn't utf8, or
5775 * if it is it's an invariant */
5778 assert(! UTF_PATTERN || UNI_IS_INVARIANT(c));
5780 if ((! utf8_target) || UNI_IS_INVARIANT(c)) {
5782 /* Here, the string isn't utf8, or the character in the EXACT
5783 * node is the same in utf8 as not, so can just do equality.
5784 * Each matching char must be 1 byte long */
5787 while (scan < loceol && UCHARAT(scan) == c) {
5792 while (scan < loceol &&
5793 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold[c]))
5799 while (scan < loceol &&
5800 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold_locale[c]))
5806 Perl_croak(aTHX_ "panic: Unexpected op %u", OP(p));
5811 /* Here, the string is utf8, and the pattern char is different
5812 * in utf8 than not. */
5817 /* Fastest to find the two utf8 bytes that represent c, and
5818 * then look for those in sequence in the utf8 string */
5819 U8 high = UTF8_TWO_BYTE_HI(c);
5820 U8 low = UTF8_TWO_BYTE_LO(c);
5823 while (hardcount < max
5824 && scan + 1 < loceol
5825 && UCHARAT(scan) == high
5826 && UCHARAT(scan + 1) == low)
5833 case EXACTFL: /* Doesn't really make sense, but is best we can
5834 do. The documents warn against mixing locale
5837 { /* utf8 string, so use utf8 foldEQ */
5838 char *tmpeol = loceol;
5839 while (hardcount < max
5840 && foldEQ_utf8(scan, &tmpeol, 0, utf8_target,
5841 STRING(p), NULL, 1, UTF_PATTERN))
5848 /* XXX Note that the above handles properly the German
5849 * sharp ss in the pattern matching ss in the string. But
5850 * it doesn't handle properly cases where the string
5851 * contains say 'LIGATURE ff' and the pattern is 'f+'.
5852 * This would require, say, a new function or revised
5853 * interface to foldEQ_utf8(), in which the maximum number
5854 * of characters to match could be passed and it would
5855 * return how many actually did. This is just one of many
5856 * cases where multi-char folds don't work properly, and so
5857 * the fix is being deferred */
5861 Perl_croak(aTHX_ "panic: Unexpected op %u", OP(p));
5868 while (hardcount < max && scan < loceol &&
5869 reginclass(prog, p, (U8*)scan, 0, utf8_target)) {
5870 scan += UTF8SKIP(scan);
5874 while (scan < loceol && REGINCLASS(prog, p, (U8*)scan))
5881 LOAD_UTF8_CHARCLASS_ALNUM();
5882 while (hardcount < max && scan < loceol &&
5883 swash_fetch(PL_utf8_alnum, (U8*)scan, utf8_target))
5885 scan += UTF8SKIP(scan);
5888 } else if (FLAGS(p) & USE_UNI) {
5889 while (scan < loceol && isWORDCHAR_L1((U8) *scan)) {
5893 while (scan < loceol && isALNUM((U8) *scan)) {
5899 PL_reg_flags |= RF_tainted;
5902 while (hardcount < max && scan < loceol &&
5903 isALNUM_LC_utf8((U8*)scan)) {
5904 scan += UTF8SKIP(scan);
5908 while (scan < loceol && isALNUM_LC(*scan))
5915 LOAD_UTF8_CHARCLASS_ALNUM();
5916 while (hardcount < max && scan < loceol &&
5917 !swash_fetch(PL_utf8_alnum, (U8*)scan, utf8_target))
5919 scan += UTF8SKIP(scan);
5922 } else if (FLAGS(p) & USE_UNI) {
5923 while (scan < loceol && ! isWORDCHAR_L1((U8) *scan)) {
5927 while (scan < loceol && ! isALNUM((U8) *scan)) {
5933 PL_reg_flags |= RF_tainted;
5936 while (hardcount < max && scan < loceol &&
5937 !isALNUM_LC_utf8((U8*)scan)) {
5938 scan += UTF8SKIP(scan);
5942 while (scan < loceol && !isALNUM_LC(*scan))
5949 LOAD_UTF8_CHARCLASS_SPACE();
5950 while (hardcount < max && scan < loceol &&
5952 swash_fetch(PL_utf8_space,(U8*)scan, utf8_target)))
5954 scan += UTF8SKIP(scan);
5957 } else if (FLAGS(p) & USE_UNI) {
5958 while (scan < loceol && isSPACE_L1((U8) *scan)) {
5962 while (scan < loceol && isSPACE((U8) *scan))
5967 PL_reg_flags |= RF_tainted;
5970 while (hardcount < max && scan < loceol &&
5971 (*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
5972 scan += UTF8SKIP(scan);
5976 while (scan < loceol && isSPACE_LC(*scan))
5983 LOAD_UTF8_CHARCLASS_SPACE();
5984 while (hardcount < max && scan < loceol &&
5986 swash_fetch(PL_utf8_space,(U8*)scan, utf8_target)))
5988 scan += UTF8SKIP(scan);
5991 } else if (FLAGS(p) & USE_UNI) {
5992 while (scan < loceol && ! isSPACE_L1((U8) *scan)) {
5996 while (scan < loceol && ! isSPACE((U8) *scan)) {
6002 PL_reg_flags |= RF_tainted;
6005 while (hardcount < max && scan < loceol &&
6006 !(*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
6007 scan += UTF8SKIP(scan);
6011 while (scan < loceol && !isSPACE_LC(*scan))
6018 LOAD_UTF8_CHARCLASS_DIGIT();
6019 while (hardcount < max && scan < loceol &&
6020 swash_fetch(PL_utf8_digit, (U8*)scan, utf8_target)) {
6021 scan += UTF8SKIP(scan);
6025 while (scan < loceol && isDIGIT(*scan))
6032 LOAD_UTF8_CHARCLASS_DIGIT();
6033 while (hardcount < max && scan < loceol &&
6034 !swash_fetch(PL_utf8_digit, (U8*)scan, utf8_target)) {
6035 scan += UTF8SKIP(scan);
6039 while (scan < loceol && !isDIGIT(*scan))
6045 while (hardcount < max && scan < loceol && (c=is_LNBREAK_utf8(scan))) {
6051 LNBREAK can match two latin chars, which is ok,
6052 because we have a null terminated string, but we
6053 have to use hardcount in this situation
6055 while (scan < loceol && (c=is_LNBREAK_latin1(scan))) {
6064 while (hardcount < max && scan < loceol && (c=is_HORIZWS_utf8(scan))) {
6069 while (scan < loceol && is_HORIZWS_latin1(scan))
6076 while (hardcount < max && scan < loceol && !is_HORIZWS_utf8(scan)) {
6077 scan += UTF8SKIP(scan);
6081 while (scan < loceol && !is_HORIZWS_latin1(scan))
6089 while (hardcount < max && scan < loceol && (c=is_VERTWS_utf8(scan))) {
6094 while (scan < loceol && is_VERTWS_latin1(scan))
6102 while (hardcount < max && scan < loceol && !is_VERTWS_utf8(scan)) {
6103 scan += UTF8SKIP(scan);
6107 while (scan < loceol && !is_VERTWS_latin1(scan))
6113 default: /* Called on something of 0 width. */
6114 break; /* So match right here or not at all. */
6120 c = scan - PL_reginput;
6124 GET_RE_DEBUG_FLAGS_DECL;
6126 SV * const prop = sv_newmortal();
6127 regprop(prog, prop, p);
6128 PerlIO_printf(Perl_debug_log,
6129 "%*s %s can match %"IVdf" times out of %"IVdf"...\n",
6130 REPORT_CODE_OFF + depth*2, "", SvPVX_const(prop),(IV)c,(IV)max);
6138 #if !defined(PERL_IN_XSUB_RE) || defined(PLUGGABLE_RE_EXTENSION)
6140 - regclass_swash - prepare the utf8 swash
6144 Perl_regclass_swash(pTHX_ const regexp *prog, register const regnode* node, bool doinit, SV** listsvp, SV **altsvp)
6150 RXi_GET_DECL(prog,progi);
6151 const struct reg_data * const data = prog ? progi->data : NULL;
6153 PERL_ARGS_ASSERT_REGCLASS_SWASH;
6155 if (data && data->count) {
6156 const U32 n = ARG(node);
6158 if (data->what[n] == 's') {
6159 SV * const rv = MUTABLE_SV(data->data[n]);
6160 AV * const av = MUTABLE_AV(SvRV(rv));
6161 SV **const ary = AvARRAY(av);
6164 /* See the end of regcomp.c:S_regclass() for
6165 * documentation of these array elements. */
6168 a = SvROK(ary[1]) ? &ary[1] : NULL;
6169 b = SvTYPE(ary[2]) == SVt_PVAV ? &ary[2] : NULL;
6173 else if (si && doinit) {
6174 sw = swash_init("utf8", "", si, 1, 0);
6175 (void)av_store(av, 1, sw);
6192 - reginclass - determine if a character falls into a character class
6194 n is the ANYOF regnode
6195 p is the target string
6196 lenp is pointer to the maximum number of bytes of how far to go in p
6197 (This is assumed wthout checking to always be at least the current
6199 utf8_target tells whether p is in UTF-8.
6201 Returns true if matched; false otherwise. If lenp is not NULL, on return
6202 from a successful match, the value it points to will be updated to how many
6203 bytes in p were matched. If there was no match, the value is undefined,
6204 possibly changed from the input.
6209 S_reginclass(pTHX_ const regexp * const prog, register const regnode * const n, register const U8* const p, STRLEN* lenp, register const bool utf8_target)
6212 const char flags = ANYOF_FLAGS(n);
6218 PERL_ARGS_ASSERT_REGINCLASS;
6220 /* If c is not already the code point, get it */
6221 if (utf8_target && !UTF8_IS_INVARIANT(c)) {
6222 c = utf8n_to_uvchr(p, UTF8_MAXBYTES, &c_len,
6223 (UTF8_ALLOW_DEFAULT & UTF8_ALLOW_ANYUV)
6224 | UTF8_ALLOW_FFFF | UTF8_CHECK_ONLY);
6225 /* see [perl #37836] for UTF8_ALLOW_ANYUV; [perl #38293] for
6226 * UTF8_ALLOW_FFFF */
6227 if (c_len == (STRLEN)-1)
6228 Perl_croak(aTHX_ "Malformed UTF-8 character (fatal)");
6234 /* Use passed in max length, or one character if none passed in or less
6235 * than one character. And assume will match just one character. This is
6236 * overwritten later if matched more. */
6238 maxlen = (*lenp > c_len) ? *lenp : c_len;
6246 /* If this character is potentially in the bitmap, check it */
6248 if (ANYOF_BITMAP_TEST(n, c))
6250 else if (flags & ANYOF_FOLD) {
6253 if (flags & ANYOF_LOCALE) {
6254 PL_reg_flags |= RF_tainted;
6255 f = PL_fold_locale[c];
6259 if (f != c && ANYOF_BITMAP_TEST(n, f))
6263 if (!match && (flags & ANYOF_CLASS)) {
6264 PL_reg_flags |= RF_tainted;
6266 (ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) ||
6267 (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) ||
6268 (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) ||
6269 (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) ||
6270 (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) ||
6271 (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) ||
6272 (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) ||
6273 (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) ||
6274 (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) ||
6275 (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) ||
6276 (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII(c)) ||
6277 (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII(c)) ||
6278 (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) ||
6279 (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) ||
6280 (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) ||
6281 (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) ||
6282 (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) ||
6283 (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) ||
6284 (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) ||
6285 (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) ||
6286 (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) ||
6287 (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) ||
6288 (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) ||
6289 (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) ||
6290 (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) ||
6291 (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) ||
6292 (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) ||
6293 (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) ||
6294 (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK(c)) ||
6295 (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK(c))
6296 ) /* How's that for a conditional? */
6303 /* If the bitmap didn't (or couldn't) match, and something outside the
6304 * bitmap could match, try that */
6305 if (!match && (utf8_target || (flags & ANYOF_UNICODE))) {
6306 if (utf8_target && (flags & ANYOF_UNICODE_ALL) && c >= 256) {
6311 SV * const sw = regclass_swash(prog, n, TRUE, 0, (SV**)&av);
6319 utf8_p = bytes_to_utf8(p, &len);
6321 if (swash_fetch(sw, utf8_p, 1))
6323 else if (flags & ANYOF_FOLD) {
6324 if (!match && lenp && av) {
6326 for (i = 0; i <= av_len(av); i++) {
6327 SV* const sv = *av_fetch(av, i, FALSE);
6329 const char * const s = SvPV_const(sv, len);
6330 if (len <= maxlen && memEQ(s, (char*)utf8_p, len)) {
6338 U8 folded[UTF8_MAXBYTES_CASE+1];
6340 /* See if the folded version matches */
6342 to_utf8_fold(utf8_p, folded, &foldlen);
6343 if (swash_fetch(sw, folded, 1)) { /* 1 => is utf8 */
6349 /* Consider "k" =~ /[K]/i. The line above would
6350 * have just folded the 'k' to itself, and that
6351 * isn't going to match 'K'. So we look through
6352 * the closure of everything that folds to 'k'.
6353 * That will find the 'K'. Initialize the list, if
6355 if (! PL_utf8_foldclosures) {
6357 /* If the folds haven't been read in, call a fold
6358 * function to force that */
6359 if (! PL_utf8_tofold) {
6360 U8 dummy[UTF8_MAXBYTES+1];
6362 to_utf8_fold((U8*) "A", dummy, &dummy_len);
6364 PL_utf8_foldclosures =
6365 _swash_inversion_hash(PL_utf8_tofold);
6368 /* The data structure is a hash with the keys every
6369 * character that is folded to, like 'k', and the
6370 * values each an array of everything that folds to
6371 * its key. e.g. [ 'k', 'K', KELVIN_SIGN ] */
6372 if ((listp = hv_fetch(PL_utf8_foldclosures,
6373 (char *) folded, foldlen, FALSE)))
6375 AV* list = (AV*) *listp;
6377 for (i = 0; i <= av_len(list); i++) {
6378 SV** try_p = av_fetch(list, i, FALSE);
6379 if (try_p == NULL) {
6380 Perl_croak(aTHX_ "panic: invalid PL_utf8_foldclosures structure");
6382 /* Don't have to worry about embeded nulls
6383 * since NULL isn't folded or foldable */
6384 if (swash_fetch(sw, (U8*) SvPVX(*try_p),1)) {
6394 /* If we allocated a string above, free it */
6395 if (! utf8_target) Safefree(utf8_p);
6400 return (flags & ANYOF_INVERT) ? !match : match;
6404 S_reghop3(U8 *s, I32 off, const U8* lim)
6408 PERL_ARGS_ASSERT_REGHOP3;
6411 while (off-- && s < lim) {
6412 /* XXX could check well-formedness here */
6417 while (off++ && s > lim) {
6419 if (UTF8_IS_CONTINUED(*s)) {
6420 while (s > lim && UTF8_IS_CONTINUATION(*s))
6423 /* XXX could check well-formedness here */
6430 /* there are a bunch of places where we use two reghop3's that should
6431 be replaced with this routine. but since thats not done yet
6432 we ifdef it out - dmq
6435 S_reghop4(U8 *s, I32 off, const U8* llim, const U8* rlim)
6439 PERL_ARGS_ASSERT_REGHOP4;
6442 while (off-- && s < rlim) {
6443 /* XXX could check well-formedness here */
6448 while (off++ && s > llim) {
6450 if (UTF8_IS_CONTINUED(*s)) {
6451 while (s > llim && UTF8_IS_CONTINUATION(*s))
6454 /* XXX could check well-formedness here */
6462 S_reghopmaybe3(U8* s, I32 off, const U8* lim)
6466 PERL_ARGS_ASSERT_REGHOPMAYBE3;
6469 while (off-- && s < lim) {
6470 /* XXX could check well-formedness here */
6477 while (off++ && s > lim) {
6479 if (UTF8_IS_CONTINUED(*s)) {
6480 while (s > lim && UTF8_IS_CONTINUATION(*s))
6483 /* XXX could check well-formedness here */
6492 restore_pos(pTHX_ void *arg)
6495 regexp * const rex = (regexp *)arg;
6496 if (PL_reg_eval_set) {
6497 if (PL_reg_oldsaved) {
6498 rex->subbeg = PL_reg_oldsaved;
6499 rex->sublen = PL_reg_oldsavedlen;
6500 #ifdef PERL_OLD_COPY_ON_WRITE
6501 rex->saved_copy = PL_nrs;
6503 RXp_MATCH_COPIED_on(rex);
6505 PL_reg_magic->mg_len = PL_reg_oldpos;
6506 PL_reg_eval_set = 0;
6507 PL_curpm = PL_reg_oldcurpm;
6512 S_to_utf8_substr(pTHX_ register regexp *prog)
6516 PERL_ARGS_ASSERT_TO_UTF8_SUBSTR;
6519 if (prog->substrs->data[i].substr
6520 && !prog->substrs->data[i].utf8_substr) {
6521 SV* const sv = newSVsv(prog->substrs->data[i].substr);
6522 prog->substrs->data[i].utf8_substr = sv;
6523 sv_utf8_upgrade(sv);
6524 if (SvVALID(prog->substrs->data[i].substr)) {
6525 const U8 flags = BmFLAGS(prog->substrs->data[i].substr);
6526 if (flags & FBMcf_TAIL) {
6527 /* Trim the trailing \n that fbm_compile added last
6529 SvCUR_set(sv, SvCUR(sv) - 1);
6530 /* Whilst this makes the SV technically "invalid" (as its
6531 buffer is no longer followed by "\0") when fbm_compile()
6532 adds the "\n" back, a "\0" is restored. */
6534 fbm_compile(sv, flags);
6536 if (prog->substrs->data[i].substr == prog->check_substr)
6537 prog->check_utf8 = sv;
6543 S_to_byte_substr(pTHX_ register regexp *prog)
6548 PERL_ARGS_ASSERT_TO_BYTE_SUBSTR;
6551 if (prog->substrs->data[i].utf8_substr
6552 && !prog->substrs->data[i].substr) {
6553 SV* sv = newSVsv(prog->substrs->data[i].utf8_substr);
6554 if (sv_utf8_downgrade(sv, TRUE)) {
6555 if (SvVALID(prog->substrs->data[i].utf8_substr)) {
6557 = BmFLAGS(prog->substrs->data[i].utf8_substr);
6558 if (flags & FBMcf_TAIL) {
6559 /* Trim the trailing \n that fbm_compile added last
6561 SvCUR_set(sv, SvCUR(sv) - 1);
6563 fbm_compile(sv, flags);
6569 prog->substrs->data[i].substr = sv;
6570 if (prog->substrs->data[i].utf8_substr == prog->check_utf8)
6571 prog->check_substr = sv;
6578 * c-indentation-style: bsd
6580 * indent-tabs-mode: t
6583 * ex: set ts=8 sts=4 sw=4 noet: