5 * "One Ring to rule them all, One Ring to find them..."
8 /* This file contains functions for executing a regular expression. See
9 * also regcomp.c which funnily enough, contains functions for compiling
10 * a regular expression.
12 * This file is also copied at build time to ext/re/re_exec.c, where
13 * it's built with -DPERL_EXT_RE_BUILD -DPERL_EXT_RE_DEBUG -DPERL_EXT.
14 * This causes the main functions to be compiled under new names and with
15 * debugging support added, which makes "use re 'debug'" work.
19 /* NOTE: this is derived from Henry Spencer's regexp code, and should not
20 * confused with the original package (see point 3 below). Thanks, Henry!
23 /* Additional note: this code is very heavily munged from Henry's version
24 * in places. In some spots I've traded clarity for efficiency, so don't
25 * blame Henry for some of the lack of readability.
28 /* The names of the functions have been changed from regcomp and
29 * regexec to pregcomp and pregexec in order to avoid conflicts
30 * with the POSIX routines of the same names.
33 #ifdef PERL_EXT_RE_BUILD
34 /* need to replace pregcomp et al, so enable that */
35 # ifndef PERL_IN_XSUB_RE
36 # define PERL_IN_XSUB_RE
38 /* need access to debugger hooks */
39 # if defined(PERL_EXT_RE_DEBUG) && !defined(DEBUGGING)
44 #ifdef PERL_IN_XSUB_RE
45 /* We *really* need to overwrite these symbols: */
46 # define Perl_regexec_flags my_regexec
47 # define Perl_regdump my_regdump
48 # define Perl_regprop my_regprop
49 # define Perl_re_intuit_start my_re_intuit_start
50 /* *These* symbols are masked to allow static link. */
51 # define Perl_pregexec my_pregexec
52 # define Perl_reginitcolors my_reginitcolors
53 # define Perl_regclass_swash my_regclass_swash
55 # define PERL_NO_GET_CONTEXT
59 * pregcomp and pregexec -- regsub and regerror are not used in perl
61 * Copyright (c) 1986 by University of Toronto.
62 * Written by Henry Spencer. Not derived from licensed software.
64 * Permission is granted to anyone to use this software for any
65 * purpose on any computer system, and to redistribute it freely,
66 * subject to the following restrictions:
68 * 1. The author is not responsible for the consequences of use of
69 * this software, no matter how awful, even if they arise
72 * 2. The origin of this software must not be misrepresented, either
73 * by explicit claim or by omission.
75 * 3. Altered versions must be plainly marked as such, and must not
76 * be misrepresented as being the original software.
78 **** Alterations to Henry's code are...
80 **** Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
81 **** 2000, 2001, 2002, 2003, 2004, 2005, 2006, by Larry Wall and others
83 **** You may distribute under the terms of either the GNU General Public
84 **** License or the Artistic License, as specified in the README file.
86 * Beware that some of this code is subtly aware of the way operator
87 * precedence is structured in regular expressions. Serious changes in
88 * regular-expression syntax might require a total rethink.
91 #define PERL_IN_REGEXEC_C
96 #define RF_tainted 1 /* tainted information used? */
97 #define RF_warned 2 /* warned about big count? */
98 #define RF_evaled 4 /* Did an EVAL with setting? */
99 #define RF_utf8 8 /* String contains multibyte chars? */
101 #define UTF ((PL_reg_flags & RF_utf8) != 0)
103 #define RS_init 1 /* eval environment created */
104 #define RS_set 2 /* replsv value is set */
107 #define STATIC static
110 #define REGINCLASS(prog,p,c) (ANYOF_FLAGS(p) ? reginclass(prog,p,c,0,0) : ANYOF_BITMAP_TEST(p,*(c)))
116 #define CHR_SVLEN(sv) (do_utf8 ? sv_len_utf8(sv) : SvCUR(sv))
117 #define CHR_DIST(a,b) (PL_reg_match_utf8 ? utf8_distance(a,b) : a - b)
119 #define HOPc(pos,off) ((char *)(PL_reg_match_utf8 \
120 ? reghop3((U8*)pos, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr)) \
122 #define HOPBACKc(pos, off) ((char*) \
123 ((PL_reg_match_utf8) \
124 ? reghopmaybe3((U8*)pos, -off, ((U8*)(off < 0 ? PL_regeol : PL_bostr))) \
125 : (pos - off >= PL_bostr) \
130 #define reghopmaybe3_c(pos,off,lim) ((char*)reghopmaybe3((U8*)pos, off, (U8*)lim))
131 #define HOP3(pos,off,lim) (PL_reg_match_utf8 ? reghop3((U8*)pos, off, (U8*)lim) : (U8*)(pos + off))
132 #define HOP3c(pos,off,lim) ((char*)HOP3(pos,off,lim))
134 #define LOAD_UTF8_CHARCLASS(class,str) STMT_START { \
135 if (!CAT2(PL_utf8_,class)) { bool ok; ENTER; save_re_context(); ok=CAT2(is_utf8_,class)((const U8*)str); assert(ok); LEAVE; } } STMT_END
136 #define LOAD_UTF8_CHARCLASS_ALNUM() LOAD_UTF8_CHARCLASS(alnum,"a")
137 #define LOAD_UTF8_CHARCLASS_DIGIT() LOAD_UTF8_CHARCLASS(digit,"0")
138 #define LOAD_UTF8_CHARCLASS_SPACE() LOAD_UTF8_CHARCLASS(space," ")
139 #define LOAD_UTF8_CHARCLASS_MARK() LOAD_UTF8_CHARCLASS(mark, "\xcd\x86")
141 /* for use after a quantifier and before an EXACT-like node -- japhy */
142 #define JUMPABLE(rn) ( \
143 OP(rn) == OPEN || OP(rn) == CLOSE || OP(rn) == EVAL || \
144 OP(rn) == SUSPEND || OP(rn) == IFMATCH || \
145 OP(rn) == PLUS || OP(rn) == MINMOD || \
146 (PL_regkind[(U8)OP(rn)] == CURLY && ARG1(rn) > 0) \
149 #define HAS_TEXT(rn) ( \
150 PL_regkind[(U8)OP(rn)] == EXACT || PL_regkind[(U8)OP(rn)] == REF \
154 Search for mandatory following text node; for lookahead, the text must
155 follow but for lookbehind (rn->flags != 0) we skip to the next step.
157 #define FIND_NEXT_IMPT(rn) STMT_START { \
158 while (JUMPABLE(rn)) \
159 if (OP(rn) == SUSPEND || PL_regkind[(U8)OP(rn)] == CURLY) \
160 rn = NEXTOPER(NEXTOPER(rn)); \
161 else if (OP(rn) == PLUS) \
163 else if (OP(rn) == IFMATCH) \
164 rn = (rn->flags == 0) ? NEXTOPER(NEXTOPER(rn)) : rn + ARG(rn); \
165 else rn += NEXT_OFF(rn); \
168 static void restore_pos(pTHX_ void *arg);
171 S_regcppush(pTHX_ I32 parenfloor)
174 const int retval = PL_savestack_ix;
175 #define REGCP_PAREN_ELEMS 4
176 const int paren_elems_to_push = (PL_regsize - parenfloor) * REGCP_PAREN_ELEMS;
179 if (paren_elems_to_push < 0)
180 Perl_croak(aTHX_ "panic: paren_elems_to_push < 0");
182 #define REGCP_OTHER_ELEMS 6
183 SSGROW(paren_elems_to_push + REGCP_OTHER_ELEMS);
184 for (p = PL_regsize; p > parenfloor; p--) {
185 /* REGCP_PARENS_ELEMS are pushed per pairs of parentheses. */
186 SSPUSHINT(PL_regendp[p]);
187 SSPUSHINT(PL_regstartp[p]);
188 SSPUSHPTR(PL_reg_start_tmp[p]);
191 /* REGCP_OTHER_ELEMS are pushed in any case, parentheses or no. */
192 SSPUSHINT(PL_regsize);
193 SSPUSHINT(*PL_reglastparen);
194 SSPUSHINT(*PL_reglastcloseparen);
195 SSPUSHPTR(PL_reginput);
196 #define REGCP_FRAME_ELEMS 2
197 /* REGCP_FRAME_ELEMS are part of the REGCP_OTHER_ELEMS and
198 * are needed for the regexp context stack bookkeeping. */
199 SSPUSHINT(paren_elems_to_push + REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS);
200 SSPUSHINT(SAVEt_REGCONTEXT); /* Magic cookie. */
205 /* These are needed since we do not localize EVAL nodes: */
206 # define REGCP_SET(cp) DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, \
207 " Setting an EVAL scope, savestack=%"IVdf"\n", \
208 (IV)PL_savestack_ix)); cp = PL_savestack_ix
210 # define REGCP_UNWIND(cp) DEBUG_EXECUTE_r(cp != PL_savestack_ix ? \
211 PerlIO_printf(Perl_debug_log, \
212 " Clearing an EVAL scope, savestack=%"IVdf"..%"IVdf"\n", \
213 (IV)(cp), (IV)PL_savestack_ix) : 0); regcpblow(cp)
216 S_regcppop(pTHX_ const regexp *rex)
222 GET_RE_DEBUG_FLAGS_DECL;
224 /* Pop REGCP_OTHER_ELEMS before the parentheses loop starts. */
226 assert(i == SAVEt_REGCONTEXT); /* Check that the magic cookie is there. */
227 i = SSPOPINT; /* Parentheses elements to pop. */
228 input = (char *) SSPOPPTR;
229 *PL_reglastcloseparen = SSPOPINT;
230 *PL_reglastparen = SSPOPINT;
231 PL_regsize = SSPOPINT;
233 /* Now restore the parentheses context. */
234 for (i -= (REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS);
235 i > 0; i -= REGCP_PAREN_ELEMS) {
237 U32 paren = (U32)SSPOPINT;
238 PL_reg_start_tmp[paren] = (char *) SSPOPPTR;
239 PL_regstartp[paren] = SSPOPINT;
241 if (paren <= *PL_reglastparen)
242 PL_regendp[paren] = tmps;
244 PerlIO_printf(Perl_debug_log,
245 " restoring \\%"UVuf" to %"IVdf"(%"IVdf")..%"IVdf"%s\n",
246 (UV)paren, (IV)PL_regstartp[paren],
247 (IV)(PL_reg_start_tmp[paren] - PL_bostr),
248 (IV)PL_regendp[paren],
249 (paren > *PL_reglastparen ? "(no)" : ""));
253 if ((I32)(*PL_reglastparen + 1) <= rex->nparens) {
254 PerlIO_printf(Perl_debug_log,
255 " restoring \\%"IVdf"..\\%"IVdf" to undef\n",
256 (IV)(*PL_reglastparen + 1), (IV)rex->nparens);
260 /* It would seem that the similar code in regtry()
261 * already takes care of this, and in fact it is in
262 * a better location to since this code can #if 0-ed out
263 * but the code in regtry() is needed or otherwise tests
264 * requiring null fields (pat.t#187 and split.t#{13,14}
265 * (as of patchlevel 7877) will fail. Then again,
266 * this code seems to be necessary or otherwise
267 * building DynaLoader will fail:
268 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
270 for (i = *PL_reglastparen + 1; i <= rex->nparens; i++) {
272 PL_regstartp[i] = -1;
279 #define regcpblow(cp) LEAVE_SCOPE(cp) /* Ignores regcppush()ed data. */
281 #define TRYPAREN(paren, n, input, where) { \
284 PL_regstartp[paren] = HOPc(input, -1) - PL_bostr; \
285 PL_regendp[paren] = input - PL_bostr; \
288 PL_regendp[paren] = -1; \
290 REGMATCH(next, where); \
294 PL_regendp[paren] = -1; \
299 * pregexec and friends
303 - pregexec - match a regexp against a string
306 Perl_pregexec(pTHX_ register regexp *prog, char *stringarg, register char *strend,
307 char *strbeg, I32 minend, SV *screamer, U32 nosave)
308 /* strend: pointer to null at end of string */
309 /* strbeg: real beginning of string */
310 /* minend: end of match must be >=minend after stringarg. */
311 /* nosave: For optimizations. */
314 regexec_flags(prog, stringarg, strend, strbeg, minend, screamer, NULL,
315 nosave ? 0 : REXEC_COPY_STR);
320 * Need to implement the following flags for reg_anch:
322 * USE_INTUIT_NOML - Useful to call re_intuit_start() first
324 * INTUIT_AUTORITATIVE_NOML - Can trust a positive answer
325 * INTUIT_AUTORITATIVE_ML
326 * INTUIT_ONCE_NOML - Intuit can match in one location only.
329 * Another flag for this function: SECOND_TIME (so that float substrs
330 * with giant delta may be not rechecked).
333 /* Assumptions: if ANCH_GPOS, then strpos is anchored. XXXX Check GPOS logic */
335 /* If SCREAM, then SvPVX_const(sv) should be compatible with strpos and strend.
336 Otherwise, only SvCUR(sv) is used to get strbeg. */
338 /* XXXX We assume that strpos is strbeg unless sv. */
340 /* XXXX Some places assume that there is a fixed substring.
341 An update may be needed if optimizer marks as "INTUITable"
342 RExen without fixed substrings. Similarly, it is assumed that
343 lengths of all the strings are no more than minlen, thus they
344 cannot come from lookahead.
345 (Or minlen should take into account lookahead.) */
347 /* A failure to find a constant substring means that there is no need to make
348 an expensive call to REx engine, thus we celebrate a failure. Similarly,
349 finding a substring too deep into the string means that less calls to
350 regtry() should be needed.
352 REx compiler's optimizer found 4 possible hints:
353 a) Anchored substring;
355 c) Whether we are anchored (beginning-of-line or \G);
356 d) First node (of those at offset 0) which may distingush positions;
357 We use a)b)d) and multiline-part of c), and try to find a position in the
358 string which does not contradict any of them.
361 /* Most of decisions we do here should have been done at compile time.
362 The nodes of the REx which we used for the search should have been
363 deleted from the finite automaton. */
366 Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos,
367 char *strend, U32 flags, re_scream_pos_data *data)
370 register I32 start_shift = 0;
371 /* Should be nonnegative! */
372 register I32 end_shift = 0;
377 const int do_utf8 = sv ? SvUTF8(sv) : 0; /* if no sv we have to assume bytes */
379 register char *other_last = NULL; /* other substr checked before this */
380 char *check_at = NULL; /* check substr found at this pos */
381 const I32 multiline = prog->reganch & PMf_MULTILINE;
383 const char * const i_strpos = strpos;
384 SV * const dsv = PERL_DEBUG_PAD_ZERO(0);
387 GET_RE_DEBUG_FLAGS_DECL;
389 RX_MATCH_UTF8_set(prog,do_utf8);
391 if (prog->reganch & ROPT_UTF8) {
392 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
393 "UTF-8 regex...\n"));
394 PL_reg_flags |= RF_utf8;
398 const char *s = PL_reg_match_utf8 ?
399 sv_uni_display(dsv, sv, 60, UNI_DISPLAY_REGEX) :
401 const int len = PL_reg_match_utf8 ?
402 strlen(s) : strend - strpos;
405 if (PL_reg_match_utf8)
406 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
407 "UTF-8 target...\n"));
408 PerlIO_printf(Perl_debug_log,
409 "%sGuessing start of match, REx%s \"%s%.60s%s%s\" against \"%s%.*s%s%s\"...\n",
410 PL_colors[4], PL_colors[5], PL_colors[0],
413 (strlen(prog->precomp) > 60 ? "..." : ""),
415 (int)(len > 60 ? 60 : len),
417 (len > 60 ? "..." : "")
421 /* CHR_DIST() would be more correct here but it makes things slow. */
422 if (prog->minlen > strend - strpos) {
423 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
424 "String too short... [re_intuit_start]\n"));
427 strbeg = (sv && SvPOK(sv)) ? strend - SvCUR(sv) : strpos;
430 if (!prog->check_utf8 && prog->check_substr)
431 to_utf8_substr(prog);
432 check = prog->check_utf8;
434 if (!prog->check_substr && prog->check_utf8)
435 to_byte_substr(prog);
436 check = prog->check_substr;
438 if (check == &PL_sv_undef) {
439 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
440 "Non-utf string cannot match utf check string\n"));
443 if (prog->reganch & ROPT_ANCH) { /* Match at beg-of-str or after \n */
444 ml_anch = !( (prog->reganch & ROPT_ANCH_SINGLE)
445 || ( (prog->reganch & ROPT_ANCH_BOL)
446 && !multiline ) ); /* Check after \n? */
449 if ( !(prog->reganch & (ROPT_ANCH_GPOS /* Checked by the caller */
450 | ROPT_IMPLICIT)) /* not a real BOL */
451 /* SvCUR is not set on references: SvRV and SvPVX_const overlap */
453 && (strpos != strbeg)) {
454 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Not at start...\n"));
457 if (prog->check_offset_min == prog->check_offset_max &&
458 !(prog->reganch & ROPT_CANY_SEEN)) {
459 /* Substring at constant offset from beg-of-str... */
462 s = HOP3c(strpos, prog->check_offset_min, strend);
464 slen = SvCUR(check); /* >= 1 */
466 if ( strend - s > slen || strend - s < slen - 1
467 || (strend - s == slen && strend[-1] != '\n')) {
468 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "String too long...\n"));
471 /* Now should match s[0..slen-2] */
473 if (slen && (*SvPVX_const(check) != *s
475 && memNE(SvPVX_const(check), s, slen)))) {
477 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "String not equal...\n"));
481 else if (*SvPVX_const(check) != *s
482 || ((slen = SvCUR(check)) > 1
483 && memNE(SvPVX_const(check), s, slen)))
486 goto success_at_start;
489 /* Match is anchored, but substr is not anchored wrt beg-of-str. */
491 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
492 end_shift = prog->minlen - start_shift -
493 CHR_SVLEN(check) + (SvTAIL(check) != 0);
495 const I32 end = prog->check_offset_max + CHR_SVLEN(check)
496 - (SvTAIL(check) != 0);
497 const I32 eshift = CHR_DIST((U8*)strend, (U8*)s) - end;
499 if (end_shift < eshift)
503 else { /* Can match at random position */
506 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
507 /* Should be nonnegative! */
508 end_shift = prog->minlen - start_shift -
509 CHR_SVLEN(check) + (SvTAIL(check) != 0);
512 #ifdef DEBUGGING /* 7/99: reports of failure (with the older version) */
514 Perl_croak(aTHX_ "panic: end_shift");
518 /* Find a possible match in the region s..strend by looking for
519 the "check" substring in the region corrected by start/end_shift. */
520 if (flags & REXEC_SCREAM) {
521 I32 p = -1; /* Internal iterator of scream. */
522 I32 * const pp = data ? data->scream_pos : &p;
524 if (PL_screamfirst[BmRARE(check)] >= 0
525 || ( BmRARE(check) == '\n'
526 && (BmPREVIOUS(check) == SvCUR(check) - 1)
528 s = screaminstr(sv, check,
529 start_shift + (s - strbeg), end_shift, pp, 0);
532 /* we may be pointing at the wrong string */
533 if (s && RX_MATCH_COPIED(prog))
534 s = strbeg + (s - SvPVX_const(sv));
536 *data->scream_olds = s;
538 else if (prog->reganch & ROPT_CANY_SEEN)
539 s = fbm_instr((U8*)(s + start_shift),
540 (U8*)(strend - end_shift),
541 check, multiline ? FBMrf_MULTILINE : 0);
543 s = fbm_instr(HOP3(s, start_shift, strend),
544 HOP3(strend, -end_shift, strbeg),
545 check, multiline ? FBMrf_MULTILINE : 0);
547 /* Update the count-of-usability, remove useless subpatterns,
550 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%s %s substr \"%s%.*s%s\"%s%s",
551 (s ? "Found" : "Did not find"),
552 (check == (do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) ? "anchored" : "floating"),
554 (int)(SvCUR(check) - (SvTAIL(check)!=0)),
556 PL_colors[1], (SvTAIL(check) ? "$" : ""),
557 (s ? " at offset " : "...\n") ) );
564 /* Finish the diagnostic message */
565 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%ld...\n", (long)(s - i_strpos)) );
567 /* Got a candidate. Check MBOL anchoring, and the *other* substr.
568 Start with the other substr.
569 XXXX no SCREAM optimization yet - and a very coarse implementation
570 XXXX /ttx+/ results in anchored="ttx", floating="x". floating will
571 *always* match. Probably should be marked during compile...
572 Probably it is right to do no SCREAM here...
575 if (do_utf8 ? (prog->float_utf8 && prog->anchored_utf8) : (prog->float_substr && prog->anchored_substr)) {
576 /* Take into account the "other" substring. */
577 /* XXXX May be hopelessly wrong for UTF... */
580 if (check == (do_utf8 ? prog->float_utf8 : prog->float_substr)) {
583 char * const last = HOP3c(s, -start_shift, strbeg);
588 t = s - prog->check_offset_max;
589 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
591 || ((t = reghopmaybe3_c(s, -(prog->check_offset_max), strpos))
596 t = HOP3c(t, prog->anchored_offset, strend);
597 if (t < other_last) /* These positions already checked */
599 last2 = last1 = HOP3c(strend, -prog->minlen, strbeg);
602 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
603 /* On end-of-str: see comment below. */
604 must = do_utf8 ? prog->anchored_utf8 : prog->anchored_substr;
605 if (must == &PL_sv_undef) {
607 DEBUG_EXECUTE_r(must = prog->anchored_utf8); /* for debug */
612 HOP3(HOP3(last1, prog->anchored_offset, strend)
613 + SvCUR(must), -(SvTAIL(must)!=0), strbeg),
615 multiline ? FBMrf_MULTILINE : 0
617 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
618 "%s anchored substr \"%s%.*s%s\"%s",
619 (s ? "Found" : "Contradicts"),
622 - (SvTAIL(must)!=0)),
624 PL_colors[1], (SvTAIL(must) ? "$" : "")));
626 if (last1 >= last2) {
627 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
628 ", giving up...\n"));
631 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
632 ", trying floating at offset %ld...\n",
633 (long)(HOP3c(s1, 1, strend) - i_strpos)));
634 other_last = HOP3c(last1, prog->anchored_offset+1, strend);
635 s = HOP3c(last, 1, strend);
639 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
640 (long)(s - i_strpos)));
641 t = HOP3c(s, -prog->anchored_offset, strbeg);
642 other_last = HOP3c(s, 1, strend);
650 else { /* Take into account the floating substring. */
655 t = HOP3c(s, -start_shift, strbeg);
657 HOP3c(strend, -prog->minlen + prog->float_min_offset, strbeg);
658 if (CHR_DIST((U8*)last, (U8*)t) > prog->float_max_offset)
659 last = HOP3c(t, prog->float_max_offset, strend);
660 s = HOP3c(t, prog->float_min_offset, strend);
663 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
664 must = do_utf8 ? prog->float_utf8 : prog->float_substr;
665 /* fbm_instr() takes into account exact value of end-of-str
666 if the check is SvTAIL(ed). Since false positives are OK,
667 and end-of-str is not later than strend we are OK. */
668 if (must == &PL_sv_undef) {
670 DEBUG_EXECUTE_r(must = prog->float_utf8); /* for debug message */
673 s = fbm_instr((unsigned char*)s,
674 (unsigned char*)last + SvCUR(must)
676 must, multiline ? FBMrf_MULTILINE : 0);
677 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%s floating substr \"%s%.*s%s\"%s",
678 (s ? "Found" : "Contradicts"),
680 (int)(SvCUR(must) - (SvTAIL(must)!=0)),
682 PL_colors[1], (SvTAIL(must) ? "$" : "")));
685 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
686 ", giving up...\n"));
689 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
690 ", trying anchored starting at offset %ld...\n",
691 (long)(s1 + 1 - i_strpos)));
693 s = HOP3c(t, 1, strend);
697 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
698 (long)(s - i_strpos)));
699 other_last = s; /* Fix this later. --Hugo */
708 t = s - prog->check_offset_max;
709 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
711 || ((t = reghopmaybe3_c(s, -prog->check_offset_max, strpos))
713 /* Fixed substring is found far enough so that the match
714 cannot start at strpos. */
716 if (ml_anch && t[-1] != '\n') {
717 /* Eventually fbm_*() should handle this, but often
718 anchored_offset is not 0, so this check will not be wasted. */
719 /* XXXX In the code below we prefer to look for "^" even in
720 presence of anchored substrings. And we search even
721 beyond the found float position. These pessimizations
722 are historical artefacts only. */
724 while (t < strend - prog->minlen) {
726 if (t < check_at - prog->check_offset_min) {
727 if (do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) {
728 /* Since we moved from the found position,
729 we definitely contradict the found anchored
730 substr. Due to the above check we do not
731 contradict "check" substr.
732 Thus we can arrive here only if check substr
733 is float. Redo checking for "other"=="fixed".
736 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld, rescanning for anchored from offset %ld...\n",
737 PL_colors[0], PL_colors[1], (long)(strpos - i_strpos), (long)(strpos - i_strpos + prog->anchored_offset)));
738 goto do_other_anchored;
740 /* We don't contradict the found floating substring. */
741 /* XXXX Why not check for STCLASS? */
743 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld...\n",
744 PL_colors[0], PL_colors[1], (long)(s - i_strpos)));
747 /* Position contradicts check-string */
748 /* XXXX probably better to look for check-string
749 than for "\n", so one should lower the limit for t? */
750 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m, restarting lookup for check-string at offset %ld...\n",
751 PL_colors[0], PL_colors[1], (long)(t + 1 - i_strpos)));
752 other_last = strpos = s = t + 1;
757 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Did not find /%s^%s/m...\n",
758 PL_colors[0], PL_colors[1]));
762 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Starting position does not contradict /%s^%s/m...\n",
763 PL_colors[0], PL_colors[1]));
767 ++BmUSEFUL(do_utf8 ? prog->check_utf8 : prog->check_substr); /* hooray/5 */
770 /* The found string does not prohibit matching at strpos,
771 - no optimization of calling REx engine can be performed,
772 unless it was an MBOL and we are not after MBOL,
773 or a future STCLASS check will fail this. */
775 /* Even in this situation we may use MBOL flag if strpos is offset
776 wrt the start of the string. */
777 if (ml_anch && sv && !SvROK(sv) /* See prev comment on SvROK */
778 && (strpos != strbeg) && strpos[-1] != '\n'
779 /* May be due to an implicit anchor of m{.*foo} */
780 && !(prog->reganch & ROPT_IMPLICIT))
785 DEBUG_EXECUTE_r( if (ml_anch)
786 PerlIO_printf(Perl_debug_log, "Position at offset %ld does not contradict /%s^%s/m...\n",
787 (long)(strpos - i_strpos), PL_colors[0], PL_colors[1]);
790 if (!(prog->reganch & ROPT_NAUGHTY) /* XXXX If strpos moved? */
792 prog->check_utf8 /* Could be deleted already */
793 && --BmUSEFUL(prog->check_utf8) < 0
794 && (prog->check_utf8 == prog->float_utf8)
796 prog->check_substr /* Could be deleted already */
797 && --BmUSEFUL(prog->check_substr) < 0
798 && (prog->check_substr == prog->float_substr)
801 /* If flags & SOMETHING - do not do it many times on the same match */
802 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "... Disabling check substring...\n"));
803 SvREFCNT_dec(do_utf8 ? prog->check_utf8 : prog->check_substr);
804 if (do_utf8 ? prog->check_substr : prog->check_utf8)
805 SvREFCNT_dec(do_utf8 ? prog->check_substr : prog->check_utf8);
806 prog->check_substr = prog->check_utf8 = NULL; /* disable */
807 prog->float_substr = prog->float_utf8 = NULL; /* clear */
808 check = NULL; /* abort */
810 /* XXXX This is a remnant of the old implementation. It
811 looks wasteful, since now INTUIT can use many
813 prog->reganch &= ~RE_USE_INTUIT;
820 /* XXXX BmUSEFUL already changed, maybe multiple change is meaningful... */
821 if (prog->regstclass) {
822 /* minlen == 0 is possible if regstclass is \b or \B,
823 and the fixed substr is ''$.
824 Since minlen is already taken into account, s+1 is before strend;
825 accidentally, minlen >= 1 guaranties no false positives at s + 1
826 even for \b or \B. But (minlen? 1 : 0) below assumes that
827 regstclass does not come from lookahead... */
828 /* If regstclass takes bytelength more than 1: If charlength==1, OK.
829 This leaves EXACTF only, which is dealt with in find_byclass(). */
830 const U8* const str = (U8*)STRING(prog->regstclass);
831 const int cl_l = (PL_regkind[(U8)OP(prog->regstclass)] == EXACT
832 ? CHR_DIST(str+STR_LEN(prog->regstclass), str)
834 const char * const endpos = (prog->anchored_substr || prog->anchored_utf8 || ml_anch)
835 ? HOP3c(s, (prog->minlen ? cl_l : 0), strend)
836 : (prog->float_substr || prog->float_utf8
837 ? HOP3c(HOP3c(check_at, -start_shift, strbeg),
842 s = find_byclass(prog, prog->regstclass, s, endpos, 1);
845 const char *what = NULL;
847 if (endpos == strend) {
848 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
849 "Could not match STCLASS...\n") );
852 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
853 "This position contradicts STCLASS...\n") );
854 if ((prog->reganch & ROPT_ANCH) && !ml_anch)
856 /* Contradict one of substrings */
857 if (prog->anchored_substr || prog->anchored_utf8) {
858 if ((do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) == check) {
859 DEBUG_EXECUTE_r( what = "anchored" );
861 s = HOP3c(t, 1, strend);
862 if (s + start_shift + end_shift > strend) {
863 /* XXXX Should be taken into account earlier? */
864 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
865 "Could not match STCLASS...\n") );
870 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
871 "Looking for %s substr starting at offset %ld...\n",
872 what, (long)(s + start_shift - i_strpos)) );
875 /* Have both, check_string is floating */
876 if (t + start_shift >= check_at) /* Contradicts floating=check */
877 goto retry_floating_check;
878 /* Recheck anchored substring, but not floating... */
882 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
883 "Looking for anchored substr starting at offset %ld...\n",
884 (long)(other_last - i_strpos)) );
885 goto do_other_anchored;
887 /* Another way we could have checked stclass at the
888 current position only: */
893 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
894 "Looking for /%s^%s/m starting at offset %ld...\n",
895 PL_colors[0], PL_colors[1], (long)(t - i_strpos)) );
898 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr)) /* Could have been deleted */
900 /* Check is floating subtring. */
901 retry_floating_check:
902 t = check_at - start_shift;
903 DEBUG_EXECUTE_r( what = "floating" );
904 goto hop_and_restart;
907 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
908 "By STCLASS: moving %ld --> %ld\n",
909 (long)(t - i_strpos), (long)(s - i_strpos))
913 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
914 "Does not contradict STCLASS...\n");
919 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%s%s:%s match at offset %ld\n",
920 PL_colors[4], (check ? "Guessed" : "Giving up"),
921 PL_colors[5], (long)(s - i_strpos)) );
924 fail_finish: /* Substring not found */
925 if (prog->check_substr || prog->check_utf8) /* could be removed already */
926 BmUSEFUL(do_utf8 ? prog->check_utf8 : prog->check_substr) += 5; /* hooray */
928 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch rejected by optimizer%s\n",
929 PL_colors[4], PL_colors[5]));
933 /* We know what class REx starts with. Try to find this position... */
935 S_find_byclass(pTHX_ regexp * prog, const regnode *c, char *s, const char *strend, I32 norun)
938 const I32 doevery = (prog->reganch & ROPT_SKIP) == 0;
942 register STRLEN uskip;
946 register I32 tmp = 1; /* Scratch variable? */
947 register const bool do_utf8 = PL_reg_match_utf8;
949 /* We know what class it must start with. */
953 while (s + (uskip = UTF8SKIP(s)) <= strend) {
954 if ((ANYOF_FLAGS(c) & ANYOF_UNICODE) ||
955 !UTF8_IS_INVARIANT((U8)s[0]) ?
956 reginclass(prog, c, (U8*)s, 0, do_utf8) :
957 REGINCLASS(prog, c, (U8*)s)) {
958 if (tmp && (norun || regtry(prog, s)))
972 if (REGINCLASS(prog, c, (U8*)s) ||
973 (ANYOF_FOLD_SHARP_S(c, s, strend) &&
974 /* The assignment of 2 is intentional:
975 * for the folded sharp s, the skip is 2. */
976 (skip = SHARP_S_SKIP))) {
977 if (tmp && (norun || regtry(prog, s)))
990 if (tmp && (norun || regtry(prog, s)))
999 ln = STR_LEN(c); /* length to match in octets/bytes */
1000 lnc = (I32) ln; /* length to match in characters */
1002 STRLEN ulen1, ulen2;
1004 U8 tmpbuf1[UTF8_MAXBYTES_CASE+1];
1005 U8 tmpbuf2[UTF8_MAXBYTES_CASE+1];
1006 const U32 uniflags = UTF8_ALLOW_DEFAULT;
1008 to_utf8_lower((U8*)m, tmpbuf1, &ulen1);
1009 to_utf8_upper((U8*)m, tmpbuf2, &ulen2);
1011 c1 = utf8n_to_uvchr(tmpbuf1, UTF8_MAXBYTES_CASE,
1013 c2 = utf8n_to_uvchr(tmpbuf2, UTF8_MAXBYTES_CASE,
1016 while (sm < ((U8 *) m + ln)) {
1031 c2 = PL_fold_locale[c1];
1033 e = HOP3c(strend, -((I32)lnc), s);
1036 e = s; /* Due to minlen logic of intuit() */
1038 /* The idea in the EXACTF* cases is to first find the
1039 * first character of the EXACTF* node and then, if
1040 * necessary, case-insensitively compare the full
1041 * text of the node. The c1 and c2 are the first
1042 * characters (though in Unicode it gets a bit
1043 * more complicated because there are more cases
1044 * than just upper and lower: one needs to use
1045 * the so-called folding case for case-insensitive
1046 * matching (called "loose matching" in Unicode).
1047 * ibcmp_utf8() will do just that. */
1051 U8 tmpbuf [UTF8_MAXBYTES+1];
1052 STRLEN len, foldlen;
1053 const U32 uniflags = UTF8_ALLOW_DEFAULT;
1055 /* Upper and lower of 1st char are equal -
1056 * probably not a "letter". */
1058 c = utf8n_to_uvchr((U8*)s, UTF8_MAXBYTES, &len,
1062 ibcmp_utf8(s, (char **)0, 0, do_utf8,
1063 m, (char **)0, ln, (bool)UTF))
1064 && (norun || regtry(prog, s)) )
1067 U8 foldbuf[UTF8_MAXBYTES_CASE+1];
1068 uvchr_to_utf8(tmpbuf, c);
1069 f = to_utf8_fold(tmpbuf, foldbuf, &foldlen);
1071 && (f == c1 || f == c2)
1072 && (ln == foldlen ||
1073 !ibcmp_utf8((char *) foldbuf,
1074 (char **)0, foldlen, do_utf8,
1076 (char **)0, ln, (bool)UTF))
1077 && (norun || regtry(prog, s)) )
1085 c = utf8n_to_uvchr((U8*)s, UTF8_MAXBYTES, &len,
1088 /* Handle some of the three Greek sigmas cases.
1089 * Note that not all the possible combinations
1090 * are handled here: some of them are handled
1091 * by the standard folding rules, and some of
1092 * them (the character class or ANYOF cases)
1093 * are handled during compiletime in
1094 * regexec.c:S_regclass(). */
1095 if (c == (UV)UNICODE_GREEK_CAPITAL_LETTER_SIGMA ||
1096 c == (UV)UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA)
1097 c = (UV)UNICODE_GREEK_SMALL_LETTER_SIGMA;
1099 if ( (c == c1 || c == c2)
1101 ibcmp_utf8(s, (char **)0, 0, do_utf8,
1102 m, (char **)0, ln, (bool)UTF))
1103 && (norun || regtry(prog, s)) )
1106 U8 foldbuf[UTF8_MAXBYTES_CASE+1];
1107 uvchr_to_utf8(tmpbuf, c);
1108 f = to_utf8_fold(tmpbuf, foldbuf, &foldlen);
1110 && (f == c1 || f == c2)
1111 && (ln == foldlen ||
1112 !ibcmp_utf8((char *) foldbuf,
1113 (char **)0, foldlen, do_utf8,
1115 (char **)0, ln, (bool)UTF))
1116 && (norun || regtry(prog, s)) )
1127 && (ln == 1 || !(OP(c) == EXACTF
1129 : ibcmp_locale(s, m, ln)))
1130 && (norun || regtry(prog, s)) )
1136 if ( (*(U8*)s == c1 || *(U8*)s == c2)
1137 && (ln == 1 || !(OP(c) == EXACTF
1139 : ibcmp_locale(s, m, ln)))
1140 && (norun || regtry(prog, s)) )
1147 PL_reg_flags |= RF_tainted;
1154 U8 * const r = reghop3((U8*)s, -1, (U8*)PL_bostr);
1155 tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, UTF8_ALLOW_DEFAULT);
1157 tmp = ((OP(c) == BOUND ?
1158 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1159 LOAD_UTF8_CHARCLASS_ALNUM();
1160 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1161 if (tmp == !(OP(c) == BOUND ?
1162 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1163 isALNUM_LC_utf8((U8*)s)))
1166 if ((norun || regtry(prog, s)))
1173 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1174 tmp = ((OP(c) == BOUND ? isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1175 while (s < strend) {
1177 !(OP(c) == BOUND ? isALNUM(*s) : isALNUM_LC(*s))) {
1179 if ((norun || regtry(prog, s)))
1185 if ((!prog->minlen && tmp) && (norun || regtry(prog, s)))
1189 PL_reg_flags |= RF_tainted;
1196 U8 * const r = reghop3((U8*)s, -1, (U8*)PL_bostr);
1197 tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, UTF8_ALLOW_DEFAULT);
1199 tmp = ((OP(c) == NBOUND ?
1200 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1201 LOAD_UTF8_CHARCLASS_ALNUM();
1202 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1203 if (tmp == !(OP(c) == NBOUND ?
1204 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1205 isALNUM_LC_utf8((U8*)s)))
1207 else if ((norun || regtry(prog, s)))
1213 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1214 tmp = ((OP(c) == NBOUND ?
1215 isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1216 while (s < strend) {
1218 !(OP(c) == NBOUND ? isALNUM(*s) : isALNUM_LC(*s)))
1220 else if ((norun || regtry(prog, s)))
1225 if ((!prog->minlen && !tmp) && (norun || regtry(prog, s)))
1230 LOAD_UTF8_CHARCLASS_ALNUM();
1231 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1232 if (swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1233 if (tmp && (norun || regtry(prog, s)))
1244 while (s < strend) {
1246 if (tmp && (norun || regtry(prog, s)))
1258 PL_reg_flags |= RF_tainted;
1260 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1261 if (isALNUM_LC_utf8((U8*)s)) {
1262 if (tmp && (norun || regtry(prog, s)))
1273 while (s < strend) {
1274 if (isALNUM_LC(*s)) {
1275 if (tmp && (norun || regtry(prog, s)))
1288 LOAD_UTF8_CHARCLASS_ALNUM();
1289 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1290 if (!swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1291 if (tmp && (norun || regtry(prog, s)))
1302 while (s < strend) {
1304 if (tmp && (norun || regtry(prog, s)))
1316 PL_reg_flags |= RF_tainted;
1318 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1319 if (!isALNUM_LC_utf8((U8*)s)) {
1320 if (tmp && (norun || regtry(prog, s)))
1331 while (s < strend) {
1332 if (!isALNUM_LC(*s)) {
1333 if (tmp && (norun || regtry(prog, s)))
1346 LOAD_UTF8_CHARCLASS_SPACE();
1347 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1348 if (*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8)) {
1349 if (tmp && (norun || regtry(prog, s)))
1360 while (s < strend) {
1362 if (tmp && (norun || regtry(prog, s)))
1374 PL_reg_flags |= RF_tainted;
1376 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1377 if (*s == ' ' || isSPACE_LC_utf8((U8*)s)) {
1378 if (tmp && (norun || regtry(prog, s)))
1389 while (s < strend) {
1390 if (isSPACE_LC(*s)) {
1391 if (tmp && (norun || regtry(prog, s)))
1404 LOAD_UTF8_CHARCLASS_SPACE();
1405 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1406 if (!(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8))) {
1407 if (tmp && (norun || regtry(prog, s)))
1418 while (s < strend) {
1420 if (tmp && (norun || regtry(prog, s)))
1432 PL_reg_flags |= RF_tainted;
1434 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1435 if (!(*s == ' ' || isSPACE_LC_utf8((U8*)s))) {
1436 if (tmp && (norun || regtry(prog, s)))
1447 while (s < strend) {
1448 if (!isSPACE_LC(*s)) {
1449 if (tmp && (norun || regtry(prog, s)))
1462 LOAD_UTF8_CHARCLASS_DIGIT();
1463 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1464 if (swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1465 if (tmp && (norun || regtry(prog, s)))
1476 while (s < strend) {
1478 if (tmp && (norun || regtry(prog, s)))
1490 PL_reg_flags |= RF_tainted;
1492 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1493 if (isDIGIT_LC_utf8((U8*)s)) {
1494 if (tmp && (norun || regtry(prog, s)))
1505 while (s < strend) {
1506 if (isDIGIT_LC(*s)) {
1507 if (tmp && (norun || regtry(prog, s)))
1520 LOAD_UTF8_CHARCLASS_DIGIT();
1521 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1522 if (!swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1523 if (tmp && (norun || regtry(prog, s)))
1534 while (s < strend) {
1536 if (tmp && (norun || regtry(prog, s)))
1548 PL_reg_flags |= RF_tainted;
1550 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1551 if (!isDIGIT_LC_utf8((U8*)s)) {
1552 if (tmp && (norun || regtry(prog, s)))
1563 while (s < strend) {
1564 if (!isDIGIT_LC(*s)) {
1565 if (tmp && (norun || regtry(prog, s)))
1577 Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c));
1586 - regexec_flags - match a regexp against a string
1589 Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char *strend,
1590 char *strbeg, I32 minend, SV *sv, void *data, U32 flags)
1591 /* strend: pointer to null at end of string */
1592 /* strbeg: real beginning of string */
1593 /* minend: end of match must be >=minend after stringarg. */
1594 /* data: May be used for some additional optimizations. */
1595 /* nosave: For optimizations. */
1599 register regnode *c;
1600 register char *startpos = stringarg;
1601 I32 minlen; /* must match at least this many chars */
1602 I32 dontbother = 0; /* how many characters not to try at end */
1603 I32 end_shift = 0; /* Same for the end. */ /* CC */
1604 I32 scream_pos = -1; /* Internal iterator of scream. */
1605 char *scream_olds = NULL;
1606 SV* oreplsv = GvSV(PL_replgv);
1607 const bool do_utf8 = DO_UTF8(sv);
1608 const I32 multiline = prog->reganch & PMf_MULTILINE;
1610 SV * const dsv0 = PERL_DEBUG_PAD_ZERO(0);
1611 SV * const dsv1 = PERL_DEBUG_PAD_ZERO(1);
1614 GET_RE_DEBUG_FLAGS_DECL;
1616 PERL_UNUSED_ARG(data);
1618 /* Be paranoid... */
1619 if (prog == NULL || startpos == NULL) {
1620 Perl_croak(aTHX_ "NULL regexp parameter");
1625 PL_regnarrate = DEBUG_r_TEST;
1628 RX_MATCH_UTF8_set(prog, do_utf8);
1630 minlen = prog->minlen;
1631 if (strend - startpos < minlen) {
1632 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
1633 "String too short [regexec_flags]...\n"));
1637 /* Check validity of program. */
1638 if (UCHARAT(prog->program) != REG_MAGIC) {
1639 Perl_croak(aTHX_ "corrupted regexp program");
1643 PL_reg_eval_set = 0;
1646 if (prog->reganch & ROPT_UTF8)
1647 PL_reg_flags |= RF_utf8;
1649 /* Mark beginning of line for ^ and lookbehind. */
1650 PL_regbol = startpos;
1654 /* Mark end of line for $ (and such) */
1657 /* see how far we have to get to not match where we matched before */
1658 PL_regtill = startpos+minend;
1660 /* If there is a "must appear" string, look for it. */
1663 if (prog->reganch & ROPT_GPOS_SEEN) { /* Need to have PL_reg_ganch */
1666 if (flags & REXEC_IGNOREPOS) /* Means: check only at start */
1667 PL_reg_ganch = startpos;
1668 else if (sv && SvTYPE(sv) >= SVt_PVMG
1670 && (mg = mg_find(sv, PERL_MAGIC_regex_global))
1671 && mg->mg_len >= 0) {
1672 PL_reg_ganch = strbeg + mg->mg_len; /* Defined pos() */
1673 if (prog->reganch & ROPT_ANCH_GPOS) {
1674 if (s > PL_reg_ganch)
1679 else /* pos() not defined */
1680 PL_reg_ganch = strbeg;
1683 if (!(flags & REXEC_CHECKED) && (prog->check_substr != NULL || prog->check_utf8 != NULL)) {
1684 re_scream_pos_data d;
1686 d.scream_olds = &scream_olds;
1687 d.scream_pos = &scream_pos;
1688 s = re_intuit_start(prog, sv, s, strend, flags, &d);
1690 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Not present...\n"));
1691 goto phooey; /* not present */
1696 const char * const s0 = UTF
1697 ? pv_uni_display(dsv0, (U8*)prog->precomp, prog->prelen, 60,
1700 const int len0 = UTF ? SvCUR(dsv0) : prog->prelen;
1701 const char * const s1 = do_utf8 ? sv_uni_display(dsv1, sv, 60,
1702 UNI_DISPLAY_REGEX) : startpos;
1703 const int len1 = do_utf8 ? SvCUR(dsv1) : strend - startpos;
1706 PerlIO_printf(Perl_debug_log,
1707 "%sMatching REx%s \"%s%*.*s%s%s\" against \"%s%.*s%s%s\"\n",
1708 PL_colors[4], PL_colors[5], PL_colors[0],
1711 len0 > 60 ? "..." : "",
1713 (int)(len1 > 60 ? 60 : len1),
1715 (len1 > 60 ? "..." : "")
1719 /* Simplest case: anchored match need be tried only once. */
1720 /* [unless only anchor is BOL and multiline is set] */
1721 if (prog->reganch & (ROPT_ANCH & ~ROPT_ANCH_GPOS)) {
1722 if (s == startpos && regtry(prog, startpos))
1724 else if (multiline || (prog->reganch & ROPT_IMPLICIT)
1725 || (prog->reganch & ROPT_ANCH_MBOL)) /* XXXX SBOL? */
1730 dontbother = minlen - 1;
1731 end = HOP3c(strend, -dontbother, strbeg) - 1;
1732 /* for multiline we only have to try after newlines */
1733 if (prog->check_substr || prog->check_utf8) {
1737 if (regtry(prog, s))
1742 if (prog->reganch & RE_USE_INTUIT) {
1743 s = re_intuit_start(prog, sv, s + 1, strend, flags, NULL);
1754 if (*s++ == '\n') { /* don't need PL_utf8skip here */
1755 if (regtry(prog, s))
1762 } else if (prog->reganch & ROPT_ANCH_GPOS) {
1763 if (regtry(prog, PL_reg_ganch))
1768 /* Messy cases: unanchored match. */
1769 if ((prog->anchored_substr || prog->anchored_utf8) && prog->reganch & ROPT_SKIP) {
1770 /* we have /x+whatever/ */
1771 /* it must be a one character string (XXXX Except UTF?) */
1776 if (!(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr))
1777 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1778 ch = SvPVX_const(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr)[0];
1781 while (s < strend) {
1783 DEBUG_EXECUTE_r( did_match = 1 );
1784 if (regtry(prog, s)) goto got_it;
1786 while (s < strend && *s == ch)
1793 while (s < strend) {
1795 DEBUG_EXECUTE_r( did_match = 1 );
1796 if (regtry(prog, s)) goto got_it;
1798 while (s < strend && *s == ch)
1804 DEBUG_EXECUTE_r(if (!did_match)
1805 PerlIO_printf(Perl_debug_log,
1806 "Did not find anchored character...\n")
1809 else if (prog->anchored_substr != NULL
1810 || prog->anchored_utf8 != NULL
1811 || ((prog->float_substr != NULL || prog->float_utf8 != NULL)
1812 && prog->float_max_offset < strend - s)) {
1817 char *last1; /* Last position checked before */
1821 if (prog->anchored_substr || prog->anchored_utf8) {
1822 if (!(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr))
1823 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1824 must = do_utf8 ? prog->anchored_utf8 : prog->anchored_substr;
1825 back_max = back_min = prog->anchored_offset;
1827 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr))
1828 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1829 must = do_utf8 ? prog->float_utf8 : prog->float_substr;
1830 back_max = prog->float_max_offset;
1831 back_min = prog->float_min_offset;
1833 if (must == &PL_sv_undef)
1834 /* could not downgrade utf8 check substring, so must fail */
1837 last = HOP3c(strend, /* Cannot start after this */
1838 -(I32)(CHR_SVLEN(must)
1839 - (SvTAIL(must) != 0) + back_min), strbeg);
1842 last1 = HOPc(s, -1);
1844 last1 = s - 1; /* bogus */
1846 /* XXXX check_substr already used to find "s", can optimize if
1847 check_substr==must. */
1849 dontbother = end_shift;
1850 strend = HOPc(strend, -dontbother);
1851 while ( (s <= last) &&
1852 ((flags & REXEC_SCREAM)
1853 ? (s = screaminstr(sv, must, HOP3c(s, back_min, strend) - strbeg,
1854 end_shift, &scream_pos, 0))
1855 : (s = fbm_instr((unsigned char*)HOP3(s, back_min, strend),
1856 (unsigned char*)strend, must,
1857 multiline ? FBMrf_MULTILINE : 0))) ) {
1858 /* we may be pointing at the wrong string */
1859 if ((flags & REXEC_SCREAM) && RX_MATCH_COPIED(prog))
1860 s = strbeg + (s - SvPVX_const(sv));
1861 DEBUG_EXECUTE_r( did_match = 1 );
1862 if (HOPc(s, -back_max) > last1) {
1863 last1 = HOPc(s, -back_min);
1864 s = HOPc(s, -back_max);
1867 char * const t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1;
1869 last1 = HOPc(s, -back_min);
1873 while (s <= last1) {
1874 if (regtry(prog, s))
1880 while (s <= last1) {
1881 if (regtry(prog, s))
1887 DEBUG_EXECUTE_r(if (!did_match)
1888 PerlIO_printf(Perl_debug_log,
1889 "Did not find %s substr \"%s%.*s%s\"%s...\n",
1890 ((must == prog->anchored_substr || must == prog->anchored_utf8)
1891 ? "anchored" : "floating"),
1893 (int)(SvCUR(must) - (SvTAIL(must)!=0)),
1895 PL_colors[1], (SvTAIL(must) ? "$" : ""))
1899 else if ((c = prog->regstclass)) {
1901 I32 op = (U8)OP(prog->regstclass);
1902 /* don't bother with what can't match */
1903 if (PL_regkind[op] != EXACT && op != CANY)
1904 strend = HOPc(strend, -(minlen - 1));
1907 SV *prop = sv_newmortal();
1913 regprop(prog, prop, c);
1915 pv_uni_display(dsv0, (U8*)SvPVX_const(prop), SvCUR(prop), 60,
1916 UNI_DISPLAY_REGEX) :
1918 len0 = UTF ? SvCUR(dsv0) : SvCUR(prop);
1920 sv_uni_display(dsv1, sv, 60, UNI_DISPLAY_REGEX) : s;
1921 len1 = UTF ? SvCUR(dsv1) : strend - s;
1922 PerlIO_printf(Perl_debug_log,
1923 "Matching stclass \"%*.*s\" against \"%*.*s\"\n",
1927 if (find_byclass(prog, c, s, strend, 0))
1929 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass...\n"));
1933 if (prog->float_substr != NULL || prog->float_utf8 != NULL) {
1938 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr))
1939 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1940 float_real = do_utf8 ? prog->float_utf8 : prog->float_substr;
1942 if (flags & REXEC_SCREAM) {
1943 last = screaminstr(sv, float_real, s - strbeg,
1944 end_shift, &scream_pos, 1); /* last one */
1946 last = scream_olds; /* Only one occurrence. */
1947 /* we may be pointing at the wrong string */
1948 else if (RX_MATCH_COPIED(prog))
1949 s = strbeg + (s - SvPVX_const(sv));
1953 const char * const little = SvPV_const(float_real, len);
1955 if (SvTAIL(float_real)) {
1956 if (memEQ(strend - len + 1, little, len - 1))
1957 last = strend - len + 1;
1958 else if (!multiline)
1959 last = memEQ(strend - len, little, len)
1960 ? strend - len : NULL;
1966 last = rninstr(s, strend, little, little + len);
1968 last = strend; /* matching "$" */
1972 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
1973 "%sCan't trim the tail, match fails (should not happen)%s\n",
1974 PL_colors[4], PL_colors[5]));
1975 goto phooey; /* Should not happen! */
1977 dontbother = strend - last + prog->float_min_offset;
1979 if (minlen && (dontbother < minlen))
1980 dontbother = minlen - 1;
1981 strend -= dontbother; /* this one's always in bytes! */
1982 /* We don't know much -- general case. */
1985 if (regtry(prog, s))
1994 if (regtry(prog, s))
1996 } while (s++ < strend);
2004 RX_MATCH_TAINTED_set(prog, PL_reg_flags & RF_tainted);
2006 if (PL_reg_eval_set) {
2007 /* Preserve the current value of $^R */
2008 if (oreplsv != GvSV(PL_replgv))
2009 sv_setsv(oreplsv, GvSV(PL_replgv));/* So that when GvSV(replgv) is
2010 restored, the value remains
2012 restore_pos(aTHX_ prog);
2015 /* make sure $`, $&, $', and $digit will work later */
2016 if ( !(flags & REXEC_NOT_FIRST) ) {
2017 RX_MATCH_COPY_FREE(prog);
2018 if (flags & REXEC_COPY_STR) {
2019 I32 i = PL_regeol - startpos + (stringarg - strbeg);
2020 #ifdef PERL_OLD_COPY_ON_WRITE
2022 || (SvFLAGS(sv) & CAN_COW_MASK) == CAN_COW_FLAGS)) {
2024 PerlIO_printf(Perl_debug_log,
2025 "Copy on write: regexp capture, type %d\n",
2028 prog->saved_copy = sv_setsv_cow(prog->saved_copy, sv);
2029 prog->subbeg = (char *)SvPVX_const(prog->saved_copy);
2030 assert (SvPOKp(prog->saved_copy));
2034 RX_MATCH_COPIED_on(prog);
2035 s = savepvn(strbeg, i);
2041 prog->subbeg = strbeg;
2042 prog->sublen = PL_regeol - strbeg; /* strend may have been modified */
2049 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n",
2050 PL_colors[4], PL_colors[5]));
2051 if (PL_reg_eval_set)
2052 restore_pos(aTHX_ prog);
2057 - regtry - try match at specific point
2059 STATIC I32 /* 0 failure, 1 success */
2060 S_regtry(pTHX_ regexp *prog, char *startpos)
2066 GET_RE_DEBUG_FLAGS_DECL;
2069 PL_regindent = 0; /* XXXX Not good when matches are reenterable... */
2071 if ((prog->reganch & ROPT_EVAL_SEEN) && !PL_reg_eval_set) {
2074 PL_reg_eval_set = RS_init;
2075 DEBUG_EXECUTE_r(DEBUG_s(
2076 PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %"IVdf"\n",
2077 (IV)(PL_stack_sp - PL_stack_base));
2079 SAVEI32(cxstack[cxstack_ix].blk_oldsp);
2080 cxstack[cxstack_ix].blk_oldsp = PL_stack_sp - PL_stack_base;
2081 /* Otherwise OP_NEXTSTATE will free whatever on stack now. */
2083 /* Apparently this is not needed, judging by wantarray. */
2084 /* SAVEI8(cxstack[cxstack_ix].blk_gimme);
2085 cxstack[cxstack_ix].blk_gimme = G_SCALAR; */
2088 /* Make $_ available to executed code. */
2089 if (PL_reg_sv != DEFSV) {
2094 if (!(SvTYPE(PL_reg_sv) >= SVt_PVMG && SvMAGIC(PL_reg_sv)
2095 && (mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global)))) {
2096 /* prepare for quick setting of pos */
2097 #ifdef PERL_OLD_COPY_ON_WRITE
2099 sv_force_normal_flags(sv, 0);
2101 mg = sv_magicext(PL_reg_sv, (SV*)0, PERL_MAGIC_regex_global,
2102 &PL_vtbl_mglob, NULL, 0);
2106 PL_reg_oldpos = mg->mg_len;
2107 SAVEDESTRUCTOR_X(restore_pos, prog);
2109 if (!PL_reg_curpm) {
2110 Newxz(PL_reg_curpm, 1, PMOP);
2113 SV* repointer = newSViv(0);
2114 /* so we know which PL_regex_padav element is PL_reg_curpm */
2115 SvFLAGS(repointer) |= SVf_BREAK;
2116 av_push(PL_regex_padav,repointer);
2117 PL_reg_curpm->op_pmoffset = av_len(PL_regex_padav);
2118 PL_regex_pad = AvARRAY(PL_regex_padav);
2122 PM_SETRE(PL_reg_curpm, prog);
2123 PL_reg_oldcurpm = PL_curpm;
2124 PL_curpm = PL_reg_curpm;
2125 if (RX_MATCH_COPIED(prog)) {
2126 /* Here is a serious problem: we cannot rewrite subbeg,
2127 since it may be needed if this match fails. Thus
2128 $` inside (?{}) could fail... */
2129 PL_reg_oldsaved = prog->subbeg;
2130 PL_reg_oldsavedlen = prog->sublen;
2131 #ifdef PERL_OLD_COPY_ON_WRITE
2132 PL_nrs = prog->saved_copy;
2134 RX_MATCH_COPIED_off(prog);
2137 PL_reg_oldsaved = NULL;
2138 prog->subbeg = PL_bostr;
2139 prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */
2141 prog->startp[0] = startpos - PL_bostr;
2142 PL_reginput = startpos;
2143 PL_regstartp = prog->startp;
2144 PL_regendp = prog->endp;
2145 PL_reglastparen = &prog->lastparen;
2146 PL_reglastcloseparen = &prog->lastcloseparen;
2147 prog->lastparen = 0;
2148 prog->lastcloseparen = 0;
2150 DEBUG_EXECUTE_r(PL_reg_starttry = startpos);
2151 if (PL_reg_start_tmpl <= prog->nparens) {
2152 PL_reg_start_tmpl = prog->nparens*3/2 + 3;
2153 if(PL_reg_start_tmp)
2154 Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
2156 Newx(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
2159 /* XXXX What this code is doing here?!!! There should be no need
2160 to do this again and again, PL_reglastparen should take care of
2163 /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code.
2164 * Actually, the code in regcppop() (which Ilya may be meaning by
2165 * PL_reglastparen), is not needed at all by the test suite
2166 * (op/regexp, op/pat, op/split), but that code is needed, oddly
2167 * enough, for building DynaLoader, or otherwise this
2168 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
2169 * will happen. Meanwhile, this code *is* needed for the
2170 * above-mentioned test suite tests to succeed. The common theme
2171 * on those tests seems to be returning null fields from matches.
2176 if (prog->nparens) {
2178 for (i = prog->nparens; i > (I32)*PL_reglastparen; i--) {
2185 if (regmatch(prog, prog->program + 1)) {
2186 prog->endp[0] = PL_reginput - PL_bostr;
2189 REGCP_UNWIND(lastcp);
2193 #define RE_UNWIND_BRANCH 1
2194 #define RE_UNWIND_BRANCHJ 2
2198 typedef struct { /* XX: makes sense to enlarge it... */
2202 } re_unwind_generic_t;
2216 } re_unwind_branch_t;
2218 typedef union re_unwind_t {
2220 re_unwind_generic_t generic;
2221 re_unwind_branch_t branch;
2224 #define sayYES goto yes
2225 #define sayNO goto no
2226 #define sayNO_ANYOF goto no_anyof
2227 #define sayYES_FINAL goto yes_final
2228 #define sayYES_LOUD goto yes_loud
2229 #define sayNO_FINAL goto no_final
2230 #define sayNO_SILENT goto do_no
2231 #define saySAME(x) if (x) goto yes; else goto no
2233 #define POSCACHE_SUCCESS 0 /* caching success rather than failure */
2234 #define POSCACHE_SEEN 1 /* we know what we're caching */
2235 #define POSCACHE_START 2 /* the real cache: this bit maps to pos 0 */
2236 #define CACHEsayYES STMT_START { \
2237 if (st->u.whilem.cache_offset | st->u.whilem.cache_bit) { \
2238 if (!(PL_reg_poscache[0] & (1<<POSCACHE_SEEN))) \
2239 PL_reg_poscache[0] |= (1<<POSCACHE_SUCCESS) || (1<<POSCACHE_SEEN); \
2240 else if (!(PL_reg_poscache[0] & (1<<POSCACHE_SUCCESS))) { \
2241 /* cache records failure, but this is success */ \
2243 PerlIO_printf(Perl_debug_log, \
2244 "%*s (remove success from failure cache)\n", \
2245 REPORT_CODE_OFF+PL_regindent*2, "") \
2247 PL_reg_poscache[st->u.whilem.cache_offset] &= ~(1<<st->u.whilem.cache_bit); \
2252 #define CACHEsayNO STMT_START { \
2253 if (st->u.whilem.cache_offset | st->u.whilem.cache_bit) { \
2254 if (!(PL_reg_poscache[0] & (1<<POSCACHE_SEEN))) \
2255 PL_reg_poscache[0] |= (1<<POSCACHE_SEEN); \
2256 else if ((PL_reg_poscache[0] & (1<<POSCACHE_SUCCESS))) { \
2257 /* cache records success, but this is failure */ \
2259 PerlIO_printf(Perl_debug_log, \
2260 "%*s (remove failure from success cache)\n", \
2261 REPORT_CODE_OFF+PL_regindent*2, "") \
2263 PL_reg_poscache[st->u.whilem.cache_offset] &= ~(1<<st->u.whilem.cache_bit); \
2269 /* this is used to determine how far from the left messages like
2270 'failed...' are printed. Currently 29 makes these messages line
2271 up with the opcode they refer to. Earlier perls used 25 which
2272 left these messages outdented making reviewing a debug output
2275 #define REPORT_CODE_OFF 29
2278 /* Make sure there is a test for this +1 options in re_tests */
2279 #define TRIE_INITAL_ACCEPT_BUFFLEN 4;
2281 /* grab a new slab and return the first slot in it */
2283 STATIC regmatch_state *
2286 regmatch_slab *s = PL_regmatch_slab->next;
2288 Newx(s, 1, regmatch_slab);
2289 s->prev = PL_regmatch_slab;
2291 PL_regmatch_slab->next = s;
2293 PL_regmatch_slab = s;
2294 return &s->states[0];
2297 /* simulate a recursive call to regmatch */
2299 #define REGMATCH(ns, where) \
2302 st->resume_state = resume_##where; \
2303 goto start_recurse; \
2304 resume_point_##where:
2307 /* push a new regex state. Set newst to point to it */
2309 #define PUSH_STATE(newst, resume) \
2311 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "PUSH STATE(%d)\n", depth)); \
2315 st->locinput = locinput; \
2316 st->resume_state = resume; \
2318 if (newst > &(PL_regmatch_slab->states[PERL_REGMATCH_SLAB_SLOTS-1])) \
2319 newst = S_push_slab(aTHX); \
2320 PL_regmatch_state = newst; \
2322 newst->minmod = 0; \
2324 newst->logical = 0; \
2325 newst->unwind = 0; \
2326 locinput = PL_reginput; \
2327 nextchr = UCHARAT(locinput);
2330 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "POP STATE(%d)\n", depth)); \
2333 if (st < &PL_regmatch_slab->states[0]) { \
2334 PL_regmatch_slab = PL_regmatch_slab->prev; \
2335 st = &PL_regmatch_slab->states[PERL_REGMATCH_SLAB_SLOTS-1]; \
2337 PL_regmatch_state = st; \
2341 locinput = st->locinput; \
2342 nextchr = UCHARAT(locinput);
2345 - regmatch - main matching routine
2347 * Conceptually the strategy is simple: check to see whether the current
2348 * node matches, call self recursively to see whether the rest matches,
2349 * and then act accordingly. In practice we make some effort to avoid
2350 * recursion, in particular by going through "ordinary" nodes (that don't
2351 * need to know whether the rest of the match failed) by a loop instead of
2354 /* [lwall] I've hoisted the register declarations to the outer block in order to
2355 * maybe save a little bit of pushing and popping on the stack. It also takes
2356 * advantage of machines that use a register save mask on subroutine entry.
2358 * This function used to be heavily recursive, but since this had the
2359 * effect of blowing the CPU stack on complex regexes, it has been
2360 * restructured to be iterative, and to save state onto the heap rather
2361 * than the stack. Essentially whereever regmatch() used to be called, it
2362 * pushes the current state, notes where to return, then jumps back into
2365 * Originally the structure of this function used to look something like
2370 while (scan != NULL) {
2371 a++; // do stuff with a and b
2377 if (regmatch(...)) // recurse
2387 * Now it looks something like this:
2395 regmatch_state *st = new();
2397 st->a++; // do stuff with a and b
2399 while (scan != NULL) {
2407 st->resume_state = resume_FOO;
2408 goto start_recurse; // recurse
2417 st = new(); push a new state
2418 st->a = 1; st->b = 2;
2425 switch (resume_state) {
2427 goto resume_point_FOO;
2434 * WARNING: this means that any line in this function that contains a
2435 * REGMATCH() or TRYPAREN() is actually simulating a recursive call to
2436 * regmatch() using gotos instead. Thus the values of any local variables
2437 * not saved in the regmatch_state structure will have been lost when
2438 * execution resumes on the next line .
2440 * States (ie the st pointer) are allocated in slabs of about 4K in size.
2441 * PL_regmatch_state always points to the currently active state, and
2442 * PL_regmatch_slab points to the slab currently containing PL_regmatch_state.
2443 * The first time regmatch is called, the first slab is allocated, and is
2444 * never freed until interpreter desctruction. When the slab is full,
2445 * a new one is allocated chained to the end. At exit from regmatch, slabs
2446 * allocated since entry are freed.
2450 STATIC I32 /* 0 failure, 1 success */
2451 S_regmatch(pTHX_ regexp *rex, regnode *prog)
2454 register const bool do_utf8 = PL_reg_match_utf8;
2455 const U32 uniflags = UTF8_ALLOW_DEFAULT;
2457 regmatch_slab *orig_slab;
2458 regmatch_state *orig_state;
2460 /* the current state. This is a cached copy of PL_regmatch_state */
2461 register regmatch_state *st;
2463 /* cache heavy used fields of st in registers */
2464 register regnode *scan;
2465 register regnode *next;
2466 register I32 n = 0; /* initialize to shut up compiler warning */
2467 register char *locinput = PL_reginput;
2469 /* these variables are NOT saved during a recusive RFEGMATCH: */
2470 register I32 nextchr; /* is always set to UCHARAT(locinput) */
2471 bool result; /* return value of S_regmatch */
2472 regnode *inner; /* Next node in internal branch. */
2473 int depth = 0; /* depth of recursion */
2474 regmatch_state *newst; /* when pushing a state, this is the new one */
2475 regmatch_state *cur_eval = NULL; /* most recent (??{}) state */
2478 SV *re_debug_flags = NULL;
2483 /* on first ever call to regmatch, allocate first slab */
2484 if (!PL_regmatch_slab) {
2485 Newx(PL_regmatch_slab, 1, regmatch_slab);
2486 PL_regmatch_slab->prev = NULL;
2487 PL_regmatch_slab->next = NULL;
2488 PL_regmatch_state = &PL_regmatch_slab->states[0] - 1;
2491 /* remember current high-water mark for exit */
2492 /* XXX this should be done with SAVE* instead */
2493 orig_slab = PL_regmatch_slab;
2494 orig_state = PL_regmatch_state;
2496 /* grab next free state slot */
2497 st = ++PL_regmatch_state;
2498 if (st > &(PL_regmatch_slab->states[PERL_REGMATCH_SLAB_SLOTS-1]))
2499 st = PL_regmatch_state = S_push_slab(aTHX);
2506 /* Note that nextchr is a byte even in UTF */
2507 nextchr = UCHARAT(locinput);
2509 while (scan != NULL) {
2512 SV * const prop = sv_newmortal();
2513 const int docolor = *PL_colors[0];
2514 const int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */
2515 int l = (PL_regeol - locinput) > taill ? taill : (PL_regeol - locinput);
2516 /* The part of the string before starttry has one color
2517 (pref0_len chars), between starttry and current
2518 position another one (pref_len - pref0_len chars),
2519 after the current position the third one.
2520 We assume that pref0_len <= pref_len, otherwise we
2521 decrease pref0_len. */
2522 int pref_len = (locinput - PL_bostr) > (5 + taill) - l
2523 ? (5 + taill) - l : locinput - PL_bostr;
2526 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len)))
2528 pref0_len = pref_len - (locinput - PL_reg_starttry);
2529 if (l + pref_len < (5 + taill) && l < PL_regeol - locinput)
2530 l = ( PL_regeol - locinput > (5 + taill) - pref_len
2531 ? (5 + taill) - pref_len : PL_regeol - locinput);
2532 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput + l)))
2536 if (pref0_len > pref_len)
2537 pref0_len = pref_len;
2538 regprop(rex, prop, scan);
2540 const char * const s0 =
2541 do_utf8 && OP(scan) != CANY ?
2542 pv_uni_display(PERL_DEBUG_PAD(0), (U8*)(locinput - pref_len),
2543 pref0_len, 60, UNI_DISPLAY_REGEX) :
2544 locinput - pref_len;
2545 const int len0 = do_utf8 ? strlen(s0) : pref0_len;
2546 const char * const s1 = do_utf8 && OP(scan) != CANY ?
2547 pv_uni_display(PERL_DEBUG_PAD(1),
2548 (U8*)(locinput - pref_len + pref0_len),
2549 pref_len - pref0_len, 60, UNI_DISPLAY_REGEX) :
2550 locinput - pref_len + pref0_len;
2551 const int len1 = do_utf8 ? strlen(s1) : pref_len - pref0_len;
2552 const char * const s2 = do_utf8 && OP(scan) != CANY ?
2553 pv_uni_display(PERL_DEBUG_PAD(2), (U8*)locinput,
2554 PL_regeol - locinput, 60, UNI_DISPLAY_REGEX) :
2556 const int len2 = do_utf8 ? strlen(s2) : l;
2557 PerlIO_printf(Perl_debug_log,
2558 "%4"IVdf" <%s%.*s%s%s%.*s%s%s%s%.*s%s>%*s|%3"IVdf":%*s%s\n",
2559 (IV)(locinput - PL_bostr),
2566 (docolor ? "" : "> <"),
2570 15 - l - pref_len + 1,
2572 (IV)(scan - rex->program), PL_regindent*2, "",
2577 next = scan + NEXT_OFF(scan);
2583 if (locinput == PL_bostr)
2585 /* regtill = regbol; */
2590 if (locinput == PL_bostr ||
2591 ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n'))
2597 if (locinput == PL_bostr)
2601 if (locinput == PL_reg_ganch)
2607 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2612 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2614 if (PL_regeol - locinput > 1)
2618 if (PL_regeol != locinput)
2622 if (!nextchr && locinput >= PL_regeol)
2625 locinput += PL_utf8skip[nextchr];
2626 if (locinput > PL_regeol)
2628 nextchr = UCHARAT(locinput);
2631 nextchr = UCHARAT(++locinput);
2634 if (!nextchr && locinput >= PL_regeol)
2636 nextchr = UCHARAT(++locinput);
2639 if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n')
2642 locinput += PL_utf8skip[nextchr];
2643 if (locinput > PL_regeol)
2645 nextchr = UCHARAT(locinput);
2648 nextchr = UCHARAT(++locinput);
2654 traverse the TRIE keeping track of all accepting states
2655 we transition through until we get to a failing node.
2663 U8 *uc = ( U8* )locinput;
2670 U8 *uscan = (U8*)NULL;
2672 SV *sv_accept_buff = NULL;
2673 const enum { trie_plain, trie_utf8, trie_uft8_fold }
2674 trie_type = do_utf8 ?
2675 (OP(scan) == TRIE ? trie_utf8 : trie_uft8_fold)
2678 /* what trie are we using right now */
2680 = (reg_trie_data*)rex->data->data[ ARG( scan ) ];
2681 st->u.trie.accepted = 0; /* how many accepting states we have seen */
2684 while ( state && uc <= (U8*)PL_regeol ) {
2686 if (trie->states[ state ].wordnum) {
2687 if (!st->u.trie.accepted ) {
2690 bufflen = TRIE_INITAL_ACCEPT_BUFFLEN;
2691 sv_accept_buff=newSV(bufflen *
2692 sizeof(reg_trie_accepted) - 1);
2693 SvCUR_set(sv_accept_buff,
2694 sizeof(reg_trie_accepted));
2695 SvPOK_on(sv_accept_buff);
2696 sv_2mortal(sv_accept_buff);
2697 st->u.trie.accept_buff =
2698 (reg_trie_accepted*)SvPV_nolen(sv_accept_buff );
2701 if (st->u.trie.accepted >= bufflen) {
2703 st->u.trie.accept_buff =(reg_trie_accepted*)
2704 SvGROW(sv_accept_buff,
2705 bufflen * sizeof(reg_trie_accepted));
2707 SvCUR_set(sv_accept_buff,SvCUR(sv_accept_buff)
2708 + sizeof(reg_trie_accepted));
2710 st->u.trie.accept_buff[st->u.trie.accepted].wordnum = trie->states[state].wordnum;
2711 st->u.trie.accept_buff[st->u.trie.accepted].endpos = uc;
2712 ++st->u.trie.accepted;
2715 base = trie->states[ state ].trans.base;
2717 DEBUG_TRIE_EXECUTE_r(
2718 PerlIO_printf( Perl_debug_log,
2719 "%*s %sState: %4"UVxf", Base: %4"UVxf", Accepted: %4"UVxf" ",
2720 REPORT_CODE_OFF + PL_regindent * 2, "", PL_colors[4],
2721 (UV)state, (UV)base, (UV)st->u.trie.accepted );
2725 switch (trie_type) {
2726 case trie_uft8_fold:
2728 uvc = utf8n_to_uvuni( uscan, UTF8_MAXLEN, &len, uniflags );
2733 U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ];
2734 uvc = utf8n_to_uvuni( (U8*)uc, UTF8_MAXLEN, &len, uniflags );
2735 uvc = to_uni_fold( uvc, foldbuf, &foldlen );
2736 foldlen -= UNISKIP( uvc );
2737 uscan = foldbuf + UNISKIP( uvc );
2741 uvc = utf8n_to_uvuni( (U8*)uc, UTF8_MAXLEN,
2750 charid = trie->charmap[ uvc ];
2754 if (trie->widecharmap) {
2755 SV** svpp = (SV**)NULL;
2756 svpp = hv_fetch(trie->widecharmap,
2757 (char*)&uvc, sizeof(UV), 0);
2759 charid = (U16)SvIV(*svpp);
2764 (base + charid > trie->uniquecharcount )
2765 && (base + charid - 1 - trie->uniquecharcount
2767 && trie->trans[base + charid - 1 -
2768 trie->uniquecharcount].check == state)
2770 state = trie->trans[base + charid - 1 -
2771 trie->uniquecharcount ].next;
2782 DEBUG_TRIE_EXECUTE_r(
2783 PerlIO_printf( Perl_debug_log,
2784 "Charid:%3x CV:%4"UVxf" After State: %4"UVxf"%s\n",
2785 charid, uvc, (UV)state, PL_colors[5] );
2788 if (!st->u.trie.accepted )
2792 There was at least one accepting state that we
2793 transitioned through. Presumably the number of accepting
2794 states is going to be low, typically one or two. So we
2795 simply scan through to find the one with lowest wordnum.
2796 Once we find it, we swap the last state into its place
2797 and decrement the size. We then try to match the rest of
2798 the pattern at the point where the word ends, if we
2799 succeed then we end the loop, otherwise the loop
2800 eventually terminates once all of the accepting states
2804 if ( st->u.trie.accepted == 1 ) {
2806 SV ** const tmp = av_fetch( trie->words, st->u.trie.accept_buff[ 0 ].wordnum-1, 0 );
2807 PerlIO_printf( Perl_debug_log,
2808 "%*s %sonly one match : #%d <%s>%s\n",
2809 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4],
2810 st->u.trie.accept_buff[ 0 ].wordnum,
2811 tmp ? SvPV_nolen_const( *tmp ) : "not compiled under -Dr",
2814 PL_reginput = (char *)st->u.trie.accept_buff[ 0 ].endpos;
2815 /* in this case we free tmps/leave before we call regmatch
2816 as we wont be using accept_buff again. */
2819 REGMATCH(scan + NEXT_OFF(scan), TRIE1);
2820 /*** all unsaved local vars undefined at this point */
2823 PerlIO_printf( Perl_debug_log,"%*s %sgot %"IVdf" possible matches%s\n",
2824 REPORT_CODE_OFF + PL_regindent * 2, "", PL_colors[4], (IV)st->u.trie.accepted,
2827 while ( !result && st->u.trie.accepted-- ) {
2830 for( cur = 1 ; cur <= st->u.trie.accepted ; cur++ ) {
2831 DEBUG_TRIE_EXECUTE_r(
2832 PerlIO_printf( Perl_debug_log,
2833 "%*s %sgot %"IVdf" (%d) as best, looking at %"IVdf" (%d)%s\n",
2834 REPORT_CODE_OFF + PL_regindent * 2, "", PL_colors[4],
2835 (IV)best, st->u.trie.accept_buff[ best ].wordnum, (IV)cur,
2836 st->u.trie.accept_buff[ cur ].wordnum, PL_colors[5] );
2839 if (st->u.trie.accept_buff[cur].wordnum <
2840 st->u.trie.accept_buff[best].wordnum)
2844 reg_trie_data * const trie = (reg_trie_data*)
2845 rex->data->data[ARG(scan)];
2846 SV ** const tmp = av_fetch( trie->words, st->u.trie.accept_buff[ best ].wordnum - 1, 0 );
2847 PerlIO_printf( Perl_debug_log, "%*s %strying alternation #%d <%s> at 0x%p%s\n",
2848 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4],
2849 st->u.trie.accept_buff[best].wordnum,
2850 tmp ? SvPV_nolen_const( *tmp ) : "not compiled under -Dr",scan,
2853 if ( best<st->u.trie.accepted ) {
2854 reg_trie_accepted tmp = st->u.trie.accept_buff[ best ];
2855 st->u.trie.accept_buff[ best ] = st->u.trie.accept_buff[ st->u.trie.accepted ];
2856 st->u.trie.accept_buff[ st->u.trie.accepted ] = tmp;
2857 best = st->u.trie.accepted;
2859 PL_reginput = (char *)st->u.trie.accept_buff[ best ].endpos;
2862 as far as I can tell we only need the SAVETMPS/FREETMPS
2863 for re's with EVAL in them but I'm leaving them in for
2864 all until I can be sure.
2867 REGMATCH(scan + NEXT_OFF(scan), TRIE2);
2868 /*** all unsaved local vars undefined at this point */
2881 /* unreached codepoint */
2883 char *s = STRING(scan);
2884 st->ln = STR_LEN(scan);
2885 if (do_utf8 != UTF) {
2886 /* The target and the pattern have differing utf8ness. */
2888 const char *e = s + st->ln;
2891 /* The target is utf8, the pattern is not utf8. */
2896 if (NATIVE_TO_UNI(*(U8*)s) !=
2897 utf8n_to_uvuni((U8*)l, UTF8_MAXBYTES, &ulen,
2905 /* The target is not utf8, the pattern is utf8. */
2910 if (NATIVE_TO_UNI(*((U8*)l)) !=
2911 utf8n_to_uvuni((U8*)s, UTF8_MAXBYTES, &ulen,
2919 nextchr = UCHARAT(locinput);
2922 /* The target and the pattern have the same utf8ness. */
2923 /* Inline the first character, for speed. */
2924 if (UCHARAT(s) != nextchr)
2926 if (PL_regeol - locinput < st->ln)
2928 if (st->ln > 1 && memNE(s, locinput, st->ln))
2931 nextchr = UCHARAT(locinput);
2935 PL_reg_flags |= RF_tainted;
2938 char *s = STRING(scan);
2939 st->ln = STR_LEN(scan);
2941 if (do_utf8 || UTF) {
2942 /* Either target or the pattern are utf8. */
2944 char *e = PL_regeol;
2946 if (ibcmp_utf8(s, 0, st->ln, (bool)UTF,
2947 l, &e, 0, do_utf8)) {
2948 /* One more case for the sharp s:
2949 * pack("U0U*", 0xDF) =~ /ss/i,
2950 * the 0xC3 0x9F are the UTF-8
2951 * byte sequence for the U+00DF. */
2953 toLOWER(s[0]) == 's' &&
2955 toLOWER(s[1]) == 's' &&
2962 nextchr = UCHARAT(locinput);
2966 /* Neither the target and the pattern are utf8. */
2968 /* Inline the first character, for speed. */
2969 if (UCHARAT(s) != nextchr &&
2970 UCHARAT(s) != ((OP(scan) == EXACTF)
2971 ? PL_fold : PL_fold_locale)[nextchr])
2973 if (PL_regeol - locinput < st->ln)
2975 if (st->ln > 1 && (OP(scan) == EXACTF
2976 ? ibcmp(s, locinput, st->ln)
2977 : ibcmp_locale(s, locinput, st->ln)))
2980 nextchr = UCHARAT(locinput);
2985 STRLEN inclasslen = PL_regeol - locinput;
2987 if (!reginclass(rex, scan, (U8*)locinput, &inclasslen, do_utf8))
2989 if (locinput >= PL_regeol)
2991 locinput += inclasslen ? inclasslen : UTF8SKIP(locinput);
2992 nextchr = UCHARAT(locinput);
2997 nextchr = UCHARAT(locinput);
2998 if (!REGINCLASS(rex, scan, (U8*)locinput))
3000 if (!nextchr && locinput >= PL_regeol)
3002 nextchr = UCHARAT(++locinput);
3006 /* If we might have the case of the German sharp s
3007 * in a casefolding Unicode character class. */
3009 if (ANYOF_FOLD_SHARP_S(scan, locinput, PL_regeol)) {
3010 locinput += SHARP_S_SKIP;
3011 nextchr = UCHARAT(locinput);
3017 PL_reg_flags |= RF_tainted;
3023 LOAD_UTF8_CHARCLASS_ALNUM();
3024 if (!(OP(scan) == ALNUM
3025 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
3026 : isALNUM_LC_utf8((U8*)locinput)))
3030 locinput += PL_utf8skip[nextchr];
3031 nextchr = UCHARAT(locinput);
3034 if (!(OP(scan) == ALNUM
3035 ? isALNUM(nextchr) : isALNUM_LC(nextchr)))
3037 nextchr = UCHARAT(++locinput);
3040 PL_reg_flags |= RF_tainted;
3043 if (!nextchr && locinput >= PL_regeol)
3046 LOAD_UTF8_CHARCLASS_ALNUM();
3047 if (OP(scan) == NALNUM
3048 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
3049 : isALNUM_LC_utf8((U8*)locinput))
3053 locinput += PL_utf8skip[nextchr];
3054 nextchr = UCHARAT(locinput);
3057 if (OP(scan) == NALNUM
3058 ? isALNUM(nextchr) : isALNUM_LC(nextchr))
3060 nextchr = UCHARAT(++locinput);
3064 PL_reg_flags |= RF_tainted;
3068 /* was last char in word? */
3070 if (locinput == PL_bostr)
3073 const U8 * const r = reghop3((U8*)locinput, -1, (U8*)PL_bostr);
3075 st->ln = utf8n_to_uvchr(r, UTF8SKIP(r), 0, uniflags);
3077 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
3078 st->ln = isALNUM_uni(st->ln);
3079 LOAD_UTF8_CHARCLASS_ALNUM();
3080 n = swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8);
3083 st->ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(st->ln));
3084 n = isALNUM_LC_utf8((U8*)locinput);
3088 st->ln = (locinput != PL_bostr) ?
3089 UCHARAT(locinput - 1) : '\n';
3090 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
3091 st->ln = isALNUM(st->ln);
3092 n = isALNUM(nextchr);
3095 st->ln = isALNUM_LC(st->ln);
3096 n = isALNUM_LC(nextchr);
3099 if (((!st->ln) == (!n)) == (OP(scan) == BOUND ||
3100 OP(scan) == BOUNDL))
3104 PL_reg_flags |= RF_tainted;
3110 if (UTF8_IS_CONTINUED(nextchr)) {
3111 LOAD_UTF8_CHARCLASS_SPACE();
3112 if (!(OP(scan) == SPACE
3113 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
3114 : isSPACE_LC_utf8((U8*)locinput)))
3118 locinput += PL_utf8skip[nextchr];
3119 nextchr = UCHARAT(locinput);
3122 if (!(OP(scan) == SPACE
3123 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
3125 nextchr = UCHARAT(++locinput);
3128 if (!(OP(scan) == SPACE
3129 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
3131 nextchr = UCHARAT(++locinput);
3135 PL_reg_flags |= RF_tainted;
3138 if (!nextchr && locinput >= PL_regeol)
3141 LOAD_UTF8_CHARCLASS_SPACE();
3142 if (OP(scan) == NSPACE
3143 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
3144 : isSPACE_LC_utf8((U8*)locinput))
3148 locinput += PL_utf8skip[nextchr];
3149 nextchr = UCHARAT(locinput);
3152 if (OP(scan) == NSPACE
3153 ? isSPACE(nextchr) : isSPACE_LC(nextchr))
3155 nextchr = UCHARAT(++locinput);
3158 PL_reg_flags |= RF_tainted;
3164 LOAD_UTF8_CHARCLASS_DIGIT();
3165 if (!(OP(scan) == DIGIT
3166 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
3167 : isDIGIT_LC_utf8((U8*)locinput)))
3171 locinput += PL_utf8skip[nextchr];
3172 nextchr = UCHARAT(locinput);
3175 if (!(OP(scan) == DIGIT
3176 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr)))
3178 nextchr = UCHARAT(++locinput);
3181 PL_reg_flags |= RF_tainted;
3184 if (!nextchr && locinput >= PL_regeol)
3187 LOAD_UTF8_CHARCLASS_DIGIT();
3188 if (OP(scan) == NDIGIT
3189 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
3190 : isDIGIT_LC_utf8((U8*)locinput))
3194 locinput += PL_utf8skip[nextchr];
3195 nextchr = UCHARAT(locinput);
3198 if (OP(scan) == NDIGIT
3199 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr))
3201 nextchr = UCHARAT(++locinput);
3204 if (locinput >= PL_regeol)
3207 LOAD_UTF8_CHARCLASS_MARK();
3208 if (swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
3210 locinput += PL_utf8skip[nextchr];
3211 while (locinput < PL_regeol &&
3212 swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
3213 locinput += UTF8SKIP(locinput);
3214 if (locinput > PL_regeol)
3219 nextchr = UCHARAT(locinput);
3222 PL_reg_flags |= RF_tainted;
3227 n = ARG(scan); /* which paren pair */
3228 st->ln = PL_regstartp[n];
3229 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
3230 if ((I32)*PL_reglastparen < n || st->ln == -1)
3231 sayNO; /* Do not match unless seen CLOSEn. */
3232 if (st->ln == PL_regendp[n])
3235 s = PL_bostr + st->ln;
3236 if (do_utf8 && OP(scan) != REF) { /* REF can do byte comparison */
3238 const char *e = PL_bostr + PL_regendp[n];
3240 * Note that we can't do the "other character" lookup trick as
3241 * in the 8-bit case (no pun intended) because in Unicode we
3242 * have to map both upper and title case to lower case.
3244 if (OP(scan) == REFF) {
3246 STRLEN ulen1, ulen2;
3247 U8 tmpbuf1[UTF8_MAXBYTES_CASE+1];
3248 U8 tmpbuf2[UTF8_MAXBYTES_CASE+1];
3252 toLOWER_utf8((U8*)s, tmpbuf1, &ulen1);
3253 toLOWER_utf8((U8*)l, tmpbuf2, &ulen2);
3254 if (ulen1 != ulen2 || memNE((char *)tmpbuf1, (char *)tmpbuf2, ulen1))
3261 nextchr = UCHARAT(locinput);
3265 /* Inline the first character, for speed. */
3266 if (UCHARAT(s) != nextchr &&
3268 (UCHARAT(s) != ((OP(scan) == REFF
3269 ? PL_fold : PL_fold_locale)[nextchr]))))
3271 st->ln = PL_regendp[n] - st->ln;
3272 if (locinput + st->ln > PL_regeol)
3274 if (st->ln > 1 && (OP(scan) == REF
3275 ? memNE(s, locinput, st->ln)
3277 ? ibcmp(s, locinput, st->ln)
3278 : ibcmp_locale(s, locinput, st->ln))))
3281 nextchr = UCHARAT(locinput);
3294 /* execute the code in the {...} */
3296 SV ** const before = SP;
3297 OP_4tree * const oop = PL_op;
3298 COP * const ocurcop = PL_curcop;
3302 PL_op = (OP_4tree*)rex->data->data[n];
3303 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) );
3304 PAD_SAVE_LOCAL(old_comppad, (PAD*)rex->data->data[n + 2]);
3305 PL_regendp[0] = PL_reg_magic->mg_len = locinput - PL_bostr;
3307 CALLRUNOPS(aTHX); /* Scalar context. */
3310 ret = &PL_sv_undef; /* protect against empty (?{}) blocks. */
3317 PAD_RESTORE_LOCAL(old_comppad);
3318 PL_curcop = ocurcop;
3321 sv_setsv(save_scalar(PL_replgv), ret);
3325 if (st->logical == 2) { /* Postponed subexpression: /(??{...})/ */
3328 /* extract RE object from returned value; compiling if
3333 if(SvROK(ret) && SvSMAGICAL(sv = SvRV(ret)))
3334 mg = mg_find(sv, PERL_MAGIC_qr);
3335 else if (SvSMAGICAL(ret)) {
3336 if (SvGMAGICAL(ret))
3337 sv_unmagic(ret, PERL_MAGIC_qr);
3339 mg = mg_find(ret, PERL_MAGIC_qr);
3343 re = (regexp *)mg->mg_obj;
3344 (void)ReREFCNT_inc(re);
3348 const char * const t = SvPV_const(ret, len);
3350 const I32 osize = PL_regsize;
3353 if (DO_UTF8(ret)) pm.op_pmdynflags |= PMdf_DYN_UTF8;
3354 re = CALLREGCOMP(aTHX_ (char*)t, (char*)t + len, &pm);
3356 & (SVs_TEMP | SVs_PADTMP | SVf_READONLY
3358 sv_magic(ret,(SV*)ReREFCNT_inc(re),
3364 /* run the pattern returned from (??{...}) */
3367 PerlIO_printf(Perl_debug_log,
3368 "Entering embedded \"%s%.60s%s%s\"\n",
3372 (strlen(re->precomp) > 60 ? "..." : ""))
3375 st->u.eval.cp = regcppush(0); /* Save *all* the positions. */
3376 REGCP_SET(st->u.eval.lastcp);
3377 *PL_reglastparen = 0;
3378 *PL_reglastcloseparen = 0;
3379 PL_reginput = locinput;
3381 /* XXXX This is too dramatic a measure... */
3385 st->u.eval.toggleutf = ((PL_reg_flags & RF_utf8) != 0) ^
3386 ((re->reganch & ROPT_UTF8) != 0);
3387 if (st->u.eval.toggleutf) PL_reg_flags ^= RF_utf8;
3388 st->u.eval.prev_rex = rex;
3391 st->u.eval.prev_eval = cur_eval;
3392 st->u.eval.prev_slab = PL_regmatch_slab;
3393 st->u.eval.depth = depth;
3395 PUSH_STATE(newst, resume_EVAL);
3398 /* now continue from first node in postoned RE */
3399 next = re->program + 1;
3403 /* /(?(?{...})X|Y)/ */
3404 st->sw = SvTRUE(ret);
3409 n = ARG(scan); /* which paren pair */
3410 PL_reg_start_tmp[n] = locinput;
3415 n = ARG(scan); /* which paren pair */
3416 PL_regstartp[n] = PL_reg_start_tmp[n] - PL_bostr;
3417 PL_regendp[n] = locinput - PL_bostr;
3418 if (n > (I32)*PL_reglastparen)
3419 *PL_reglastparen = n;
3420 *PL_reglastcloseparen = n;
3423 n = ARG(scan); /* which paren pair */
3424 st->sw = ((I32)*PL_reglastparen >= n && PL_regendp[n] != -1);
3427 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
3429 next = NEXTOPER(NEXTOPER(scan));
3431 next = scan + ARG(scan);
3432 if (OP(next) == IFTHEN) /* Fake one. */
3433 next = NEXTOPER(NEXTOPER(next));
3437 st->logical = scan->flags;
3439 /*******************************************************************
3440 cc points to the regmatch_state associated with the most recent CURLYX.
3441 This struct contains info about the innermost (...)* loop (an
3442 "infoblock"), and a pointer to the next outer cc.
3444 Here is how Y(A)*Z is processed (if it is compiled into CURLYX/WHILEM):
3446 1) After matching Y, regnode for CURLYX is processed;
3448 2) This regnode populates cc, and calls regmatch() recursively
3449 with the starting point at WHILEM node;
3451 3) Each hit of WHILEM node tries to match A and Z (in the order
3452 depending on the current iteration, min/max of {min,max} and
3453 greediness). The information about where are nodes for "A"
3454 and "Z" is read from cc, as is info on how many times "A"
3455 was already matched, and greediness.
3457 4) After A matches, the same WHILEM node is hit again.
3459 5) Each time WHILEM is hit, cc is the infoblock created by CURLYX
3460 of the same pair. Thus when WHILEM tries to match Z, it temporarily
3461 resets cc, since this Y(A)*Z can be a part of some other loop:
3462 as in (Y(A)*Z)*. If Z matches, the automaton will hit the WHILEM node
3463 of the external loop.
3465 Currently present infoblocks form a tree with a stem formed by st->cc
3466 and whatever it mentions via ->next, and additional attached trees
3467 corresponding to temporarily unset infoblocks as in "5" above.
3469 In the following picture, infoblocks for outer loop of
3470 (Y(A)*?Z)*?T are denoted O, for inner I. NULL starting block
3471 is denoted by x. The matched string is YAAZYAZT. Temporarily postponed
3472 infoblocks are drawn below the "reset" infoblock.
3474 In fact in the picture below we do not show failed matches for Z and T
3475 by WHILEM blocks. [We illustrate minimal matches, since for them it is
3476 more obvious *why* one needs to *temporary* unset infoblocks.]
3478 Matched REx position InfoBlocks Comment
3482 Y A)*?Z)*?T x <- O <- I
3483 YA )*?Z)*?T x <- O <- I
3484 YA A)*?Z)*?T x <- O <- I
3485 YAA )*?Z)*?T x <- O <- I
3486 YAA Z)*?T x <- O # Temporary unset I
3489 YAAZ Y(A)*?Z)*?T x <- O
3492 YAAZY (A)*?Z)*?T x <- O
3495 YAAZY A)*?Z)*?T x <- O <- I
3498 YAAZYA )*?Z)*?T x <- O <- I
3501 YAAZYA Z)*?T x <- O # Temporary unset I
3507 YAAZYAZ T x # Temporary unset O
3514 *******************************************************************/
3517 /* No need to save/restore up to this paren */
3518 I32 parenfloor = scan->flags;
3520 if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */
3522 /* XXXX Probably it is better to teach regpush to support
3523 parenfloor > PL_regsize... */
3524 if (parenfloor > (I32)*PL_reglastparen)
3525 parenfloor = *PL_reglastparen; /* Pessimization... */
3527 st->u.curlyx.cp = PL_savestack_ix;
3528 st->u.curlyx.outercc = st->cc;
3530 /* these fields contain the state of the current curly.
3531 * they are accessed by subsequent WHILEMs;
3532 * cur and lastloc are also updated by WHILEM */
3533 st->u.curlyx.parenfloor = parenfloor;
3534 st->u.curlyx.cur = -1; /* this will be updated by WHILEM */
3535 st->u.curlyx.min = ARG1(scan);
3536 st->u.curlyx.max = ARG2(scan);
3537 st->u.curlyx.scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
3538 st->u.curlyx.lastloc = 0;
3539 /* st->next and st->minmod are also read by WHILEM */
3541 PL_reginput = locinput;
3542 REGMATCH(PREVOPER(next), CURLYX); /* start on the WHILEM */
3543 /*** all unsaved local vars undefined at this point */
3544 regcpblow(st->u.curlyx.cp);
3545 st->cc = st->u.curlyx.outercc;
3551 * This is really hard to understand, because after we match
3552 * what we're trying to match, we must make sure the rest of
3553 * the REx is going to match for sure, and to do that we have
3554 * to go back UP the parse tree by recursing ever deeper. And
3555 * if it fails, we have to reset our parent's current state
3556 * that we can try again after backing off.
3559 st->u.whilem.lastloc = st->cc->u.curlyx.lastloc; /* Detection of 0-len. */
3560 st->u.whilem.cache_offset = 0;
3561 st->u.whilem.cache_bit = 0;
3563 n = st->cc->u.curlyx.cur + 1; /* how many we know we matched */
3564 PL_reginput = locinput;
3567 PerlIO_printf(Perl_debug_log,
3568 "%*s %ld out of %ld..%ld cc=%"UVxf"\n",
3569 REPORT_CODE_OFF+PL_regindent*2, "",
3570 (long)n, (long)st->cc->u.curlyx.min,
3571 (long)st->cc->u.curlyx.max, PTR2UV(st->cc))
3574 /* If degenerate scan matches "", assume scan done. */
3576 if (locinput == st->cc->u.curlyx.lastloc && n >= st->cc->u.curlyx.min) {
3577 st->u.whilem.savecc = st->cc;
3578 st->cc = st->cc->u.curlyx.outercc;
3580 st->ln = st->cc->u.curlyx.cur;
3582 PerlIO_printf(Perl_debug_log,
3583 "%*s empty match detected, try continuation...\n",
3584 REPORT_CODE_OFF+PL_regindent*2, "")
3586 REGMATCH(st->u.whilem.savecc->next, WHILEM1);
3587 /*** all unsaved local vars undefined at this point */
3588 st->cc = st->u.whilem.savecc;
3591 if (st->cc->u.curlyx.outercc)
3592 st->cc->u.curlyx.outercc->u.curlyx.cur = st->ln;
3596 /* First just match a string of min scans. */
3598 if (n < st->cc->u.curlyx.min) {
3599 st->cc->u.curlyx.cur = n;
3600 st->cc->u.curlyx.lastloc = locinput;
3601 REGMATCH(st->cc->u.curlyx.scan, WHILEM2);
3602 /*** all unsaved local vars undefined at this point */
3605 st->cc->u.curlyx.cur = n - 1;
3606 st->cc->u.curlyx.lastloc = st->u.whilem.lastloc;
3611 /* Check whether we already were at this position.
3612 Postpone detection until we know the match is not
3613 *that* much linear. */
3614 if (!PL_reg_maxiter) {
3615 PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4);
3616 PL_reg_leftiter = PL_reg_maxiter;
3618 if (PL_reg_leftiter-- == 0) {
3619 const I32 size = (PL_reg_maxiter + 7 + POSCACHE_START)/8;
3620 if (PL_reg_poscache) {
3621 if ((I32)PL_reg_poscache_size < size) {
3622 Renew(PL_reg_poscache, size, char);
3623 PL_reg_poscache_size = size;
3625 Zero(PL_reg_poscache, size, char);
3628 PL_reg_poscache_size = size;
3629 Newxz(PL_reg_poscache, size, char);
3632 PerlIO_printf(Perl_debug_log,
3633 "%sDetected a super-linear match, switching on caching%s...\n",
3634 PL_colors[4], PL_colors[5])
3637 if (PL_reg_leftiter < 0) {
3638 st->u.whilem.cache_offset = locinput - PL_bostr;
3640 st->u.whilem.cache_offset = (scan->flags & 0xf) - 1 + POSCACHE_START
3641 + st->u.whilem.cache_offset * (scan->flags>>4);
3642 st->u.whilem.cache_bit = st->u.whilem.cache_offset % 8;
3643 st->u.whilem.cache_offset /= 8;
3644 if (PL_reg_poscache[st->u.whilem.cache_offset] & (1<<st->u.whilem.cache_bit)) {
3646 PerlIO_printf(Perl_debug_log,
3647 "%*s already tried at this position...\n",
3648 REPORT_CODE_OFF+PL_regindent*2, "")
3650 if (PL_reg_poscache[0] & (1<<POSCACHE_SUCCESS))
3651 /* cache records success */
3654 /* cache records failure */
3657 PL_reg_poscache[st->u.whilem.cache_offset] |= (1<<st->u.whilem.cache_bit);
3661 /* Prefer next over scan for minimal matching. */
3663 if (st->cc->minmod) {
3664 st->u.whilem.savecc = st->cc;
3665 st->cc = st->cc->u.curlyx.outercc;
3667 st->ln = st->cc->u.curlyx.cur;
3668 st->u.whilem.cp = regcppush(st->u.whilem.savecc->u.curlyx.parenfloor);
3669 REGCP_SET(st->u.whilem.lastcp);
3670 REGMATCH(st->u.whilem.savecc->next, WHILEM3);
3671 /*** all unsaved local vars undefined at this point */
3672 st->cc = st->u.whilem.savecc;
3674 regcpblow(st->u.whilem.cp);
3675 CACHEsayYES; /* All done. */
3677 REGCP_UNWIND(st->u.whilem.lastcp);
3679 if (st->cc->u.curlyx.outercc)
3680 st->cc->u.curlyx.outercc->u.curlyx.cur = st->ln;
3682 if (n >= st->cc->u.curlyx.max) { /* Maximum greed exceeded? */
3683 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3684 && !(PL_reg_flags & RF_warned)) {
3685 PL_reg_flags |= RF_warned;
3686 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s limit (%d) exceeded",
3687 "Complex regular subexpression recursion",
3694 PerlIO_printf(Perl_debug_log,
3695 "%*s trying longer...\n",
3696 REPORT_CODE_OFF+PL_regindent*2, "")
3698 /* Try scanning more and see if it helps. */
3699 PL_reginput = locinput;
3700 st->cc->u.curlyx.cur = n;
3701 st->cc->u.curlyx.lastloc = locinput;
3702 st->u.whilem.cp = regcppush(st->cc->u.curlyx.parenfloor);
3703 REGCP_SET(st->u.whilem.lastcp);
3704 REGMATCH(st->cc->u.curlyx.scan, WHILEM4);
3705 /*** all unsaved local vars undefined at this point */
3707 regcpblow(st->u.whilem.cp);
3710 REGCP_UNWIND(st->u.whilem.lastcp);
3712 st->cc->u.curlyx.cur = n - 1;
3713 st->cc->u.curlyx.lastloc = st->u.whilem.lastloc;
3717 /* Prefer scan over next for maximal matching. */
3719 if (n < st->cc->u.curlyx.max) { /* More greed allowed? */
3720 st->u.whilem.cp = regcppush(st->cc->u.curlyx.parenfloor);
3721 st->cc->u.curlyx.cur = n;
3722 st->cc->u.curlyx.lastloc = locinput;
3723 REGCP_SET(st->u.whilem.lastcp);
3724 REGMATCH(st->cc->u.curlyx.scan, WHILEM5);
3725 /*** all unsaved local vars undefined at this point */
3727 regcpblow(st->u.whilem.cp);
3730 REGCP_UNWIND(st->u.whilem.lastcp);
3731 regcppop(rex); /* Restore some previous $<digit>s? */
3732 PL_reginput = locinput;
3734 PerlIO_printf(Perl_debug_log,
3735 "%*s failed, try continuation...\n",
3736 REPORT_CODE_OFF+PL_regindent*2, "")
3739 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3740 && !(PL_reg_flags & RF_warned)) {
3741 PL_reg_flags |= RF_warned;
3742 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s limit (%d) exceeded",
3743 "Complex regular subexpression recursion",
3747 /* Failed deeper matches of scan, so see if this one works. */
3748 st->u.whilem.savecc = st->cc;
3749 st->cc = st->cc->u.curlyx.outercc;
3751 st->ln = st->cc->u.curlyx.cur;
3752 REGMATCH(st->u.whilem.savecc->next, WHILEM6);
3753 /*** all unsaved local vars undefined at this point */
3754 st->cc = st->u.whilem.savecc;
3757 if (st->cc->u.curlyx.outercc)
3758 st->cc->u.curlyx.outercc->u.curlyx.cur = st->ln;
3759 st->cc->u.curlyx.cur = n - 1;
3760 st->cc->u.curlyx.lastloc = st->u.whilem.lastloc;
3765 next = scan + ARG(scan);
3768 inner = NEXTOPER(NEXTOPER(scan));
3771 inner = NEXTOPER(scan);
3776 if (OP(next) != type) /* No choice. */
3777 next = inner; /* Avoid recursion. */
3779 const I32 lastparen = *PL_reglastparen;
3780 /* Put unwinding data on stack */
3781 const I32 unwind1 = SSNEWt(1,re_unwind_branch_t);
3782 re_unwind_branch_t * const uw = SSPTRt(unwind1,re_unwind_branch_t);
3784 uw->prev = st->unwind;
3785 st->unwind = unwind1;
3786 uw->type = ((type == BRANCH)
3788 : RE_UNWIND_BRANCHJ);
3789 uw->lastparen = lastparen;
3791 uw->locinput = locinput;
3792 uw->nextchr = nextchr;
3793 uw->minmod = st->minmod;
3795 uw->regindent = ++PL_regindent;
3798 REGCP_SET(uw->lastcp);
3800 /* Now go into the first branch */
3810 st->u.curlym.l = st->u.curlym.matches = 0;
3812 /* We suppose that the next guy does not need
3813 backtracking: in particular, it is of constant non-zero length,
3814 and has no parenths to influence future backrefs. */
3815 st->ln = ARG1(scan); /* min to match */
3816 n = ARG2(scan); /* max to match */
3817 st->u.curlym.paren = scan->flags;
3818 if (st->u.curlym.paren) {
3819 if (st->u.curlym.paren > PL_regsize)
3820 PL_regsize = st->u.curlym.paren;
3821 if (st->u.curlym.paren > (I32)*PL_reglastparen)
3822 *PL_reglastparen = st->u.curlym.paren;
3824 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3825 if (st->u.curlym.paren)
3826 scan += NEXT_OFF(scan); /* Skip former OPEN. */
3827 PL_reginput = locinput;
3828 st->u.curlym.maxwanted = st->minmod ? st->ln : n;
3829 if (st->u.curlym.maxwanted) {
3830 while (PL_reginput < PL_regeol && st->u.curlym.matches < st->u.curlym.maxwanted) {
3831 REGMATCH(scan, CURLYM1);
3832 /*** all unsaved local vars undefined at this point */
3835 /* on first match, determine length, u.curlym.l */
3836 if (!st->u.curlym.matches++) {
3837 if (PL_reg_match_utf8) {
3839 while (s < PL_reginput) {
3845 st->u.curlym.l = PL_reginput - locinput;
3847 if (st->u.curlym.l == 0) {
3848 st->u.curlym.matches = st->u.curlym.maxwanted;
3852 locinput = PL_reginput;
3856 PL_reginput = locinput;
3860 if (st->ln && st->u.curlym.matches < st->ln)
3862 if (HAS_TEXT(next) || JUMPABLE(next)) {
3863 regnode *text_node = next;
3865 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3867 if (! HAS_TEXT(text_node)) st->u.curlym.c1 = st->u.curlym.c2 = -1000;
3869 if (PL_regkind[(U8)OP(text_node)] == REF) {
3870 st->u.curlym.c1 = st->u.curlym.c2 = -1000;
3873 else { st->u.curlym.c1 = (U8)*STRING(text_node); }
3874 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3875 st->u.curlym.c2 = PL_fold[st->u.curlym.c1];
3876 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3877 st->u.curlym.c2 = PL_fold_locale[st->u.curlym.c1];
3879 st->u.curlym.c2 = st->u.curlym.c1;
3883 st->u.curlym.c1 = st->u.curlym.c2 = -1000;
3885 REGCP_SET(st->u.curlym.lastcp);
3886 while (n >= st->ln || (n == REG_INFTY && st->ln > 0)) { /* ln overflow ? */
3887 /* If it could work, try it. */
3888 if (st->u.curlym.c1 == -1000 ||
3889 UCHARAT(PL_reginput) == st->u.curlym.c1 ||
3890 UCHARAT(PL_reginput) == st->u.curlym.c2)
3892 if (st->u.curlym.paren) {
3894 PL_regstartp[st->u.curlym.paren] =
3895 HOPc(PL_reginput, -st->u.curlym.l) - PL_bostr;
3896 PL_regendp[st->u.curlym.paren] = PL_reginput - PL_bostr;
3899 PL_regendp[st->u.curlym.paren] = -1;
3901 REGMATCH(next, CURLYM2);
3902 /*** all unsaved local vars undefined at this point */
3905 REGCP_UNWIND(st->u.curlym.lastcp);
3907 /* Couldn't or didn't -- move forward. */
3908 PL_reginput = locinput;
3909 REGMATCH(scan, CURLYM3);
3910 /*** all unsaved local vars undefined at this point */
3913 locinput = PL_reginput;
3921 PerlIO_printf(Perl_debug_log,
3922 "%*s matched %"IVdf" times, len=%"IVdf"...\n",
3923 (int)(REPORT_CODE_OFF+PL_regindent*2), "",
3924 (IV) st->u.curlym.matches, (IV)st->u.curlym.l)
3926 if (st->u.curlym.matches >= st->ln) {
3927 if (HAS_TEXT(next) || JUMPABLE(next)) {
3928 regnode *text_node = next;
3930 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3932 if (! HAS_TEXT(text_node)) st->u.curlym.c1 = st->u.curlym.c2 = -1000;
3934 if (PL_regkind[(U8)OP(text_node)] == REF) {
3935 st->u.curlym.c1 = st->u.curlym.c2 = -1000;
3938 else { st->u.curlym.c1 = (U8)*STRING(text_node); }
3940 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3941 st->u.curlym.c2 = PL_fold[st->u.curlym.c1];
3942 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3943 st->u.curlym.c2 = PL_fold_locale[st->u.curlym.c1];
3945 st->u.curlym.c2 = st->u.curlym.c1;
3949 st->u.curlym.c1 = st->u.curlym.c2 = -1000;
3952 REGCP_SET(st->u.curlym.lastcp);
3953 while (st->u.curlym.matches >= st->ln) {
3954 /* If it could work, try it. */
3955 if (st->u.curlym.c1 == -1000 ||
3956 UCHARAT(PL_reginput) == st->u.curlym.c1 ||
3957 UCHARAT(PL_reginput) == st->u.curlym.c2)
3960 PerlIO_printf(Perl_debug_log,
3961 "%*s trying tail with matches=%"IVdf"...\n",
3962 (int)(REPORT_CODE_OFF+PL_regindent*2),
3963 "", (IV)st->u.curlym.matches)
3965 if (st->u.curlym.paren) {
3966 if (st->u.curlym.matches) {
3967 PL_regstartp[st->u.curlym.paren]
3968 = HOPc(PL_reginput, -st->u.curlym.l) - PL_bostr;
3969 PL_regendp[st->u.curlym.paren] = PL_reginput - PL_bostr;
3972 PL_regendp[st->u.curlym.paren] = -1;
3974 REGMATCH(next, CURLYM4);
3975 /*** all unsaved local vars undefined at this point */
3978 REGCP_UNWIND(st->u.curlym.lastcp);
3980 /* Couldn't or didn't -- back up. */
3981 st->u.curlym.matches--;
3982 locinput = HOPc(locinput, -st->u.curlym.l);
3983 PL_reginput = locinput;
3991 st->u.plus.paren = scan->flags; /* Which paren to set */
3992 if (st->u.plus.paren > PL_regsize)
3993 PL_regsize = st->u.plus.paren;
3994 if (st->u.plus.paren > (I32)*PL_reglastparen)
3995 *PL_reglastparen = st->u.plus.paren;
3996 st->ln = ARG1(scan); /* min to match */
3997 n = ARG2(scan); /* max to match */
3998 scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE);
4001 st->u.plus.paren = 0;
4002 st->ln = ARG1(scan); /* min to match */
4003 n = ARG2(scan); /* max to match */
4004 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
4009 scan = NEXTOPER(scan);
4010 st->u.plus.paren = 0;
4015 scan = NEXTOPER(scan);
4016 st->u.plus.paren = 0;
4019 * Lookahead to avoid useless match attempts
4020 * when we know what character comes next.
4024 * Used to only do .*x and .*?x, but now it allows
4025 * for )'s, ('s and (?{ ... })'s to be in the way
4026 * of the quantifier and the EXACT-like node. -- japhy
4029 if (HAS_TEXT(next) || JUMPABLE(next)) {
4031 regnode *text_node = next;
4033 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
4035 if (! HAS_TEXT(text_node)) st->u.plus.c1 = st->u.plus.c2 = -1000;
4037 if (PL_regkind[(U8)OP(text_node)] == REF) {
4038 st->u.plus.c1 = st->u.plus.c2 = -1000;
4039 goto assume_ok_easy;
4041 else { s = (U8*)STRING(text_node); }
4044 st->u.plus.c2 = st->u.plus.c1 = *s;
4045 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
4046 st->u.plus.c2 = PL_fold[st->u.plus.c1];
4047 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
4048 st->u.plus.c2 = PL_fold_locale[st->u.plus.c1];
4051 if (OP(text_node) == EXACTF || OP(text_node) == REFF) {
4052 STRLEN ulen1, ulen2;
4053 U8 tmpbuf1[UTF8_MAXBYTES_CASE+1];
4054 U8 tmpbuf2[UTF8_MAXBYTES_CASE+1];
4056 to_utf8_lower((U8*)s, tmpbuf1, &ulen1);
4057 to_utf8_upper((U8*)s, tmpbuf2, &ulen2);
4059 st->u.plus.c1 = utf8n_to_uvuni(tmpbuf1, UTF8_MAXBYTES, 0,
4061 st->u.plus.c2 = utf8n_to_uvuni(tmpbuf2, UTF8_MAXBYTES, 0,
4065 st->u.plus.c2 = st->u.plus.c1 = utf8n_to_uvchr(s, UTF8_MAXBYTES, 0,
4072 st->u.plus.c1 = st->u.plus.c2 = -1000;
4074 PL_reginput = locinput;
4077 if (st->ln && regrepeat(rex, scan, st->ln) < st->ln)
4079 locinput = PL_reginput;
4080 REGCP_SET(st->u.plus.lastcp);
4081 if (st->u.plus.c1 != -1000) {
4082 st->u.plus.old = locinput;
4083 st->u.plus.count = 0;
4085 if (n == REG_INFTY) {
4086 st->u.plus.e = PL_regeol - 1;
4088 while (UTF8_IS_CONTINUATION(*(U8*)st->u.plus.e))
4093 for (st->u.plus.e = locinput;
4094 m >0 && st->u.plus.e + UTF8SKIP(st->u.plus.e) <= PL_regeol; m--)
4095 st->u.plus.e += UTF8SKIP(st->u.plus.e);
4098 st->u.plus.e = locinput + n - st->ln;
4099 if (st->u.plus.e >= PL_regeol)
4100 st->u.plus.e = PL_regeol - 1;
4103 /* Find place 'next' could work */
4105 if (st->u.plus.c1 == st->u.plus.c2) {
4106 while (locinput <= st->u.plus.e &&
4107 UCHARAT(locinput) != st->u.plus.c1)
4110 while (locinput <= st->u.plus.e
4111 && UCHARAT(locinput) != st->u.plus.c1
4112 && UCHARAT(locinput) != st->u.plus.c2)
4115 st->u.plus.count = locinput - st->u.plus.old;
4118 if (st->u.plus.c1 == st->u.plus.c2) {
4120 /* count initialised to
4121 * utf8_distance(old, locinput) */
4122 while (locinput <= st->u.plus.e &&
4123 utf8n_to_uvchr((U8*)locinput,
4124 UTF8_MAXBYTES, &len,
4125 uniflags) != (UV)st->u.plus.c1) {
4131 /* count initialised to
4132 * utf8_distance(old, locinput) */
4133 while (locinput <= st->u.plus.e) {
4134 UV c = utf8n_to_uvchr((U8*)locinput,
4135 UTF8_MAXBYTES, &len,
4137 if (c == (UV)st->u.plus.c1 || c == (UV)st->u.plus.c2)
4144 if (locinput > st->u.plus.e)
4146 /* PL_reginput == old now */
4147 if (locinput != st->u.plus.old) {
4148 st->ln = 1; /* Did some */
4149 if (regrepeat(rex, scan, st->u.plus.count) < st->u.plus.count)
4152 /* PL_reginput == locinput now */
4153 TRYPAREN(st->u.plus.paren, st->ln, locinput, PLUS1);
4154 /*** all unsaved local vars undefined at this point */
4155 PL_reginput = locinput; /* Could be reset... */
4156 REGCP_UNWIND(st->u.plus.lastcp);
4157 /* Couldn't or didn't -- move forward. */
4158 st->u.plus.old = locinput;
4160 locinput += UTF8SKIP(locinput);
4163 st->u.plus.count = 1;
4167 while (n >= st->ln || (n == REG_INFTY && st->ln > 0)) { /* ln overflow ? */
4169 if (st->u.plus.c1 != -1000) {
4171 c = utf8n_to_uvchr((U8*)PL_reginput,
4175 c = UCHARAT(PL_reginput);
4176 /* If it could work, try it. */
4177 if (c == (UV)st->u.plus.c1 || c == (UV)st->u.plus.c2)
4179 TRYPAREN(st->u.plus.paren, st->ln, PL_reginput, PLUS2);
4180 /*** all unsaved local vars undefined at this point */
4181 REGCP_UNWIND(st->u.plus.lastcp);
4184 /* If it could work, try it. */
4185 else if (st->u.plus.c1 == -1000)
4187 TRYPAREN(st->u.plus.paren, st->ln, PL_reginput, PLUS3);
4188 /*** all unsaved local vars undefined at this point */
4189 REGCP_UNWIND(st->u.plus.lastcp);
4191 /* Couldn't or didn't -- move forward. */
4192 PL_reginput = locinput;
4193 if (regrepeat(rex, scan, 1)) {
4195 locinput = PL_reginput;
4202 n = regrepeat(rex, scan, n);
4203 locinput = PL_reginput;
4204 if (st->ln < n && PL_regkind[(U8)OP(next)] == EOL &&
4205 (OP(next) != MEOL ||
4206 OP(next) == SEOL || OP(next) == EOS))
4208 st->ln = n; /* why back off? */
4209 /* ...because $ and \Z can match before *and* after
4210 newline at the end. Consider "\n\n" =~ /\n+\Z\n/.
4211 We should back off by one in this case. */
4212 if (UCHARAT(PL_reginput - 1) == '\n' && OP(next) != EOS)
4215 REGCP_SET(st->u.plus.lastcp);
4218 while (n >= st->ln) {
4219 if (st->u.plus.c1 != -1000) {
4221 c = utf8n_to_uvchr((U8*)PL_reginput,
4225 c = UCHARAT(PL_reginput);
4227 /* If it could work, try it. */
4228 if (st->u.plus.c1 == -1000 || c == (UV)st->u.plus.c1 || c == (UV)st->u.plus.c2)
4230 TRYPAREN(st->u.plus.paren, n, PL_reginput, PLUS4);
4231 /*** all unsaved local vars undefined at this point */
4232 REGCP_UNWIND(st->u.plus.lastcp);
4234 /* Couldn't or didn't -- back up. */
4236 PL_reginput = locinput = HOPc(locinput, -1);
4244 /* we have successfully completed the execution of a
4245 * postponed re. Pop all states back to the last EVAL
4246 * then continue with the node following the (??{...})
4249 /* this simulates a POP_STATE, except that it pops several
4250 * levels, and doesn't restore locinput */
4253 PL_regmatch_slab = st->u.eval.prev_slab;
4254 cur_eval = st->u.eval.prev_eval;
4255 depth = st->u.eval.depth;
4257 PL_regmatch_state = st;
4262 if (st->u.eval.toggleutf)
4263 PL_reg_flags ^= RF_utf8;
4265 rex = st->u.eval.prev_rex;
4266 /* XXXX This is too dramatic a measure... */
4269 /* Restore parens of the caller without popping the
4272 I32 tmp = PL_savestack_ix;
4273 PL_savestack_ix = st->u.eval.lastcp;
4275 PL_savestack_ix = tmp;
4279 PL_reginput = locinput;
4280 /* resume at node following the (??{...}) */
4285 if (locinput < PL_regtill) {
4286 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
4287 "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n",
4289 (long)(locinput - PL_reg_starttry),
4290 (long)(PL_regtill - PL_reg_starttry),
4292 sayNO_FINAL; /* Cannot match: too short. */
4294 PL_reginput = locinput; /* put where regtry can find it */
4295 sayYES_FINAL; /* Success! */
4297 PL_reginput = locinput; /* put where regtry can find it */
4298 sayYES_LOUD; /* Success! */
4301 PL_reginput = locinput;
4306 char * const s = HOPBACKc(locinput, scan->flags);
4312 PL_reginput = locinput;
4317 char * const s = HOPBACKc(locinput, scan->flags);
4323 PL_reginput = locinput;
4326 REGMATCH(NEXTOPER(NEXTOPER(scan)), IFMATCH);
4327 /*** all unsaved local vars undefined at this point */
4343 if (OP(scan) == SUSPEND) {
4344 locinput = PL_reginput;
4345 nextchr = UCHARAT(locinput);
4350 next = scan + ARG(scan);
4355 PerlIO_printf(Perl_error_log, "%"UVxf" %d\n",
4356 PTR2UV(scan), OP(scan));
4357 Perl_croak(aTHX_ "regexp memory corruption");
4365 /* simulate recursively calling regmatch(), but without actually
4366 * recursing - ie save the current state on the heap rather than on
4367 * the stack, then re-enter the loop. This avoids complex regexes
4368 * blowing the processor stack */
4372 /* push new state */
4373 regmatch_state *oldst = st;
4377 /* grab the next free state slot */
4379 if (st > &(PL_regmatch_slab->states[PERL_REGMATCH_SLAB_SLOTS-1]))
4380 st = S_push_slab(aTHX);
4381 PL_regmatch_state = st;
4385 oldst->locinput = locinput;
4388 locinput = PL_reginput;
4389 nextchr = UCHARAT(locinput);
4403 * We get here only if there's trouble -- normally "case END" is
4404 * the terminating point.
4406 Perl_croak(aTHX_ "corrupted regexp pointers");
4412 PerlIO_printf(Perl_debug_log,
4413 "%*s %scould match...%s\n",
4414 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4], PL_colors[5])
4418 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n",
4419 PL_colors[4], PL_colors[5]));
4426 /* XXX this is duplicate(ish) code to that in the do_no section.
4427 * eventually a yes should just pop the whole stack */
4429 /* restore previous state and re-enter */
4432 switch (st->resume_state) {
4434 goto resume_point_TRIE1;
4436 goto resume_point_TRIE2;
4440 goto resume_point_CURLYX;
4441 case resume_WHILEM1:
4442 goto resume_point_WHILEM1;
4443 case resume_WHILEM2:
4444 goto resume_point_WHILEM2;
4445 case resume_WHILEM3:
4446 goto resume_point_WHILEM3;
4447 case resume_WHILEM4:
4448 goto resume_point_WHILEM4;
4449 case resume_WHILEM5:
4450 goto resume_point_WHILEM5;
4451 case resume_WHILEM6:
4452 goto resume_point_WHILEM6;
4453 case resume_CURLYM1:
4454 goto resume_point_CURLYM1;
4455 case resume_CURLYM2:
4456 goto resume_point_CURLYM2;
4457 case resume_CURLYM3:
4458 goto resume_point_CURLYM3;
4459 case resume_CURLYM4:
4460 goto resume_point_CURLYM4;
4461 case resume_IFMATCH:
4462 goto resume_point_IFMATCH;
4464 goto resume_point_PLUS1;
4466 goto resume_point_PLUS2;
4468 goto resume_point_PLUS3;
4470 goto resume_point_PLUS4;
4472 Perl_croak(aTHX_ "regexp resume memory corruption");
4479 PerlIO_printf(Perl_debug_log,
4480 "%*s %sfailed...%s\n",
4481 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4], PL_colors[5])
4487 re_unwind_t * const uw = SSPTRt(st->unwind,re_unwind_t);
4490 case RE_UNWIND_BRANCH:
4491 case RE_UNWIND_BRANCHJ:
4493 re_unwind_branch_t * const uwb = &(uw->branch);
4494 const I32 lastparen = uwb->lastparen;
4496 REGCP_UNWIND(uwb->lastcp);
4497 for (n = *PL_reglastparen; n > lastparen; n--)
4499 *PL_reglastparen = n;
4500 scan = next = uwb->next;
4501 st->minmod = uwb->minmod;
4503 OP(scan) != (uwb->type == RE_UNWIND_BRANCH
4504 ? BRANCH : BRANCHJ) ) { /* Failure */
4505 st->unwind = uwb->prev;
4511 /* Have more choice yet. Reuse the same uwb. */
4512 if ((n = (uwb->type == RE_UNWIND_BRANCH
4513 ? NEXT_OFF(next) : ARG(next))))
4516 next = NULL; /* XXXX Needn't unwinding in this case... */
4518 next = NEXTOPER(scan);
4519 if (uwb->type == RE_UNWIND_BRANCHJ)
4520 next = NEXTOPER(next);
4521 locinput = uwb->locinput;
4522 nextchr = uwb->nextchr;
4524 PL_regindent = uwb->regindent;
4531 Perl_croak(aTHX_ "regexp unwind memory corruption");
4542 /* there's a previous state to backtrack to */
4544 switch (st->resume_state) {
4546 goto resume_point_TRIE1;
4548 goto resume_point_TRIE2;
4550 /* we have failed an (??{...}). Restore state to the outer re
4551 * then re-throw the failure */
4552 if (st->u.eval.toggleutf)
4553 PL_reg_flags ^= RF_utf8;
4555 rex = st->u.eval.prev_rex;
4556 cur_eval = st->u.eval.prev_eval;
4558 /* XXXX This is too dramatic a measure... */
4561 PL_reginput = locinput;
4562 REGCP_UNWIND(st->u.eval.lastcp);
4567 goto resume_point_CURLYX;
4568 case resume_WHILEM1:
4569 goto resume_point_WHILEM1;
4570 case resume_WHILEM2:
4571 goto resume_point_WHILEM2;
4572 case resume_WHILEM3:
4573 goto resume_point_WHILEM3;
4574 case resume_WHILEM4:
4575 goto resume_point_WHILEM4;
4576 case resume_WHILEM5:
4577 goto resume_point_WHILEM5;
4578 case resume_WHILEM6:
4579 goto resume_point_WHILEM6;
4580 case resume_CURLYM1:
4581 goto resume_point_CURLYM1;
4582 case resume_CURLYM2:
4583 goto resume_point_CURLYM2;
4584 case resume_CURLYM3:
4585 goto resume_point_CURLYM3;
4586 case resume_CURLYM4:
4587 goto resume_point_CURLYM4;
4588 case resume_IFMATCH:
4589 goto resume_point_IFMATCH;
4591 goto resume_point_PLUS1;
4593 goto resume_point_PLUS2;
4595 goto resume_point_PLUS3;
4597 goto resume_point_PLUS4;
4599 Perl_croak(aTHX_ "regexp resume memory corruption");
4605 /* restore original high-water mark */
4606 PL_regmatch_slab = orig_slab;
4607 PL_regmatch_state = orig_state;
4609 /* free all slabs above current one */
4610 if (orig_slab->next) {
4611 regmatch_slab *osl, *sl = orig_slab->next;
4612 orig_slab->next = NULL;
4625 - regrepeat - repeatedly match something simple, report how many
4628 * [This routine now assumes that it will only match on things of length 1.
4629 * That was true before, but now we assume scan - reginput is the count,
4630 * rather than incrementing count on every character. [Er, except utf8.]]
4633 S_regrepeat(pTHX_ const regexp *prog, const regnode *p, I32 max)
4636 register char *scan;
4638 register char *loceol = PL_regeol;
4639 register I32 hardcount = 0;
4640 register bool do_utf8 = PL_reg_match_utf8;
4643 if (max == REG_INFTY)
4645 else if (max < loceol - scan)
4646 loceol = scan + max;
4651 while (scan < loceol && hardcount < max && *scan != '\n') {
4652 scan += UTF8SKIP(scan);
4656 while (scan < loceol && *scan != '\n')
4663 while (scan < loceol && hardcount < max) {
4664 scan += UTF8SKIP(scan);
4674 case EXACT: /* length of string is 1 */
4676 while (scan < loceol && UCHARAT(scan) == c)
4679 case EXACTF: /* length of string is 1 */
4681 while (scan < loceol &&
4682 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold[c]))
4685 case EXACTFL: /* length of string is 1 */
4686 PL_reg_flags |= RF_tainted;
4688 while (scan < loceol &&
4689 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold_locale[c]))
4695 while (hardcount < max && scan < loceol &&
4696 reginclass(prog, p, (U8*)scan, 0, do_utf8)) {
4697 scan += UTF8SKIP(scan);
4701 while (scan < loceol && REGINCLASS(prog, p, (U8*)scan))
4708 LOAD_UTF8_CHARCLASS_ALNUM();
4709 while (hardcount < max && scan < loceol &&
4710 swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
4711 scan += UTF8SKIP(scan);
4715 while (scan < loceol && isALNUM(*scan))
4720 PL_reg_flags |= RF_tainted;
4723 while (hardcount < max && scan < loceol &&
4724 isALNUM_LC_utf8((U8*)scan)) {
4725 scan += UTF8SKIP(scan);
4729 while (scan < loceol && isALNUM_LC(*scan))
4736 LOAD_UTF8_CHARCLASS_ALNUM();
4737 while (hardcount < max && scan < loceol &&
4738 !swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
4739 scan += UTF8SKIP(scan);
4743 while (scan < loceol && !isALNUM(*scan))
4748 PL_reg_flags |= RF_tainted;
4751 while (hardcount < max && scan < loceol &&
4752 !isALNUM_LC_utf8((U8*)scan)) {
4753 scan += UTF8SKIP(scan);
4757 while (scan < loceol && !isALNUM_LC(*scan))
4764 LOAD_UTF8_CHARCLASS_SPACE();
4765 while (hardcount < max && scan < loceol &&
4767 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
4768 scan += UTF8SKIP(scan);
4772 while (scan < loceol && isSPACE(*scan))
4777 PL_reg_flags |= RF_tainted;
4780 while (hardcount < max && scan < loceol &&
4781 (*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
4782 scan += UTF8SKIP(scan);
4786 while (scan < loceol && isSPACE_LC(*scan))
4793 LOAD_UTF8_CHARCLASS_SPACE();
4794 while (hardcount < max && scan < loceol &&
4796 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
4797 scan += UTF8SKIP(scan);
4801 while (scan < loceol && !isSPACE(*scan))
4806 PL_reg_flags |= RF_tainted;
4809 while (hardcount < max && scan < loceol &&
4810 !(*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
4811 scan += UTF8SKIP(scan);
4815 while (scan < loceol && !isSPACE_LC(*scan))
4822 LOAD_UTF8_CHARCLASS_DIGIT();
4823 while (hardcount < max && scan < loceol &&
4824 swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
4825 scan += UTF8SKIP(scan);
4829 while (scan < loceol && isDIGIT(*scan))
4836 LOAD_UTF8_CHARCLASS_DIGIT();
4837 while (hardcount < max && scan < loceol &&
4838 !swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
4839 scan += UTF8SKIP(scan);
4843 while (scan < loceol && !isDIGIT(*scan))
4847 default: /* Called on something of 0 width. */
4848 break; /* So match right here or not at all. */
4854 c = scan - PL_reginput;
4858 SV *re_debug_flags = NULL;
4859 SV * const prop = sv_newmortal();
4862 regprop(prog, prop, p);
4863 PerlIO_printf(Perl_debug_log,
4864 "%*s %s can match %"IVdf" times out of %"IVdf"...\n",
4865 REPORT_CODE_OFF+1, "", SvPVX_const(prop),(IV)c,(IV)max);
4874 - regclass_swash - prepare the utf8 swash
4878 Perl_regclass_swash(pTHX_ const regexp *prog, register const regnode* node, bool doinit, SV** listsvp, SV **altsvp)
4884 const struct reg_data *data = prog ? prog->data : NULL;
4886 if (data && data->count) {
4887 const U32 n = ARG(node);
4889 if (data->what[n] == 's') {
4890 SV * const rv = (SV*)data->data[n];
4891 AV * const av = (AV*)SvRV((SV*)rv);
4892 SV **const ary = AvARRAY(av);
4895 /* See the end of regcomp.c:S_regclass() for
4896 * documentation of these array elements. */
4899 a = SvROK(ary[1]) ? &ary[1] : 0;
4900 b = SvTYPE(ary[2]) == SVt_PVAV ? &ary[2] : 0;
4904 else if (si && doinit) {
4905 sw = swash_init("utf8", "", si, 1, 0);
4906 (void)av_store(av, 1, sw);
4922 - reginclass - determine if a character falls into a character class
4924 The n is the ANYOF regnode, the p is the target string, lenp
4925 is pointer to the maximum length of how far to go in the p
4926 (if the lenp is zero, UTF8SKIP(p) is used),
4927 do_utf8 tells whether the target string is in UTF-8.
4932 S_reginclass(pTHX_ const regexp *prog, register const regnode *n, register const U8* p, STRLEN* lenp, register bool do_utf8)
4935 const char flags = ANYOF_FLAGS(n);
4941 if (do_utf8 && !UTF8_IS_INVARIANT(c)) {
4942 c = utf8n_to_uvchr(p, UTF8_MAXBYTES, &len,
4943 (UTF8_ALLOW_DEFAULT & UTF8_ALLOW_ANYUV) | UTF8_CHECK_ONLY);
4944 /* see [perl #37836] for UTF8_ALLOW_ANYUV */
4945 if (len == (STRLEN)-1)
4946 Perl_croak(aTHX_ "Malformed UTF-8 character (fatal)");
4949 plen = lenp ? *lenp : UNISKIP(NATIVE_TO_UNI(c));
4950 if (do_utf8 || (flags & ANYOF_UNICODE)) {
4953 if (do_utf8 && !ANYOF_RUNTIME(n)) {
4954 if (len != (STRLEN)-1 && c < 256 && ANYOF_BITMAP_TEST(n, c))
4957 if (!match && do_utf8 && (flags & ANYOF_UNICODE_ALL) && c >= 256)
4961 SV * const sw = regclass_swash(prog, n, TRUE, 0, (SV**)&av);
4964 if (swash_fetch(sw, p, do_utf8))
4966 else if (flags & ANYOF_FOLD) {
4967 if (!match && lenp && av) {
4969 for (i = 0; i <= av_len(av); i++) {
4970 SV* const sv = *av_fetch(av, i, FALSE);
4972 const char * const s = SvPV_const(sv, len);
4974 if (len <= plen && memEQ(s, (char*)p, len)) {
4982 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
4985 to_utf8_fold(p, tmpbuf, &tmplen);
4986 if (swash_fetch(sw, tmpbuf, do_utf8))
4992 if (match && lenp && *lenp == 0)
4993 *lenp = UNISKIP(NATIVE_TO_UNI(c));
4995 if (!match && c < 256) {
4996 if (ANYOF_BITMAP_TEST(n, c))
4998 else if (flags & ANYOF_FOLD) {
5001 if (flags & ANYOF_LOCALE) {
5002 PL_reg_flags |= RF_tainted;
5003 f = PL_fold_locale[c];
5007 if (f != c && ANYOF_BITMAP_TEST(n, f))
5011 if (!match && (flags & ANYOF_CLASS)) {
5012 PL_reg_flags |= RF_tainted;
5014 (ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) ||
5015 (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) ||
5016 (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) ||
5017 (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) ||
5018 (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) ||
5019 (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) ||
5020 (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) ||
5021 (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) ||
5022 (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) ||
5023 (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) ||
5024 (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII(c)) ||
5025 (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII(c)) ||
5026 (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) ||
5027 (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) ||
5028 (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) ||
5029 (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) ||
5030 (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) ||
5031 (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) ||
5032 (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) ||
5033 (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) ||
5034 (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) ||
5035 (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) ||
5036 (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) ||
5037 (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) ||
5038 (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) ||
5039 (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) ||
5040 (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) ||
5041 (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) ||
5042 (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK(c)) ||
5043 (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK(c))
5044 ) /* How's that for a conditional? */
5051 return (flags & ANYOF_INVERT) ? !match : match;
5055 S_reghop3(U8 *s, I32 off, U8* lim)
5059 while (off-- && s < lim) {
5060 /* XXX could check well-formedness here */
5068 if (UTF8_IS_CONTINUED(*s)) {
5069 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
5072 /* XXX could check well-formedness here */
5080 S_reghopmaybe3(U8* s, I32 off, U8* lim)
5084 while (off-- && s < lim) {
5085 /* XXX could check well-formedness here */
5095 if (UTF8_IS_CONTINUED(*s)) {
5096 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
5099 /* XXX could check well-formedness here */
5111 restore_pos(pTHX_ void *arg)
5114 regexp * const rex = (regexp *)arg;
5115 if (PL_reg_eval_set) {
5116 if (PL_reg_oldsaved) {
5117 rex->subbeg = PL_reg_oldsaved;
5118 rex->sublen = PL_reg_oldsavedlen;
5119 #ifdef PERL_OLD_COPY_ON_WRITE
5120 rex->saved_copy = PL_nrs;
5122 RX_MATCH_COPIED_on(rex);
5124 PL_reg_magic->mg_len = PL_reg_oldpos;
5125 PL_reg_eval_set = 0;
5126 PL_curpm = PL_reg_oldcurpm;
5131 S_to_utf8_substr(pTHX_ register regexp *prog)
5133 if (prog->float_substr && !prog->float_utf8) {
5134 SV* const sv = newSVsv(prog->float_substr);
5135 prog->float_utf8 = sv;
5136 sv_utf8_upgrade(sv);
5137 if (SvTAIL(prog->float_substr))
5139 if (prog->float_substr == prog->check_substr)
5140 prog->check_utf8 = sv;
5142 if (prog->anchored_substr && !prog->anchored_utf8) {
5143 SV* const sv = newSVsv(prog->anchored_substr);
5144 prog->anchored_utf8 = sv;
5145 sv_utf8_upgrade(sv);
5146 if (SvTAIL(prog->anchored_substr))
5148 if (prog->anchored_substr == prog->check_substr)
5149 prog->check_utf8 = sv;
5154 S_to_byte_substr(pTHX_ register regexp *prog)
5157 if (prog->float_utf8 && !prog->float_substr) {
5158 SV* sv = newSVsv(prog->float_utf8);
5159 prog->float_substr = sv;
5160 if (sv_utf8_downgrade(sv, TRUE)) {
5161 if (SvTAIL(prog->float_utf8))
5165 prog->float_substr = sv = &PL_sv_undef;
5167 if (prog->float_utf8 == prog->check_utf8)
5168 prog->check_substr = sv;
5170 if (prog->anchored_utf8 && !prog->anchored_substr) {
5171 SV* sv = newSVsv(prog->anchored_utf8);
5172 prog->anchored_substr = sv;
5173 if (sv_utf8_downgrade(sv, TRUE)) {
5174 if (SvTAIL(prog->anchored_utf8))
5178 prog->anchored_substr = sv = &PL_sv_undef;
5180 if (prog->anchored_utf8 == prog->check_utf8)
5181 prog->check_substr = sv;
5187 * c-indentation-style: bsd
5189 * indent-tabs-mode: t
5192 * ex: set ts=8 sts=4 sw=4 noet: