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);
1854 Newz(22,PL_reg_curpm, 1, PMOP);
1855 PM_SETRE(PL_reg_curpm, prog);
1856 PL_reg_oldcurpm = PL_curpm;
1857 PL_curpm = PL_reg_curpm;
1858 if (RX_MATCH_COPIED(prog)) {
1859 /* Here is a serious problem: we cannot rewrite subbeg,
1860 since it may be needed if this match fails. Thus
1861 $` inside (?{}) could fail... */
1862 PL_reg_oldsaved = prog->subbeg;
1863 PL_reg_oldsavedlen = prog->sublen;
1864 RX_MATCH_COPIED_off(prog);
1867 PL_reg_oldsaved = Nullch;
1868 prog->subbeg = PL_bostr;
1869 prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */
1871 prog->startp[0] = startpos - PL_bostr;
1872 PL_reginput = startpos;
1873 PL_regstartp = prog->startp;
1874 PL_regendp = prog->endp;
1875 PL_reglastparen = &prog->lastparen;
1876 PL_reglastcloseparen = &prog->lastcloseparen;
1877 prog->lastparen = 0;
1879 DEBUG_r(PL_reg_starttry = startpos);
1880 if (PL_reg_start_tmpl <= prog->nparens) {
1881 PL_reg_start_tmpl = prog->nparens*3/2 + 3;
1882 if(PL_reg_start_tmp)
1883 Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
1885 New(22,PL_reg_start_tmp, PL_reg_start_tmpl, char*);
1888 /* XXXX What this code is doing here?!!! There should be no need
1889 to do this again and again, PL_reglastparen should take care of
1892 /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code.
1893 * Actually, the code in regcppop() (which Ilya may be meaning by
1894 * PL_reglastparen), is not needed at all by the test suite
1895 * (op/regexp, op/pat, op/split), but that code is needed, oddly
1896 * enough, for building DynaLoader, or otherwise this
1897 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
1898 * will happen. Meanwhile, this code *is* needed for the
1899 * above-mentioned test suite tests to succeed. The common theme
1900 * on those tests seems to be returning null fields from matches.
1905 if (prog->nparens) {
1906 for (i = prog->nparens; i > *PL_reglastparen; i--) {
1913 if (regmatch(prog->program + 1)) {
1914 prog->endp[0] = PL_reginput - PL_bostr;
1917 REGCP_UNWIND(lastcp);
1921 #define RE_UNWIND_BRANCH 1
1922 #define RE_UNWIND_BRANCHJ 2
1926 typedef struct { /* XX: makes sense to enlarge it... */
1930 } re_unwind_generic_t;
1943 } re_unwind_branch_t;
1945 typedef union re_unwind_t {
1947 re_unwind_generic_t generic;
1948 re_unwind_branch_t branch;
1952 - regmatch - main matching routine
1954 * Conceptually the strategy is simple: check to see whether the current
1955 * node matches, call self recursively to see whether the rest matches,
1956 * and then act accordingly. In practice we make some effort to avoid
1957 * recursion, in particular by going through "ordinary" nodes (that don't
1958 * need to know whether the rest of the match failed) by a loop instead of
1961 /* [lwall] I've hoisted the register declarations to the outer block in order to
1962 * maybe save a little bit of pushing and popping on the stack. It also takes
1963 * advantage of machines that use a register save mask on subroutine entry.
1965 STATIC I32 /* 0 failure, 1 success */
1966 S_regmatch(pTHX_ regnode *prog)
1968 register regnode *scan; /* Current node. */
1969 regnode *next; /* Next node. */
1970 regnode *inner; /* Next node in internal branch. */
1971 register I32 nextchr; /* renamed nextchr - nextchar colides with
1972 function of same name */
1973 register I32 n; /* no or next */
1974 register I32 ln = 0; /* len or last */
1975 register char *s = Nullch; /* operand or save */
1976 register char *locinput = PL_reginput;
1977 register I32 c1 = 0, c2 = 0, paren; /* case fold search, parenth */
1978 int minmod = 0, sw = 0, logical = 0;
1981 I32 firstcp = PL_savestack_ix;
1983 register bool do_utf8 = DO_UTF8(PL_reg_sv);
1989 /* Note that nextchr is a byte even in UTF */
1990 nextchr = UCHARAT(locinput);
1992 while (scan != NULL) {
1993 #define sayNO_L (logical ? (logical = 0, sw = 0, goto cont) : sayNO)
1995 # define sayYES goto yes
1996 # define sayNO goto no
1997 # define sayYES_FINAL goto yes_final
1998 # define sayYES_LOUD goto yes_loud
1999 # define sayNO_FINAL goto no_final
2000 # define sayNO_SILENT goto do_no
2001 # define saySAME(x) if (x) goto yes; else goto no
2002 # define REPORT_CODE_OFF 24
2004 # define sayYES return 1
2005 # define sayNO return 0
2006 # define sayYES_FINAL return 1
2007 # define sayYES_LOUD return 1
2008 # define sayNO_FINAL return 0
2009 # define sayNO_SILENT return 0
2010 # define saySAME(x) return x
2013 SV *prop = sv_newmortal();
2014 int docolor = *PL_colors[0];
2015 int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */
2016 int l = (PL_regeol - locinput) > taill ? taill : (PL_regeol - locinput);
2017 /* The part of the string before starttry has one color
2018 (pref0_len chars), between starttry and current
2019 position another one (pref_len - pref0_len chars),
2020 after the current position the third one.
2021 We assume that pref0_len <= pref_len, otherwise we
2022 decrease pref0_len. */
2023 int pref_len = (locinput - PL_bostr) > (5 + taill) - l
2024 ? (5 + taill) - l : locinput - PL_bostr;
2027 while (UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len)))
2029 pref0_len = pref_len - (locinput - PL_reg_starttry);
2030 if (l + pref_len < (5 + taill) && l < PL_regeol - locinput)
2031 l = ( PL_regeol - locinput > (5 + taill) - pref_len
2032 ? (5 + taill) - pref_len : PL_regeol - locinput);
2033 while (UTF8_IS_CONTINUATION(*(U8*)(locinput + l)))
2037 if (pref0_len > pref_len)
2038 pref0_len = pref_len;
2039 regprop(prop, scan);
2040 PerlIO_printf(Perl_debug_log,
2041 "%4"IVdf" <%s%.*s%s%s%.*s%s%s%s%.*s%s>%*s|%3"IVdf":%*s%s\n",
2042 (IV)(locinput - PL_bostr),
2043 PL_colors[4], pref0_len,
2044 locinput - pref_len, PL_colors[5],
2045 PL_colors[2], pref_len - pref0_len,
2046 locinput - pref_len + pref0_len, PL_colors[3],
2047 (docolor ? "" : "> <"),
2048 PL_colors[0], l, locinput, PL_colors[1],
2049 15 - l - pref_len + 1,
2051 (IV)(scan - PL_regprogram), PL_regindent*2, "",
2055 next = scan + NEXT_OFF(scan);
2061 if (locinput == PL_bostr || (PL_multiline &&
2062 (nextchr || locinput < PL_regeol) && locinput[-1] == '\n') )
2064 /* regtill = regbol; */
2069 if (locinput == PL_bostr ||
2070 ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n'))
2076 if (locinput == PL_bostr)
2080 if (locinput == PL_reg_ganch)
2090 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2095 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2097 if (PL_regeol - locinput > 1)
2101 if (PL_regeol != locinput)
2105 if (!nextchr && locinput >= PL_regeol)
2107 nextchr = UCHARAT(++locinput);
2110 if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n')
2113 locinput += PL_utf8skip[nextchr];
2114 if (locinput > PL_regeol)
2116 nextchr = UCHARAT(locinput);
2119 nextchr = UCHARAT(++locinput);
2124 if (do_utf8 != (UTF!=0)) {
2132 if (*((U8*)s) != utf8_to_uvchr((U8*)l, &len))
2141 if (*((U8*)l) != utf8_to_uvchr((U8*)s, &len))
2147 nextchr = UCHARAT(locinput);
2150 /* Inline the first character, for speed. */
2151 if (UCHARAT(s) != nextchr)
2153 if (PL_regeol - locinput < ln)
2155 if (ln > 1 && memNE(s, locinput, ln))
2158 nextchr = UCHARAT(locinput);
2161 PL_reg_flags |= RF_tainted;
2171 c1 = OP(scan) == EXACTF;
2173 if (l >= PL_regeol) {
2176 if ((UTF ? utf8n_to_uvchr((U8*)s, e - s, 0, 0) : *((U8*)s)) !=
2177 (c1 ? toLOWER_utf8((U8*)l) : toLOWER_LC_utf8((U8*)l)))
2179 s += UTF ? UTF8SKIP(s) : 1;
2183 nextchr = UCHARAT(locinput);
2187 /* Inline the first character, for speed. */
2188 if (UCHARAT(s) != nextchr &&
2189 UCHARAT(s) != ((OP(scan) == EXACTF)
2190 ? PL_fold : PL_fold_locale)[nextchr])
2192 if (PL_regeol - locinput < ln)
2194 if (ln > 1 && (OP(scan) == EXACTF
2195 ? ibcmp(s, locinput, ln)
2196 : ibcmp_locale(s, locinput, ln)))
2199 nextchr = UCHARAT(locinput);
2203 if (!reginclass(scan, (U8*)locinput, do_utf8))
2205 if (locinput >= PL_regeol)
2207 locinput += PL_utf8skip[nextchr];
2208 nextchr = UCHARAT(locinput);
2212 nextchr = UCHARAT(locinput);
2213 if (!reginclass(scan, (U8*)locinput, do_utf8))
2215 if (!nextchr && locinput >= PL_regeol)
2217 nextchr = UCHARAT(++locinput);
2221 PL_reg_flags |= RF_tainted;
2227 LOAD_UTF8_CHARCLASS(alnum,"a");
2228 if (!(OP(scan) == ALNUM
2229 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2230 : isALNUM_LC_utf8((U8*)locinput)))
2234 locinput += PL_utf8skip[nextchr];
2235 nextchr = UCHARAT(locinput);
2238 if (!(OP(scan) == ALNUM
2239 ? isALNUM(nextchr) : isALNUM_LC(nextchr)))
2241 nextchr = UCHARAT(++locinput);
2244 PL_reg_flags |= RF_tainted;
2247 if (!nextchr && locinput >= PL_regeol)
2250 LOAD_UTF8_CHARCLASS(alnum,"a");
2251 if (OP(scan) == NALNUM
2252 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2253 : isALNUM_LC_utf8((U8*)locinput))
2257 locinput += PL_utf8skip[nextchr];
2258 nextchr = UCHARAT(locinput);
2261 if (OP(scan) == NALNUM
2262 ? isALNUM(nextchr) : isALNUM_LC(nextchr))
2264 nextchr = UCHARAT(++locinput);
2268 PL_reg_flags |= RF_tainted;
2272 /* was last char in word? */
2274 if (locinput == PL_bostr)
2277 U8 *r = reghop((U8*)locinput, -1);
2279 ln = utf8n_to_uvchr(r, s - (char*)r, 0, 0);
2281 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2282 ln = isALNUM_uni(ln);
2283 LOAD_UTF8_CHARCLASS(alnum,"a");
2284 n = swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8);
2287 ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(ln));
2288 n = isALNUM_LC_utf8((U8*)locinput);
2292 ln = (locinput != PL_bostr) ?
2293 UCHARAT(locinput - 1) : '\n';
2294 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2296 n = isALNUM(nextchr);
2299 ln = isALNUM_LC(ln);
2300 n = isALNUM_LC(nextchr);
2303 if (((!ln) == (!n)) == (OP(scan) == BOUND ||
2304 OP(scan) == BOUNDL))
2308 PL_reg_flags |= RF_tainted;
2314 if (UTF8_IS_CONTINUED(nextchr)) {
2315 LOAD_UTF8_CHARCLASS(space," ");
2316 if (!(OP(scan) == SPACE
2317 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2318 : isSPACE_LC_utf8((U8*)locinput)))
2322 locinput += PL_utf8skip[nextchr];
2323 nextchr = UCHARAT(locinput);
2326 if (!(OP(scan) == SPACE
2327 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2329 nextchr = UCHARAT(++locinput);
2332 if (!(OP(scan) == SPACE
2333 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2335 nextchr = UCHARAT(++locinput);
2339 PL_reg_flags |= RF_tainted;
2342 if (!nextchr && locinput >= PL_regeol)
2345 LOAD_UTF8_CHARCLASS(space," ");
2346 if (OP(scan) == NSPACE
2347 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2348 : isSPACE_LC_utf8((U8*)locinput))
2352 locinput += PL_utf8skip[nextchr];
2353 nextchr = UCHARAT(locinput);
2356 if (OP(scan) == NSPACE
2357 ? isSPACE(nextchr) : isSPACE_LC(nextchr))
2359 nextchr = UCHARAT(++locinput);
2362 PL_reg_flags |= RF_tainted;
2368 LOAD_UTF8_CHARCLASS(digit,"0");
2369 if (!(OP(scan) == DIGIT
2370 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2371 : isDIGIT_LC_utf8((U8*)locinput)))
2375 locinput += PL_utf8skip[nextchr];
2376 nextchr = UCHARAT(locinput);
2379 if (!(OP(scan) == DIGIT
2380 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr)))
2382 nextchr = UCHARAT(++locinput);
2385 PL_reg_flags |= RF_tainted;
2388 if (!nextchr && locinput >= PL_regeol)
2391 LOAD_UTF8_CHARCLASS(digit,"0");
2392 if (OP(scan) == NDIGIT
2393 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2394 : isDIGIT_LC_utf8((U8*)locinput))
2398 locinput += PL_utf8skip[nextchr];
2399 nextchr = UCHARAT(locinput);
2402 if (OP(scan) == NDIGIT
2403 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr))
2405 nextchr = UCHARAT(++locinput);
2408 LOAD_UTF8_CHARCLASS(mark,"~");
2409 if (locinput >= PL_regeol ||
2410 swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2412 locinput += PL_utf8skip[nextchr];
2413 while (locinput < PL_regeol &&
2414 swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2415 locinput += UTF8SKIP(locinput);
2416 if (locinput > PL_regeol)
2418 nextchr = UCHARAT(locinput);
2421 PL_reg_flags |= RF_tainted;
2425 n = ARG(scan); /* which paren pair */
2426 ln = PL_regstartp[n];
2427 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2428 if (*PL_reglastparen < n || ln == -1)
2429 sayNO; /* Do not match unless seen CLOSEn. */
2430 if (ln == PL_regendp[n])
2434 if (do_utf8 && OP(scan) != REF) { /* REF can do byte comparison */
2436 char *e = PL_bostr + PL_regendp[n];
2438 * Note that we can't do the "other character" lookup trick as
2439 * in the 8-bit case (no pun intended) because in Unicode we
2440 * have to map both upper and title case to lower case.
2442 if (OP(scan) == REFF) {
2446 if (toLOWER_utf8((U8*)s) != toLOWER_utf8((U8*)l))
2456 if (toLOWER_LC_utf8((U8*)s) != toLOWER_LC_utf8((U8*)l))
2463 nextchr = UCHARAT(locinput);
2467 /* Inline the first character, for speed. */
2468 if (UCHARAT(s) != nextchr &&
2470 (UCHARAT(s) != ((OP(scan) == REFF
2471 ? PL_fold : PL_fold_locale)[nextchr]))))
2473 ln = PL_regendp[n] - ln;
2474 if (locinput + ln > PL_regeol)
2476 if (ln > 1 && (OP(scan) == REF
2477 ? memNE(s, locinput, ln)
2479 ? ibcmp(s, locinput, ln)
2480 : ibcmp_locale(s, locinput, ln))))
2483 nextchr = UCHARAT(locinput);
2494 OP_4tree *oop = PL_op;
2495 COP *ocurcop = PL_curcop;
2496 SV **ocurpad = PL_curpad;
2500 PL_op = (OP_4tree*)PL_regdata->data[n];
2501 DEBUG_r( PerlIO_printf(Perl_debug_log, " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) );
2502 PL_curpad = AvARRAY((AV*)PL_regdata->data[n + 2]);
2503 PL_regendp[0] = PL_reg_magic->mg_len = locinput - PL_bostr;
2505 CALLRUNOPS(aTHX); /* Scalar context. */
2511 PL_curpad = ocurpad;
2512 PL_curcop = ocurcop;
2514 if (logical == 2) { /* Postponed subexpression. */
2516 MAGIC *mg = Null(MAGIC*);
2518 CHECKPOINT cp, lastcp;
2520 if(SvROK(ret) || SvRMAGICAL(ret)) {
2521 SV *sv = SvROK(ret) ? SvRV(ret) : ret;
2524 mg = mg_find(sv, PERL_MAGIC_qr);
2527 re = (regexp *)mg->mg_obj;
2528 (void)ReREFCNT_inc(re);
2532 char *t = SvPV(ret, len);
2534 char *oprecomp = PL_regprecomp;
2535 I32 osize = PL_regsize;
2536 I32 onpar = PL_regnpar;
2539 re = CALLREGCOMP(aTHX_ t, t + len, &pm);
2541 & (SVs_TEMP | SVs_PADTMP | SVf_READONLY)))
2542 sv_magic(ret,(SV*)ReREFCNT_inc(re),
2544 PL_regprecomp = oprecomp;
2549 PerlIO_printf(Perl_debug_log,
2550 "Entering embedded `%s%.60s%s%s'\n",
2554 (strlen(re->precomp) > 60 ? "..." : ""))
2557 state.prev = PL_reg_call_cc;
2558 state.cc = PL_regcc;
2559 state.re = PL_reg_re;
2563 cp = regcppush(0); /* Save *all* the positions. */
2566 state.ss = PL_savestack_ix;
2567 *PL_reglastparen = 0;
2568 *PL_reglastcloseparen = 0;
2569 PL_reg_call_cc = &state;
2570 PL_reginput = locinput;
2572 /* XXXX This is too dramatic a measure... */
2575 if (regmatch(re->program + 1)) {
2576 /* Even though we succeeded, we need to restore
2577 global variables, since we may be wrapped inside
2578 SUSPEND, thus the match may be not finished yet. */
2580 /* XXXX Do this only if SUSPENDed? */
2581 PL_reg_call_cc = state.prev;
2582 PL_regcc = state.cc;
2583 PL_reg_re = state.re;
2584 cache_re(PL_reg_re);
2586 /* XXXX This is too dramatic a measure... */
2589 /* These are needed even if not SUSPEND. */
2595 REGCP_UNWIND(lastcp);
2597 PL_reg_call_cc = state.prev;
2598 PL_regcc = state.cc;
2599 PL_reg_re = state.re;
2600 cache_re(PL_reg_re);
2602 /* XXXX This is too dramatic a measure... */
2611 sv_setsv(save_scalar(PL_replgv), ret);
2615 n = ARG(scan); /* which paren pair */
2616 PL_reg_start_tmp[n] = locinput;
2621 n = ARG(scan); /* which paren pair */
2622 PL_regstartp[n] = PL_reg_start_tmp[n] - PL_bostr;
2623 PL_regendp[n] = locinput - PL_bostr;
2624 if (n > *PL_reglastparen)
2625 *PL_reglastparen = n;
2626 *PL_reglastcloseparen = n;
2629 n = ARG(scan); /* which paren pair */
2630 sw = (*PL_reglastparen >= n && PL_regendp[n] != -1);
2633 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2635 next = NEXTOPER(NEXTOPER(scan));
2637 next = scan + ARG(scan);
2638 if (OP(next) == IFTHEN) /* Fake one. */
2639 next = NEXTOPER(NEXTOPER(next));
2643 logical = scan->flags;
2645 /*******************************************************************
2646 PL_regcc contains infoblock about the innermost (...)* loop, and
2647 a pointer to the next outer infoblock.
2649 Here is how Y(A)*Z is processed (if it is compiled into CURLYX/WHILEM):
2651 1) After matching X, regnode for CURLYX is processed;
2653 2) This regnode creates infoblock on the stack, and calls
2654 regmatch() recursively with the starting point at WHILEM node;
2656 3) Each hit of WHILEM node tries to match A and Z (in the order
2657 depending on the current iteration, min/max of {min,max} and
2658 greediness). The information about where are nodes for "A"
2659 and "Z" is read from the infoblock, as is info on how many times "A"
2660 was already matched, and greediness.
2662 4) After A matches, the same WHILEM node is hit again.
2664 5) Each time WHILEM is hit, PL_regcc is the infoblock created by CURLYX
2665 of the same pair. Thus when WHILEM tries to match Z, it temporarily
2666 resets PL_regcc, since this Y(A)*Z can be a part of some other loop:
2667 as in (Y(A)*Z)*. If Z matches, the automaton will hit the WHILEM node
2668 of the external loop.
2670 Currently present infoblocks form a tree with a stem formed by PL_curcc
2671 and whatever it mentions via ->next, and additional attached trees
2672 corresponding to temporarily unset infoblocks as in "5" above.
2674 In the following picture infoblocks for outer loop of
2675 (Y(A)*?Z)*?T are denoted O, for inner I. NULL starting block
2676 is denoted by x. The matched string is YAAZYAZT. Temporarily postponed
2677 infoblocks are drawn below the "reset" infoblock.
2679 In fact in the picture below we do not show failed matches for Z and T
2680 by WHILEM blocks. [We illustrate minimal matches, since for them it is
2681 more obvious *why* one needs to *temporary* unset infoblocks.]
2683 Matched REx position InfoBlocks Comment
2687 Y A)*?Z)*?T x <- O <- I
2688 YA )*?Z)*?T x <- O <- I
2689 YA A)*?Z)*?T x <- O <- I
2690 YAA )*?Z)*?T x <- O <- I
2691 YAA Z)*?T x <- O # Temporary unset I
2694 YAAZ Y(A)*?Z)*?T x <- O
2697 YAAZY (A)*?Z)*?T x <- O
2700 YAAZY A)*?Z)*?T x <- O <- I
2703 YAAZYA )*?Z)*?T x <- O <- I
2706 YAAZYA Z)*?T x <- O # Temporary unset I
2712 YAAZYAZ T x # Temporary unset O
2719 *******************************************************************/
2722 CHECKPOINT cp = PL_savestack_ix;
2723 /* No need to save/restore up to this paren */
2724 I32 parenfloor = scan->flags;
2726 if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */
2728 cc.oldcc = PL_regcc;
2730 /* XXXX Probably it is better to teach regpush to support
2731 parenfloor > PL_regsize... */
2732 if (parenfloor > *PL_reglastparen)
2733 parenfloor = *PL_reglastparen; /* Pessimization... */
2734 cc.parenfloor = parenfloor;
2736 cc.min = ARG1(scan);
2737 cc.max = ARG2(scan);
2738 cc.scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
2742 PL_reginput = locinput;
2743 n = regmatch(PREVOPER(next)); /* start on the WHILEM */
2745 PL_regcc = cc.oldcc;
2751 * This is really hard to understand, because after we match
2752 * what we're trying to match, we must make sure the rest of
2753 * the REx is going to match for sure, and to do that we have
2754 * to go back UP the parse tree by recursing ever deeper. And
2755 * if it fails, we have to reset our parent's current state
2756 * that we can try again after backing off.
2759 CHECKPOINT cp, lastcp;
2760 CURCUR* cc = PL_regcc;
2761 char *lastloc = cc->lastloc; /* Detection of 0-len. */
2763 n = cc->cur + 1; /* how many we know we matched */
2764 PL_reginput = locinput;
2767 PerlIO_printf(Perl_debug_log,
2768 "%*s %ld out of %ld..%ld cc=%lx\n",
2769 REPORT_CODE_OFF+PL_regindent*2, "",
2770 (long)n, (long)cc->min,
2771 (long)cc->max, (long)cc)
2774 /* If degenerate scan matches "", assume scan done. */
2776 if (locinput == cc->lastloc && n >= cc->min) {
2777 PL_regcc = cc->oldcc;
2781 PerlIO_printf(Perl_debug_log,
2782 "%*s empty match detected, try continuation...\n",
2783 REPORT_CODE_OFF+PL_regindent*2, "")
2785 if (regmatch(cc->next))
2793 /* First just match a string of min scans. */
2797 cc->lastloc = locinput;
2798 if (regmatch(cc->scan))
2801 cc->lastloc = lastloc;
2806 /* Check whether we already were at this position.
2807 Postpone detection until we know the match is not
2808 *that* much linear. */
2809 if (!PL_reg_maxiter) {
2810 PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4);
2811 PL_reg_leftiter = PL_reg_maxiter;
2813 if (PL_reg_leftiter-- == 0) {
2814 I32 size = (PL_reg_maxiter + 7)/8;
2815 if (PL_reg_poscache) {
2816 if (PL_reg_poscache_size < size) {
2817 Renew(PL_reg_poscache, size, char);
2818 PL_reg_poscache_size = size;
2820 Zero(PL_reg_poscache, size, char);
2823 PL_reg_poscache_size = size;
2824 Newz(29, PL_reg_poscache, size, char);
2827 PerlIO_printf(Perl_debug_log,
2828 "%sDetected a super-linear match, switching on caching%s...\n",
2829 PL_colors[4], PL_colors[5])
2832 if (PL_reg_leftiter < 0) {
2833 I32 o = locinput - PL_bostr, b;
2835 o = (scan->flags & 0xf) - 1 + o * (scan->flags>>4);
2838 if (PL_reg_poscache[o] & (1<<b)) {
2840 PerlIO_printf(Perl_debug_log,
2841 "%*s already tried at this position...\n",
2842 REPORT_CODE_OFF+PL_regindent*2, "")
2846 PL_reg_poscache[o] |= (1<<b);
2850 /* Prefer next over scan for minimal matching. */
2853 PL_regcc = cc->oldcc;
2856 cp = regcppush(cc->parenfloor);
2858 if (regmatch(cc->next)) {
2860 sayYES; /* All done. */
2862 REGCP_UNWIND(lastcp);
2868 if (n >= cc->max) { /* Maximum greed exceeded? */
2869 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
2870 && !(PL_reg_flags & RF_warned)) {
2871 PL_reg_flags |= RF_warned;
2872 Perl_warner(aTHX_ WARN_REGEXP, "%s limit (%d) exceeded",
2873 "Complex regular subexpression recursion",
2880 PerlIO_printf(Perl_debug_log,
2881 "%*s trying longer...\n",
2882 REPORT_CODE_OFF+PL_regindent*2, "")
2884 /* Try scanning more and see if it helps. */
2885 PL_reginput = locinput;
2887 cc->lastloc = locinput;
2888 cp = regcppush(cc->parenfloor);
2890 if (regmatch(cc->scan)) {
2894 REGCP_UNWIND(lastcp);
2897 cc->lastloc = lastloc;
2901 /* Prefer scan over next for maximal matching. */
2903 if (n < cc->max) { /* More greed allowed? */
2904 cp = regcppush(cc->parenfloor);
2906 cc->lastloc = locinput;
2908 if (regmatch(cc->scan)) {
2912 REGCP_UNWIND(lastcp);
2913 regcppop(); /* Restore some previous $<digit>s? */
2914 PL_reginput = locinput;
2916 PerlIO_printf(Perl_debug_log,
2917 "%*s failed, try continuation...\n",
2918 REPORT_CODE_OFF+PL_regindent*2, "")
2921 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
2922 && !(PL_reg_flags & RF_warned)) {
2923 PL_reg_flags |= RF_warned;
2924 Perl_warner(aTHX_ WARN_REGEXP, "%s limit (%d) exceeded",
2925 "Complex regular subexpression recursion",
2929 /* Failed deeper matches of scan, so see if this one works. */
2930 PL_regcc = cc->oldcc;
2933 if (regmatch(cc->next))
2939 cc->lastloc = lastloc;
2944 next = scan + ARG(scan);
2947 inner = NEXTOPER(NEXTOPER(scan));
2950 inner = NEXTOPER(scan);
2954 if (OP(next) != c1) /* No choice. */
2955 next = inner; /* Avoid recursion. */
2957 I32 lastparen = *PL_reglastparen;
2959 re_unwind_branch_t *uw;
2961 /* Put unwinding data on stack */
2962 unwind1 = SSNEWt(1,re_unwind_branch_t);
2963 uw = SSPTRt(unwind1,re_unwind_branch_t);
2966 uw->type = ((c1 == BRANCH)
2968 : RE_UNWIND_BRANCHJ);
2969 uw->lastparen = lastparen;
2971 uw->locinput = locinput;
2972 uw->nextchr = nextchr;
2974 uw->regindent = ++PL_regindent;
2977 REGCP_SET(uw->lastcp);
2979 /* Now go into the first branch */
2992 /* We suppose that the next guy does not need
2993 backtracking: in particular, it is of constant length,
2994 and has no parenths to influence future backrefs. */
2995 ln = ARG1(scan); /* min to match */
2996 n = ARG2(scan); /* max to match */
2997 paren = scan->flags;
2999 if (paren > PL_regsize)
3001 if (paren > *PL_reglastparen)
3002 *PL_reglastparen = paren;
3004 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3006 scan += NEXT_OFF(scan); /* Skip former OPEN. */
3007 PL_reginput = locinput;
3010 if (ln && regrepeat_hard(scan, ln, &l) < ln)
3012 if (ln && l == 0 && n >= ln
3013 /* In fact, this is tricky. If paren, then the
3014 fact that we did/didnot match may influence
3015 future execution. */
3016 && !(paren && ln == 0))
3018 locinput = PL_reginput;
3019 if (PL_regkind[(U8)OP(next)] == EXACT) {
3020 c1 = (U8)*STRING(next);
3021 if (OP(next) == EXACTF)
3023 else if (OP(next) == EXACTFL)
3024 c2 = PL_fold_locale[c1];
3031 /* This may be improved if l == 0. */
3032 while (n >= ln || (n == REG_INFTY && ln > 0 && l)) { /* ln overflow ? */
3033 /* If it could work, try it. */
3035 UCHARAT(PL_reginput) == c1 ||
3036 UCHARAT(PL_reginput) == c2)
3040 PL_regstartp[paren] =
3041 HOPc(PL_reginput, -l) - PL_bostr;
3042 PL_regendp[paren] = PL_reginput - PL_bostr;
3045 PL_regendp[paren] = -1;
3049 REGCP_UNWIND(lastcp);
3051 /* Couldn't or didn't -- move forward. */
3052 PL_reginput = locinput;
3053 if (regrepeat_hard(scan, 1, &l)) {
3055 locinput = PL_reginput;
3062 n = regrepeat_hard(scan, n, &l);
3063 if (n != 0 && l == 0
3064 /* In fact, this is tricky. If paren, then the
3065 fact that we did/didnot match may influence
3066 future execution. */
3067 && !(paren && ln == 0))
3069 locinput = PL_reginput;
3071 PerlIO_printf(Perl_debug_log,
3072 "%*s matched %"IVdf" times, len=%"IVdf"...\n",
3073 (int)(REPORT_CODE_OFF+PL_regindent*2), "",
3077 if (PL_regkind[(U8)OP(next)] == EXACT) {
3078 c1 = (U8)*STRING(next);
3079 if (OP(next) == EXACTF)
3081 else if (OP(next) == EXACTFL)
3082 c2 = PL_fold_locale[c1];
3091 /* If it could work, try it. */
3093 UCHARAT(PL_reginput) == c1 ||
3094 UCHARAT(PL_reginput) == c2)
3097 PerlIO_printf(Perl_debug_log,
3098 "%*s trying tail with n=%"IVdf"...\n",
3099 (int)(REPORT_CODE_OFF+PL_regindent*2), "", (IV)n)
3103 PL_regstartp[paren] = HOPc(PL_reginput, -l) - PL_bostr;
3104 PL_regendp[paren] = PL_reginput - PL_bostr;
3107 PL_regendp[paren] = -1;
3111 REGCP_UNWIND(lastcp);
3113 /* Couldn't or didn't -- back up. */
3115 locinput = HOPc(locinput, -l);
3116 PL_reginput = locinput;
3123 paren = scan->flags; /* Which paren to set */
3124 if (paren > PL_regsize)
3126 if (paren > *PL_reglastparen)
3127 *PL_reglastparen = paren;
3128 ln = ARG1(scan); /* min to match */
3129 n = ARG2(scan); /* max to match */
3130 scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE);
3134 ln = ARG1(scan); /* min to match */
3135 n = ARG2(scan); /* max to match */
3136 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3141 scan = NEXTOPER(scan);
3147 scan = NEXTOPER(scan);
3151 * Lookahead to avoid useless match attempts
3152 * when we know what character comes next.
3154 if (PL_regkind[(U8)OP(next)] == EXACT) {
3155 U8 *s = (U8*)STRING(next);
3158 if (OP(next) == EXACTF)
3160 else if (OP(next) == EXACTFL)
3161 c2 = PL_fold_locale[c1];
3164 if (OP(next) == EXACTF) {
3165 c1 = to_utf8_lower(s);
3166 c2 = to_utf8_upper(s);
3169 c2 = c1 = utf8_to_uvchr(s, NULL);
3175 PL_reginput = locinput;
3179 if (ln && regrepeat(scan, ln) < ln)
3181 locinput = PL_reginput;
3184 char *e; /* Should not check after this */
3185 char *old = locinput;
3187 if (n == REG_INFTY) {
3190 while (UTF8_IS_CONTINUATION(*(U8*)e))
3196 m >0 && e + UTF8SKIP(e) <= PL_regeol; m--)
3200 e = locinput + n - ln;
3206 /* Find place 'next' could work */
3209 while (locinput <= e && *locinput != c1)
3212 while (locinput <= e
3217 count = locinput - old;
3224 utf8_to_uvchr((U8*)locinput, &len) != c1;
3229 for (count = 0; locinput <= e; count++) {
3230 UV c = utf8_to_uvchr((U8*)locinput, &len);
3231 if (c == c1 || c == c2)
3239 /* PL_reginput == old now */
3240 if (locinput != old) {
3241 ln = 1; /* Did some */
3242 if (regrepeat(scan, count) < count)
3245 /* PL_reginput == locinput now */
3246 TRYPAREN(paren, ln, locinput);
3247 PL_reginput = locinput; /* Could be reset... */
3248 REGCP_UNWIND(lastcp);
3249 /* Couldn't or didn't -- move forward. */
3252 locinput += UTF8SKIP(locinput);
3258 while (n >= ln || (n == REG_INFTY && ln > 0)) { /* ln overflow ? */
3262 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3264 c = UCHARAT(PL_reginput);
3265 /* If it could work, try it. */
3266 if (c == c1 || c == c2)
3268 TRYPAREN(paren, n, PL_reginput);
3269 REGCP_UNWIND(lastcp);
3272 /* If it could work, try it. */
3273 else if (c1 == -1000)
3275 TRYPAREN(paren, n, PL_reginput);
3276 REGCP_UNWIND(lastcp);
3278 /* Couldn't or didn't -- move forward. */
3279 PL_reginput = locinput;
3280 if (regrepeat(scan, 1)) {
3282 locinput = PL_reginput;
3290 n = regrepeat(scan, n);
3291 locinput = PL_reginput;
3292 if (ln < n && PL_regkind[(U8)OP(next)] == EOL &&
3293 (!PL_multiline || OP(next) == SEOL || OP(next) == EOS)) {
3294 ln = n; /* why back off? */
3295 /* ...because $ and \Z can match before *and* after
3296 newline at the end. Consider "\n\n" =~ /\n+\Z\n/.
3297 We should back off by one in this case. */
3298 if (UCHARAT(PL_reginput - 1) == '\n' && OP(next) != EOS)
3307 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3309 c = UCHARAT(PL_reginput);
3311 /* If it could work, try it. */
3312 if (c1 == -1000 || c == c1 || c == c2)
3314 TRYPAREN(paren, n, PL_reginput);
3315 REGCP_UNWIND(lastcp);
3317 /* Couldn't or didn't -- back up. */
3319 PL_reginput = locinput = HOPc(locinput, -1);
3327 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3329 c = UCHARAT(PL_reginput);
3331 /* If it could work, try it. */
3332 if (c1 == -1000 || c == c1 || c == c2)
3334 TRYPAREN(paren, n, PL_reginput);
3335 REGCP_UNWIND(lastcp);
3337 /* Couldn't or didn't -- back up. */
3339 PL_reginput = locinput = HOPc(locinput, -1);
3346 if (PL_reg_call_cc) {
3347 re_cc_state *cur_call_cc = PL_reg_call_cc;
3348 CURCUR *cctmp = PL_regcc;
3349 regexp *re = PL_reg_re;
3350 CHECKPOINT cp, lastcp;
3352 cp = regcppush(0); /* Save *all* the positions. */
3354 regcp_set_to(PL_reg_call_cc->ss); /* Restore parens of
3356 PL_reginput = locinput; /* Make position available to
3358 cache_re(PL_reg_call_cc->re);
3359 PL_regcc = PL_reg_call_cc->cc;
3360 PL_reg_call_cc = PL_reg_call_cc->prev;
3361 if (regmatch(cur_call_cc->node)) {
3362 PL_reg_call_cc = cur_call_cc;
3366 REGCP_UNWIND(lastcp);
3368 PL_reg_call_cc = cur_call_cc;
3374 PerlIO_printf(Perl_debug_log,
3375 "%*s continuation failed...\n",
3376 REPORT_CODE_OFF+PL_regindent*2, "")
3380 if (locinput < PL_regtill) {
3381 DEBUG_r(PerlIO_printf(Perl_debug_log,
3382 "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n",
3384 (long)(locinput - PL_reg_starttry),
3385 (long)(PL_regtill - PL_reg_starttry),
3387 sayNO_FINAL; /* Cannot match: too short. */
3389 PL_reginput = locinput; /* put where regtry can find it */
3390 sayYES_FINAL; /* Success! */
3392 PL_reginput = locinput; /* put where regtry can find it */
3393 sayYES_LOUD; /* Success! */
3396 PL_reginput = locinput;
3401 s = HOPBACKc(locinput, scan->flags);
3407 PL_reginput = locinput;
3412 s = HOPBACKc(locinput, scan->flags);
3418 PL_reginput = locinput;
3421 inner = NEXTOPER(NEXTOPER(scan));
3422 if (regmatch(inner) != n) {
3437 if (OP(scan) == SUSPEND) {
3438 locinput = PL_reginput;
3439 nextchr = UCHARAT(locinput);
3444 next = scan + ARG(scan);
3449 PerlIO_printf(Perl_error_log, "%"UVxf" %d\n",
3450 PTR2UV(scan), OP(scan));
3451 Perl_croak(aTHX_ "regexp memory corruption");
3458 * We get here only if there's trouble -- normally "case END" is
3459 * the terminating point.
3461 Perl_croak(aTHX_ "corrupted regexp pointers");
3467 PerlIO_printf(Perl_debug_log,
3468 "%*s %scould match...%s\n",
3469 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4],PL_colors[5])
3473 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n",
3474 PL_colors[4],PL_colors[5]));
3480 #if 0 /* Breaks $^R */
3488 PerlIO_printf(Perl_debug_log,
3489 "%*s %sfailed...%s\n",
3490 REPORT_CODE_OFF+PL_regindent*2, "",PL_colors[4],PL_colors[5])
3496 re_unwind_t *uw = SSPTRt(unwind,re_unwind_t);
3499 case RE_UNWIND_BRANCH:
3500 case RE_UNWIND_BRANCHJ:
3502 re_unwind_branch_t *uwb = &(uw->branch);
3503 I32 lastparen = uwb->lastparen;
3505 REGCP_UNWIND(uwb->lastcp);
3506 for (n = *PL_reglastparen; n > lastparen; n--)
3508 *PL_reglastparen = n;
3509 scan = next = uwb->next;
3511 OP(scan) != (uwb->type == RE_UNWIND_BRANCH
3512 ? BRANCH : BRANCHJ) ) { /* Failure */
3519 /* Have more choice yet. Reuse the same uwb. */
3521 if ((n = (uwb->type == RE_UNWIND_BRANCH
3522 ? NEXT_OFF(next) : ARG(next))))
3525 next = NULL; /* XXXX Needn't unwinding in this case... */
3527 next = NEXTOPER(scan);
3528 if (uwb->type == RE_UNWIND_BRANCHJ)
3529 next = NEXTOPER(next);
3530 locinput = uwb->locinput;
3531 nextchr = uwb->nextchr;
3533 PL_regindent = uwb->regindent;
3540 Perl_croak(aTHX_ "regexp unwind memory corruption");
3551 - regrepeat - repeatedly match something simple, report how many
3554 * [This routine now assumes that it will only match on things of length 1.
3555 * That was true before, but now we assume scan - reginput is the count,
3556 * rather than incrementing count on every character. [Er, except utf8.]]
3559 S_regrepeat(pTHX_ regnode *p, I32 max)
3561 register char *scan;
3563 register char *loceol = PL_regeol;
3564 register I32 hardcount = 0;
3565 register bool do_utf8 = DO_UTF8(PL_reg_sv);
3568 if (max != REG_INFTY && max < loceol - scan)
3569 loceol = scan + max;
3574 while (scan < loceol && hardcount < max && *scan != '\n') {
3575 scan += UTF8SKIP(scan);
3579 while (scan < loceol && *scan != '\n')
3586 case EXACT: /* length of string is 1 */
3588 while (scan < loceol && UCHARAT(scan) == c)
3591 case EXACTF: /* length of string is 1 */
3593 while (scan < loceol &&
3594 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold[c]))
3597 case EXACTFL: /* length of string is 1 */
3598 PL_reg_flags |= RF_tainted;
3600 while (scan < loceol &&
3601 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold_locale[c]))
3607 while (hardcount < max && scan < loceol &&
3608 reginclass(p, (U8*)scan, do_utf8)) {
3609 scan += UTF8SKIP(scan);
3613 while (scan < loceol && reginclass(p, (U8*)scan, do_utf8))
3620 LOAD_UTF8_CHARCLASS(alnum,"a");
3621 while (hardcount < max && scan < loceol &&
3622 swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
3623 scan += UTF8SKIP(scan);
3627 while (scan < loceol && isALNUM(*scan))
3632 PL_reg_flags |= RF_tainted;
3635 while (hardcount < max && scan < loceol &&
3636 isALNUM_LC_utf8((U8*)scan)) {
3637 scan += UTF8SKIP(scan);
3641 while (scan < loceol && isALNUM_LC(*scan))
3648 LOAD_UTF8_CHARCLASS(alnum,"a");
3649 while (hardcount < max && scan < loceol &&
3650 !swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
3651 scan += UTF8SKIP(scan);
3655 while (scan < loceol && !isALNUM(*scan))
3660 PL_reg_flags |= RF_tainted;
3663 while (hardcount < max && scan < loceol &&
3664 !isALNUM_LC_utf8((U8*)scan)) {
3665 scan += UTF8SKIP(scan);
3669 while (scan < loceol && !isALNUM_LC(*scan))
3676 LOAD_UTF8_CHARCLASS(space," ");
3677 while (hardcount < max && scan < loceol &&
3679 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
3680 scan += UTF8SKIP(scan);
3684 while (scan < loceol && isSPACE(*scan))
3689 PL_reg_flags |= RF_tainted;
3692 while (hardcount < max && scan < loceol &&
3693 (*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
3694 scan += UTF8SKIP(scan);
3698 while (scan < loceol && isSPACE_LC(*scan))
3705 LOAD_UTF8_CHARCLASS(space," ");
3706 while (hardcount < max && scan < loceol &&
3708 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
3709 scan += UTF8SKIP(scan);
3713 while (scan < loceol && !isSPACE(*scan))
3718 PL_reg_flags |= RF_tainted;
3721 while (hardcount < max && scan < loceol &&
3722 !(*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
3723 scan += UTF8SKIP(scan);
3727 while (scan < loceol && !isSPACE_LC(*scan))
3734 LOAD_UTF8_CHARCLASS(digit,"0");
3735 while (hardcount < max && scan < loceol &&
3736 swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
3737 scan += UTF8SKIP(scan);
3741 while (scan < loceol && isDIGIT(*scan))
3748 LOAD_UTF8_CHARCLASS(digit,"0");
3749 while (hardcount < max && scan < loceol &&
3750 !swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
3751 scan += UTF8SKIP(scan);
3755 while (scan < loceol && !isDIGIT(*scan))
3759 default: /* Called on something of 0 width. */
3760 break; /* So match right here or not at all. */
3766 c = scan - PL_reginput;
3771 SV *prop = sv_newmortal();
3774 PerlIO_printf(Perl_debug_log,
3775 "%*s %s can match %"IVdf" times out of %"IVdf"...\n",
3776 REPORT_CODE_OFF+1, "", SvPVX(prop),(IV)c,(IV)max);
3783 - regrepeat_hard - repeatedly match something, report total lenth and length
3785 * The repeater is supposed to have constant length.
3789 S_regrepeat_hard(pTHX_ regnode *p, I32 max, I32 *lp)
3791 register char *scan = Nullch;
3792 register char *start;
3793 register char *loceol = PL_regeol;
3795 I32 count = 0, res = 1;
3800 start = PL_reginput;
3801 if (DO_UTF8(PL_reg_sv)) {
3802 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
3805 while (start < PL_reginput) {
3807 start += UTF8SKIP(start);
3818 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
3820 *lp = l = PL_reginput - start;
3821 if (max != REG_INFTY && l*max < loceol - scan)
3822 loceol = scan + l*max;
3835 - regclass_swash - prepare the utf8 swash
3839 Perl_regclass_swash(pTHX_ register regnode* node, bool doinit, SV** initsvp)
3844 if (PL_regdata && PL_regdata->count) {
3847 if (PL_regdata->what[n] == 's') {
3848 SV *rv = (SV*)PL_regdata->data[n];
3849 AV *av = (AV*)SvRV((SV*)rv);
3852 si = *av_fetch(av, 0, FALSE);
3853 a = av_fetch(av, 1, FALSE);
3857 else if (si && doinit) {
3858 sw = swash_init("utf8", "", si, 1, 0);
3859 (void)av_store(av, 1, sw);
3871 - reginclass - determine if a character falls into a character class
3875 S_reginclass(pTHX_ register regnode *n, register U8* p, register bool do_utf8)
3877 char flags = ANYOF_FLAGS(n);
3882 c = do_utf8 ? utf8_to_uvchr(p, &len) : *p;
3884 if (do_utf8 || (flags & ANYOF_UNICODE)) {
3885 if (do_utf8 && !ANYOF_RUNTIME(n)) {
3886 if (len != (STRLEN)-1 && c < 256 && ANYOF_BITMAP_TEST(n, c))
3889 if (!match && do_utf8 && (flags & ANYOF_UNICODE_ALL) && c >= 256)
3892 SV *sw = regclass_swash(n, TRUE, 0);
3895 if (swash_fetch(sw, p, do_utf8))
3897 else if (flags & ANYOF_FOLD) {
3898 U8 tmpbuf[UTF8_MAXLEN+1];
3900 if (flags & ANYOF_LOCALE) {
3901 PL_reg_flags |= RF_tainted;
3902 uvchr_to_utf8(tmpbuf, toLOWER_LC_utf8(p));
3905 uvchr_to_utf8(tmpbuf, toLOWER_utf8(p));
3906 if (swash_fetch(sw, tmpbuf, do_utf8))
3912 if (!match && c < 256) {
3913 if (ANYOF_BITMAP_TEST(n, c))
3915 else if (flags & ANYOF_FOLD) {
3918 if (flags & ANYOF_LOCALE) {
3919 PL_reg_flags |= RF_tainted;
3920 f = PL_fold_locale[c];
3924 if (f != c && ANYOF_BITMAP_TEST(n, f))
3928 if (!match && (flags & ANYOF_CLASS)) {
3929 PL_reg_flags |= RF_tainted;
3931 (ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) ||
3932 (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) ||
3933 (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) ||
3934 (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) ||
3935 (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) ||
3936 (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) ||
3937 (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) ||
3938 (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) ||
3939 (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) ||
3940 (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) ||
3941 (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII(c)) ||
3942 (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII(c)) ||
3943 (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) ||
3944 (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) ||
3945 (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) ||
3946 (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) ||
3947 (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) ||
3948 (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) ||
3949 (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) ||
3950 (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) ||
3951 (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) ||
3952 (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) ||
3953 (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) ||
3954 (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) ||
3955 (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) ||
3956 (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) ||
3957 (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) ||
3958 (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) ||
3959 (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK(c)) ||
3960 (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK(c))
3961 ) /* How's that for a conditional? */
3968 return (flags & ANYOF_INVERT) ? !match : match;
3972 S_reghop(pTHX_ U8 *s, I32 off)
3974 return S_reghop3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
3978 S_reghop3(pTHX_ U8 *s, I32 off, U8* lim)
3981 while (off-- && s < lim) {
3982 /* XXX could check well-formedness here */
3990 if (UTF8_IS_CONTINUED(*s)) {
3991 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
3994 /* XXX could check well-formedness here */
4002 S_reghopmaybe(pTHX_ U8 *s, I32 off)
4004 return S_reghopmaybe3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4008 S_reghopmaybe3(pTHX_ U8* s, I32 off, U8* lim)
4011 while (off-- && s < lim) {
4012 /* XXX could check well-formedness here */
4022 if (UTF8_IS_CONTINUED(*s)) {
4023 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4026 /* XXX could check well-formedness here */
4042 restore_pos(pTHXo_ void *arg)
4044 if (PL_reg_eval_set) {
4045 if (PL_reg_oldsaved) {
4046 PL_reg_re->subbeg = PL_reg_oldsaved;
4047 PL_reg_re->sublen = PL_reg_oldsavedlen;
4048 RX_MATCH_COPIED_on(PL_reg_re);
4050 PL_reg_magic->mg_len = PL_reg_oldpos;
4051 PL_reg_eval_set = 0;
4052 PL_curpm = PL_reg_oldcurpm;