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? e.g. locale */
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)
94 /* Valid for non-utf8 strings, non-ANYOFV nodes only: avoids the reginclass
95 * call if there are no complications: i.e., if everything matchable is
96 * straight forward in the bitmap */
97 #define REGINCLASS(prog,p,c) (ANYOF_FLAGS(p) ? reginclass(prog,p,c,0,0) \
98 : ANYOF_BITMAP_TEST(p,*(c)))
104 #define CHR_SVLEN(sv) (utf8_target ? sv_len_utf8(sv) : SvCUR(sv))
105 #define CHR_DIST(a,b) (PL_reg_match_utf8 ? utf8_distance(a,b) : a - b)
107 #define HOPc(pos,off) \
108 (char *)(PL_reg_match_utf8 \
109 ? reghop3((U8*)pos, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr)) \
111 #define HOPBACKc(pos, off) \
112 (char*)(PL_reg_match_utf8\
113 ? reghopmaybe3((U8*)pos, -off, (U8*)PL_bostr) \
114 : (pos - off >= PL_bostr) \
118 #define HOP3(pos,off,lim) (PL_reg_match_utf8 ? reghop3((U8*)(pos), off, (U8*)(lim)) : (U8*)(pos + off))
119 #define HOP3c(pos,off,lim) ((char*)HOP3(pos,off,lim))
121 /* these are unrolled below in the CCC_TRY_XXX defined */
123 /* Often 'str' is a hard-coded utf8 string instead of utfebcdic. so just
124 * skip the check on EBCDIC platforms */
125 # define LOAD_UTF8_CHARCLASS(class,str) LOAD_UTF8_CHARCLASS_NO_CHECK(class)
127 # define LOAD_UTF8_CHARCLASS(class,str) STMT_START { \
128 if (!CAT2(PL_utf8_,class)) { \
130 ENTER; save_re_context(); \
131 ok=CAT2(is_utf8_,class)((const U8*)str); \
132 assert(ok); assert(CAT2(PL_utf8_,class)); LEAVE; } } STMT_END
135 /* Doesn't do an assert to verify that is correct */
136 #define LOAD_UTF8_CHARCLASS_NO_CHECK(class) STMT_START { \
137 if (!CAT2(PL_utf8_,class)) { \
138 bool throw_away PERL_UNUSED_DECL; \
139 ENTER; save_re_context(); \
140 throw_away = CAT2(is_utf8_,class)((const U8*)" "); \
143 #define LOAD_UTF8_CHARCLASS_ALNUM() LOAD_UTF8_CHARCLASS(alnum,"a")
144 #define LOAD_UTF8_CHARCLASS_DIGIT() LOAD_UTF8_CHARCLASS(digit,"0")
145 #define LOAD_UTF8_CHARCLASS_SPACE() LOAD_UTF8_CHARCLASS(space," ")
147 #define LOAD_UTF8_CHARCLASS_GCB() /* Grapheme cluster boundaries */ \
148 LOAD_UTF8_CHARCLASS(X_begin, " "); \
149 LOAD_UTF8_CHARCLASS(X_non_hangul, "A"); \
150 /* These are utf8 constants, and not utf-ebcdic constants, so the \
151 * assert should likely and hopefully fail on an EBCDIC machine */ \
152 LOAD_UTF8_CHARCLASS(X_extend, "\xcc\x80"); /* U+0300 */ \
154 /* No asserts are done for these, in case called on an early \
155 * Unicode version in which they map to nothing */ \
156 LOAD_UTF8_CHARCLASS_NO_CHECK(X_prepend);/* U+0E40 "\xe0\xb9\x80" */ \
157 LOAD_UTF8_CHARCLASS_NO_CHECK(X_L); /* U+1100 "\xe1\x84\x80" */ \
158 LOAD_UTF8_CHARCLASS_NO_CHECK(X_LV); /* U+AC00 "\xea\xb0\x80" */ \
159 LOAD_UTF8_CHARCLASS_NO_CHECK(X_LVT); /* U+AC01 "\xea\xb0\x81" */ \
160 LOAD_UTF8_CHARCLASS_NO_CHECK(X_LV_LVT_V);/* U+AC01 "\xea\xb0\x81" */\
161 LOAD_UTF8_CHARCLASS_NO_CHECK(X_T); /* U+11A8 "\xe1\x86\xa8" */ \
162 LOAD_UTF8_CHARCLASS_NO_CHECK(X_V) /* U+1160 "\xe1\x85\xa0" */
164 #define PLACEHOLDER /* Something for the preprocessor to grab onto */
166 /* The actual code for CCC_TRY, which uses several variables from the routine
167 * it's callable from. It is designed to be the bulk of a case statement.
168 * FUNC is the macro or function to call on non-utf8 targets that indicate if
169 * nextchr matches the class.
170 * UTF8_TEST is the whole test string to use for utf8 targets
171 * LOAD is what to use to test, and if not present to load in the swash for the
173 * POS_OR_NEG is either empty or ! to complement the results of FUNC or
175 * The logic is: Fail if we're at the end-of-string; otherwise if the target is
176 * utf8 and a variant, load the swash if necessary and test using the utf8
177 * test. Advance to the next character if test is ok, otherwise fail; If not
178 * utf8 or an invariant under utf8, use the non-utf8 test, and fail if it
179 * fails, or advance to the next character */
181 #define _CCC_TRY_CODE(POS_OR_NEG, FUNC, UTF8_TEST, CLASS, STR) \
182 if (locinput >= PL_regeol) { \
185 if (utf8_target && UTF8_IS_CONTINUED(nextchr)) { \
186 LOAD_UTF8_CHARCLASS(CLASS, STR); \
187 if (POS_OR_NEG (UTF8_TEST)) { \
190 locinput += PL_utf8skip[nextchr]; \
191 nextchr = UCHARAT(locinput); \
194 if (POS_OR_NEG (FUNC(nextchr))) { \
197 nextchr = UCHARAT(++locinput); \
200 /* Handle the non-locale cases for a character class and its complement. It
201 * calls _CCC_TRY_CODE with a ! to complement the test for the character class.
202 * This is because that code fails when the test succeeds, so we want to have
203 * the test fail so that the code succeeds. The swash is stored in a
204 * predictable PL_ place */
205 #define _CCC_TRY_NONLOCALE(NAME, NNAME, FUNC, \
208 _CCC_TRY_CODE( !, FUNC, \
209 cBOOL(swash_fetch(CAT2(PL_utf8_,CLASS), \
210 (U8*)locinput, TRUE)), \
213 _CCC_TRY_CODE( PLACEHOLDER , FUNC, \
214 cBOOL(swash_fetch(CAT2(PL_utf8_,CLASS), \
215 (U8*)locinput, TRUE)), \
218 /* Generate the case statements for both locale and non-locale character
219 * classes in regmatch for classes that don't have special unicode semantics.
220 * Locales don't use an immediate swash, but an intermediary special locale
221 * function that is called on the pointer to the current place in the input
222 * string. That function will resolve to needing the same swash. One might
223 * think that because we don't know what the locale will match, we shouldn't
224 * check with the swash loading function that it loaded properly; ie, that we
225 * should use LOAD_UTF8_CHARCLASS_NO_CHECK for those, but what is passed to the
226 * regular LOAD_UTF8_CHARCLASS is in non-locale terms, and so locale is
228 #define CCC_TRY(NAME, NNAME, FUNC, \
229 NAMEL, NNAMEL, LCFUNC, LCFUNC_utf8, \
230 NAMEA, NNAMEA, FUNCA, \
233 PL_reg_flags |= RF_tainted; \
234 _CCC_TRY_CODE( !, LCFUNC, LCFUNC_utf8((U8*)locinput), CLASS, STR) \
236 PL_reg_flags |= RF_tainted; \
237 _CCC_TRY_CODE( PLACEHOLDER, LCFUNC, LCFUNC_utf8((U8*)locinput), \
240 if (locinput >= PL_regeol || ! FUNCA(nextchr)) { \
243 /* Matched a utf8-invariant, so don't have to worry about utf8 */ \
244 nextchr = UCHARAT(++locinput); \
247 if (locinput >= PL_regeol || FUNCA(nextchr)) { \
251 locinput += PL_utf8skip[nextchr]; \
252 nextchr = UCHARAT(locinput); \
255 nextchr = UCHARAT(++locinput); \
258 /* Generate the non-locale cases */ \
259 _CCC_TRY_NONLOCALE(NAME, NNAME, FUNC, CLASS, STR)
261 /* This is like CCC_TRY, but has an extra set of parameters for generating case
262 * statements to handle separate Unicode semantics nodes */
263 #define CCC_TRY_U(NAME, NNAME, FUNC, \
264 NAMEL, NNAMEL, LCFUNC, LCFUNC_utf8, \
265 NAMEU, NNAMEU, FUNCU, \
266 NAMEA, NNAMEA, FUNCA, \
268 CCC_TRY(NAME, NNAME, FUNC, \
269 NAMEL, NNAMEL, LCFUNC, LCFUNC_utf8, \
270 NAMEA, NNAMEA, FUNCA, \
272 _CCC_TRY_NONLOCALE(NAMEU, NNAMEU, FUNCU, CLASS, STR)
274 /* TODO: Combine JUMPABLE and HAS_TEXT to cache OP(rn) */
276 /* for use after a quantifier and before an EXACT-like node -- japhy */
277 /* it would be nice to rework regcomp.sym to generate this stuff. sigh
279 * NOTE that *nothing* that affects backtracking should be in here, specifically
280 * VERBS must NOT be included. JUMPABLE is used to determine if we can ignore a
281 * node that is in between two EXACT like nodes when ascertaining what the required
282 * "follow" character is. This should probably be moved to regex compile time
283 * although it may be done at run time beause of the REF possibility - more
284 * investigation required. -- demerphq
286 #define JUMPABLE(rn) ( \
288 (OP(rn) == CLOSE && (!cur_eval || cur_eval->u.eval.close_paren != ARG(rn))) || \
290 OP(rn) == SUSPEND || OP(rn) == IFMATCH || \
291 OP(rn) == PLUS || OP(rn) == MINMOD || \
293 (PL_regkind[OP(rn)] == CURLY && ARG1(rn) > 0) \
295 #define IS_EXACT(rn) (PL_regkind[OP(rn)] == EXACT)
297 #define HAS_TEXT(rn) ( IS_EXACT(rn) || PL_regkind[OP(rn)] == REF )
300 /* Currently these are only used when PL_regkind[OP(rn)] == EXACT so
301 we don't need this definition. */
302 #define IS_TEXT(rn) ( OP(rn)==EXACT || OP(rn)==REF || OP(rn)==NREF )
303 #define IS_TEXTF(rn) ( OP(rn)==EXACTFU || OP(rn)==EXACTFU_SS || OP(rn)==EXACTFU_TRICKYFOLD || OP(rn)==EXACTFA || OP(rn)==EXACTF || OP(rn)==REFF || OP(rn)==NREFF )
304 #define IS_TEXTFL(rn) ( OP(rn)==EXACTFL || OP(rn)==REFFL || OP(rn)==NREFFL )
307 /* ... so we use this as its faster. */
308 #define IS_TEXT(rn) ( OP(rn)==EXACT )
309 #define IS_TEXTFU(rn) ( OP(rn)==EXACTFU || OP(rn)==EXACTFU_SS || OP(rn)==EXACTFU_TRICKYFOLD || OP(rn) == EXACTFA)
310 #define IS_TEXTF(rn) ( OP(rn)==EXACTF )
311 #define IS_TEXTFL(rn) ( OP(rn)==EXACTFL )
316 Search for mandatory following text node; for lookahead, the text must
317 follow but for lookbehind (rn->flags != 0) we skip to the next step.
319 #define FIND_NEXT_IMPT(rn) STMT_START { \
320 while (JUMPABLE(rn)) { \
321 const OPCODE type = OP(rn); \
322 if (type == SUSPEND || PL_regkind[type] == CURLY) \
323 rn = NEXTOPER(NEXTOPER(rn)); \
324 else if (type == PLUS) \
326 else if (type == IFMATCH) \
327 rn = (rn->flags == 0) ? NEXTOPER(NEXTOPER(rn)) : rn + ARG(rn); \
328 else rn += NEXT_OFF(rn); \
333 static void restore_pos(pTHX_ void *arg);
335 #define REGCP_PAREN_ELEMS 4
336 #define REGCP_OTHER_ELEMS 3
337 #define REGCP_FRAME_ELEMS 1
338 /* REGCP_FRAME_ELEMS are not part of the REGCP_OTHER_ELEMS and
339 * are needed for the regexp context stack bookkeeping. */
342 S_regcppush(pTHX_ const regexp *rex, I32 parenfloor)
345 const int retval = PL_savestack_ix;
346 const int paren_elems_to_push = (PL_regsize - parenfloor) * REGCP_PAREN_ELEMS;
347 const UV total_elems = paren_elems_to_push + REGCP_OTHER_ELEMS;
348 const UV elems_shifted = total_elems << SAVE_TIGHT_SHIFT;
350 GET_RE_DEBUG_FLAGS_DECL;
352 PERL_ARGS_ASSERT_REGCPPUSH;
354 if (paren_elems_to_push < 0)
355 Perl_croak(aTHX_ "panic: paren_elems_to_push, %i < 0",
356 paren_elems_to_push);
358 if ((elems_shifted >> SAVE_TIGHT_SHIFT) != total_elems)
359 Perl_croak(aTHX_ "panic: paren_elems_to_push offset %"UVuf
360 " out of range (%lu-%ld)",
361 total_elems, (unsigned long)PL_regsize, (long)parenfloor);
363 SSGROW(total_elems + REGCP_FRAME_ELEMS);
366 if ((int)PL_regsize > (int)parenfloor)
367 PerlIO_printf(Perl_debug_log,
368 "rex=0x%"UVxf" offs=0x%"UVxf": saving capture indices:\n",
373 for (p = PL_regsize; p > parenfloor; p--) {
374 /* REGCP_PARENS_ELEMS are pushed per pairs of parentheses. */
375 SSPUSHINT(rex->offs[p].end);
376 SSPUSHINT(rex->offs[p].start);
377 SSPUSHINT(rex->offs[p].start_tmp);
379 DEBUG_BUFFERS_r(PerlIO_printf(Perl_debug_log,
380 " \\%"UVuf": %"IVdf"(%"IVdf")..%"IVdf"\n",
382 (IV)rex->offs[p].start,
383 (IV)rex->offs[p].start_tmp,
387 /* REGCP_OTHER_ELEMS are pushed in any case, parentheses or no. */
388 SSPUSHINT(PL_regsize);
389 SSPUSHINT(rex->lastparen);
390 SSPUSHINT(rex->lastcloseparen);
391 SSPUSHUV(SAVEt_REGCONTEXT | elems_shifted); /* Magic cookie. */
396 /* These are needed since we do not localize EVAL nodes: */
397 #define REGCP_SET(cp) \
399 PerlIO_printf(Perl_debug_log, \
400 " Setting an EVAL scope, savestack=%"IVdf"\n", \
401 (IV)PL_savestack_ix)); \
404 #define REGCP_UNWIND(cp) \
406 if (cp != PL_savestack_ix) \
407 PerlIO_printf(Perl_debug_log, \
408 " Clearing an EVAL scope, savestack=%"IVdf"..%"IVdf"\n", \
409 (IV)(cp), (IV)PL_savestack_ix)); \
413 S_regcppop(pTHX_ regexp *rex)
417 GET_RE_DEBUG_FLAGS_DECL;
419 PERL_ARGS_ASSERT_REGCPPOP;
421 /* Pop REGCP_OTHER_ELEMS before the parentheses loop starts. */
423 assert((i & SAVE_MASK) == SAVEt_REGCONTEXT); /* Check that the magic cookie is there. */
424 i >>= SAVE_TIGHT_SHIFT; /* Parentheses elements to pop. */
425 rex->lastcloseparen = SSPOPINT;
426 rex->lastparen = SSPOPINT;
427 PL_regsize = SSPOPINT;
429 i -= REGCP_OTHER_ELEMS;
430 /* Now restore the parentheses context. */
432 if (i || rex->lastparen + 1 <= rex->nparens)
433 PerlIO_printf(Perl_debug_log,
434 "rex=0x%"UVxf" offs=0x%"UVxf": restoring capture indices to:\n",
439 for ( ; i > 0; i -= REGCP_PAREN_ELEMS) {
441 U32 paren = (U32)SSPOPINT;
442 rex->offs[paren].start_tmp = SSPOPINT;
443 rex->offs[paren].start = SSPOPINT;
445 if (paren <= rex->lastparen)
446 rex->offs[paren].end = tmps;
447 DEBUG_BUFFERS_r( PerlIO_printf(Perl_debug_log,
448 " \\%"UVuf": %"IVdf"(%"IVdf")..%"IVdf"%s\n",
450 (IV)rex->offs[paren].start,
451 (IV)rex->offs[paren].start_tmp,
452 (IV)rex->offs[paren].end,
453 (paren > rex->lastparen ? "(skipped)" : ""));
457 /* It would seem that the similar code in regtry()
458 * already takes care of this, and in fact it is in
459 * a better location to since this code can #if 0-ed out
460 * but the code in regtry() is needed or otherwise tests
461 * requiring null fields (pat.t#187 and split.t#{13,14}
462 * (as of patchlevel 7877) will fail. Then again,
463 * this code seems to be necessary or otherwise
464 * this erroneously leaves $1 defined: "1" =~ /^(?:(\d)x)?\d$/
465 * --jhi updated by dapm */
466 for (i = rex->lastparen + 1; i <= rex->nparens; i++) {
468 rex->offs[i].start = -1;
469 rex->offs[i].end = -1;
470 DEBUG_BUFFERS_r( PerlIO_printf(Perl_debug_log,
471 " \\%"UVuf": %s ..-1 undeffing\n",
473 (i > PL_regsize) ? "-1" : " "
479 #define regcpblow(cp) LEAVE_SCOPE(cp) /* Ignores regcppush()ed data. */
482 * pregexec and friends
485 #ifndef PERL_IN_XSUB_RE
487 - pregexec - match a regexp against a string
490 Perl_pregexec(pTHX_ REGEXP * const prog, char* stringarg, register char *strend,
491 char *strbeg, I32 minend, SV *screamer, U32 nosave)
492 /* strend: pointer to null at end of string */
493 /* strbeg: real beginning of string */
494 /* minend: end of match must be >=minend after stringarg. */
495 /* nosave: For optimizations. */
497 PERL_ARGS_ASSERT_PREGEXEC;
500 regexec_flags(prog, stringarg, strend, strbeg, minend, screamer, NULL,
501 nosave ? 0 : REXEC_COPY_STR);
506 * Need to implement the following flags for reg_anch:
508 * USE_INTUIT_NOML - Useful to call re_intuit_start() first
510 * INTUIT_AUTORITATIVE_NOML - Can trust a positive answer
511 * INTUIT_AUTORITATIVE_ML
512 * INTUIT_ONCE_NOML - Intuit can match in one location only.
515 * Another flag for this function: SECOND_TIME (so that float substrs
516 * with giant delta may be not rechecked).
519 /* Assumptions: if ANCH_GPOS, then strpos is anchored. XXXX Check GPOS logic */
521 /* If SCREAM, then SvPVX_const(sv) should be compatible with strpos and strend.
522 Otherwise, only SvCUR(sv) is used to get strbeg. */
524 /* XXXX We assume that strpos is strbeg unless sv. */
526 /* XXXX Some places assume that there is a fixed substring.
527 An update may be needed if optimizer marks as "INTUITable"
528 RExen without fixed substrings. Similarly, it is assumed that
529 lengths of all the strings are no more than minlen, thus they
530 cannot come from lookahead.
531 (Or minlen should take into account lookahead.)
532 NOTE: Some of this comment is not correct. minlen does now take account
533 of lookahead/behind. Further research is required. -- demerphq
537 /* A failure to find a constant substring means that there is no need to make
538 an expensive call to REx engine, thus we celebrate a failure. Similarly,
539 finding a substring too deep into the string means that less calls to
540 regtry() should be needed.
542 REx compiler's optimizer found 4 possible hints:
543 a) Anchored substring;
545 c) Whether we are anchored (beginning-of-line or \G);
546 d) First node (of those at offset 0) which may distinguish positions;
547 We use a)b)d) and multiline-part of c), and try to find a position in the
548 string which does not contradict any of them.
551 /* Most of decisions we do here should have been done at compile time.
552 The nodes of the REx which we used for the search should have been
553 deleted from the finite automaton. */
556 Perl_re_intuit_start(pTHX_ REGEXP * const rx, SV *sv, char *strpos,
557 char *strend, const U32 flags, re_scream_pos_data *data)
560 struct regexp *const prog = (struct regexp *)SvANY(rx);
561 register I32 start_shift = 0;
562 /* Should be nonnegative! */
563 register I32 end_shift = 0;
568 const bool utf8_target = (sv && SvUTF8(sv)) ? 1 : 0; /* if no sv we have to assume bytes */
570 register char *other_last = NULL; /* other substr checked before this */
571 char *check_at = NULL; /* check substr found at this pos */
572 char *checked_upto = NULL; /* how far into the string we have already checked using find_byclass*/
573 const I32 multiline = prog->extflags & RXf_PMf_MULTILINE;
574 RXi_GET_DECL(prog,progi);
576 const char * const i_strpos = strpos;
578 GET_RE_DEBUG_FLAGS_DECL;
580 PERL_ARGS_ASSERT_RE_INTUIT_START;
581 PERL_UNUSED_ARG(flags);
582 PERL_UNUSED_ARG(data);
584 RX_MATCH_UTF8_set(rx,utf8_target);
587 PL_reg_flags |= RF_utf8;
590 debug_start_match(rx, utf8_target, strpos, strend,
591 sv ? "Guessing start of match in sv for"
592 : "Guessing start of match in string for");
595 /* CHR_DIST() would be more correct here but it makes things slow. */
596 if (prog->minlen > strend - strpos) {
597 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
598 "String too short... [re_intuit_start]\n"));
602 strbeg = (sv && SvPOK(sv)) ? strend - SvCUR(sv) : strpos;
605 if (!prog->check_utf8 && prog->check_substr)
606 to_utf8_substr(prog);
607 check = prog->check_utf8;
609 if (!prog->check_substr && prog->check_utf8)
610 to_byte_substr(prog);
611 check = prog->check_substr;
613 if (check == &PL_sv_undef) {
614 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
615 "Non-utf8 string cannot match utf8 check string\n"));
618 if (prog->extflags & RXf_ANCH) { /* Match at beg-of-str or after \n */
619 ml_anch = !( (prog->extflags & RXf_ANCH_SINGLE)
620 || ( (prog->extflags & RXf_ANCH_BOL)
621 && !multiline ) ); /* Check after \n? */
624 if ( !(prog->extflags & RXf_ANCH_GPOS) /* Checked by the caller */
625 && !(prog->intflags & PREGf_IMPLICIT) /* not a real BOL */
626 /* SvCUR is not set on references: SvRV and SvPVX_const overlap */
628 && (strpos != strbeg)) {
629 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Not at start...\n"));
632 if (prog->check_offset_min == prog->check_offset_max &&
633 !(prog->extflags & RXf_CANY_SEEN)) {
634 /* Substring at constant offset from beg-of-str... */
637 s = HOP3c(strpos, prog->check_offset_min, strend);
640 slen = SvCUR(check); /* >= 1 */
642 if ( strend - s > slen || strend - s < slen - 1
643 || (strend - s == slen && strend[-1] != '\n')) {
644 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "String too long...\n"));
647 /* Now should match s[0..slen-2] */
649 if (slen && (*SvPVX_const(check) != *s
651 && memNE(SvPVX_const(check), s, slen)))) {
653 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "String not equal...\n"));
657 else if (*SvPVX_const(check) != *s
658 || ((slen = SvCUR(check)) > 1
659 && memNE(SvPVX_const(check), s, slen)))
662 goto success_at_start;
665 /* Match is anchored, but substr is not anchored wrt beg-of-str. */
667 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
668 end_shift = prog->check_end_shift;
671 const I32 end = prog->check_offset_max + CHR_SVLEN(check)
672 - (SvTAIL(check) != 0);
673 const I32 eshift = CHR_DIST((U8*)strend, (U8*)s) - end;
675 if (end_shift < eshift)
679 else { /* Can match at random position */
682 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
683 end_shift = prog->check_end_shift;
685 /* end shift should be non negative here */
688 #ifdef QDEBUGGING /* 7/99: reports of failure (with the older version) */
690 Perl_croak(aTHX_ "panic: end_shift: %"IVdf" pattern:\n%s\n ",
691 (IV)end_shift, RX_PRECOMP(prog));
695 /* Find a possible match in the region s..strend by looking for
696 the "check" substring in the region corrected by start/end_shift. */
699 I32 srch_start_shift = start_shift;
700 I32 srch_end_shift = end_shift;
703 if (srch_start_shift < 0 && strbeg - s > srch_start_shift) {
704 srch_end_shift -= ((strbeg - s) - srch_start_shift);
705 srch_start_shift = strbeg - s;
707 DEBUG_OPTIMISE_MORE_r({
708 PerlIO_printf(Perl_debug_log, "Check offset min: %"IVdf" Start shift: %"IVdf" End shift %"IVdf" Real End Shift: %"IVdf"\n",
709 (IV)prog->check_offset_min,
710 (IV)srch_start_shift,
712 (IV)prog->check_end_shift);
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);
732 /* Update the count-of-usability, remove useless subpatterns,
736 RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0),
737 SvPVX_const(check), RE_SV_DUMPLEN(check), 30);
738 PerlIO_printf(Perl_debug_log, "%s %s substr %s%s%s",
739 (s ? "Found" : "Did not find"),
740 (check == (utf8_target ? prog->anchored_utf8 : prog->anchored_substr)
741 ? "anchored" : "floating"),
744 (s ? " at offset " : "...\n") );
749 /* Finish the diagnostic message */
750 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%ld...\n", (long)(s - i_strpos)) );
752 /* XXX dmq: first branch is for positive lookbehind...
753 Our check string is offset from the beginning of the pattern.
754 So we need to do any stclass tests offset forward from that
763 /* Got a candidate. Check MBOL anchoring, and the *other* substr.
764 Start with the other substr.
765 XXXX no SCREAM optimization yet - and a very coarse implementation
766 XXXX /ttx+/ results in anchored="ttx", floating="x". floating will
767 *always* match. Probably should be marked during compile...
768 Probably it is right to do no SCREAM here...
771 if (utf8_target ? (prog->float_utf8 && prog->anchored_utf8)
772 : (prog->float_substr && prog->anchored_substr))
774 /* Take into account the "other" substring. */
775 /* XXXX May be hopelessly wrong for UTF... */
778 if (check == (utf8_target ? prog->float_utf8 : prog->float_substr)) {
781 char * const last = HOP3c(s, -start_shift, strbeg);
783 char * const saved_s = s;
786 t = s - prog->check_offset_max;
787 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
789 || ((t = (char*)reghopmaybe3((U8*)s, -(prog->check_offset_max), (U8*)strpos))
794 t = HOP3c(t, prog->anchored_offset, strend);
795 if (t < other_last) /* These positions already checked */
797 last2 = last1 = HOP3c(strend, -prog->minlen, strbeg);
800 /* XXXX It is not documented what units *_offsets are in.
801 We assume bytes, but this is clearly wrong.
802 Meaning this code needs to be carefully reviewed for errors.
806 /* On end-of-str: see comment below. */
807 must = utf8_target ? prog->anchored_utf8 : prog->anchored_substr;
808 if (must == &PL_sv_undef) {
810 DEBUG_r(must = prog->anchored_utf8); /* for debug */
815 HOP3(HOP3(last1, prog->anchored_offset, strend)
816 + SvCUR(must), -(SvTAIL(must)!=0), strbeg),
818 multiline ? FBMrf_MULTILINE : 0
821 RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0),
822 SvPVX_const(must), RE_SV_DUMPLEN(must), 30);
823 PerlIO_printf(Perl_debug_log, "%s anchored substr %s%s",
824 (s ? "Found" : "Contradicts"),
825 quoted, RE_SV_TAIL(must));
830 if (last1 >= last2) {
831 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
832 ", giving up...\n"));
835 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
836 ", trying floating at offset %ld...\n",
837 (long)(HOP3c(saved_s, 1, strend) - i_strpos)));
838 other_last = HOP3c(last1, prog->anchored_offset+1, strend);
839 s = HOP3c(last, 1, strend);
843 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
844 (long)(s - i_strpos)));
845 t = HOP3c(s, -prog->anchored_offset, strbeg);
846 other_last = HOP3c(s, 1, strend);
854 else { /* Take into account the floating substring. */
856 char * const saved_s = s;
859 t = HOP3c(s, -start_shift, strbeg);
861 HOP3c(strend, -prog->minlen + prog->float_min_offset, strbeg);
862 if (CHR_DIST((U8*)last, (U8*)t) > prog->float_max_offset)
863 last = HOP3c(t, prog->float_max_offset, strend);
864 s = HOP3c(t, prog->float_min_offset, strend);
867 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
868 must = utf8_target ? prog->float_utf8 : prog->float_substr;
869 /* fbm_instr() takes into account exact value of end-of-str
870 if the check is SvTAIL(ed). Since false positives are OK,
871 and end-of-str is not later than strend we are OK. */
872 if (must == &PL_sv_undef) {
874 DEBUG_r(must = prog->float_utf8); /* for debug message */
877 s = fbm_instr((unsigned char*)s,
878 (unsigned char*)last + SvCUR(must)
880 must, multiline ? FBMrf_MULTILINE : 0);
882 RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0),
883 SvPVX_const(must), RE_SV_DUMPLEN(must), 30);
884 PerlIO_printf(Perl_debug_log, "%s floating substr %s%s",
885 (s ? "Found" : "Contradicts"),
886 quoted, RE_SV_TAIL(must));
890 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
891 ", giving up...\n"));
894 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
895 ", trying anchored starting at offset %ld...\n",
896 (long)(saved_s + 1 - i_strpos)));
898 s = HOP3c(t, 1, strend);
902 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
903 (long)(s - i_strpos)));
904 other_last = s; /* Fix this later. --Hugo */
914 t= (char*)HOP3( s, -prog->check_offset_max, (prog->check_offset_max<0) ? strend : strpos);
916 DEBUG_OPTIMISE_MORE_r(
917 PerlIO_printf(Perl_debug_log,
918 "Check offset min:%"IVdf" max:%"IVdf" S:%"IVdf" t:%"IVdf" D:%"IVdf" end:%"IVdf"\n",
919 (IV)prog->check_offset_min,
920 (IV)prog->check_offset_max,
928 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
930 || ((t = (char*)reghopmaybe3((U8*)s, -prog->check_offset_max, (U8*) ((prog->check_offset_max<0) ? strend : strpos)))
933 /* Fixed substring is found far enough so that the match
934 cannot start at strpos. */
936 if (ml_anch && t[-1] != '\n') {
937 /* Eventually fbm_*() should handle this, but often
938 anchored_offset is not 0, so this check will not be wasted. */
939 /* XXXX In the code below we prefer to look for "^" even in
940 presence of anchored substrings. And we search even
941 beyond the found float position. These pessimizations
942 are historical artefacts only. */
944 while (t < strend - prog->minlen) {
946 if (t < check_at - prog->check_offset_min) {
947 if (utf8_target ? prog->anchored_utf8 : prog->anchored_substr) {
948 /* Since we moved from the found position,
949 we definitely contradict the found anchored
950 substr. Due to the above check we do not
951 contradict "check" substr.
952 Thus we can arrive here only if check substr
953 is float. Redo checking for "other"=="fixed".
956 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld, rescanning for anchored from offset %ld...\n",
957 PL_colors[0], PL_colors[1], (long)(strpos - i_strpos), (long)(strpos - i_strpos + prog->anchored_offset)));
958 goto do_other_anchored;
960 /* We don't contradict the found floating substring. */
961 /* XXXX Why not check for STCLASS? */
963 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld...\n",
964 PL_colors[0], PL_colors[1], (long)(s - i_strpos)));
967 /* Position contradicts check-string */
968 /* XXXX probably better to look for check-string
969 than for "\n", so one should lower the limit for t? */
970 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m, restarting lookup for check-string at offset %ld...\n",
971 PL_colors[0], PL_colors[1], (long)(t + 1 - i_strpos)));
972 other_last = strpos = s = t + 1;
977 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Did not find /%s^%s/m...\n",
978 PL_colors[0], PL_colors[1]));
982 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Starting position does not contradict /%s^%s/m...\n",
983 PL_colors[0], PL_colors[1]));
987 ++BmUSEFUL(utf8_target ? prog->check_utf8 : prog->check_substr); /* hooray/5 */
990 /* The found string does not prohibit matching at strpos,
991 - no optimization of calling REx engine can be performed,
992 unless it was an MBOL and we are not after MBOL,
993 or a future STCLASS check will fail this. */
995 /* Even in this situation we may use MBOL flag if strpos is offset
996 wrt the start of the string. */
997 if (ml_anch && sv && !SvROK(sv) /* See prev comment on SvROK */
998 && (strpos != strbeg) && strpos[-1] != '\n'
999 /* May be due to an implicit anchor of m{.*foo} */
1000 && !(prog->intflags & PREGf_IMPLICIT))
1005 DEBUG_EXECUTE_r( if (ml_anch)
1006 PerlIO_printf(Perl_debug_log, "Position at offset %ld does not contradict /%s^%s/m...\n",
1007 (long)(strpos - i_strpos), PL_colors[0], PL_colors[1]);
1010 if (!(prog->intflags & PREGf_NAUGHTY) /* XXXX If strpos moved? */
1012 prog->check_utf8 /* Could be deleted already */
1013 && --BmUSEFUL(prog->check_utf8) < 0
1014 && (prog->check_utf8 == prog->float_utf8)
1016 prog->check_substr /* Could be deleted already */
1017 && --BmUSEFUL(prog->check_substr) < 0
1018 && (prog->check_substr == prog->float_substr)
1021 /* If flags & SOMETHING - do not do it many times on the same match */
1022 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "... Disabling check substring...\n"));
1023 /* XXX Does the destruction order has to change with utf8_target? */
1024 SvREFCNT_dec(utf8_target ? prog->check_utf8 : prog->check_substr);
1025 SvREFCNT_dec(utf8_target ? prog->check_substr : prog->check_utf8);
1026 prog->check_substr = prog->check_utf8 = NULL; /* disable */
1027 prog->float_substr = prog->float_utf8 = NULL; /* clear */
1028 check = NULL; /* abort */
1030 /* XXXX If the check string was an implicit check MBOL, then we need to unset the relevant flag
1031 see http://bugs.activestate.com/show_bug.cgi?id=87173 */
1032 if (prog->intflags & PREGf_IMPLICIT)
1033 prog->extflags &= ~RXf_ANCH_MBOL;
1034 /* XXXX This is a remnant of the old implementation. It
1035 looks wasteful, since now INTUIT can use many
1036 other heuristics. */
1037 prog->extflags &= ~RXf_USE_INTUIT;
1038 /* XXXX What other flags might need to be cleared in this branch? */
1044 /* Last resort... */
1045 /* XXXX BmUSEFUL already changed, maybe multiple change is meaningful... */
1046 /* trie stclasses are too expensive to use here, we are better off to
1047 leave it to regmatch itself */
1048 if (progi->regstclass && PL_regkind[OP(progi->regstclass)]!=TRIE) {
1049 /* minlen == 0 is possible if regstclass is \b or \B,
1050 and the fixed substr is ''$.
1051 Since minlen is already taken into account, s+1 is before strend;
1052 accidentally, minlen >= 1 guaranties no false positives at s + 1
1053 even for \b or \B. But (minlen? 1 : 0) below assumes that
1054 regstclass does not come from lookahead... */
1055 /* If regstclass takes bytelength more than 1: If charlength==1, OK.
1056 This leaves EXACTF-ish only, which are dealt with in find_byclass(). */
1057 const U8* const str = (U8*)STRING(progi->regstclass);
1058 const int cl_l = (PL_regkind[OP(progi->regstclass)] == EXACT
1059 ? CHR_DIST(str+STR_LEN(progi->regstclass), str)
1062 if (prog->anchored_substr || prog->anchored_utf8 || ml_anch)
1063 endpos= HOP3c(s, (prog->minlen ? cl_l : 0), strend);
1064 else if (prog->float_substr || prog->float_utf8)
1065 endpos= HOP3c(HOP3c(check_at, -start_shift, strbeg), cl_l, strend);
1069 if (checked_upto < s)
1071 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "start_shift: %"IVdf" check_at: %"IVdf" s: %"IVdf" endpos: %"IVdf" checked_upto: %"IVdf"\n",
1072 (IV)start_shift, (IV)(check_at - strbeg), (IV)(s - strbeg), (IV)(endpos - strbeg), (IV)(checked_upto- strbeg)));
1075 s = find_byclass(prog, progi->regstclass, checked_upto, endpos, NULL);
1080 const char *what = NULL;
1082 if (endpos == strend) {
1083 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
1084 "Could not match STCLASS...\n") );
1087 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
1088 "This position contradicts STCLASS...\n") );
1089 if ((prog->extflags & RXf_ANCH) && !ml_anch)
1091 checked_upto = HOPBACKc(endpos, start_shift);
1092 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "start_shift: %"IVdf" check_at: %"IVdf" endpos: %"IVdf" checked_upto: %"IVdf"\n",
1093 (IV)start_shift, (IV)(check_at - strbeg), (IV)(endpos - strbeg), (IV)(checked_upto- strbeg)));
1094 /* Contradict one of substrings */
1095 if (prog->anchored_substr || prog->anchored_utf8) {
1096 if ((utf8_target ? prog->anchored_utf8 : prog->anchored_substr) == check) {
1097 DEBUG_EXECUTE_r( what = "anchored" );
1099 s = HOP3c(t, 1, strend);
1100 if (s + start_shift + end_shift > strend) {
1101 /* XXXX Should be taken into account earlier? */
1102 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
1103 "Could not match STCLASS...\n") );
1108 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
1109 "Looking for %s substr starting at offset %ld...\n",
1110 what, (long)(s + start_shift - i_strpos)) );
1113 /* Have both, check_string is floating */
1114 if (t + start_shift >= check_at) /* Contradicts floating=check */
1115 goto retry_floating_check;
1116 /* Recheck anchored substring, but not floating... */
1120 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
1121 "Looking for anchored substr starting at offset %ld...\n",
1122 (long)(other_last - i_strpos)) );
1123 goto do_other_anchored;
1125 /* Another way we could have checked stclass at the
1126 current position only: */
1131 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
1132 "Looking for /%s^%s/m starting at offset %ld...\n",
1133 PL_colors[0], PL_colors[1], (long)(t - i_strpos)) );
1136 if (!(utf8_target ? prog->float_utf8 : prog->float_substr)) /* Could have been deleted */
1138 /* Check is floating substring. */
1139 retry_floating_check:
1140 t = check_at - start_shift;
1141 DEBUG_EXECUTE_r( what = "floating" );
1142 goto hop_and_restart;
1145 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
1146 "By STCLASS: moving %ld --> %ld\n",
1147 (long)(t - i_strpos), (long)(s - i_strpos))
1151 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
1152 "Does not contradict STCLASS...\n");
1157 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%s%s:%s match at offset %ld\n",
1158 PL_colors[4], (check ? "Guessed" : "Giving up"),
1159 PL_colors[5], (long)(s - i_strpos)) );
1162 fail_finish: /* Substring not found */
1163 if (prog->check_substr || prog->check_utf8) /* could be removed already */
1164 BmUSEFUL(utf8_target ? prog->check_utf8 : prog->check_substr) += 5; /* hooray */
1166 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch rejected by optimizer%s\n",
1167 PL_colors[4], PL_colors[5]));
1171 #define DECL_TRIE_TYPE(scan) \
1172 const enum { trie_plain, trie_utf8, trie_utf8_fold, trie_latin_utf8_fold } \
1173 trie_type = ((scan->flags == EXACT) \
1174 ? (utf8_target ? trie_utf8 : trie_plain) \
1175 : (utf8_target ? trie_utf8_fold : trie_latin_utf8_fold))
1177 #define REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc, uscan, len, \
1178 uvc, charid, foldlen, foldbuf, uniflags) STMT_START { \
1180 switch (trie_type) { \
1181 case trie_utf8_fold: \
1182 if ( foldlen>0 ) { \
1183 uvc = utf8n_to_uvuni( (const U8*) uscan, UTF8_MAXLEN, &len, uniflags ); \
1188 uvc = to_utf8_fold( (const U8*) uc, foldbuf, &foldlen ); \
1189 len = UTF8SKIP(uc); \
1190 skiplen = UNISKIP( uvc ); \
1191 foldlen -= skiplen; \
1192 uscan = foldbuf + skiplen; \
1195 case trie_latin_utf8_fold: \
1196 if ( foldlen>0 ) { \
1197 uvc = utf8n_to_uvuni( (const U8*) uscan, UTF8_MAXLEN, &len, uniflags ); \
1203 uvc = _to_fold_latin1( (U8) *uc, foldbuf, &foldlen, 1); \
1204 skiplen = UNISKIP( uvc ); \
1205 foldlen -= skiplen; \
1206 uscan = foldbuf + skiplen; \
1210 uvc = utf8n_to_uvuni( (const U8*) uc, UTF8_MAXLEN, &len, uniflags ); \
1217 charid = trie->charmap[ uvc ]; \
1221 if (widecharmap) { \
1222 SV** const svpp = hv_fetch(widecharmap, \
1223 (char*)&uvc, sizeof(UV), 0); \
1225 charid = (U16)SvIV(*svpp); \
1230 #define REXEC_FBC_EXACTISH_SCAN(CoNd) \
1234 && (ln == 1 || folder(s, pat_string, ln)) \
1235 && (!reginfo || regtry(reginfo, &s)) ) \
1241 #define REXEC_FBC_UTF8_SCAN(CoDe) \
1243 while (s + (uskip = UTF8SKIP(s)) <= strend) { \
1249 #define REXEC_FBC_SCAN(CoDe) \
1251 while (s < strend) { \
1257 #define REXEC_FBC_UTF8_CLASS_SCAN(CoNd) \
1258 REXEC_FBC_UTF8_SCAN( \
1260 if (tmp && (!reginfo || regtry(reginfo, &s))) \
1269 #define REXEC_FBC_CLASS_SCAN(CoNd) \
1272 if (tmp && (!reginfo || regtry(reginfo, &s))) \
1281 #define REXEC_FBC_TRYIT \
1282 if ((!reginfo || regtry(reginfo, &s))) \
1285 #define REXEC_FBC_CSCAN(CoNdUtF8,CoNd) \
1286 if (utf8_target) { \
1287 REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \
1290 REXEC_FBC_CLASS_SCAN(CoNd); \
1293 #define REXEC_FBC_CSCAN_PRELOAD(UtFpReLoAd,CoNdUtF8,CoNd) \
1294 if (utf8_target) { \
1296 REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \
1299 REXEC_FBC_CLASS_SCAN(CoNd); \
1302 #define REXEC_FBC_CSCAN_TAINT(CoNdUtF8,CoNd) \
1303 PL_reg_flags |= RF_tainted; \
1304 if (utf8_target) { \
1305 REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \
1308 REXEC_FBC_CLASS_SCAN(CoNd); \
1311 #define DUMP_EXEC_POS(li,s,doutf8) \
1312 dump_exec_pos(li,s,(PL_regeol),(PL_bostr),(PL_reg_starttry),doutf8)
1315 #define UTF8_NOLOAD(TEST_NON_UTF8, IF_SUCCESS, IF_FAIL) \
1316 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n'; \
1317 tmp = TEST_NON_UTF8(tmp); \
1318 REXEC_FBC_UTF8_SCAN( \
1319 if (tmp == ! TEST_NON_UTF8((U8) *s)) { \
1328 #define UTF8_LOAD(TeSt1_UtF8, TeSt2_UtF8, IF_SUCCESS, IF_FAIL) \
1329 if (s == PL_bostr) { \
1333 U8 * const r = reghop3((U8*)s, -1, (U8*)PL_bostr); \
1334 tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, UTF8_ALLOW_DEFAULT); \
1337 LOAD_UTF8_CHARCLASS_ALNUM(); \
1338 REXEC_FBC_UTF8_SCAN( \
1339 if (tmp == ! (TeSt2_UtF8)) { \
1348 /* The only difference between the BOUND and NBOUND cases is that
1349 * REXEC_FBC_TRYIT is called when matched in BOUND, and when non-matched in
1350 * NBOUND. This is accomplished by passing it in either the if or else clause,
1351 * with the other one being empty */
1352 #define FBC_BOUND(TEST_NON_UTF8, TEST1_UTF8, TEST2_UTF8) \
1353 FBC_BOUND_COMMON(UTF8_LOAD(TEST1_UTF8, TEST2_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER), TEST_NON_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER)
1355 #define FBC_BOUND_NOLOAD(TEST_NON_UTF8, TEST1_UTF8, TEST2_UTF8) \
1356 FBC_BOUND_COMMON(UTF8_NOLOAD(TEST_NON_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER), TEST_NON_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER)
1358 #define FBC_NBOUND(TEST_NON_UTF8, TEST1_UTF8, TEST2_UTF8) \
1359 FBC_BOUND_COMMON(UTF8_LOAD(TEST1_UTF8, TEST2_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT), TEST_NON_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT)
1361 #define FBC_NBOUND_NOLOAD(TEST_NON_UTF8, TEST1_UTF8, TEST2_UTF8) \
1362 FBC_BOUND_COMMON(UTF8_NOLOAD(TEST_NON_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT), TEST_NON_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT)
1365 /* Common to the BOUND and NBOUND cases. Unfortunately the UTF8 tests need to
1366 * be passed in completely with the variable name being tested, which isn't
1367 * such a clean interface, but this is easier to read than it was before. We
1368 * are looking for the boundary (or non-boundary between a word and non-word
1369 * character. The utf8 and non-utf8 cases have the same logic, but the details
1370 * must be different. Find the "wordness" of the character just prior to this
1371 * one, and compare it with the wordness of this one. If they differ, we have
1372 * a boundary. At the beginning of the string, pretend that the previous
1373 * character was a new-line */
1374 #define FBC_BOUND_COMMON(UTF8_CODE, TEST_NON_UTF8, IF_SUCCESS, IF_FAIL) \
1375 if (utf8_target) { \
1378 else { /* Not utf8 */ \
1379 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n'; \
1380 tmp = TEST_NON_UTF8(tmp); \
1382 if (tmp == ! TEST_NON_UTF8((U8) *s)) { \
1391 if ((!prog->minlen && tmp) && (!reginfo || regtry(reginfo, &s))) \
1394 /* We know what class REx starts with. Try to find this position... */
1395 /* if reginfo is NULL, its a dryrun */
1396 /* annoyingly all the vars in this routine have different names from their counterparts
1397 in regmatch. /grrr */
1400 S_find_byclass(pTHX_ regexp * prog, const regnode *c, char *s,
1401 const char *strend, regmatch_info *reginfo)
1404 const I32 doevery = (prog->intflags & PREGf_SKIP) == 0;
1405 char *pat_string; /* The pattern's exactish string */
1406 char *pat_end; /* ptr to end char of pat_string */
1407 re_fold_t folder; /* Function for computing non-utf8 folds */
1408 const U8 *fold_array; /* array for folding ords < 256 */
1411 register STRLEN uskip;
1415 register I32 tmp = 1; /* Scratch variable? */
1416 register const bool utf8_target = PL_reg_match_utf8;
1417 UV utf8_fold_flags = 0;
1418 RXi_GET_DECL(prog,progi);
1420 PERL_ARGS_ASSERT_FIND_BYCLASS;
1422 /* We know what class it must start with. */
1426 if (utf8_target || OP(c) == ANYOFV) {
1427 STRLEN inclasslen = strend - s;
1428 REXEC_FBC_UTF8_CLASS_SCAN(
1429 reginclass(prog, c, (U8*)s, &inclasslen, utf8_target));
1432 REXEC_FBC_CLASS_SCAN(REGINCLASS(prog, c, (U8*)s));
1437 if (tmp && (!reginfo || regtry(reginfo, &s)))
1445 if (UTF_PATTERN || utf8_target) {
1446 utf8_fold_flags = FOLDEQ_UTF8_NOMIX_ASCII;
1447 goto do_exactf_utf8;
1449 fold_array = PL_fold_latin1; /* Latin1 folds are not affected by */
1450 folder = foldEQ_latin1; /* /a, except the sharp s one which */
1451 goto do_exactf_non_utf8; /* isn't dealt with by these */
1456 /* regcomp.c already folded this if pattern is in UTF-8 */
1457 utf8_fold_flags = 0;
1458 goto do_exactf_utf8;
1460 fold_array = PL_fold;
1462 goto do_exactf_non_utf8;
1465 if (UTF_PATTERN || utf8_target) {
1466 utf8_fold_flags = FOLDEQ_UTF8_LOCALE;
1467 goto do_exactf_utf8;
1469 fold_array = PL_fold_locale;
1470 folder = foldEQ_locale;
1471 goto do_exactf_non_utf8;
1475 utf8_fold_flags = FOLDEQ_S2_ALREADY_FOLDED;
1477 goto do_exactf_utf8;
1479 case EXACTFU_TRICKYFOLD:
1481 if (UTF_PATTERN || utf8_target) {
1482 utf8_fold_flags = (UTF_PATTERN) ? FOLDEQ_S2_ALREADY_FOLDED : 0;
1483 goto do_exactf_utf8;
1486 /* Any 'ss' in the pattern should have been replaced by regcomp,
1487 * so we don't have to worry here about this single special case
1488 * in the Latin1 range */
1489 fold_array = PL_fold_latin1;
1490 folder = foldEQ_latin1;
1494 do_exactf_non_utf8: /* Neither pattern nor string are UTF8, and there
1495 are no glitches with fold-length differences
1496 between the target string and pattern */
1498 /* The idea in the non-utf8 EXACTF* cases is to first find the
1499 * first character of the EXACTF* node and then, if necessary,
1500 * case-insensitively compare the full text of the node. c1 is the
1501 * first character. c2 is its fold. This logic will not work for
1502 * Unicode semantics and the german sharp ss, which hence should
1503 * not be compiled into a node that gets here. */
1504 pat_string = STRING(c);
1505 ln = STR_LEN(c); /* length to match in octets/bytes */
1507 /* We know that we have to match at least 'ln' bytes (which is the
1508 * same as characters, since not utf8). If we have to match 3
1509 * characters, and there are only 2 availabe, we know without
1510 * trying that it will fail; so don't start a match past the
1511 * required minimum number from the far end */
1512 e = HOP3c(strend, -((I32)ln), s);
1514 if (!reginfo && e < s) {
1515 e = s; /* Due to minlen logic of intuit() */
1519 c2 = fold_array[c1];
1520 if (c1 == c2) { /* If char and fold are the same */
1521 REXEC_FBC_EXACTISH_SCAN(*(U8*)s == c1);
1524 REXEC_FBC_EXACTISH_SCAN(*(U8*)s == c1 || *(U8*)s == c2);
1533 /* If one of the operands is in utf8, we can't use the simpler
1534 * folding above, due to the fact that many different characters
1535 * can have the same fold, or portion of a fold, or different-
1537 pat_string = STRING(c);
1538 ln = STR_LEN(c); /* length to match in octets/bytes */
1539 pat_end = pat_string + ln;
1540 lnc = (UTF_PATTERN) /* length to match in characters */
1541 ? utf8_length((U8 *) pat_string, (U8 *) pat_end)
1544 /* We have 'lnc' characters to match in the pattern, but because of
1545 * multi-character folding, each character in the target can match
1546 * up to 3 characters (Unicode guarantees it will never exceed
1547 * this) if it is utf8-encoded; and up to 2 if not (based on the
1548 * fact that the Latin 1 folds are already determined, and the
1549 * only multi-char fold in that range is the sharp-s folding to
1550 * 'ss'. Thus, a pattern character can match as little as 1/3 of a
1551 * string character. Adjust lnc accordingly, rounding up, so that
1552 * if we need to match at least 4+1/3 chars, that really is 5. */
1553 expansion = (utf8_target) ? UTF8_MAX_FOLD_CHAR_EXPAND : 2;
1554 lnc = (lnc + expansion - 1) / expansion;
1556 /* As in the non-UTF8 case, if we have to match 3 characters, and
1557 * only 2 are left, it's guaranteed to fail, so don't start a
1558 * match that would require us to go beyond the end of the string
1560 e = HOP3c(strend, -((I32)lnc), s);
1562 if (!reginfo && e < s) {
1563 e = s; /* Due to minlen logic of intuit() */
1566 /* XXX Note that we could recalculate e to stop the loop earlier,
1567 * as the worst case expansion above will rarely be met, and as we
1568 * go along we would usually find that e moves further to the left.
1569 * This would happen only after we reached the point in the loop
1570 * where if there were no expansion we should fail. Unclear if
1571 * worth the expense */
1574 char *my_strend= (char *)strend;
1575 if (foldEQ_utf8_flags(s, &my_strend, 0, utf8_target,
1576 pat_string, NULL, ln, cBOOL(UTF_PATTERN), utf8_fold_flags)
1577 && (!reginfo || regtry(reginfo, &s)) )
1581 s += (utf8_target) ? UTF8SKIP(s) : 1;
1586 PL_reg_flags |= RF_tainted;
1587 FBC_BOUND(isALNUM_LC,
1588 isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp)),
1589 isALNUM_LC_utf8((U8*)s));
1592 PL_reg_flags |= RF_tainted;
1593 FBC_NBOUND(isALNUM_LC,
1594 isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp)),
1595 isALNUM_LC_utf8((U8*)s));
1598 FBC_BOUND(isWORDCHAR,
1600 cBOOL(swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target)));
1603 FBC_BOUND_NOLOAD(isWORDCHAR_A,
1605 isWORDCHAR_A((U8*)s));
1608 FBC_NBOUND(isWORDCHAR,
1610 cBOOL(swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target)));
1613 FBC_NBOUND_NOLOAD(isWORDCHAR_A,
1615 isWORDCHAR_A((U8*)s));
1618 FBC_BOUND(isWORDCHAR_L1,
1620 cBOOL(swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target)));
1623 FBC_NBOUND(isWORDCHAR_L1,
1625 cBOOL(swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target)));
1628 REXEC_FBC_CSCAN_TAINT(
1629 isALNUM_LC_utf8((U8*)s),
1634 REXEC_FBC_CSCAN_PRELOAD(
1635 LOAD_UTF8_CHARCLASS_ALNUM(),
1636 swash_fetch(PL_utf8_alnum,(U8*)s, utf8_target),
1637 isWORDCHAR_L1((U8) *s)
1641 REXEC_FBC_CSCAN_PRELOAD(
1642 LOAD_UTF8_CHARCLASS_ALNUM(),
1643 swash_fetch(PL_utf8_alnum,(U8*)s, utf8_target),
1648 /* Don't need to worry about utf8, as it can match only a single
1649 * byte invariant character */
1650 REXEC_FBC_CLASS_SCAN( isWORDCHAR_A(*s));
1653 REXEC_FBC_CSCAN_PRELOAD(
1654 LOAD_UTF8_CHARCLASS_ALNUM(),
1655 !swash_fetch(PL_utf8_alnum,(U8*)s, utf8_target),
1656 ! isWORDCHAR_L1((U8) *s)
1660 REXEC_FBC_CSCAN_PRELOAD(
1661 LOAD_UTF8_CHARCLASS_ALNUM(),
1662 !swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target),
1673 REXEC_FBC_CSCAN_TAINT(
1674 !isALNUM_LC_utf8((U8*)s),
1679 REXEC_FBC_CSCAN_PRELOAD(
1680 LOAD_UTF8_CHARCLASS_SPACE(),
1681 *s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, utf8_target),
1686 REXEC_FBC_CSCAN_PRELOAD(
1687 LOAD_UTF8_CHARCLASS_SPACE(),
1688 *s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, utf8_target),
1693 /* Don't need to worry about utf8, as it can match only a single
1694 * byte invariant character */
1695 REXEC_FBC_CLASS_SCAN( isSPACE_A(*s));
1698 REXEC_FBC_CSCAN_TAINT(
1699 isSPACE_LC_utf8((U8*)s),
1704 REXEC_FBC_CSCAN_PRELOAD(
1705 LOAD_UTF8_CHARCLASS_SPACE(),
1706 !( *s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, utf8_target)),
1707 ! isSPACE_L1((U8) *s)
1711 REXEC_FBC_CSCAN_PRELOAD(
1712 LOAD_UTF8_CHARCLASS_SPACE(),
1713 !(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, utf8_target)),
1724 REXEC_FBC_CSCAN_TAINT(
1725 !isSPACE_LC_utf8((U8*)s),
1730 REXEC_FBC_CSCAN_PRELOAD(
1731 LOAD_UTF8_CHARCLASS_DIGIT(),
1732 swash_fetch(PL_utf8_digit,(U8*)s, utf8_target),
1737 /* Don't need to worry about utf8, as it can match only a single
1738 * byte invariant character */
1739 REXEC_FBC_CLASS_SCAN( isDIGIT_A(*s));
1742 REXEC_FBC_CSCAN_TAINT(
1743 isDIGIT_LC_utf8((U8*)s),
1748 REXEC_FBC_CSCAN_PRELOAD(
1749 LOAD_UTF8_CHARCLASS_DIGIT(),
1750 !swash_fetch(PL_utf8_digit,(U8*)s, utf8_target),
1761 REXEC_FBC_CSCAN_TAINT(
1762 !isDIGIT_LC_utf8((U8*)s),
1769 is_LNBREAK_latin1(s)
1781 !is_VERTWS_latin1(s)
1787 is_HORIZWS_latin1(s)
1792 !is_HORIZWS_utf8(s),
1793 !is_HORIZWS_latin1(s)
1800 /* what trie are we using right now */
1802 = (reg_ac_data*)progi->data->data[ ARG( c ) ];
1804 = (reg_trie_data*)progi->data->data[ aho->trie ];
1805 HV *widecharmap = MUTABLE_HV(progi->data->data[ aho->trie + 1 ]);
1807 const char *last_start = strend - trie->minlen;
1809 const char *real_start = s;
1811 STRLEN maxlen = trie->maxlen;
1813 U8 **points; /* map of where we were in the input string
1814 when reading a given char. For ASCII this
1815 is unnecessary overhead as the relationship
1816 is always 1:1, but for Unicode, especially
1817 case folded Unicode this is not true. */
1818 U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ];
1822 GET_RE_DEBUG_FLAGS_DECL;
1824 /* We can't just allocate points here. We need to wrap it in
1825 * an SV so it gets freed properly if there is a croak while
1826 * running the match */
1829 sv_points=newSV(maxlen * sizeof(U8 *));
1830 SvCUR_set(sv_points,
1831 maxlen * sizeof(U8 *));
1832 SvPOK_on(sv_points);
1833 sv_2mortal(sv_points);
1834 points=(U8**)SvPV_nolen(sv_points );
1835 if ( trie_type != trie_utf8_fold
1836 && (trie->bitmap || OP(c)==AHOCORASICKC) )
1839 bitmap=(U8*)trie->bitmap;
1841 bitmap=(U8*)ANYOF_BITMAP(c);
1843 /* this is the Aho-Corasick algorithm modified a touch
1844 to include special handling for long "unknown char"
1845 sequences. The basic idea being that we use AC as long
1846 as we are dealing with a possible matching char, when
1847 we encounter an unknown char (and we have not encountered
1848 an accepting state) we scan forward until we find a legal
1850 AC matching is basically that of trie matching, except
1851 that when we encounter a failing transition, we fall back
1852 to the current states "fail state", and try the current char
1853 again, a process we repeat until we reach the root state,
1854 state 1, or a legal transition. If we fail on the root state
1855 then we can either terminate if we have reached an accepting
1856 state previously, or restart the entire process from the beginning
1860 while (s <= last_start) {
1861 const U32 uniflags = UTF8_ALLOW_DEFAULT;
1869 U8 *uscan = (U8*)NULL;
1870 U8 *leftmost = NULL;
1872 U32 accepted_word= 0;
1876 while ( state && uc <= (U8*)strend ) {
1878 U32 word = aho->states[ state ].wordnum;
1882 DEBUG_TRIE_EXECUTE_r(
1883 if ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) {
1884 dump_exec_pos( (char *)uc, c, strend, real_start,
1885 (char *)uc, utf8_target );
1886 PerlIO_printf( Perl_debug_log,
1887 " Scanning for legal start char...\n");
1891 while ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) {
1895 while ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) {
1901 if (uc >(U8*)last_start) break;
1905 U8 *lpos= points[ (pointpos - trie->wordinfo[word].len) % maxlen ];
1906 if (!leftmost || lpos < leftmost) {
1907 DEBUG_r(accepted_word=word);
1913 points[pointpos++ % maxlen]= uc;
1914 REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc,
1915 uscan, len, uvc, charid, foldlen,
1917 DEBUG_TRIE_EXECUTE_r({
1918 dump_exec_pos( (char *)uc, c, strend, real_start,
1920 PerlIO_printf(Perl_debug_log,
1921 " Charid:%3u CP:%4"UVxf" ",
1927 word = aho->states[ state ].wordnum;
1929 base = aho->states[ state ].trans.base;
1931 DEBUG_TRIE_EXECUTE_r({
1933 dump_exec_pos( (char *)uc, c, strend, real_start,
1935 PerlIO_printf( Perl_debug_log,
1936 "%sState: %4"UVxf", word=%"UVxf,
1937 failed ? " Fail transition to " : "",
1938 (UV)state, (UV)word);
1944 ( ((offset = base + charid
1945 - 1 - trie->uniquecharcount)) >= 0)
1946 && ((U32)offset < trie->lasttrans)
1947 && trie->trans[offset].check == state
1948 && (tmp=trie->trans[offset].next))
1950 DEBUG_TRIE_EXECUTE_r(
1951 PerlIO_printf( Perl_debug_log," - legal\n"));
1956 DEBUG_TRIE_EXECUTE_r(
1957 PerlIO_printf( Perl_debug_log," - fail\n"));
1959 state = aho->fail[state];
1963 /* we must be accepting here */
1964 DEBUG_TRIE_EXECUTE_r(
1965 PerlIO_printf( Perl_debug_log," - accepting\n"));
1974 if (!state) state = 1;
1977 if ( aho->states[ state ].wordnum ) {
1978 U8 *lpos = points[ (pointpos - trie->wordinfo[aho->states[ state ].wordnum].len) % maxlen ];
1979 if (!leftmost || lpos < leftmost) {
1980 DEBUG_r(accepted_word=aho->states[ state ].wordnum);
1985 s = (char*)leftmost;
1986 DEBUG_TRIE_EXECUTE_r({
1988 Perl_debug_log,"Matches word #%"UVxf" at position %"IVdf". Trying full pattern...\n",
1989 (UV)accepted_word, (IV)(s - real_start)
1992 if (!reginfo || regtry(reginfo, &s)) {
1998 DEBUG_TRIE_EXECUTE_r({
1999 PerlIO_printf( Perl_debug_log,"Pattern failed. Looking for new start point...\n");
2002 DEBUG_TRIE_EXECUTE_r(
2003 PerlIO_printf( Perl_debug_log,"No match.\n"));
2012 Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c));
2022 - regexec_flags - match a regexp against a string
2025 Perl_regexec_flags(pTHX_ REGEXP * const rx, char *stringarg, register char *strend,
2026 char *strbeg, I32 minend, SV *sv, void *data, U32 flags)
2027 /* strend: pointer to null at end of string */
2028 /* strbeg: real beginning of string */
2029 /* minend: end of match must be >=minend after stringarg. */
2030 /* data: May be used for some additional optimizations.
2031 Currently its only used, with a U32 cast, for transmitting
2032 the ganch offset when doing a /g match. This will change */
2033 /* nosave: For optimizations. */
2036 struct regexp *const prog = (struct regexp *)SvANY(rx);
2037 /*register*/ char *s;
2038 register regnode *c;
2039 /*register*/ char *startpos = stringarg;
2040 I32 minlen; /* must match at least this many chars */
2041 I32 dontbother = 0; /* how many characters not to try at end */
2042 I32 end_shift = 0; /* Same for the end. */ /* CC */
2043 I32 scream_pos = -1; /* Internal iterator of scream. */
2044 char *scream_olds = NULL;
2045 const bool utf8_target = cBOOL(DO_UTF8(sv));
2047 RXi_GET_DECL(prog,progi);
2048 regmatch_info reginfo; /* create some info to pass to regtry etc */
2049 regexp_paren_pair *swap = NULL;
2050 GET_RE_DEBUG_FLAGS_DECL;
2052 PERL_ARGS_ASSERT_REGEXEC_FLAGS;
2053 PERL_UNUSED_ARG(data);
2055 /* Be paranoid... */
2056 if (prog == NULL || startpos == NULL) {
2057 Perl_croak(aTHX_ "NULL regexp parameter");
2061 multiline = prog->extflags & RXf_PMf_MULTILINE;
2062 reginfo.prog = rx; /* Yes, sorry that this is confusing. */
2064 RX_MATCH_UTF8_set(rx, utf8_target);
2066 debug_start_match(rx, utf8_target, startpos, strend,
2070 minlen = prog->minlen;
2072 if (strend - startpos < (minlen+(prog->check_offset_min<0?prog->check_offset_min:0))) {
2073 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
2074 "String too short [regexec_flags]...\n"));
2079 /* Check validity of program. */
2080 if (UCHARAT(progi->program) != REG_MAGIC) {
2081 Perl_croak(aTHX_ "corrupted regexp program");
2085 PL_reg_state.re_state_eval_setup_done = FALSE;
2089 PL_reg_flags |= RF_utf8;
2091 /* Mark beginning of line for ^ and lookbehind. */
2092 reginfo.bol = startpos; /* XXX not used ??? */
2096 /* Mark end of line for $ (and such) */
2099 /* see how far we have to get to not match where we matched before */
2100 reginfo.till = startpos+minend;
2102 /* If there is a "must appear" string, look for it. */
2105 if (prog->extflags & RXf_GPOS_SEEN) { /* Need to set reginfo->ganch */
2107 if (flags & REXEC_IGNOREPOS){ /* Means: check only at start */
2108 reginfo.ganch = startpos + prog->gofs;
2109 DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log,
2110 "GPOS IGNOREPOS: reginfo.ganch = startpos + %"UVxf"\n",(UV)prog->gofs));
2111 } else if (sv && SvTYPE(sv) >= SVt_PVMG
2113 && (mg = mg_find(sv, PERL_MAGIC_regex_global))
2114 && mg->mg_len >= 0) {
2115 reginfo.ganch = strbeg + mg->mg_len; /* Defined pos() */
2116 DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log,
2117 "GPOS MAGIC: reginfo.ganch = strbeg + %"IVdf"\n",(IV)mg->mg_len));
2119 if (prog->extflags & RXf_ANCH_GPOS) {
2120 if (s > reginfo.ganch)
2122 s = reginfo.ganch - prog->gofs;
2123 DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log,
2124 "GPOS ANCH_GPOS: s = ganch - %"UVxf"\n",(UV)prog->gofs));
2130 reginfo.ganch = strbeg + PTR2UV(data);
2131 DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log,
2132 "GPOS DATA: reginfo.ganch= strbeg + %"UVxf"\n",PTR2UV(data)));
2134 } else { /* pos() not defined */
2135 reginfo.ganch = strbeg;
2136 DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log,
2137 "GPOS: reginfo.ganch = strbeg\n"));
2140 if (PL_curpm && (PM_GETRE(PL_curpm) == rx)) {
2141 /* We have to be careful. If the previous successful match
2142 was from this regex we don't want a subsequent partially
2143 successful match to clobber the old results.
2144 So when we detect this possibility we add a swap buffer
2145 to the re, and switch the buffer each match. If we fail
2146 we switch it back, otherwise we leave it swapped.
2149 /* do we need a save destructor here for eval dies? */
2150 Newxz(prog->offs, (prog->nparens + 1), regexp_paren_pair);
2151 DEBUG_BUFFERS_r(PerlIO_printf(Perl_debug_log,
2152 "rex=0x%"UVxf" saving offs: orig=0x%"UVxf" new=0x%"UVxf"\n",
2158 if (!(flags & REXEC_CHECKED) && (prog->check_substr != NULL || prog->check_utf8 != NULL)) {
2159 re_scream_pos_data d;
2161 d.scream_olds = &scream_olds;
2162 d.scream_pos = &scream_pos;
2163 s = re_intuit_start(rx, sv, s, strend, flags, &d);
2165 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Not present...\n"));
2166 goto phooey; /* not present */
2172 /* Simplest case: anchored match need be tried only once. */
2173 /* [unless only anchor is BOL and multiline is set] */
2174 if (prog->extflags & (RXf_ANCH & ~RXf_ANCH_GPOS)) {
2175 if (s == startpos && regtry(®info, &startpos))
2177 else if (multiline || (prog->intflags & PREGf_IMPLICIT)
2178 || (prog->extflags & RXf_ANCH_MBOL)) /* XXXX SBOL? */
2183 dontbother = minlen - 1;
2184 end = HOP3c(strend, -dontbother, strbeg) - 1;
2185 /* for multiline we only have to try after newlines */
2186 if (prog->check_substr || prog->check_utf8) {
2187 /* because of the goto we can not easily reuse the macros for bifurcating the
2188 unicode/non-unicode match modes here like we do elsewhere - demerphq */
2191 goto after_try_utf8;
2193 if (regtry(®info, &s)) {
2200 if (prog->extflags & RXf_USE_INTUIT) {
2201 s = re_intuit_start(rx, sv, s + UTF8SKIP(s), strend, flags, NULL);
2210 } /* end search for check string in unicode */
2212 if (s == startpos) {
2213 goto after_try_latin;
2216 if (regtry(®info, &s)) {
2223 if (prog->extflags & RXf_USE_INTUIT) {
2224 s = re_intuit_start(rx, sv, s + 1, strend, flags, NULL);
2233 } /* end search for check string in latin*/
2234 } /* end search for check string */
2235 else { /* search for newline */
2237 /*XXX: The s-- is almost definitely wrong here under unicode - demeprhq*/
2240 /* We can use a more efficient search as newlines are the same in unicode as they are in latin */
2241 while (s <= end) { /* note it could be possible to match at the end of the string */
2242 if (*s++ == '\n') { /* don't need PL_utf8skip here */
2243 if (regtry(®info, &s))
2247 } /* end search for newline */
2248 } /* end anchored/multiline check string search */
2250 } else if (RXf_GPOS_CHECK == (prog->extflags & RXf_GPOS_CHECK))
2252 /* the warning about reginfo.ganch being used without initialization
2253 is bogus -- we set it above, when prog->extflags & RXf_GPOS_SEEN
2254 and we only enter this block when the same bit is set. */
2255 char *tmp_s = reginfo.ganch - prog->gofs;
2257 if (tmp_s >= strbeg && regtry(®info, &tmp_s))
2262 /* Messy cases: unanchored match. */
2263 if ((prog->anchored_substr || prog->anchored_utf8) && prog->intflags & PREGf_SKIP) {
2264 /* we have /x+whatever/ */
2265 /* it must be a one character string (XXXX Except UTF_PATTERN?) */
2270 if (!(utf8_target ? prog->anchored_utf8 : prog->anchored_substr))
2271 utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog);
2272 ch = SvPVX_const(utf8_target ? prog->anchored_utf8 : prog->anchored_substr)[0];
2277 DEBUG_EXECUTE_r( did_match = 1 );
2278 if (regtry(®info, &s)) goto got_it;
2280 while (s < strend && *s == ch)
2288 DEBUG_EXECUTE_r( did_match = 1 );
2289 if (regtry(®info, &s)) goto got_it;
2291 while (s < strend && *s == ch)
2296 DEBUG_EXECUTE_r(if (!did_match)
2297 PerlIO_printf(Perl_debug_log,
2298 "Did not find anchored character...\n")
2301 else if (prog->anchored_substr != NULL
2302 || prog->anchored_utf8 != NULL
2303 || ((prog->float_substr != NULL || prog->float_utf8 != NULL)
2304 && prog->float_max_offset < strend - s)) {
2309 char *last1; /* Last position checked before */
2313 if (prog->anchored_substr || prog->anchored_utf8) {
2314 if (!(utf8_target ? prog->anchored_utf8 : prog->anchored_substr))
2315 utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog);
2316 must = utf8_target ? prog->anchored_utf8 : prog->anchored_substr;
2317 back_max = back_min = prog->anchored_offset;
2319 if (!(utf8_target ? prog->float_utf8 : prog->float_substr))
2320 utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog);
2321 must = utf8_target ? prog->float_utf8 : prog->float_substr;
2322 back_max = prog->float_max_offset;
2323 back_min = prog->float_min_offset;
2327 if (must == &PL_sv_undef)
2328 /* could not downgrade utf8 check substring, so must fail */
2334 last = HOP3c(strend, /* Cannot start after this */
2335 -(I32)(CHR_SVLEN(must)
2336 - (SvTAIL(must) != 0) + back_min), strbeg);
2339 last1 = HOPc(s, -1);
2341 last1 = s - 1; /* bogus */
2343 /* XXXX check_substr already used to find "s", can optimize if
2344 check_substr==must. */
2346 dontbother = end_shift;
2347 strend = HOPc(strend, -dontbother);
2348 while ( (s <= last) &&
2349 (s = fbm_instr((unsigned char*)HOP3(s, back_min, (back_min<0 ? strbeg : strend)),
2350 (unsigned char*)strend, must,
2351 multiline ? FBMrf_MULTILINE : 0)) ) {
2352 DEBUG_EXECUTE_r( did_match = 1 );
2353 if (HOPc(s, -back_max) > last1) {
2354 last1 = HOPc(s, -back_min);
2355 s = HOPc(s, -back_max);
2358 char * const t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1;
2360 last1 = HOPc(s, -back_min);
2364 while (s <= last1) {
2365 if (regtry(®info, &s))
2371 while (s <= last1) {
2372 if (regtry(®info, &s))
2378 DEBUG_EXECUTE_r(if (!did_match) {
2379 RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0),
2380 SvPVX_const(must), RE_SV_DUMPLEN(must), 30);
2381 PerlIO_printf(Perl_debug_log, "Did not find %s substr %s%s...\n",
2382 ((must == prog->anchored_substr || must == prog->anchored_utf8)
2383 ? "anchored" : "floating"),
2384 quoted, RE_SV_TAIL(must));
2388 else if ( (c = progi->regstclass) ) {
2390 const OPCODE op = OP(progi->regstclass);
2391 /* don't bother with what can't match */
2392 if (PL_regkind[op] != EXACT && op != CANY && PL_regkind[op] != TRIE)
2393 strend = HOPc(strend, -(minlen - 1));
2396 SV * const prop = sv_newmortal();
2397 regprop(prog, prop, c);
2399 RE_PV_QUOTED_DECL(quoted,utf8_target,PERL_DEBUG_PAD_ZERO(1),
2401 PerlIO_printf(Perl_debug_log,
2402 "Matching stclass %.*s against %s (%d bytes)\n",
2403 (int)SvCUR(prop), SvPVX_const(prop),
2404 quoted, (int)(strend - s));
2407 if (find_byclass(prog, c, s, strend, ®info))
2409 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass... [regexec_flags]\n"));
2413 if (prog->float_substr != NULL || prog->float_utf8 != NULL) {
2420 if (!(utf8_target ? prog->float_utf8 : prog->float_substr))
2421 utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog);
2422 float_real = utf8_target ? prog->float_utf8 : prog->float_substr;
2424 little = SvPV_const(float_real, len);
2425 if (SvTAIL(float_real)) {
2426 /* This means that float_real contains an artificial \n on the end
2427 * due to the presence of something like this: /foo$/
2428 * where we can match both "foo" and "foo\n" at the end of the string.
2429 * So we have to compare the end of the string first against the float_real
2430 * without the \n and then against the full float_real with the string.
2431 * We have to watch out for cases where the string might be smaller
2432 * than the float_real or the float_real without the \n.
2434 char *checkpos= strend - len;
2436 PerlIO_printf(Perl_debug_log,
2437 "%sChecking for float_real.%s\n",
2438 PL_colors[4], PL_colors[5]));
2439 if (checkpos + 1 < strbeg) {
2440 /* can't match, even if we remove the trailing \n string is too short to match */
2442 PerlIO_printf(Perl_debug_log,
2443 "%sString shorter than required trailing substring, cannot match.%s\n",
2444 PL_colors[4], PL_colors[5]));
2446 } else if (memEQ(checkpos + 1, little, len - 1)) {
2447 /* can match, the end of the string matches without the "\n" */
2448 last = checkpos + 1;
2449 } else if (checkpos < strbeg) {
2450 /* cant match, string is too short when the "\n" is included */
2452 PerlIO_printf(Perl_debug_log,
2453 "%sString does not contain required trailing substring, cannot match.%s\n",
2454 PL_colors[4], PL_colors[5]));
2456 } else if (!multiline) {
2457 /* non multiline match, so compare with the "\n" at the end of the string */
2458 if (memEQ(checkpos, little, len)) {
2462 PerlIO_printf(Perl_debug_log,
2463 "%sString does not contain required trailing substring, cannot match.%s\n",
2464 PL_colors[4], PL_colors[5]));
2468 /* multiline match, so we have to search for a place where the full string is located */
2474 last = rninstr(s, strend, little, little + len);
2476 last = strend; /* matching "$" */
2479 /* at one point this block contained a comment which was probably
2480 * incorrect, which said that this was a "should not happen" case.
2481 * Even if it was true when it was written I am pretty sure it is
2482 * not anymore, so I have removed the comment and replaced it with
2485 PerlIO_printf(Perl_debug_log,
2486 "String does not contain required substring, cannot match.\n"
2490 dontbother = strend - last + prog->float_min_offset;
2492 if (minlen && (dontbother < minlen))
2493 dontbother = minlen - 1;
2494 strend -= dontbother; /* this one's always in bytes! */
2495 /* We don't know much -- general case. */
2498 if (regtry(®info, &s))
2507 if (regtry(®info, &s))
2509 } while (s++ < strend);
2519 PerlIO_printf(Perl_debug_log,
2520 "rex=0x%"UVxf" freeing offs: 0x%"UVxf"\n",
2526 RX_MATCH_TAINTED_set(rx, PL_reg_flags & RF_tainted);
2528 if (PL_reg_state.re_state_eval_setup_done)
2529 restore_pos(aTHX_ prog);
2530 if (RXp_PAREN_NAMES(prog))
2531 (void)hv_iterinit(RXp_PAREN_NAMES(prog));
2533 /* make sure $`, $&, $', and $digit will work later */
2534 if ( !(flags & REXEC_NOT_FIRST) ) {
2535 RX_MATCH_COPY_FREE(rx);
2536 if (flags & REXEC_COPY_STR) {
2537 const I32 i = PL_regeol - startpos + (stringarg - strbeg);
2538 #ifdef PERL_OLD_COPY_ON_WRITE
2540 || (SvFLAGS(sv) & CAN_COW_MASK) == CAN_COW_FLAGS)) {
2542 PerlIO_printf(Perl_debug_log,
2543 "Copy on write: regexp capture, type %d\n",
2546 prog->saved_copy = sv_setsv_cow(prog->saved_copy, sv);
2547 prog->subbeg = (char *)SvPVX_const(prog->saved_copy);
2548 assert (SvPOKp(prog->saved_copy));
2552 RX_MATCH_COPIED_on(rx);
2553 s = savepvn(strbeg, i);
2559 prog->subbeg = strbeg;
2560 prog->sublen = PL_regeol - strbeg; /* strend may have been modified */
2567 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n",
2568 PL_colors[4], PL_colors[5]));
2569 if (PL_reg_state.re_state_eval_setup_done)
2570 restore_pos(aTHX_ prog);
2572 /* we failed :-( roll it back */
2573 DEBUG_BUFFERS_r(PerlIO_printf(Perl_debug_log,
2574 "rex=0x%"UVxf" rolling back offs: freeing=0x%"UVxf" restoring=0x%"UVxf"\n",
2579 Safefree(prog->offs);
2588 - regtry - try match at specific point
2590 STATIC I32 /* 0 failure, 1 success */
2591 S_regtry(pTHX_ regmatch_info *reginfo, char **startpos)
2595 REGEXP *const rx = reginfo->prog;
2596 regexp *const prog = (struct regexp *)SvANY(rx);
2597 RXi_GET_DECL(prog,progi);
2598 GET_RE_DEBUG_FLAGS_DECL;
2600 PERL_ARGS_ASSERT_REGTRY;
2602 reginfo->cutpoint=NULL;
2604 if ((prog->extflags & RXf_EVAL_SEEN)
2605 && !PL_reg_state.re_state_eval_setup_done)
2609 PL_reg_state.re_state_eval_setup_done = TRUE;
2611 /* Make $_ available to executed code. */
2612 if (reginfo->sv != DEFSV) {
2614 DEFSV_set(reginfo->sv);
2617 if (!(SvTYPE(reginfo->sv) >= SVt_PVMG && SvMAGIC(reginfo->sv)
2618 && (mg = mg_find(reginfo->sv, PERL_MAGIC_regex_global)))) {
2619 /* prepare for quick setting of pos */
2620 #ifdef PERL_OLD_COPY_ON_WRITE
2621 if (SvIsCOW(reginfo->sv))
2622 sv_force_normal_flags(reginfo->sv, 0);
2624 mg = sv_magicext(reginfo->sv, NULL, PERL_MAGIC_regex_global,
2625 &PL_vtbl_mglob, NULL, 0);
2629 PL_reg_oldpos = mg->mg_len;
2630 SAVEDESTRUCTOR_X(restore_pos, prog);
2632 if (!PL_reg_curpm) {
2633 Newxz(PL_reg_curpm, 1, PMOP);
2636 SV* const repointer = &PL_sv_undef;
2637 /* this regexp is also owned by the new PL_reg_curpm, which
2638 will try to free it. */
2639 av_push(PL_regex_padav, repointer);
2640 PL_reg_curpm->op_pmoffset = av_len(PL_regex_padav);
2641 PL_regex_pad = AvARRAY(PL_regex_padav);
2646 /* It seems that non-ithreads works both with and without this code.
2647 So for efficiency reasons it seems best not to have the code
2648 compiled when it is not needed. */
2649 /* This is safe against NULLs: */
2650 ReREFCNT_dec(PM_GETRE(PL_reg_curpm));
2651 /* PM_reg_curpm owns a reference to this regexp. */
2652 (void)ReREFCNT_inc(rx);
2654 PM_SETRE(PL_reg_curpm, rx);
2655 PL_reg_oldcurpm = PL_curpm;
2656 PL_curpm = PL_reg_curpm;
2657 if (RXp_MATCH_COPIED(prog)) {
2658 /* Here is a serious problem: we cannot rewrite subbeg,
2659 since it may be needed if this match fails. Thus
2660 $` inside (?{}) could fail... */
2661 PL_reg_oldsaved = prog->subbeg;
2662 PL_reg_oldsavedlen = prog->sublen;
2663 #ifdef PERL_OLD_COPY_ON_WRITE
2664 PL_nrs = prog->saved_copy;
2666 RXp_MATCH_COPIED_off(prog);
2669 PL_reg_oldsaved = NULL;
2670 prog->subbeg = PL_bostr;
2671 prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */
2674 PL_reg_starttry = *startpos;
2676 prog->offs[0].start = *startpos - PL_bostr;
2677 PL_reginput = *startpos;
2678 prog->lastparen = 0;
2679 prog->lastcloseparen = 0;
2682 /* XXXX What this code is doing here?!!! There should be no need
2683 to do this again and again, prog->lastparen should take care of
2686 /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code.
2687 * Actually, the code in regcppop() (which Ilya may be meaning by
2688 * prog->lastparen), is not needed at all by the test suite
2689 * (op/regexp, op/pat, op/split), but that code is needed otherwise
2690 * this erroneously leaves $1 defined: "1" =~ /^(?:(\d)x)?\d$/
2691 * Meanwhile, this code *is* needed for the
2692 * above-mentioned test suite tests to succeed. The common theme
2693 * on those tests seems to be returning null fields from matches.
2694 * --jhi updated by dapm */
2696 if (prog->nparens) {
2697 regexp_paren_pair *pp = prog->offs;
2699 for (i = prog->nparens; i > (I32)prog->lastparen; i--) {
2707 if (regmatch(reginfo, progi->program + 1)) {
2708 prog->offs[0].end = PL_reginput - PL_bostr;
2711 if (reginfo->cutpoint)
2712 *startpos= reginfo->cutpoint;
2713 REGCP_UNWIND(lastcp);
2718 #define sayYES goto yes
2719 #define sayNO goto no
2720 #define sayNO_SILENT goto no_silent
2722 /* we dont use STMT_START/END here because it leads to
2723 "unreachable code" warnings, which are bogus, but distracting. */
2724 #define CACHEsayNO \
2725 if (ST.cache_mask) \
2726 PL_reg_poscache[ST.cache_offset] |= ST.cache_mask; \
2729 /* this is used to determine how far from the left messages like
2730 'failed...' are printed. It should be set such that messages
2731 are inline with the regop output that created them.
2733 #define REPORT_CODE_OFF 32
2736 #define CHRTEST_UNINIT -1001 /* c1/c2 haven't been calculated yet */
2737 #define CHRTEST_VOID -1000 /* the c1/c2 "next char" test should be skipped */
2739 #define SLAB_FIRST(s) (&(s)->states[0])
2740 #define SLAB_LAST(s) (&(s)->states[PERL_REGMATCH_SLAB_SLOTS-1])
2742 /* grab a new slab and return the first slot in it */
2744 STATIC regmatch_state *
2747 #if PERL_VERSION < 9 && !defined(PERL_CORE)
2750 regmatch_slab *s = PL_regmatch_slab->next;
2752 Newx(s, 1, regmatch_slab);
2753 s->prev = PL_regmatch_slab;
2755 PL_regmatch_slab->next = s;
2757 PL_regmatch_slab = s;
2758 return SLAB_FIRST(s);
2762 /* push a new state then goto it */
2764 #define PUSH_STATE_GOTO(state, node) \
2766 st->resume_state = state; \
2769 /* push a new state with success backtracking, then goto it */
2771 #define PUSH_YES_STATE_GOTO(state, node) \
2773 st->resume_state = state; \
2774 goto push_yes_state;
2780 regmatch() - main matching routine
2782 This is basically one big switch statement in a loop. We execute an op,
2783 set 'next' to point the next op, and continue. If we come to a point which
2784 we may need to backtrack to on failure such as (A|B|C), we push a
2785 backtrack state onto the backtrack stack. On failure, we pop the top
2786 state, and re-enter the loop at the state indicated. If there are no more
2787 states to pop, we return failure.
2789 Sometimes we also need to backtrack on success; for example /A+/, where
2790 after successfully matching one A, we need to go back and try to
2791 match another one; similarly for lookahead assertions: if the assertion
2792 completes successfully, we backtrack to the state just before the assertion
2793 and then carry on. In these cases, the pushed state is marked as
2794 'backtrack on success too'. This marking is in fact done by a chain of
2795 pointers, each pointing to the previous 'yes' state. On success, we pop to
2796 the nearest yes state, discarding any intermediate failure-only states.
2797 Sometimes a yes state is pushed just to force some cleanup code to be
2798 called at the end of a successful match or submatch; e.g. (??{$re}) uses
2799 it to free the inner regex.
2801 Note that failure backtracking rewinds the cursor position, while
2802 success backtracking leaves it alone.
2804 A pattern is complete when the END op is executed, while a subpattern
2805 such as (?=foo) is complete when the SUCCESS op is executed. Both of these
2806 ops trigger the "pop to last yes state if any, otherwise return true"
2809 A common convention in this function is to use A and B to refer to the two
2810 subpatterns (or to the first nodes thereof) in patterns like /A*B/: so A is
2811 the subpattern to be matched possibly multiple times, while B is the entire
2812 rest of the pattern. Variable and state names reflect this convention.
2814 The states in the main switch are the union of ops and failure/success of
2815 substates associated with with that op. For example, IFMATCH is the op
2816 that does lookahead assertions /(?=A)B/ and so the IFMATCH state means
2817 'execute IFMATCH'; while IFMATCH_A is a state saying that we have just
2818 successfully matched A and IFMATCH_A_fail is a state saying that we have
2819 just failed to match A. Resume states always come in pairs. The backtrack
2820 state we push is marked as 'IFMATCH_A', but when that is popped, we resume
2821 at IFMATCH_A or IFMATCH_A_fail, depending on whether we are backtracking
2822 on success or failure.
2824 The struct that holds a backtracking state is actually a big union, with
2825 one variant for each major type of op. The variable st points to the
2826 top-most backtrack struct. To make the code clearer, within each
2827 block of code we #define ST to alias the relevant union.
2829 Here's a concrete example of a (vastly oversimplified) IFMATCH
2835 #define ST st->u.ifmatch
2837 case IFMATCH: // we are executing the IFMATCH op, (?=A)B
2838 ST.foo = ...; // some state we wish to save
2840 // push a yes backtrack state with a resume value of
2841 // IFMATCH_A/IFMATCH_A_fail, then continue execution at the
2843 PUSH_YES_STATE_GOTO(IFMATCH_A, A);
2846 case IFMATCH_A: // we have successfully executed A; now continue with B
2848 bar = ST.foo; // do something with the preserved value
2851 case IFMATCH_A_fail: // A failed, so the assertion failed
2852 ...; // do some housekeeping, then ...
2853 sayNO; // propagate the failure
2860 For any old-timers reading this who are familiar with the old recursive
2861 approach, the code above is equivalent to:
2863 case IFMATCH: // we are executing the IFMATCH op, (?=A)B
2872 ...; // do some housekeeping, then ...
2873 sayNO; // propagate the failure
2876 The topmost backtrack state, pointed to by st, is usually free. If you
2877 want to claim it, populate any ST.foo fields in it with values you wish to
2878 save, then do one of
2880 PUSH_STATE_GOTO(resume_state, node);
2881 PUSH_YES_STATE_GOTO(resume_state, node);
2883 which sets that backtrack state's resume value to 'resume_state', pushes a
2884 new free entry to the top of the backtrack stack, then goes to 'node'.
2885 On backtracking, the free slot is popped, and the saved state becomes the
2886 new free state. An ST.foo field in this new top state can be temporarily
2887 accessed to retrieve values, but once the main loop is re-entered, it
2888 becomes available for reuse.
2890 Note that the depth of the backtrack stack constantly increases during the
2891 left-to-right execution of the pattern, rather than going up and down with
2892 the pattern nesting. For example the stack is at its maximum at Z at the
2893 end of the pattern, rather than at X in the following:
2895 /(((X)+)+)+....(Y)+....Z/
2897 The only exceptions to this are lookahead/behind assertions and the cut,
2898 (?>A), which pop all the backtrack states associated with A before
2901 Backtrack state structs are allocated in slabs of about 4K in size.
2902 PL_regmatch_state and st always point to the currently active state,
2903 and PL_regmatch_slab points to the slab currently containing
2904 PL_regmatch_state. The first time regmatch() is called, the first slab is
2905 allocated, and is never freed until interpreter destruction. When the slab
2906 is full, a new one is allocated and chained to the end. At exit from
2907 regmatch(), slabs allocated since entry are freed.
2912 #define DEBUG_STATE_pp(pp) \
2914 DUMP_EXEC_POS(locinput, scan, utf8_target); \
2915 PerlIO_printf(Perl_debug_log, \
2916 " %*s"pp" %s%s%s%s%s\n", \
2918 PL_reg_name[st->resume_state], \
2919 ((st==yes_state||st==mark_state) ? "[" : ""), \
2920 ((st==yes_state) ? "Y" : ""), \
2921 ((st==mark_state) ? "M" : ""), \
2922 ((st==yes_state||st==mark_state) ? "]" : "") \
2927 #define REG_NODE_NUM(x) ((x) ? (int)((x)-prog) : -1)
2932 S_debug_start_match(pTHX_ const REGEXP *prog, const bool utf8_target,
2933 const char *start, const char *end, const char *blurb)
2935 const bool utf8_pat = RX_UTF8(prog) ? 1 : 0;
2937 PERL_ARGS_ASSERT_DEBUG_START_MATCH;
2942 RE_PV_QUOTED_DECL(s0, utf8_pat, PERL_DEBUG_PAD_ZERO(0),
2943 RX_PRECOMP_const(prog), RX_PRELEN(prog), 60);
2945 RE_PV_QUOTED_DECL(s1, utf8_target, PERL_DEBUG_PAD_ZERO(1),
2946 start, end - start, 60);
2948 PerlIO_printf(Perl_debug_log,
2949 "%s%s REx%s %s against %s\n",
2950 PL_colors[4], blurb, PL_colors[5], s0, s1);
2952 if (utf8_target||utf8_pat)
2953 PerlIO_printf(Perl_debug_log, "UTF-8 %s%s%s...\n",
2954 utf8_pat ? "pattern" : "",
2955 utf8_pat && utf8_target ? " and " : "",
2956 utf8_target ? "string" : ""
2962 S_dump_exec_pos(pTHX_ const char *locinput,
2963 const regnode *scan,
2964 const char *loc_regeol,
2965 const char *loc_bostr,
2966 const char *loc_reg_starttry,
2967 const bool utf8_target)
2969 const int docolor = *PL_colors[0] || *PL_colors[2] || *PL_colors[4];
2970 const int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */
2971 int l = (loc_regeol - locinput) > taill ? taill : (loc_regeol - locinput);
2972 /* The part of the string before starttry has one color
2973 (pref0_len chars), between starttry and current
2974 position another one (pref_len - pref0_len chars),
2975 after the current position the third one.
2976 We assume that pref0_len <= pref_len, otherwise we
2977 decrease pref0_len. */
2978 int pref_len = (locinput - loc_bostr) > (5 + taill) - l
2979 ? (5 + taill) - l : locinput - loc_bostr;
2982 PERL_ARGS_ASSERT_DUMP_EXEC_POS;
2984 while (utf8_target && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len)))
2986 pref0_len = pref_len - (locinput - loc_reg_starttry);
2987 if (l + pref_len < (5 + taill) && l < loc_regeol - locinput)
2988 l = ( loc_regeol - locinput > (5 + taill) - pref_len
2989 ? (5 + taill) - pref_len : loc_regeol - locinput);
2990 while (utf8_target && UTF8_IS_CONTINUATION(*(U8*)(locinput + l)))
2994 if (pref0_len > pref_len)
2995 pref0_len = pref_len;
2997 const int is_uni = (utf8_target && OP(scan) != CANY) ? 1 : 0;
2999 RE_PV_COLOR_DECL(s0,len0,is_uni,PERL_DEBUG_PAD(0),
3000 (locinput - pref_len),pref0_len, 60, 4, 5);
3002 RE_PV_COLOR_DECL(s1,len1,is_uni,PERL_DEBUG_PAD(1),
3003 (locinput - pref_len + pref0_len),
3004 pref_len - pref0_len, 60, 2, 3);
3006 RE_PV_COLOR_DECL(s2,len2,is_uni,PERL_DEBUG_PAD(2),
3007 locinput, loc_regeol - locinput, 10, 0, 1);
3009 const STRLEN tlen=len0+len1+len2;
3010 PerlIO_printf(Perl_debug_log,
3011 "%4"IVdf" <%.*s%.*s%s%.*s>%*s|",
3012 (IV)(locinput - loc_bostr),
3015 (docolor ? "" : "> <"),
3017 (int)(tlen > 19 ? 0 : 19 - tlen),
3024 /* reg_check_named_buff_matched()
3025 * Checks to see if a named buffer has matched. The data array of
3026 * buffer numbers corresponding to the buffer is expected to reside
3027 * in the regexp->data->data array in the slot stored in the ARG() of
3028 * node involved. Note that this routine doesn't actually care about the
3029 * name, that information is not preserved from compilation to execution.
3030 * Returns the index of the leftmost defined buffer with the given name
3031 * or 0 if non of the buffers matched.
3034 S_reg_check_named_buff_matched(pTHX_ const regexp *rex, const regnode *scan)
3037 RXi_GET_DECL(rex,rexi);
3038 SV *sv_dat= MUTABLE_SV(rexi->data->data[ ARG( scan ) ]);
3039 I32 *nums=(I32*)SvPVX(sv_dat);
3041 PERL_ARGS_ASSERT_REG_CHECK_NAMED_BUFF_MATCHED;
3043 for ( n=0; n<SvIVX(sv_dat); n++ ) {
3044 if ((I32)rex->lastparen >= nums[n] &&
3045 rex->offs[nums[n]].end != -1)
3054 /* free all slabs above current one - called during LEAVE_SCOPE */
3057 S_clear_backtrack_stack(pTHX_ void *p)
3059 regmatch_slab *s = PL_regmatch_slab->next;
3064 PL_regmatch_slab->next = NULL;
3066 regmatch_slab * const osl = s;
3073 #define SETREX(Re1,Re2) \
3074 if (PL_reg_state.re_state_eval_setup_done) \
3075 PM_SETRE((PL_reg_curpm), (Re2)); \
3078 STATIC I32 /* 0 failure, 1 success */
3079 S_regmatch(pTHX_ regmatch_info *reginfo, regnode *prog)
3081 #if PERL_VERSION < 9 && !defined(PERL_CORE)
3085 register const bool utf8_target = PL_reg_match_utf8;
3086 const U32 uniflags = UTF8_ALLOW_DEFAULT;
3087 REGEXP *rex_sv = reginfo->prog;
3088 regexp *rex = (struct regexp *)SvANY(rex_sv);
3089 RXi_GET_DECL(rex,rexi);
3091 /* the current state. This is a cached copy of PL_regmatch_state */
3092 register regmatch_state *st;
3093 /* cache heavy used fields of st in registers */
3094 register regnode *scan;
3095 register regnode *next;
3096 register U32 n = 0; /* general value; init to avoid compiler warning */
3097 register I32 ln = 0; /* len or last; init to avoid compiler warning */
3098 register char *locinput = PL_reginput;
3099 register I32 nextchr; /* is always set to UCHARAT(locinput) */
3101 bool result = 0; /* return value of S_regmatch */
3102 int depth = 0; /* depth of backtrack stack */
3103 U32 nochange_depth = 0; /* depth of GOSUB recursion with nochange */
3104 const U32 max_nochange_depth =
3105 (3 * rex->nparens > MAX_RECURSE_EVAL_NOCHANGE_DEPTH) ?
3106 3 * rex->nparens : MAX_RECURSE_EVAL_NOCHANGE_DEPTH;
3107 regmatch_state *yes_state = NULL; /* state to pop to on success of
3109 /* mark_state piggy backs on the yes_state logic so that when we unwind
3110 the stack on success we can update the mark_state as we go */
3111 regmatch_state *mark_state = NULL; /* last mark state we have seen */
3112 regmatch_state *cur_eval = NULL; /* most recent EVAL_AB state */
3113 struct regmatch_state *cur_curlyx = NULL; /* most recent curlyx */
3115 bool no_final = 0; /* prevent failure from backtracking? */
3116 bool do_cutgroup = 0; /* no_final only until next branch/trie entry */
3117 char *startpoint = PL_reginput;
3118 SV *popmark = NULL; /* are we looking for a mark? */
3119 SV *sv_commit = NULL; /* last mark name seen in failure */
3120 SV *sv_yes_mark = NULL; /* last mark name we have seen
3121 during a successful match */
3122 U32 lastopen = 0; /* last open we saw */
3123 bool has_cutgroup = RX_HAS_CUTGROUP(rex) ? 1 : 0;
3124 SV* const oreplsv = GvSV(PL_replgv);
3125 /* these three flags are set by various ops to signal information to
3126 * the very next op. They have a useful lifetime of exactly one loop
3127 * iteration, and are not preserved or restored by state pushes/pops
3129 bool sw = 0; /* the condition value in (?(cond)a|b) */
3130 bool minmod = 0; /* the next "{n,m}" is a "{n,m}?" */
3131 int logical = 0; /* the following EVAL is:
3135 or the following IFMATCH/UNLESSM is:
3136 false: plain (?=foo)
3137 true: used as a condition: (?(?=foo))
3139 PAD* last_pad = NULL;
3141 I32 gimme = G_SCALAR;
3142 CV *caller_cv = NULL; /* who called us */
3143 CV *last_pushed_cv = NULL; /* most recently called (?{}) CV */
3146 GET_RE_DEBUG_FLAGS_DECL;
3149 /* shut up 'may be used uninitialized' compiler warnings for dMULTICALL */
3150 multicall_oldcatch = 0;
3151 multicall_cv = NULL;
3155 PERL_ARGS_ASSERT_REGMATCH;
3157 DEBUG_OPTIMISE_r( DEBUG_EXECUTE_r({
3158 PerlIO_printf(Perl_debug_log,"regmatch start\n");
3160 /* on first ever call to regmatch, allocate first slab */
3161 if (!PL_regmatch_slab) {
3162 Newx(PL_regmatch_slab, 1, regmatch_slab);
3163 PL_regmatch_slab->prev = NULL;
3164 PL_regmatch_slab->next = NULL;
3165 PL_regmatch_state = SLAB_FIRST(PL_regmatch_slab);
3168 oldsave = PL_savestack_ix;
3169 SAVEDESTRUCTOR_X(S_clear_backtrack_stack, NULL);
3170 SAVEVPTR(PL_regmatch_slab);
3171 SAVEVPTR(PL_regmatch_state);
3173 /* grab next free state slot */
3174 st = ++PL_regmatch_state;
3175 if (st > SLAB_LAST(PL_regmatch_slab))
3176 st = PL_regmatch_state = S_push_slab(aTHX);
3178 /* Note that nextchr is a byte even in UTF */
3179 nextchr = UCHARAT(locinput);
3181 while (scan != NULL) {
3184 SV * const prop = sv_newmortal();
3185 regnode *rnext=regnext(scan);
3186 DUMP_EXEC_POS( locinput, scan, utf8_target );
3187 regprop(rex, prop, scan);
3189 PerlIO_printf(Perl_debug_log,
3190 "%3"IVdf":%*s%s(%"IVdf")\n",
3191 (IV)(scan - rexi->program), depth*2, "",
3193 (PL_regkind[OP(scan)] == END || !rnext) ?
3194 0 : (IV)(rnext - rexi->program));
3197 next = scan + NEXT_OFF(scan);
3200 state_num = OP(scan);
3204 switch (state_num) {
3206 if (locinput == PL_bostr)
3208 /* reginfo->till = reginfo->bol; */
3213 if (locinput == PL_bostr ||
3214 ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n'))
3220 if (locinput == PL_bostr)
3224 if (locinput == reginfo->ganch)
3229 /* update the startpoint */
3230 st->u.keeper.val = rex->offs[0].start;
3231 PL_reginput = locinput;
3232 rex->offs[0].start = locinput - PL_bostr;
3233 PUSH_STATE_GOTO(KEEPS_next, next);
3235 case KEEPS_next_fail:
3236 /* rollback the start point change */
3237 rex->offs[0].start = st->u.keeper.val;
3243 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
3248 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
3250 if (PL_regeol - locinput > 1)
3254 if (PL_regeol != locinput)
3258 if (!nextchr && locinput >= PL_regeol)
3261 locinput += PL_utf8skip[nextchr];
3262 if (locinput > PL_regeol)
3264 nextchr = UCHARAT(locinput);
3267 nextchr = UCHARAT(++locinput);
3270 if (!nextchr && locinput >= PL_regeol)
3272 nextchr = UCHARAT(++locinput);
3275 if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n')
3278 locinput += PL_utf8skip[nextchr];
3279 if (locinput > PL_regeol)
3281 nextchr = UCHARAT(locinput);
3284 nextchr = UCHARAT(++locinput);
3288 #define ST st->u.trie
3290 /* In this case the charclass data is available inline so
3291 we can fail fast without a lot of extra overhead.
3293 if(!ANYOF_BITMAP_TEST(scan, *locinput)) {
3295 PerlIO_printf(Perl_debug_log,
3296 "%*s %sfailed to match trie start class...%s\n",
3297 REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5])
3304 /* the basic plan of execution of the trie is:
3305 * At the beginning, run though all the states, and
3306 * find the longest-matching word. Also remember the position
3307 * of the shortest matching word. For example, this pattern:
3310 * when matched against the string "abcde", will generate
3311 * accept states for all words except 3, with the longest
3312 * matching word being 4, and the shortest being 1 (with
3313 * the position being after char 1 of the string).
3315 * Then for each matching word, in word order (i.e. 1,2,4,5),
3316 * we run the remainder of the pattern; on each try setting
3317 * the current position to the character following the word,
3318 * returning to try the next word on failure.
3320 * We avoid having to build a list of words at runtime by
3321 * using a compile-time structure, wordinfo[].prev, which
3322 * gives, for each word, the previous accepting word (if any).
3323 * In the case above it would contain the mappings 1->2, 2->0,
3324 * 3->0, 4->5, 5->1. We can use this table to generate, from
3325 * the longest word (4 above), a list of all words, by
3326 * following the list of prev pointers; this gives us the
3327 * unordered list 4,5,1,2. Then given the current word we have
3328 * just tried, we can go through the list and find the
3329 * next-biggest word to try (so if we just failed on word 2,
3330 * the next in the list is 4).
3332 * Since at runtime we don't record the matching position in
3333 * the string for each word, we have to work that out for
3334 * each word we're about to process. The wordinfo table holds
3335 * the character length of each word; given that we recorded
3336 * at the start: the position of the shortest word and its
3337 * length in chars, we just need to move the pointer the
3338 * difference between the two char lengths. Depending on
3339 * Unicode status and folding, that's cheap or expensive.
3341 * This algorithm is optimised for the case where are only a
3342 * small number of accept states, i.e. 0,1, or maybe 2.
3343 * With lots of accepts states, and having to try all of them,
3344 * it becomes quadratic on number of accept states to find all
3349 /* what type of TRIE am I? (utf8 makes this contextual) */
3350 DECL_TRIE_TYPE(scan);
3352 /* what trie are we using right now */
3353 reg_trie_data * const trie
3354 = (reg_trie_data*)rexi->data->data[ ARG( scan ) ];
3355 HV * widecharmap = MUTABLE_HV(rexi->data->data[ ARG( scan ) + 1 ]);
3356 U32 state = trie->startstate;
3358 if (trie->bitmap && !TRIE_BITMAP_TEST(trie,*locinput) ) {
3359 if (trie->states[ state ].wordnum) {
3361 PerlIO_printf(Perl_debug_log,
3362 "%*s %smatched empty string...%s\n",
3363 REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5])
3369 PerlIO_printf(Perl_debug_log,
3370 "%*s %sfailed to match trie start class...%s\n",
3371 REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5])
3378 U8 *uc = ( U8* )locinput;
3382 U8 *uscan = (U8*)NULL;
3383 U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ];
3384 U32 charcount = 0; /* how many input chars we have matched */
3385 U32 accepted = 0; /* have we seen any accepting states? */
3388 ST.jump = trie->jump;
3391 ST.longfold = FALSE; /* char longer if folded => it's harder */
3394 /* fully traverse the TRIE; note the position of the
3395 shortest accept state and the wordnum of the longest
3398 while ( state && uc <= (U8*)PL_regeol ) {
3399 U32 base = trie->states[ state ].trans.base;
3403 wordnum = trie->states[ state ].wordnum;
3405 if (wordnum) { /* it's an accept state */
3408 /* record first match position */
3410 ST.firstpos = (U8*)locinput;
3415 ST.firstchars = charcount;
3418 if (!ST.nextword || wordnum < ST.nextword)
3419 ST.nextword = wordnum;
3420 ST.topword = wordnum;
3423 DEBUG_TRIE_EXECUTE_r({
3424 DUMP_EXEC_POS( (char *)uc, scan, utf8_target );
3425 PerlIO_printf( Perl_debug_log,
3426 "%*s %sState: %4"UVxf" Accepted: %c ",
3427 2+depth * 2, "", PL_colors[4],
3428 (UV)state, (accepted ? 'Y' : 'N'));
3431 /* read a char and goto next state */
3434 REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc,
3435 uscan, len, uvc, charid, foldlen,
3442 base + charid - 1 - trie->uniquecharcount)) >= 0)
3444 && ((U32)offset < trie->lasttrans)
3445 && trie->trans[offset].check == state)
3447 state = trie->trans[offset].next;
3458 DEBUG_TRIE_EXECUTE_r(
3459 PerlIO_printf( Perl_debug_log,
3460 "Charid:%3x CP:%4"UVxf" After State: %4"UVxf"%s\n",
3461 charid, uvc, (UV)state, PL_colors[5] );
3467 /* calculate total number of accept states */
3472 w = trie->wordinfo[w].prev;
3475 ST.accepted = accepted;
3479 PerlIO_printf( Perl_debug_log,
3480 "%*s %sgot %"IVdf" possible matches%s\n",
3481 REPORT_CODE_OFF + depth * 2, "",
3482 PL_colors[4], (IV)ST.accepted, PL_colors[5] );
3484 goto trie_first_try; /* jump into the fail handler */
3488 case TRIE_next_fail: /* we failed - try next alternative */
3490 REGCP_UNWIND(ST.cp);
3491 for (n = rex->lastparen; n > ST.lastparen; n--)
3492 rex->offs[n].end = -1;
3495 if (!--ST.accepted) {
3497 PerlIO_printf( Perl_debug_log,
3498 "%*s %sTRIE failed...%s\n",
3499 REPORT_CODE_OFF+depth*2, "",
3506 /* Find next-highest word to process. Note that this code
3507 * is O(N^2) per trie run (O(N) per branch), so keep tight */
3508 register U16 min = 0;
3510 register U16 const nextword = ST.nextword;
3511 register reg_trie_wordinfo * const wordinfo
3512 = ((reg_trie_data*)rexi->data->data[ARG(ST.me)])->wordinfo;
3513 for (word=ST.topword; word; word=wordinfo[word].prev) {
3514 if (word > nextword && (!min || word < min))
3527 ST.lastparen = rex->lastparen;
3531 /* find start char of end of current word */
3533 U32 chars; /* how many chars to skip */
3534 U8 *uc = ST.firstpos;
3535 reg_trie_data * const trie
3536 = (reg_trie_data*)rexi->data->data[ARG(ST.me)];
3538 assert((trie->wordinfo[ST.nextword].len - trie->prefixlen)
3540 chars = (trie->wordinfo[ST.nextword].len - trie->prefixlen)
3544 /* the hard option - fold each char in turn and find
3545 * its folded length (which may be different */
3546 U8 foldbuf[UTF8_MAXBYTES_CASE + 1];
3554 uvc = utf8n_to_uvuni((U8*)uc, UTF8_MAXLEN, &len,
3562 uvc = to_uni_fold(uvc, foldbuf, &foldlen);
3567 uvc = utf8n_to_uvuni(uscan, UTF8_MAXLEN, &len,
3581 PL_reginput = (char *)uc;
3584 scan = (ST.jump && ST.jump[ST.nextword])
3585 ? ST.me + ST.jump[ST.nextword]
3589 PerlIO_printf( Perl_debug_log,
3590 "%*s %sTRIE matched word #%d, continuing%s\n",
3591 REPORT_CODE_OFF+depth*2, "",
3598 if (ST.accepted > 1 || has_cutgroup) {
3599 PUSH_STATE_GOTO(TRIE_next, scan);
3602 /* only one choice left - just continue */
3604 AV *const trie_words
3605 = MUTABLE_AV(rexi->data->data[ARG(ST.me)+TRIE_WORDS_OFFSET]);
3606 SV ** const tmp = av_fetch( trie_words,
3608 SV *sv= tmp ? sv_newmortal() : NULL;
3610 PerlIO_printf( Perl_debug_log,
3611 "%*s %sonly one match left, short-circuiting: #%d <%s>%s\n",
3612 REPORT_CODE_OFF+depth*2, "", PL_colors[4],
3614 tmp ? pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), 0,
3615 PL_colors[0], PL_colors[1],
3616 (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0)|PERL_PV_ESCAPE_NONASCII
3618 : "not compiled under -Dr",
3622 locinput = PL_reginput;
3623 nextchr = UCHARAT(locinput);
3624 continue; /* execute rest of RE */
3629 char *s = STRING(scan);
3631 if (utf8_target != UTF_PATTERN) {
3632 /* The target and the pattern have differing utf8ness. */
3634 const char * const e = s + ln;
3637 /* The target is utf8, the pattern is not utf8. */
3642 if (NATIVE_TO_UNI(*(U8*)s) !=
3643 utf8n_to_uvuni((U8*)l, UTF8_MAXBYTES, &ulen,
3651 /* The target is not utf8, the pattern is utf8. */
3656 if (NATIVE_TO_UNI(*((U8*)l)) !=
3657 utf8n_to_uvuni((U8*)s, UTF8_MAXBYTES, &ulen,
3665 nextchr = UCHARAT(locinput);
3668 /* The target and the pattern have the same utf8ness. */
3669 /* Inline the first character, for speed. */
3670 if (UCHARAT(s) != nextchr)
3672 if (PL_regeol - locinput < ln)
3674 if (ln > 1 && memNE(s, locinput, ln))
3677 nextchr = UCHARAT(locinput);
3682 const U8 * fold_array;
3684 U32 fold_utf8_flags;
3686 PL_reg_flags |= RF_tainted;
3687 folder = foldEQ_locale;
3688 fold_array = PL_fold_locale;
3689 fold_utf8_flags = FOLDEQ_UTF8_LOCALE;
3693 case EXACTFU_TRICKYFOLD:
3695 folder = foldEQ_latin1;
3696 fold_array = PL_fold_latin1;
3697 fold_utf8_flags = (UTF_PATTERN) ? FOLDEQ_S1_ALREADY_FOLDED : 0;
3701 folder = foldEQ_latin1;
3702 fold_array = PL_fold_latin1;
3703 fold_utf8_flags = FOLDEQ_UTF8_NOMIX_ASCII;
3708 fold_array = PL_fold;
3709 fold_utf8_flags = 0;
3715 if (utf8_target || UTF_PATTERN || state_num == EXACTFU_SS) {
3716 /* Either target or the pattern are utf8, or has the issue where
3717 * the fold lengths may differ. */
3718 const char * const l = locinput;
3719 char *e = PL_regeol;
3721 if (! foldEQ_utf8_flags(s, 0, ln, cBOOL(UTF_PATTERN),
3722 l, &e, 0, utf8_target, fold_utf8_flags))
3727 nextchr = UCHARAT(locinput);
3731 /* Neither the target nor the pattern are utf8 */
3732 if (UCHARAT(s) != nextchr &&
3733 UCHARAT(s) != fold_array[nextchr])
3737 if (PL_regeol - locinput < ln)
3739 if (ln > 1 && ! folder(s, locinput, ln))
3742 nextchr = UCHARAT(locinput);
3746 /* XXX Could improve efficiency by separating these all out using a
3747 * macro or in-line function. At that point regcomp.c would no longer
3748 * have to set the FLAGS fields of these */
3751 PL_reg_flags |= RF_tainted;
3759 /* was last char in word? */
3761 && FLAGS(scan) != REGEX_ASCII_RESTRICTED_CHARSET
3762 && FLAGS(scan) != REGEX_ASCII_MORE_RESTRICTED_CHARSET)
3764 if (locinput == PL_bostr)
3767 const U8 * const r = reghop3((U8*)locinput, -1, (U8*)PL_bostr);
3769 ln = utf8n_to_uvchr(r, UTF8SKIP(r), 0, uniflags);
3771 if (FLAGS(scan) != REGEX_LOCALE_CHARSET) {
3772 ln = isALNUM_uni(ln);
3773 LOAD_UTF8_CHARCLASS_ALNUM();
3774 n = swash_fetch(PL_utf8_alnum, (U8*)locinput, utf8_target);
3777 ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(ln));
3778 n = isALNUM_LC_utf8((U8*)locinput);
3783 /* Here the string isn't utf8, or is utf8 and only ascii
3784 * characters are to match \w. In the latter case looking at
3785 * the byte just prior to the current one may be just the final
3786 * byte of a multi-byte character. This is ok. There are two
3788 * 1) it is a single byte character, and then the test is doing
3789 * just what it's supposed to.
3790 * 2) it is a multi-byte character, in which case the final
3791 * byte is never mistakable for ASCII, and so the test
3792 * will say it is not a word character, which is the
3793 * correct answer. */
3794 ln = (locinput != PL_bostr) ?
3795 UCHARAT(locinput - 1) : '\n';
3796 switch (FLAGS(scan)) {
3797 case REGEX_UNICODE_CHARSET:
3798 ln = isWORDCHAR_L1(ln);
3799 n = isWORDCHAR_L1(nextchr);
3801 case REGEX_LOCALE_CHARSET:
3802 ln = isALNUM_LC(ln);
3803 n = isALNUM_LC(nextchr);
3805 case REGEX_DEPENDS_CHARSET:
3807 n = isALNUM(nextchr);
3809 case REGEX_ASCII_RESTRICTED_CHARSET:
3810 case REGEX_ASCII_MORE_RESTRICTED_CHARSET:
3811 ln = isWORDCHAR_A(ln);
3812 n = isWORDCHAR_A(nextchr);
3815 Perl_croak(aTHX_ "panic: Unexpected FLAGS %u in op %u", FLAGS(scan), OP(scan));
3819 /* Note requires that all BOUNDs be lower than all NBOUNDs in
3821 if (((!ln) == (!n)) == (OP(scan) < NBOUND))
3826 if (utf8_target || state_num == ANYOFV) {
3827 STRLEN inclasslen = PL_regeol - locinput;
3828 if (locinput >= PL_regeol)
3831 if (!reginclass(rex, scan, (U8*)locinput, &inclasslen, utf8_target))
3833 locinput += inclasslen;
3834 nextchr = UCHARAT(locinput);
3839 nextchr = UCHARAT(locinput);
3840 if (!nextchr && locinput >= PL_regeol)
3842 if (!REGINCLASS(rex, scan, (U8*)locinput))
3844 nextchr = UCHARAT(++locinput);
3848 /* Special char classes - The defines start on line 129 or so */
3849 CCC_TRY_U(ALNUM, NALNUM, isWORDCHAR,
3850 ALNUML, NALNUML, isALNUM_LC, isALNUM_LC_utf8,
3851 ALNUMU, NALNUMU, isWORDCHAR_L1,
3852 ALNUMA, NALNUMA, isWORDCHAR_A,
3855 CCC_TRY_U(SPACE, NSPACE, isSPACE,
3856 SPACEL, NSPACEL, isSPACE_LC, isSPACE_LC_utf8,
3857 SPACEU, NSPACEU, isSPACE_L1,
3858 SPACEA, NSPACEA, isSPACE_A,
3861 CCC_TRY(DIGIT, NDIGIT, isDIGIT,
3862 DIGITL, NDIGITL, isDIGIT_LC, isDIGIT_LC_utf8,
3863 DIGITA, NDIGITA, isDIGIT_A,
3866 case CLUMP: /* Match \X: logical Unicode character. This is defined as
3867 a Unicode extended Grapheme Cluster */
3868 /* From http://www.unicode.org/reports/tr29 (5.2 version). An
3869 extended Grapheme Cluster is:
3872 | Prepend* Begin Extend*
3875 Begin is (Hangul-syllable | ! Control)
3876 Extend is (Grapheme_Extend | Spacing_Mark)
3877 Control is [ GCB_Control CR LF ]
3879 The discussion below shows how the code for CLUMP is derived
3880 from this regex. Note that most of these concepts are from
3881 property values of the Grapheme Cluster Boundary (GCB) property.
3882 No code point can have multiple property values for a given
3883 property. Thus a code point in Prepend can't be in Control, but
3884 it must be in !Control. This is why Control above includes
3885 GCB_Control plus CR plus LF. The latter two are used in the GCB
3886 property separately, and so can't be in GCB_Control, even though
3887 they logically are controls. Control is not the same as gc=cc,
3888 but includes format and other characters as well.
3890 The Unicode definition of Hangul-syllable is:
3892 | (L* ( ( V | LV ) V* | LVT ) T*)
3895 Each of these is a value for the GCB property, and hence must be
3896 disjoint, so the order they are tested is immaterial, so the
3897 above can safely be changed to
3900 | (L* ( LVT | ( V | LV ) V*) T*)
3902 The last two terms can be combined like this:
3904 | (( LVT | ( V | LV ) V*) T*))
3906 And refactored into this:
3907 L* (L | LVT T* | V V* T* | LV V* T*)
3909 That means that if we have seen any L's at all we can quit
3910 there, but if the next character is an LVT, a V, or an LV we
3913 There is a subtlety with Prepend* which showed up in testing.
3914 Note that the Begin, and only the Begin is required in:
3915 | Prepend* Begin Extend*
3916 Also, Begin contains '! Control'. A Prepend must be a
3917 '! Control', which means it must also be a Begin. What it
3918 comes down to is that if we match Prepend* and then find no
3919 suitable Begin afterwards, that if we backtrack the last
3920 Prepend, that one will be a suitable Begin.
3923 if (locinput >= PL_regeol)
3925 if (! utf8_target) {
3927 /* Match either CR LF or '.', as all the other possibilities
3929 locinput++; /* Match the . or CR */
3930 if (nextchr == '\r' /* And if it was CR, and the next is LF,
3932 && locinput < PL_regeol
3933 && UCHARAT(locinput) == '\n') locinput++;
3937 /* Utf8: See if is ( CR LF ); already know that locinput <
3938 * PL_regeol, so locinput+1 is in bounds */
3939 if (nextchr == '\r' && UCHARAT(locinput + 1) == '\n') {
3943 /* In case have to backtrack to beginning, then match '.' */
3944 char *starting = locinput;
3946 /* In case have to backtrack the last prepend */
3947 char *previous_prepend = 0;
3949 LOAD_UTF8_CHARCLASS_GCB();
3951 /* Match (prepend)* */
3952 while (locinput < PL_regeol
3953 && swash_fetch(PL_utf8_X_prepend,
3954 (U8*)locinput, utf8_target))
3956 previous_prepend = locinput;
3957 locinput += UTF8SKIP(locinput);
3960 /* As noted above, if we matched a prepend character, but
3961 * the next thing won't match, back off the last prepend we
3962 * matched, as it is guaranteed to match the begin */
3963 if (previous_prepend
3964 && (locinput >= PL_regeol
3965 || ! swash_fetch(PL_utf8_X_begin,
3966 (U8*)locinput, utf8_target)))
3968 locinput = previous_prepend;
3971 /* Note that here we know PL_regeol > locinput, as we
3972 * tested that upon input to this switch case, and if we
3973 * moved locinput forward, we tested the result just above
3974 * and it either passed, or we backed off so that it will
3976 if (! swash_fetch(PL_utf8_X_begin, (U8*)locinput, utf8_target)) {
3978 /* Here did not match the required 'Begin' in the
3979 * second term. So just match the very first
3980 * character, the '.' of the final term of the regex */
3981 locinput = starting + UTF8SKIP(starting);
3984 /* Here is the beginning of a character that can have
3985 * an extender. It is either a hangul syllable, or a
3987 if (swash_fetch(PL_utf8_X_non_hangul,
3988 (U8*)locinput, utf8_target))
3991 /* Here not a Hangul syllable, must be a
3992 * ('! * Control') */
3993 locinput += UTF8SKIP(locinput);
3996 /* Here is a Hangul syllable. It can be composed
3997 * of several individual characters. One
3998 * possibility is T+ */
3999 if (swash_fetch(PL_utf8_X_T,
4000 (U8*)locinput, utf8_target))
4002 while (locinput < PL_regeol
4003 && swash_fetch(PL_utf8_X_T,
4004 (U8*)locinput, utf8_target))
4006 locinput += UTF8SKIP(locinput);
4010 /* Here, not T+, but is a Hangul. That means
4011 * it is one of the others: L, LV, LVT or V,
4013 * L* (L | LVT T* | V V* T* | LV V* T*) */
4016 while (locinput < PL_regeol
4017 && swash_fetch(PL_utf8_X_L,
4018 (U8*)locinput, utf8_target))
4020 locinput += UTF8SKIP(locinput);
4023 /* Here, have exhausted L*. If the next
4024 * character is not an LV, LVT nor V, it means
4025 * we had to have at least one L, so matches L+
4026 * in the original equation, we have a complete
4027 * hangul syllable. Are done. */
4029 if (locinput < PL_regeol
4030 && swash_fetch(PL_utf8_X_LV_LVT_V,
4031 (U8*)locinput, utf8_target))
4034 /* Otherwise keep going. Must be LV, LVT
4035 * or V. See if LVT */
4036 if (swash_fetch(PL_utf8_X_LVT,
4037 (U8*)locinput, utf8_target))
4039 locinput += UTF8SKIP(locinput);
4042 /* Must be V or LV. Take it, then
4044 locinput += UTF8SKIP(locinput);
4045 while (locinput < PL_regeol
4046 && swash_fetch(PL_utf8_X_V,
4047 (U8*)locinput, utf8_target))
4049 locinput += UTF8SKIP(locinput);
4053 /* And any of LV, LVT, or V can be followed
4055 while (locinput < PL_regeol
4056 && swash_fetch(PL_utf8_X_T,
4060 locinput += UTF8SKIP(locinput);
4066 /* Match any extender */
4067 while (locinput < PL_regeol
4068 && swash_fetch(PL_utf8_X_extend,
4069 (U8*)locinput, utf8_target))
4071 locinput += UTF8SKIP(locinput);
4075 if (locinput > PL_regeol) sayNO;
4077 nextchr = UCHARAT(locinput);
4081 { /* The capture buffer cases. The ones beginning with N for the
4082 named buffers just convert to the equivalent numbered and
4083 pretend they were called as the corresponding numbered buffer
4085 /* don't initialize these in the declaration, it makes C++
4090 const U8 *fold_array;
4093 PL_reg_flags |= RF_tainted;
4094 folder = foldEQ_locale;
4095 fold_array = PL_fold_locale;
4097 utf8_fold_flags = FOLDEQ_UTF8_LOCALE;
4101 folder = foldEQ_latin1;
4102 fold_array = PL_fold_latin1;
4104 utf8_fold_flags = FOLDEQ_UTF8_NOMIX_ASCII;
4108 folder = foldEQ_latin1;
4109 fold_array = PL_fold_latin1;
4111 utf8_fold_flags = 0;
4116 fold_array = PL_fold;
4118 utf8_fold_flags = 0;
4125 utf8_fold_flags = 0;
4128 /* For the named back references, find the corresponding buffer
4130 n = reg_check_named_buff_matched(rex,scan);
4135 goto do_nref_ref_common;
4138 PL_reg_flags |= RF_tainted;
4139 folder = foldEQ_locale;
4140 fold_array = PL_fold_locale;
4141 utf8_fold_flags = FOLDEQ_UTF8_LOCALE;
4145 folder = foldEQ_latin1;
4146 fold_array = PL_fold_latin1;
4147 utf8_fold_flags = FOLDEQ_UTF8_NOMIX_ASCII;
4151 folder = foldEQ_latin1;
4152 fold_array = PL_fold_latin1;
4153 utf8_fold_flags = 0;
4158 fold_array = PL_fold;
4159 utf8_fold_flags = 0;
4165 utf8_fold_flags = 0;
4169 n = ARG(scan); /* which paren pair */
4172 ln = rex->offs[n].start;
4173 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
4174 if (rex->lastparen < n || ln == -1)
4175 sayNO; /* Do not match unless seen CLOSEn. */
4176 if (ln == rex->offs[n].end)
4180 if (type != REF /* REF can do byte comparison */
4181 && (utf8_target || type == REFFU))
4182 { /* XXX handle REFFL better */
4183 char * limit = PL_regeol;
4185 /* This call case insensitively compares the entire buffer
4186 * at s, with the current input starting at locinput, but
4187 * not going off the end given by PL_regeol, and returns in
4188 * limit upon success, how much of the current input was
4190 if (! foldEQ_utf8_flags(s, NULL, rex->offs[n].end - ln, utf8_target,
4191 locinput, &limit, 0, utf8_target, utf8_fold_flags))
4196 nextchr = UCHARAT(locinput);
4200 /* Not utf8: Inline the first character, for speed. */
4201 if (UCHARAT(s) != nextchr &&
4203 UCHARAT(s) != fold_array[nextchr]))
4205 ln = rex->offs[n].end - ln;
4206 if (locinput + ln > PL_regeol)
4208 if (ln > 1 && (type == REF
4209 ? memNE(s, locinput, ln)
4210 : ! folder(s, locinput, ln)))
4213 nextchr = UCHARAT(locinput);
4223 #define ST st->u.eval
4228 regexp_internal *rei;
4229 regnode *startpoint;
4232 case GOSUB: /* /(...(?1))/ /(...(?&foo))/ */
4233 if (cur_eval && cur_eval->locinput==locinput) {
4234 if (cur_eval->u.eval.close_paren == (U32)ARG(scan))
4235 Perl_croak(aTHX_ "Infinite recursion in regex");
4236 if ( ++nochange_depth > max_nochange_depth )
4238 "Pattern subroutine nesting without pos change"
4239 " exceeded limit in regex");
4246 (void)ReREFCNT_inc(rex_sv);
4247 if (OP(scan)==GOSUB) {
4248 startpoint = scan + ARG2L(scan);
4249 ST.close_paren = ARG(scan);
4251 startpoint = rei->program+1;
4254 goto eval_recurse_doit;
4256 case EVAL: /* /(?{A})B/ /(??{A})B/ and /(?(?{A})X|Y)B/ */
4257 if (cur_eval && cur_eval->locinput==locinput) {
4258 if ( ++nochange_depth > max_nochange_depth )
4259 Perl_croak(aTHX_ "EVAL without pos change exceeded limit in regex");
4264 /* execute the code in the {...} */
4268 OP * const oop = PL_op;
4269 COP * const ocurcop = PL_curcop;
4271 char *saved_regeol = PL_regeol;
4272 struct re_save_state saved_state;
4275 /* To not corrupt the existing regex state while executing the
4276 * eval we would normally put it on the save stack, like with
4277 * save_re_context. However, re-evals have a weird scoping so we
4278 * can't just add ENTER/LEAVE here. With that, things like
4280 * (?{$a=2})(a(?{local$a=$a+1}))*aak*c(?{$b=$a})
4282 * would break, as they expect the localisation to be unwound
4283 * only when the re-engine backtracks through the bit that
4286 * What we do instead is just saving the state in a local c
4289 Copy(&PL_reg_state, &saved_state, 1, struct re_save_state);
4291 PL_reg_state.re_reparsing = FALSE;
4294 caller_cv = find_runcv(NULL);
4298 if (rexi->data->what[n] == 'r') { /* code from an external qr */
4299 newcv = ((struct regexp *)SvANY(
4300 (REGEXP*)(rexi->data->data[n])
4303 nop = (OP*)rexi->data->data[n+1];
4305 else if (rexi->data->what[n] == 'l') { /* literal code */
4307 nop = (OP*)rexi->data->data[n];
4308 assert(CvDEPTH(newcv));
4311 /* literal with own CV */
4312 assert(rexi->data->what[n] == 'L');
4313 newcv = rex->qr_anoncv;
4314 nop = (OP*)rexi->data->data[n];
4317 /* the initial nextstate you would normally execute
4318 * at the start of an eval (which would cause error
4319 * messages to come from the eval), may be optimised
4320 * away from the execution path in the regex code blocks;
4321 * so manually set PL_curcop to it initially */
4323 OP *o = cUNOPx(nop)->op_first;
4324 assert(o->op_type == OP_NULL);
4325 if (o->op_targ == OP_SCOPE) {
4326 o = cUNOPo->op_first;
4329 assert(o->op_targ == OP_LEAVE);
4330 o = cUNOPo->op_first;
4331 assert(o->op_type == OP_ENTER);
4335 if (o->op_type != OP_STUB) {
4336 assert( o->op_type == OP_NEXTSTATE
4337 || o->op_type == OP_DBSTATE
4338 || (o->op_type == OP_NULL
4339 && ( o->op_targ == OP_NEXTSTATE
4340 || o->op_targ == OP_DBSTATE
4344 PL_curcop = (COP*)o;
4349 DEBUG_STATE_r( PerlIO_printf(Perl_debug_log,
4350 " re EVAL PL_op=0x%"UVxf"\n", PTR2UV(nop)) );
4352 /* normally if we're about to execute code from the same
4353 * CV that we used previously, we just use the existing
4354 * CX stack entry. However, its possible that in the
4355 * meantime we may have backtracked, popped from the save
4356 * stack, and undone the SAVECOMPPAD(s) associated with
4357 * PUSH_MULTICALL; in which case PL_comppad no longer
4358 * points to newcv's pad. */
4359 if (newcv != last_pushed_cv || PL_comppad != last_pad)
4361 I32 depth = (newcv == caller_cv) ? 0 : 1;
4362 if (last_pushed_cv) {
4363 CHANGE_MULTICALL_WITHDEPTH(newcv, depth);
4366 PUSH_MULTICALL_WITHDEPTH(newcv, depth);
4368 last_pushed_cv = newcv;
4370 last_pad = PL_comppad;
4372 rex->offs[0].end = PL_reg_magic->mg_len = locinput - PL_bostr;
4375 SV *sv_mrk = get_sv("REGMARK", 1);
4376 sv_setsv(sv_mrk, sv_yes_mark);
4379 /* we don't use MULTICALL here as we want to call the
4380 * first op of the block of interest, rather than the
4381 * first op of the sub */
4384 CALLRUNOPS(aTHX); /* Scalar context. */
4387 ret = &PL_sv_undef; /* protect against empty (?{}) blocks. */
4393 Copy(&saved_state, &PL_reg_state, 1, struct re_save_state);
4395 /* *** Note that at this point we don't restore
4396 * PL_comppad, (or pop the CxSUB) on the assumption it may
4397 * be used again soon. This is safe as long as nothing
4398 * in the regexp code uses the pad ! */
4400 PL_curcop = ocurcop;
4401 PL_regeol = saved_regeol;
4404 sv_setsv(save_scalar(PL_replgv), ret);
4408 if (logical == 2) { /* Postponed subexpression: /(??{...})/ */
4411 /* extract RE object from returned value; compiling if
4417 SV *const sv = SvRV(ret);
4419 if (SvTYPE(sv) == SVt_REGEXP) {
4421 } else if (SvSMAGICAL(sv)) {
4422 mg = mg_find(sv, PERL_MAGIC_qr);
4425 } else if (SvTYPE(ret) == SVt_REGEXP) {
4427 } else if (SvSMAGICAL(ret)) {
4428 if (SvGMAGICAL(ret)) {
4429 /* I don't believe that there is ever qr magic
4431 assert(!mg_find(ret, PERL_MAGIC_qr));
4432 sv_unmagic(ret, PERL_MAGIC_qr);
4435 mg = mg_find(ret, PERL_MAGIC_qr);
4436 /* testing suggests mg only ends up non-NULL for
4437 scalars who were upgraded and compiled in the
4438 else block below. In turn, this is only
4439 triggered in the "postponed utf8 string" tests
4445 rx = (REGEXP *) mg->mg_obj; /*XXX:dmq*/
4449 rx = reg_temp_copy(NULL, rx);
4453 const I32 osize = PL_regsize;
4456 assert (SvUTF8(ret));
4457 } else if (SvUTF8(ret)) {
4458 /* Not doing UTF-8, despite what the SV says. Is
4459 this only if we're trapped in use 'bytes'? */
4460 /* Make a copy of the octet sequence, but without
4461 the flag on, as the compiler now honours the
4462 SvUTF8 flag on ret. */
4464 const char *const p = SvPV(ret, len);
4465 ret = newSVpvn_flags(p, len, SVs_TEMP);
4467 rx = CALLREGCOMP(ret, pm_flags);
4469 & (SVs_TEMP | SVs_PADTMP | SVf_READONLY
4471 /* This isn't a first class regexp. Instead, it's
4472 caching a regexp onto an existing, Perl visible
4474 sv_magic(ret, MUTABLE_SV(rx), PERL_MAGIC_qr, 0, 0);
4479 re = (struct regexp *)SvANY(rx);
4481 RXp_MATCH_COPIED_off(re);
4482 re->subbeg = rex->subbeg;
4483 re->sublen = rex->sublen;
4486 debug_start_match(re_sv, utf8_target, locinput, PL_regeol,
4487 "Matching embedded");
4489 startpoint = rei->program + 1;
4490 ST.close_paren = 0; /* only used for GOSUB */
4492 eval_recurse_doit: /* Share code with GOSUB below this line */
4493 /* run the pattern returned from (??{...}) */
4494 ST.cp = regcppush(rex, 0); /* Save *all* the positions. */
4495 REGCP_SET(ST.lastcp);
4497 /* see regtry, specifically PL_reglast(?:close)?paren is a pointer! (i dont know why) :dmq */
4499 re->lastcloseparen = 0;
4501 PL_reginput = locinput;
4504 /* XXXX This is too dramatic a measure... */
4507 ST.toggle_reg_flags = PL_reg_flags;
4509 PL_reg_flags |= RF_utf8;
4511 PL_reg_flags &= ~RF_utf8;
4512 ST.toggle_reg_flags ^= PL_reg_flags; /* diff of old and new */
4514 ST.prev_rex = rex_sv;
4515 ST.prev_curlyx = cur_curlyx;
4516 SETREX(rex_sv,re_sv);
4521 ST.prev_eval = cur_eval;
4523 /* now continue from first node in postoned RE */
4524 PUSH_YES_STATE_GOTO(EVAL_AB, startpoint);
4527 /* logical is 1, /(?(?{...})X|Y)/ */
4528 sw = cBOOL(SvTRUE(ret));
4533 case EVAL_AB: /* cleanup after a successful (??{A})B */
4534 /* note: this is called twice; first after popping B, then A */
4535 PL_reg_flags ^= ST.toggle_reg_flags;
4536 ReREFCNT_dec(rex_sv);
4537 SETREX(rex_sv,ST.prev_rex);
4538 rex = (struct regexp *)SvANY(rex_sv);
4539 rexi = RXi_GET(rex);
4541 cur_eval = ST.prev_eval;
4542 cur_curlyx = ST.prev_curlyx;
4544 /* XXXX This is too dramatic a measure... */
4546 if ( nochange_depth )
4551 case EVAL_AB_fail: /* unsuccessfully ran A or B in (??{A})B */
4552 /* note: this is called twice; first after popping B, then A */
4553 PL_reg_flags ^= ST.toggle_reg_flags;
4554 ReREFCNT_dec(rex_sv);
4555 SETREX(rex_sv,ST.prev_rex);
4556 rex = (struct regexp *)SvANY(rex_sv);
4557 rexi = RXi_GET(rex);
4559 PL_reginput = locinput;
4560 REGCP_UNWIND(ST.lastcp);
4562 cur_eval = ST.prev_eval;
4563 cur_curlyx = ST.prev_curlyx;
4564 /* XXXX This is too dramatic a measure... */
4566 if ( nochange_depth )
4572 n = ARG(scan); /* which paren pair */
4573 rex->offs[n].start_tmp = locinput - PL_bostr;
4576 DEBUG_BUFFERS_r(PerlIO_printf(Perl_debug_log,
4577 "rex=0x%"UVxf" offs=0x%"UVxf": \\%"UVuf": set %"IVdf" tmp; regsize=%"UVuf"\n",
4581 (IV)rex->offs[n].start_tmp,
4587 /* XXX really need to log other places start/end are set too */
4588 #define CLOSE_CAPTURE \
4589 rex->offs[n].start = rex->offs[n].start_tmp; \
4590 rex->offs[n].end = locinput - PL_bostr; \
4591 DEBUG_BUFFERS_r(PerlIO_printf(Perl_debug_log, \
4592 "rex=0x%"UVxf" offs=0x%"UVxf": \\%"UVuf": set %"IVdf"..%"IVdf"\n", \
4594 PTR2UV(rex->offs), \
4596 (IV)rex->offs[n].start, \
4597 (IV)rex->offs[n].end \
4601 n = ARG(scan); /* which paren pair */
4603 /*if (n > PL_regsize)
4605 if (n > rex->lastparen)
4607 rex->lastcloseparen = n;
4608 if (cur_eval && cur_eval->u.eval.close_paren == n) {
4616 cursor && OP(cursor)!=END;
4617 cursor=regnext(cursor))
4619 if ( OP(cursor)==CLOSE ){
4621 if ( n <= lastopen ) {
4623 /*if (n > PL_regsize)
4625 if (n > rex->lastparen)
4627 rex->lastcloseparen = n;
4628 if ( n == ARG(scan) || (cur_eval &&
4629 cur_eval->u.eval.close_paren == n))
4638 n = ARG(scan); /* which paren pair */
4639 sw = cBOOL(rex->lastparen >= n && rex->offs[n].end != -1);
4642 /* reg_check_named_buff_matched returns 0 for no match */
4643 sw = cBOOL(0 < reg_check_named_buff_matched(rex,scan));
4647 sw = (cur_eval && (!n || cur_eval->u.eval.close_paren == n));
4653 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
4655 next = NEXTOPER(NEXTOPER(scan));
4657 next = scan + ARG(scan);
4658 if (OP(next) == IFTHEN) /* Fake one. */
4659 next = NEXTOPER(NEXTOPER(next));
4663 logical = scan->flags;
4666 /*******************************************************************
4668 The CURLYX/WHILEM pair of ops handle the most generic case of the /A*B/
4669 pattern, where A and B are subpatterns. (For simple A, CURLYM or
4670 STAR/PLUS/CURLY/CURLYN are used instead.)
4672 A*B is compiled as <CURLYX><A><WHILEM><B>
4674 On entry to the subpattern, CURLYX is called. This pushes a CURLYX
4675 state, which contains the current count, initialised to -1. It also sets
4676 cur_curlyx to point to this state, with any previous value saved in the
4679 CURLYX then jumps straight to the WHILEM op, rather than executing A,
4680 since the pattern may possibly match zero times (i.e. it's a while {} loop
4681 rather than a do {} while loop).
4683 Each entry to WHILEM represents a successful match of A. The count in the
4684 CURLYX block is incremented, another WHILEM state is pushed, and execution
4685 passes to A or B depending on greediness and the current count.
4687 For example, if matching against the string a1a2a3b (where the aN are
4688 substrings that match /A/), then the match progresses as follows: (the
4689 pushed states are interspersed with the bits of strings matched so far):
4692 <CURLYX cnt=0><WHILEM>
4693 <CURLYX cnt=1><WHILEM> a1 <WHILEM>
4694 <CURLYX cnt=2><WHILEM> a1 <WHILEM> a2 <WHILEM>
4695 <CURLYX cnt=3><WHILEM> a1 <WHILEM> a2 <WHILEM> a3 <WHILEM>
4696 <CURLYX cnt=3><WHILEM> a1 <WHILEM> a2 <WHILEM> a3 <WHILEM> b
4698 (Contrast this with something like CURLYM, which maintains only a single
4702 a1 <CURLYM cnt=1> a2
4703 a1 a2 <CURLYM cnt=2> a3
4704 a1 a2 a3 <CURLYM cnt=3> b
4707 Each WHILEM state block marks a point to backtrack to upon partial failure
4708 of A or B, and also contains some minor state data related to that
4709 iteration. The CURLYX block, pointed to by cur_curlyx, contains the
4710 overall state, such as the count, and pointers to the A and B ops.
4712 This is complicated slightly by nested CURLYX/WHILEM's. Since cur_curlyx
4713 must always point to the *current* CURLYX block, the rules are:
4715 When executing CURLYX, save the old cur_curlyx in the CURLYX state block,
4716 and set cur_curlyx to point the new block.
4718 When popping the CURLYX block after a successful or unsuccessful match,
4719 restore the previous cur_curlyx.
4721 When WHILEM is about to execute B, save the current cur_curlyx, and set it
4722 to the outer one saved in the CURLYX block.
4724 When popping the WHILEM block after a successful or unsuccessful B match,
4725 restore the previous cur_curlyx.
4727 Here's an example for the pattern (AI* BI)*BO
4728 I and O refer to inner and outer, C and W refer to CURLYX and WHILEM:
4731 curlyx backtrack stack
4732 ------ ---------------
4734 CO <CO prev=NULL> <WO>
4735 CI <CO prev=NULL> <WO> <CI prev=CO> <WI> ai
4736 CO <CO prev=NULL> <WO> <CI prev=CO> <WI> ai <WI prev=CI> bi
4737 NULL <CO prev=NULL> <WO> <CI prev=CO> <WI> ai <WI prev=CI> bi <WO prev=CO> bo
4739 At this point the pattern succeeds, and we work back down the stack to
4740 clean up, restoring as we go:
4742 CO <CO prev=NULL> <WO> <CI prev=CO> <WI> ai <WI prev=CI> bi
4743 CI <CO prev=NULL> <WO> <CI prev=CO> <WI> ai
4744 CO <CO prev=NULL> <WO>
4747 *******************************************************************/
4749 #define ST st->u.curlyx
4751 case CURLYX: /* start of /A*B/ (for complex A) */
4753 /* No need to save/restore up to this paren */
4754 I32 parenfloor = scan->flags;
4756 assert(next); /* keep Coverity happy */
4757 if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */
4760 /* XXXX Probably it is better to teach regpush to support
4761 parenfloor > PL_regsize... */
4762 if (parenfloor > (I32)rex->lastparen)
4763 parenfloor = rex->lastparen; /* Pessimization... */
4765 ST.prev_curlyx= cur_curlyx;
4767 ST.cp = PL_savestack_ix;
4769 /* these fields contain the state of the current curly.
4770 * they are accessed by subsequent WHILEMs */
4771 ST.parenfloor = parenfloor;
4776 ST.count = -1; /* this will be updated by WHILEM */
4777 ST.lastloc = NULL; /* this will be updated by WHILEM */
4779 PL_reginput = locinput;
4780 PUSH_YES_STATE_GOTO(CURLYX_end, PREVOPER(next));
4784 case CURLYX_end: /* just finished matching all of A*B */
4785 cur_curlyx = ST.prev_curlyx;
4789 case CURLYX_end_fail: /* just failed to match all of A*B */
4791 cur_curlyx = ST.prev_curlyx;
4797 #define ST st->u.whilem
4799 case WHILEM: /* just matched an A in /A*B/ (for complex A) */
4801 /* see the discussion above about CURLYX/WHILEM */
4803 int min = ARG1(cur_curlyx->u.curlyx.me);
4804 int max = ARG2(cur_curlyx->u.curlyx.me);
4805 regnode *A = NEXTOPER(cur_curlyx->u.curlyx.me) + EXTRA_STEP_2ARGS;
4807 assert(cur_curlyx); /* keep Coverity happy */
4808 n = ++cur_curlyx->u.curlyx.count; /* how many A's matched */
4809 ST.save_lastloc = cur_curlyx->u.curlyx.lastloc;
4810 ST.cache_offset = 0;
4813 PL_reginput = locinput;
4815 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
4816 "%*s whilem: matched %ld out of %d..%d\n",
4817 REPORT_CODE_OFF+depth*2, "", (long)n, min, max)
4820 /* First just match a string of min A's. */
4823 ST.cp = regcppush(rex, cur_curlyx->u.curlyx.parenfloor);
4824 cur_curlyx->u.curlyx.lastloc = locinput;
4825 REGCP_SET(ST.lastcp);
4827 PUSH_STATE_GOTO(WHILEM_A_pre, A);
4831 /* If degenerate A matches "", assume A done. */
4833 if (locinput == cur_curlyx->u.curlyx.lastloc) {
4834 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
4835 "%*s whilem: empty match detected, trying continuation...\n",
4836 REPORT_CODE_OFF+depth*2, "")
4838 goto do_whilem_B_max;
4841 /* super-linear cache processing */
4845 if (!PL_reg_maxiter) {
4846 /* start the countdown: Postpone detection until we
4847 * know the match is not *that* much linear. */
4848 PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4);
4849 /* possible overflow for long strings and many CURLYX's */
4850 if (PL_reg_maxiter < 0)
4851 PL_reg_maxiter = I32_MAX;
4852 PL_reg_leftiter = PL_reg_maxiter;
4855 if (PL_reg_leftiter-- == 0) {
4856 /* initialise cache */
4857 const I32 size = (PL_reg_maxiter + 7)/8;
4858 if (PL_reg_poscache) {
4859 if ((I32)PL_reg_poscache_size < size) {
4860 Renew(PL_reg_poscache, size, char);
4861 PL_reg_poscache_size = size;
4863 Zero(PL_reg_poscache, size, char);
4866 PL_reg_poscache_size = size;
4867 Newxz(PL_reg_poscache, size, char);
4869 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
4870 "%swhilem: Detected a super-linear match, switching on caching%s...\n",
4871 PL_colors[4], PL_colors[5])
4875 if (PL_reg_leftiter < 0) {
4876 /* have we already failed at this position? */
4878 offset = (scan->flags & 0xf) - 1
4879 + (locinput - PL_bostr) * (scan->flags>>4);
4880 mask = 1 << (offset % 8);
4882 if (PL_reg_poscache[offset] & mask) {
4883 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
4884 "%*s whilem: (cache) already tried at this position...\n",
4885 REPORT_CODE_OFF+depth*2, "")
4887 sayNO; /* cache records failure */
4889 ST.cache_offset = offset;
4890 ST.cache_mask = mask;
4894 /* Prefer B over A for minimal matching. */
4896 if (cur_curlyx->u.curlyx.minmod) {
4897 ST.save_curlyx = cur_curlyx;
4898 cur_curlyx = cur_curlyx->u.curlyx.prev_curlyx;
4899 ST.cp = regcppush(rex, ST.save_curlyx->u.curlyx.parenfloor);
4900 REGCP_SET(ST.lastcp);
4901 PUSH_YES_STATE_GOTO(WHILEM_B_min, ST.save_curlyx->u.curlyx.B);
4905 /* Prefer A over B for maximal matching. */
4907 if (n < max) { /* More greed allowed? */
4908 ST.cp = regcppush(rex, cur_curlyx->u.curlyx.parenfloor);
4909 cur_curlyx->u.curlyx.lastloc = locinput;
4910 REGCP_SET(ST.lastcp);
4911 PUSH_STATE_GOTO(WHILEM_A_max, A);
4914 goto do_whilem_B_max;
4918 case WHILEM_B_min: /* just matched B in a minimal match */
4919 case WHILEM_B_max: /* just matched B in a maximal match */
4920 cur_curlyx = ST.save_curlyx;
4924 case WHILEM_B_max_fail: /* just failed to match B in a maximal match */
4925 cur_curlyx = ST.save_curlyx;
4926 cur_curlyx->u.curlyx.lastloc = ST.save_lastloc;
4927 cur_curlyx->u.curlyx.count--;
4931 case WHILEM_A_min_fail: /* just failed to match A in a minimal match */
4933 case WHILEM_A_pre_fail: /* just failed to match even minimal A */
4934 REGCP_UNWIND(ST.lastcp);
4936 cur_curlyx->u.curlyx.lastloc = ST.save_lastloc;
4937 cur_curlyx->u.curlyx.count--;
4941 case WHILEM_A_max_fail: /* just failed to match A in a maximal match */
4942 REGCP_UNWIND(ST.lastcp);
4943 regcppop(rex); /* Restore some previous $<digit>s? */
4944 PL_reginput = locinput;
4945 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
4946 "%*s whilem: failed, trying continuation...\n",
4947 REPORT_CODE_OFF+depth*2, "")
4950 if (cur_curlyx->u.curlyx.count >= REG_INFTY
4951 && ckWARN(WARN_REGEXP)
4952 && !(PL_reg_flags & RF_warned))
4954 PL_reg_flags |= RF_warned;
4955 Perl_warner(aTHX_ packWARN(WARN_REGEXP),
4956 "Complex regular subexpression recursion limit (%d) "
4962 ST.save_curlyx = cur_curlyx;
4963 cur_curlyx = cur_curlyx->u.curlyx.prev_curlyx;
4964 PUSH_YES_STATE_GOTO(WHILEM_B_max, ST.save_curlyx->u.curlyx.B);
4967 case WHILEM_B_min_fail: /* just failed to match B in a minimal match */
4968 cur_curlyx = ST.save_curlyx;
4969 REGCP_UNWIND(ST.lastcp);
4972 if (cur_curlyx->u.curlyx.count >= /*max*/ARG2(cur_curlyx->u.curlyx.me)) {
4973 /* Maximum greed exceeded */
4974 if (cur_curlyx->u.curlyx.count >= REG_INFTY
4975 && ckWARN(WARN_REGEXP)
4976 && !(PL_reg_flags & RF_warned))
4978 PL_reg_flags |= RF_warned;
4979 Perl_warner(aTHX_ packWARN(WARN_REGEXP),
4980 "Complex regular subexpression recursion "
4981 "limit (%d) exceeded",
4984 cur_curlyx->u.curlyx.count--;
4988 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
4989 "%*s trying longer...\n", REPORT_CODE_OFF+depth*2, "")
4991 /* Try grabbing another A and see if it helps. */
4992 PL_reginput = locinput;
4993 cur_curlyx->u.curlyx.lastloc = locinput;
4994 ST.cp = regcppush(rex, cur_curlyx->u.curlyx.parenfloor);
4995 REGCP_SET(ST.lastcp);
4996 PUSH_STATE_GOTO(WHILEM_A_min,
4997 /*A*/ NEXTOPER(ST.save_curlyx->u.curlyx.me) + EXTRA_STEP_2ARGS);
5001 #define ST st->u.branch
5003 case BRANCHJ: /* /(...|A|...)/ with long next pointer */
5004 next = scan + ARG(scan);
5007 scan = NEXTOPER(scan);
5010 case BRANCH: /* /(...|A|...)/ */
5011 scan = NEXTOPER(scan); /* scan now points to inner node */
5012 ST.lastparen = rex->lastparen;
5013 ST.next_branch = next;
5015 PL_reginput = locinput;
5017 /* Now go into the branch */
5019 PUSH_YES_STATE_GOTO(BRANCH_next, scan);
5021 PUSH_STATE_GOTO(BRANCH_next, scan);
5025 PL_reginput = locinput;
5026 sv_yes_mark = st->u.mark.mark_name = scan->flags ? NULL :
5027 MUTABLE_SV(rexi->data->data[ ARG( scan ) ]);
5028 PUSH_STATE_GOTO(CUTGROUP_next,next);
5030 case CUTGROUP_next_fail:
5033 if (st->u.mark.mark_name)
5034 sv_commit = st->u.mark.mark_name;
5040 case BRANCH_next_fail: /* that branch failed; try the next, if any */
5045 REGCP_UNWIND(ST.cp);
5046 for (n = rex->lastparen; n > ST.lastparen; n--)
5047 rex->offs[n].end = -1;
5049 /*dmq: rex->lastcloseparen = n; */
5050 scan = ST.next_branch;
5051 /* no more branches? */
5052 if (!scan || (OP(scan) != BRANCH && OP(scan) != BRANCHJ)) {
5054 PerlIO_printf( Perl_debug_log,
5055 "%*s %sBRANCH failed...%s\n",
5056 REPORT_CODE_OFF+depth*2, "",
5062 continue; /* execute next BRANCH[J] op */
5070 #define ST st->u.curlym
5072 case CURLYM: /* /A{m,n}B/ where A is fixed-length */
5074 /* This is an optimisation of CURLYX that enables us to push
5075 * only a single backtracking state, no matter how many matches
5076 * there are in {m,n}. It relies on the pattern being constant
5077 * length, with no parens to influence future backrefs
5081 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
5083 /* if paren positive, emulate an OPEN/CLOSE around A */
5085 U32 paren = ST.me->flags;
5086 if (paren > PL_regsize)
5088 if (paren > rex->lastparen)
5089 rex->lastparen = paren;
5090 scan += NEXT_OFF(scan); /* Skip former OPEN. */
5098 ST.c1 = CHRTEST_UNINIT;
5101 if (!(ST.minmod ? ARG1(ST.me) : ARG2(ST.me))) /* min/max */
5104 curlym_do_A: /* execute the A in /A{m,n}B/ */
5105 PL_reginput = locinput;
5106 PUSH_YES_STATE_GOTO(CURLYM_A, ST.A); /* match A */
5109 case CURLYM_A: /* we've just matched an A */
5110 locinput = st->locinput;
5111 nextchr = UCHARAT(locinput);
5114 /* after first match, determine A's length: u.curlym.alen */
5115 if (ST.count == 1) {
5116 if (PL_reg_match_utf8) {
5118 while (s < PL_reginput) {
5124 ST.alen = PL_reginput - locinput;
5127 ST.count = ST.minmod ? ARG1(ST.me) : ARG2(ST.me);
5130 PerlIO_printf(Perl_debug_log,
5131 "%*s CURLYM now matched %"IVdf" times, len=%"IVdf"...\n",
5132 (int)(REPORT_CODE_OFF+(depth*2)), "",
5133 (IV) ST.count, (IV)ST.alen)
5136 locinput = PL_reginput;
5138 if (cur_eval && cur_eval->u.eval.close_paren &&
5139 cur_eval->u.eval.close_paren == (U32)ST.me->flags)
5143 I32 max = (ST.minmod ? ARG1(ST.me) : ARG2(ST.me));
5144 if ( max == REG_INFTY || ST.count < max )
5145 goto curlym_do_A; /* try to match another A */
5147 goto curlym_do_B; /* try to match B */
5149 case CURLYM_A_fail: /* just failed to match an A */
5150 REGCP_UNWIND(ST.cp);
5152 if (ST.minmod || ST.count < ARG1(ST.me) /* min*/
5153 || (cur_eval && cur_eval->u.eval.close_paren &&
5154 cur_eval->u.eval.close_paren == (U32)ST.me->flags))
5157 curlym_do_B: /* execute the B in /A{m,n}B/ */
5158 PL_reginput = locinput;
5159 if (ST.c1 == CHRTEST_UNINIT) {
5160 /* calculate c1 and c2 for possible match of 1st char
5161 * following curly */
5162 ST.c1 = ST.c2 = CHRTEST_VOID;
5163 if (HAS_TEXT(ST.B) || JUMPABLE(ST.B)) {
5164 regnode *text_node = ST.B;
5165 if (! HAS_TEXT(text_node))
5166 FIND_NEXT_IMPT(text_node);
5169 (HAS_TEXT(text_node) && PL_regkind[OP(text_node)] == EXACT)
5171 But the former is redundant in light of the latter.
5173 if this changes back then the macro for
5174 IS_TEXT and friends need to change.
5176 if (PL_regkind[OP(text_node)] == EXACT)
5179 ST.c1 = (U8)*STRING(text_node);
5180 switch (OP(text_node)) {
5181 case EXACTF: ST.c2 = PL_fold[ST.c1]; break;
5184 case EXACTFU_TRICKYFOLD:
5185 case EXACTFU: ST.c2 = PL_fold_latin1[ST.c1]; break;
5186 case EXACTFL: ST.c2 = PL_fold_locale[ST.c1]; break;
5187 default: ST.c2 = ST.c1;
5194 PerlIO_printf(Perl_debug_log,
5195 "%*s CURLYM trying tail with matches=%"IVdf"...\n",
5196 (int)(REPORT_CODE_OFF+(depth*2)),
5199 if (ST.c1 != CHRTEST_VOID
5200 && UCHARAT(PL_reginput) != ST.c1
5201 && UCHARAT(PL_reginput) != ST.c2)
5203 /* simulate B failing */
5205 PerlIO_printf(Perl_debug_log,
5206 "%*s CURLYM Fast bail c1=%"IVdf" c2=%"IVdf"\n",
5207 (int)(REPORT_CODE_OFF+(depth*2)),"",
5210 state_num = CURLYM_B_fail;
5211 goto reenter_switch;
5215 /* mark current A as captured */
5216 I32 paren = ST.me->flags;
5218 rex->offs[paren].start
5219 = HOPc(PL_reginput, -ST.alen) - PL_bostr;
5220 rex->offs[paren].end = PL_reginput - PL_bostr;
5221 /*dmq: rex->lastcloseparen = paren; */
5224 rex->offs[paren].end = -1;
5225 if (cur_eval && cur_eval->u.eval.close_paren &&
5226 cur_eval->u.eval.close_paren == (U32)ST.me->flags)
5235 PUSH_STATE_GOTO(CURLYM_B, ST.B); /* match B */
5238 case CURLYM_B_fail: /* just failed to match a B */
5239 REGCP_UNWIND(ST.cp);
5241 I32 max = ARG2(ST.me);
5242 if (max != REG_INFTY && ST.count == max)
5244 goto curlym_do_A; /* try to match a further A */
5246 /* backtrack one A */
5247 if (ST.count == ARG1(ST.me) /* min */)
5250 locinput = HOPc(locinput, -ST.alen);
5251 goto curlym_do_B; /* try to match B */
5254 #define ST st->u.curly
5256 #define CURLY_SETPAREN(paren, success) \
5259 rex->offs[paren].start = HOPc(locinput, -1) - PL_bostr; \
5260 rex->offs[paren].end = locinput - PL_bostr; \
5261 rex->lastcloseparen = paren; \
5264 rex->offs[paren].end = -1; \
5267 case STAR: /* /A*B/ where A is width 1 */
5271 scan = NEXTOPER(scan);
5273 case PLUS: /* /A+B/ where A is width 1 */
5277 scan = NEXTOPER(scan);
5279 case CURLYN: /* /(A){m,n}B/ where A is width 1 */
5280 ST.paren = scan->flags; /* Which paren to set */
5281 if (ST.paren > PL_regsize)
5282 PL_regsize = ST.paren;
5283 if (ST.paren > rex->lastparen)
5284 rex->lastparen = ST.paren;
5285 ST.min = ARG1(scan); /* min to match */
5286 ST.max = ARG2(scan); /* max to match */
5287 if (cur_eval && cur_eval->u.eval.close_paren &&
5288 cur_eval->u.eval.close_paren == (U32)ST.paren) {
5292 scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE);
5294 case CURLY: /* /A{m,n}B/ where A is width 1 */
5296 ST.min = ARG1(scan); /* min to match */
5297 ST.max = ARG2(scan); /* max to match */
5298 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
5301 * Lookahead to avoid useless match attempts
5302 * when we know what character comes next.
5304 * Used to only do .*x and .*?x, but now it allows
5305 * for )'s, ('s and (?{ ... })'s to be in the way
5306 * of the quantifier and the EXACT-like node. -- japhy
5309 if (ST.min > ST.max) /* XXX make this a compile-time check? */
5311 if (HAS_TEXT(next) || JUMPABLE(next)) {
5313 regnode *text_node = next;
5315 if (! HAS_TEXT(text_node))
5316 FIND_NEXT_IMPT(text_node);
5318 if (! HAS_TEXT(text_node))
5319 ST.c1 = ST.c2 = CHRTEST_VOID;
5321 if ( PL_regkind[OP(text_node)] != EXACT ) {
5322 ST.c1 = ST.c2 = CHRTEST_VOID;
5323 goto assume_ok_easy;
5326 s = (U8*)STRING(text_node);
5328 /* Currently we only get here when
5330 PL_rekind[OP(text_node)] == EXACT
5332 if this changes back then the macro for IS_TEXT and
5333 friends need to change. */
5336 switch (OP(text_node)) {
5337 case EXACTF: ST.c2 = PL_fold[ST.c1]; break;
5340 case EXACTFU_TRICKYFOLD:
5341 case EXACTFU: ST.c2 = PL_fold_latin1[ST.c1]; break;
5342 case EXACTFL: ST.c2 = PL_fold_locale[ST.c1]; break;
5343 default: ST.c2 = ST.c1; break;
5346 else { /* UTF_PATTERN */
5347 if (IS_TEXTFU(text_node) || IS_TEXTF(text_node)) {
5349 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
5351 to_utf8_fold((U8*)s, tmpbuf, &ulen);
5352 ST.c1 = ST.c2 = utf8n_to_uvchr(tmpbuf, UTF8_MAXLEN, 0,
5356 ST.c2 = ST.c1 = utf8n_to_uvchr(s, UTF8_MAXBYTES, 0,
5363 ST.c1 = ST.c2 = CHRTEST_VOID;
5368 PL_reginput = locinput;
5371 if (ST.min && regrepeat(rex, ST.A, ST.min, depth) < ST.min)
5374 locinput = PL_reginput;
5376 if (ST.c1 == CHRTEST_VOID)
5377 goto curly_try_B_min;
5379 ST.oldloc = locinput;
5381 /* set ST.maxpos to the furthest point along the
5382 * string that could possibly match */
5383 if (ST.max == REG_INFTY) {
5384 ST.maxpos = PL_regeol - 1;
5386 while (UTF8_IS_CONTINUATION(*(U8*)ST.maxpos))
5389 else if (utf8_target) {
5390 int m = ST.max - ST.min;
5391 for (ST.maxpos = locinput;
5392 m >0 && ST.maxpos + UTF8SKIP(ST.maxpos) <= PL_regeol; m--)
5393 ST.maxpos += UTF8SKIP(ST.maxpos);
5396 ST.maxpos = locinput + ST.max - ST.min;
5397 if (ST.maxpos >= PL_regeol)
5398 ST.maxpos = PL_regeol - 1;
5400 goto curly_try_B_min_known;
5404 ST.count = regrepeat(rex, ST.A, ST.max, depth);
5405 locinput = PL_reginput;
5406 if (ST.count < ST.min)
5408 if ((ST.count > ST.min)
5409 && (PL_regkind[OP(ST.B)] == EOL) && (OP(ST.B) != MEOL))
5411 /* A{m,n} must come at the end of the string, there's
5412 * no point in backing off ... */
5414 /* ...except that $ and \Z can match before *and* after
5415 newline at the end. Consider "\n\n" =~ /\n+\Z\n/.
5416 We may back off by one in this case. */
5417 if (UCHARAT(PL_reginput - 1) == '\n' && OP(ST.B) != EOS)
5421 goto curly_try_B_max;
5426 case CURLY_B_min_known_fail:
5427 /* failed to find B in a non-greedy match where c1,c2 valid */
5428 if (ST.paren && ST.count)
5429 rex->offs[ST.paren].end = -1;
5431 PL_reginput = locinput; /* Could be reset... */
5432 REGCP_UNWIND(ST.cp);
5433 /* Couldn't or didn't -- move forward. */
5434 ST.oldloc = locinput;
5436 locinput += UTF8SKIP(locinput);
5440 curly_try_B_min_known:
5441 /* find the next place where 'B' could work, then call B */
5445 n = (ST.oldloc == locinput) ? 0 : 1;
5446 if (ST.c1 == ST.c2) {
5448 /* set n to utf8_distance(oldloc, locinput) */
5449 while (locinput <= ST.maxpos &&
5450 utf8n_to_uvchr((U8*)locinput,
5451 UTF8_MAXBYTES, &len,
5452 uniflags) != (UV)ST.c1) {
5458 /* set n to utf8_distance(oldloc, locinput) */
5459 while (locinput <= ST.maxpos) {
5461 const UV c = utf8n_to_uvchr((U8*)locinput,
5462 UTF8_MAXBYTES, &len,
5464 if (c == (UV)ST.c1 || c == (UV)ST.c2)
5472 if (ST.c1 == ST.c2) {
5473 while (locinput <= ST.maxpos &&
5474 UCHARAT(locinput) != ST.c1)
5478 while (locinput <= ST.maxpos
5479 && UCHARAT(locinput) != ST.c1
5480 && UCHARAT(locinput) != ST.c2)
5483 n = locinput - ST.oldloc;
5485 if (locinput > ST.maxpos)
5487 /* PL_reginput == oldloc now */
5490 if (regrepeat(rex, ST.A, n, depth) < n)
5493 PL_reginput = locinput;
5494 CURLY_SETPAREN(ST.paren, ST.count);
5495 if (cur_eval && cur_eval->u.eval.close_paren &&
5496 cur_eval->u.eval.close_paren == (U32)ST.paren) {
5499 PUSH_STATE_GOTO(CURLY_B_min_known, ST.B);
5504 case CURLY_B_min_fail:
5505 /* failed to find B in a non-greedy match where c1,c2 invalid */
5506 if (ST.paren && ST.count)
5507 rex->offs[ST.paren].end = -1;
5509 REGCP_UNWIND(ST.cp);
5510 /* failed -- move forward one */
5511 PL_reginput = locinput;
5512 if (regrepeat(rex, ST.A, 1, depth)) {
5514 locinput = PL_reginput;
5515 if (ST.count <= ST.max || (ST.max == REG_INFTY &&
5516 ST.count > 0)) /* count overflow ? */
5519 CURLY_SETPAREN(ST.paren, ST.count);
5520 if (cur_eval && cur_eval->u.eval.close_paren &&
5521 cur_eval->u.eval.close_paren == (U32)ST.paren) {
5524 PUSH_STATE_GOTO(CURLY_B_min, ST.B);
5532 /* a successful greedy match: now try to match B */
5533 if (cur_eval && cur_eval->u.eval.close_paren &&
5534 cur_eval->u.eval.close_paren == (U32)ST.paren) {
5539 if (ST.c1 != CHRTEST_VOID)
5540 c = utf8_target ? utf8n_to_uvchr((U8*)PL_reginput,
5541 UTF8_MAXBYTES, 0, uniflags)
5542 : (UV) UCHARAT(PL_reginput);
5543 /* If it could work, try it. */
5544 if (ST.c1 == CHRTEST_VOID || c == (UV)ST.c1 || c == (UV)ST.c2) {
5545 CURLY_SETPAREN(ST.paren, ST.count);
5546 PUSH_STATE_GOTO(CURLY_B_max, ST.B);
5551 case CURLY_B_max_fail:
5552 /* failed to find B in a greedy match */
5553 if (ST.paren && ST.count)
5554 rex->offs[ST.paren].end = -1;
5556 REGCP_UNWIND(ST.cp);
5558 if (--ST.count < ST.min)
5560 PL_reginput = locinput = HOPc(locinput, -1);
5561 goto curly_try_B_max;
5568 /* we've just finished A in /(??{A})B/; now continue with B */
5570 st->u.eval.toggle_reg_flags
5571 = cur_eval->u.eval.toggle_reg_flags;
5572 PL_reg_flags ^= st->u.eval.toggle_reg_flags;
5574 st->u.eval.prev_rex = rex_sv; /* inner */
5575 st->u.eval.cp = regcppush(rex, 0); /* Save *all* the positions. */
5576 SETREX(rex_sv,cur_eval->u.eval.prev_rex);
5577 rex = (struct regexp *)SvANY(rex_sv);
5578 rexi = RXi_GET(rex);
5579 cur_curlyx = cur_eval->u.eval.prev_curlyx;
5580 (void)ReREFCNT_inc(rex_sv);
5582 REGCP_SET(st->u.eval.lastcp);
5583 PL_reginput = locinput;
5585 /* Restore parens of the outer rex without popping the
5587 tmpix = PL_savestack_ix;
5588 PL_savestack_ix = cur_eval->u.eval.lastcp;
5590 PL_savestack_ix = tmpix;
5592 st->u.eval.prev_eval = cur_eval;
5593 cur_eval = cur_eval->u.eval.prev_eval;
5595 PerlIO_printf(Perl_debug_log, "%*s EVAL trying tail ... %"UVxf"\n",
5596 REPORT_CODE_OFF+depth*2, "",PTR2UV(cur_eval)););
5597 if ( nochange_depth )
5600 PUSH_YES_STATE_GOTO(EVAL_AB,
5601 st->u.eval.prev_eval->u.eval.B); /* match B */
5604 if (locinput < reginfo->till) {
5605 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
5606 "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n",
5608 (long)(locinput - PL_reg_starttry),
5609 (long)(reginfo->till - PL_reg_starttry),
5612 sayNO_SILENT; /* Cannot match: too short. */
5614 PL_reginput = locinput; /* put where regtry can find it */
5615 sayYES; /* Success! */
5617 case SUCCEED: /* successful SUSPEND/UNLESSM/IFMATCH/CURLYM */
5619 PerlIO_printf(Perl_debug_log,
5620 "%*s %ssubpattern success...%s\n",
5621 REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5]));
5622 PL_reginput = locinput; /* put where regtry can find it */
5623 sayYES; /* Success! */
5626 #define ST st->u.ifmatch
5628 case SUSPEND: /* (?>A) */
5630 PL_reginput = locinput;
5633 case UNLESSM: /* -ve lookaround: (?!A), or with flags, (?<!A) */
5635 goto ifmatch_trivial_fail_test;
5637 case IFMATCH: /* +ve lookaround: (?=A), or with flags, (?<=A) */
5639 ifmatch_trivial_fail_test:
5641 char * const s = HOPBACKc(locinput, scan->flags);
5646 sw = 1 - cBOOL(ST.wanted);
5650 next = scan + ARG(scan);
5658 PL_reginput = locinput;
5662 ST.logical = logical;
5663 logical = 0; /* XXX: reset state of logical once it has been saved into ST */
5665 /* execute body of (?...A) */
5666 PUSH_YES_STATE_GOTO(IFMATCH_A, NEXTOPER(NEXTOPER(scan)));
5669 case IFMATCH_A_fail: /* body of (?...A) failed */
5670 ST.wanted = !ST.wanted;
5673 case IFMATCH_A: /* body of (?...A) succeeded */
5675 sw = cBOOL(ST.wanted);
5677 else if (!ST.wanted)
5680 if (OP(ST.me) == SUSPEND)
5681 locinput = PL_reginput;
5683 locinput = PL_reginput = st->locinput;
5684 nextchr = UCHARAT(locinput);
5686 scan = ST.me + ARG(ST.me);
5689 continue; /* execute B */
5694 next = scan + ARG(scan);
5699 reginfo->cutpoint = PL_regeol;
5702 PL_reginput = locinput;
5704 sv_yes_mark = sv_commit = MUTABLE_SV(rexi->data->data[ ARG( scan ) ]);
5705 PUSH_STATE_GOTO(COMMIT_next,next);
5707 case COMMIT_next_fail:
5714 #define ST st->u.mark
5716 ST.prev_mark = mark_state;
5717 ST.mark_name = sv_commit = sv_yes_mark
5718 = MUTABLE_SV(rexi->data->data[ ARG( scan ) ]);
5720 ST.mark_loc = PL_reginput = locinput;
5721 PUSH_YES_STATE_GOTO(MARKPOINT_next,next);
5723 case MARKPOINT_next:
5724 mark_state = ST.prev_mark;
5727 case MARKPOINT_next_fail:
5728 if (popmark && sv_eq(ST.mark_name,popmark))
5730 if (ST.mark_loc > startpoint)
5731 reginfo->cutpoint = HOPBACKc(ST.mark_loc, 1);
5732 popmark = NULL; /* we found our mark */
5733 sv_commit = ST.mark_name;
5736 PerlIO_printf(Perl_debug_log,
5737 "%*s %ssetting cutpoint to mark:%"SVf"...%s\n",
5738 REPORT_CODE_OFF+depth*2, "",
5739 PL_colors[4], SVfARG(sv_commit), PL_colors[5]);
5742 mark_state = ST.prev_mark;
5743 sv_yes_mark = mark_state ?
5744 mark_state->u.mark.mark_name : NULL;
5748 PL_reginput = locinput;
5750 /* (*SKIP) : if we fail we cut here*/
5751 ST.mark_name = NULL;
5752 ST.mark_loc = locinput;
5753 PUSH_STATE_GOTO(SKIP_next,next);
5755 /* (*SKIP:NAME) : if there is a (*MARK:NAME) fail where it was,
5756 otherwise do nothing. Meaning we need to scan
5758 regmatch_state *cur = mark_state;
5759 SV *find = MUTABLE_SV(rexi->data->data[ ARG( scan ) ]);
5762 if ( sv_eq( cur->u.mark.mark_name,
5765 ST.mark_name = find;
5766 PUSH_STATE_GOTO( SKIP_next, next );
5768 cur = cur->u.mark.prev_mark;
5771 /* Didn't find our (*MARK:NAME) so ignore this (*SKIP:NAME) */
5773 case SKIP_next_fail:
5775 /* (*CUT:NAME) - Set up to search for the name as we
5776 collapse the stack*/
5777 popmark = ST.mark_name;
5779 /* (*CUT) - No name, we cut here.*/
5780 if (ST.mark_loc > startpoint)
5781 reginfo->cutpoint = HOPBACKc(ST.mark_loc, 1);
5782 /* but we set sv_commit to latest mark_name if there
5783 is one so they can test to see how things lead to this
5786 sv_commit=mark_state->u.mark.mark_name;
5793 if ((n=is_LNBREAK(locinput,utf8_target))) {
5795 nextchr = UCHARAT(locinput);
5800 #define CASE_CLASS(nAmE) \
5802 if (locinput >= PL_regeol) \
5804 if ((n=is_##nAmE(locinput,utf8_target))) { \
5806 nextchr = UCHARAT(locinput); \
5811 if (locinput >= PL_regeol) \
5813 if ((n=is_##nAmE(locinput,utf8_target))) { \
5816 locinput += UTF8SKIP(locinput); \
5817 nextchr = UCHARAT(locinput); \
5822 CASE_CLASS(HORIZWS);
5826 PerlIO_printf(Perl_error_log, "%"UVxf" %d\n",
5827 PTR2UV(scan), OP(scan));
5828 Perl_croak(aTHX_ "regexp memory corruption");
5832 /* switch break jumps here */
5833 scan = next; /* prepare to execute the next op and ... */
5834 continue; /* ... jump back to the top, reusing st */
5838 /* push a state that backtracks on success */
5839 st->u.yes.prev_yes_state = yes_state;
5843 /* push a new regex state, then continue at scan */
5845 regmatch_state *newst;
5848 regmatch_state *cur = st;
5849 regmatch_state *curyes = yes_state;
5851 regmatch_slab *slab = PL_regmatch_slab;
5852 for (;curd > -1;cur--,curd--) {
5853 if (cur < SLAB_FIRST(slab)) {
5855 cur = SLAB_LAST(slab);
5857 PerlIO_printf(Perl_error_log, "%*s#%-3d %-10s %s\n",
5858 REPORT_CODE_OFF + 2 + depth * 2,"",
5859 curd, PL_reg_name[cur->resume_state],
5860 (curyes == cur) ? "yes" : ""
5863 curyes = cur->u.yes.prev_yes_state;
5866 DEBUG_STATE_pp("push")
5869 st->locinput = locinput;
5871 if (newst > SLAB_LAST(PL_regmatch_slab))
5872 newst = S_push_slab(aTHX);
5873 PL_regmatch_state = newst;
5875 locinput = PL_reginput;
5876 nextchr = UCHARAT(locinput);
5884 * We get here only if there's trouble -- normally "case END" is
5885 * the terminating point.
5887 Perl_croak(aTHX_ "corrupted regexp pointers");
5893 /* we have successfully completed a subexpression, but we must now
5894 * pop to the state marked by yes_state and continue from there */
5895 assert(st != yes_state);
5897 while (st != yes_state) {
5899 if (st < SLAB_FIRST(PL_regmatch_slab)) {
5900 PL_regmatch_slab = PL_regmatch_slab->prev;
5901 st = SLAB_LAST(PL_regmatch_slab);
5905 DEBUG_STATE_pp("pop (no final)");
5907 DEBUG_STATE_pp("pop (yes)");
5913 while (yes_state < SLAB_FIRST(PL_regmatch_slab)
5914 || yes_state > SLAB_LAST(PL_regmatch_slab))
5916 /* not in this slab, pop slab */
5917 depth -= (st - SLAB_FIRST(PL_regmatch_slab) + 1);
5918 PL_regmatch_slab = PL_regmatch_slab->prev;
5919 st = SLAB_LAST(PL_regmatch_slab);
5921 depth -= (st - yes_state);
5924 yes_state = st->u.yes.prev_yes_state;
5925 PL_regmatch_state = st;
5928 locinput= st->locinput;
5929 nextchr = UCHARAT(locinput);
5931 state_num = st->resume_state + no_final;
5932 goto reenter_switch;
5935 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n",
5936 PL_colors[4], PL_colors[5]));
5938 if (PL_reg_state.re_state_eval_setup_done) {
5939 /* each successfully executed (?{...}) block does the equivalent of
5940 * local $^R = do {...}
5941 * When popping the save stack, all these locals would be undone;
5942 * bypass this by setting the outermost saved $^R to the latest
5944 if (oreplsv != GvSV(PL_replgv))
5945 sv_setsv(oreplsv, GvSV(PL_replgv));
5952 PerlIO_printf(Perl_debug_log,
5953 "%*s %sfailed...%s\n",
5954 REPORT_CODE_OFF+depth*2, "",
5955 PL_colors[4], PL_colors[5])
5967 /* there's a previous state to backtrack to */
5969 if (st < SLAB_FIRST(PL_regmatch_slab)) {
5970 PL_regmatch_slab = PL_regmatch_slab->prev;
5971 st = SLAB_LAST(PL_regmatch_slab);
5973 PL_regmatch_state = st;
5974 locinput= st->locinput;
5975 nextchr = UCHARAT(locinput);
5977 DEBUG_STATE_pp("pop");
5979 if (yes_state == st)
5980 yes_state = st->u.yes.prev_yes_state;
5982 state_num = st->resume_state + 1; /* failure = success + 1 */
5983 goto reenter_switch;
5988 if (rex->intflags & PREGf_VERBARG_SEEN) {
5989 SV *sv_err = get_sv("REGERROR", 1);
5990 SV *sv_mrk = get_sv("REGMARK", 1);
5992 sv_commit = &PL_sv_no;
5994 sv_yes_mark = &PL_sv_yes;
5997 sv_commit = &PL_sv_yes;
5998 sv_yes_mark = &PL_sv_no;
6000 sv_setsv(sv_err, sv_commit);
6001 sv_setsv(sv_mrk, sv_yes_mark);
6005 if (last_pushed_cv) {
6010 /* clean up; in particular, free all slabs above current one */
6011 LEAVE_SCOPE(oldsave);
6017 - regrepeat - repeatedly match something simple, report how many
6020 * [This routine now assumes that it will only match on things of length 1.
6021 * That was true before, but now we assume scan - reginput is the count,
6022 * rather than incrementing count on every character. [Er, except utf8.]]
6025 S_regrepeat(pTHX_ const regexp *prog, const regnode *p, I32 max, int depth)
6028 register char *scan;
6030 register char *loceol = PL_regeol;
6031 register I32 hardcount = 0;
6032 register bool utf8_target = PL_reg_match_utf8;
6035 PERL_UNUSED_ARG(depth);
6038 PERL_ARGS_ASSERT_REGREPEAT;
6041 if (max == REG_INFTY)
6043 else if (max < loceol - scan)
6044 loceol = scan + max;
6049 while (scan < loceol && hardcount < max && *scan != '\n') {
6050 scan += UTF8SKIP(scan);
6054 while (scan < loceol && *scan != '\n')
6061 while (scan < loceol && hardcount < max) {
6062 scan += UTF8SKIP(scan);
6073 /* To get here, EXACTish nodes must have *byte* length == 1. That
6074 * means they match only characters in the string that can be expressed
6075 * as a single byte. For non-utf8 strings, that means a simple match.
6076 * For utf8 strings, the character matched must be an invariant, or
6077 * downgradable to a single byte. The pattern's utf8ness is
6078 * irrelevant, as since it's a single byte, it either isn't utf8, or if
6079 * it is, it's an invariant */
6082 assert(! UTF_PATTERN || UNI_IS_INVARIANT(c));
6084 if (! utf8_target || UNI_IS_INVARIANT(c)) {
6085 while (scan < loceol && UCHARAT(scan) == c) {
6091 /* Here, the string is utf8, and the pattern char is different
6092 * in utf8 than not, so can't compare them directly. Outside the
6093 * loop, find the two utf8 bytes that represent c, and then
6094 * look for those in sequence in the utf8 string */
6095 U8 high = UTF8_TWO_BYTE_HI(c);
6096 U8 low = UTF8_TWO_BYTE_LO(c);
6099 while (hardcount < max
6100 && scan + 1 < loceol
6101 && UCHARAT(scan) == high
6102 && UCHARAT(scan + 1) == low)
6110 utf8_flags = FOLDEQ_UTF8_NOMIX_ASCII;
6114 PL_reg_flags |= RF_tainted;
6115 utf8_flags = FOLDEQ_UTF8_LOCALE;
6123 case EXACTFU_TRICKYFOLD:
6125 utf8_flags = (UTF_PATTERN) ? FOLDEQ_S2_ALREADY_FOLDED : 0;
6127 /* The comments for the EXACT case above apply as well to these fold
6132 assert(! UTF_PATTERN || UNI_IS_INVARIANT(c));
6134 if (utf8_target || OP(p) == EXACTFU_SS) { /* Use full Unicode fold matching */
6135 char *tmpeol = loceol;
6136 while (hardcount < max
6137 && foldEQ_utf8_flags(scan, &tmpeol, 0, utf8_target,
6138 STRING(p), NULL, 1, cBOOL(UTF_PATTERN), utf8_flags))
6145 /* XXX Note that the above handles properly the German sharp s in
6146 * the pattern matching ss in the string. But it doesn't handle
6147 * properly cases where the string contains say 'LIGATURE ff' and
6148 * the pattern is 'f+'. This would require, say, a new function or
6149 * revised interface to foldEQ_utf8(), in which the maximum number
6150 * of characters to match could be passed and it would return how
6151 * many actually did. This is just one of many cases where
6152 * multi-char folds don't work properly, and so the fix is being
6158 /* Here, the string isn't utf8 and c is a single byte; and either
6159 * the pattern isn't utf8 or c is an invariant, so its utf8ness
6160 * doesn't affect c. Can just do simple comparisons for exact or
6163 case EXACTF: folded = PL_fold[c]; break;
6165 case EXACTFU_TRICKYFOLD:
6166 case EXACTFU: folded = PL_fold_latin1[c]; break;
6167 case EXACTFL: folded = PL_fold_locale[c]; break;
6168 default: Perl_croak(aTHX_ "panic: Unexpected op %u", OP(p));
6170 while (scan < loceol &&
6171 (UCHARAT(scan) == c || UCHARAT(scan) == folded))
6179 if (utf8_target || OP(p) == ANYOFV) {
6182 inclasslen = loceol - scan;
6183 while (hardcount < max
6184 && ((inclasslen = loceol - scan) > 0)
6185 && reginclass(prog, p, (U8*)scan, &inclasslen, utf8_target))
6191 while (scan < loceol && REGINCLASS(prog, p, (U8*)scan))
6199 LOAD_UTF8_CHARCLASS_ALNUM();
6200 while (hardcount < max && scan < loceol &&
6201 swash_fetch(PL_utf8_alnum, (U8*)scan, utf8_target))
6203 scan += UTF8SKIP(scan);
6207 while (scan < loceol && isWORDCHAR_L1((U8) *scan)) {
6215 while (scan < loceol && isALNUM((U8) *scan)) {
6220 while (scan < loceol && isWORDCHAR_A((U8) *scan)) {
6225 PL_reg_flags |= RF_tainted;
6228 while (hardcount < max && scan < loceol &&
6229 isALNUM_LC_utf8((U8*)scan)) {
6230 scan += UTF8SKIP(scan);
6234 while (scan < loceol && isALNUM_LC(*scan))
6244 LOAD_UTF8_CHARCLASS_ALNUM();
6245 while (hardcount < max && scan < loceol &&
6246 ! swash_fetch(PL_utf8_alnum, (U8*)scan, utf8_target))
6248 scan += UTF8SKIP(scan);
6252 while (scan < loceol && ! isWORDCHAR_L1((U8) *scan)) {
6259 goto utf8_Nwordchar;
6260 while (scan < loceol && ! isALNUM((U8) *scan)) {
6266 while (scan < loceol && ! isWORDCHAR_A((U8) *scan)) {
6267 scan += UTF8SKIP(scan);
6271 while (scan < loceol && ! isWORDCHAR_A((U8) *scan)) {
6277 PL_reg_flags |= RF_tainted;
6280 while (hardcount < max && scan < loceol &&
6281 !isALNUM_LC_utf8((U8*)scan)) {
6282 scan += UTF8SKIP(scan);
6286 while (scan < loceol && !isALNUM_LC(*scan))
6296 LOAD_UTF8_CHARCLASS_SPACE();
6297 while (hardcount < max && scan < loceol &&
6299 swash_fetch(PL_utf8_space,(U8*)scan, utf8_target)))
6301 scan += UTF8SKIP(scan);
6307 while (scan < loceol && isSPACE_L1((U8) *scan)) {
6316 while (scan < loceol && isSPACE((U8) *scan)) {
6321 while (scan < loceol && isSPACE_A((U8) *scan)) {
6326 PL_reg_flags |= RF_tainted;
6329 while (hardcount < max && scan < loceol &&
6330 isSPACE_LC_utf8((U8*)scan)) {
6331 scan += UTF8SKIP(scan);
6335 while (scan < loceol && isSPACE_LC(*scan))
6345 LOAD_UTF8_CHARCLASS_SPACE();
6346 while (hardcount < max && scan < loceol &&
6348 swash_fetch(PL_utf8_space,(U8*)scan, utf8_target)))
6350 scan += UTF8SKIP(scan);
6356 while (scan < loceol && ! isSPACE_L1((U8) *scan)) {
6365 while (scan < loceol && ! isSPACE((U8) *scan)) {
6371 while (scan < loceol && ! isSPACE_A((U8) *scan)) {
6372 scan += UTF8SKIP(scan);
6376 while (scan < loceol && ! isSPACE_A((U8) *scan)) {
6382 PL_reg_flags |= RF_tainted;
6385 while (hardcount < max && scan < loceol &&
6386 !isSPACE_LC_utf8((U8*)scan)) {
6387 scan += UTF8SKIP(scan);
6391 while (scan < loceol && !isSPACE_LC(*scan))
6398 LOAD_UTF8_CHARCLASS_DIGIT();
6399 while (hardcount < max && scan < loceol &&
6400 swash_fetch(PL_utf8_digit, (U8*)scan, utf8_target)) {
6401 scan += UTF8SKIP(scan);
6405 while (scan < loceol && isDIGIT(*scan))
6410 while (scan < loceol && isDIGIT_A((U8) *scan)) {
6415 PL_reg_flags |= RF_tainted;
6418 while (hardcount < max && scan < loceol &&
6419 isDIGIT_LC_utf8((U8*)scan)) {
6420 scan += UTF8SKIP(scan);
6424 while (scan < loceol && isDIGIT_LC(*scan))
6431 LOAD_UTF8_CHARCLASS_DIGIT();
6432 while (hardcount < max && scan < loceol &&
6433 !swash_fetch(PL_utf8_digit, (U8*)scan, utf8_target)) {
6434 scan += UTF8SKIP(scan);
6438 while (scan < loceol && !isDIGIT(*scan))
6444 while (scan < loceol && ! isDIGIT_A((U8) *scan)) {
6445 scan += UTF8SKIP(scan);
6449 while (scan < loceol && ! isDIGIT_A((U8) *scan)) {
6455 PL_reg_flags |= RF_tainted;
6458 while (hardcount < max && scan < loceol &&
6459 !isDIGIT_LC_utf8((U8*)scan)) {
6460 scan += UTF8SKIP(scan);
6464 while (scan < loceol && !isDIGIT_LC(*scan))
6471 while (hardcount < max && scan < loceol && (c=is_LNBREAK_utf8(scan))) {
6477 LNBREAK can match two latin chars, which is ok,
6478 because we have a null terminated string, but we
6479 have to use hardcount in this situation
6481 while (scan < loceol && (c=is_LNBREAK_latin1(scan))) {
6490 while (hardcount < max && scan < loceol && (c=is_HORIZWS_utf8(scan))) {
6495 while (scan < loceol && is_HORIZWS_latin1(scan))
6502 while (hardcount < max && scan < loceol && !is_HORIZWS_utf8(scan)) {
6503 scan += UTF8SKIP(scan);
6507 while (scan < loceol && !is_HORIZWS_latin1(scan))
6515 while (hardcount < max && scan < loceol && (c=is_VERTWS_utf8(scan))) {
6520 while (scan < loceol && is_VERTWS_latin1(scan))
6528 while (hardcount < max && scan < loceol && !is_VERTWS_utf8(scan)) {
6529 scan += UTF8SKIP(scan);
6533 while (scan < loceol && !is_VERTWS_latin1(scan))
6539 default: /* Called on something of 0 width. */
6540 break; /* So match right here or not at all. */
6546 c = scan - PL_reginput;
6550 GET_RE_DEBUG_FLAGS_DECL;
6552 SV * const prop = sv_newmortal();
6553 regprop(prog, prop, p);
6554 PerlIO_printf(Perl_debug_log,
6555 "%*s %s can match %"IVdf" times out of %"IVdf"...\n",
6556 REPORT_CODE_OFF + depth*2, "", SvPVX_const(prop),(IV)c,(IV)max);
6564 #if !defined(PERL_IN_XSUB_RE) || defined(PLUGGABLE_RE_EXTENSION)
6566 - regclass_swash - prepare the utf8 swash. Wraps the shared core version to
6567 create a copy so that changes the caller makes won't change the shared one
6570 Perl_regclass_swash(pTHX_ const regexp *prog, register const regnode* node, bool doinit, SV** listsvp, SV **altsvp)
6572 PERL_ARGS_ASSERT_REGCLASS_SWASH;
6573 return newSVsv(core_regclass_swash(prog, node, doinit, listsvp, altsvp));
6578 S_core_regclass_swash(pTHX_ const regexp *prog, register const regnode* node, bool doinit, SV** listsvp, SV **altsvp)
6580 /* Returns the swash for the input 'node' in the regex 'prog'.
6581 * If <doinit> is true, will attempt to create the swash if not already
6583 * If <listsvp> is non-null, will return the swash initialization string in
6585 * If <altsvp> is non-null, will return the alternates to the regular swash
6587 * Tied intimately to how regcomp.c sets up the data structure */
6595 RXi_GET_DECL(prog,progi);
6596 const struct reg_data * const data = prog ? progi->data : NULL;
6598 PERL_ARGS_ASSERT_CORE_REGCLASS_SWASH;
6600 assert(ANYOF_NONBITMAP(node));
6602 if (data && data->count) {
6603 const U32 n = ARG(node);
6605 if (data->what[n] == 's') {
6606 SV * const rv = MUTABLE_SV(data->data[n]);
6607 AV * const av = MUTABLE_AV(SvRV(rv));
6608 SV **const ary = AvARRAY(av);
6609 bool invlist_has_user_defined_property;
6611 si = *ary; /* ary[0] = the string to initialize the swash with */
6613 /* Elements 3 and 4 are either both present or both absent. [3] is
6614 * any inversion list generated at compile time; [4] indicates if
6615 * that inversion list has any user-defined properties in it. */
6616 if (av_len(av) >= 3) {
6618 invlist_has_user_defined_property = cBOOL(SvUV(ary[4]));
6622 invlist_has_user_defined_property = FALSE;
6625 /* Element [1] is reserved for the set-up swash. If already there,
6626 * return it; if not, create it and store it there */
6627 if (SvROK(ary[1])) {
6630 else if (si && doinit) {
6632 sw = _core_swash_init("utf8", /* the utf8 package */
6636 0, /* not from tr/// */
6637 FALSE, /* is error if can't find
6640 invlist_has_user_defined_property);
6641 (void)av_store(av, 1, sw);
6644 /* Element [2] is for any multi-char folds. Note that is a
6645 * fundamentally flawed design, because can't backtrack and try
6646 * again. See [perl #89774] */
6647 if (SvTYPE(ary[2]) == SVt_PVAV) {
6654 SV* matches_string = newSVpvn("", 0);
6657 /* Use the swash, if any, which has to have incorporated into it all
6661 && SvTYPE(SvRV(sw)) == SVt_PVHV
6662 && (invlistsvp = hv_fetchs(MUTABLE_HV(SvRV(sw)), "INVLIST", FALSE)))
6664 invlist = *invlistsvp;
6666 else if (si && si != &PL_sv_undef) {
6668 /* If no swash, use the input nitialization string, if available */
6669 sv_catsv(matches_string, si);
6672 /* Add the inversion list to whatever we have. This may have come from
6673 * the swash, or from an input parameter */
6675 sv_catsv(matches_string, _invlist_contents(invlist));
6677 *listsvp = matches_string;
6687 - reginclass - determine if a character falls into a character class
6689 n is the ANYOF regnode
6690 p is the target string
6691 lenp is pointer to the maximum number of bytes of how far to go in p
6692 (This is assumed wthout checking to always be at least the current
6694 utf8_target tells whether p is in UTF-8.
6696 Returns true if matched; false otherwise. If lenp is not NULL, on return
6697 from a successful match, the value it points to will be updated to how many
6698 bytes in p were matched. If there was no match, the value is undefined,
6699 possibly changed from the input.
6701 Note that this can be a synthetic start class, a combination of various
6702 nodes, so things you think might be mutually exclusive, such as locale,
6703 aren't. It can match both locale and non-locale
6708 S_reginclass(pTHX_ const regexp * const prog, register const regnode * const n, register const U8* const p, STRLEN* lenp, register const bool utf8_target)
6711 const char flags = ANYOF_FLAGS(n);
6717 PERL_ARGS_ASSERT_REGINCLASS;
6719 /* If c is not already the code point, get it */
6720 if (utf8_target && !UTF8_IS_INVARIANT(c)) {
6721 c = utf8n_to_uvchr(p, UTF8_MAXBYTES, &c_len,
6722 (UTF8_ALLOW_DEFAULT & UTF8_ALLOW_ANYUV)
6723 | UTF8_ALLOW_FFFF | UTF8_CHECK_ONLY);
6724 /* see [perl #37836] for UTF8_ALLOW_ANYUV; [perl #38293] for
6725 * UTF8_ALLOW_FFFF */
6726 if (c_len == (STRLEN)-1)
6727 Perl_croak(aTHX_ "Malformed UTF-8 character (fatal)");
6733 /* Use passed in max length, or one character if none passed in or less
6734 * than one character. And assume will match just one character. This is
6735 * overwritten later if matched more. */
6737 maxlen = (*lenp > c_len) ? *lenp : c_len;
6745 /* If this character is potentially in the bitmap, check it */
6747 if (ANYOF_BITMAP_TEST(n, c))
6749 else if (flags & ANYOF_NON_UTF8_LATIN1_ALL
6756 else if (flags & ANYOF_LOCALE) {
6757 PL_reg_flags |= RF_tainted;
6759 if ((flags & ANYOF_LOC_NONBITMAP_FOLD)
6760 && ANYOF_BITMAP_TEST(n, PL_fold_locale[c]))
6764 else if (ANYOF_CLASS_TEST_ANY_SET(n) &&
6765 ((ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) ||
6766 (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) ||
6767 (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) ||
6768 (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) ||
6769 (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) ||
6770 (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) ||
6771 (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) ||
6772 (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) ||
6773 (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) ||
6774 (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) ||
6775 (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII_LC(c)) ||
6776 (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII_LC(c)) ||
6777 (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) ||
6778 (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) ||
6779 (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) ||
6780 (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) ||
6781 (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) ||
6782 (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) ||
6783 (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) ||
6784 (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) ||
6785 (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) ||
6786 (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) ||
6787 (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) ||
6788 (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) ||
6789 (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) ||
6790 (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) ||
6791 (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) ||
6792 (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) ||
6793 (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK_LC(c)) ||
6794 (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK_LC(c))
6795 ) /* How's that for a conditional? */
6802 /* If the bitmap didn't (or couldn't) match, and something outside the
6803 * bitmap could match, try that. Locale nodes specifiy completely the
6804 * behavior of code points in the bit map (otherwise, a utf8 target would
6805 * cause them to be treated as Unicode and not locale), except in
6806 * the very unlikely event when this node is a synthetic start class, which
6807 * could be a combination of locale and non-locale nodes. So allow locale
6808 * to match for the synthetic start class, which will give a false
6809 * positive that will be resolved when the match is done again as not part
6810 * of the synthetic start class */
6812 if (utf8_target && (flags & ANYOF_UNICODE_ALL) && c >= 256) {
6813 match = TRUE; /* Everything above 255 matches */
6815 else if (ANYOF_NONBITMAP(n)
6816 && ((flags & ANYOF_NONBITMAP_NON_UTF8)
6819 || (! (flags & ANYOF_LOCALE))
6820 || (flags & ANYOF_IS_SYNTHETIC)))))
6823 SV * const sw = core_regclass_swash(prog, n, TRUE, 0, (SV**)&av);
6831 /* Not utf8. Convert as much of the string as available up
6832 * to the limit of how far the (single) character in the
6833 * pattern can possibly match (no need to go further). If
6834 * the node is a straight ANYOF or not folding, it can't
6835 * match more than one. Otherwise, It can match up to how
6836 * far a single char can fold to. Since not utf8, each
6837 * character is a single byte, so the max it can be in
6838 * bytes is the same as the max it can be in characters */
6839 STRLEN len = (OP(n) == ANYOF
6840 || ! (flags & ANYOF_LOC_NONBITMAP_FOLD))
6842 : (maxlen < UTF8_MAX_FOLD_CHAR_EXPAND)
6844 : UTF8_MAX_FOLD_CHAR_EXPAND;
6845 utf8_p = bytes_to_utf8(p, &len);
6848 if (swash_fetch(sw, utf8_p, TRUE))
6850 else if (flags & ANYOF_LOC_NONBITMAP_FOLD) {
6852 /* Here, we need to test if the fold of the target string
6853 * matches. The non-multi char folds have all been moved to
6854 * the compilation phase, and the multi-char folds have
6855 * been stored by regcomp into 'av'; we linearly check to
6856 * see if any match the target string (folded). We know
6857 * that the originals were each one character, but we don't
6858 * currently know how many characters/bytes each folded to,
6859 * except we do know that there are small limits imposed by
6860 * Unicode. XXX A performance enhancement would be to have
6861 * regcomp.c store the max number of chars/bytes that are
6862 * in an av entry, as, say the 0th element. Even better
6863 * would be to have a hash of the few characters that can
6864 * start a multi-char fold to the max number of chars of
6867 * If there is a match, we will need to advance (if lenp is
6868 * specified) the match pointer in the target string. But
6869 * what we are comparing here isn't that string directly,
6870 * but its fold, whose length may differ from the original.
6871 * As we go along in constructing the fold, therefore, we
6872 * create a map so that we know how many bytes in the
6873 * source to advance given that we have matched a certain
6874 * number of bytes in the fold. This map is stored in
6875 * 'map_fold_len_back'. Let n mean the number of bytes in
6876 * the fold of the first character that we are folding.
6877 * Then map_fold_len_back[n] is set to the number of bytes
6878 * in that first character. Similarly let m be the
6879 * corresponding number for the second character to be
6880 * folded. Then map_fold_len_back[n+m] is set to the
6881 * number of bytes occupied by the first two source
6882 * characters. ... */
6883 U8 map_fold_len_back[UTF8_MAXBYTES_CASE+1] = { 0 };
6884 U8 folded[UTF8_MAXBYTES_CASE+1];
6885 STRLEN foldlen = 0; /* num bytes in fold of 1st char */
6886 STRLEN total_foldlen = 0; /* num bytes in fold of all
6889 if (OP(n) == ANYOF || maxlen == 1 || ! lenp || ! av) {
6891 /* Here, only need to fold the first char of the target
6892 * string. It the source wasn't utf8, is 1 byte long */
6893 to_utf8_fold(utf8_p, folded, &foldlen);
6894 total_foldlen = foldlen;
6895 map_fold_len_back[foldlen] = (utf8_target)
6901 /* Here, need to fold more than the first char. Do so
6902 * up to the limits */
6903 U8* source_ptr = utf8_p; /* The source for the fold
6906 U8* folded_ptr = folded;
6907 U8* e = utf8_p + maxlen; /* Can't go beyond last
6908 available byte in the
6912 i < UTF8_MAX_FOLD_CHAR_EXPAND && source_ptr < e;
6916 /* Fold the next character */
6917 U8 this_char_folded[UTF8_MAXBYTES_CASE+1];
6918 STRLEN this_char_foldlen;
6919 to_utf8_fold(source_ptr,
6921 &this_char_foldlen);
6923 /* Bail if it would exceed the byte limit for
6924 * folding a single char. */
6925 if (this_char_foldlen + folded_ptr - folded >
6931 /* Add the fold of this character */
6932 Copy(this_char_folded,
6936 source_ptr += UTF8SKIP(source_ptr);
6937 folded_ptr += this_char_foldlen;
6938 total_foldlen = folded_ptr - folded;
6940 /* Create map from the number of bytes in the fold
6941 * back to the number of bytes in the source. If
6942 * the source isn't utf8, the byte count is just
6943 * the number of characters so far */
6944 map_fold_len_back[total_foldlen]
6946 ? source_ptr - utf8_p
6953 /* Do the linear search to see if the fold is in the list
6954 * of multi-char folds. */
6957 for (i = 0; i <= av_len(av); i++) {
6958 SV* const sv = *av_fetch(av, i, FALSE);
6960 const char * const s = SvPV_const(sv, len);
6962 if (len <= total_foldlen
6963 && memEQ(s, (char*)folded, len)
6965 /* If 0, means matched a partial char. See
6967 && map_fold_len_back[len])
6970 /* Advance the target string ptr to account for
6971 * this fold, but have to translate from the
6972 * folded length to the corresponding source
6975 *lenp = map_fold_len_back[len];
6984 /* If we allocated a string above, free it */
6985 if (! utf8_target) Safefree(utf8_p);
6990 return (flags & ANYOF_INVERT) ? !match : match;
6994 S_reghop3(U8 *s, I32 off, const U8* lim)
6996 /* return the position 'off' UTF-8 characters away from 's', forward if
6997 * 'off' >= 0, backwards if negative. But don't go outside of position
6998 * 'lim', which better be < s if off < 0 */
7002 PERL_ARGS_ASSERT_REGHOP3;
7005 while (off-- && s < lim) {
7006 /* XXX could check well-formedness here */
7011 while (off++ && s > lim) {
7013 if (UTF8_IS_CONTINUED(*s)) {
7014 while (s > lim && UTF8_IS_CONTINUATION(*s))
7017 /* XXX could check well-formedness here */
7024 /* there are a bunch of places where we use two reghop3's that should
7025 be replaced with this routine. but since thats not done yet
7026 we ifdef it out - dmq
7029 S_reghop4(U8 *s, I32 off, const U8* llim, const U8* rlim)
7033 PERL_ARGS_ASSERT_REGHOP4;
7036 while (off-- && s < rlim) {
7037 /* XXX could check well-formedness here */
7042 while (off++ && s > llim) {
7044 if (UTF8_IS_CONTINUED(*s)) {
7045 while (s > llim && UTF8_IS_CONTINUATION(*s))
7048 /* XXX could check well-formedness here */
7056 S_reghopmaybe3(U8* s, I32 off, const U8* lim)
7060 PERL_ARGS_ASSERT_REGHOPMAYBE3;
7063 while (off-- && s < lim) {
7064 /* XXX could check well-formedness here */
7071 while (off++ && s > lim) {
7073 if (UTF8_IS_CONTINUED(*s)) {
7074 while (s > lim && UTF8_IS_CONTINUATION(*s))
7077 /* XXX could check well-formedness here */
7086 restore_pos(pTHX_ void *arg)
7089 regexp * const rex = (regexp *)arg;
7090 if (PL_reg_state.re_state_eval_setup_done) {
7091 if (PL_reg_oldsaved) {
7092 rex->subbeg = PL_reg_oldsaved;
7093 rex->sublen = PL_reg_oldsavedlen;
7094 #ifdef PERL_OLD_COPY_ON_WRITE
7095 rex->saved_copy = PL_nrs;
7097 RXp_MATCH_COPIED_on(rex);
7099 PL_reg_magic->mg_len = PL_reg_oldpos;
7100 PL_reg_state.re_state_eval_setup_done = FALSE;
7101 PL_curpm = PL_reg_oldcurpm;
7106 S_to_utf8_substr(pTHX_ register regexp *prog)
7110 PERL_ARGS_ASSERT_TO_UTF8_SUBSTR;
7113 if (prog->substrs->data[i].substr
7114 && !prog->substrs->data[i].utf8_substr) {
7115 SV* const sv = newSVsv(prog->substrs->data[i].substr);
7116 prog->substrs->data[i].utf8_substr = sv;
7117 sv_utf8_upgrade(sv);
7118 if (SvVALID(prog->substrs->data[i].substr)) {
7119 if (SvTAIL(prog->substrs->data[i].substr)) {
7120 /* Trim the trailing \n that fbm_compile added last
7122 SvCUR_set(sv, SvCUR(sv) - 1);
7123 /* Whilst this makes the SV technically "invalid" (as its
7124 buffer is no longer followed by "\0") when fbm_compile()
7125 adds the "\n" back, a "\0" is restored. */
7126 fbm_compile(sv, FBMcf_TAIL);
7130 if (prog->substrs->data[i].substr == prog->check_substr)
7131 prog->check_utf8 = sv;
7137 S_to_byte_substr(pTHX_ register regexp *prog)
7142 PERL_ARGS_ASSERT_TO_BYTE_SUBSTR;
7145 if (prog->substrs->data[i].utf8_substr
7146 && !prog->substrs->data[i].substr) {
7147 SV* sv = newSVsv(prog->substrs->data[i].utf8_substr);
7148 if (sv_utf8_downgrade(sv, TRUE)) {
7149 if (SvVALID(prog->substrs->data[i].utf8_substr)) {
7150 if (SvTAIL(prog->substrs->data[i].utf8_substr)) {
7151 /* Trim the trailing \n that fbm_compile added last
7153 SvCUR_set(sv, SvCUR(sv) - 1);
7154 fbm_compile(sv, FBMcf_TAIL);
7162 prog->substrs->data[i].substr = sv;
7163 if (prog->substrs->data[i].utf8_substr == prog->check_utf8)
7164 prog->check_substr = sv;
7171 * c-indentation-style: bsd
7173 * indent-tabs-mode: nil
7176 * ex: set ts=8 sts=4 sw=4 et: