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;
1961 - regmatch - main matching routine
1963 * Conceptually the strategy is simple: check to see whether the current
1964 * node matches, call self recursively to see whether the rest matches,
1965 * and then act accordingly. In practice we make some effort to avoid
1966 * recursion, in particular by going through "ordinary" nodes (that don't
1967 * need to know whether the rest of the match failed) by a loop instead of
1970 /* [lwall] I've hoisted the register declarations to the outer block in order to
1971 * maybe save a little bit of pushing and popping on the stack. It also takes
1972 * advantage of machines that use a register save mask on subroutine entry.
1974 STATIC I32 /* 0 failure, 1 success */
1975 S_regmatch(pTHX_ regnode *prog)
1977 register regnode *scan; /* Current node. */
1978 regnode *next; /* Next node. */
1979 regnode *inner; /* Next node in internal branch. */
1980 register I32 nextchr; /* renamed nextchr - nextchar colides with
1981 function of same name */
1982 register I32 n; /* no or next */
1983 register I32 ln = 0; /* len or last */
1984 register char *s = Nullch; /* operand or save */
1985 register char *locinput = PL_reginput;
1986 register I32 c1 = 0, c2 = 0, paren; /* case fold search, parenth */
1987 int minmod = 0, sw = 0, logical = 0;
1990 I32 firstcp = PL_savestack_ix;
1992 register bool do_utf8 = DO_UTF8(PL_reg_sv);
1998 /* Note that nextchr is a byte even in UTF */
1999 nextchr = UCHARAT(locinput);
2001 while (scan != NULL) {
2002 #define sayNO_L (logical ? (logical = 0, sw = 0, goto cont) : sayNO)
2004 # define sayYES goto yes
2005 # define sayNO goto no
2006 # define sayYES_FINAL goto yes_final
2007 # define sayYES_LOUD goto yes_loud
2008 # define sayNO_FINAL goto no_final
2009 # define sayNO_SILENT goto do_no
2010 # define saySAME(x) if (x) goto yes; else goto no
2011 # define REPORT_CODE_OFF 24
2013 # define sayYES return 1
2014 # define sayNO return 0
2015 # define sayYES_FINAL return 1
2016 # define sayYES_LOUD return 1
2017 # define sayNO_FINAL return 0
2018 # define sayNO_SILENT return 0
2019 # define saySAME(x) return x
2022 SV *prop = sv_newmortal();
2023 int docolor = *PL_colors[0];
2024 int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */
2025 int l = (PL_regeol - locinput) > taill ? taill : (PL_regeol - locinput);
2026 /* The part of the string before starttry has one color
2027 (pref0_len chars), between starttry and current
2028 position another one (pref_len - pref0_len chars),
2029 after the current position the third one.
2030 We assume that pref0_len <= pref_len, otherwise we
2031 decrease pref0_len. */
2032 int pref_len = (locinput - PL_bostr) > (5 + taill) - l
2033 ? (5 + taill) - l : locinput - PL_bostr;
2036 while (UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len)))
2038 pref0_len = pref_len - (locinput - PL_reg_starttry);
2039 if (l + pref_len < (5 + taill) && l < PL_regeol - locinput)
2040 l = ( PL_regeol - locinput > (5 + taill) - pref_len
2041 ? (5 + taill) - pref_len : PL_regeol - locinput);
2042 while (UTF8_IS_CONTINUATION(*(U8*)(locinput + l)))
2046 if (pref0_len > pref_len)
2047 pref0_len = pref_len;
2048 regprop(prop, scan);
2049 PerlIO_printf(Perl_debug_log,
2050 "%4"IVdf" <%s%.*s%s%s%.*s%s%s%s%.*s%s>%*s|%3"IVdf":%*s%s\n",
2051 (IV)(locinput - PL_bostr),
2052 PL_colors[4], pref0_len,
2053 locinput - pref_len, PL_colors[5],
2054 PL_colors[2], pref_len - pref0_len,
2055 locinput - pref_len + pref0_len, PL_colors[3],
2056 (docolor ? "" : "> <"),
2057 PL_colors[0], l, locinput, PL_colors[1],
2058 15 - l - pref_len + 1,
2060 (IV)(scan - PL_regprogram), PL_regindent*2, "",
2064 next = scan + NEXT_OFF(scan);
2070 if (locinput == PL_bostr || (PL_multiline &&
2071 (nextchr || locinput < PL_regeol) && locinput[-1] == '\n') )
2073 /* regtill = regbol; */
2078 if (locinput == PL_bostr ||
2079 ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n'))
2085 if (locinput == PL_bostr)
2089 if (locinput == PL_reg_ganch)
2099 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2104 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2106 if (PL_regeol - locinput > 1)
2110 if (PL_regeol != locinput)
2114 if (!nextchr && locinput >= PL_regeol)
2116 nextchr = UCHARAT(++locinput);
2119 if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n')
2122 locinput += PL_utf8skip[nextchr];
2123 if (locinput > PL_regeol)
2125 nextchr = UCHARAT(locinput);
2128 nextchr = UCHARAT(++locinput);
2133 if (do_utf8 != (UTF!=0)) {
2141 if (*((U8*)s) != utf8_to_uvchr((U8*)l, &len))
2150 if (*((U8*)l) != utf8_to_uvchr((U8*)s, &len))
2156 nextchr = UCHARAT(locinput);
2159 /* Inline the first character, for speed. */
2160 if (UCHARAT(s) != nextchr)
2162 if (PL_regeol - locinput < ln)
2164 if (ln > 1 && memNE(s, locinput, ln))
2167 nextchr = UCHARAT(locinput);
2170 PL_reg_flags |= RF_tainted;
2180 c1 = OP(scan) == EXACTF;
2182 if (l >= PL_regeol) {
2185 if ((UTF ? utf8n_to_uvchr((U8*)s, e - s, 0, 0) : *((U8*)s)) !=
2186 (c1 ? toLOWER_utf8((U8*)l) : toLOWER_LC_utf8((U8*)l)))
2188 s += UTF ? UTF8SKIP(s) : 1;
2192 nextchr = UCHARAT(locinput);
2196 /* Inline the first character, for speed. */
2197 if (UCHARAT(s) != nextchr &&
2198 UCHARAT(s) != ((OP(scan) == EXACTF)
2199 ? PL_fold : PL_fold_locale)[nextchr])
2201 if (PL_regeol - locinput < ln)
2203 if (ln > 1 && (OP(scan) == EXACTF
2204 ? ibcmp(s, locinput, ln)
2205 : ibcmp_locale(s, locinput, ln)))
2208 nextchr = UCHARAT(locinput);
2212 if (!reginclass(scan, (U8*)locinput, do_utf8))
2214 if (locinput >= PL_regeol)
2216 locinput += PL_utf8skip[nextchr];
2217 nextchr = UCHARAT(locinput);
2221 nextchr = UCHARAT(locinput);
2222 if (!reginclass(scan, (U8*)locinput, do_utf8))
2224 if (!nextchr && locinput >= PL_regeol)
2226 nextchr = UCHARAT(++locinput);
2230 PL_reg_flags |= RF_tainted;
2236 LOAD_UTF8_CHARCLASS(alnum,"a");
2237 if (!(OP(scan) == ALNUM
2238 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2239 : isALNUM_LC_utf8((U8*)locinput)))
2243 locinput += PL_utf8skip[nextchr];
2244 nextchr = UCHARAT(locinput);
2247 if (!(OP(scan) == ALNUM
2248 ? isALNUM(nextchr) : isALNUM_LC(nextchr)))
2250 nextchr = UCHARAT(++locinput);
2253 PL_reg_flags |= RF_tainted;
2256 if (!nextchr && locinput >= PL_regeol)
2259 LOAD_UTF8_CHARCLASS(alnum,"a");
2260 if (OP(scan) == NALNUM
2261 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2262 : isALNUM_LC_utf8((U8*)locinput))
2266 locinput += PL_utf8skip[nextchr];
2267 nextchr = UCHARAT(locinput);
2270 if (OP(scan) == NALNUM
2271 ? isALNUM(nextchr) : isALNUM_LC(nextchr))
2273 nextchr = UCHARAT(++locinput);
2277 PL_reg_flags |= RF_tainted;
2281 /* was last char in word? */
2283 if (locinput == PL_bostr)
2286 U8 *r = reghop((U8*)locinput, -1);
2288 ln = utf8n_to_uvchr(r, s - (char*)r, 0, 0);
2290 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2291 ln = isALNUM_uni(ln);
2292 LOAD_UTF8_CHARCLASS(alnum,"a");
2293 n = swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8);
2296 ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(ln));
2297 n = isALNUM_LC_utf8((U8*)locinput);
2301 ln = (locinput != PL_bostr) ?
2302 UCHARAT(locinput - 1) : '\n';
2303 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2305 n = isALNUM(nextchr);
2308 ln = isALNUM_LC(ln);
2309 n = isALNUM_LC(nextchr);
2312 if (((!ln) == (!n)) == (OP(scan) == BOUND ||
2313 OP(scan) == BOUNDL))
2317 PL_reg_flags |= RF_tainted;
2323 if (UTF8_IS_CONTINUED(nextchr)) {
2324 LOAD_UTF8_CHARCLASS(space," ");
2325 if (!(OP(scan) == SPACE
2326 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2327 : isSPACE_LC_utf8((U8*)locinput)))
2331 locinput += PL_utf8skip[nextchr];
2332 nextchr = UCHARAT(locinput);
2335 if (!(OP(scan) == SPACE
2336 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2338 nextchr = UCHARAT(++locinput);
2341 if (!(OP(scan) == SPACE
2342 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2344 nextchr = UCHARAT(++locinput);
2348 PL_reg_flags |= RF_tainted;
2351 if (!nextchr && locinput >= PL_regeol)
2354 LOAD_UTF8_CHARCLASS(space," ");
2355 if (OP(scan) == NSPACE
2356 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2357 : isSPACE_LC_utf8((U8*)locinput))
2361 locinput += PL_utf8skip[nextchr];
2362 nextchr = UCHARAT(locinput);
2365 if (OP(scan) == NSPACE
2366 ? isSPACE(nextchr) : isSPACE_LC(nextchr))
2368 nextchr = UCHARAT(++locinput);
2371 PL_reg_flags |= RF_tainted;
2377 LOAD_UTF8_CHARCLASS(digit,"0");
2378 if (!(OP(scan) == DIGIT
2379 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2380 : isDIGIT_LC_utf8((U8*)locinput)))
2384 locinput += PL_utf8skip[nextchr];
2385 nextchr = UCHARAT(locinput);
2388 if (!(OP(scan) == DIGIT
2389 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr)))
2391 nextchr = UCHARAT(++locinput);
2394 PL_reg_flags |= RF_tainted;
2397 if (!nextchr && locinput >= PL_regeol)
2400 LOAD_UTF8_CHARCLASS(digit,"0");
2401 if (OP(scan) == NDIGIT
2402 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2403 : isDIGIT_LC_utf8((U8*)locinput))
2407 locinput += PL_utf8skip[nextchr];
2408 nextchr = UCHARAT(locinput);
2411 if (OP(scan) == NDIGIT
2412 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr))
2414 nextchr = UCHARAT(++locinput);
2417 LOAD_UTF8_CHARCLASS(mark,"~");
2418 if (locinput >= PL_regeol ||
2419 swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2421 locinput += PL_utf8skip[nextchr];
2422 while (locinput < PL_regeol &&
2423 swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2424 locinput += UTF8SKIP(locinput);
2425 if (locinput > PL_regeol)
2427 nextchr = UCHARAT(locinput);
2430 PL_reg_flags |= RF_tainted;
2434 n = ARG(scan); /* which paren pair */
2435 ln = PL_regstartp[n];
2436 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2437 if (*PL_reglastparen < n || ln == -1)
2438 sayNO; /* Do not match unless seen CLOSEn. */
2439 if (ln == PL_regendp[n])
2443 if (do_utf8 && OP(scan) != REF) { /* REF can do byte comparison */
2445 char *e = PL_bostr + PL_regendp[n];
2447 * Note that we can't do the "other character" lookup trick as
2448 * in the 8-bit case (no pun intended) because in Unicode we
2449 * have to map both upper and title case to lower case.
2451 if (OP(scan) == REFF) {
2455 if (toLOWER_utf8((U8*)s) != toLOWER_utf8((U8*)l))
2465 if (toLOWER_LC_utf8((U8*)s) != toLOWER_LC_utf8((U8*)l))
2472 nextchr = UCHARAT(locinput);
2476 /* Inline the first character, for speed. */
2477 if (UCHARAT(s) != nextchr &&
2479 (UCHARAT(s) != ((OP(scan) == REFF
2480 ? PL_fold : PL_fold_locale)[nextchr]))))
2482 ln = PL_regendp[n] - ln;
2483 if (locinput + ln > PL_regeol)
2485 if (ln > 1 && (OP(scan) == REF
2486 ? memNE(s, locinput, ln)
2488 ? ibcmp(s, locinput, ln)
2489 : ibcmp_locale(s, locinput, ln))))
2492 nextchr = UCHARAT(locinput);
2503 OP_4tree *oop = PL_op;
2504 COP *ocurcop = PL_curcop;
2505 SV **ocurpad = PL_curpad;
2509 PL_op = (OP_4tree*)PL_regdata->data[n];
2510 DEBUG_r( PerlIO_printf(Perl_debug_log, " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) );
2511 PL_curpad = AvARRAY((AV*)PL_regdata->data[n + 2]);
2512 PL_regendp[0] = PL_reg_magic->mg_len = locinput - PL_bostr;
2514 CALLRUNOPS(aTHX); /* Scalar context. */
2520 PL_curpad = ocurpad;
2521 PL_curcop = ocurcop;
2523 if (logical == 2) { /* Postponed subexpression. */
2525 MAGIC *mg = Null(MAGIC*);
2527 CHECKPOINT cp, lastcp;
2529 if(SvROK(ret) || SvRMAGICAL(ret)) {
2530 SV *sv = SvROK(ret) ? SvRV(ret) : ret;
2533 mg = mg_find(sv, PERL_MAGIC_qr);
2536 re = (regexp *)mg->mg_obj;
2537 (void)ReREFCNT_inc(re);
2541 char *t = SvPV(ret, len);
2543 char *oprecomp = PL_regprecomp;
2544 I32 osize = PL_regsize;
2545 I32 onpar = PL_regnpar;
2548 re = CALLREGCOMP(aTHX_ t, t + len, &pm);
2550 & (SVs_TEMP | SVs_PADTMP | SVf_READONLY)))
2551 sv_magic(ret,(SV*)ReREFCNT_inc(re),
2553 PL_regprecomp = oprecomp;
2558 PerlIO_printf(Perl_debug_log,
2559 "Entering embedded `%s%.60s%s%s'\n",
2563 (strlen(re->precomp) > 60 ? "..." : ""))
2566 state.prev = PL_reg_call_cc;
2567 state.cc = PL_regcc;
2568 state.re = PL_reg_re;
2572 cp = regcppush(0); /* Save *all* the positions. */
2575 state.ss = PL_savestack_ix;
2576 *PL_reglastparen = 0;
2577 *PL_reglastcloseparen = 0;
2578 PL_reg_call_cc = &state;
2579 PL_reginput = locinput;
2581 /* XXXX This is too dramatic a measure... */
2584 if (regmatch(re->program + 1)) {
2585 /* Even though we succeeded, we need to restore
2586 global variables, since we may be wrapped inside
2587 SUSPEND, thus the match may be not finished yet. */
2589 /* XXXX Do this only if SUSPENDed? */
2590 PL_reg_call_cc = state.prev;
2591 PL_regcc = state.cc;
2592 PL_reg_re = state.re;
2593 cache_re(PL_reg_re);
2595 /* XXXX This is too dramatic a measure... */
2598 /* These are needed even if not SUSPEND. */
2604 REGCP_UNWIND(lastcp);
2606 PL_reg_call_cc = state.prev;
2607 PL_regcc = state.cc;
2608 PL_reg_re = state.re;
2609 cache_re(PL_reg_re);
2611 /* XXXX This is too dramatic a measure... */
2620 sv_setsv(save_scalar(PL_replgv), ret);
2624 n = ARG(scan); /* which paren pair */
2625 PL_reg_start_tmp[n] = locinput;
2630 n = ARG(scan); /* which paren pair */
2631 PL_regstartp[n] = PL_reg_start_tmp[n] - PL_bostr;
2632 PL_regendp[n] = locinput - PL_bostr;
2633 if (n > *PL_reglastparen)
2634 *PL_reglastparen = n;
2635 *PL_reglastcloseparen = n;
2638 n = ARG(scan); /* which paren pair */
2639 sw = (*PL_reglastparen >= n && PL_regendp[n] != -1);
2642 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2644 next = NEXTOPER(NEXTOPER(scan));
2646 next = scan + ARG(scan);
2647 if (OP(next) == IFTHEN) /* Fake one. */
2648 next = NEXTOPER(NEXTOPER(next));
2652 logical = scan->flags;
2654 /*******************************************************************
2655 PL_regcc contains infoblock about the innermost (...)* loop, and
2656 a pointer to the next outer infoblock.
2658 Here is how Y(A)*Z is processed (if it is compiled into CURLYX/WHILEM):
2660 1) After matching X, regnode for CURLYX is processed;
2662 2) This regnode creates infoblock on the stack, and calls
2663 regmatch() recursively with the starting point at WHILEM node;
2665 3) Each hit of WHILEM node tries to match A and Z (in the order
2666 depending on the current iteration, min/max of {min,max} and
2667 greediness). The information about where are nodes for "A"
2668 and "Z" is read from the infoblock, as is info on how many times "A"
2669 was already matched, and greediness.
2671 4) After A matches, the same WHILEM node is hit again.
2673 5) Each time WHILEM is hit, PL_regcc is the infoblock created by CURLYX
2674 of the same pair. Thus when WHILEM tries to match Z, it temporarily
2675 resets PL_regcc, since this Y(A)*Z can be a part of some other loop:
2676 as in (Y(A)*Z)*. If Z matches, the automaton will hit the WHILEM node
2677 of the external loop.
2679 Currently present infoblocks form a tree with a stem formed by PL_curcc
2680 and whatever it mentions via ->next, and additional attached trees
2681 corresponding to temporarily unset infoblocks as in "5" above.
2683 In the following picture infoblocks for outer loop of
2684 (Y(A)*?Z)*?T are denoted O, for inner I. NULL starting block
2685 is denoted by x. The matched string is YAAZYAZT. Temporarily postponed
2686 infoblocks are drawn below the "reset" infoblock.
2688 In fact in the picture below we do not show failed matches for Z and T
2689 by WHILEM blocks. [We illustrate minimal matches, since for them it is
2690 more obvious *why* one needs to *temporary* unset infoblocks.]
2692 Matched REx position InfoBlocks Comment
2696 Y A)*?Z)*?T x <- O <- I
2697 YA )*?Z)*?T x <- O <- I
2698 YA A)*?Z)*?T x <- O <- I
2699 YAA )*?Z)*?T x <- O <- I
2700 YAA Z)*?T x <- O # Temporary unset I
2703 YAAZ Y(A)*?Z)*?T x <- O
2706 YAAZY (A)*?Z)*?T x <- O
2709 YAAZY A)*?Z)*?T x <- O <- I
2712 YAAZYA )*?Z)*?T x <- O <- I
2715 YAAZYA Z)*?T x <- O # Temporary unset I
2721 YAAZYAZ T x # Temporary unset O
2728 *******************************************************************/
2731 CHECKPOINT cp = PL_savestack_ix;
2732 /* No need to save/restore up to this paren */
2733 I32 parenfloor = scan->flags;
2735 if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */
2737 cc.oldcc = PL_regcc;
2739 /* XXXX Probably it is better to teach regpush to support
2740 parenfloor > PL_regsize... */
2741 if (parenfloor > *PL_reglastparen)
2742 parenfloor = *PL_reglastparen; /* Pessimization... */
2743 cc.parenfloor = parenfloor;
2745 cc.min = ARG1(scan);
2746 cc.max = ARG2(scan);
2747 cc.scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
2751 PL_reginput = locinput;
2752 n = regmatch(PREVOPER(next)); /* start on the WHILEM */
2754 PL_regcc = cc.oldcc;
2760 * This is really hard to understand, because after we match
2761 * what we're trying to match, we must make sure the rest of
2762 * the REx is going to match for sure, and to do that we have
2763 * to go back UP the parse tree by recursing ever deeper. And
2764 * if it fails, we have to reset our parent's current state
2765 * that we can try again after backing off.
2768 CHECKPOINT cp, lastcp;
2769 CURCUR* cc = PL_regcc;
2770 char *lastloc = cc->lastloc; /* Detection of 0-len. */
2772 n = cc->cur + 1; /* how many we know we matched */
2773 PL_reginput = locinput;
2776 PerlIO_printf(Perl_debug_log,
2777 "%*s %ld out of %ld..%ld cc=%lx\n",
2778 REPORT_CODE_OFF+PL_regindent*2, "",
2779 (long)n, (long)cc->min,
2780 (long)cc->max, (long)cc)
2783 /* If degenerate scan matches "", assume scan done. */
2785 if (locinput == cc->lastloc && n >= cc->min) {
2786 PL_regcc = cc->oldcc;
2790 PerlIO_printf(Perl_debug_log,
2791 "%*s empty match detected, try continuation...\n",
2792 REPORT_CODE_OFF+PL_regindent*2, "")
2794 if (regmatch(cc->next))
2802 /* First just match a string of min scans. */
2806 cc->lastloc = locinput;
2807 if (regmatch(cc->scan))
2810 cc->lastloc = lastloc;
2815 /* Check whether we already were at this position.
2816 Postpone detection until we know the match is not
2817 *that* much linear. */
2818 if (!PL_reg_maxiter) {
2819 PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4);
2820 PL_reg_leftiter = PL_reg_maxiter;
2822 if (PL_reg_leftiter-- == 0) {
2823 I32 size = (PL_reg_maxiter + 7)/8;
2824 if (PL_reg_poscache) {
2825 if (PL_reg_poscache_size < size) {
2826 Renew(PL_reg_poscache, size, char);
2827 PL_reg_poscache_size = size;
2829 Zero(PL_reg_poscache, size, char);
2832 PL_reg_poscache_size = size;
2833 Newz(29, PL_reg_poscache, size, char);
2836 PerlIO_printf(Perl_debug_log,
2837 "%sDetected a super-linear match, switching on caching%s...\n",
2838 PL_colors[4], PL_colors[5])
2841 if (PL_reg_leftiter < 0) {
2842 I32 o = locinput - PL_bostr, b;
2844 o = (scan->flags & 0xf) - 1 + o * (scan->flags>>4);
2847 if (PL_reg_poscache[o] & (1<<b)) {
2849 PerlIO_printf(Perl_debug_log,
2850 "%*s already tried at this position...\n",
2851 REPORT_CODE_OFF+PL_regindent*2, "")
2855 PL_reg_poscache[o] |= (1<<b);
2859 /* Prefer next over scan for minimal matching. */
2862 PL_regcc = cc->oldcc;
2865 cp = regcppush(cc->parenfloor);
2867 if (regmatch(cc->next)) {
2869 sayYES; /* All done. */
2871 REGCP_UNWIND(lastcp);
2877 if (n >= cc->max) { /* Maximum greed exceeded? */
2878 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
2879 && !(PL_reg_flags & RF_warned)) {
2880 PL_reg_flags |= RF_warned;
2881 Perl_warner(aTHX_ WARN_REGEXP, "%s limit (%d) exceeded",
2882 "Complex regular subexpression recursion",
2889 PerlIO_printf(Perl_debug_log,
2890 "%*s trying longer...\n",
2891 REPORT_CODE_OFF+PL_regindent*2, "")
2893 /* Try scanning more and see if it helps. */
2894 PL_reginput = locinput;
2896 cc->lastloc = locinput;
2897 cp = regcppush(cc->parenfloor);
2899 if (regmatch(cc->scan)) {
2903 REGCP_UNWIND(lastcp);
2906 cc->lastloc = lastloc;
2910 /* Prefer scan over next for maximal matching. */
2912 if (n < cc->max) { /* More greed allowed? */
2913 cp = regcppush(cc->parenfloor);
2915 cc->lastloc = locinput;
2917 if (regmatch(cc->scan)) {
2921 REGCP_UNWIND(lastcp);
2922 regcppop(); /* Restore some previous $<digit>s? */
2923 PL_reginput = locinput;
2925 PerlIO_printf(Perl_debug_log,
2926 "%*s failed, try continuation...\n",
2927 REPORT_CODE_OFF+PL_regindent*2, "")
2930 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
2931 && !(PL_reg_flags & RF_warned)) {
2932 PL_reg_flags |= RF_warned;
2933 Perl_warner(aTHX_ WARN_REGEXP, "%s limit (%d) exceeded",
2934 "Complex regular subexpression recursion",
2938 /* Failed deeper matches of scan, so see if this one works. */
2939 PL_regcc = cc->oldcc;
2942 if (regmatch(cc->next))
2948 cc->lastloc = lastloc;
2953 next = scan + ARG(scan);
2956 inner = NEXTOPER(NEXTOPER(scan));
2959 inner = NEXTOPER(scan);
2963 if (OP(next) != c1) /* No choice. */
2964 next = inner; /* Avoid recursion. */
2966 I32 lastparen = *PL_reglastparen;
2968 re_unwind_branch_t *uw;
2970 /* Put unwinding data on stack */
2971 unwind1 = SSNEWt(1,re_unwind_branch_t);
2972 uw = SSPTRt(unwind1,re_unwind_branch_t);
2975 uw->type = ((c1 == BRANCH)
2977 : RE_UNWIND_BRANCHJ);
2978 uw->lastparen = lastparen;
2980 uw->locinput = locinput;
2981 uw->nextchr = nextchr;
2983 uw->regindent = ++PL_regindent;
2986 REGCP_SET(uw->lastcp);
2988 /* Now go into the first branch */
3001 /* We suppose that the next guy does not need
3002 backtracking: in particular, it is of constant length,
3003 and has no parenths to influence future backrefs. */
3004 ln = ARG1(scan); /* min to match */
3005 n = ARG2(scan); /* max to match */
3006 paren = scan->flags;
3008 if (paren > PL_regsize)
3010 if (paren > *PL_reglastparen)
3011 *PL_reglastparen = paren;
3013 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3015 scan += NEXT_OFF(scan); /* Skip former OPEN. */
3016 PL_reginput = locinput;
3019 if (ln && regrepeat_hard(scan, ln, &l) < ln)
3021 if (ln && l == 0 && n >= ln
3022 /* In fact, this is tricky. If paren, then the
3023 fact that we did/didnot match may influence
3024 future execution. */
3025 && !(paren && ln == 0))
3027 locinput = PL_reginput;
3028 if (PL_regkind[(U8)OP(next)] == EXACT) {
3029 c1 = (U8)*STRING(next);
3030 if (OP(next) == EXACTF)
3032 else if (OP(next) == EXACTFL)
3033 c2 = PL_fold_locale[c1];
3040 /* This may be improved if l == 0. */
3041 while (n >= ln || (n == REG_INFTY && ln > 0 && l)) { /* ln overflow ? */
3042 /* If it could work, try it. */
3044 UCHARAT(PL_reginput) == c1 ||
3045 UCHARAT(PL_reginput) == c2)
3049 PL_regstartp[paren] =
3050 HOPc(PL_reginput, -l) - PL_bostr;
3051 PL_regendp[paren] = PL_reginput - PL_bostr;
3054 PL_regendp[paren] = -1;
3058 REGCP_UNWIND(lastcp);
3060 /* Couldn't or didn't -- move forward. */
3061 PL_reginput = locinput;
3062 if (regrepeat_hard(scan, 1, &l)) {
3064 locinput = PL_reginput;
3071 n = regrepeat_hard(scan, n, &l);
3072 if (n != 0 && l == 0
3073 /* In fact, this is tricky. If paren, then the
3074 fact that we did/didnot match may influence
3075 future execution. */
3076 && !(paren && ln == 0))
3078 locinput = PL_reginput;
3080 PerlIO_printf(Perl_debug_log,
3081 "%*s matched %"IVdf" times, len=%"IVdf"...\n",
3082 (int)(REPORT_CODE_OFF+PL_regindent*2), "",
3086 if (PL_regkind[(U8)OP(next)] == EXACT) {
3087 c1 = (U8)*STRING(next);
3088 if (OP(next) == EXACTF)
3090 else if (OP(next) == EXACTFL)
3091 c2 = PL_fold_locale[c1];
3100 /* If it could work, try it. */
3102 UCHARAT(PL_reginput) == c1 ||
3103 UCHARAT(PL_reginput) == c2)
3106 PerlIO_printf(Perl_debug_log,
3107 "%*s trying tail with n=%"IVdf"...\n",
3108 (int)(REPORT_CODE_OFF+PL_regindent*2), "", (IV)n)
3112 PL_regstartp[paren] = HOPc(PL_reginput, -l) - PL_bostr;
3113 PL_regendp[paren] = PL_reginput - PL_bostr;
3116 PL_regendp[paren] = -1;
3120 REGCP_UNWIND(lastcp);
3122 /* Couldn't or didn't -- back up. */
3124 locinput = HOPc(locinput, -l);
3125 PL_reginput = locinput;
3132 paren = scan->flags; /* Which paren to set */
3133 if (paren > PL_regsize)
3135 if (paren > *PL_reglastparen)
3136 *PL_reglastparen = paren;
3137 ln = ARG1(scan); /* min to match */
3138 n = ARG2(scan); /* max to match */
3139 scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE);
3143 ln = ARG1(scan); /* min to match */
3144 n = ARG2(scan); /* max to match */
3145 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3150 scan = NEXTOPER(scan);
3156 scan = NEXTOPER(scan);
3160 * Lookahead to avoid useless match attempts
3161 * when we know what character comes next.
3163 if (PL_regkind[(U8)OP(next)] == EXACT) {
3164 U8 *s = (U8*)STRING(next);
3167 if (OP(next) == EXACTF)
3169 else if (OP(next) == EXACTFL)
3170 c2 = PL_fold_locale[c1];
3173 if (OP(next) == EXACTF) {
3174 c1 = to_utf8_lower(s);
3175 c2 = to_utf8_upper(s);
3178 c2 = c1 = utf8_to_uvchr(s, NULL);
3184 PL_reginput = locinput;
3188 if (ln && regrepeat(scan, ln) < ln)
3190 locinput = PL_reginput;
3193 char *e; /* Should not check after this */
3194 char *old = locinput;
3196 if (n == REG_INFTY) {
3199 while (UTF8_IS_CONTINUATION(*(U8*)e))
3205 m >0 && e + UTF8SKIP(e) <= PL_regeol; m--)
3209 e = locinput + n - ln;
3215 /* Find place 'next' could work */
3218 while (locinput <= e && *locinput != c1)
3221 while (locinput <= e
3226 count = locinput - old;
3233 utf8_to_uvchr((U8*)locinput, &len) != c1;
3238 for (count = 0; locinput <= e; count++) {
3239 UV c = utf8_to_uvchr((U8*)locinput, &len);
3240 if (c == c1 || c == c2)
3248 /* PL_reginput == old now */
3249 if (locinput != old) {
3250 ln = 1; /* Did some */
3251 if (regrepeat(scan, count) < count)
3254 /* PL_reginput == locinput now */
3255 TRYPAREN(paren, ln, locinput);
3256 PL_reginput = locinput; /* Could be reset... */
3257 REGCP_UNWIND(lastcp);
3258 /* Couldn't or didn't -- move forward. */
3261 locinput += UTF8SKIP(locinput);
3267 while (n >= ln || (n == REG_INFTY && ln > 0)) { /* ln overflow ? */
3271 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3273 c = UCHARAT(PL_reginput);
3274 /* If it could work, try it. */
3275 if (c == c1 || c == c2)
3277 TRYPAREN(paren, n, PL_reginput);
3278 REGCP_UNWIND(lastcp);
3281 /* If it could work, try it. */
3282 else if (c1 == -1000)
3284 TRYPAREN(paren, n, PL_reginput);
3285 REGCP_UNWIND(lastcp);
3287 /* Couldn't or didn't -- move forward. */
3288 PL_reginput = locinput;
3289 if (regrepeat(scan, 1)) {
3291 locinput = PL_reginput;
3299 n = regrepeat(scan, n);
3300 locinput = PL_reginput;
3301 if (ln < n && PL_regkind[(U8)OP(next)] == EOL &&
3302 (!PL_multiline || OP(next) == SEOL || OP(next) == EOS)) {
3303 ln = n; /* why back off? */
3304 /* ...because $ and \Z can match before *and* after
3305 newline at the end. Consider "\n\n" =~ /\n+\Z\n/.
3306 We should back off by one in this case. */
3307 if (UCHARAT(PL_reginput - 1) == '\n' && OP(next) != EOS)
3316 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3318 c = UCHARAT(PL_reginput);
3320 /* If it could work, try it. */
3321 if (c1 == -1000 || c == c1 || c == c2)
3323 TRYPAREN(paren, n, PL_reginput);
3324 REGCP_UNWIND(lastcp);
3326 /* Couldn't or didn't -- back up. */
3328 PL_reginput = locinput = HOPc(locinput, -1);
3336 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3338 c = UCHARAT(PL_reginput);
3340 /* If it could work, try it. */
3341 if (c1 == -1000 || c == c1 || c == c2)
3343 TRYPAREN(paren, n, PL_reginput);
3344 REGCP_UNWIND(lastcp);
3346 /* Couldn't or didn't -- back up. */
3348 PL_reginput = locinput = HOPc(locinput, -1);
3355 if (PL_reg_call_cc) {
3356 re_cc_state *cur_call_cc = PL_reg_call_cc;
3357 CURCUR *cctmp = PL_regcc;
3358 regexp *re = PL_reg_re;
3359 CHECKPOINT cp, lastcp;
3361 cp = regcppush(0); /* Save *all* the positions. */
3363 regcp_set_to(PL_reg_call_cc->ss); /* Restore parens of
3365 PL_reginput = locinput; /* Make position available to
3367 cache_re(PL_reg_call_cc->re);
3368 PL_regcc = PL_reg_call_cc->cc;
3369 PL_reg_call_cc = PL_reg_call_cc->prev;
3370 if (regmatch(cur_call_cc->node)) {
3371 PL_reg_call_cc = cur_call_cc;
3375 REGCP_UNWIND(lastcp);
3377 PL_reg_call_cc = cur_call_cc;
3383 PerlIO_printf(Perl_debug_log,
3384 "%*s continuation failed...\n",
3385 REPORT_CODE_OFF+PL_regindent*2, "")
3389 if (locinput < PL_regtill) {
3390 DEBUG_r(PerlIO_printf(Perl_debug_log,
3391 "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n",
3393 (long)(locinput - PL_reg_starttry),
3394 (long)(PL_regtill - PL_reg_starttry),
3396 sayNO_FINAL; /* Cannot match: too short. */
3398 PL_reginput = locinput; /* put where regtry can find it */
3399 sayYES_FINAL; /* Success! */
3401 PL_reginput = locinput; /* put where regtry can find it */
3402 sayYES_LOUD; /* Success! */
3405 PL_reginput = locinput;
3410 s = HOPBACKc(locinput, scan->flags);
3416 PL_reginput = locinput;
3421 s = HOPBACKc(locinput, scan->flags);
3427 PL_reginput = locinput;
3430 inner = NEXTOPER(NEXTOPER(scan));
3431 if (regmatch(inner) != n) {
3446 if (OP(scan) == SUSPEND) {
3447 locinput = PL_reginput;
3448 nextchr = UCHARAT(locinput);
3453 next = scan + ARG(scan);
3458 PerlIO_printf(Perl_error_log, "%"UVxf" %d\n",
3459 PTR2UV(scan), OP(scan));
3460 Perl_croak(aTHX_ "regexp memory corruption");
3467 * We get here only if there's trouble -- normally "case END" is
3468 * the terminating point.
3470 Perl_croak(aTHX_ "corrupted regexp pointers");
3476 PerlIO_printf(Perl_debug_log,
3477 "%*s %scould match...%s\n",
3478 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4],PL_colors[5])
3482 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n",
3483 PL_colors[4],PL_colors[5]));
3489 #if 0 /* Breaks $^R */
3497 PerlIO_printf(Perl_debug_log,
3498 "%*s %sfailed...%s\n",
3499 REPORT_CODE_OFF+PL_regindent*2, "",PL_colors[4],PL_colors[5])
3505 re_unwind_t *uw = SSPTRt(unwind,re_unwind_t);
3508 case RE_UNWIND_BRANCH:
3509 case RE_UNWIND_BRANCHJ:
3511 re_unwind_branch_t *uwb = &(uw->branch);
3512 I32 lastparen = uwb->lastparen;
3514 REGCP_UNWIND(uwb->lastcp);
3515 for (n = *PL_reglastparen; n > lastparen; n--)
3517 *PL_reglastparen = n;
3518 scan = next = uwb->next;
3520 OP(scan) != (uwb->type == RE_UNWIND_BRANCH
3521 ? BRANCH : BRANCHJ) ) { /* Failure */
3528 /* Have more choice yet. Reuse the same uwb. */
3530 if ((n = (uwb->type == RE_UNWIND_BRANCH
3531 ? NEXT_OFF(next) : ARG(next))))
3534 next = NULL; /* XXXX Needn't unwinding in this case... */
3536 next = NEXTOPER(scan);
3537 if (uwb->type == RE_UNWIND_BRANCHJ)
3538 next = NEXTOPER(next);
3539 locinput = uwb->locinput;
3540 nextchr = uwb->nextchr;
3542 PL_regindent = uwb->regindent;
3549 Perl_croak(aTHX_ "regexp unwind memory corruption");
3560 - regrepeat - repeatedly match something simple, report how many
3563 * [This routine now assumes that it will only match on things of length 1.
3564 * That was true before, but now we assume scan - reginput is the count,
3565 * rather than incrementing count on every character. [Er, except utf8.]]
3568 S_regrepeat(pTHX_ regnode *p, I32 max)
3570 register char *scan;
3572 register char *loceol = PL_regeol;
3573 register I32 hardcount = 0;
3574 register bool do_utf8 = DO_UTF8(PL_reg_sv);
3577 if (max != REG_INFTY && max < loceol - scan)
3578 loceol = scan + max;
3583 while (scan < loceol && hardcount < max && *scan != '\n') {
3584 scan += UTF8SKIP(scan);
3588 while (scan < loceol && *scan != '\n')
3595 case EXACT: /* length of string is 1 */
3597 while (scan < loceol && UCHARAT(scan) == c)
3600 case EXACTF: /* length of string is 1 */
3602 while (scan < loceol &&
3603 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold[c]))
3606 case EXACTFL: /* length of string is 1 */
3607 PL_reg_flags |= RF_tainted;
3609 while (scan < loceol &&
3610 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold_locale[c]))
3616 while (hardcount < max && scan < loceol &&
3617 reginclass(p, (U8*)scan, do_utf8)) {
3618 scan += UTF8SKIP(scan);
3622 while (scan < loceol && reginclass(p, (U8*)scan, do_utf8))
3629 LOAD_UTF8_CHARCLASS(alnum,"a");
3630 while (hardcount < max && scan < loceol &&
3631 swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
3632 scan += UTF8SKIP(scan);
3636 while (scan < loceol && isALNUM(*scan))
3641 PL_reg_flags |= RF_tainted;
3644 while (hardcount < max && scan < loceol &&
3645 isALNUM_LC_utf8((U8*)scan)) {
3646 scan += UTF8SKIP(scan);
3650 while (scan < loceol && isALNUM_LC(*scan))
3657 LOAD_UTF8_CHARCLASS(alnum,"a");
3658 while (hardcount < max && scan < loceol &&
3659 !swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
3660 scan += UTF8SKIP(scan);
3664 while (scan < loceol && !isALNUM(*scan))
3669 PL_reg_flags |= RF_tainted;
3672 while (hardcount < max && scan < loceol &&
3673 !isALNUM_LC_utf8((U8*)scan)) {
3674 scan += UTF8SKIP(scan);
3678 while (scan < loceol && !isALNUM_LC(*scan))
3685 LOAD_UTF8_CHARCLASS(space," ");
3686 while (hardcount < max && scan < loceol &&
3688 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
3689 scan += UTF8SKIP(scan);
3693 while (scan < loceol && isSPACE(*scan))
3698 PL_reg_flags |= RF_tainted;
3701 while (hardcount < max && scan < loceol &&
3702 (*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
3703 scan += UTF8SKIP(scan);
3707 while (scan < loceol && isSPACE_LC(*scan))
3714 LOAD_UTF8_CHARCLASS(space," ");
3715 while (hardcount < max && scan < loceol &&
3717 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
3718 scan += UTF8SKIP(scan);
3722 while (scan < loceol && !isSPACE(*scan))
3727 PL_reg_flags |= RF_tainted;
3730 while (hardcount < max && scan < loceol &&
3731 !(*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
3732 scan += UTF8SKIP(scan);
3736 while (scan < loceol && !isSPACE_LC(*scan))
3743 LOAD_UTF8_CHARCLASS(digit,"0");
3744 while (hardcount < max && scan < loceol &&
3745 swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
3746 scan += UTF8SKIP(scan);
3750 while (scan < loceol && isDIGIT(*scan))
3757 LOAD_UTF8_CHARCLASS(digit,"0");
3758 while (hardcount < max && scan < loceol &&
3759 !swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
3760 scan += UTF8SKIP(scan);
3764 while (scan < loceol && !isDIGIT(*scan))
3768 default: /* Called on something of 0 width. */
3769 break; /* So match right here or not at all. */
3775 c = scan - PL_reginput;
3780 SV *prop = sv_newmortal();
3783 PerlIO_printf(Perl_debug_log,
3784 "%*s %s can match %"IVdf" times out of %"IVdf"...\n",
3785 REPORT_CODE_OFF+1, "", SvPVX(prop),(IV)c,(IV)max);
3792 - regrepeat_hard - repeatedly match something, report total lenth and length
3794 * The repeater is supposed to have constant length.
3798 S_regrepeat_hard(pTHX_ regnode *p, I32 max, I32 *lp)
3800 register char *scan = Nullch;
3801 register char *start;
3802 register char *loceol = PL_regeol;
3804 I32 count = 0, res = 1;
3809 start = PL_reginput;
3810 if (DO_UTF8(PL_reg_sv)) {
3811 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
3814 while (start < PL_reginput) {
3816 start += UTF8SKIP(start);
3827 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
3829 *lp = l = PL_reginput - start;
3830 if (max != REG_INFTY && l*max < loceol - scan)
3831 loceol = scan + l*max;
3844 - regclass_swash - prepare the utf8 swash
3848 Perl_regclass_swash(pTHX_ register regnode* node, bool doinit, SV** initsvp)
3853 if (PL_regdata && PL_regdata->count) {
3856 if (PL_regdata->what[n] == 's') {
3857 SV *rv = (SV*)PL_regdata->data[n];
3858 AV *av = (AV*)SvRV((SV*)rv);
3861 si = *av_fetch(av, 0, FALSE);
3862 a = av_fetch(av, 1, FALSE);
3866 else if (si && doinit) {
3867 sw = swash_init("utf8", "", si, 1, 0);
3868 (void)av_store(av, 1, sw);
3880 - reginclass - determine if a character falls into a character class
3884 S_reginclass(pTHX_ register regnode *n, register U8* p, register bool do_utf8)
3886 char flags = ANYOF_FLAGS(n);
3891 c = do_utf8 ? utf8_to_uvchr(p, &len) : *p;
3893 if (do_utf8 || (flags & ANYOF_UNICODE)) {
3894 if (do_utf8 && !ANYOF_RUNTIME(n)) {
3895 if (len != (STRLEN)-1 && c < 256 && ANYOF_BITMAP_TEST(n, c))
3898 if (!match && do_utf8 && (flags & ANYOF_UNICODE_ALL) && c >= 256)
3901 SV *sw = regclass_swash(n, TRUE, 0);
3904 if (swash_fetch(sw, p, do_utf8))
3906 else if (flags & ANYOF_FOLD) {
3907 U8 tmpbuf[UTF8_MAXLEN+1];
3909 if (flags & ANYOF_LOCALE) {
3910 PL_reg_flags |= RF_tainted;
3911 uvchr_to_utf8(tmpbuf, toLOWER_LC_utf8(p));
3914 uvchr_to_utf8(tmpbuf, toLOWER_utf8(p));
3915 if (swash_fetch(sw, tmpbuf, do_utf8))
3921 if (!match && c < 256) {
3922 if (ANYOF_BITMAP_TEST(n, c))
3924 else if (flags & ANYOF_FOLD) {
3927 if (flags & ANYOF_LOCALE) {
3928 PL_reg_flags |= RF_tainted;
3929 f = PL_fold_locale[c];
3933 if (f != c && ANYOF_BITMAP_TEST(n, f))
3937 if (!match && (flags & ANYOF_CLASS)) {
3938 PL_reg_flags |= RF_tainted;
3940 (ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) ||
3941 (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) ||
3942 (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) ||
3943 (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) ||
3944 (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) ||
3945 (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) ||
3946 (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) ||
3947 (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) ||
3948 (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) ||
3949 (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) ||
3950 (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII(c)) ||
3951 (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII(c)) ||
3952 (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) ||
3953 (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) ||
3954 (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) ||
3955 (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) ||
3956 (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) ||
3957 (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) ||
3958 (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) ||
3959 (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) ||
3960 (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) ||
3961 (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) ||
3962 (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) ||
3963 (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) ||
3964 (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) ||
3965 (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) ||
3966 (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) ||
3967 (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) ||
3968 (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK(c)) ||
3969 (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK(c))
3970 ) /* How's that for a conditional? */
3977 return (flags & ANYOF_INVERT) ? !match : match;
3981 S_reghop(pTHX_ U8 *s, I32 off)
3983 return S_reghop3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
3987 S_reghop3(pTHX_ U8 *s, I32 off, U8* lim)
3990 while (off-- && s < lim) {
3991 /* XXX could check well-formedness here */
3999 if (UTF8_IS_CONTINUED(*s)) {
4000 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4003 /* XXX could check well-formedness here */
4011 S_reghopmaybe(pTHX_ U8 *s, I32 off)
4013 return S_reghopmaybe3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4017 S_reghopmaybe3(pTHX_ U8* s, I32 off, U8* lim)
4020 while (off-- && s < lim) {
4021 /* XXX could check well-formedness here */
4031 if (UTF8_IS_CONTINUED(*s)) {
4032 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4035 /* XXX could check well-formedness here */
4051 restore_pos(pTHXo_ void *arg)
4053 if (PL_reg_eval_set) {
4054 if (PL_reg_oldsaved) {
4055 PL_reg_re->subbeg = PL_reg_oldsaved;
4056 PL_reg_re->sublen = PL_reg_oldsavedlen;
4057 RX_MATCH_COPIED_on(PL_reg_re);
4059 PL_reg_magic->mg_len = PL_reg_oldpos;
4060 PL_reg_eval_set = 0;
4061 PL_curpm = PL_reg_oldcurpm;