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 5
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 SSPUSHPTR(PL_reginput);
163 #define REGCP_FRAME_ELEMS 2
164 /* REGCP_FRAME_ELEMS are part of the REGCP_OTHER_ELEMS and
165 * are needed for the regexp context stack bookkeeping. */
166 SSPUSHINT(paren_elems_to_push + REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS);
167 SSPUSHINT(SAVEt_REGCONTEXT); /* Magic cookie. */
172 /* These are needed since we do not localize EVAL nodes: */
173 # define REGCP_SET(cp) DEBUG_r(PerlIO_printf(Perl_debug_log, \
174 " Setting an EVAL scope, savestack=%"IVdf"\n", \
175 (IV)PL_savestack_ix)); cp = PL_savestack_ix
177 # define REGCP_UNWIND(cp) DEBUG_r(cp != PL_savestack_ix ? \
178 PerlIO_printf(Perl_debug_log, \
179 " Clearing an EVAL scope, savestack=%"IVdf"..%"IVdf"\n", \
180 (IV)(cp), (IV)PL_savestack_ix) : 0); regcpblow(cp)
190 /* Pop REGCP_OTHER_ELEMS before the parentheses loop starts. */
192 assert(i == SAVEt_REGCONTEXT); /* Check that the magic cookie is there. */
193 i = SSPOPINT; /* Parentheses elements to pop. */
194 input = (char *) SSPOPPTR;
195 *PL_reglastparen = SSPOPINT;
196 PL_regsize = SSPOPINT;
198 /* Now restore the parentheses context. */
199 for (i -= (REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS);
200 i > 0; i -= REGCP_PAREN_ELEMS) {
201 paren = (U32)SSPOPINT;
202 PL_reg_start_tmp[paren] = (char *) SSPOPPTR;
203 PL_regstartp[paren] = SSPOPINT;
205 if (paren <= *PL_reglastparen)
206 PL_regendp[paren] = tmps;
208 PerlIO_printf(Perl_debug_log,
209 " restoring \\%"UVuf" to %"IVdf"(%"IVdf")..%"IVdf"%s\n",
210 (UV)paren, (IV)PL_regstartp[paren],
211 (IV)(PL_reg_start_tmp[paren] - PL_bostr),
212 (IV)PL_regendp[paren],
213 (paren > *PL_reglastparen ? "(no)" : ""));
217 if (*PL_reglastparen + 1 <= PL_regnpar) {
218 PerlIO_printf(Perl_debug_log,
219 " restoring \\%"IVdf"..\\%"IVdf" to undef\n",
220 (IV)(*PL_reglastparen + 1), (IV)PL_regnpar);
224 /* It would seem that the similar code in regtry()
225 * already takes care of this, and in fact it is in
226 * a better location to since this code can #if 0-ed out
227 * but the code in regtry() is needed or otherwise tests
228 * requiring null fields (pat.t#187 and split.t#{13,14}
229 * (as of patchlevel 7877) will fail. Then again,
230 * this code seems to be necessary or otherwise
231 * building DynaLoader will fail:
232 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
234 for (paren = *PL_reglastparen + 1; paren <= PL_regnpar; paren++) {
235 if (paren > PL_regsize)
236 PL_regstartp[paren] = -1;
237 PL_regendp[paren] = -1;
244 S_regcp_set_to(pTHX_ I32 ss)
246 I32 tmp = PL_savestack_ix;
248 PL_savestack_ix = ss;
250 PL_savestack_ix = tmp;
254 typedef struct re_cc_state
258 struct re_cc_state *prev;
263 #define regcpblow(cp) LEAVE_SCOPE(cp) /* Ignores regcppush()ed data. */
265 #define TRYPAREN(paren, n, input) { \
268 PL_regstartp[paren] = HOPc(input, -1) - PL_bostr; \
269 PL_regendp[paren] = input - PL_bostr; \
272 PL_regendp[paren] = -1; \
274 if (regmatch(next)) \
277 PL_regendp[paren] = -1; \
282 * pregexec and friends
286 - pregexec - match a regexp against a string
289 Perl_pregexec(pTHX_ register regexp *prog, char *stringarg, register char *strend,
290 char *strbeg, I32 minend, SV *screamer, U32 nosave)
291 /* strend: pointer to null at end of string */
292 /* strbeg: real beginning of string */
293 /* minend: end of match must be >=minend after stringarg. */
294 /* nosave: For optimizations. */
297 regexec_flags(prog, stringarg, strend, strbeg, minend, screamer, NULL,
298 nosave ? 0 : REXEC_COPY_STR);
302 S_cache_re(pTHX_ regexp *prog)
304 PL_regprecomp = prog->precomp; /* Needed for FAIL. */
306 PL_regprogram = prog->program;
308 PL_regnpar = prog->nparens;
309 PL_regdata = prog->data;
314 * Need to implement the following flags for reg_anch:
316 * USE_INTUIT_NOML - Useful to call re_intuit_start() first
318 * INTUIT_AUTORITATIVE_NOML - Can trust a positive answer
319 * INTUIT_AUTORITATIVE_ML
320 * INTUIT_ONCE_NOML - Intuit can match in one location only.
323 * Another flag for this function: SECOND_TIME (so that float substrs
324 * with giant delta may be not rechecked).
327 /* Assumptions: if ANCH_GPOS, then strpos is anchored. XXXX Check GPOS logic */
329 /* If SCREAM, then SvPVX(sv) should be compatible with strpos and strend.
330 Otherwise, only SvCUR(sv) is used to get strbeg. */
332 /* XXXX We assume that strpos is strbeg unless sv. */
334 /* XXXX Some places assume that there is a fixed substring.
335 An update may be needed if optimizer marks as "INTUITable"
336 RExen without fixed substrings. Similarly, it is assumed that
337 lengths of all the strings are no more than minlen, thus they
338 cannot come from lookahead.
339 (Or minlen should take into account lookahead.) */
341 /* A failure to find a constant substring means that there is no need to make
342 an expensive call to REx engine, thus we celebrate a failure. Similarly,
343 finding a substring too deep into the string means that less calls to
344 regtry() should be needed.
346 REx compiler's optimizer found 4 possible hints:
347 a) Anchored substring;
349 c) Whether we are anchored (beginning-of-line or \G);
350 d) First node (of those at offset 0) which may distingush positions;
351 We use a)b)d) and multiline-part of c), and try to find a position in the
352 string which does not contradict any of them.
355 /* Most of decisions we do here should have been done at compile time.
356 The nodes of the REx which we used for the search should have been
357 deleted from the finite automaton. */
360 Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos,
361 char *strend, U32 flags, re_scream_pos_data *data)
363 register I32 start_shift;
364 /* Should be nonnegative! */
365 register I32 end_shift;
372 register char *other_last = Nullch; /* other substr checked before this */
373 char *check_at; /* check substr found at this pos */
375 char *i_strpos = strpos;
378 DEBUG_r( if (!PL_colorset) reginitcolors() );
379 DEBUG_r(PerlIO_printf(Perl_debug_log,
380 "%sGuessing start of match, REx%s `%s%.60s%s%s' against `%s%.*s%s%s'...\n",
381 PL_colors[4],PL_colors[5],PL_colors[0],
384 (strlen(prog->precomp) > 60 ? "..." : ""),
386 (int)(strend - strpos > 60 ? 60 : strend - strpos),
387 strpos, PL_colors[1],
388 (strend - strpos > 60 ? "..." : ""))
391 if (prog->reganch & ROPT_UTF8)
392 PL_reg_flags |= RF_utf8;
394 if (prog->minlen > CHR_DIST((U8*)strend, (U8*)strpos)) {
395 DEBUG_r(PerlIO_printf(Perl_debug_log, "String too short...\n"));
398 strbeg = (sv && SvPOK(sv)) ? strend - SvCUR(sv) : strpos;
400 check = prog->check_substr;
401 if (prog->reganch & ROPT_ANCH) { /* Match at beg-of-str or after \n */
402 ml_anch = !( (prog->reganch & ROPT_ANCH_SINGLE)
403 || ( (prog->reganch & ROPT_ANCH_BOL)
404 && !PL_multiline ) ); /* Check after \n? */
407 if ( !(prog->reganch & (ROPT_ANCH_GPOS /* Checked by the caller */
408 | ROPT_IMPLICIT)) /* not a real BOL */
409 /* SvCUR is not set on references: SvRV and SvPVX overlap */
411 && (strpos != strbeg)) {
412 DEBUG_r(PerlIO_printf(Perl_debug_log, "Not at start...\n"));
415 if (prog->check_offset_min == prog->check_offset_max &&
416 !(prog->reganch & ROPT_SANY_SEEN)) {
417 /* Substring at constant offset from beg-of-str... */
420 s = HOP3c(strpos, prog->check_offset_min, strend);
422 slen = SvCUR(check); /* >= 1 */
424 if ( strend - s > slen || strend - s < slen - 1
425 || (strend - s == slen && strend[-1] != '\n')) {
426 DEBUG_r(PerlIO_printf(Perl_debug_log, "String too long...\n"));
429 /* Now should match s[0..slen-2] */
431 if (slen && (*SvPVX(check) != *s
433 && memNE(SvPVX(check), s, slen)))) {
435 DEBUG_r(PerlIO_printf(Perl_debug_log, "String not equal...\n"));
439 else if (*SvPVX(check) != *s
440 || ((slen = SvCUR(check)) > 1
441 && memNE(SvPVX(check), s, slen)))
443 goto success_at_start;
446 /* Match is anchored, but substr is not anchored wrt beg-of-str. */
448 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
449 end_shift = prog->minlen - start_shift -
450 CHR_SVLEN(check) + (SvTAIL(check) != 0);
452 I32 end = prog->check_offset_max + CHR_SVLEN(check)
453 - (SvTAIL(check) != 0);
454 I32 eshift = CHR_DIST((U8*)strend, (U8*)s) - end;
456 if (end_shift < eshift)
460 else { /* Can match at random position */
463 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
464 /* Should be nonnegative! */
465 end_shift = prog->minlen - start_shift -
466 CHR_SVLEN(check) + (SvTAIL(check) != 0);
469 #ifdef DEBUGGING /* 7/99: reports of failure (with the older version) */
471 Perl_croak(aTHX_ "panic: end_shift");
475 /* Find a possible match in the region s..strend by looking for
476 the "check" substring in the region corrected by start/end_shift. */
477 if (flags & REXEC_SCREAM) {
478 I32 p = -1; /* Internal iterator of scream. */
479 I32 *pp = data ? data->scream_pos : &p;
481 if (PL_screamfirst[BmRARE(check)] >= 0
482 || ( BmRARE(check) == '\n'
483 && (BmPREVIOUS(check) == SvCUR(check) - 1)
485 s = screaminstr(sv, check,
486 start_shift + (s - strbeg), end_shift, pp, 0);
490 *data->scream_olds = s;
492 else if (prog->reganch & ROPT_SANY_SEEN)
493 s = fbm_instr((U8*)(s + start_shift),
494 (U8*)(strend - end_shift),
495 check, PL_multiline ? FBMrf_MULTILINE : 0);
497 s = fbm_instr(HOP3(s, start_shift, strend),
498 HOP3(strend, -end_shift, strbeg),
499 check, PL_multiline ? FBMrf_MULTILINE : 0);
501 /* Update the count-of-usability, remove useless subpatterns,
504 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s %s substr `%s%.*s%s'%s%s",
505 (s ? "Found" : "Did not find"),
506 ((check == prog->anchored_substr) ? "anchored" : "floating"),
508 (int)(SvCUR(check) - (SvTAIL(check)!=0)),
510 PL_colors[1], (SvTAIL(check) ? "$" : ""),
511 (s ? " at offset " : "...\n") ) );
518 /* Finish the diagnostic message */
519 DEBUG_r(PerlIO_printf(Perl_debug_log, "%ld...\n", (long)(s - i_strpos)) );
521 /* Got a candidate. Check MBOL anchoring, and the *other* substr.
522 Start with the other substr.
523 XXXX no SCREAM optimization yet - and a very coarse implementation
524 XXXX /ttx+/ results in anchored=`ttx', floating=`x'. floating will
525 *always* match. Probably should be marked during compile...
526 Probably it is right to do no SCREAM here...
529 if (prog->float_substr && prog->anchored_substr) {
530 /* Take into account the "other" substring. */
531 /* XXXX May be hopelessly wrong for UTF... */
534 if (check == prog->float_substr) {
537 char *last = HOP3c(s, -start_shift, strbeg), *last1, *last2;
540 t = s - prog->check_offset_max;
541 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
542 && (!(prog->reganch & ROPT_UTF8)
543 || ((t = reghopmaybe3_c(s, -(prog->check_offset_max), strpos))
548 t = HOP3c(t, prog->anchored_offset, strend);
549 if (t < other_last) /* These positions already checked */
551 last2 = last1 = HOP3c(strend, -prog->minlen, strbeg);
554 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
555 /* On end-of-str: see comment below. */
556 s = fbm_instr((unsigned char*)t,
557 HOP3(HOP3(last1, prog->anchored_offset, strend)
558 + SvCUR(prog->anchored_substr),
559 -(SvTAIL(prog->anchored_substr)!=0), strbeg),
560 prog->anchored_substr,
561 PL_multiline ? FBMrf_MULTILINE : 0);
562 DEBUG_r(PerlIO_printf(Perl_debug_log,
563 "%s anchored substr `%s%.*s%s'%s",
564 (s ? "Found" : "Contradicts"),
566 (int)(SvCUR(prog->anchored_substr)
567 - (SvTAIL(prog->anchored_substr)!=0)),
568 SvPVX(prog->anchored_substr),
569 PL_colors[1], (SvTAIL(prog->anchored_substr) ? "$" : "")));
571 if (last1 >= last2) {
572 DEBUG_r(PerlIO_printf(Perl_debug_log,
573 ", giving up...\n"));
576 DEBUG_r(PerlIO_printf(Perl_debug_log,
577 ", trying floating at offset %ld...\n",
578 (long)(HOP3c(s1, 1, strend) - i_strpos)));
579 other_last = HOP3c(last1, prog->anchored_offset+1, strend);
580 s = HOP3c(last, 1, strend);
584 DEBUG_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
585 (long)(s - i_strpos)));
586 t = HOP3c(s, -prog->anchored_offset, strbeg);
587 other_last = HOP3c(s, 1, strend);
595 else { /* Take into account the floating substring. */
599 t = HOP3c(s, -start_shift, strbeg);
601 HOP3c(strend, -prog->minlen + prog->float_min_offset, strbeg);
602 if (CHR_DIST((U8*)last, (U8*)t) > prog->float_max_offset)
603 last = HOP3c(t, prog->float_max_offset, strend);
604 s = HOP3c(t, prog->float_min_offset, strend);
607 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
608 /* fbm_instr() takes into account exact value of end-of-str
609 if the check is SvTAIL(ed). Since false positives are OK,
610 and end-of-str is not later than strend we are OK. */
611 s = fbm_instr((unsigned char*)s,
612 (unsigned char*)last + SvCUR(prog->float_substr)
613 - (SvTAIL(prog->float_substr)!=0),
614 prog->float_substr, PL_multiline ? FBMrf_MULTILINE : 0);
615 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s floating substr `%s%.*s%s'%s",
616 (s ? "Found" : "Contradicts"),
618 (int)(SvCUR(prog->float_substr)
619 - (SvTAIL(prog->float_substr)!=0)),
620 SvPVX(prog->float_substr),
621 PL_colors[1], (SvTAIL(prog->float_substr) ? "$" : "")));
624 DEBUG_r(PerlIO_printf(Perl_debug_log,
625 ", giving up...\n"));
628 DEBUG_r(PerlIO_printf(Perl_debug_log,
629 ", trying anchored starting at offset %ld...\n",
630 (long)(s1 + 1 - i_strpos)));
632 s = HOP3c(t, 1, strend);
636 DEBUG_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
637 (long)(s - i_strpos)));
638 other_last = s; /* Fix this later. --Hugo */
647 t = s - prog->check_offset_max;
648 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
649 && (!(prog->reganch & ROPT_UTF8)
650 || ((t = reghopmaybe3_c(s, -prog->check_offset_max, strpos))
652 /* Fixed substring is found far enough so that the match
653 cannot start at strpos. */
655 if (ml_anch && t[-1] != '\n') {
656 /* Eventually fbm_*() should handle this, but often
657 anchored_offset is not 0, so this check will not be wasted. */
658 /* XXXX In the code below we prefer to look for "^" even in
659 presence of anchored substrings. And we search even
660 beyond the found float position. These pessimizations
661 are historical artefacts only. */
663 while (t < strend - prog->minlen) {
665 if (t < check_at - prog->check_offset_min) {
666 if (prog->anchored_substr) {
667 /* Since we moved from the found position,
668 we definitely contradict the found anchored
669 substr. Due to the above check we do not
670 contradict "check" substr.
671 Thus we can arrive here only if check substr
672 is float. Redo checking for "other"=="fixed".
675 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld, rescanning for anchored from offset %ld...\n",
676 PL_colors[0],PL_colors[1], (long)(strpos - i_strpos), (long)(strpos - i_strpos + prog->anchored_offset)));
677 goto do_other_anchored;
679 /* We don't contradict the found floating substring. */
680 /* XXXX Why not check for STCLASS? */
682 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld...\n",
683 PL_colors[0],PL_colors[1], (long)(s - i_strpos)));
686 /* Position contradicts check-string */
687 /* XXXX probably better to look for check-string
688 than for "\n", so one should lower the limit for t? */
689 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m, restarting lookup for check-string at offset %ld...\n",
690 PL_colors[0],PL_colors[1], (long)(t + 1 - i_strpos)));
691 other_last = strpos = s = t + 1;
696 DEBUG_r(PerlIO_printf(Perl_debug_log, "Did not find /%s^%s/m...\n",
697 PL_colors[0],PL_colors[1]));
701 DEBUG_r(PerlIO_printf(Perl_debug_log, "Starting position does not contradict /%s^%s/m...\n",
702 PL_colors[0],PL_colors[1]));
706 ++BmUSEFUL(prog->check_substr); /* hooray/5 */
709 /* The found string does not prohibit matching at strpos,
710 - no optimization of calling REx engine can be performed,
711 unless it was an MBOL and we are not after MBOL,
712 or a future STCLASS check will fail this. */
714 /* Even in this situation we may use MBOL flag if strpos is offset
715 wrt the start of the string. */
716 if (ml_anch && sv && !SvROK(sv) /* See prev comment on SvROK */
717 && (strpos != strbeg) && strpos[-1] != '\n'
718 /* May be due to an implicit anchor of m{.*foo} */
719 && !(prog->reganch & ROPT_IMPLICIT))
724 DEBUG_r( if (ml_anch)
725 PerlIO_printf(Perl_debug_log, "Position at offset %ld does not contradict /%s^%s/m...\n",
726 (long)(strpos - i_strpos), PL_colors[0],PL_colors[1]);
729 if (!(prog->reganch & ROPT_NAUGHTY) /* XXXX If strpos moved? */
730 && prog->check_substr /* Could be deleted already */
731 && --BmUSEFUL(prog->check_substr) < 0
732 && prog->check_substr == prog->float_substr)
734 /* If flags & SOMETHING - do not do it many times on the same match */
735 DEBUG_r(PerlIO_printf(Perl_debug_log, "... Disabling check substring...\n"));
736 SvREFCNT_dec(prog->check_substr);
737 prog->check_substr = Nullsv; /* disable */
738 prog->float_substr = Nullsv; /* clear */
739 check = Nullsv; /* abort */
741 /* XXXX This is a remnant of the old implementation. It
742 looks wasteful, since now INTUIT can use many
744 prog->reganch &= ~RE_USE_INTUIT;
751 /* XXXX BmUSEFUL already changed, maybe multiple change is meaningful... */
752 if (prog->regstclass) {
753 /* minlen == 0 is possible if regstclass is \b or \B,
754 and the fixed substr is ''$.
755 Since minlen is already taken into account, s+1 is before strend;
756 accidentally, minlen >= 1 guaranties no false positives at s + 1
757 even for \b or \B. But (minlen? 1 : 0) below assumes that
758 regstclass does not come from lookahead... */
759 /* If regstclass takes bytelength more than 1: If charlength==1, OK.
760 This leaves EXACTF only, which is dealt with in find_byclass(). */
761 U8* str = (U8*)STRING(prog->regstclass);
762 int cl_l = (PL_regkind[(U8)OP(prog->regstclass)] == EXACT
763 ? CHR_DIST(str+STR_LEN(prog->regstclass), str)
765 char *endpos = (prog->anchored_substr || ml_anch)
766 ? HOP3c(s, (prog->minlen ? cl_l : 0), strend)
767 : (prog->float_substr
768 ? HOP3c(HOP3c(check_at, -start_shift, strbeg),
771 char *startpos = strbeg;
774 if (prog->reganch & ROPT_UTF8) {
775 PL_regdata = prog->data;
778 s = find_byclass(prog, prog->regstclass, s, endpos, startpos, 1);
783 if (endpos == strend) {
784 DEBUG_r( PerlIO_printf(Perl_debug_log,
785 "Could not match STCLASS...\n") );
788 DEBUG_r( PerlIO_printf(Perl_debug_log,
789 "This position contradicts STCLASS...\n") );
790 if ((prog->reganch & ROPT_ANCH) && !ml_anch)
792 /* Contradict one of substrings */
793 if (prog->anchored_substr) {
794 if (prog->anchored_substr == check) {
795 DEBUG_r( what = "anchored" );
797 s = HOP3c(t, 1, strend);
798 if (s + start_shift + end_shift > strend) {
799 /* XXXX Should be taken into account earlier? */
800 DEBUG_r( PerlIO_printf(Perl_debug_log,
801 "Could not match STCLASS...\n") );
806 DEBUG_r( PerlIO_printf(Perl_debug_log,
807 "Looking for %s substr starting at offset %ld...\n",
808 what, (long)(s + start_shift - i_strpos)) );
811 /* Have both, check_string is floating */
812 if (t + start_shift >= check_at) /* Contradicts floating=check */
813 goto retry_floating_check;
814 /* Recheck anchored substring, but not floating... */
818 DEBUG_r( PerlIO_printf(Perl_debug_log,
819 "Looking for anchored substr starting at offset %ld...\n",
820 (long)(other_last - i_strpos)) );
821 goto do_other_anchored;
823 /* Another way we could have checked stclass at the
824 current position only: */
829 DEBUG_r( PerlIO_printf(Perl_debug_log,
830 "Looking for /%s^%s/m starting at offset %ld...\n",
831 PL_colors[0],PL_colors[1], (long)(t - i_strpos)) );
834 if (!prog->float_substr) /* Could have been deleted */
836 /* Check is floating subtring. */
837 retry_floating_check:
838 t = check_at - start_shift;
839 DEBUG_r( what = "floating" );
840 goto hop_and_restart;
843 PerlIO_printf(Perl_debug_log,
844 "By STCLASS: moving %ld --> %ld\n",
845 (long)(t - i_strpos), (long)(s - i_strpos));
847 PerlIO_printf(Perl_debug_log,
848 "Does not contradict STCLASS...\n") );
851 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s%s:%s match at offset %ld\n",
852 PL_colors[4], (check ? "Guessed" : "Giving up"),
853 PL_colors[5], (long)(s - i_strpos)) );
856 fail_finish: /* Substring not found */
857 if (prog->check_substr) /* could be removed already */
858 BmUSEFUL(prog->check_substr) += 5; /* hooray */
860 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch rejected by optimizer%s\n",
861 PL_colors[4],PL_colors[5]));
865 /* We know what class REx starts with. Try to find this position... */
867 S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, char *strend, char *startpos, I32 norun)
869 I32 doevery = (prog->reganch & ROPT_SKIP) == 0;
875 register I32 tmp = 1; /* Scratch variable? */
876 register bool do_utf8 = DO_UTF8(PL_reg_sv);
878 /* We know what class it must start with. */
882 if (reginclass(c, (U8*)s, do_utf8)) {
883 if (tmp && (norun || regtry(prog, s)))
890 s += do_utf8 ? UTF8SKIP(s) : 1;
897 c1 = to_utf8_lower((U8*)m);
898 c2 = to_utf8_upper((U8*)m);
909 c2 = PL_fold_locale[c1];
914 e = s; /* Due to minlen logic of intuit() */
920 if ( utf8_to_uvchr((U8*)s, &len) == c1
927 UV c = utf8_to_uvchr((U8*)s, &len);
928 if ( (c == c1 || c == c2) && regtry(prog, s) )
937 && (ln == 1 || !(OP(c) == EXACTF
939 : ibcmp_locale(s, m, ln)))
940 && (norun || regtry(prog, s)) )
946 if ( (*(U8*)s == c1 || *(U8*)s == c2)
947 && (ln == 1 || !(OP(c) == EXACTF
949 : ibcmp_locale(s, m, ln)))
950 && (norun || regtry(prog, s)) )
957 PL_reg_flags |= RF_tainted;
964 U8 *r = reghop3((U8*)s, -1, (U8*)startpos);
966 tmp = (I32)utf8n_to_uvchr(r, s - (char*)r, 0, 0);
968 tmp = ((OP(c) == BOUND ?
969 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
970 LOAD_UTF8_CHARCLASS(alnum,"a");
972 if (tmp == !(OP(c) == BOUND ?
973 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
974 isALNUM_LC_utf8((U8*)s)))
977 if ((norun || regtry(prog, s)))
984 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
985 tmp = ((OP(c) == BOUND ? isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
988 !(OP(c) == BOUND ? isALNUM(*s) : isALNUM_LC(*s))) {
990 if ((norun || regtry(prog, s)))
996 if ((!prog->minlen && tmp) && (norun || regtry(prog, s)))
1000 PL_reg_flags |= RF_tainted;
1007 U8 *r = reghop3((U8*)s, -1, (U8*)startpos);
1009 tmp = (I32)utf8n_to_uvchr(r, s - (char*)r, 0, 0);
1011 tmp = ((OP(c) == NBOUND ?
1012 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1013 LOAD_UTF8_CHARCLASS(alnum,"a");
1014 while (s < strend) {
1015 if (tmp == !(OP(c) == NBOUND ?
1016 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1017 isALNUM_LC_utf8((U8*)s)))
1019 else if ((norun || regtry(prog, s)))
1025 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1026 tmp = ((OP(c) == NBOUND ?
1027 isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1028 while (s < strend) {
1030 !(OP(c) == NBOUND ? isALNUM(*s) : isALNUM_LC(*s)))
1032 else if ((norun || regtry(prog, s)))
1037 if ((!prog->minlen && !tmp) && (norun || regtry(prog, s)))
1042 LOAD_UTF8_CHARCLASS(alnum,"a");
1043 while (s < strend) {
1044 if (swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1045 if (tmp && (norun || regtry(prog, s)))
1056 while (s < strend) {
1058 if (tmp && (norun || regtry(prog, s)))
1070 PL_reg_flags |= RF_tainted;
1072 while (s < strend) {
1073 if (isALNUM_LC_utf8((U8*)s)) {
1074 if (tmp && (norun || regtry(prog, s)))
1085 while (s < strend) {
1086 if (isALNUM_LC(*s)) {
1087 if (tmp && (norun || regtry(prog, s)))
1100 LOAD_UTF8_CHARCLASS(alnum,"a");
1101 while (s < strend) {
1102 if (!swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1103 if (tmp && (norun || regtry(prog, s)))
1114 while (s < strend) {
1116 if (tmp && (norun || regtry(prog, s)))
1128 PL_reg_flags |= RF_tainted;
1130 while (s < strend) {
1131 if (!isALNUM_LC_utf8((U8*)s)) {
1132 if (tmp && (norun || regtry(prog, s)))
1143 while (s < strend) {
1144 if (!isALNUM_LC(*s)) {
1145 if (tmp && (norun || regtry(prog, s)))
1158 LOAD_UTF8_CHARCLASS(space," ");
1159 while (s < strend) {
1160 if (*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8)) {
1161 if (tmp && (norun || regtry(prog, s)))
1172 while (s < strend) {
1174 if (tmp && (norun || regtry(prog, s)))
1186 PL_reg_flags |= RF_tainted;
1188 while (s < strend) {
1189 if (*s == ' ' || isSPACE_LC_utf8((U8*)s)) {
1190 if (tmp && (norun || regtry(prog, s)))
1201 while (s < strend) {
1202 if (isSPACE_LC(*s)) {
1203 if (tmp && (norun || regtry(prog, s)))
1216 LOAD_UTF8_CHARCLASS(space," ");
1217 while (s < strend) {
1218 if (!(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8))) {
1219 if (tmp && (norun || regtry(prog, s)))
1230 while (s < strend) {
1232 if (tmp && (norun || regtry(prog, s)))
1244 PL_reg_flags |= RF_tainted;
1246 while (s < strend) {
1247 if (!(*s == ' ' || isSPACE_LC_utf8((U8*)s))) {
1248 if (tmp && (norun || regtry(prog, s)))
1259 while (s < strend) {
1260 if (!isSPACE_LC(*s)) {
1261 if (tmp && (norun || regtry(prog, s)))
1274 LOAD_UTF8_CHARCLASS(digit,"0");
1275 while (s < strend) {
1276 if (swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1277 if (tmp && (norun || regtry(prog, s)))
1288 while (s < strend) {
1290 if (tmp && (norun || regtry(prog, s)))
1302 PL_reg_flags |= RF_tainted;
1304 while (s < strend) {
1305 if (isDIGIT_LC_utf8((U8*)s)) {
1306 if (tmp && (norun || regtry(prog, s)))
1317 while (s < strend) {
1318 if (isDIGIT_LC(*s)) {
1319 if (tmp && (norun || regtry(prog, s)))
1332 LOAD_UTF8_CHARCLASS(digit,"0");
1333 while (s < strend) {
1334 if (!swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1335 if (tmp && (norun || regtry(prog, s)))
1346 while (s < strend) {
1348 if (tmp && (norun || regtry(prog, s)))
1360 PL_reg_flags |= RF_tainted;
1362 while (s < strend) {
1363 if (!isDIGIT_LC_utf8((U8*)s)) {
1364 if (tmp && (norun || regtry(prog, s)))
1375 while (s < strend) {
1376 if (!isDIGIT_LC(*s)) {
1377 if (tmp && (norun || regtry(prog, s)))
1389 Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c));
1398 - regexec_flags - match a regexp against a string
1401 Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char *strend,
1402 char *strbeg, I32 minend, SV *sv, void *data, U32 flags)
1403 /* strend: pointer to null at end of string */
1404 /* strbeg: real beginning of string */
1405 /* minend: end of match must be >=minend after stringarg. */
1406 /* data: May be used for some additional optimizations. */
1407 /* nosave: For optimizations. */
1410 register regnode *c;
1411 register char *startpos = stringarg;
1412 I32 minlen; /* must match at least this many chars */
1413 I32 dontbother = 0; /* how many characters not to try at end */
1414 /* I32 start_shift = 0; */ /* Offset of the start to find
1415 constant substr. */ /* CC */
1416 I32 end_shift = 0; /* Same for the end. */ /* CC */
1417 I32 scream_pos = -1; /* Internal iterator of scream. */
1419 SV* oreplsv = GvSV(PL_replgv);
1420 bool do_utf8 = DO_UTF8(sv);
1426 PL_regnarrate = DEBUG_r_TEST;
1429 /* Be paranoid... */
1430 if (prog == NULL || startpos == NULL) {
1431 Perl_croak(aTHX_ "NULL regexp parameter");
1435 minlen = prog->minlen;
1437 if (!(prog->reganch & ROPT_SANY_SEEN))
1438 if (utf8_distance((U8*)strend, (U8*)startpos) < minlen) goto phooey;
1441 if (strend - startpos < minlen) goto phooey;
1444 /* Check validity of program. */
1445 if (UCHARAT(prog->program) != REG_MAGIC) {
1446 Perl_croak(aTHX_ "corrupted regexp program");
1450 PL_reg_eval_set = 0;
1453 if (prog->reganch & ROPT_UTF8)
1454 PL_reg_flags |= RF_utf8;
1456 /* Mark beginning of line for ^ and lookbehind. */
1457 PL_regbol = startpos;
1461 /* Mark end of line for $ (and such) */
1464 /* see how far we have to get to not match where we matched before */
1465 PL_regtill = startpos+minend;
1467 /* We start without call_cc context. */
1470 /* If there is a "must appear" string, look for it. */
1473 if (prog->reganch & ROPT_GPOS_SEEN) { /* Need to have PL_reg_ganch */
1476 if (flags & REXEC_IGNOREPOS) /* Means: check only at start */
1477 PL_reg_ganch = startpos;
1478 else if (sv && SvTYPE(sv) >= SVt_PVMG
1480 && (mg = mg_find(sv, PERL_MAGIC_regex_global))
1481 && mg->mg_len >= 0) {
1482 PL_reg_ganch = strbeg + mg->mg_len; /* Defined pos() */
1483 if (prog->reganch & ROPT_ANCH_GPOS) {
1484 if (s > PL_reg_ganch)
1489 else /* pos() not defined */
1490 PL_reg_ganch = strbeg;
1493 if (do_utf8 == (UTF!=0) &&
1494 !(flags & REXEC_CHECKED) && prog->check_substr != Nullsv) {
1495 re_scream_pos_data d;
1497 d.scream_olds = &scream_olds;
1498 d.scream_pos = &scream_pos;
1499 s = re_intuit_start(prog, sv, s, strend, flags, &d);
1501 goto phooey; /* not present */
1504 DEBUG_r( if (!PL_colorset) reginitcolors() );
1505 DEBUG_r(PerlIO_printf(Perl_debug_log,
1506 "%sMatching REx%s `%s%.60s%s%s' against `%s%.*s%s%s'\n",
1507 PL_colors[4],PL_colors[5],PL_colors[0],
1510 (strlen(prog->precomp) > 60 ? "..." : ""),
1512 (int)(strend - startpos > 60 ? 60 : strend - startpos),
1513 startpos, PL_colors[1],
1514 (strend - startpos > 60 ? "..." : ""))
1517 /* Simplest case: anchored match need be tried only once. */
1518 /* [unless only anchor is BOL and multiline is set] */
1519 if (prog->reganch & (ROPT_ANCH & ~ROPT_ANCH_GPOS)) {
1520 if (s == startpos && regtry(prog, startpos))
1522 else if (PL_multiline || (prog->reganch & ROPT_IMPLICIT)
1523 || (prog->reganch & ROPT_ANCH_MBOL)) /* XXXX SBOL? */
1528 dontbother = minlen - 1;
1529 end = HOP3c(strend, -dontbother, strbeg) - 1;
1530 /* for multiline we only have to try after newlines */
1531 if (prog->check_substr) {
1535 if (regtry(prog, s))
1540 if (prog->reganch & RE_USE_INTUIT) {
1541 s = re_intuit_start(prog, sv, s + 1, strend, flags, NULL);
1552 if (*s++ == '\n') { /* don't need PL_utf8skip here */
1553 if (regtry(prog, s))
1560 } else if (prog->reganch & ROPT_ANCH_GPOS) {
1561 if (regtry(prog, PL_reg_ganch))
1566 /* Messy cases: unanchored match. */
1567 if (prog->anchored_substr && prog->reganch & ROPT_SKIP) {
1568 /* we have /x+whatever/ */
1569 /* it must be a one character string (XXXX Except UTF?) */
1570 char ch = SvPVX(prog->anchored_substr)[0];
1576 while (s < strend) {
1578 DEBUG_r( did_match = 1 );
1579 if (regtry(prog, s)) goto got_it;
1581 while (s < strend && *s == ch)
1588 while (s < strend) {
1590 DEBUG_r( did_match = 1 );
1591 if (regtry(prog, s)) goto got_it;
1593 while (s < strend && *s == ch)
1599 DEBUG_r(did_match ||
1600 PerlIO_printf(Perl_debug_log,
1601 "Did not find anchored character...\n"));
1604 else if (do_utf8 == (UTF!=0) &&
1605 (prog->anchored_substr != Nullsv
1606 || (prog->float_substr != Nullsv
1607 && prog->float_max_offset < strend - s))) {
1608 SV *must = prog->anchored_substr
1609 ? prog->anchored_substr : prog->float_substr;
1611 prog->anchored_substr ? prog->anchored_offset : prog->float_max_offset;
1613 prog->anchored_substr ? prog->anchored_offset : prog->float_min_offset;
1614 char *last = HOP3c(strend, /* Cannot start after this */
1615 -(I32)(CHR_SVLEN(must)
1616 - (SvTAIL(must) != 0) + back_min), strbeg);
1617 char *last1; /* Last position checked before */
1623 last1 = HOPc(s, -1);
1625 last1 = s - 1; /* bogus */
1627 /* XXXX check_substr already used to find `s', can optimize if
1628 check_substr==must. */
1630 dontbother = end_shift;
1631 strend = HOPc(strend, -dontbother);
1632 while ( (s <= last) &&
1633 ((flags & REXEC_SCREAM)
1634 ? (s = screaminstr(sv, must, HOP3c(s, back_min, strend) - strbeg,
1635 end_shift, &scream_pos, 0))
1636 : (s = fbm_instr((unsigned char*)HOP3(s, back_min, strend),
1637 (unsigned char*)strend, must,
1638 PL_multiline ? FBMrf_MULTILINE : 0))) ) {
1639 DEBUG_r( did_match = 1 );
1640 if (HOPc(s, -back_max) > last1) {
1641 last1 = HOPc(s, -back_min);
1642 s = HOPc(s, -back_max);
1645 char *t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1;
1647 last1 = HOPc(s, -back_min);
1651 while (s <= last1) {
1652 if (regtry(prog, s))
1658 while (s <= last1) {
1659 if (regtry(prog, s))
1665 DEBUG_r(did_match ||
1666 PerlIO_printf(Perl_debug_log, "Did not find %s substr `%s%.*s%s'%s...\n",
1667 ((must == prog->anchored_substr)
1668 ? "anchored" : "floating"),
1670 (int)(SvCUR(must) - (SvTAIL(must)!=0)),
1672 PL_colors[1], (SvTAIL(must) ? "$" : "")));
1675 else if ((c = prog->regstclass)) {
1676 if (minlen && PL_regkind[(U8)OP(prog->regstclass)] != EXACT)
1677 /* don't bother with what can't match */
1678 strend = HOPc(strend, -(minlen - 1));
1680 SV *prop = sv_newmortal();
1682 PerlIO_printf(Perl_debug_log, "Matching stclass `%s' against `%s'\n", SvPVX(prop), s);
1684 if (find_byclass(prog, c, s, strend, startpos, 0))
1686 DEBUG_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass...\n"));
1690 if (prog->float_substr != Nullsv) { /* Trim the end. */
1693 if (flags & REXEC_SCREAM) {
1694 last = screaminstr(sv, prog->float_substr, s - strbeg,
1695 end_shift, &scream_pos, 1); /* last one */
1697 last = scream_olds; /* Only one occurrence. */
1701 char *little = SvPV(prog->float_substr, len);
1703 if (SvTAIL(prog->float_substr)) {
1704 if (memEQ(strend - len + 1, little, len - 1))
1705 last = strend - len + 1;
1706 else if (!PL_multiline)
1707 last = memEQ(strend - len, little, len)
1708 ? strend - len : Nullch;
1714 last = rninstr(s, strend, little, little + len);
1716 last = strend; /* matching `$' */
1720 DEBUG_r(PerlIO_printf(Perl_debug_log,
1721 "%sCan't trim the tail, match fails (should not happen)%s\n",
1722 PL_colors[4],PL_colors[5]));
1723 goto phooey; /* Should not happen! */
1725 dontbother = strend - last + prog->float_min_offset;
1727 if (minlen && (dontbother < minlen))
1728 dontbother = minlen - 1;
1729 strend -= dontbother; /* this one's always in bytes! */
1730 /* We don't know much -- general case. */
1733 if (regtry(prog, s))
1742 if (regtry(prog, s))
1744 } while (s++ < strend);
1752 RX_MATCH_TAINTED_set(prog, PL_reg_flags & RF_tainted);
1754 if (PL_reg_eval_set) {
1755 /* Preserve the current value of $^R */
1756 if (oreplsv != GvSV(PL_replgv))
1757 sv_setsv(oreplsv, GvSV(PL_replgv));/* So that when GvSV(replgv) is
1758 restored, the value remains
1760 restore_pos(aTHXo_ 0);
1763 /* make sure $`, $&, $', and $digit will work later */
1764 if ( !(flags & REXEC_NOT_FIRST) ) {
1765 if (RX_MATCH_COPIED(prog)) {
1766 Safefree(prog->subbeg);
1767 RX_MATCH_COPIED_off(prog);
1769 if (flags & REXEC_COPY_STR) {
1770 I32 i = PL_regeol - startpos + (stringarg - strbeg);
1772 s = savepvn(strbeg, i);
1775 RX_MATCH_COPIED_on(prog);
1778 prog->subbeg = strbeg;
1779 prog->sublen = PL_regeol - strbeg; /* strend may have been modified */
1786 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n",
1787 PL_colors[4],PL_colors[5]));
1788 if (PL_reg_eval_set)
1789 restore_pos(aTHXo_ 0);
1794 - regtry - try match at specific point
1796 STATIC I32 /* 0 failure, 1 success */
1797 S_regtry(pTHX_ regexp *prog, char *startpos)
1805 PL_regindent = 0; /* XXXX Not good when matches are reenterable... */
1807 if ((prog->reganch & ROPT_EVAL_SEEN) && !PL_reg_eval_set) {
1810 PL_reg_eval_set = RS_init;
1812 PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %"IVdf"\n",
1813 (IV)(PL_stack_sp - PL_stack_base));
1815 SAVEI32(cxstack[cxstack_ix].blk_oldsp);
1816 cxstack[cxstack_ix].blk_oldsp = PL_stack_sp - PL_stack_base;
1817 /* Otherwise OP_NEXTSTATE will free whatever on stack now. */
1819 /* Apparently this is not needed, judging by wantarray. */
1820 /* SAVEI8(cxstack[cxstack_ix].blk_gimme);
1821 cxstack[cxstack_ix].blk_gimme = G_SCALAR; */
1824 /* Make $_ available to executed code. */
1825 if (PL_reg_sv != DEFSV) {
1826 /* SAVE_DEFSV does *not* suffice here for USE_THREADS */
1831 if (!(SvTYPE(PL_reg_sv) >= SVt_PVMG && SvMAGIC(PL_reg_sv)
1832 && (mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global)))) {
1833 /* prepare for quick setting of pos */
1834 sv_magic(PL_reg_sv, (SV*)0,
1835 PERL_MAGIC_regex_global, Nullch, 0);
1836 mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global);
1840 PL_reg_oldpos = mg->mg_len;
1841 SAVEDESTRUCTOR_X(restore_pos, 0);
1844 Newz(22,PL_reg_curpm, 1, PMOP);
1845 PL_reg_curpm->op_pmregexp = prog;
1846 PL_reg_oldcurpm = PL_curpm;
1847 PL_curpm = PL_reg_curpm;
1848 if (RX_MATCH_COPIED(prog)) {
1849 /* Here is a serious problem: we cannot rewrite subbeg,
1850 since it may be needed if this match fails. Thus
1851 $` inside (?{}) could fail... */
1852 PL_reg_oldsaved = prog->subbeg;
1853 PL_reg_oldsavedlen = prog->sublen;
1854 RX_MATCH_COPIED_off(prog);
1857 PL_reg_oldsaved = Nullch;
1858 prog->subbeg = PL_bostr;
1859 prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */
1861 prog->startp[0] = startpos - PL_bostr;
1862 PL_reginput = startpos;
1863 PL_regstartp = prog->startp;
1864 PL_regendp = prog->endp;
1865 PL_reglastparen = &prog->lastparen;
1866 prog->lastparen = 0;
1868 DEBUG_r(PL_reg_starttry = startpos);
1869 if (PL_reg_start_tmpl <= prog->nparens) {
1870 PL_reg_start_tmpl = prog->nparens*3/2 + 3;
1871 if(PL_reg_start_tmp)
1872 Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
1874 New(22,PL_reg_start_tmp, PL_reg_start_tmpl, char*);
1877 /* XXXX What this code is doing here?!!! There should be no need
1878 to do this again and again, PL_reglastparen should take care of
1881 /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code.
1882 * Actually, the code in regcppop() (which Ilya may be meaning by
1883 * PL_reglastparen), is not needed at all by the test suite
1884 * (op/regexp, op/pat, op/split), but that code is needed, oddly
1885 * enough, for building DynaLoader, or otherwise this
1886 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
1887 * will happen. Meanwhile, this code *is* needed for the
1888 * above-mentioned test suite tests to succeed. The common theme
1889 * on those tests seems to be returning null fields from matches.
1894 if (prog->nparens) {
1895 for (i = prog->nparens; i > *PL_reglastparen; i--) {
1902 if (regmatch(prog->program + 1)) {
1903 prog->endp[0] = PL_reginput - PL_bostr;
1906 REGCP_UNWIND(lastcp);
1910 #define RE_UNWIND_BRANCH 1
1911 #define RE_UNWIND_BRANCHJ 2
1915 typedef struct { /* XX: makes sense to enlarge it... */
1919 } re_unwind_generic_t;
1932 } re_unwind_branch_t;
1934 typedef union re_unwind_t {
1936 re_unwind_generic_t generic;
1937 re_unwind_branch_t branch;
1941 - regmatch - main matching routine
1943 * Conceptually the strategy is simple: check to see whether the current
1944 * node matches, call self recursively to see whether the rest matches,
1945 * and then act accordingly. In practice we make some effort to avoid
1946 * recursion, in particular by going through "ordinary" nodes (that don't
1947 * need to know whether the rest of the match failed) by a loop instead of
1950 /* [lwall] I've hoisted the register declarations to the outer block in order to
1951 * maybe save a little bit of pushing and popping on the stack. It also takes
1952 * advantage of machines that use a register save mask on subroutine entry.
1954 STATIC I32 /* 0 failure, 1 success */
1955 S_regmatch(pTHX_ regnode *prog)
1957 register regnode *scan; /* Current node. */
1958 regnode *next; /* Next node. */
1959 regnode *inner; /* Next node in internal branch. */
1960 register I32 nextchr; /* renamed nextchr - nextchar colides with
1961 function of same name */
1962 register I32 n; /* no or next */
1963 register I32 ln; /* len or last */
1964 register char *s; /* operand or save */
1965 register char *locinput = PL_reginput;
1966 register I32 c1, c2, paren; /* case fold search, parenth */
1967 int minmod = 0, sw = 0, logical = 0;
1969 I32 firstcp = PL_savestack_ix;
1970 register bool do_utf8 = DO_UTF8(PL_reg_sv);
1976 /* Note that nextchr is a byte even in UTF */
1977 nextchr = UCHARAT(locinput);
1979 while (scan != NULL) {
1980 #define sayNO_L (logical ? (logical = 0, sw = 0, goto cont) : sayNO)
1982 # define sayYES goto yes
1983 # define sayNO goto no
1984 # define sayYES_FINAL goto yes_final
1985 # define sayYES_LOUD goto yes_loud
1986 # define sayNO_FINAL goto no_final
1987 # define sayNO_SILENT goto do_no
1988 # define saySAME(x) if (x) goto yes; else goto no
1989 # define REPORT_CODE_OFF 24
1991 # define sayYES return 1
1992 # define sayNO return 0
1993 # define sayYES_FINAL return 1
1994 # define sayYES_LOUD return 1
1995 # define sayNO_FINAL return 0
1996 # define sayNO_SILENT return 0
1997 # define saySAME(x) return x
2000 SV *prop = sv_newmortal();
2001 int docolor = *PL_colors[0];
2002 int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */
2003 int l = (PL_regeol - locinput) > taill ? taill : (PL_regeol - locinput);
2004 /* The part of the string before starttry has one color
2005 (pref0_len chars), between starttry and current
2006 position another one (pref_len - pref0_len chars),
2007 after the current position the third one.
2008 We assume that pref0_len <= pref_len, otherwise we
2009 decrease pref0_len. */
2010 int pref_len = (locinput - PL_bostr) > (5 + taill) - l
2011 ? (5 + taill) - l : locinput - PL_bostr;
2014 while (UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len)))
2016 pref0_len = pref_len - (locinput - PL_reg_starttry);
2017 if (l + pref_len < (5 + taill) && l < PL_regeol - locinput)
2018 l = ( PL_regeol - locinput > (5 + taill) - pref_len
2019 ? (5 + taill) - pref_len : PL_regeol - locinput);
2020 while (UTF8_IS_CONTINUATION(*(U8*)(locinput + l)))
2024 if (pref0_len > pref_len)
2025 pref0_len = pref_len;
2026 regprop(prop, scan);
2027 PerlIO_printf(Perl_debug_log,
2028 "%4"IVdf" <%s%.*s%s%s%.*s%s%s%s%.*s%s>%*s|%3"IVdf":%*s%s\n",
2029 (IV)(locinput - PL_bostr),
2030 PL_colors[4], pref0_len,
2031 locinput - pref_len, PL_colors[5],
2032 PL_colors[2], pref_len - pref0_len,
2033 locinput - pref_len + pref0_len, PL_colors[3],
2034 (docolor ? "" : "> <"),
2035 PL_colors[0], l, locinput, PL_colors[1],
2036 15 - l - pref_len + 1,
2038 (IV)(scan - PL_regprogram), PL_regindent*2, "",
2042 next = scan + NEXT_OFF(scan);
2048 if (locinput == PL_bostr || (PL_multiline &&
2049 (nextchr || locinput < PL_regeol) && locinput[-1] == '\n') )
2051 /* regtill = regbol; */
2056 if (locinput == PL_bostr ||
2057 ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n'))
2063 if (locinput == PL_bostr)
2067 if (locinput == PL_reg_ganch)
2077 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2082 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2084 if (PL_regeol - locinput > 1)
2088 if (PL_regeol != locinput)
2092 if (!nextchr && locinput >= PL_regeol)
2094 nextchr = UCHARAT(++locinput);
2097 if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n')
2100 locinput += PL_utf8skip[nextchr];
2101 if (locinput > PL_regeol)
2103 nextchr = UCHARAT(locinput);
2106 nextchr = UCHARAT(++locinput);
2111 if (do_utf8 != (UTF!=0)) {
2119 if (*((U8*)s) != utf8_to_uvchr((U8*)l, &len))
2128 if (*((U8*)l) != utf8_to_uvchr((U8*)s, &len))
2134 nextchr = UCHARAT(locinput);
2137 /* Inline the first character, for speed. */
2138 if (UCHARAT(s) != nextchr)
2140 if (PL_regeol - locinput < ln)
2142 if (ln > 1 && memNE(s, locinput, ln))
2145 nextchr = UCHARAT(locinput);
2148 PL_reg_flags |= RF_tainted;
2158 c1 = OP(scan) == EXACTF;
2160 if (l >= PL_regeol) {
2163 if ((UTF ? utf8n_to_uvchr((U8*)s, e - s, 0, 0) : *((U8*)s)) !=
2164 (c1 ? toLOWER_utf8((U8*)l) : toLOWER_LC_utf8((U8*)l)))
2166 s += UTF ? UTF8SKIP(s) : 1;
2170 nextchr = UCHARAT(locinput);
2174 /* Inline the first character, for speed. */
2175 if (UCHARAT(s) != nextchr &&
2176 UCHARAT(s) != ((OP(scan) == EXACTF)
2177 ? PL_fold : PL_fold_locale)[nextchr])
2179 if (PL_regeol - locinput < ln)
2181 if (ln > 1 && (OP(scan) == EXACTF
2182 ? ibcmp(s, locinput, ln)
2183 : ibcmp_locale(s, locinput, ln)))
2186 nextchr = UCHARAT(locinput);
2190 if (!reginclass(scan, (U8*)locinput, do_utf8))
2192 if (locinput >= PL_regeol)
2194 locinput += PL_utf8skip[nextchr];
2195 nextchr = UCHARAT(locinput);
2199 nextchr = UCHARAT(locinput);
2200 if (!reginclass(scan, (U8*)locinput, do_utf8))
2202 if (!nextchr && locinput >= PL_regeol)
2204 nextchr = UCHARAT(++locinput);
2208 PL_reg_flags |= RF_tainted;
2214 if (!(OP(scan) == ALNUM
2215 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2216 : isALNUM_LC_utf8((U8*)locinput)))
2220 locinput += PL_utf8skip[nextchr];
2221 nextchr = UCHARAT(locinput);
2224 if (!(OP(scan) == ALNUM
2225 ? isALNUM(nextchr) : isALNUM_LC(nextchr)))
2227 nextchr = UCHARAT(++locinput);
2230 PL_reg_flags |= RF_tainted;
2233 if (!nextchr && locinput >= PL_regeol)
2236 LOAD_UTF8_CHARCLASS(alnum,"a");
2237 if (OP(scan) == NALNUM
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) == NALNUM
2248 ? isALNUM(nextchr) : isALNUM_LC(nextchr))
2250 nextchr = UCHARAT(++locinput);
2254 PL_reg_flags |= RF_tainted;
2258 /* was last char in word? */
2260 if (locinput == PL_bostr)
2263 U8 *r = reghop((U8*)locinput, -1);
2265 ln = utf8n_to_uvchr(r, s - (char*)r, 0, 0);
2267 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2268 ln = isALNUM_uni(ln);
2269 LOAD_UTF8_CHARCLASS(alnum,"a");
2270 n = swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8);
2273 ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(ln));
2274 n = isALNUM_LC_utf8((U8*)locinput);
2278 ln = (locinput != PL_bostr) ?
2279 UCHARAT(locinput - 1) : '\n';
2280 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2282 n = isALNUM(nextchr);
2285 ln = isALNUM_LC(ln);
2286 n = isALNUM_LC(nextchr);
2289 if (((!ln) == (!n)) == (OP(scan) == BOUND ||
2290 OP(scan) == BOUNDL))
2294 PL_reg_flags |= RF_tainted;
2300 if (UTF8_IS_CONTINUED(nextchr)) {
2301 LOAD_UTF8_CHARCLASS(space," ");
2302 if (!(OP(scan) == SPACE
2303 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2304 : isSPACE_LC_utf8((U8*)locinput)))
2308 locinput += PL_utf8skip[nextchr];
2309 nextchr = UCHARAT(locinput);
2312 if (!(OP(scan) == SPACE
2313 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2315 nextchr = UCHARAT(++locinput);
2318 if (!(OP(scan) == SPACE
2319 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2321 nextchr = UCHARAT(++locinput);
2325 PL_reg_flags |= RF_tainted;
2328 if (!nextchr && locinput >= PL_regeol)
2331 LOAD_UTF8_CHARCLASS(space," ");
2332 if (OP(scan) == NSPACE
2333 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2334 : isSPACE_LC_utf8((U8*)locinput))
2338 locinput += PL_utf8skip[nextchr];
2339 nextchr = UCHARAT(locinput);
2342 if (OP(scan) == NSPACE
2343 ? isSPACE(nextchr) : isSPACE_LC(nextchr))
2345 nextchr = UCHARAT(++locinput);
2348 PL_reg_flags |= RF_tainted;
2354 LOAD_UTF8_CHARCLASS(digit,"0");
2355 if (!(OP(scan) == DIGIT
2356 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2357 : isDIGIT_LC_utf8((U8*)locinput)))
2361 locinput += PL_utf8skip[nextchr];
2362 nextchr = UCHARAT(locinput);
2365 if (!(OP(scan) == DIGIT
2366 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr)))
2368 nextchr = UCHARAT(++locinput);
2371 PL_reg_flags |= RF_tainted;
2374 if (!nextchr && locinput >= PL_regeol)
2377 LOAD_UTF8_CHARCLASS(digit,"0");
2378 if (OP(scan) == NDIGIT
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) == NDIGIT
2389 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr))
2391 nextchr = UCHARAT(++locinput);
2394 LOAD_UTF8_CHARCLASS(mark,"~");
2395 if (locinput >= PL_regeol ||
2396 swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2398 locinput += PL_utf8skip[nextchr];
2399 while (locinput < PL_regeol &&
2400 swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2401 locinput += UTF8SKIP(locinput);
2402 if (locinput > PL_regeol)
2404 nextchr = UCHARAT(locinput);
2407 PL_reg_flags |= RF_tainted;
2411 n = ARG(scan); /* which paren pair */
2412 ln = PL_regstartp[n];
2413 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2414 if (*PL_reglastparen < n || ln == -1)
2415 sayNO; /* Do not match unless seen CLOSEn. */
2416 if (ln == PL_regendp[n])
2420 if (do_utf8 && OP(scan) != REF) { /* REF can do byte comparison */
2422 char *e = PL_bostr + PL_regendp[n];
2424 * Note that we can't do the "other character" lookup trick as
2425 * in the 8-bit case (no pun intended) because in Unicode we
2426 * have to map both upper and title case to lower case.
2428 if (OP(scan) == REFF) {
2432 if (toLOWER_utf8((U8*)s) != toLOWER_utf8((U8*)l))
2442 if (toLOWER_LC_utf8((U8*)s) != toLOWER_LC_utf8((U8*)l))
2449 nextchr = UCHARAT(locinput);
2453 /* Inline the first character, for speed. */
2454 if (UCHARAT(s) != nextchr &&
2456 (UCHARAT(s) != ((OP(scan) == REFF
2457 ? PL_fold : PL_fold_locale)[nextchr]))))
2459 ln = PL_regendp[n] - ln;
2460 if (locinput + ln > PL_regeol)
2462 if (ln > 1 && (OP(scan) == REF
2463 ? memNE(s, locinput, ln)
2465 ? ibcmp(s, locinput, ln)
2466 : ibcmp_locale(s, locinput, ln))))
2469 nextchr = UCHARAT(locinput);
2480 OP_4tree *oop = PL_op;
2481 COP *ocurcop = PL_curcop;
2482 SV **ocurpad = PL_curpad;
2486 PL_op = (OP_4tree*)PL_regdata->data[n];
2487 DEBUG_r( PerlIO_printf(Perl_debug_log, " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) );
2488 PL_curpad = AvARRAY((AV*)PL_regdata->data[n + 2]);
2489 PL_regendp[0] = PL_reg_magic->mg_len = locinput - PL_bostr;
2491 CALLRUNOPS(aTHX); /* Scalar context. */
2497 PL_curpad = ocurpad;
2498 PL_curcop = ocurcop;
2500 if (logical == 2) { /* Postponed subexpression. */
2502 MAGIC *mg = Null(MAGIC*);
2504 CHECKPOINT cp, lastcp;
2506 if(SvROK(ret) || SvRMAGICAL(ret)) {
2507 SV *sv = SvROK(ret) ? SvRV(ret) : ret;
2510 mg = mg_find(sv, PERL_MAGIC_qr);
2513 re = (regexp *)mg->mg_obj;
2514 (void)ReREFCNT_inc(re);
2518 char *t = SvPV(ret, len);
2520 char *oprecomp = PL_regprecomp;
2521 I32 osize = PL_regsize;
2522 I32 onpar = PL_regnpar;
2525 re = CALLREGCOMP(aTHX_ t, t + len, &pm);
2527 & (SVs_TEMP | SVs_PADTMP | SVf_READONLY)))
2528 sv_magic(ret,(SV*)ReREFCNT_inc(re),
2530 PL_regprecomp = oprecomp;
2535 PerlIO_printf(Perl_debug_log,
2536 "Entering embedded `%s%.60s%s%s'\n",
2540 (strlen(re->precomp) > 60 ? "..." : ""))
2543 state.prev = PL_reg_call_cc;
2544 state.cc = PL_regcc;
2545 state.re = PL_reg_re;
2549 cp = regcppush(0); /* Save *all* the positions. */
2552 state.ss = PL_savestack_ix;
2553 *PL_reglastparen = 0;
2554 PL_reg_call_cc = &state;
2555 PL_reginput = locinput;
2557 /* XXXX This is too dramatic a measure... */
2560 if (regmatch(re->program + 1)) {
2561 /* Even though we succeeded, we need to restore
2562 global variables, since we may be wrapped inside
2563 SUSPEND, thus the match may be not finished yet. */
2565 /* XXXX Do this only if SUSPENDed? */
2566 PL_reg_call_cc = state.prev;
2567 PL_regcc = state.cc;
2568 PL_reg_re = state.re;
2569 cache_re(PL_reg_re);
2571 /* XXXX This is too dramatic a measure... */
2574 /* These are needed even if not SUSPEND. */
2580 REGCP_UNWIND(lastcp);
2582 PL_reg_call_cc = state.prev;
2583 PL_regcc = state.cc;
2584 PL_reg_re = state.re;
2585 cache_re(PL_reg_re);
2587 /* XXXX This is too dramatic a measure... */
2596 sv_setsv(save_scalar(PL_replgv), ret);
2600 n = ARG(scan); /* which paren pair */
2601 PL_reg_start_tmp[n] = locinput;
2606 n = ARG(scan); /* which paren pair */
2607 PL_regstartp[n] = PL_reg_start_tmp[n] - PL_bostr;
2608 PL_regendp[n] = locinput - PL_bostr;
2609 if (n > *PL_reglastparen)
2610 *PL_reglastparen = n;
2613 n = ARG(scan); /* which paren pair */
2614 sw = (*PL_reglastparen >= n && PL_regendp[n] != -1);
2617 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2619 next = NEXTOPER(NEXTOPER(scan));
2621 next = scan + ARG(scan);
2622 if (OP(next) == IFTHEN) /* Fake one. */
2623 next = NEXTOPER(NEXTOPER(next));
2627 logical = scan->flags;
2629 /*******************************************************************
2630 PL_regcc contains infoblock about the innermost (...)* loop, and
2631 a pointer to the next outer infoblock.
2633 Here is how Y(A)*Z is processed (if it is compiled into CURLYX/WHILEM):
2635 1) After matching X, regnode for CURLYX is processed;
2637 2) This regnode creates infoblock on the stack, and calls
2638 regmatch() recursively with the starting point at WHILEM node;
2640 3) Each hit of WHILEM node tries to match A and Z (in the order
2641 depending on the current iteration, min/max of {min,max} and
2642 greediness). The information about where are nodes for "A"
2643 and "Z" is read from the infoblock, as is info on how many times "A"
2644 was already matched, and greediness.
2646 4) After A matches, the same WHILEM node is hit again.
2648 5) Each time WHILEM is hit, PL_regcc is the infoblock created by CURLYX
2649 of the same pair. Thus when WHILEM tries to match Z, it temporarily
2650 resets PL_regcc, since this Y(A)*Z can be a part of some other loop:
2651 as in (Y(A)*Z)*. If Z matches, the automaton will hit the WHILEM node
2652 of the external loop.
2654 Currently present infoblocks form a tree with a stem formed by PL_curcc
2655 and whatever it mentions via ->next, and additional attached trees
2656 corresponding to temporarily unset infoblocks as in "5" above.
2658 In the following picture infoblocks for outer loop of
2659 (Y(A)*?Z)*?T are denoted O, for inner I. NULL starting block
2660 is denoted by x. The matched string is YAAZYAZT. Temporarily postponed
2661 infoblocks are drawn below the "reset" infoblock.
2663 In fact in the picture below we do not show failed matches for Z and T
2664 by WHILEM blocks. [We illustrate minimal matches, since for them it is
2665 more obvious *why* one needs to *temporary* unset infoblocks.]
2667 Matched REx position InfoBlocks Comment
2671 Y A)*?Z)*?T x <- O <- I
2672 YA )*?Z)*?T x <- O <- I
2673 YA A)*?Z)*?T x <- O <- I
2674 YAA )*?Z)*?T x <- O <- I
2675 YAA Z)*?T x <- O # Temporary unset I
2678 YAAZ Y(A)*?Z)*?T x <- O
2681 YAAZY (A)*?Z)*?T x <- O
2684 YAAZY A)*?Z)*?T x <- O <- I
2687 YAAZYA )*?Z)*?T x <- O <- I
2690 YAAZYA Z)*?T x <- O # Temporary unset I
2696 YAAZYAZ T x # Temporary unset O
2703 *******************************************************************/
2706 CHECKPOINT cp = PL_savestack_ix;
2707 /* No need to save/restore up to this paren */
2708 I32 parenfloor = scan->flags;
2710 if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */
2712 cc.oldcc = PL_regcc;
2714 /* XXXX Probably it is better to teach regpush to support
2715 parenfloor > PL_regsize... */
2716 if (parenfloor > *PL_reglastparen)
2717 parenfloor = *PL_reglastparen; /* Pessimization... */
2718 cc.parenfloor = parenfloor;
2720 cc.min = ARG1(scan);
2721 cc.max = ARG2(scan);
2722 cc.scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
2726 PL_reginput = locinput;
2727 n = regmatch(PREVOPER(next)); /* start on the WHILEM */
2729 PL_regcc = cc.oldcc;
2735 * This is really hard to understand, because after we match
2736 * what we're trying to match, we must make sure the rest of
2737 * the REx is going to match for sure, and to do that we have
2738 * to go back UP the parse tree by recursing ever deeper. And
2739 * if it fails, we have to reset our parent's current state
2740 * that we can try again after backing off.
2743 CHECKPOINT cp, lastcp;
2744 CURCUR* cc = PL_regcc;
2745 char *lastloc = cc->lastloc; /* Detection of 0-len. */
2747 n = cc->cur + 1; /* how many we know we matched */
2748 PL_reginput = locinput;
2751 PerlIO_printf(Perl_debug_log,
2752 "%*s %ld out of %ld..%ld cc=%lx\n",
2753 REPORT_CODE_OFF+PL_regindent*2, "",
2754 (long)n, (long)cc->min,
2755 (long)cc->max, (long)cc)
2758 /* If degenerate scan matches "", assume scan done. */
2760 if (locinput == cc->lastloc && n >= cc->min) {
2761 PL_regcc = cc->oldcc;
2765 PerlIO_printf(Perl_debug_log,
2766 "%*s empty match detected, try continuation...\n",
2767 REPORT_CODE_OFF+PL_regindent*2, "")
2769 if (regmatch(cc->next))
2777 /* First just match a string of min scans. */
2781 cc->lastloc = locinput;
2782 if (regmatch(cc->scan))
2785 cc->lastloc = lastloc;
2790 /* Check whether we already were at this position.
2791 Postpone detection until we know the match is not
2792 *that* much linear. */
2793 if (!PL_reg_maxiter) {
2794 PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4);
2795 PL_reg_leftiter = PL_reg_maxiter;
2797 if (PL_reg_leftiter-- == 0) {
2798 I32 size = (PL_reg_maxiter + 7)/8;
2799 if (PL_reg_poscache) {
2800 if (PL_reg_poscache_size < size) {
2801 Renew(PL_reg_poscache, size, char);
2802 PL_reg_poscache_size = size;
2804 Zero(PL_reg_poscache, size, char);
2807 PL_reg_poscache_size = size;
2808 Newz(29, PL_reg_poscache, size, char);
2811 PerlIO_printf(Perl_debug_log,
2812 "%sDetected a super-linear match, switching on caching%s...\n",
2813 PL_colors[4], PL_colors[5])
2816 if (PL_reg_leftiter < 0) {
2817 I32 o = locinput - PL_bostr, b;
2819 o = (scan->flags & 0xf) - 1 + o * (scan->flags>>4);
2822 if (PL_reg_poscache[o] & (1<<b)) {
2824 PerlIO_printf(Perl_debug_log,
2825 "%*s already tried at this position...\n",
2826 REPORT_CODE_OFF+PL_regindent*2, "")
2830 PL_reg_poscache[o] |= (1<<b);
2834 /* Prefer next over scan for minimal matching. */
2837 PL_regcc = cc->oldcc;
2840 cp = regcppush(cc->parenfloor);
2842 if (regmatch(cc->next)) {
2844 sayYES; /* All done. */
2846 REGCP_UNWIND(lastcp);
2852 if (n >= cc->max) { /* Maximum greed exceeded? */
2853 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
2854 && !(PL_reg_flags & RF_warned)) {
2855 PL_reg_flags |= RF_warned;
2856 Perl_warner(aTHX_ WARN_REGEXP, "%s limit (%d) exceeded",
2857 "Complex regular subexpression recursion",
2864 PerlIO_printf(Perl_debug_log,
2865 "%*s trying longer...\n",
2866 REPORT_CODE_OFF+PL_regindent*2, "")
2868 /* Try scanning more and see if it helps. */
2869 PL_reginput = locinput;
2871 cc->lastloc = locinput;
2872 cp = regcppush(cc->parenfloor);
2874 if (regmatch(cc->scan)) {
2878 REGCP_UNWIND(lastcp);
2881 cc->lastloc = lastloc;
2885 /* Prefer scan over next for maximal matching. */
2887 if (n < cc->max) { /* More greed allowed? */
2888 cp = regcppush(cc->parenfloor);
2890 cc->lastloc = locinput;
2892 if (regmatch(cc->scan)) {
2896 REGCP_UNWIND(lastcp);
2897 regcppop(); /* Restore some previous $<digit>s? */
2898 PL_reginput = locinput;
2900 PerlIO_printf(Perl_debug_log,
2901 "%*s failed, try continuation...\n",
2902 REPORT_CODE_OFF+PL_regindent*2, "")
2905 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
2906 && !(PL_reg_flags & RF_warned)) {
2907 PL_reg_flags |= RF_warned;
2908 Perl_warner(aTHX_ WARN_REGEXP, "%s limit (%d) exceeded",
2909 "Complex regular subexpression recursion",
2913 /* Failed deeper matches of scan, so see if this one works. */
2914 PL_regcc = cc->oldcc;
2917 if (regmatch(cc->next))
2923 cc->lastloc = lastloc;
2928 next = scan + ARG(scan);
2931 inner = NEXTOPER(NEXTOPER(scan));
2934 inner = NEXTOPER(scan);
2939 if (OP(next) != c1) /* No choice. */
2940 next = inner; /* Avoid recursion. */
2942 I32 lastparen = *PL_reglastparen;
2944 re_unwind_branch_t *uw;
2946 /* Put unwinding data on stack */
2947 unwind1 = SSNEWt(1,re_unwind_branch_t);
2948 uw = SSPTRt(unwind1,re_unwind_branch_t);
2951 uw->type = ((c1 == BRANCH)
2953 : RE_UNWIND_BRANCHJ);
2954 uw->lastparen = lastparen;
2956 uw->locinput = locinput;
2957 uw->nextchr = nextchr;
2959 uw->regindent = ++PL_regindent;
2962 REGCP_SET(uw->lastcp);
2964 /* Now go into the first branch */
2977 /* We suppose that the next guy does not need
2978 backtracking: in particular, it is of constant length,
2979 and has no parenths to influence future backrefs. */
2980 ln = ARG1(scan); /* min to match */
2981 n = ARG2(scan); /* max to match */
2982 paren = scan->flags;
2984 if (paren > PL_regsize)
2986 if (paren > *PL_reglastparen)
2987 *PL_reglastparen = paren;
2989 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
2991 scan += NEXT_OFF(scan); /* Skip former OPEN. */
2992 PL_reginput = locinput;
2995 if (ln && regrepeat_hard(scan, ln, &l) < ln)
2997 if (ln && l == 0 && n >= ln
2998 /* In fact, this is tricky. If paren, then the
2999 fact that we did/didnot match may influence
3000 future execution. */
3001 && !(paren && ln == 0))
3003 locinput = PL_reginput;
3004 if (PL_regkind[(U8)OP(next)] == EXACT) {
3005 c1 = (U8)*STRING(next);
3006 if (OP(next) == EXACTF)
3008 else if (OP(next) == EXACTFL)
3009 c2 = PL_fold_locale[c1];
3016 /* This may be improved if l == 0. */
3017 while (n >= ln || (n == REG_INFTY && ln > 0 && l)) { /* ln overflow ? */
3018 /* If it could work, try it. */
3020 UCHARAT(PL_reginput) == c1 ||
3021 UCHARAT(PL_reginput) == c2)
3025 PL_regstartp[paren] =
3026 HOPc(PL_reginput, -l) - PL_bostr;
3027 PL_regendp[paren] = PL_reginput - PL_bostr;
3030 PL_regendp[paren] = -1;
3034 REGCP_UNWIND(lastcp);
3036 /* Couldn't or didn't -- move forward. */
3037 PL_reginput = locinput;
3038 if (regrepeat_hard(scan, 1, &l)) {
3040 locinput = PL_reginput;
3047 n = regrepeat_hard(scan, n, &l);
3048 if (n != 0 && l == 0
3049 /* In fact, this is tricky. If paren, then the
3050 fact that we did/didnot match may influence
3051 future execution. */
3052 && !(paren && ln == 0))
3054 locinput = PL_reginput;
3056 PerlIO_printf(Perl_debug_log,
3057 "%*s matched %"IVdf" times, len=%"IVdf"...\n",
3058 (int)(REPORT_CODE_OFF+PL_regindent*2), "",
3062 if (PL_regkind[(U8)OP(next)] == EXACT) {
3063 c1 = (U8)*STRING(next);
3064 if (OP(next) == EXACTF)
3066 else if (OP(next) == EXACTFL)
3067 c2 = PL_fold_locale[c1];
3076 /* If it could work, try it. */
3078 UCHARAT(PL_reginput) == c1 ||
3079 UCHARAT(PL_reginput) == c2)
3082 PerlIO_printf(Perl_debug_log,
3083 "%*s trying tail with n=%"IVdf"...\n",
3084 (int)(REPORT_CODE_OFF+PL_regindent*2), "", (IV)n)
3088 PL_regstartp[paren] = HOPc(PL_reginput, -l) - PL_bostr;
3089 PL_regendp[paren] = PL_reginput - PL_bostr;
3092 PL_regendp[paren] = -1;
3096 REGCP_UNWIND(lastcp);
3098 /* Couldn't or didn't -- back up. */
3100 locinput = HOPc(locinput, -l);
3101 PL_reginput = locinput;
3108 paren = scan->flags; /* Which paren to set */
3109 if (paren > PL_regsize)
3111 if (paren > *PL_reglastparen)
3112 *PL_reglastparen = paren;
3113 ln = ARG1(scan); /* min to match */
3114 n = ARG2(scan); /* max to match */
3115 scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE);
3119 ln = ARG1(scan); /* min to match */
3120 n = ARG2(scan); /* max to match */
3121 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3126 scan = NEXTOPER(scan);
3132 scan = NEXTOPER(scan);
3136 * Lookahead to avoid useless match attempts
3137 * when we know what character comes next.
3139 if (PL_regkind[(U8)OP(next)] == EXACT) {
3140 U8 *s = (U8*)STRING(next);
3143 if (OP(next) == EXACTF)
3145 else if (OP(next) == EXACTFL)
3146 c2 = PL_fold_locale[c1];
3149 if (OP(next) == EXACTF) {
3150 c1 = to_utf8_lower(s);
3151 c2 = to_utf8_upper(s);
3154 c2 = c1 = utf8_to_uvchr(s, NULL);
3160 PL_reginput = locinput;
3164 if (ln && regrepeat(scan, ln) < ln)
3166 locinput = PL_reginput;
3169 char *e; /* Should not check after this */
3170 char *old = locinput;
3172 if (n == REG_INFTY) {
3175 while (UTF8_IS_CONTINUATION(*(U8*)e))
3181 m >0 && e + UTF8SKIP(e) <= PL_regeol; m--)
3185 e = locinput + n - ln;
3191 /* Find place 'next' could work */
3194 while (locinput <= e && *locinput != c1)
3197 while (locinput <= e
3202 count = locinput - old;
3209 utf8_to_uvchr((U8*)locinput, &len) != c1;
3214 for (count = 0; locinput <= e; count++) {
3215 UV c = utf8_to_uvchr((U8*)locinput, &len);
3216 if (c == c1 || c == c2)
3224 /* PL_reginput == old now */
3225 if (locinput != old) {
3226 ln = 1; /* Did some */
3227 if (regrepeat(scan, count) < count)
3230 /* PL_reginput == locinput now */
3231 TRYPAREN(paren, ln, locinput);
3232 PL_reginput = locinput; /* Could be reset... */
3233 REGCP_UNWIND(lastcp);
3234 /* Couldn't or didn't -- move forward. */
3237 locinput += UTF8SKIP(locinput);
3243 while (n >= ln || (n == REG_INFTY && ln > 0)) { /* ln overflow ? */
3247 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3249 c = UCHARAT(PL_reginput);
3250 /* If it could work, try it. */
3251 if (c == c1 || c == c2)
3253 TRYPAREN(paren, n, PL_reginput);
3254 REGCP_UNWIND(lastcp);
3257 /* If it could work, try it. */
3258 else if (c1 == -1000)
3260 TRYPAREN(paren, n, PL_reginput);
3261 REGCP_UNWIND(lastcp);
3263 /* Couldn't or didn't -- move forward. */
3264 PL_reginput = locinput;
3265 if (regrepeat(scan, 1)) {
3267 locinput = PL_reginput;
3275 n = regrepeat(scan, n);
3276 locinput = PL_reginput;
3277 if (ln < n && PL_regkind[(U8)OP(next)] == EOL &&
3278 (!PL_multiline || OP(next) == SEOL || OP(next) == EOS)) {
3279 ln = n; /* why back off? */
3280 /* ...because $ and \Z can match before *and* after
3281 newline at the end. Consider "\n\n" =~ /\n+\Z\n/.
3282 We should back off by one in this case. */
3283 if (UCHARAT(PL_reginput - 1) == '\n' && OP(next) != EOS)
3292 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3294 c = UCHARAT(PL_reginput);
3296 /* If it could work, try it. */
3297 if (c1 == -1000 || c == c1 || c == c2)
3299 TRYPAREN(paren, n, PL_reginput);
3300 REGCP_UNWIND(lastcp);
3302 /* Couldn't or didn't -- back up. */
3304 PL_reginput = locinput = HOPc(locinput, -1);
3312 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3314 c = UCHARAT(PL_reginput);
3316 /* If it could work, try it. */
3317 if (c1 == -1000 || c == c1 || c == c2)
3319 TRYPAREN(paren, n, PL_reginput);
3320 REGCP_UNWIND(lastcp);
3322 /* Couldn't or didn't -- back up. */
3324 PL_reginput = locinput = HOPc(locinput, -1);
3331 if (PL_reg_call_cc) {
3332 re_cc_state *cur_call_cc = PL_reg_call_cc;
3333 CURCUR *cctmp = PL_regcc;
3334 regexp *re = PL_reg_re;
3335 CHECKPOINT cp, lastcp;
3337 cp = regcppush(0); /* Save *all* the positions. */
3339 regcp_set_to(PL_reg_call_cc->ss); /* Restore parens of
3341 PL_reginput = locinput; /* Make position available to
3343 cache_re(PL_reg_call_cc->re);
3344 PL_regcc = PL_reg_call_cc->cc;
3345 PL_reg_call_cc = PL_reg_call_cc->prev;
3346 if (regmatch(cur_call_cc->node)) {
3347 PL_reg_call_cc = cur_call_cc;
3351 REGCP_UNWIND(lastcp);
3353 PL_reg_call_cc = cur_call_cc;
3359 PerlIO_printf(Perl_debug_log,
3360 "%*s continuation failed...\n",
3361 REPORT_CODE_OFF+PL_regindent*2, "")
3365 if (locinput < PL_regtill) {
3366 DEBUG_r(PerlIO_printf(Perl_debug_log,
3367 "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n",
3369 (long)(locinput - PL_reg_starttry),
3370 (long)(PL_regtill - PL_reg_starttry),
3372 sayNO_FINAL; /* Cannot match: too short. */
3374 PL_reginput = locinput; /* put where regtry can find it */
3375 sayYES_FINAL; /* Success! */
3377 PL_reginput = locinput; /* put where regtry can find it */
3378 sayYES_LOUD; /* Success! */
3381 PL_reginput = locinput;
3386 s = HOPBACKc(locinput, scan->flags);
3392 PL_reginput = locinput;
3397 s = HOPBACKc(locinput, scan->flags);
3403 PL_reginput = locinput;
3406 inner = NEXTOPER(NEXTOPER(scan));
3407 if (regmatch(inner) != n) {
3422 if (OP(scan) == SUSPEND) {
3423 locinput = PL_reginput;
3424 nextchr = UCHARAT(locinput);
3429 next = scan + ARG(scan);
3434 PerlIO_printf(Perl_error_log, "%"UVxf" %d\n",
3435 PTR2UV(scan), OP(scan));
3436 Perl_croak(aTHX_ "regexp memory corruption");
3443 * We get here only if there's trouble -- normally "case END" is
3444 * the terminating point.
3446 Perl_croak(aTHX_ "corrupted regexp pointers");
3452 PerlIO_printf(Perl_debug_log,
3453 "%*s %scould match...%s\n",
3454 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4],PL_colors[5])
3458 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n",
3459 PL_colors[4],PL_colors[5]));
3465 #if 0 /* Breaks $^R */
3473 PerlIO_printf(Perl_debug_log,
3474 "%*s %sfailed...%s\n",
3475 REPORT_CODE_OFF+PL_regindent*2, "",PL_colors[4],PL_colors[5])
3481 re_unwind_t *uw = SSPTRt(unwind,re_unwind_t);
3484 case RE_UNWIND_BRANCH:
3485 case RE_UNWIND_BRANCHJ:
3487 re_unwind_branch_t *uwb = &(uw->branch);
3488 I32 lastparen = uwb->lastparen;
3490 REGCP_UNWIND(uwb->lastcp);
3491 for (n = *PL_reglastparen; n > lastparen; n--)
3493 *PL_reglastparen = n;
3494 scan = next = uwb->next;
3496 OP(scan) != (uwb->type == RE_UNWIND_BRANCH
3497 ? BRANCH : BRANCHJ) ) { /* Failure */
3504 /* Have more choice yet. Reuse the same uwb. */
3506 if ((n = (uwb->type == RE_UNWIND_BRANCH
3507 ? NEXT_OFF(next) : ARG(next))))
3510 next = NULL; /* XXXX Needn't unwinding in this case... */
3512 next = NEXTOPER(scan);
3513 if (uwb->type == RE_UNWIND_BRANCHJ)
3514 next = NEXTOPER(next);
3515 locinput = uwb->locinput;
3516 nextchr = uwb->nextchr;
3518 PL_regindent = uwb->regindent;
3525 Perl_croak(aTHX_ "regexp unwind memory corruption");
3536 - regrepeat - repeatedly match something simple, report how many
3539 * [This routine now assumes that it will only match on things of length 1.
3540 * That was true before, but now we assume scan - reginput is the count,
3541 * rather than incrementing count on every character. [Er, except utf8.]]
3544 S_regrepeat(pTHX_ regnode *p, I32 max)
3546 register char *scan;
3548 register char *loceol = PL_regeol;
3549 register I32 hardcount = 0;
3550 register bool do_utf8 = DO_UTF8(PL_reg_sv);
3553 if (max != REG_INFTY && max < loceol - scan)
3554 loceol = scan + max;
3559 while (scan < loceol && hardcount < max && *scan != '\n') {
3560 scan += UTF8SKIP(scan);
3564 while (scan < loceol && *scan != '\n')
3571 case EXACT: /* length of string is 1 */
3573 while (scan < loceol && UCHARAT(scan) == c)
3576 case EXACTF: /* length of string is 1 */
3578 while (scan < loceol &&
3579 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold[c]))
3582 case EXACTFL: /* length of string is 1 */
3583 PL_reg_flags |= RF_tainted;
3585 while (scan < loceol &&
3586 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold_locale[c]))
3592 while (hardcount < max && scan < loceol &&
3593 reginclass(p, (U8*)scan, do_utf8)) {
3594 scan += UTF8SKIP(scan);
3598 while (scan < loceol && reginclass(p, (U8*)scan, do_utf8))
3605 LOAD_UTF8_CHARCLASS(alnum,"a");
3606 while (hardcount < max && scan < loceol &&
3607 swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
3608 scan += UTF8SKIP(scan);
3612 while (scan < loceol && isALNUM(*scan))
3617 PL_reg_flags |= RF_tainted;
3620 while (hardcount < max && scan < loceol &&
3621 isALNUM_LC_utf8((U8*)scan)) {
3622 scan += UTF8SKIP(scan);
3626 while (scan < loceol && isALNUM_LC(*scan))
3633 LOAD_UTF8_CHARCLASS(alnum,"a");
3634 while (hardcount < max && scan < loceol &&
3635 !swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
3636 scan += UTF8SKIP(scan);
3640 while (scan < loceol && !isALNUM(*scan))
3645 PL_reg_flags |= RF_tainted;
3648 while (hardcount < max && scan < loceol &&
3649 !isALNUM_LC_utf8((U8*)scan)) {
3650 scan += UTF8SKIP(scan);
3654 while (scan < loceol && !isALNUM_LC(*scan))
3661 LOAD_UTF8_CHARCLASS(space," ");
3662 while (hardcount < max && scan < loceol &&
3664 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
3665 scan += UTF8SKIP(scan);
3669 while (scan < loceol && isSPACE(*scan))
3674 PL_reg_flags |= RF_tainted;
3677 while (hardcount < max && scan < loceol &&
3678 (*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
3679 scan += UTF8SKIP(scan);
3683 while (scan < loceol && isSPACE_LC(*scan))
3690 LOAD_UTF8_CHARCLASS(space," ");
3691 while (hardcount < max && scan < loceol &&
3693 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
3694 scan += UTF8SKIP(scan);
3698 while (scan < loceol && !isSPACE(*scan))
3703 PL_reg_flags |= RF_tainted;
3706 while (hardcount < max && scan < loceol &&
3707 !(*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
3708 scan += UTF8SKIP(scan);
3712 while (scan < loceol && !isSPACE_LC(*scan))
3719 LOAD_UTF8_CHARCLASS(digit,"0");
3720 while (hardcount < max && scan < loceol &&
3721 swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
3722 scan += UTF8SKIP(scan);
3726 while (scan < loceol && isDIGIT(*scan))
3733 LOAD_UTF8_CHARCLASS(digit,"0");
3734 while (hardcount < max && scan < loceol &&
3735 !swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
3736 scan += UTF8SKIP(scan);
3740 while (scan < loceol && !isDIGIT(*scan))
3744 default: /* Called on something of 0 width. */
3745 break; /* So match right here or not at all. */
3751 c = scan - PL_reginput;
3756 SV *prop = sv_newmortal();
3759 PerlIO_printf(Perl_debug_log,
3760 "%*s %s can match %"IVdf" times out of %"IVdf"...\n",
3761 REPORT_CODE_OFF+1, "", SvPVX(prop),(IV)c,(IV)max);
3768 - regrepeat_hard - repeatedly match something, report total lenth and length
3770 * The repeater is supposed to have constant length.
3774 S_regrepeat_hard(pTHX_ regnode *p, I32 max, I32 *lp)
3776 register char *scan;
3777 register char *start;
3778 register char *loceol = PL_regeol;
3780 I32 count = 0, res = 1;
3785 start = PL_reginput;
3786 if (DO_UTF8(PL_reg_sv)) {
3787 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
3790 while (start < PL_reginput) {
3792 start += UTF8SKIP(start);
3803 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
3805 *lp = l = PL_reginput - start;
3806 if (max != REG_INFTY && l*max < loceol - scan)
3807 loceol = scan + l*max;
3820 - regclass_swash - prepare the utf8 swash
3824 Perl_regclass_swash(pTHX_ register regnode* node, bool doinit, SV** initsvp)
3829 if (PL_regdata && PL_regdata->count) {
3832 if (PL_regdata->what[n] == 's') {
3833 SV *rv = (SV*)PL_regdata->data[n];
3834 AV *av = (AV*)SvRV((SV*)rv);
3837 si = *av_fetch(av, 0, FALSE);
3838 a = av_fetch(av, 1, FALSE);
3842 else if (si && doinit) {
3843 sw = swash_init("utf8", "", si, 1, 0);
3844 (void)av_store(av, 1, sw);
3856 - reginclass - determine if a character falls into a character class
3860 S_reginclass(pTHX_ register regnode *n, register U8* p, register bool do_utf8)
3862 char flags = ANYOF_FLAGS(n);
3867 c = do_utf8 ? utf8_to_uvchr(p, &len) : *p;
3869 if (do_utf8 || (flags & ANYOF_UNICODE)) {
3870 if (do_utf8 && !ANYOF_RUNTIME(n)) {
3871 if (len != (STRLEN)-1 && c < 256 && ANYOF_BITMAP_TEST(n, c))
3874 if (!match && do_utf8 && (flags & ANYOF_UNICODE_ALL) && c >= 256)
3877 SV *sw = regclass_swash(n, TRUE, 0);
3880 if (swash_fetch(sw, p, do_utf8))
3882 else if (flags & ANYOF_FOLD) {
3883 U8 tmpbuf[UTF8_MAXLEN+1];
3885 if (flags & ANYOF_LOCALE) {
3886 PL_reg_flags |= RF_tainted;
3887 uvchr_to_utf8(tmpbuf, toLOWER_LC_utf8(p));
3890 uvchr_to_utf8(tmpbuf, toLOWER_utf8(p));
3891 if (swash_fetch(sw, tmpbuf, do_utf8))
3897 if (!match && c < 256) {
3898 if (ANYOF_BITMAP_TEST(n, c))
3900 else if (flags & ANYOF_FOLD) {
3903 if (flags & ANYOF_LOCALE) {
3904 PL_reg_flags |= RF_tainted;
3905 f = PL_fold_locale[c];
3909 if (f != c && ANYOF_BITMAP_TEST(n, f))
3913 if (!match && (flags & ANYOF_CLASS)) {
3914 PL_reg_flags |= RF_tainted;
3916 (ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) ||
3917 (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) ||
3918 (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) ||
3919 (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) ||
3920 (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) ||
3921 (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) ||
3922 (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) ||
3923 (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) ||
3924 (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) ||
3925 (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) ||
3926 (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII(c)) ||
3927 (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII(c)) ||
3928 (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) ||
3929 (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) ||
3930 (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) ||
3931 (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) ||
3932 (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) ||
3933 (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) ||
3934 (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) ||
3935 (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) ||
3936 (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) ||
3937 (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) ||
3938 (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) ||
3939 (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) ||
3940 (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) ||
3941 (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) ||
3942 (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) ||
3943 (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) ||
3944 (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK(c)) ||
3945 (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK(c))
3946 ) /* How's that for a conditional? */
3953 return (flags & ANYOF_INVERT) ? !match : match;
3957 S_reghop(pTHX_ U8 *s, I32 off)
3959 return S_reghop3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
3963 S_reghop3(pTHX_ U8 *s, I32 off, U8* lim)
3966 while (off-- && s < lim) {
3967 /* XXX could check well-formedness here */
3975 if (UTF8_IS_CONTINUED(*s)) {
3976 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
3979 /* XXX could check well-formedness here */
3987 S_reghopmaybe(pTHX_ U8 *s, I32 off)
3989 return S_reghopmaybe3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
3993 S_reghopmaybe3(pTHX_ U8* s, I32 off, U8* lim)
3996 while (off-- && s < lim) {
3997 /* XXX could check well-formedness here */
4007 if (UTF8_IS_CONTINUED(*s)) {
4008 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4011 /* XXX could check well-formedness here */
4027 restore_pos(pTHXo_ void *arg)
4029 if (PL_reg_eval_set) {
4030 if (PL_reg_oldsaved) {
4031 PL_reg_re->subbeg = PL_reg_oldsaved;
4032 PL_reg_re->sublen = PL_reg_oldsavedlen;
4033 RX_MATCH_COPIED_on(PL_reg_re);
4035 PL_reg_magic->mg_len = PL_reg_oldpos;
4036 PL_reg_eval_set = 0;
4037 PL_curpm = PL_reg_oldcurpm;