5 * "One Ring to rule them all, One Ring to find them..."
8 /* NOTE: this is derived from Henry Spencer's regexp code, and should not
9 * confused with the original package (see point 3 below). Thanks, Henry!
12 /* Additional note: this code is very heavily munged from Henry's version
13 * in places. In some spots I've traded clarity for efficiency, so don't
14 * blame Henry for some of the lack of readability.
17 /* The names of the functions have been changed from regcomp and
18 * regexec to pregcomp and pregexec in order to avoid conflicts
19 * with the POSIX routines of the same names.
22 #ifdef PERL_EXT_RE_BUILD
23 /* need to replace pregcomp et al, so enable that */
24 # ifndef PERL_IN_XSUB_RE
25 # define PERL_IN_XSUB_RE
27 /* need access to debugger hooks */
28 # if defined(PERL_EXT_RE_DEBUG) && !defined(DEBUGGING)
33 #ifdef PERL_IN_XSUB_RE
34 /* We *really* need to overwrite these symbols: */
35 # define Perl_regexec_flags my_regexec
36 # define Perl_regdump my_regdump
37 # define Perl_regprop my_regprop
38 # define Perl_re_intuit_start my_re_intuit_start
39 /* *These* symbols are masked to allow static link. */
40 # define Perl_pregexec my_pregexec
41 # define Perl_reginitcolors my_reginitcolors
42 # define Perl_regclass_swash my_regclass_swash
44 # define PERL_NO_GET_CONTEXT
49 * pregcomp and pregexec -- regsub and regerror are not used in perl
51 * Copyright (c) 1986 by University of Toronto.
52 * Written by Henry Spencer. Not derived from licensed software.
54 * Permission is granted to anyone to use this software for any
55 * purpose on any computer system, and to redistribute it freely,
56 * subject to the following restrictions:
58 * 1. The author is not responsible for the consequences of use of
59 * this software, no matter how awful, even if they arise
62 * 2. The origin of this software must not be misrepresented, either
63 * by explicit claim or by omission.
65 * 3. Altered versions must be plainly marked as such, and must not
66 * be misrepresented as being the original software.
68 **** Alterations to Henry's code are...
70 **** Copyright (c) 1991-2001, Larry Wall
72 **** You may distribute under the terms of either the GNU General Public
73 **** License or the Artistic License, as specified in the README file.
75 * Beware that some of this code is subtly aware of the way operator
76 * precedence is structured in regular expressions. Serious changes in
77 * regular-expression syntax might require a total rethink.
80 #define PERL_IN_REGEXEC_C
83 #ifdef PERL_IN_XSUB_RE
84 # if defined(PERL_CAPI) || defined(PERL_OBJECT)
91 #define RF_tainted 1 /* tainted information used? */
92 #define RF_warned 2 /* warned about big count? */
93 #define RF_evaled 4 /* Did an EVAL with setting? */
94 #define RF_utf8 8 /* String contains multibyte chars? */
96 #define UTF (PL_reg_flags & RF_utf8)
98 #define RS_init 1 /* eval environment created */
99 #define RS_set 2 /* replsv value is set */
102 #define STATIC static
109 #define CHR_SVLEN(sv) (UTF ? sv_len_utf8(sv) : SvCUR(sv))
110 #define CHR_DIST(a,b) (DO_UTF8(PL_reg_sv) ? utf8_distance(a,b) : a - b)
112 #define reghop_c(pos,off) ((char*)reghop((U8*)pos, off))
113 #define reghopmaybe_c(pos,off) ((char*)reghopmaybe((U8*)pos, off))
114 #define HOP(pos,off) (DO_UTF8(PL_reg_sv) ? reghop((U8*)pos, off) : (U8*)(pos + off))
115 #define HOPMAYBE(pos,off) (DO_UTF8(PL_reg_sv) ? reghopmaybe((U8*)pos, off) : (U8*)(pos + off))
116 #define HOPc(pos,off) ((char*)HOP(pos,off))
117 #define HOPMAYBEc(pos,off) ((char*)HOPMAYBE(pos,off))
119 #define HOPBACK(pos, off) ( \
120 (UTF && DO_UTF8(PL_reg_sv)) \
121 ? reghopmaybe((U8*)pos, -off) \
122 : (pos - off >= PL_bostr) \
126 #define HOPBACKc(pos, off) (char*)HOPBACK(pos, off)
128 #define reghop3_c(pos,off,lim) ((char*)reghop3((U8*)pos, off, (U8*)lim))
129 #define reghopmaybe3_c(pos,off,lim) ((char*)reghopmaybe3((U8*)pos, off, (U8*)lim))
130 #define HOP3(pos,off,lim) (DO_UTF8(PL_reg_sv) ? reghop3((U8*)pos, off, (U8*)lim) : (U8*)(pos + off))
131 #define HOPMAYBE3(pos,off,lim) (DO_UTF8(PL_reg_sv) ? reghopmaybe3((U8*)pos, off, (U8*)lim) : (U8*)(pos + off))
132 #define HOP3c(pos,off,lim) ((char*)HOP3(pos,off,lim))
133 #define HOPMAYBE3c(pos,off,lim) ((char*)HOPMAYBE3(pos,off,lim))
135 #define LOAD_UTF8_CHARCLASS(a,b) STMT_START { if (!CAT2(PL_utf8_,a)) (void)CAT2(is_utf8_, a)((U8*)b); } STMT_END
137 static void restore_pos(pTHXo_ void *arg);
140 S_regcppush(pTHX_ I32 parenfloor)
142 int retval = PL_savestack_ix;
143 #define REGCP_PAREN_ELEMS 4
144 int paren_elems_to_push = (PL_regsize - parenfloor) * REGCP_PAREN_ELEMS;
147 if (paren_elems_to_push < 0)
148 Perl_croak(aTHX_ "panic: paren_elems_to_push < 0");
150 #define REGCP_OTHER_ELEMS 6
151 SSCHECK(paren_elems_to_push + REGCP_OTHER_ELEMS);
152 for (p = PL_regsize; p > parenfloor; p--) {
153 /* REGCP_PARENS_ELEMS are pushed per pairs of parentheses. */
154 SSPUSHINT(PL_regendp[p]);
155 SSPUSHINT(PL_regstartp[p]);
156 SSPUSHPTR(PL_reg_start_tmp[p]);
159 /* REGCP_OTHER_ELEMS are pushed in any case, parentheses or no. */
160 SSPUSHINT(PL_regsize);
161 SSPUSHINT(*PL_reglastparen);
162 SSPUSHINT(*PL_reglastcloseparen);
163 SSPUSHPTR(PL_reginput);
164 #define REGCP_FRAME_ELEMS 2
165 /* REGCP_FRAME_ELEMS are part of the REGCP_OTHER_ELEMS and
166 * are needed for the regexp context stack bookkeeping. */
167 SSPUSHINT(paren_elems_to_push + REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS);
168 SSPUSHINT(SAVEt_REGCONTEXT); /* Magic cookie. */
173 /* These are needed since we do not localize EVAL nodes: */
174 # define REGCP_SET(cp) DEBUG_r(PerlIO_printf(Perl_debug_log, \
175 " Setting an EVAL scope, savestack=%"IVdf"\n", \
176 (IV)PL_savestack_ix)); cp = PL_savestack_ix
178 # define REGCP_UNWIND(cp) DEBUG_r(cp != PL_savestack_ix ? \
179 PerlIO_printf(Perl_debug_log, \
180 " Clearing an EVAL scope, savestack=%"IVdf"..%"IVdf"\n", \
181 (IV)(cp), (IV)PL_savestack_ix) : 0); regcpblow(cp)
191 /* Pop REGCP_OTHER_ELEMS before the parentheses loop starts. */
193 assert(i == SAVEt_REGCONTEXT); /* Check that the magic cookie is there. */
194 i = SSPOPINT; /* Parentheses elements to pop. */
195 input = (char *) SSPOPPTR;
196 *PL_reglastcloseparen = SSPOPINT;
197 *PL_reglastparen = SSPOPINT;
198 PL_regsize = SSPOPINT;
200 /* Now restore the parentheses context. */
201 for (i -= (REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS);
202 i > 0; i -= REGCP_PAREN_ELEMS) {
203 paren = (U32)SSPOPINT;
204 PL_reg_start_tmp[paren] = (char *) SSPOPPTR;
205 PL_regstartp[paren] = SSPOPINT;
207 if (paren <= *PL_reglastparen)
208 PL_regendp[paren] = tmps;
210 PerlIO_printf(Perl_debug_log,
211 " restoring \\%"UVuf" to %"IVdf"(%"IVdf")..%"IVdf"%s\n",
212 (UV)paren, (IV)PL_regstartp[paren],
213 (IV)(PL_reg_start_tmp[paren] - PL_bostr),
214 (IV)PL_regendp[paren],
215 (paren > *PL_reglastparen ? "(no)" : ""));
219 if (*PL_reglastparen + 1 <= PL_regnpar) {
220 PerlIO_printf(Perl_debug_log,
221 " restoring \\%"IVdf"..\\%"IVdf" to undef\n",
222 (IV)(*PL_reglastparen + 1), (IV)PL_regnpar);
226 /* It would seem that the similar code in regtry()
227 * already takes care of this, and in fact it is in
228 * a better location to since this code can #if 0-ed out
229 * but the code in regtry() is needed or otherwise tests
230 * requiring null fields (pat.t#187 and split.t#{13,14}
231 * (as of patchlevel 7877) will fail. Then again,
232 * this code seems to be necessary or otherwise
233 * building DynaLoader will fail:
234 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
236 for (paren = *PL_reglastparen + 1; paren <= PL_regnpar; paren++) {
237 if (paren > PL_regsize)
238 PL_regstartp[paren] = -1;
239 PL_regendp[paren] = -1;
246 S_regcp_set_to(pTHX_ I32 ss)
248 I32 tmp = PL_savestack_ix;
250 PL_savestack_ix = ss;
252 PL_savestack_ix = tmp;
256 typedef struct re_cc_state
260 struct re_cc_state *prev;
265 #define regcpblow(cp) LEAVE_SCOPE(cp) /* Ignores regcppush()ed data. */
267 #define TRYPAREN(paren, n, input) { \
270 PL_regstartp[paren] = HOPc(input, -1) - PL_bostr; \
271 PL_regendp[paren] = input - PL_bostr; \
274 PL_regendp[paren] = -1; \
276 if (regmatch(next)) \
279 PL_regendp[paren] = -1; \
284 * pregexec and friends
288 - pregexec - match a regexp against a string
291 Perl_pregexec(pTHX_ register regexp *prog, char *stringarg, register char *strend,
292 char *strbeg, I32 minend, SV *screamer, U32 nosave)
293 /* strend: pointer to null at end of string */
294 /* strbeg: real beginning of string */
295 /* minend: end of match must be >=minend after stringarg. */
296 /* nosave: For optimizations. */
299 regexec_flags(prog, stringarg, strend, strbeg, minend, screamer, NULL,
300 nosave ? 0 : REXEC_COPY_STR);
304 S_cache_re(pTHX_ regexp *prog)
306 PL_regprecomp = prog->precomp; /* Needed for FAIL. */
308 PL_regprogram = prog->program;
310 PL_regnpar = prog->nparens;
311 PL_regdata = prog->data;
316 * Need to implement the following flags for reg_anch:
318 * USE_INTUIT_NOML - Useful to call re_intuit_start() first
320 * INTUIT_AUTORITATIVE_NOML - Can trust a positive answer
321 * INTUIT_AUTORITATIVE_ML
322 * INTUIT_ONCE_NOML - Intuit can match in one location only.
325 * Another flag for this function: SECOND_TIME (so that float substrs
326 * with giant delta may be not rechecked).
329 /* Assumptions: if ANCH_GPOS, then strpos is anchored. XXXX Check GPOS logic */
331 /* If SCREAM, then SvPVX(sv) should be compatible with strpos and strend.
332 Otherwise, only SvCUR(sv) is used to get strbeg. */
334 /* XXXX We assume that strpos is strbeg unless sv. */
336 /* XXXX Some places assume that there is a fixed substring.
337 An update may be needed if optimizer marks as "INTUITable"
338 RExen without fixed substrings. Similarly, it is assumed that
339 lengths of all the strings are no more than minlen, thus they
340 cannot come from lookahead.
341 (Or minlen should take into account lookahead.) */
343 /* A failure to find a constant substring means that there is no need to make
344 an expensive call to REx engine, thus we celebrate a failure. Similarly,
345 finding a substring too deep into the string means that less calls to
346 regtry() should be needed.
348 REx compiler's optimizer found 4 possible hints:
349 a) Anchored substring;
351 c) Whether we are anchored (beginning-of-line or \G);
352 d) First node (of those at offset 0) which may distingush positions;
353 We use a)b)d) and multiline-part of c), and try to find a position in the
354 string which does not contradict any of them.
357 /* Most of decisions we do here should have been done at compile time.
358 The nodes of the REx which we used for the search should have been
359 deleted from the finite automaton. */
362 Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos,
363 char *strend, U32 flags, re_scream_pos_data *data)
365 register I32 start_shift = 0;
366 /* Should be nonnegative! */
367 register I32 end_shift = 0;
373 register char *other_last = Nullch; /* other substr checked before this */
374 char *check_at = Nullch; /* check substr found at this pos */
376 char *i_strpos = strpos;
379 DEBUG_r( if (!PL_colorset) reginitcolors() );
380 DEBUG_r(PerlIO_printf(Perl_debug_log,
381 "%sGuessing start of match, REx%s `%s%.60s%s%s' against `%s%.*s%s%s'...\n",
382 PL_colors[4],PL_colors[5],PL_colors[0],
385 (strlen(prog->precomp) > 60 ? "..." : ""),
387 (int)(strend - strpos > 60 ? 60 : strend - strpos),
388 strpos, PL_colors[1],
389 (strend - strpos > 60 ? "..." : ""))
392 if (prog->reganch & ROPT_UTF8)
393 PL_reg_flags |= RF_utf8;
395 if (prog->minlen > CHR_DIST((U8*)strend, (U8*)strpos)) {
396 DEBUG_r(PerlIO_printf(Perl_debug_log, "String too short...\n"));
399 strbeg = (sv && SvPOK(sv)) ? strend - SvCUR(sv) : strpos;
401 check = prog->check_substr;
402 if (prog->reganch & ROPT_ANCH) { /* Match at beg-of-str or after \n */
403 ml_anch = !( (prog->reganch & ROPT_ANCH_SINGLE)
404 || ( (prog->reganch & ROPT_ANCH_BOL)
405 && !PL_multiline ) ); /* Check after \n? */
408 if ( !(prog->reganch & (ROPT_ANCH_GPOS /* Checked by the caller */
409 | ROPT_IMPLICIT)) /* not a real BOL */
410 /* SvCUR is not set on references: SvRV and SvPVX overlap */
412 && (strpos != strbeg)) {
413 DEBUG_r(PerlIO_printf(Perl_debug_log, "Not at start...\n"));
416 if (prog->check_offset_min == prog->check_offset_max &&
417 !(prog->reganch & ROPT_SANY_SEEN)) {
418 /* Substring at constant offset from beg-of-str... */
421 s = HOP3c(strpos, prog->check_offset_min, strend);
423 slen = SvCUR(check); /* >= 1 */
425 if ( strend - s > slen || strend - s < slen - 1
426 || (strend - s == slen && strend[-1] != '\n')) {
427 DEBUG_r(PerlIO_printf(Perl_debug_log, "String too long...\n"));
430 /* Now should match s[0..slen-2] */
432 if (slen && (*SvPVX(check) != *s
434 && memNE(SvPVX(check), s, slen)))) {
436 DEBUG_r(PerlIO_printf(Perl_debug_log, "String not equal...\n"));
440 else if (*SvPVX(check) != *s
441 || ((slen = SvCUR(check)) > 1
442 && memNE(SvPVX(check), s, slen)))
444 goto success_at_start;
447 /* Match is anchored, but substr is not anchored wrt beg-of-str. */
449 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
450 end_shift = prog->minlen - start_shift -
451 CHR_SVLEN(check) + (SvTAIL(check) != 0);
453 I32 end = prog->check_offset_max + CHR_SVLEN(check)
454 - (SvTAIL(check) != 0);
455 I32 eshift = CHR_DIST((U8*)strend, (U8*)s) - end;
457 if (end_shift < eshift)
461 else { /* Can match at random position */
464 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
465 /* Should be nonnegative! */
466 end_shift = prog->minlen - start_shift -
467 CHR_SVLEN(check) + (SvTAIL(check) != 0);
470 #ifdef DEBUGGING /* 7/99: reports of failure (with the older version) */
472 Perl_croak(aTHX_ "panic: end_shift");
476 /* Find a possible match in the region s..strend by looking for
477 the "check" substring in the region corrected by start/end_shift. */
478 if (flags & REXEC_SCREAM) {
479 I32 p = -1; /* Internal iterator of scream. */
480 I32 *pp = data ? data->scream_pos : &p;
482 if (PL_screamfirst[BmRARE(check)] >= 0
483 || ( BmRARE(check) == '\n'
484 && (BmPREVIOUS(check) == SvCUR(check) - 1)
486 s = screaminstr(sv, check,
487 start_shift + (s - strbeg), end_shift, pp, 0);
491 *data->scream_olds = s;
493 else if (prog->reganch & ROPT_SANY_SEEN)
494 s = fbm_instr((U8*)(s + start_shift),
495 (U8*)(strend - end_shift),
496 check, PL_multiline ? FBMrf_MULTILINE : 0);
498 s = fbm_instr(HOP3(s, start_shift, strend),
499 HOP3(strend, -end_shift, strbeg),
500 check, PL_multiline ? FBMrf_MULTILINE : 0);
502 /* Update the count-of-usability, remove useless subpatterns,
505 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s %s substr `%s%.*s%s'%s%s",
506 (s ? "Found" : "Did not find"),
507 ((check == prog->anchored_substr) ? "anchored" : "floating"),
509 (int)(SvCUR(check) - (SvTAIL(check)!=0)),
511 PL_colors[1], (SvTAIL(check) ? "$" : ""),
512 (s ? " at offset " : "...\n") ) );
519 /* Finish the diagnostic message */
520 DEBUG_r(PerlIO_printf(Perl_debug_log, "%ld...\n", (long)(s - i_strpos)) );
522 /* Got a candidate. Check MBOL anchoring, and the *other* substr.
523 Start with the other substr.
524 XXXX no SCREAM optimization yet - and a very coarse implementation
525 XXXX /ttx+/ results in anchored=`ttx', floating=`x'. floating will
526 *always* match. Probably should be marked during compile...
527 Probably it is right to do no SCREAM here...
530 if (prog->float_substr && prog->anchored_substr) {
531 /* Take into account the "other" substring. */
532 /* XXXX May be hopelessly wrong for UTF... */
535 if (check == prog->float_substr) {
538 char *last = HOP3c(s, -start_shift, strbeg), *last1, *last2;
541 t = s - prog->check_offset_max;
542 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
543 && (!(prog->reganch & ROPT_UTF8)
544 || ((t = reghopmaybe3_c(s, -(prog->check_offset_max), strpos))
549 t = HOP3c(t, prog->anchored_offset, strend);
550 if (t < other_last) /* These positions already checked */
552 last2 = last1 = HOP3c(strend, -prog->minlen, strbeg);
555 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
556 /* On end-of-str: see comment below. */
557 s = fbm_instr((unsigned char*)t,
558 HOP3(HOP3(last1, prog->anchored_offset, strend)
559 + SvCUR(prog->anchored_substr),
560 -(SvTAIL(prog->anchored_substr)!=0), strbeg),
561 prog->anchored_substr,
562 PL_multiline ? FBMrf_MULTILINE : 0);
563 DEBUG_r(PerlIO_printf(Perl_debug_log,
564 "%s anchored substr `%s%.*s%s'%s",
565 (s ? "Found" : "Contradicts"),
567 (int)(SvCUR(prog->anchored_substr)
568 - (SvTAIL(prog->anchored_substr)!=0)),
569 SvPVX(prog->anchored_substr),
570 PL_colors[1], (SvTAIL(prog->anchored_substr) ? "$" : "")));
572 if (last1 >= last2) {
573 DEBUG_r(PerlIO_printf(Perl_debug_log,
574 ", giving up...\n"));
577 DEBUG_r(PerlIO_printf(Perl_debug_log,
578 ", trying floating at offset %ld...\n",
579 (long)(HOP3c(s1, 1, strend) - i_strpos)));
580 other_last = HOP3c(last1, prog->anchored_offset+1, strend);
581 s = HOP3c(last, 1, strend);
585 DEBUG_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
586 (long)(s - i_strpos)));
587 t = HOP3c(s, -prog->anchored_offset, strbeg);
588 other_last = HOP3c(s, 1, strend);
596 else { /* Take into account the floating substring. */
600 t = HOP3c(s, -start_shift, strbeg);
602 HOP3c(strend, -prog->minlen + prog->float_min_offset, strbeg);
603 if (CHR_DIST((U8*)last, (U8*)t) > prog->float_max_offset)
604 last = HOP3c(t, prog->float_max_offset, strend);
605 s = HOP3c(t, prog->float_min_offset, strend);
608 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
609 /* fbm_instr() takes into account exact value of end-of-str
610 if the check is SvTAIL(ed). Since false positives are OK,
611 and end-of-str is not later than strend we are OK. */
612 s = fbm_instr((unsigned char*)s,
613 (unsigned char*)last + SvCUR(prog->float_substr)
614 - (SvTAIL(prog->float_substr)!=0),
615 prog->float_substr, PL_multiline ? FBMrf_MULTILINE : 0);
616 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s floating substr `%s%.*s%s'%s",
617 (s ? "Found" : "Contradicts"),
619 (int)(SvCUR(prog->float_substr)
620 - (SvTAIL(prog->float_substr)!=0)),
621 SvPVX(prog->float_substr),
622 PL_colors[1], (SvTAIL(prog->float_substr) ? "$" : "")));
625 DEBUG_r(PerlIO_printf(Perl_debug_log,
626 ", giving up...\n"));
629 DEBUG_r(PerlIO_printf(Perl_debug_log,
630 ", trying anchored starting at offset %ld...\n",
631 (long)(s1 + 1 - i_strpos)));
633 s = HOP3c(t, 1, strend);
637 DEBUG_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
638 (long)(s - i_strpos)));
639 other_last = s; /* Fix this later. --Hugo */
648 t = s - prog->check_offset_max;
649 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
650 && (!(prog->reganch & ROPT_UTF8)
651 || ((t = reghopmaybe3_c(s, -prog->check_offset_max, strpos))
653 /* Fixed substring is found far enough so that the match
654 cannot start at strpos. */
656 if (ml_anch && t[-1] != '\n') {
657 /* Eventually fbm_*() should handle this, but often
658 anchored_offset is not 0, so this check will not be wasted. */
659 /* XXXX In the code below we prefer to look for "^" even in
660 presence of anchored substrings. And we search even
661 beyond the found float position. These pessimizations
662 are historical artefacts only. */
664 while (t < strend - prog->minlen) {
666 if (t < check_at - prog->check_offset_min) {
667 if (prog->anchored_substr) {
668 /* Since we moved from the found position,
669 we definitely contradict the found anchored
670 substr. Due to the above check we do not
671 contradict "check" substr.
672 Thus we can arrive here only if check substr
673 is float. Redo checking for "other"=="fixed".
676 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld, rescanning for anchored from offset %ld...\n",
677 PL_colors[0],PL_colors[1], (long)(strpos - i_strpos), (long)(strpos - i_strpos + prog->anchored_offset)));
678 goto do_other_anchored;
680 /* We don't contradict the found floating substring. */
681 /* XXXX Why not check for STCLASS? */
683 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld...\n",
684 PL_colors[0],PL_colors[1], (long)(s - i_strpos)));
687 /* Position contradicts check-string */
688 /* XXXX probably better to look for check-string
689 than for "\n", so one should lower the limit for t? */
690 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m, restarting lookup for check-string at offset %ld...\n",
691 PL_colors[0],PL_colors[1], (long)(t + 1 - i_strpos)));
692 other_last = strpos = s = t + 1;
697 DEBUG_r(PerlIO_printf(Perl_debug_log, "Did not find /%s^%s/m...\n",
698 PL_colors[0],PL_colors[1]));
702 DEBUG_r(PerlIO_printf(Perl_debug_log, "Starting position does not contradict /%s^%s/m...\n",
703 PL_colors[0],PL_colors[1]));
707 ++BmUSEFUL(prog->check_substr); /* hooray/5 */
710 /* The found string does not prohibit matching at strpos,
711 - no optimization of calling REx engine can be performed,
712 unless it was an MBOL and we are not after MBOL,
713 or a future STCLASS check will fail this. */
715 /* Even in this situation we may use MBOL flag if strpos is offset
716 wrt the start of the string. */
717 if (ml_anch && sv && !SvROK(sv) /* See prev comment on SvROK */
718 && (strpos != strbeg) && strpos[-1] != '\n'
719 /* May be due to an implicit anchor of m{.*foo} */
720 && !(prog->reganch & ROPT_IMPLICIT))
725 DEBUG_r( if (ml_anch)
726 PerlIO_printf(Perl_debug_log, "Position at offset %ld does not contradict /%s^%s/m...\n",
727 (long)(strpos - i_strpos), PL_colors[0],PL_colors[1]);
730 if (!(prog->reganch & ROPT_NAUGHTY) /* XXXX If strpos moved? */
731 && prog->check_substr /* Could be deleted already */
732 && --BmUSEFUL(prog->check_substr) < 0
733 && prog->check_substr == prog->float_substr)
735 /* If flags & SOMETHING - do not do it many times on the same match */
736 DEBUG_r(PerlIO_printf(Perl_debug_log, "... Disabling check substring...\n"));
737 SvREFCNT_dec(prog->check_substr);
738 prog->check_substr = Nullsv; /* disable */
739 prog->float_substr = Nullsv; /* clear */
740 check = Nullsv; /* abort */
742 /* XXXX This is a remnant of the old implementation. It
743 looks wasteful, since now INTUIT can use many
745 prog->reganch &= ~RE_USE_INTUIT;
752 /* XXXX BmUSEFUL already changed, maybe multiple change is meaningful... */
753 if (prog->regstclass) {
754 /* minlen == 0 is possible if regstclass is \b or \B,
755 and the fixed substr is ''$.
756 Since minlen is already taken into account, s+1 is before strend;
757 accidentally, minlen >= 1 guaranties no false positives at s + 1
758 even for \b or \B. But (minlen? 1 : 0) below assumes that
759 regstclass does not come from lookahead... */
760 /* If regstclass takes bytelength more than 1: If charlength==1, OK.
761 This leaves EXACTF only, which is dealt with in find_byclass(). */
762 U8* str = (U8*)STRING(prog->regstclass);
763 int cl_l = (PL_regkind[(U8)OP(prog->regstclass)] == EXACT
764 ? CHR_DIST(str+STR_LEN(prog->regstclass), str)
766 char *endpos = (prog->anchored_substr || ml_anch)
767 ? HOP3c(s, (prog->minlen ? cl_l : 0), strend)
768 : (prog->float_substr
769 ? HOP3c(HOP3c(check_at, -start_shift, strbeg),
772 char *startpos = strbeg;
775 if (prog->reganch & ROPT_UTF8) {
776 PL_regdata = prog->data;
779 s = find_byclass(prog, prog->regstclass, s, endpos, startpos, 1);
784 if (endpos == strend) {
785 DEBUG_r( PerlIO_printf(Perl_debug_log,
786 "Could not match STCLASS...\n") );
789 DEBUG_r( PerlIO_printf(Perl_debug_log,
790 "This position contradicts STCLASS...\n") );
791 if ((prog->reganch & ROPT_ANCH) && !ml_anch)
793 /* Contradict one of substrings */
794 if (prog->anchored_substr) {
795 if (prog->anchored_substr == check) {
796 DEBUG_r( what = "anchored" );
798 s = HOP3c(t, 1, strend);
799 if (s + start_shift + end_shift > strend) {
800 /* XXXX Should be taken into account earlier? */
801 DEBUG_r( PerlIO_printf(Perl_debug_log,
802 "Could not match STCLASS...\n") );
807 DEBUG_r( PerlIO_printf(Perl_debug_log,
808 "Looking for %s substr starting at offset %ld...\n",
809 what, (long)(s + start_shift - i_strpos)) );
812 /* Have both, check_string is floating */
813 if (t + start_shift >= check_at) /* Contradicts floating=check */
814 goto retry_floating_check;
815 /* Recheck anchored substring, but not floating... */
819 DEBUG_r( PerlIO_printf(Perl_debug_log,
820 "Looking for anchored substr starting at offset %ld...\n",
821 (long)(other_last - i_strpos)) );
822 goto do_other_anchored;
824 /* Another way we could have checked stclass at the
825 current position only: */
830 DEBUG_r( PerlIO_printf(Perl_debug_log,
831 "Looking for /%s^%s/m starting at offset %ld...\n",
832 PL_colors[0],PL_colors[1], (long)(t - i_strpos)) );
835 if (!prog->float_substr) /* Could have been deleted */
837 /* Check is floating subtring. */
838 retry_floating_check:
839 t = check_at - start_shift;
840 DEBUG_r( what = "floating" );
841 goto hop_and_restart;
844 DEBUG_r(PerlIO_printf(Perl_debug_log,
845 "By STCLASS: moving %ld --> %ld\n",
846 (long)(t - i_strpos), (long)(s - i_strpos))
850 DEBUG_r(PerlIO_printf(Perl_debug_log,
851 "Does not contradict STCLASS...\n");
856 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s%s:%s match at offset %ld\n",
857 PL_colors[4], (check ? "Guessed" : "Giving up"),
858 PL_colors[5], (long)(s - i_strpos)) );
861 fail_finish: /* Substring not found */
862 if (prog->check_substr) /* could be removed already */
863 BmUSEFUL(prog->check_substr) += 5; /* hooray */
865 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch rejected by optimizer%s\n",
866 PL_colors[4],PL_colors[5]));
870 /* We know what class REx starts with. Try to find this position... */
872 S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, char *strend, char *startpos, I32 norun)
874 I32 doevery = (prog->reganch & ROPT_SKIP) == 0;
880 register I32 tmp = 1; /* Scratch variable? */
881 register bool do_utf8 = DO_UTF8(PL_reg_sv);
883 /* We know what class it must start with. */
887 if (reginclass(c, (U8*)s, do_utf8)) {
888 if (tmp && (norun || regtry(prog, s)))
895 s += do_utf8 ? UTF8SKIP(s) : 1;
902 c1 = to_utf8_lower((U8*)m);
903 c2 = to_utf8_upper((U8*)m);
914 c2 = PL_fold_locale[c1];
919 e = s; /* Due to minlen logic of intuit() */
925 if ( utf8_to_uvchr((U8*)s, &len) == c1
932 UV c = utf8_to_uvchr((U8*)s, &len);
933 if ( (c == c1 || c == c2) && regtry(prog, s) )
942 && (ln == 1 || !(OP(c) == EXACTF
944 : ibcmp_locale(s, m, ln)))
945 && (norun || regtry(prog, s)) )
951 if ( (*(U8*)s == c1 || *(U8*)s == c2)
952 && (ln == 1 || !(OP(c) == EXACTF
954 : ibcmp_locale(s, m, ln)))
955 && (norun || regtry(prog, s)) )
962 PL_reg_flags |= RF_tainted;
969 U8 *r = reghop3((U8*)s, -1, (U8*)startpos);
972 tmp = (I32)utf8n_to_uvchr(r, s - (char*)r, 0, 0);
974 tmp = ((OP(c) == BOUND ?
975 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
976 LOAD_UTF8_CHARCLASS(alnum,"a");
978 if (tmp == !(OP(c) == BOUND ?
979 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
980 isALNUM_LC_utf8((U8*)s)))
983 if ((norun || regtry(prog, s)))
990 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
991 tmp = ((OP(c) == BOUND ? isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
994 !(OP(c) == BOUND ? isALNUM(*s) : isALNUM_LC(*s))) {
996 if ((norun || regtry(prog, s)))
1002 if ((!prog->minlen && tmp) && (norun || regtry(prog, s)))
1006 PL_reg_flags |= RF_tainted;
1013 U8 *r = reghop3((U8*)s, -1, (U8*)startpos);
1016 tmp = (I32)utf8n_to_uvchr(r, s - (char*)r, 0, 0);
1018 tmp = ((OP(c) == NBOUND ?
1019 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1020 LOAD_UTF8_CHARCLASS(alnum,"a");
1021 while (s < strend) {
1022 if (tmp == !(OP(c) == NBOUND ?
1023 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1024 isALNUM_LC_utf8((U8*)s)))
1026 else if ((norun || regtry(prog, s)))
1032 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1033 tmp = ((OP(c) == NBOUND ?
1034 isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1035 while (s < strend) {
1037 !(OP(c) == NBOUND ? isALNUM(*s) : isALNUM_LC(*s)))
1039 else if ((norun || regtry(prog, s)))
1044 if ((!prog->minlen && !tmp) && (norun || regtry(prog, s)))
1049 LOAD_UTF8_CHARCLASS(alnum,"a");
1050 while (s < strend) {
1051 if (swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1052 if (tmp && (norun || regtry(prog, s)))
1063 while (s < strend) {
1065 if (tmp && (norun || regtry(prog, s)))
1077 PL_reg_flags |= RF_tainted;
1079 while (s < strend) {
1080 if (isALNUM_LC_utf8((U8*)s)) {
1081 if (tmp && (norun || regtry(prog, s)))
1092 while (s < strend) {
1093 if (isALNUM_LC(*s)) {
1094 if (tmp && (norun || regtry(prog, s)))
1107 LOAD_UTF8_CHARCLASS(alnum,"a");
1108 while (s < strend) {
1109 if (!swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1110 if (tmp && (norun || regtry(prog, s)))
1121 while (s < strend) {
1123 if (tmp && (norun || regtry(prog, s)))
1135 PL_reg_flags |= RF_tainted;
1137 while (s < strend) {
1138 if (!isALNUM_LC_utf8((U8*)s)) {
1139 if (tmp && (norun || regtry(prog, s)))
1150 while (s < strend) {
1151 if (!isALNUM_LC(*s)) {
1152 if (tmp && (norun || regtry(prog, s)))
1165 LOAD_UTF8_CHARCLASS(space," ");
1166 while (s < strend) {
1167 if (*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8)) {
1168 if (tmp && (norun || regtry(prog, s)))
1179 while (s < strend) {
1181 if (tmp && (norun || regtry(prog, s)))
1193 PL_reg_flags |= RF_tainted;
1195 while (s < strend) {
1196 if (*s == ' ' || isSPACE_LC_utf8((U8*)s)) {
1197 if (tmp && (norun || regtry(prog, s)))
1208 while (s < strend) {
1209 if (isSPACE_LC(*s)) {
1210 if (tmp && (norun || regtry(prog, s)))
1223 LOAD_UTF8_CHARCLASS(space," ");
1224 while (s < strend) {
1225 if (!(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8))) {
1226 if (tmp && (norun || regtry(prog, s)))
1237 while (s < strend) {
1239 if (tmp && (norun || regtry(prog, s)))
1251 PL_reg_flags |= RF_tainted;
1253 while (s < strend) {
1254 if (!(*s == ' ' || isSPACE_LC_utf8((U8*)s))) {
1255 if (tmp && (norun || regtry(prog, s)))
1266 while (s < strend) {
1267 if (!isSPACE_LC(*s)) {
1268 if (tmp && (norun || regtry(prog, s)))
1281 LOAD_UTF8_CHARCLASS(digit,"0");
1282 while (s < strend) {
1283 if (swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1284 if (tmp && (norun || regtry(prog, s)))
1295 while (s < strend) {
1297 if (tmp && (norun || regtry(prog, s)))
1309 PL_reg_flags |= RF_tainted;
1311 while (s < strend) {
1312 if (isDIGIT_LC_utf8((U8*)s)) {
1313 if (tmp && (norun || regtry(prog, s)))
1324 while (s < strend) {
1325 if (isDIGIT_LC(*s)) {
1326 if (tmp && (norun || regtry(prog, s)))
1339 LOAD_UTF8_CHARCLASS(digit,"0");
1340 while (s < strend) {
1341 if (!swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1342 if (tmp && (norun || regtry(prog, s)))
1353 while (s < strend) {
1355 if (tmp && (norun || regtry(prog, s)))
1367 PL_reg_flags |= RF_tainted;
1369 while (s < strend) {
1370 if (!isDIGIT_LC_utf8((U8*)s)) {
1371 if (tmp && (norun || regtry(prog, s)))
1382 while (s < strend) {
1383 if (!isDIGIT_LC(*s)) {
1384 if (tmp && (norun || regtry(prog, s)))
1396 Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c));
1405 - regexec_flags - match a regexp against a string
1408 Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char *strend,
1409 char *strbeg, I32 minend, SV *sv, void *data, U32 flags)
1410 /* strend: pointer to null at end of string */
1411 /* strbeg: real beginning of string */
1412 /* minend: end of match must be >=minend after stringarg. */
1413 /* data: May be used for some additional optimizations. */
1414 /* nosave: For optimizations. */
1417 register regnode *c;
1418 register char *startpos = stringarg;
1419 I32 minlen; /* must match at least this many chars */
1420 I32 dontbother = 0; /* how many characters not to try at end */
1421 /* I32 start_shift = 0; */ /* Offset of the start to find
1422 constant substr. */ /* CC */
1423 I32 end_shift = 0; /* Same for the end. */ /* CC */
1424 I32 scream_pos = -1; /* Internal iterator of scream. */
1426 SV* oreplsv = GvSV(PL_replgv);
1427 bool do_utf8 = DO_UTF8(sv);
1433 PL_regnarrate = DEBUG_r_TEST;
1436 /* Be paranoid... */
1437 if (prog == NULL || startpos == NULL) {
1438 Perl_croak(aTHX_ "NULL regexp parameter");
1442 minlen = prog->minlen;
1444 if (!(prog->reganch & ROPT_SANY_SEEN))
1445 if (utf8_distance((U8*)strend, (U8*)startpos) < minlen) goto phooey;
1448 if (strend - startpos < minlen) goto phooey;
1451 /* Check validity of program. */
1452 if (UCHARAT(prog->program) != REG_MAGIC) {
1453 Perl_croak(aTHX_ "corrupted regexp program");
1457 PL_reg_eval_set = 0;
1460 if (prog->reganch & ROPT_UTF8)
1461 PL_reg_flags |= RF_utf8;
1463 /* Mark beginning of line for ^ and lookbehind. */
1464 PL_regbol = startpos;
1468 /* Mark end of line for $ (and such) */
1471 /* see how far we have to get to not match where we matched before */
1472 PL_regtill = startpos+minend;
1474 /* We start without call_cc context. */
1477 /* If there is a "must appear" string, look for it. */
1480 if (prog->reganch & ROPT_GPOS_SEEN) { /* Need to have PL_reg_ganch */
1483 if (flags & REXEC_IGNOREPOS) /* Means: check only at start */
1484 PL_reg_ganch = startpos;
1485 else if (sv && SvTYPE(sv) >= SVt_PVMG
1487 && (mg = mg_find(sv, PERL_MAGIC_regex_global))
1488 && mg->mg_len >= 0) {
1489 PL_reg_ganch = strbeg + mg->mg_len; /* Defined pos() */
1490 if (prog->reganch & ROPT_ANCH_GPOS) {
1491 if (s > PL_reg_ganch)
1496 else /* pos() not defined */
1497 PL_reg_ganch = strbeg;
1500 if (do_utf8 == (UTF!=0) &&
1501 !(flags & REXEC_CHECKED) && prog->check_substr != Nullsv) {
1502 re_scream_pos_data d;
1504 d.scream_olds = &scream_olds;
1505 d.scream_pos = &scream_pos;
1506 s = re_intuit_start(prog, sv, s, strend, flags, &d);
1508 goto phooey; /* not present */
1511 DEBUG_r( if (!PL_colorset) reginitcolors() );
1512 DEBUG_r(PerlIO_printf(Perl_debug_log,
1513 "%sMatching REx%s `%s%.60s%s%s' against `%s%.*s%s%s'\n",
1514 PL_colors[4],PL_colors[5],PL_colors[0],
1517 (strlen(prog->precomp) > 60 ? "..." : ""),
1519 (int)(strend - startpos > 60 ? 60 : strend - startpos),
1520 startpos, PL_colors[1],
1521 (strend - startpos > 60 ? "..." : ""))
1524 /* Simplest case: anchored match need be tried only once. */
1525 /* [unless only anchor is BOL and multiline is set] */
1526 if (prog->reganch & (ROPT_ANCH & ~ROPT_ANCH_GPOS)) {
1527 if (s == startpos && regtry(prog, startpos))
1529 else if (PL_multiline || (prog->reganch & ROPT_IMPLICIT)
1530 || (prog->reganch & ROPT_ANCH_MBOL)) /* XXXX SBOL? */
1535 dontbother = minlen - 1;
1536 end = HOP3c(strend, -dontbother, strbeg) - 1;
1537 /* for multiline we only have to try after newlines */
1538 if (prog->check_substr) {
1542 if (regtry(prog, s))
1547 if (prog->reganch & RE_USE_INTUIT) {
1548 s = re_intuit_start(prog, sv, s + 1, strend, flags, NULL);
1559 if (*s++ == '\n') { /* don't need PL_utf8skip here */
1560 if (regtry(prog, s))
1567 } else if (prog->reganch & ROPT_ANCH_GPOS) {
1568 if (regtry(prog, PL_reg_ganch))
1573 /* Messy cases: unanchored match. */
1574 if (prog->anchored_substr && prog->reganch & ROPT_SKIP) {
1575 /* we have /x+whatever/ */
1576 /* it must be a one character string (XXXX Except UTF?) */
1577 char ch = SvPVX(prog->anchored_substr)[0];
1583 while (s < strend) {
1585 DEBUG_r( did_match = 1 );
1586 if (regtry(prog, s)) goto got_it;
1588 while (s < strend && *s == ch)
1595 while (s < strend) {
1597 DEBUG_r( did_match = 1 );
1598 if (regtry(prog, s)) goto got_it;
1600 while (s < strend && *s == ch)
1606 DEBUG_r(if (!did_match)
1607 PerlIO_printf(Perl_debug_log,
1608 "Did not find anchored character...\n")
1612 else if (do_utf8 == (UTF!=0) &&
1613 (prog->anchored_substr != Nullsv
1614 || (prog->float_substr != Nullsv
1615 && prog->float_max_offset < strend - s))) {
1616 SV *must = prog->anchored_substr
1617 ? prog->anchored_substr : prog->float_substr;
1619 prog->anchored_substr ? prog->anchored_offset : prog->float_max_offset;
1621 prog->anchored_substr ? prog->anchored_offset : prog->float_min_offset;
1622 char *last = HOP3c(strend, /* Cannot start after this */
1623 -(I32)(CHR_SVLEN(must)
1624 - (SvTAIL(must) != 0) + back_min), strbeg);
1625 char *last1; /* Last position checked before */
1631 last1 = HOPc(s, -1);
1633 last1 = s - 1; /* bogus */
1635 /* XXXX check_substr already used to find `s', can optimize if
1636 check_substr==must. */
1638 dontbother = end_shift;
1639 strend = HOPc(strend, -dontbother);
1640 while ( (s <= last) &&
1641 ((flags & REXEC_SCREAM)
1642 ? (s = screaminstr(sv, must, HOP3c(s, back_min, strend) - strbeg,
1643 end_shift, &scream_pos, 0))
1644 : (s = fbm_instr((unsigned char*)HOP3(s, back_min, strend),
1645 (unsigned char*)strend, must,
1646 PL_multiline ? FBMrf_MULTILINE : 0))) ) {
1647 DEBUG_r( did_match = 1 );
1648 if (HOPc(s, -back_max) > last1) {
1649 last1 = HOPc(s, -back_min);
1650 s = HOPc(s, -back_max);
1653 char *t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1;
1655 last1 = HOPc(s, -back_min);
1659 while (s <= last1) {
1660 if (regtry(prog, s))
1666 while (s <= last1) {
1667 if (regtry(prog, s))
1673 DEBUG_r(if (!did_match)
1674 PerlIO_printf(Perl_debug_log,
1675 "Did not find %s substr `%s%.*s%s'%s...\n",
1676 ((must == prog->anchored_substr)
1677 ? "anchored" : "floating"),
1679 (int)(SvCUR(must) - (SvTAIL(must)!=0)),
1681 PL_colors[1], (SvTAIL(must) ? "$" : ""))
1685 else if ((c = prog->regstclass)) {
1686 if (minlen && PL_regkind[(U8)OP(prog->regstclass)] != EXACT)
1687 /* don't bother with what can't match */
1688 strend = HOPc(strend, -(minlen - 1));
1690 SV *prop = sv_newmortal();
1692 PerlIO_printf(Perl_debug_log, "Matching stclass `%s' against `%s'\n", SvPVX(prop), s);
1694 if (find_byclass(prog, c, s, strend, startpos, 0))
1696 DEBUG_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass...\n"));
1700 if (prog->float_substr != Nullsv) { /* Trim the end. */
1703 if (flags & REXEC_SCREAM) {
1704 last = screaminstr(sv, prog->float_substr, s - strbeg,
1705 end_shift, &scream_pos, 1); /* last one */
1707 last = scream_olds; /* Only one occurrence. */
1711 char *little = SvPV(prog->float_substr, len);
1713 if (SvTAIL(prog->float_substr)) {
1714 if (memEQ(strend - len + 1, little, len - 1))
1715 last = strend - len + 1;
1716 else if (!PL_multiline)
1717 last = memEQ(strend - len, little, len)
1718 ? strend - len : Nullch;
1724 last = rninstr(s, strend, little, little + len);
1726 last = strend; /* matching `$' */
1730 DEBUG_r(PerlIO_printf(Perl_debug_log,
1731 "%sCan't trim the tail, match fails (should not happen)%s\n",
1732 PL_colors[4],PL_colors[5]));
1733 goto phooey; /* Should not happen! */
1735 dontbother = strend - last + prog->float_min_offset;
1737 if (minlen && (dontbother < minlen))
1738 dontbother = minlen - 1;
1739 strend -= dontbother; /* this one's always in bytes! */
1740 /* We don't know much -- general case. */
1743 if (regtry(prog, s))
1752 if (regtry(prog, s))
1754 } while (s++ < strend);
1762 RX_MATCH_TAINTED_set(prog, PL_reg_flags & RF_tainted);
1764 if (PL_reg_eval_set) {
1765 /* Preserve the current value of $^R */
1766 if (oreplsv != GvSV(PL_replgv))
1767 sv_setsv(oreplsv, GvSV(PL_replgv));/* So that when GvSV(replgv) is
1768 restored, the value remains
1770 restore_pos(aTHXo_ 0);
1773 /* make sure $`, $&, $', and $digit will work later */
1774 if ( !(flags & REXEC_NOT_FIRST) ) {
1775 if (RX_MATCH_COPIED(prog)) {
1776 Safefree(prog->subbeg);
1777 RX_MATCH_COPIED_off(prog);
1779 if (flags & REXEC_COPY_STR) {
1780 I32 i = PL_regeol - startpos + (stringarg - strbeg);
1782 s = savepvn(strbeg, i);
1785 RX_MATCH_COPIED_on(prog);
1788 prog->subbeg = strbeg;
1789 prog->sublen = PL_regeol - strbeg; /* strend may have been modified */
1796 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n",
1797 PL_colors[4],PL_colors[5]));
1798 if (PL_reg_eval_set)
1799 restore_pos(aTHXo_ 0);
1804 - regtry - try match at specific point
1806 STATIC I32 /* 0 failure, 1 success */
1807 S_regtry(pTHX_ regexp *prog, char *startpos)
1815 PL_regindent = 0; /* XXXX Not good when matches are reenterable... */
1817 if ((prog->reganch & ROPT_EVAL_SEEN) && !PL_reg_eval_set) {
1820 PL_reg_eval_set = RS_init;
1822 PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %"IVdf"\n",
1823 (IV)(PL_stack_sp - PL_stack_base));
1825 SAVEI32(cxstack[cxstack_ix].blk_oldsp);
1826 cxstack[cxstack_ix].blk_oldsp = PL_stack_sp - PL_stack_base;
1827 /* Otherwise OP_NEXTSTATE will free whatever on stack now. */
1829 /* Apparently this is not needed, judging by wantarray. */
1830 /* SAVEI8(cxstack[cxstack_ix].blk_gimme);
1831 cxstack[cxstack_ix].blk_gimme = G_SCALAR; */
1834 /* Make $_ available to executed code. */
1835 if (PL_reg_sv != DEFSV) {
1836 /* SAVE_DEFSV does *not* suffice here for USE_THREADS */
1841 if (!(SvTYPE(PL_reg_sv) >= SVt_PVMG && SvMAGIC(PL_reg_sv)
1842 && (mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global)))) {
1843 /* prepare for quick setting of pos */
1844 sv_magic(PL_reg_sv, (SV*)0,
1845 PERL_MAGIC_regex_global, Nullch, 0);
1846 mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global);
1850 PL_reg_oldpos = mg->mg_len;
1851 SAVEDESTRUCTOR_X(restore_pos, 0);
1853 if (!PL_reg_curpm) {
1854 Newz(22,PL_reg_curpm, 1, PMOP);
1857 SV* repointer = newSViv(0);
1858 av_push(PL_regex_padav,repointer);
1859 PL_reg_curpm->op_pmoffset = av_len(PL_regex_padav);
1860 PL_regex_pad = AvARRAY(PL_regex_padav);
1864 PM_SETRE(PL_reg_curpm, prog);
1865 PL_reg_oldcurpm = PL_curpm;
1866 PL_curpm = PL_reg_curpm;
1867 if (RX_MATCH_COPIED(prog)) {
1868 /* Here is a serious problem: we cannot rewrite subbeg,
1869 since it may be needed if this match fails. Thus
1870 $` inside (?{}) could fail... */
1871 PL_reg_oldsaved = prog->subbeg;
1872 PL_reg_oldsavedlen = prog->sublen;
1873 RX_MATCH_COPIED_off(prog);
1876 PL_reg_oldsaved = Nullch;
1877 prog->subbeg = PL_bostr;
1878 prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */
1880 prog->startp[0] = startpos - PL_bostr;
1881 PL_reginput = startpos;
1882 PL_regstartp = prog->startp;
1883 PL_regendp = prog->endp;
1884 PL_reglastparen = &prog->lastparen;
1885 PL_reglastcloseparen = &prog->lastcloseparen;
1886 prog->lastparen = 0;
1888 DEBUG_r(PL_reg_starttry = startpos);
1889 if (PL_reg_start_tmpl <= prog->nparens) {
1890 PL_reg_start_tmpl = prog->nparens*3/2 + 3;
1891 if(PL_reg_start_tmp)
1892 Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
1894 New(22,PL_reg_start_tmp, PL_reg_start_tmpl, char*);
1897 /* XXXX What this code is doing here?!!! There should be no need
1898 to do this again and again, PL_reglastparen should take care of
1901 /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code.
1902 * Actually, the code in regcppop() (which Ilya may be meaning by
1903 * PL_reglastparen), is not needed at all by the test suite
1904 * (op/regexp, op/pat, op/split), but that code is needed, oddly
1905 * enough, for building DynaLoader, or otherwise this
1906 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
1907 * will happen. Meanwhile, this code *is* needed for the
1908 * above-mentioned test suite tests to succeed. The common theme
1909 * on those tests seems to be returning null fields from matches.
1914 if (prog->nparens) {
1915 for (i = prog->nparens; i > *PL_reglastparen; i--) {
1922 if (regmatch(prog->program + 1)) {
1923 prog->endp[0] = PL_reginput - PL_bostr;
1926 REGCP_UNWIND(lastcp);
1930 #define RE_UNWIND_BRANCH 1
1931 #define RE_UNWIND_BRANCHJ 2
1935 typedef struct { /* XX: makes sense to enlarge it... */
1939 } re_unwind_generic_t;
1952 } re_unwind_branch_t;
1954 typedef union re_unwind_t {
1956 re_unwind_generic_t generic;
1957 re_unwind_branch_t branch;
1960 #define sayYES goto yes
1961 #define sayNO goto no
1962 #define sayYES_FINAL goto yes_final
1963 #define sayYES_LOUD goto yes_loud
1964 #define sayNO_FINAL goto no_final
1965 #define sayNO_SILENT goto do_no
1966 #define saySAME(x) if (x) goto yes; else goto no
1968 #define REPORT_CODE_OFF 24
1971 - regmatch - main matching routine
1973 * Conceptually the strategy is simple: check to see whether the current
1974 * node matches, call self recursively to see whether the rest matches,
1975 * and then act accordingly. In practice we make some effort to avoid
1976 * recursion, in particular by going through "ordinary" nodes (that don't
1977 * need to know whether the rest of the match failed) by a loop instead of
1980 /* [lwall] I've hoisted the register declarations to the outer block in order to
1981 * maybe save a little bit of pushing and popping on the stack. It also takes
1982 * advantage of machines that use a register save mask on subroutine entry.
1984 STATIC I32 /* 0 failure, 1 success */
1985 S_regmatch(pTHX_ regnode *prog)
1987 register regnode *scan; /* Current node. */
1988 regnode *next; /* Next node. */
1989 regnode *inner; /* Next node in internal branch. */
1990 register I32 nextchr; /* renamed nextchr - nextchar colides with
1991 function of same name */
1992 register I32 n; /* no or next */
1993 register I32 ln = 0; /* len or last */
1994 register char *s = Nullch; /* operand or save */
1995 register char *locinput = PL_reginput;
1996 register I32 c1 = 0, c2 = 0, paren; /* case fold search, parenth */
1997 int minmod = 0, sw = 0, logical = 0;
2000 I32 firstcp = PL_savestack_ix;
2002 register bool do_utf8 = DO_UTF8(PL_reg_sv);
2008 /* Note that nextchr is a byte even in UTF */
2009 nextchr = UCHARAT(locinput);
2011 while (scan != NULL) {
2014 SV *prop = sv_newmortal();
2015 int docolor = *PL_colors[0];
2016 int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */
2017 int l = (PL_regeol - locinput) > taill ? taill : (PL_regeol - locinput);
2018 /* The part of the string before starttry has one color
2019 (pref0_len chars), between starttry and current
2020 position another one (pref_len - pref0_len chars),
2021 after the current position the third one.
2022 We assume that pref0_len <= pref_len, otherwise we
2023 decrease pref0_len. */
2024 int pref_len = (locinput - PL_bostr) > (5 + taill) - l
2025 ? (5 + taill) - l : locinput - PL_bostr;
2028 while (UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len)))
2030 pref0_len = pref_len - (locinput - PL_reg_starttry);
2031 if (l + pref_len < (5 + taill) && l < PL_regeol - locinput)
2032 l = ( PL_regeol - locinput > (5 + taill) - pref_len
2033 ? (5 + taill) - pref_len : PL_regeol - locinput);
2034 while (UTF8_IS_CONTINUATION(*(U8*)(locinput + l)))
2038 if (pref0_len > pref_len)
2039 pref0_len = pref_len;
2040 regprop(prop, scan);
2041 PerlIO_printf(Perl_debug_log,
2042 "%4"IVdf" <%s%.*s%s%s%.*s%s%s%s%.*s%s>%*s|%3"IVdf":%*s%s\n",
2043 (IV)(locinput - PL_bostr),
2044 PL_colors[4], pref0_len,
2045 locinput - pref_len, PL_colors[5],
2046 PL_colors[2], pref_len - pref0_len,
2047 locinput - pref_len + pref0_len, PL_colors[3],
2048 (docolor ? "" : "> <"),
2049 PL_colors[0], l, locinput, PL_colors[1],
2050 15 - l - pref_len + 1,
2052 (IV)(scan - PL_regprogram), PL_regindent*2, "",
2056 next = scan + NEXT_OFF(scan);
2062 if (locinput == PL_bostr || (PL_multiline &&
2063 (nextchr || locinput < PL_regeol) && locinput[-1] == '\n') )
2065 /* regtill = regbol; */
2070 if (locinput == PL_bostr ||
2071 ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n'))
2077 if (locinput == PL_bostr)
2081 if (locinput == PL_reg_ganch)
2091 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2096 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2098 if (PL_regeol - locinput > 1)
2102 if (PL_regeol != locinput)
2106 if (!nextchr && locinput >= PL_regeol)
2108 nextchr = UCHARAT(++locinput);
2111 if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n')
2114 locinput += PL_utf8skip[nextchr];
2115 if (locinput > PL_regeol)
2117 nextchr = UCHARAT(locinput);
2120 nextchr = UCHARAT(++locinput);
2125 if (do_utf8 != (UTF!=0)) {
2133 if (*((U8*)s) != utf8_to_uvchr((U8*)l, &len))
2142 if (*((U8*)l) != utf8_to_uvchr((U8*)s, &len))
2148 nextchr = UCHARAT(locinput);
2151 /* Inline the first character, for speed. */
2152 if (UCHARAT(s) != nextchr)
2154 if (PL_regeol - locinput < ln)
2156 if (ln > 1 && memNE(s, locinput, ln))
2159 nextchr = UCHARAT(locinput);
2162 PL_reg_flags |= RF_tainted;
2172 c1 = OP(scan) == EXACTF;
2174 if (l >= PL_regeol) {
2177 if ((UTF ? utf8n_to_uvchr((U8*)s, e - s, 0, 0) : *((U8*)s)) !=
2178 (c1 ? toLOWER_utf8((U8*)l) : toLOWER_LC_utf8((U8*)l)))
2180 s += UTF ? UTF8SKIP(s) : 1;
2184 nextchr = UCHARAT(locinput);
2188 /* Inline the first character, for speed. */
2189 if (UCHARAT(s) != nextchr &&
2190 UCHARAT(s) != ((OP(scan) == EXACTF)
2191 ? PL_fold : PL_fold_locale)[nextchr])
2193 if (PL_regeol - locinput < ln)
2195 if (ln > 1 && (OP(scan) == EXACTF
2196 ? ibcmp(s, locinput, ln)
2197 : ibcmp_locale(s, locinput, ln)))
2200 nextchr = UCHARAT(locinput);
2204 if (!reginclass(scan, (U8*)locinput, do_utf8))
2206 if (locinput >= PL_regeol)
2208 locinput += PL_utf8skip[nextchr];
2209 nextchr = UCHARAT(locinput);
2213 nextchr = UCHARAT(locinput);
2214 if (!reginclass(scan, (U8*)locinput, do_utf8))
2216 if (!nextchr && locinput >= PL_regeol)
2218 nextchr = UCHARAT(++locinput);
2222 PL_reg_flags |= RF_tainted;
2228 LOAD_UTF8_CHARCLASS(alnum,"a");
2229 if (!(OP(scan) == ALNUM
2230 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2231 : isALNUM_LC_utf8((U8*)locinput)))
2235 locinput += PL_utf8skip[nextchr];
2236 nextchr = UCHARAT(locinput);
2239 if (!(OP(scan) == ALNUM
2240 ? isALNUM(nextchr) : isALNUM_LC(nextchr)))
2242 nextchr = UCHARAT(++locinput);
2245 PL_reg_flags |= RF_tainted;
2248 if (!nextchr && locinput >= PL_regeol)
2251 LOAD_UTF8_CHARCLASS(alnum,"a");
2252 if (OP(scan) == NALNUM
2253 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2254 : isALNUM_LC_utf8((U8*)locinput))
2258 locinput += PL_utf8skip[nextchr];
2259 nextchr = UCHARAT(locinput);
2262 if (OP(scan) == NALNUM
2263 ? isALNUM(nextchr) : isALNUM_LC(nextchr))
2265 nextchr = UCHARAT(++locinput);
2269 PL_reg_flags |= RF_tainted;
2273 /* was last char in word? */
2275 if (locinput == PL_bostr)
2278 U8 *r = reghop((U8*)locinput, -1);
2280 ln = utf8n_to_uvchr(r, s - (char*)r, 0, 0);
2282 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2283 ln = isALNUM_uni(ln);
2284 LOAD_UTF8_CHARCLASS(alnum,"a");
2285 n = swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8);
2288 ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(ln));
2289 n = isALNUM_LC_utf8((U8*)locinput);
2293 ln = (locinput != PL_bostr) ?
2294 UCHARAT(locinput - 1) : '\n';
2295 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2297 n = isALNUM(nextchr);
2300 ln = isALNUM_LC(ln);
2301 n = isALNUM_LC(nextchr);
2304 if (((!ln) == (!n)) == (OP(scan) == BOUND ||
2305 OP(scan) == BOUNDL))
2309 PL_reg_flags |= RF_tainted;
2315 if (UTF8_IS_CONTINUED(nextchr)) {
2316 LOAD_UTF8_CHARCLASS(space," ");
2317 if (!(OP(scan) == SPACE
2318 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2319 : isSPACE_LC_utf8((U8*)locinput)))
2323 locinput += PL_utf8skip[nextchr];
2324 nextchr = UCHARAT(locinput);
2327 if (!(OP(scan) == SPACE
2328 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2330 nextchr = UCHARAT(++locinput);
2333 if (!(OP(scan) == SPACE
2334 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2336 nextchr = UCHARAT(++locinput);
2340 PL_reg_flags |= RF_tainted;
2343 if (!nextchr && locinput >= PL_regeol)
2346 LOAD_UTF8_CHARCLASS(space," ");
2347 if (OP(scan) == NSPACE
2348 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2349 : isSPACE_LC_utf8((U8*)locinput))
2353 locinput += PL_utf8skip[nextchr];
2354 nextchr = UCHARAT(locinput);
2357 if (OP(scan) == NSPACE
2358 ? isSPACE(nextchr) : isSPACE_LC(nextchr))
2360 nextchr = UCHARAT(++locinput);
2363 PL_reg_flags |= RF_tainted;
2369 LOAD_UTF8_CHARCLASS(digit,"0");
2370 if (!(OP(scan) == DIGIT
2371 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2372 : isDIGIT_LC_utf8((U8*)locinput)))
2376 locinput += PL_utf8skip[nextchr];
2377 nextchr = UCHARAT(locinput);
2380 if (!(OP(scan) == DIGIT
2381 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr)))
2383 nextchr = UCHARAT(++locinput);
2386 PL_reg_flags |= RF_tainted;
2389 if (!nextchr && locinput >= PL_regeol)
2392 LOAD_UTF8_CHARCLASS(digit,"0");
2393 if (OP(scan) == NDIGIT
2394 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2395 : isDIGIT_LC_utf8((U8*)locinput))
2399 locinput += PL_utf8skip[nextchr];
2400 nextchr = UCHARAT(locinput);
2403 if (OP(scan) == NDIGIT
2404 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr))
2406 nextchr = UCHARAT(++locinput);
2409 LOAD_UTF8_CHARCLASS(mark,"~");
2410 if (locinput >= PL_regeol ||
2411 swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2413 locinput += PL_utf8skip[nextchr];
2414 while (locinput < PL_regeol &&
2415 swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2416 locinput += UTF8SKIP(locinput);
2417 if (locinput > PL_regeol)
2419 nextchr = UCHARAT(locinput);
2422 PL_reg_flags |= RF_tainted;
2426 n = ARG(scan); /* which paren pair */
2427 ln = PL_regstartp[n];
2428 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2429 if (*PL_reglastparen < n || ln == -1)
2430 sayNO; /* Do not match unless seen CLOSEn. */
2431 if (ln == PL_regendp[n])
2435 if (do_utf8 && OP(scan) != REF) { /* REF can do byte comparison */
2437 char *e = PL_bostr + PL_regendp[n];
2439 * Note that we can't do the "other character" lookup trick as
2440 * in the 8-bit case (no pun intended) because in Unicode we
2441 * have to map both upper and title case to lower case.
2443 if (OP(scan) == REFF) {
2447 if (toLOWER_utf8((U8*)s) != toLOWER_utf8((U8*)l))
2457 if (toLOWER_LC_utf8((U8*)s) != toLOWER_LC_utf8((U8*)l))
2464 nextchr = UCHARAT(locinput);
2468 /* Inline the first character, for speed. */
2469 if (UCHARAT(s) != nextchr &&
2471 (UCHARAT(s) != ((OP(scan) == REFF
2472 ? PL_fold : PL_fold_locale)[nextchr]))))
2474 ln = PL_regendp[n] - ln;
2475 if (locinput + ln > PL_regeol)
2477 if (ln > 1 && (OP(scan) == REF
2478 ? memNE(s, locinput, ln)
2480 ? ibcmp(s, locinput, ln)
2481 : ibcmp_locale(s, locinput, ln))))
2484 nextchr = UCHARAT(locinput);
2495 OP_4tree *oop = PL_op;
2496 COP *ocurcop = PL_curcop;
2497 SV **ocurpad = PL_curpad;
2501 PL_op = (OP_4tree*)PL_regdata->data[n];
2502 DEBUG_r( PerlIO_printf(Perl_debug_log, " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) );
2503 PL_curpad = AvARRAY((AV*)PL_regdata->data[n + 2]);
2504 PL_regendp[0] = PL_reg_magic->mg_len = locinput - PL_bostr;
2506 CALLRUNOPS(aTHX); /* Scalar context. */
2512 PL_curpad = ocurpad;
2513 PL_curcop = ocurcop;
2515 if (logical == 2) { /* Postponed subexpression. */
2517 MAGIC *mg = Null(MAGIC*);
2519 CHECKPOINT cp, lastcp;
2521 if(SvROK(ret) || SvRMAGICAL(ret)) {
2522 SV *sv = SvROK(ret) ? SvRV(ret) : ret;
2525 mg = mg_find(sv, PERL_MAGIC_qr);
2528 re = (regexp *)mg->mg_obj;
2529 (void)ReREFCNT_inc(re);
2533 char *t = SvPV(ret, len);
2535 char *oprecomp = PL_regprecomp;
2536 I32 osize = PL_regsize;
2537 I32 onpar = PL_regnpar;
2540 re = CALLREGCOMP(aTHX_ t, t + len, &pm);
2542 & (SVs_TEMP | SVs_PADTMP | SVf_READONLY)))
2543 sv_magic(ret,(SV*)ReREFCNT_inc(re),
2545 PL_regprecomp = oprecomp;
2550 PerlIO_printf(Perl_debug_log,
2551 "Entering embedded `%s%.60s%s%s'\n",
2555 (strlen(re->precomp) > 60 ? "..." : ""))
2558 state.prev = PL_reg_call_cc;
2559 state.cc = PL_regcc;
2560 state.re = PL_reg_re;
2564 cp = regcppush(0); /* Save *all* the positions. */
2567 state.ss = PL_savestack_ix;
2568 *PL_reglastparen = 0;
2569 *PL_reglastcloseparen = 0;
2570 PL_reg_call_cc = &state;
2571 PL_reginput = locinput;
2573 /* XXXX This is too dramatic a measure... */
2576 if (regmatch(re->program + 1)) {
2577 /* Even though we succeeded, we need to restore
2578 global variables, since we may be wrapped inside
2579 SUSPEND, thus the match may be not finished yet. */
2581 /* XXXX Do this only if SUSPENDed? */
2582 PL_reg_call_cc = state.prev;
2583 PL_regcc = state.cc;
2584 PL_reg_re = state.re;
2585 cache_re(PL_reg_re);
2587 /* XXXX This is too dramatic a measure... */
2590 /* These are needed even if not SUSPEND. */
2596 REGCP_UNWIND(lastcp);
2598 PL_reg_call_cc = state.prev;
2599 PL_regcc = state.cc;
2600 PL_reg_re = state.re;
2601 cache_re(PL_reg_re);
2603 /* XXXX This is too dramatic a measure... */
2612 sv_setsv(save_scalar(PL_replgv), ret);
2616 n = ARG(scan); /* which paren pair */
2617 PL_reg_start_tmp[n] = locinput;
2622 n = ARG(scan); /* which paren pair */
2623 PL_regstartp[n] = PL_reg_start_tmp[n] - PL_bostr;
2624 PL_regendp[n] = locinput - PL_bostr;
2625 if (n > *PL_reglastparen)
2626 *PL_reglastparen = n;
2627 *PL_reglastcloseparen = n;
2630 n = ARG(scan); /* which paren pair */
2631 sw = (*PL_reglastparen >= n && PL_regendp[n] != -1);
2634 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2636 next = NEXTOPER(NEXTOPER(scan));
2638 next = scan + ARG(scan);
2639 if (OP(next) == IFTHEN) /* Fake one. */
2640 next = NEXTOPER(NEXTOPER(next));
2644 logical = scan->flags;
2646 /*******************************************************************
2647 PL_regcc contains infoblock about the innermost (...)* loop, and
2648 a pointer to the next outer infoblock.
2650 Here is how Y(A)*Z is processed (if it is compiled into CURLYX/WHILEM):
2652 1) After matching X, regnode for CURLYX is processed;
2654 2) This regnode creates infoblock on the stack, and calls
2655 regmatch() recursively with the starting point at WHILEM node;
2657 3) Each hit of WHILEM node tries to match A and Z (in the order
2658 depending on the current iteration, min/max of {min,max} and
2659 greediness). The information about where are nodes for "A"
2660 and "Z" is read from the infoblock, as is info on how many times "A"
2661 was already matched, and greediness.
2663 4) After A matches, the same WHILEM node is hit again.
2665 5) Each time WHILEM is hit, PL_regcc is the infoblock created by CURLYX
2666 of the same pair. Thus when WHILEM tries to match Z, it temporarily
2667 resets PL_regcc, since this Y(A)*Z can be a part of some other loop:
2668 as in (Y(A)*Z)*. If Z matches, the automaton will hit the WHILEM node
2669 of the external loop.
2671 Currently present infoblocks form a tree with a stem formed by PL_curcc
2672 and whatever it mentions via ->next, and additional attached trees
2673 corresponding to temporarily unset infoblocks as in "5" above.
2675 In the following picture infoblocks for outer loop of
2676 (Y(A)*?Z)*?T are denoted O, for inner I. NULL starting block
2677 is denoted by x. The matched string is YAAZYAZT. Temporarily postponed
2678 infoblocks are drawn below the "reset" infoblock.
2680 In fact in the picture below we do not show failed matches for Z and T
2681 by WHILEM blocks. [We illustrate minimal matches, since for them it is
2682 more obvious *why* one needs to *temporary* unset infoblocks.]
2684 Matched REx position InfoBlocks Comment
2688 Y A)*?Z)*?T x <- O <- I
2689 YA )*?Z)*?T x <- O <- I
2690 YA A)*?Z)*?T x <- O <- I
2691 YAA )*?Z)*?T x <- O <- I
2692 YAA Z)*?T x <- O # Temporary unset I
2695 YAAZ Y(A)*?Z)*?T x <- O
2698 YAAZY (A)*?Z)*?T x <- O
2701 YAAZY A)*?Z)*?T x <- O <- I
2704 YAAZYA )*?Z)*?T x <- O <- I
2707 YAAZYA Z)*?T x <- O # Temporary unset I
2713 YAAZYAZ T x # Temporary unset O
2720 *******************************************************************/
2723 CHECKPOINT cp = PL_savestack_ix;
2724 /* No need to save/restore up to this paren */
2725 I32 parenfloor = scan->flags;
2727 if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */
2729 cc.oldcc = PL_regcc;
2731 /* XXXX Probably it is better to teach regpush to support
2732 parenfloor > PL_regsize... */
2733 if (parenfloor > *PL_reglastparen)
2734 parenfloor = *PL_reglastparen; /* Pessimization... */
2735 cc.parenfloor = parenfloor;
2737 cc.min = ARG1(scan);
2738 cc.max = ARG2(scan);
2739 cc.scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
2743 PL_reginput = locinput;
2744 n = regmatch(PREVOPER(next)); /* start on the WHILEM */
2746 PL_regcc = cc.oldcc;
2752 * This is really hard to understand, because after we match
2753 * what we're trying to match, we must make sure the rest of
2754 * the REx is going to match for sure, and to do that we have
2755 * to go back UP the parse tree by recursing ever deeper. And
2756 * if it fails, we have to reset our parent's current state
2757 * that we can try again after backing off.
2760 CHECKPOINT cp, lastcp;
2761 CURCUR* cc = PL_regcc;
2762 char *lastloc = cc->lastloc; /* Detection of 0-len. */
2764 n = cc->cur + 1; /* how many we know we matched */
2765 PL_reginput = locinput;
2768 PerlIO_printf(Perl_debug_log,
2769 "%*s %ld out of %ld..%ld cc=%lx\n",
2770 REPORT_CODE_OFF+PL_regindent*2, "",
2771 (long)n, (long)cc->min,
2772 (long)cc->max, (long)cc)
2775 /* If degenerate scan matches "", assume scan done. */
2777 if (locinput == cc->lastloc && n >= cc->min) {
2778 PL_regcc = cc->oldcc;
2782 PerlIO_printf(Perl_debug_log,
2783 "%*s empty match detected, try continuation...\n",
2784 REPORT_CODE_OFF+PL_regindent*2, "")
2786 if (regmatch(cc->next))
2794 /* First just match a string of min scans. */
2798 cc->lastloc = locinput;
2799 if (regmatch(cc->scan))
2802 cc->lastloc = lastloc;
2807 /* Check whether we already were at this position.
2808 Postpone detection until we know the match is not
2809 *that* much linear. */
2810 if (!PL_reg_maxiter) {
2811 PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4);
2812 PL_reg_leftiter = PL_reg_maxiter;
2814 if (PL_reg_leftiter-- == 0) {
2815 I32 size = (PL_reg_maxiter + 7)/8;
2816 if (PL_reg_poscache) {
2817 if (PL_reg_poscache_size < size) {
2818 Renew(PL_reg_poscache, size, char);
2819 PL_reg_poscache_size = size;
2821 Zero(PL_reg_poscache, size, char);
2824 PL_reg_poscache_size = size;
2825 Newz(29, PL_reg_poscache, size, char);
2828 PerlIO_printf(Perl_debug_log,
2829 "%sDetected a super-linear match, switching on caching%s...\n",
2830 PL_colors[4], PL_colors[5])
2833 if (PL_reg_leftiter < 0) {
2834 I32 o = locinput - PL_bostr, b;
2836 o = (scan->flags & 0xf) - 1 + o * (scan->flags>>4);
2839 if (PL_reg_poscache[o] & (1<<b)) {
2841 PerlIO_printf(Perl_debug_log,
2842 "%*s already tried at this position...\n",
2843 REPORT_CODE_OFF+PL_regindent*2, "")
2847 PL_reg_poscache[o] |= (1<<b);
2851 /* Prefer next over scan for minimal matching. */
2854 PL_regcc = cc->oldcc;
2857 cp = regcppush(cc->parenfloor);
2859 if (regmatch(cc->next)) {
2861 sayYES; /* All done. */
2863 REGCP_UNWIND(lastcp);
2869 if (n >= cc->max) { /* Maximum greed exceeded? */
2870 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
2871 && !(PL_reg_flags & RF_warned)) {
2872 PL_reg_flags |= RF_warned;
2873 Perl_warner(aTHX_ WARN_REGEXP, "%s limit (%d) exceeded",
2874 "Complex regular subexpression recursion",
2881 PerlIO_printf(Perl_debug_log,
2882 "%*s trying longer...\n",
2883 REPORT_CODE_OFF+PL_regindent*2, "")
2885 /* Try scanning more and see if it helps. */
2886 PL_reginput = locinput;
2888 cc->lastloc = locinput;
2889 cp = regcppush(cc->parenfloor);
2891 if (regmatch(cc->scan)) {
2895 REGCP_UNWIND(lastcp);
2898 cc->lastloc = lastloc;
2902 /* Prefer scan over next for maximal matching. */
2904 if (n < cc->max) { /* More greed allowed? */
2905 cp = regcppush(cc->parenfloor);
2907 cc->lastloc = locinput;
2909 if (regmatch(cc->scan)) {
2913 REGCP_UNWIND(lastcp);
2914 regcppop(); /* Restore some previous $<digit>s? */
2915 PL_reginput = locinput;
2917 PerlIO_printf(Perl_debug_log,
2918 "%*s failed, try continuation...\n",
2919 REPORT_CODE_OFF+PL_regindent*2, "")
2922 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
2923 && !(PL_reg_flags & RF_warned)) {
2924 PL_reg_flags |= RF_warned;
2925 Perl_warner(aTHX_ WARN_REGEXP, "%s limit (%d) exceeded",
2926 "Complex regular subexpression recursion",
2930 /* Failed deeper matches of scan, so see if this one works. */
2931 PL_regcc = cc->oldcc;
2934 if (regmatch(cc->next))
2940 cc->lastloc = lastloc;
2945 next = scan + ARG(scan);
2948 inner = NEXTOPER(NEXTOPER(scan));
2951 inner = NEXTOPER(scan);
2955 if (OP(next) != c1) /* No choice. */
2956 next = inner; /* Avoid recursion. */
2958 I32 lastparen = *PL_reglastparen;
2960 re_unwind_branch_t *uw;
2962 /* Put unwinding data on stack */
2963 unwind1 = SSNEWt(1,re_unwind_branch_t);
2964 uw = SSPTRt(unwind1,re_unwind_branch_t);
2967 uw->type = ((c1 == BRANCH)
2969 : RE_UNWIND_BRANCHJ);
2970 uw->lastparen = lastparen;
2972 uw->locinput = locinput;
2973 uw->nextchr = nextchr;
2975 uw->regindent = ++PL_regindent;
2978 REGCP_SET(uw->lastcp);
2980 /* Now go into the first branch */
2993 /* We suppose that the next guy does not need
2994 backtracking: in particular, it is of constant length,
2995 and has no parenths to influence future backrefs. */
2996 ln = ARG1(scan); /* min to match */
2997 n = ARG2(scan); /* max to match */
2998 paren = scan->flags;
3000 if (paren > PL_regsize)
3002 if (paren > *PL_reglastparen)
3003 *PL_reglastparen = paren;
3005 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3007 scan += NEXT_OFF(scan); /* Skip former OPEN. */
3008 PL_reginput = locinput;
3011 if (ln && regrepeat_hard(scan, ln, &l) < ln)
3013 if (ln && l == 0 && n >= ln
3014 /* In fact, this is tricky. If paren, then the
3015 fact that we did/didnot match may influence
3016 future execution. */
3017 && !(paren && ln == 0))
3019 locinput = PL_reginput;
3020 if (PL_regkind[(U8)OP(next)] == EXACT) {
3021 c1 = (U8)*STRING(next);
3022 if (OP(next) == EXACTF)
3024 else if (OP(next) == EXACTFL)
3025 c2 = PL_fold_locale[c1];
3032 /* This may be improved if l == 0. */
3033 while (n >= ln || (n == REG_INFTY && ln > 0 && l)) { /* ln overflow ? */
3034 /* If it could work, try it. */
3036 UCHARAT(PL_reginput) == c1 ||
3037 UCHARAT(PL_reginput) == c2)
3041 PL_regstartp[paren] =
3042 HOPc(PL_reginput, -l) - PL_bostr;
3043 PL_regendp[paren] = PL_reginput - PL_bostr;
3046 PL_regendp[paren] = -1;
3050 REGCP_UNWIND(lastcp);
3052 /* Couldn't or didn't -- move forward. */
3053 PL_reginput = locinput;
3054 if (regrepeat_hard(scan, 1, &l)) {
3056 locinput = PL_reginput;
3063 n = regrepeat_hard(scan, n, &l);
3064 if (n != 0 && l == 0
3065 /* In fact, this is tricky. If paren, then the
3066 fact that we did/didnot match may influence
3067 future execution. */
3068 && !(paren && ln == 0))
3070 locinput = PL_reginput;
3072 PerlIO_printf(Perl_debug_log,
3073 "%*s matched %"IVdf" times, len=%"IVdf"...\n",
3074 (int)(REPORT_CODE_OFF+PL_regindent*2), "",
3078 if (PL_regkind[(U8)OP(next)] == EXACT) {
3079 c1 = (U8)*STRING(next);
3080 if (OP(next) == EXACTF)
3082 else if (OP(next) == EXACTFL)
3083 c2 = PL_fold_locale[c1];
3092 /* If it could work, try it. */
3094 UCHARAT(PL_reginput) == c1 ||
3095 UCHARAT(PL_reginput) == c2)
3098 PerlIO_printf(Perl_debug_log,
3099 "%*s trying tail with n=%"IVdf"...\n",
3100 (int)(REPORT_CODE_OFF+PL_regindent*2), "", (IV)n)
3104 PL_regstartp[paren] = HOPc(PL_reginput, -l) - PL_bostr;
3105 PL_regendp[paren] = PL_reginput - PL_bostr;
3108 PL_regendp[paren] = -1;
3112 REGCP_UNWIND(lastcp);
3114 /* Couldn't or didn't -- back up. */
3116 locinput = HOPc(locinput, -l);
3117 PL_reginput = locinput;
3124 paren = scan->flags; /* Which paren to set */
3125 if (paren > PL_regsize)
3127 if (paren > *PL_reglastparen)
3128 *PL_reglastparen = paren;
3129 ln = ARG1(scan); /* min to match */
3130 n = ARG2(scan); /* max to match */
3131 scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE);
3135 ln = ARG1(scan); /* min to match */
3136 n = ARG2(scan); /* max to match */
3137 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3142 scan = NEXTOPER(scan);
3148 scan = NEXTOPER(scan);
3152 * Lookahead to avoid useless match attempts
3153 * when we know what character comes next.
3155 if (PL_regkind[(U8)OP(next)] == EXACT) {
3156 U8 *s = (U8*)STRING(next);
3159 if (OP(next) == EXACTF)
3161 else if (OP(next) == EXACTFL)
3162 c2 = PL_fold_locale[c1];
3165 if (OP(next) == EXACTF) {
3166 c1 = to_utf8_lower(s);
3167 c2 = to_utf8_upper(s);
3170 c2 = c1 = utf8_to_uvchr(s, NULL);
3176 PL_reginput = locinput;
3180 if (ln && regrepeat(scan, ln) < ln)
3182 locinput = PL_reginput;
3185 char *e; /* Should not check after this */
3186 char *old = locinput;
3188 if (n == REG_INFTY) {
3191 while (UTF8_IS_CONTINUATION(*(U8*)e))
3197 m >0 && e + UTF8SKIP(e) <= PL_regeol; m--)
3201 e = locinput + n - ln;
3207 /* Find place 'next' could work */
3210 while (locinput <= e && *locinput != c1)
3213 while (locinput <= e
3218 count = locinput - old;
3225 utf8_to_uvchr((U8*)locinput, &len) != c1;
3230 for (count = 0; locinput <= e; count++) {
3231 UV c = utf8_to_uvchr((U8*)locinput, &len);
3232 if (c == c1 || c == c2)
3240 /* PL_reginput == old now */
3241 if (locinput != old) {
3242 ln = 1; /* Did some */
3243 if (regrepeat(scan, count) < count)
3246 /* PL_reginput == locinput now */
3247 TRYPAREN(paren, ln, locinput);
3248 PL_reginput = locinput; /* Could be reset... */
3249 REGCP_UNWIND(lastcp);
3250 /* Couldn't or didn't -- move forward. */
3253 locinput += UTF8SKIP(locinput);
3259 while (n >= ln || (n == REG_INFTY && ln > 0)) { /* ln overflow ? */
3263 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3265 c = UCHARAT(PL_reginput);
3266 /* If it could work, try it. */
3267 if (c == c1 || c == c2)
3269 TRYPAREN(paren, n, PL_reginput);
3270 REGCP_UNWIND(lastcp);
3273 /* If it could work, try it. */
3274 else if (c1 == -1000)
3276 TRYPAREN(paren, n, PL_reginput);
3277 REGCP_UNWIND(lastcp);
3279 /* Couldn't or didn't -- move forward. */
3280 PL_reginput = locinput;
3281 if (regrepeat(scan, 1)) {
3283 locinput = PL_reginput;
3291 n = regrepeat(scan, n);
3292 locinput = PL_reginput;
3293 if (ln < n && PL_regkind[(U8)OP(next)] == EOL &&
3294 (!PL_multiline || OP(next) == SEOL || OP(next) == EOS)) {
3295 ln = n; /* why back off? */
3296 /* ...because $ and \Z can match before *and* after
3297 newline at the end. Consider "\n\n" =~ /\n+\Z\n/.
3298 We should back off by one in this case. */
3299 if (UCHARAT(PL_reginput - 1) == '\n' && OP(next) != EOS)
3308 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3310 c = UCHARAT(PL_reginput);
3312 /* If it could work, try it. */
3313 if (c1 == -1000 || c == c1 || c == c2)
3315 TRYPAREN(paren, n, PL_reginput);
3316 REGCP_UNWIND(lastcp);
3318 /* Couldn't or didn't -- back up. */
3320 PL_reginput = locinput = HOPc(locinput, -1);
3328 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3330 c = UCHARAT(PL_reginput);
3332 /* If it could work, try it. */
3333 if (c1 == -1000 || c == c1 || c == c2)
3335 TRYPAREN(paren, n, PL_reginput);
3336 REGCP_UNWIND(lastcp);
3338 /* Couldn't or didn't -- back up. */
3340 PL_reginput = locinput = HOPc(locinput, -1);
3347 if (PL_reg_call_cc) {
3348 re_cc_state *cur_call_cc = PL_reg_call_cc;
3349 CURCUR *cctmp = PL_regcc;
3350 regexp *re = PL_reg_re;
3351 CHECKPOINT cp, lastcp;
3353 cp = regcppush(0); /* Save *all* the positions. */
3355 regcp_set_to(PL_reg_call_cc->ss); /* Restore parens of
3357 PL_reginput = locinput; /* Make position available to
3359 cache_re(PL_reg_call_cc->re);
3360 PL_regcc = PL_reg_call_cc->cc;
3361 PL_reg_call_cc = PL_reg_call_cc->prev;
3362 if (regmatch(cur_call_cc->node)) {
3363 PL_reg_call_cc = cur_call_cc;
3367 REGCP_UNWIND(lastcp);
3369 PL_reg_call_cc = cur_call_cc;
3375 PerlIO_printf(Perl_debug_log,
3376 "%*s continuation failed...\n",
3377 REPORT_CODE_OFF+PL_regindent*2, "")
3381 if (locinput < PL_regtill) {
3382 DEBUG_r(PerlIO_printf(Perl_debug_log,
3383 "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n",
3385 (long)(locinput - PL_reg_starttry),
3386 (long)(PL_regtill - PL_reg_starttry),
3388 sayNO_FINAL; /* Cannot match: too short. */
3390 PL_reginput = locinput; /* put where regtry can find it */
3391 sayYES_FINAL; /* Success! */
3393 PL_reginput = locinput; /* put where regtry can find it */
3394 sayYES_LOUD; /* Success! */
3397 PL_reginput = locinput;
3402 s = HOPBACKc(locinput, scan->flags);
3408 PL_reginput = locinput;
3413 s = HOPBACKc(locinput, scan->flags);
3419 PL_reginput = locinput;
3422 inner = NEXTOPER(NEXTOPER(scan));
3423 if (regmatch(inner) != n) {
3438 if (OP(scan) == SUSPEND) {
3439 locinput = PL_reginput;
3440 nextchr = UCHARAT(locinput);
3445 next = scan + ARG(scan);
3450 PerlIO_printf(Perl_error_log, "%"UVxf" %d\n",
3451 PTR2UV(scan), OP(scan));
3452 Perl_croak(aTHX_ "regexp memory corruption");
3459 * We get here only if there's trouble -- normally "case END" is
3460 * the terminating point.
3462 Perl_croak(aTHX_ "corrupted regexp pointers");
3468 PerlIO_printf(Perl_debug_log,
3469 "%*s %scould match...%s\n",
3470 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4],PL_colors[5])
3474 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n",
3475 PL_colors[4],PL_colors[5]));
3481 #if 0 /* Breaks $^R */
3489 PerlIO_printf(Perl_debug_log,
3490 "%*s %sfailed...%s\n",
3491 REPORT_CODE_OFF+PL_regindent*2, "",PL_colors[4],PL_colors[5])
3497 re_unwind_t *uw = SSPTRt(unwind,re_unwind_t);
3500 case RE_UNWIND_BRANCH:
3501 case RE_UNWIND_BRANCHJ:
3503 re_unwind_branch_t *uwb = &(uw->branch);
3504 I32 lastparen = uwb->lastparen;
3506 REGCP_UNWIND(uwb->lastcp);
3507 for (n = *PL_reglastparen; n > lastparen; n--)
3509 *PL_reglastparen = n;
3510 scan = next = uwb->next;
3512 OP(scan) != (uwb->type == RE_UNWIND_BRANCH
3513 ? BRANCH : BRANCHJ) ) { /* Failure */
3520 /* Have more choice yet. Reuse the same uwb. */
3522 if ((n = (uwb->type == RE_UNWIND_BRANCH
3523 ? NEXT_OFF(next) : ARG(next))))
3526 next = NULL; /* XXXX Needn't unwinding in this case... */
3528 next = NEXTOPER(scan);
3529 if (uwb->type == RE_UNWIND_BRANCHJ)
3530 next = NEXTOPER(next);
3531 locinput = uwb->locinput;
3532 nextchr = uwb->nextchr;
3534 PL_regindent = uwb->regindent;
3541 Perl_croak(aTHX_ "regexp unwind memory corruption");
3552 - regrepeat - repeatedly match something simple, report how many
3555 * [This routine now assumes that it will only match on things of length 1.
3556 * That was true before, but now we assume scan - reginput is the count,
3557 * rather than incrementing count on every character. [Er, except utf8.]]
3560 S_regrepeat(pTHX_ regnode *p, I32 max)
3562 register char *scan;
3564 register char *loceol = PL_regeol;
3565 register I32 hardcount = 0;
3566 register bool do_utf8 = DO_UTF8(PL_reg_sv);
3569 if (max != REG_INFTY && max < loceol - scan)
3570 loceol = scan + max;
3575 while (scan < loceol && hardcount < max && *scan != '\n') {
3576 scan += UTF8SKIP(scan);
3580 while (scan < loceol && *scan != '\n')
3587 case EXACT: /* length of string is 1 */
3589 while (scan < loceol && UCHARAT(scan) == c)
3592 case EXACTF: /* length of string is 1 */
3594 while (scan < loceol &&
3595 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold[c]))
3598 case EXACTFL: /* length of string is 1 */
3599 PL_reg_flags |= RF_tainted;
3601 while (scan < loceol &&
3602 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold_locale[c]))
3608 while (hardcount < max && scan < loceol &&
3609 reginclass(p, (U8*)scan, do_utf8)) {
3610 scan += UTF8SKIP(scan);
3614 while (scan < loceol && reginclass(p, (U8*)scan, do_utf8))
3621 LOAD_UTF8_CHARCLASS(alnum,"a");
3622 while (hardcount < max && scan < loceol &&
3623 swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
3624 scan += UTF8SKIP(scan);
3628 while (scan < loceol && isALNUM(*scan))
3633 PL_reg_flags |= RF_tainted;
3636 while (hardcount < max && scan < loceol &&
3637 isALNUM_LC_utf8((U8*)scan)) {
3638 scan += UTF8SKIP(scan);
3642 while (scan < loceol && isALNUM_LC(*scan))
3649 LOAD_UTF8_CHARCLASS(alnum,"a");
3650 while (hardcount < max && scan < loceol &&
3651 !swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
3652 scan += UTF8SKIP(scan);
3656 while (scan < loceol && !isALNUM(*scan))
3661 PL_reg_flags |= RF_tainted;
3664 while (hardcount < max && scan < loceol &&
3665 !isALNUM_LC_utf8((U8*)scan)) {
3666 scan += UTF8SKIP(scan);
3670 while (scan < loceol && !isALNUM_LC(*scan))
3677 LOAD_UTF8_CHARCLASS(space," ");
3678 while (hardcount < max && scan < loceol &&
3680 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
3681 scan += UTF8SKIP(scan);
3685 while (scan < loceol && isSPACE(*scan))
3690 PL_reg_flags |= RF_tainted;
3693 while (hardcount < max && scan < loceol &&
3694 (*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
3695 scan += UTF8SKIP(scan);
3699 while (scan < loceol && isSPACE_LC(*scan))
3706 LOAD_UTF8_CHARCLASS(space," ");
3707 while (hardcount < max && scan < loceol &&
3709 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
3710 scan += UTF8SKIP(scan);
3714 while (scan < loceol && !isSPACE(*scan))
3719 PL_reg_flags |= RF_tainted;
3722 while (hardcount < max && scan < loceol &&
3723 !(*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
3724 scan += UTF8SKIP(scan);
3728 while (scan < loceol && !isSPACE_LC(*scan))
3735 LOAD_UTF8_CHARCLASS(digit,"0");
3736 while (hardcount < max && scan < loceol &&
3737 swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
3738 scan += UTF8SKIP(scan);
3742 while (scan < loceol && isDIGIT(*scan))
3749 LOAD_UTF8_CHARCLASS(digit,"0");
3750 while (hardcount < max && scan < loceol &&
3751 !swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
3752 scan += UTF8SKIP(scan);
3756 while (scan < loceol && !isDIGIT(*scan))
3760 default: /* Called on something of 0 width. */
3761 break; /* So match right here or not at all. */
3767 c = scan - PL_reginput;
3772 SV *prop = sv_newmortal();
3775 PerlIO_printf(Perl_debug_log,
3776 "%*s %s can match %"IVdf" times out of %"IVdf"...\n",
3777 REPORT_CODE_OFF+1, "", SvPVX(prop),(IV)c,(IV)max);
3784 - regrepeat_hard - repeatedly match something, report total lenth and length
3786 * The repeater is supposed to have constant length.
3790 S_regrepeat_hard(pTHX_ regnode *p, I32 max, I32 *lp)
3792 register char *scan = Nullch;
3793 register char *start;
3794 register char *loceol = PL_regeol;
3796 I32 count = 0, res = 1;
3801 start = PL_reginput;
3802 if (DO_UTF8(PL_reg_sv)) {
3803 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
3806 while (start < PL_reginput) {
3808 start += UTF8SKIP(start);
3819 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
3821 *lp = l = PL_reginput - start;
3822 if (max != REG_INFTY && l*max < loceol - scan)
3823 loceol = scan + l*max;
3836 - regclass_swash - prepare the utf8 swash
3840 Perl_regclass_swash(pTHX_ register regnode* node, bool doinit, SV** initsvp)
3845 if (PL_regdata && PL_regdata->count) {
3848 if (PL_regdata->what[n] == 's') {
3849 SV *rv = (SV*)PL_regdata->data[n];
3850 AV *av = (AV*)SvRV((SV*)rv);
3853 si = *av_fetch(av, 0, FALSE);
3854 a = av_fetch(av, 1, FALSE);
3858 else if (si && doinit) {
3859 sw = swash_init("utf8", "", si, 1, 0);
3860 (void)av_store(av, 1, sw);
3872 - reginclass - determine if a character falls into a character class
3876 S_reginclass(pTHX_ register regnode *n, register U8* p, register bool do_utf8)
3878 char flags = ANYOF_FLAGS(n);
3883 c = do_utf8 ? utf8_to_uvchr(p, &len) : *p;
3885 if (do_utf8 || (flags & ANYOF_UNICODE)) {
3886 if (do_utf8 && !ANYOF_RUNTIME(n)) {
3887 if (len != (STRLEN)-1 && c < 256 && ANYOF_BITMAP_TEST(n, c))
3890 if (!match && do_utf8 && (flags & ANYOF_UNICODE_ALL) && c >= 256)
3893 SV *sw = regclass_swash(n, TRUE, 0);
3896 if (swash_fetch(sw, p, do_utf8))
3898 else if (flags & ANYOF_FOLD) {
3899 U8 tmpbuf[UTF8_MAXLEN+1];
3901 if (flags & ANYOF_LOCALE) {
3902 PL_reg_flags |= RF_tainted;
3903 uvchr_to_utf8(tmpbuf, toLOWER_LC_utf8(p));
3906 uvchr_to_utf8(tmpbuf, toLOWER_utf8(p));
3907 if (swash_fetch(sw, tmpbuf, do_utf8))
3913 if (!match && c < 256) {
3914 if (ANYOF_BITMAP_TEST(n, c))
3916 else if (flags & ANYOF_FOLD) {
3919 if (flags & ANYOF_LOCALE) {
3920 PL_reg_flags |= RF_tainted;
3921 f = PL_fold_locale[c];
3925 if (f != c && ANYOF_BITMAP_TEST(n, f))
3929 if (!match && (flags & ANYOF_CLASS)) {
3930 PL_reg_flags |= RF_tainted;
3932 (ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) ||
3933 (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) ||
3934 (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) ||
3935 (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) ||
3936 (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) ||
3937 (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) ||
3938 (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) ||
3939 (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) ||
3940 (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) ||
3941 (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) ||
3942 (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII(c)) ||
3943 (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII(c)) ||
3944 (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) ||
3945 (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) ||
3946 (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) ||
3947 (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) ||
3948 (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) ||
3949 (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) ||
3950 (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) ||
3951 (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) ||
3952 (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) ||
3953 (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) ||
3954 (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) ||
3955 (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) ||
3956 (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) ||
3957 (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) ||
3958 (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) ||
3959 (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) ||
3960 (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK(c)) ||
3961 (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK(c))
3962 ) /* How's that for a conditional? */
3969 return (flags & ANYOF_INVERT) ? !match : match;
3973 S_reghop(pTHX_ U8 *s, I32 off)
3975 return S_reghop3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
3979 S_reghop3(pTHX_ U8 *s, I32 off, U8* lim)
3982 while (off-- && s < lim) {
3983 /* XXX could check well-formedness here */
3991 if (UTF8_IS_CONTINUED(*s)) {
3992 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
3995 /* XXX could check well-formedness here */
4003 S_reghopmaybe(pTHX_ U8 *s, I32 off)
4005 return S_reghopmaybe3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4009 S_reghopmaybe3(pTHX_ U8* s, I32 off, U8* lim)
4012 while (off-- && s < lim) {
4013 /* XXX could check well-formedness here */
4023 if (UTF8_IS_CONTINUED(*s)) {
4024 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4027 /* XXX could check well-formedness here */
4043 restore_pos(pTHXo_ void *arg)
4045 if (PL_reg_eval_set) {
4046 if (PL_reg_oldsaved) {
4047 PL_reg_re->subbeg = PL_reg_oldsaved;
4048 PL_reg_re->sublen = PL_reg_oldsavedlen;
4049 RX_MATCH_COPIED_on(PL_reg_re);
4051 PL_reg_magic->mg_len = PL_reg_oldpos;
4052 PL_reg_eval_set = 0;
4053 PL_curpm = PL_reg_oldcurpm;