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 reghop3_c(pos,off,lim) ((char*)reghop3((U8*)pos, off, (U8*)lim))
120 #define reghopmaybe3_c(pos,off,lim) ((char*)reghopmaybe3((U8*)pos, off, (U8*)lim))
121 #define HOP3(pos,off,lim) (DO_UTF8(PL_reg_sv) ? reghop3((U8*)pos, off, (U8*)lim) : (U8*)(pos + off))
122 #define HOPMAYBE3(pos,off,lim) (DO_UTF8(PL_reg_sv) ? reghopmaybe3((U8*)pos, off, (U8*)lim) : (U8*)(pos + off))
123 #define HOP3c(pos,off,lim) ((char*)HOP3(pos,off,lim))
124 #define HOPMAYBE3c(pos,off,lim) ((char*)HOPMAYBE3(pos,off,lim))
126 #define LOAD_UTF8_CHARCLASS(a,b) STMT_START { if (!CAT2(PL_utf8_,a)) (void)CAT2(is_utf8_, a)((U8*)b); } STMT_END
128 static void restore_pos(pTHXo_ void *arg);
131 S_regcppush(pTHX_ I32 parenfloor)
133 int retval = PL_savestack_ix;
134 #define REGCP_PAREN_ELEMS 4
135 int paren_elems_to_push = (PL_regsize - parenfloor) * REGCP_PAREN_ELEMS;
138 #define REGCP_OTHER_ELEMS 5
139 SSCHECK(paren_elems_to_push + REGCP_OTHER_ELEMS);
140 for (p = PL_regsize; p > parenfloor; p--) {
141 /* REGCP_PARENS_ELEMS are pushed per pairs of parentheses. */
142 SSPUSHINT(PL_regendp[p]);
143 SSPUSHINT(PL_regstartp[p]);
144 SSPUSHPTR(PL_reg_start_tmp[p]);
147 /* REGCP_OTHER_ELEMS are pushed in any case, parentheses or no. */
148 SSPUSHINT(PL_regsize);
149 SSPUSHINT(*PL_reglastparen);
150 SSPUSHPTR(PL_reginput);
151 #define REGCP_FRAME_ELEMS 2
152 /* REGCP_FRAME_ELEMS are part of the REGCP_OTHER_ELEMS and
153 * are needed for the regexp context stack bookkeeping. */
154 SSPUSHINT(paren_elems_to_push + REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS);
155 SSPUSHINT(SAVEt_REGCONTEXT); /* Magic cookie. */
160 /* These are needed since we do not localize EVAL nodes: */
161 # define REGCP_SET(cp) DEBUG_r(PerlIO_printf(Perl_debug_log, \
162 " Setting an EVAL scope, savestack=%"IVdf"\n", \
163 (IV)PL_savestack_ix)); cp = PL_savestack_ix
165 # define REGCP_UNWIND(cp) DEBUG_r(cp != PL_savestack_ix ? \
166 PerlIO_printf(Perl_debug_log, \
167 " Clearing an EVAL scope, savestack=%"IVdf"..%"IVdf"\n", \
168 (IV)(cp), (IV)PL_savestack_ix) : 0); regcpblow(cp)
178 /* Pop REGCP_OTHER_ELEMS before the parentheses loop starts. */
180 assert(i == SAVEt_REGCONTEXT); /* Check that the magic cookie is there. */
181 i = SSPOPINT; /* Parentheses elements to pop. */
182 input = (char *) SSPOPPTR;
183 *PL_reglastparen = SSPOPINT;
184 PL_regsize = SSPOPINT;
186 /* Now restore the parentheses context. */
187 for (i -= (REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS);
188 i > 0; i -= REGCP_PAREN_ELEMS) {
189 paren = (U32)SSPOPINT;
190 PL_reg_start_tmp[paren] = (char *) SSPOPPTR;
191 PL_regstartp[paren] = SSPOPINT;
193 if (paren <= *PL_reglastparen)
194 PL_regendp[paren] = tmps;
196 PerlIO_printf(Perl_debug_log,
197 " restoring \\%"UVuf" to %"IVdf"(%"IVdf")..%"IVdf"%s\n",
198 (UV)paren, (IV)PL_regstartp[paren],
199 (IV)(PL_reg_start_tmp[paren] - PL_bostr),
200 (IV)PL_regendp[paren],
201 (paren > *PL_reglastparen ? "(no)" : ""));
205 if (*PL_reglastparen + 1 <= PL_regnpar) {
206 PerlIO_printf(Perl_debug_log,
207 " restoring \\%"IVdf"..\\%"IVdf" to undef\n",
208 (IV)(*PL_reglastparen + 1), (IV)PL_regnpar);
212 /* It would seem that the similar code in regtry()
213 * already takes care of this, and in fact it is in
214 * a better location to since this code can #if 0-ed out
215 * but the code in regtry() is needed or otherwise tests
216 * requiring null fields (pat.t#187 and split.t#{13,14}
217 * (as of patchlevel 7877) will fail. Then again,
218 * this code seems to be necessary or otherwise
219 * building DynaLoader will fail:
220 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
222 for (paren = *PL_reglastparen + 1; paren <= PL_regnpar; paren++) {
223 if (paren > PL_regsize)
224 PL_regstartp[paren] = -1;
225 PL_regendp[paren] = -1;
232 S_regcp_set_to(pTHX_ I32 ss)
234 I32 tmp = PL_savestack_ix;
236 PL_savestack_ix = ss;
238 PL_savestack_ix = tmp;
242 typedef struct re_cc_state
246 struct re_cc_state *prev;
251 #define regcpblow(cp) LEAVE_SCOPE(cp) /* Ignores regcppush()ed data. */
253 #define TRYPAREN(paren, n, input) { \
256 PL_regstartp[paren] = HOPc(input, -1) - PL_bostr; \
257 PL_regendp[paren] = input - PL_bostr; \
260 PL_regendp[paren] = -1; \
262 if (regmatch(next)) \
265 PL_regendp[paren] = -1; \
270 * pregexec and friends
274 - pregexec - match a regexp against a string
277 Perl_pregexec(pTHX_ register regexp *prog, char *stringarg, register char *strend,
278 char *strbeg, I32 minend, SV *screamer, U32 nosave)
279 /* strend: pointer to null at end of string */
280 /* strbeg: real beginning of string */
281 /* minend: end of match must be >=minend after stringarg. */
282 /* nosave: For optimizations. */
285 regexec_flags(prog, stringarg, strend, strbeg, minend, screamer, NULL,
286 nosave ? 0 : REXEC_COPY_STR);
290 S_cache_re(pTHX_ regexp *prog)
292 PL_regprecomp = prog->precomp; /* Needed for FAIL. */
294 PL_regprogram = prog->program;
296 PL_regnpar = prog->nparens;
297 PL_regdata = prog->data;
302 * Need to implement the following flags for reg_anch:
304 * USE_INTUIT_NOML - Useful to call re_intuit_start() first
306 * INTUIT_AUTORITATIVE_NOML - Can trust a positive answer
307 * INTUIT_AUTORITATIVE_ML
308 * INTUIT_ONCE_NOML - Intuit can match in one location only.
311 * Another flag for this function: SECOND_TIME (so that float substrs
312 * with giant delta may be not rechecked).
315 /* Assumptions: if ANCH_GPOS, then strpos is anchored. XXXX Check GPOS logic */
317 /* If SCREAM, then SvPVX(sv) should be compatible with strpos and strend.
318 Otherwise, only SvCUR(sv) is used to get strbeg. */
320 /* XXXX We assume that strpos is strbeg unless sv. */
322 /* XXXX Some places assume that there is a fixed substring.
323 An update may be needed if optimizer marks as "INTUITable"
324 RExen without fixed substrings. Similarly, it is assumed that
325 lengths of all the strings are no more than minlen, thus they
326 cannot come from lookahead.
327 (Or minlen should take into account lookahead.) */
329 /* A failure to find a constant substring means that there is no need to make
330 an expensive call to REx engine, thus we celebrate a failure. Similarly,
331 finding a substring too deep into the string means that less calls to
332 regtry() should be needed.
334 REx compiler's optimizer found 4 possible hints:
335 a) Anchored substring;
337 c) Whether we are anchored (beginning-of-line or \G);
338 d) First node (of those at offset 0) which may distingush positions;
339 We use a)b)d) and multiline-part of c), and try to find a position in the
340 string which does not contradict any of them.
343 /* Most of decisions we do here should have been done at compile time.
344 The nodes of the REx which we used for the search should have been
345 deleted from the finite automaton. */
348 Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos,
349 char *strend, U32 flags, re_scream_pos_data *data)
351 register I32 start_shift;
352 /* Should be nonnegative! */
353 register I32 end_shift;
360 register char *other_last = Nullch; /* other substr checked before this */
361 char *check_at; /* check substr found at this pos */
363 char *i_strpos = strpos;
366 DEBUG_r( if (!PL_colorset) reginitcolors() );
367 DEBUG_r(PerlIO_printf(Perl_debug_log,
368 "%sGuessing start of match, REx%s `%s%.60s%s%s' against `%s%.*s%s%s'...\n",
369 PL_colors[4],PL_colors[5],PL_colors[0],
372 (strlen(prog->precomp) > 60 ? "..." : ""),
374 (int)(strend - strpos > 60 ? 60 : strend - strpos),
375 strpos, PL_colors[1],
376 (strend - strpos > 60 ? "..." : ""))
379 if (prog->reganch & ROPT_UTF8)
380 PL_reg_flags |= RF_utf8;
382 if (prog->minlen > CHR_DIST((U8*)strend, (U8*)strpos)) {
383 DEBUG_r(PerlIO_printf(Perl_debug_log, "String too short...\n"));
386 strbeg = (sv && SvPOK(sv)) ? strend - SvCUR(sv) : strpos;
388 check = prog->check_substr;
389 if (prog->reganch & ROPT_ANCH) { /* Match at beg-of-str or after \n */
390 ml_anch = !( (prog->reganch & ROPT_ANCH_SINGLE)
391 || ( (prog->reganch & ROPT_ANCH_BOL)
392 && !PL_multiline ) ); /* Check after \n? */
395 if ( !(prog->reganch & (ROPT_ANCH_GPOS /* Checked by the caller */
396 | ROPT_IMPLICIT)) /* not a real BOL */
397 /* SvCUR is not set on references: SvRV and SvPVX overlap */
399 && (strpos != strbeg)) {
400 DEBUG_r(PerlIO_printf(Perl_debug_log, "Not at start...\n"));
403 if (prog->check_offset_min == prog->check_offset_max &&
404 !(prog->reganch & ROPT_SANY_SEEN)) {
405 /* Substring at constant offset from beg-of-str... */
408 s = HOP3c(strpos, prog->check_offset_min, strend);
410 slen = SvCUR(check); /* >= 1 */
412 if ( strend - s > slen || strend - s < slen - 1
413 || (strend - s == slen && strend[-1] != '\n')) {
414 DEBUG_r(PerlIO_printf(Perl_debug_log, "String too long...\n"));
417 /* Now should match s[0..slen-2] */
419 if (slen && (*SvPVX(check) != *s
421 && memNE(SvPVX(check), s, slen)))) {
423 DEBUG_r(PerlIO_printf(Perl_debug_log, "String not equal...\n"));
427 else if (*SvPVX(check) != *s
428 || ((slen = SvCUR(check)) > 1
429 && memNE(SvPVX(check), s, slen)))
431 goto success_at_start;
434 /* Match is anchored, but substr is not anchored wrt beg-of-str. */
436 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
437 end_shift = prog->minlen - start_shift -
438 CHR_SVLEN(check) + (SvTAIL(check) != 0);
440 I32 end = prog->check_offset_max + CHR_SVLEN(check)
441 - (SvTAIL(check) != 0);
442 I32 eshift = CHR_DIST((U8*)strend, (U8*)s) - end;
444 if (end_shift < eshift)
448 else { /* Can match at random position */
451 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
452 /* Should be nonnegative! */
453 end_shift = prog->minlen - start_shift -
454 CHR_SVLEN(check) + (SvTAIL(check) != 0);
457 #ifdef DEBUGGING /* 7/99: reports of failure (with the older version) */
459 Perl_croak(aTHX_ "panic: end_shift");
463 /* Find a possible match in the region s..strend by looking for
464 the "check" substring in the region corrected by start/end_shift. */
465 if (flags & REXEC_SCREAM) {
466 I32 p = -1; /* Internal iterator of scream. */
467 I32 *pp = data ? data->scream_pos : &p;
469 if (PL_screamfirst[BmRARE(check)] >= 0
470 || ( BmRARE(check) == '\n'
471 && (BmPREVIOUS(check) == SvCUR(check) - 1)
473 s = screaminstr(sv, check,
474 start_shift + (s - strbeg), end_shift, pp, 0);
478 *data->scream_olds = s;
480 else if (prog->reganch & ROPT_SANY_SEEN)
481 s = fbm_instr((U8*)(s + start_shift),
482 (U8*)(strend - end_shift),
483 check, PL_multiline ? FBMrf_MULTILINE : 0);
485 s = fbm_instr(HOP3(s, start_shift, strend),
486 HOP3(strend, -end_shift, strbeg),
487 check, PL_multiline ? FBMrf_MULTILINE : 0);
489 /* Update the count-of-usability, remove useless subpatterns,
492 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s %s substr `%s%.*s%s'%s%s",
493 (s ? "Found" : "Did not find"),
494 ((check == prog->anchored_substr) ? "anchored" : "floating"),
496 (int)(SvCUR(check) - (SvTAIL(check)!=0)),
498 PL_colors[1], (SvTAIL(check) ? "$" : ""),
499 (s ? " at offset " : "...\n") ) );
506 /* Finish the diagnostic message */
507 DEBUG_r(PerlIO_printf(Perl_debug_log, "%ld...\n", (long)(s - i_strpos)) );
509 /* Got a candidate. Check MBOL anchoring, and the *other* substr.
510 Start with the other substr.
511 XXXX no SCREAM optimization yet - and a very coarse implementation
512 XXXX /ttx+/ results in anchored=`ttx', floating=`x'. floating will
513 *always* match. Probably should be marked during compile...
514 Probably it is right to do no SCREAM here...
517 if (prog->float_substr && prog->anchored_substr) {
518 /* Take into account the "other" substring. */
519 /* XXXX May be hopelessly wrong for UTF... */
522 if (check == prog->float_substr) {
525 char *last = HOP3c(s, -start_shift, strbeg), *last1, *last2;
528 t = s - prog->check_offset_max;
529 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
530 && (!(prog->reganch & ROPT_UTF8)
531 || ((t = reghopmaybe3_c(s, -(prog->check_offset_max), strpos))
536 t = HOP3c(t, prog->anchored_offset, strend);
537 if (t < other_last) /* These positions already checked */
539 last2 = last1 = HOP3c(strend, -prog->minlen, strbeg);
542 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
543 /* On end-of-str: see comment below. */
544 s = fbm_instr((unsigned char*)t,
545 HOP3(HOP3(last1, prog->anchored_offset, strend)
546 + SvCUR(prog->anchored_substr),
547 -(SvTAIL(prog->anchored_substr)!=0), strbeg),
548 prog->anchored_substr,
549 PL_multiline ? FBMrf_MULTILINE : 0);
550 DEBUG_r(PerlIO_printf(Perl_debug_log,
551 "%s anchored substr `%s%.*s%s'%s",
552 (s ? "Found" : "Contradicts"),
554 (int)(SvCUR(prog->anchored_substr)
555 - (SvTAIL(prog->anchored_substr)!=0)),
556 SvPVX(prog->anchored_substr),
557 PL_colors[1], (SvTAIL(prog->anchored_substr) ? "$" : "")));
559 if (last1 >= last2) {
560 DEBUG_r(PerlIO_printf(Perl_debug_log,
561 ", giving up...\n"));
564 DEBUG_r(PerlIO_printf(Perl_debug_log,
565 ", trying floating at offset %ld...\n",
566 (long)(HOP3c(s1, 1, strend) - i_strpos)));
567 other_last = HOP3c(last1, prog->anchored_offset+1, strend);
568 s = HOP3c(last, 1, strend);
572 DEBUG_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
573 (long)(s - i_strpos)));
574 t = HOP3c(s, -prog->anchored_offset, strbeg);
575 other_last = HOP3c(s, 1, strend);
583 else { /* Take into account the floating substring. */
587 t = HOP3c(s, -start_shift, strbeg);
589 HOP3c(strend, -prog->minlen + prog->float_min_offset, strbeg);
590 if (CHR_DIST((U8*)last, (U8*)t) > prog->float_max_offset)
591 last = HOP3c(t, prog->float_max_offset, strend);
592 s = HOP3c(t, prog->float_min_offset, strend);
595 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
596 /* fbm_instr() takes into account exact value of end-of-str
597 if the check is SvTAIL(ed). Since false positives are OK,
598 and end-of-str is not later than strend we are OK. */
599 s = fbm_instr((unsigned char*)s,
600 (unsigned char*)last + SvCUR(prog->float_substr)
601 - (SvTAIL(prog->float_substr)!=0),
602 prog->float_substr, PL_multiline ? FBMrf_MULTILINE : 0);
603 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s floating substr `%s%.*s%s'%s",
604 (s ? "Found" : "Contradicts"),
606 (int)(SvCUR(prog->float_substr)
607 - (SvTAIL(prog->float_substr)!=0)),
608 SvPVX(prog->float_substr),
609 PL_colors[1], (SvTAIL(prog->float_substr) ? "$" : "")));
612 DEBUG_r(PerlIO_printf(Perl_debug_log,
613 ", giving up...\n"));
616 DEBUG_r(PerlIO_printf(Perl_debug_log,
617 ", trying anchored starting at offset %ld...\n",
618 (long)(s1 + 1 - i_strpos)));
620 s = HOP3c(t, 1, strend);
624 DEBUG_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
625 (long)(s - i_strpos)));
626 other_last = s; /* Fix this later. --Hugo */
635 t = s - prog->check_offset_max;
636 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
637 && (!(prog->reganch & ROPT_UTF8)
638 || ((t = reghopmaybe3_c(s, -prog->check_offset_max, strpos))
640 /* Fixed substring is found far enough so that the match
641 cannot start at strpos. */
643 if (ml_anch && t[-1] != '\n') {
644 /* Eventually fbm_*() should handle this, but often
645 anchored_offset is not 0, so this check will not be wasted. */
646 /* XXXX In the code below we prefer to look for "^" even in
647 presence of anchored substrings. And we search even
648 beyond the found float position. These pessimizations
649 are historical artefacts only. */
651 while (t < strend - prog->minlen) {
653 if (t < check_at - prog->check_offset_min) {
654 if (prog->anchored_substr) {
655 /* Since we moved from the found position,
656 we definitely contradict the found anchored
657 substr. Due to the above check we do not
658 contradict "check" substr.
659 Thus we can arrive here only if check substr
660 is float. Redo checking for "other"=="fixed".
663 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld, rescanning for anchored from offset %ld...\n",
664 PL_colors[0],PL_colors[1], (long)(strpos - i_strpos), (long)(strpos - i_strpos + prog->anchored_offset)));
665 goto do_other_anchored;
667 /* We don't contradict the found floating substring. */
668 /* XXXX Why not check for STCLASS? */
670 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld...\n",
671 PL_colors[0],PL_colors[1], (long)(s - i_strpos)));
674 /* Position contradicts check-string */
675 /* XXXX probably better to look for check-string
676 than for "\n", so one should lower the limit for t? */
677 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m, restarting lookup for check-string at offset %ld...\n",
678 PL_colors[0],PL_colors[1], (long)(t + 1 - i_strpos)));
679 other_last = strpos = s = t + 1;
684 DEBUG_r(PerlIO_printf(Perl_debug_log, "Did not find /%s^%s/m...\n",
685 PL_colors[0],PL_colors[1]));
689 DEBUG_r(PerlIO_printf(Perl_debug_log, "Starting position does not contradict /%s^%s/m...\n",
690 PL_colors[0],PL_colors[1]));
694 ++BmUSEFUL(prog->check_substr); /* hooray/5 */
697 /* The found string does not prohibit matching at strpos,
698 - no optimization of calling REx engine can be performed,
699 unless it was an MBOL and we are not after MBOL,
700 or a future STCLASS check will fail this. */
702 /* Even in this situation we may use MBOL flag if strpos is offset
703 wrt the start of the string. */
704 if (ml_anch && sv && !SvROK(sv) /* See prev comment on SvROK */
705 && (strpos != strbeg) && strpos[-1] != '\n'
706 /* May be due to an implicit anchor of m{.*foo} */
707 && !(prog->reganch & ROPT_IMPLICIT))
712 DEBUG_r( if (ml_anch)
713 PerlIO_printf(Perl_debug_log, "Position at offset %ld does not contradict /%s^%s/m...\n",
714 (long)(strpos - i_strpos), PL_colors[0],PL_colors[1]);
717 if (!(prog->reganch & ROPT_NAUGHTY) /* XXXX If strpos moved? */
718 && prog->check_substr /* Could be deleted already */
719 && --BmUSEFUL(prog->check_substr) < 0
720 && prog->check_substr == prog->float_substr)
722 /* If flags & SOMETHING - do not do it many times on the same match */
723 DEBUG_r(PerlIO_printf(Perl_debug_log, "... Disabling check substring...\n"));
724 SvREFCNT_dec(prog->check_substr);
725 prog->check_substr = Nullsv; /* disable */
726 prog->float_substr = Nullsv; /* clear */
727 check = Nullsv; /* abort */
729 /* XXXX This is a remnant of the old implementation. It
730 looks wasteful, since now INTUIT can use many
732 prog->reganch &= ~RE_USE_INTUIT;
739 /* XXXX BmUSEFUL already changed, maybe multiple change is meaningful... */
740 if (prog->regstclass) {
741 /* minlen == 0 is possible if regstclass is \b or \B,
742 and the fixed substr is ''$.
743 Since minlen is already taken into account, s+1 is before strend;
744 accidentally, minlen >= 1 guaranties no false positives at s + 1
745 even for \b or \B. But (minlen? 1 : 0) below assumes that
746 regstclass does not come from lookahead... */
747 /* If regstclass takes bytelength more than 1: If charlength==1, OK.
748 This leaves EXACTF only, which is dealt with in find_byclass(). */
749 U8* str = (U8*)STRING(prog->regstclass);
750 int cl_l = (PL_regkind[(U8)OP(prog->regstclass)] == EXACT
751 ? CHR_DIST(str+STR_LEN(prog->regstclass), str)
753 char *endpos = (prog->anchored_substr || ml_anch)
754 ? HOP3c(s, (prog->minlen ? cl_l : 0), strend)
755 : (prog->float_substr
756 ? HOP3c(HOP3c(check_at, -start_shift, strbeg),
759 char *startpos = strbeg;
762 if (prog->reganch & ROPT_UTF8) {
763 PL_regdata = prog->data;
766 s = find_byclass(prog, prog->regstclass, s, endpos, startpos, 1);
771 if (endpos == strend) {
772 DEBUG_r( PerlIO_printf(Perl_debug_log,
773 "Could not match STCLASS...\n") );
776 DEBUG_r( PerlIO_printf(Perl_debug_log,
777 "This position contradicts STCLASS...\n") );
778 if ((prog->reganch & ROPT_ANCH) && !ml_anch)
780 /* Contradict one of substrings */
781 if (prog->anchored_substr) {
782 if (prog->anchored_substr == check) {
783 DEBUG_r( what = "anchored" );
785 s = HOP3c(t, 1, strend);
786 if (s + start_shift + end_shift > strend) {
787 /* XXXX Should be taken into account earlier? */
788 DEBUG_r( PerlIO_printf(Perl_debug_log,
789 "Could not match STCLASS...\n") );
794 DEBUG_r( PerlIO_printf(Perl_debug_log,
795 "Looking for %s substr starting at offset %ld...\n",
796 what, (long)(s + start_shift - i_strpos)) );
799 /* Have both, check_string is floating */
800 if (t + start_shift >= check_at) /* Contradicts floating=check */
801 goto retry_floating_check;
802 /* Recheck anchored substring, but not floating... */
806 DEBUG_r( PerlIO_printf(Perl_debug_log,
807 "Looking for anchored substr starting at offset %ld...\n",
808 (long)(other_last - i_strpos)) );
809 goto do_other_anchored;
811 /* Another way we could have checked stclass at the
812 current position only: */
817 DEBUG_r( PerlIO_printf(Perl_debug_log,
818 "Looking for /%s^%s/m starting at offset %ld...\n",
819 PL_colors[0],PL_colors[1], (long)(t - i_strpos)) );
822 if (!prog->float_substr) /* Could have been deleted */
824 /* Check is floating subtring. */
825 retry_floating_check:
826 t = check_at - start_shift;
827 DEBUG_r( what = "floating" );
828 goto hop_and_restart;
831 PerlIO_printf(Perl_debug_log,
832 "By STCLASS: moving %ld --> %ld\n",
833 (long)(t - i_strpos), (long)(s - i_strpos));
835 PerlIO_printf(Perl_debug_log,
836 "Does not contradict STCLASS...\n") );
839 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s%s:%s match at offset %ld\n",
840 PL_colors[4], (check ? "Guessed" : "Giving up"),
841 PL_colors[5], (long)(s - i_strpos)) );
844 fail_finish: /* Substring not found */
845 if (prog->check_substr) /* could be removed already */
846 BmUSEFUL(prog->check_substr) += 5; /* hooray */
848 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch rejected by optimizer%s\n",
849 PL_colors[4],PL_colors[5]));
853 /* We know what class REx starts with. Try to find this position... */
855 S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, char *strend, char *startpos, I32 norun)
857 I32 doevery = (prog->reganch & ROPT_SKIP) == 0;
863 register I32 tmp = 1; /* Scratch variable? */
864 register bool do_utf8 = DO_UTF8(PL_reg_sv);
866 /* We know what class it must start with. */
870 if (reginclass(c, (U8*)s, do_utf8)) {
871 if (tmp && (norun || regtry(prog, s)))
878 s += do_utf8 ? UTF8SKIP(s) : 1;
885 c1 = to_utf8_lower((U8*)m);
886 c2 = to_utf8_upper((U8*)m);
897 c2 = PL_fold_locale[c1];
902 e = s; /* Due to minlen logic of intuit() */
908 if ( utf8_to_uvchr((U8*)s, &len) == c1
915 UV c = utf8_to_uvchr((U8*)s, &len);
916 if ( (c == c1 || c == c2) && regtry(prog, s) )
925 && (ln == 1 || !(OP(c) == EXACTF
927 : ibcmp_locale(s, m, ln)))
928 && (norun || regtry(prog, s)) )
934 if ( (*(U8*)s == c1 || *(U8*)s == c2)
935 && (ln == 1 || !(OP(c) == EXACTF
937 : ibcmp_locale(s, m, ln)))
938 && (norun || regtry(prog, s)) )
945 PL_reg_flags |= RF_tainted;
952 U8 *r = reghop3((U8*)s, -1, (U8*)startpos);
954 tmp = (I32)utf8n_to_uvchr(r, s - (char*)r, 0, 0);
956 tmp = ((OP(c) == BOUND ?
957 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
958 LOAD_UTF8_CHARCLASS(alnum,"a");
960 if (tmp == !(OP(c) == BOUND ?
961 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
962 isALNUM_LC_utf8((U8*)s)))
965 if ((norun || regtry(prog, s)))
972 tmp = (s != startpos) ? UCHARAT(s - 1) : '\n';
973 tmp = ((OP(c) == BOUND ? isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
976 !(OP(c) == BOUND ? isALNUM(*s) : isALNUM_LC(*s))) {
978 if ((norun || regtry(prog, s)))
984 if ((!prog->minlen && tmp) && (norun || regtry(prog, s)))
988 PL_reg_flags |= RF_tainted;
995 U8 *r = reghop3((U8*)s, -1, (U8*)startpos);
997 tmp = (I32)utf8n_to_uvchr(r, s - (char*)r, 0, 0);
999 tmp = ((OP(c) == NBOUND ?
1000 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1001 LOAD_UTF8_CHARCLASS(alnum,"a");
1002 while (s < strend) {
1003 if (tmp == !(OP(c) == NBOUND ?
1004 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1005 isALNUM_LC_utf8((U8*)s)))
1007 else if ((norun || regtry(prog, s)))
1013 tmp = (s != startpos) ? UCHARAT(s - 1) : '\n';
1014 tmp = ((OP(c) == NBOUND ?
1015 isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1016 while (s < strend) {
1018 !(OP(c) == NBOUND ? isALNUM(*s) : isALNUM_LC(*s)))
1020 else if ((norun || regtry(prog, s)))
1025 if ((!prog->minlen && !tmp) && (norun || regtry(prog, s)))
1030 LOAD_UTF8_CHARCLASS(alnum,"a");
1031 while (s < strend) {
1032 if (swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1033 if (tmp && (norun || regtry(prog, s)))
1044 while (s < strend) {
1046 if (tmp && (norun || regtry(prog, s)))
1058 PL_reg_flags |= RF_tainted;
1060 while (s < strend) {
1061 if (isALNUM_LC_utf8((U8*)s)) {
1062 if (tmp && (norun || regtry(prog, s)))
1073 while (s < strend) {
1074 if (isALNUM_LC(*s)) {
1075 if (tmp && (norun || regtry(prog, s)))
1088 LOAD_UTF8_CHARCLASS(alnum,"a");
1089 while (s < strend) {
1090 if (!swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1091 if (tmp && (norun || regtry(prog, s)))
1102 while (s < strend) {
1104 if (tmp && (norun || regtry(prog, s)))
1116 PL_reg_flags |= RF_tainted;
1118 while (s < strend) {
1119 if (!isALNUM_LC_utf8((U8*)s)) {
1120 if (tmp && (norun || regtry(prog, s)))
1131 while (s < strend) {
1132 if (!isALNUM_LC(*s)) {
1133 if (tmp && (norun || regtry(prog, s)))
1146 LOAD_UTF8_CHARCLASS(space," ");
1147 while (s < strend) {
1148 if (*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8)) {
1149 if (tmp && (norun || regtry(prog, s)))
1160 while (s < strend) {
1162 if (tmp && (norun || regtry(prog, s)))
1174 PL_reg_flags |= RF_tainted;
1176 while (s < strend) {
1177 if (*s == ' ' || isSPACE_LC_utf8((U8*)s)) {
1178 if (tmp && (norun || regtry(prog, s)))
1189 while (s < strend) {
1190 if (isSPACE_LC(*s)) {
1191 if (tmp && (norun || regtry(prog, s)))
1204 LOAD_UTF8_CHARCLASS(space," ");
1205 while (s < strend) {
1206 if (!(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8))) {
1207 if (tmp && (norun || regtry(prog, s)))
1218 while (s < strend) {
1220 if (tmp && (norun || regtry(prog, s)))
1232 PL_reg_flags |= RF_tainted;
1234 while (s < strend) {
1235 if (!(*s == ' ' || isSPACE_LC_utf8((U8*)s))) {
1236 if (tmp && (norun || regtry(prog, s)))
1247 while (s < strend) {
1248 if (!isSPACE_LC(*s)) {
1249 if (tmp && (norun || regtry(prog, s)))
1262 LOAD_UTF8_CHARCLASS(digit,"0");
1263 while (s < strend) {
1264 if (swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1265 if (tmp && (norun || regtry(prog, s)))
1276 while (s < strend) {
1278 if (tmp && (norun || regtry(prog, s)))
1290 PL_reg_flags |= RF_tainted;
1292 while (s < strend) {
1293 if (isDIGIT_LC_utf8((U8*)s)) {
1294 if (tmp && (norun || regtry(prog, s)))
1305 while (s < strend) {
1306 if (isDIGIT_LC(*s)) {
1307 if (tmp && (norun || regtry(prog, s)))
1320 LOAD_UTF8_CHARCLASS(digit,"0");
1321 while (s < strend) {
1322 if (!swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1323 if (tmp && (norun || regtry(prog, s)))
1334 while (s < strend) {
1336 if (tmp && (norun || regtry(prog, s)))
1348 PL_reg_flags |= RF_tainted;
1350 while (s < strend) {
1351 if (!isDIGIT_LC_utf8((U8*)s)) {
1352 if (tmp && (norun || regtry(prog, s)))
1363 while (s < strend) {
1364 if (!isDIGIT_LC(*s)) {
1365 if (tmp && (norun || regtry(prog, s)))
1377 Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c));
1386 - regexec_flags - match a regexp against a string
1389 Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char *strend,
1390 char *strbeg, I32 minend, SV *sv, void *data, U32 flags)
1391 /* strend: pointer to null at end of string */
1392 /* strbeg: real beginning of string */
1393 /* minend: end of match must be >=minend after stringarg. */
1394 /* data: May be used for some additional optimizations. */
1395 /* nosave: For optimizations. */
1398 register regnode *c;
1399 register char *startpos = stringarg;
1400 I32 minlen; /* must match at least this many chars */
1401 I32 dontbother = 0; /* how many characters not to try at end */
1402 /* I32 start_shift = 0; */ /* Offset of the start to find
1403 constant substr. */ /* CC */
1404 I32 end_shift = 0; /* Same for the end. */ /* CC */
1405 I32 scream_pos = -1; /* Internal iterator of scream. */
1407 SV* oreplsv = GvSV(PL_replgv);
1408 bool do_utf8 = DO_UTF8(sv);
1414 PL_regnarrate = DEBUG_r_TEST;
1417 /* Be paranoid... */
1418 if (prog == NULL || startpos == NULL) {
1419 Perl_croak(aTHX_ "NULL regexp parameter");
1423 minlen = prog->minlen;
1425 if (!(prog->reganch & ROPT_SANY_SEEN))
1426 if (utf8_distance((U8*)strend, (U8*)startpos) < minlen) goto phooey;
1429 if (strend - startpos < minlen) goto phooey;
1432 if (startpos == strbeg) /* is ^ valid at stringarg? */
1435 if (prog->reganch & ROPT_UTF8 && do_utf8) {
1436 U8 *s = reghop3((U8*)stringarg, -1, (U8*)strbeg);
1437 PL_regprev = utf8n_to_uvchr(s, (U8*)stringarg - s, NULL, 0);
1440 PL_regprev = (U32)stringarg[-1];
1441 if (!PL_multiline && PL_regprev == '\n')
1442 PL_regprev = '\0'; /* force ^ to NOT match */
1445 /* Check validity of program. */
1446 if (UCHARAT(prog->program) != REG_MAGIC) {
1447 Perl_croak(aTHX_ "corrupted regexp program");
1451 PL_reg_eval_set = 0;
1454 if (prog->reganch & ROPT_UTF8)
1455 PL_reg_flags |= RF_utf8;
1457 /* Mark beginning of line for ^ and lookbehind. */
1458 PL_regbol = startpos;
1462 /* Mark end of line for $ (and such) */
1465 /* see how far we have to get to not match where we matched before */
1466 PL_regtill = startpos+minend;
1468 /* We start without call_cc context. */
1471 /* If there is a "must appear" string, look for it. */
1474 if (prog->reganch & ROPT_GPOS_SEEN) { /* Need to have PL_reg_ganch */
1477 if (flags & REXEC_IGNOREPOS) /* Means: check only at start */
1478 PL_reg_ganch = startpos;
1479 else if (sv && SvTYPE(sv) >= SVt_PVMG
1481 && (mg = mg_find(sv, 'g')) && 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, 'g')))) {
1833 /* prepare for quick setting of pos */
1834 sv_magic(PL_reg_sv, (SV*)0, 'g', Nullch, 0);
1835 mg = mg_find(PL_reg_sv, 'g');
1839 PL_reg_oldpos = mg->mg_len;
1840 SAVEDESTRUCTOR_X(restore_pos, 0);
1843 Newz(22,PL_reg_curpm, 1, PMOP);
1844 PL_reg_curpm->op_pmregexp = prog;
1845 PL_reg_oldcurpm = PL_curpm;
1846 PL_curpm = PL_reg_curpm;
1847 if (RX_MATCH_COPIED(prog)) {
1848 /* Here is a serious problem: we cannot rewrite subbeg,
1849 since it may be needed if this match fails. Thus
1850 $` inside (?{}) could fail... */
1851 PL_reg_oldsaved = prog->subbeg;
1852 PL_reg_oldsavedlen = prog->sublen;
1853 RX_MATCH_COPIED_off(prog);
1856 PL_reg_oldsaved = Nullch;
1857 prog->subbeg = PL_bostr;
1858 prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */
1860 prog->startp[0] = startpos - PL_bostr;
1861 PL_reginput = startpos;
1862 PL_regstartp = prog->startp;
1863 PL_regendp = prog->endp;
1864 PL_reglastparen = &prog->lastparen;
1865 prog->lastparen = 0;
1867 DEBUG_r(PL_reg_starttry = startpos);
1868 if (PL_reg_start_tmpl <= prog->nparens) {
1869 PL_reg_start_tmpl = prog->nparens*3/2 + 3;
1870 if(PL_reg_start_tmp)
1871 Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
1873 New(22,PL_reg_start_tmp, PL_reg_start_tmpl, char*);
1876 /* XXXX What this code is doing here?!!! There should be no need
1877 to do this again and again, PL_reglastparen should take care of
1880 /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code.
1881 * Actually, the code in regcppop() (which Ilya may be meaning by
1882 * PL_reglastparen), is not needed at all by the test suite
1883 * (op/regexp, op/pat, op/split), but that code is needed, oddly
1884 * enough, for building DynaLoader, or otherwise this
1885 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
1886 * will happen. Meanwhile, this code *is* needed for the
1887 * above-mentioned test suite tests to succeed. The common theme
1888 * on those tests seems to be returning null fields from matches.
1893 if (prog->nparens) {
1894 for (i = prog->nparens; i > *PL_reglastparen; i--) {
1901 if (regmatch(prog->program + 1)) {
1902 prog->endp[0] = PL_reginput - PL_bostr;
1905 REGCP_UNWIND(lastcp);
1909 #define RE_UNWIND_BRANCH 1
1910 #define RE_UNWIND_BRANCHJ 2
1914 typedef struct { /* XX: makes sense to enlarge it... */
1918 } re_unwind_generic_t;
1931 } re_unwind_branch_t;
1933 typedef union re_unwind_t {
1935 re_unwind_generic_t generic;
1936 re_unwind_branch_t branch;
1940 - regmatch - main matching routine
1942 * Conceptually the strategy is simple: check to see whether the current
1943 * node matches, call self recursively to see whether the rest matches,
1944 * and then act accordingly. In practice we make some effort to avoid
1945 * recursion, in particular by going through "ordinary" nodes (that don't
1946 * need to know whether the rest of the match failed) by a loop instead of
1949 /* [lwall] I've hoisted the register declarations to the outer block in order to
1950 * maybe save a little bit of pushing and popping on the stack. It also takes
1951 * advantage of machines that use a register save mask on subroutine entry.
1953 STATIC I32 /* 0 failure, 1 success */
1954 S_regmatch(pTHX_ regnode *prog)
1956 register regnode *scan; /* Current node. */
1957 regnode *next; /* Next node. */
1958 regnode *inner; /* Next node in internal branch. */
1959 register I32 nextchr; /* renamed nextchr - nextchar colides with
1960 function of same name */
1961 register I32 n; /* no or next */
1962 register I32 ln; /* len or last */
1963 register char *s; /* operand or save */
1964 register char *locinput = PL_reginput;
1965 register I32 c1, c2, paren; /* case fold search, parenth */
1966 int minmod = 0, sw = 0, logical = 0;
1968 I32 firstcp = PL_savestack_ix;
1969 register bool do_utf8 = DO_UTF8(PL_reg_sv);
1975 /* Note that nextchr is a byte even in UTF */
1976 nextchr = UCHARAT(locinput);
1978 while (scan != NULL) {
1979 #define sayNO_L (logical ? (logical = 0, sw = 0, goto cont) : sayNO)
1981 # define sayYES goto yes
1982 # define sayNO goto no
1983 # define sayYES_FINAL goto yes_final
1984 # define sayYES_LOUD goto yes_loud
1985 # define sayNO_FINAL goto no_final
1986 # define sayNO_SILENT goto do_no
1987 # define saySAME(x) if (x) goto yes; else goto no
1988 # define REPORT_CODE_OFF 24
1990 # define sayYES return 1
1991 # define sayNO return 0
1992 # define sayYES_FINAL return 1
1993 # define sayYES_LOUD return 1
1994 # define sayNO_FINAL return 0
1995 # define sayNO_SILENT return 0
1996 # define saySAME(x) return x
1999 SV *prop = sv_newmortal();
2000 int docolor = *PL_colors[0];
2001 int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */
2002 int l = (PL_regeol - locinput) > taill ? taill : (PL_regeol - locinput);
2003 /* The part of the string before starttry has one color
2004 (pref0_len chars), between starttry and current
2005 position another one (pref_len - pref0_len chars),
2006 after the current position the third one.
2007 We assume that pref0_len <= pref_len, otherwise we
2008 decrease pref0_len. */
2009 int pref_len = (locinput - PL_bostr) > (5 + taill) - l
2010 ? (5 + taill) - l : locinput - PL_bostr;
2013 while (UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len)))
2015 pref0_len = pref_len - (locinput - PL_reg_starttry);
2016 if (l + pref_len < (5 + taill) && l < PL_regeol - locinput)
2017 l = ( PL_regeol - locinput > (5 + taill) - pref_len
2018 ? (5 + taill) - pref_len : PL_regeol - locinput);
2019 while (UTF8_IS_CONTINUATION(*(U8*)(locinput + l)))
2023 if (pref0_len > pref_len)
2024 pref0_len = pref_len;
2025 regprop(prop, scan);
2026 PerlIO_printf(Perl_debug_log,
2027 "%4"IVdf" <%s%.*s%s%s%.*s%s%s%s%.*s%s>%*s|%3"IVdf":%*s%s\n",
2028 (IV)(locinput - PL_bostr),
2029 PL_colors[4], pref0_len,
2030 locinput - pref_len, PL_colors[5],
2031 PL_colors[2], pref_len - pref0_len,
2032 locinput - pref_len + pref0_len, PL_colors[3],
2033 (docolor ? "" : "> <"),
2034 PL_colors[0], l, locinput, PL_colors[1],
2035 15 - l - pref_len + 1,
2037 (IV)(scan - PL_regprogram), PL_regindent*2, "",
2041 next = scan + NEXT_OFF(scan);
2047 if (locinput == PL_bostr
2048 ? PL_regprev == '\n'
2050 (nextchr || locinput < PL_regeol) && locinput[-1] == '\n') )
2052 /* regtill = regbol; */
2057 if (locinput == PL_bostr
2058 ? PL_regprev == '\n'
2059 : ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n') )
2065 if (locinput == PL_bostr)
2069 if (locinput == PL_reg_ganch)
2079 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2084 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2086 if (PL_regeol - locinput > 1)
2090 if (PL_regeol != locinput)
2094 if (!nextchr && locinput >= PL_regeol)
2096 nextchr = UCHARAT(++locinput);
2099 if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n')
2102 locinput += PL_utf8skip[nextchr];
2103 if (locinput > PL_regeol)
2105 nextchr = UCHARAT(locinput);
2108 nextchr = UCHARAT(++locinput);
2113 if (do_utf8 != (UTF!=0)) {
2121 if (*((U8*)s) != utf8_to_uvchr((U8*)l, &len))
2130 if (*((U8*)l) != utf8_to_uvchr((U8*)s, &len))
2136 nextchr = UCHARAT(locinput);
2139 /* Inline the first character, for speed. */
2140 if (UCHARAT(s) != nextchr)
2142 if (PL_regeol - locinput < ln)
2144 if (ln > 1 && memNE(s, locinput, ln))
2147 nextchr = UCHARAT(locinput);
2150 PL_reg_flags |= RF_tainted;
2160 c1 = OP(scan) == EXACTF;
2162 if (l >= PL_regeol) {
2165 if ((UTF ? utf8n_to_uvchr((U8*)s, e - s, 0, 0) : *((U8*)s)) !=
2166 (c1 ? toLOWER_utf8((U8*)l) : toLOWER_LC_utf8((U8*)l)))
2168 s += UTF ? UTF8SKIP(s) : 1;
2172 nextchr = UCHARAT(locinput);
2176 /* Inline the first character, for speed. */
2177 if (UCHARAT(s) != nextchr &&
2178 UCHARAT(s) != ((OP(scan) == EXACTF)
2179 ? PL_fold : PL_fold_locale)[nextchr])
2181 if (PL_regeol - locinput < ln)
2183 if (ln > 1 && (OP(scan) == EXACTF
2184 ? ibcmp(s, locinput, ln)
2185 : ibcmp_locale(s, locinput, ln)))
2188 nextchr = UCHARAT(locinput);
2192 if (!reginclass(scan, (U8*)locinput, do_utf8))
2194 if (locinput >= PL_regeol)
2196 locinput += PL_utf8skip[nextchr];
2197 nextchr = UCHARAT(locinput);
2201 nextchr = UCHARAT(locinput);
2202 if (!reginclass(scan, (U8*)locinput, do_utf8))
2204 if (!nextchr && locinput >= PL_regeol)
2206 nextchr = UCHARAT(++locinput);
2210 PL_reg_flags |= RF_tainted;
2216 if (!(OP(scan) == ALNUM
2217 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2218 : isALNUM_LC_utf8((U8*)locinput)))
2222 locinput += PL_utf8skip[nextchr];
2223 nextchr = UCHARAT(locinput);
2226 if (!(OP(scan) == ALNUM
2227 ? isALNUM(nextchr) : isALNUM_LC(nextchr)))
2229 nextchr = UCHARAT(++locinput);
2232 PL_reg_flags |= RF_tainted;
2235 if (!nextchr && locinput >= PL_regeol)
2238 LOAD_UTF8_CHARCLASS(alnum,"a");
2239 if (OP(scan) == NALNUM
2240 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2241 : isALNUM_LC_utf8((U8*)locinput))
2245 locinput += PL_utf8skip[nextchr];
2246 nextchr = UCHARAT(locinput);
2249 if (OP(scan) == NALNUM
2250 ? isALNUM(nextchr) : isALNUM_LC(nextchr))
2252 nextchr = UCHARAT(++locinput);
2256 PL_reg_flags |= RF_tainted;
2260 /* was last char in word? */
2262 if (locinput == PL_regbol)
2265 U8 *r = reghop((U8*)locinput, -1);
2267 ln = utf8n_to_uvchr(r, s - (char*)r, 0, 0);
2269 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2270 ln = isALNUM_uni(ln);
2271 LOAD_UTF8_CHARCLASS(alnum,"a");
2272 n = swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8);
2275 ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(ln));
2276 n = isALNUM_LC_utf8((U8*)locinput);
2280 ln = (locinput != PL_regbol) ?
2281 UCHARAT(locinput - 1) : PL_regprev;
2282 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2284 n = isALNUM(nextchr);
2287 ln = isALNUM_LC(ln);
2288 n = isALNUM_LC(nextchr);
2291 if (((!ln) == (!n)) == (OP(scan) == BOUND ||
2292 OP(scan) == BOUNDL))
2296 PL_reg_flags |= RF_tainted;
2302 if (UTF8_IS_CONTINUED(nextchr)) {
2303 LOAD_UTF8_CHARCLASS(space," ");
2304 if (!(OP(scan) == SPACE
2305 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2306 : isSPACE_LC_utf8((U8*)locinput)))
2310 locinput += PL_utf8skip[nextchr];
2311 nextchr = UCHARAT(locinput);
2314 if (!(OP(scan) == SPACE
2315 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2317 nextchr = UCHARAT(++locinput);
2320 if (!(OP(scan) == SPACE
2321 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2323 nextchr = UCHARAT(++locinput);
2327 PL_reg_flags |= RF_tainted;
2330 if (!nextchr && locinput >= PL_regeol)
2333 LOAD_UTF8_CHARCLASS(space," ");
2334 if (OP(scan) == NSPACE
2335 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2336 : isSPACE_LC_utf8((U8*)locinput))
2340 locinput += PL_utf8skip[nextchr];
2341 nextchr = UCHARAT(locinput);
2344 if (OP(scan) == NSPACE
2345 ? isSPACE(nextchr) : isSPACE_LC(nextchr))
2347 nextchr = UCHARAT(++locinput);
2350 PL_reg_flags |= RF_tainted;
2356 LOAD_UTF8_CHARCLASS(digit,"0");
2357 if (!(OP(scan) == DIGIT
2358 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2359 : isDIGIT_LC_utf8((U8*)locinput)))
2363 locinput += PL_utf8skip[nextchr];
2364 nextchr = UCHARAT(locinput);
2367 if (!(OP(scan) == DIGIT
2368 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr)))
2370 nextchr = UCHARAT(++locinput);
2373 PL_reg_flags |= RF_tainted;
2376 if (!nextchr && locinput >= PL_regeol)
2379 LOAD_UTF8_CHARCLASS(digit,"0");
2380 if (OP(scan) == NDIGIT
2381 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2382 : isDIGIT_LC_utf8((U8*)locinput))
2386 locinput += PL_utf8skip[nextchr];
2387 nextchr = UCHARAT(locinput);
2390 if (OP(scan) == NDIGIT
2391 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr))
2393 nextchr = UCHARAT(++locinput);
2396 LOAD_UTF8_CHARCLASS(mark,"~");
2397 if (locinput >= PL_regeol ||
2398 swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2400 locinput += PL_utf8skip[nextchr];
2401 while (locinput < PL_regeol &&
2402 swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2403 locinput += UTF8SKIP(locinput);
2404 if (locinput > PL_regeol)
2406 nextchr = UCHARAT(locinput);
2409 PL_reg_flags |= RF_tainted;
2413 n = ARG(scan); /* which paren pair */
2414 ln = PL_regstartp[n];
2415 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2416 if (*PL_reglastparen < n || ln == -1)
2417 sayNO; /* Do not match unless seen CLOSEn. */
2418 if (ln == PL_regendp[n])
2422 if (do_utf8 && OP(scan) != REF) { /* REF can do byte comparison */
2424 char *e = PL_bostr + PL_regendp[n];
2426 * Note that we can't do the "other character" lookup trick as
2427 * in the 8-bit case (no pun intended) because in Unicode we
2428 * have to map both upper and title case to lower case.
2430 if (OP(scan) == REFF) {
2434 if (toLOWER_utf8((U8*)s) != toLOWER_utf8((U8*)l))
2444 if (toLOWER_LC_utf8((U8*)s) != toLOWER_LC_utf8((U8*)l))
2451 nextchr = UCHARAT(locinput);
2455 /* Inline the first character, for speed. */
2456 if (UCHARAT(s) != nextchr &&
2458 (UCHARAT(s) != ((OP(scan) == REFF
2459 ? PL_fold : PL_fold_locale)[nextchr]))))
2461 ln = PL_regendp[n] - ln;
2462 if (locinput + ln > PL_regeol)
2464 if (ln > 1 && (OP(scan) == REF
2465 ? memNE(s, locinput, ln)
2467 ? ibcmp(s, locinput, ln)
2468 : ibcmp_locale(s, locinput, ln))))
2471 nextchr = UCHARAT(locinput);
2482 OP_4tree *oop = PL_op;
2483 COP *ocurcop = PL_curcop;
2484 SV **ocurpad = PL_curpad;
2488 PL_op = (OP_4tree*)PL_regdata->data[n];
2489 DEBUG_r( PerlIO_printf(Perl_debug_log, " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) );
2490 PL_curpad = AvARRAY((AV*)PL_regdata->data[n + 2]);
2491 PL_regendp[0] = PL_reg_magic->mg_len = locinput - PL_bostr;
2493 CALLRUNOPS(aTHX); /* Scalar context. */
2499 PL_curpad = ocurpad;
2500 PL_curcop = ocurcop;
2502 if (logical == 2) { /* Postponed subexpression. */
2504 MAGIC *mg = Null(MAGIC*);
2506 CHECKPOINT cp, lastcp;
2508 if(SvROK(ret) || SvRMAGICAL(ret)) {
2509 SV *sv = SvROK(ret) ? SvRV(ret) : ret;
2512 mg = mg_find(sv, 'r');
2515 re = (regexp *)mg->mg_obj;
2516 (void)ReREFCNT_inc(re);
2520 char *t = SvPV(ret, len);
2522 char *oprecomp = PL_regprecomp;
2523 I32 osize = PL_regsize;
2524 I32 onpar = PL_regnpar;
2527 re = CALLREGCOMP(aTHX_ t, t + len, &pm);
2529 & (SVs_TEMP | SVs_PADTMP | SVf_READONLY)))
2530 sv_magic(ret,(SV*)ReREFCNT_inc(re),'r',0,0);
2531 PL_regprecomp = oprecomp;
2536 PerlIO_printf(Perl_debug_log,
2537 "Entering embedded `%s%.60s%s%s'\n",
2541 (strlen(re->precomp) > 60 ? "..." : ""))
2544 state.prev = PL_reg_call_cc;
2545 state.cc = PL_regcc;
2546 state.re = PL_reg_re;
2550 cp = regcppush(0); /* Save *all* the positions. */
2553 state.ss = PL_savestack_ix;
2554 *PL_reglastparen = 0;
2555 PL_reg_call_cc = &state;
2556 PL_reginput = locinput;
2558 /* XXXX This is too dramatic a measure... */
2561 if (regmatch(re->program + 1)) {
2562 /* Even though we succeeded, we need to restore
2563 global variables, since we may be wrapped inside
2564 SUSPEND, thus the match may be not finished yet. */
2566 /* XXXX Do this only if SUSPENDed? */
2567 PL_reg_call_cc = state.prev;
2568 PL_regcc = state.cc;
2569 PL_reg_re = state.re;
2570 cache_re(PL_reg_re);
2572 /* XXXX This is too dramatic a measure... */
2575 /* These are needed even if not SUSPEND. */
2581 REGCP_UNWIND(lastcp);
2583 PL_reg_call_cc = state.prev;
2584 PL_regcc = state.cc;
2585 PL_reg_re = state.re;
2586 cache_re(PL_reg_re);
2588 /* XXXX This is too dramatic a measure... */
2597 sv_setsv(save_scalar(PL_replgv), ret);
2601 n = ARG(scan); /* which paren pair */
2602 PL_reg_start_tmp[n] = locinput;
2607 n = ARG(scan); /* which paren pair */
2608 PL_regstartp[n] = PL_reg_start_tmp[n] - PL_bostr;
2609 PL_regendp[n] = locinput - PL_bostr;
2610 if (n > *PL_reglastparen)
2611 *PL_reglastparen = n;
2614 n = ARG(scan); /* which paren pair */
2615 sw = (*PL_reglastparen >= n && PL_regendp[n] != -1);
2618 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2620 next = NEXTOPER(NEXTOPER(scan));
2622 next = scan + ARG(scan);
2623 if (OP(next) == IFTHEN) /* Fake one. */
2624 next = NEXTOPER(NEXTOPER(next));
2628 logical = scan->flags;
2630 /*******************************************************************
2631 PL_regcc contains infoblock about the innermost (...)* loop, and
2632 a pointer to the next outer infoblock.
2634 Here is how Y(A)*Z is processed (if it is compiled into CURLYX/WHILEM):
2636 1) After matching X, regnode for CURLYX is processed;
2638 2) This regnode creates infoblock on the stack, and calls
2639 regmatch() recursively with the starting point at WHILEM node;
2641 3) Each hit of WHILEM node tries to match A and Z (in the order
2642 depending on the current iteration, min/max of {min,max} and
2643 greediness). The information about where are nodes for "A"
2644 and "Z" is read from the infoblock, as is info on how many times "A"
2645 was already matched, and greediness.
2647 4) After A matches, the same WHILEM node is hit again.
2649 5) Each time WHILEM is hit, PL_regcc is the infoblock created by CURLYX
2650 of the same pair. Thus when WHILEM tries to match Z, it temporarily
2651 resets PL_regcc, since this Y(A)*Z can be a part of some other loop:
2652 as in (Y(A)*Z)*. If Z matches, the automaton will hit the WHILEM node
2653 of the external loop.
2655 Currently present infoblocks form a tree with a stem formed by PL_curcc
2656 and whatever it mentions via ->next, and additional attached trees
2657 corresponding to temporarily unset infoblocks as in "5" above.
2659 In the following picture infoblocks for outer loop of
2660 (Y(A)*?Z)*?T are denoted O, for inner I. NULL starting block
2661 is denoted by x. The matched string is YAAZYAZT. Temporarily postponed
2662 infoblocks are drawn below the "reset" infoblock.
2664 In fact in the picture below we do not show failed matches for Z and T
2665 by WHILEM blocks. [We illustrate minimal matches, since for them it is
2666 more obvious *why* one needs to *temporary* unset infoblocks.]
2668 Matched REx position InfoBlocks Comment
2672 Y A)*?Z)*?T x <- O <- I
2673 YA )*?Z)*?T x <- O <- I
2674 YA A)*?Z)*?T x <- O <- I
2675 YAA )*?Z)*?T x <- O <- I
2676 YAA Z)*?T x <- O # Temporary unset I
2679 YAAZ Y(A)*?Z)*?T x <- O
2682 YAAZY (A)*?Z)*?T x <- O
2685 YAAZY A)*?Z)*?T x <- O <- I
2688 YAAZYA )*?Z)*?T x <- O <- I
2691 YAAZYA Z)*?T x <- O # Temporary unset I
2697 YAAZYAZ T x # Temporary unset O
2704 *******************************************************************/
2707 CHECKPOINT cp = PL_savestack_ix;
2708 /* No need to save/restore up to this paren */
2709 I32 parenfloor = scan->flags;
2711 if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */
2713 cc.oldcc = PL_regcc;
2715 /* XXXX Probably it is better to teach regpush to support
2716 parenfloor > PL_regsize... */
2717 if (parenfloor > *PL_reglastparen)
2718 parenfloor = *PL_reglastparen; /* Pessimization... */
2719 cc.parenfloor = parenfloor;
2721 cc.min = ARG1(scan);
2722 cc.max = ARG2(scan);
2723 cc.scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
2727 PL_reginput = locinput;
2728 n = regmatch(PREVOPER(next)); /* start on the WHILEM */
2730 PL_regcc = cc.oldcc;
2736 * This is really hard to understand, because after we match
2737 * what we're trying to match, we must make sure the rest of
2738 * the REx is going to match for sure, and to do that we have
2739 * to go back UP the parse tree by recursing ever deeper. And
2740 * if it fails, we have to reset our parent's current state
2741 * that we can try again after backing off.
2744 CHECKPOINT cp, lastcp;
2745 CURCUR* cc = PL_regcc;
2746 char *lastloc = cc->lastloc; /* Detection of 0-len. */
2748 n = cc->cur + 1; /* how many we know we matched */
2749 PL_reginput = locinput;
2752 PerlIO_printf(Perl_debug_log,
2753 "%*s %ld out of %ld..%ld cc=%lx\n",
2754 REPORT_CODE_OFF+PL_regindent*2, "",
2755 (long)n, (long)cc->min,
2756 (long)cc->max, (long)cc)
2759 /* If degenerate scan matches "", assume scan done. */
2761 if (locinput == cc->lastloc && n >= cc->min) {
2762 PL_regcc = cc->oldcc;
2766 PerlIO_printf(Perl_debug_log,
2767 "%*s empty match detected, try continuation...\n",
2768 REPORT_CODE_OFF+PL_regindent*2, "")
2770 if (regmatch(cc->next))
2778 /* First just match a string of min scans. */
2782 cc->lastloc = locinput;
2783 if (regmatch(cc->scan))
2786 cc->lastloc = lastloc;
2791 /* Check whether we already were at this position.
2792 Postpone detection until we know the match is not
2793 *that* much linear. */
2794 if (!PL_reg_maxiter) {
2795 PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4);
2796 PL_reg_leftiter = PL_reg_maxiter;
2798 if (PL_reg_leftiter-- == 0) {
2799 I32 size = (PL_reg_maxiter + 7)/8;
2800 if (PL_reg_poscache) {
2801 if (PL_reg_poscache_size < size) {
2802 Renew(PL_reg_poscache, size, char);
2803 PL_reg_poscache_size = size;
2805 Zero(PL_reg_poscache, size, char);
2808 PL_reg_poscache_size = size;
2809 Newz(29, PL_reg_poscache, size, char);
2812 PerlIO_printf(Perl_debug_log,
2813 "%sDetected a super-linear match, switching on caching%s...\n",
2814 PL_colors[4], PL_colors[5])
2817 if (PL_reg_leftiter < 0) {
2818 I32 o = locinput - PL_bostr, b;
2820 o = (scan->flags & 0xf) - 1 + o * (scan->flags>>4);
2823 if (PL_reg_poscache[o] & (1<<b)) {
2825 PerlIO_printf(Perl_debug_log,
2826 "%*s already tried at this position...\n",
2827 REPORT_CODE_OFF+PL_regindent*2, "")
2831 PL_reg_poscache[o] |= (1<<b);
2835 /* Prefer next over scan for minimal matching. */
2838 PL_regcc = cc->oldcc;
2841 cp = regcppush(cc->parenfloor);
2843 if (regmatch(cc->next)) {
2845 sayYES; /* All done. */
2847 REGCP_UNWIND(lastcp);
2853 if (n >= cc->max) { /* Maximum greed exceeded? */
2854 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
2855 && !(PL_reg_flags & RF_warned)) {
2856 PL_reg_flags |= RF_warned;
2857 Perl_warner(aTHX_ WARN_REGEXP, "%s limit (%d) exceeded",
2858 "Complex regular subexpression recursion",
2865 PerlIO_printf(Perl_debug_log,
2866 "%*s trying longer...\n",
2867 REPORT_CODE_OFF+PL_regindent*2, "")
2869 /* Try scanning more and see if it helps. */
2870 PL_reginput = locinput;
2872 cc->lastloc = locinput;
2873 cp = regcppush(cc->parenfloor);
2875 if (regmatch(cc->scan)) {
2879 REGCP_UNWIND(lastcp);
2882 cc->lastloc = lastloc;
2886 /* Prefer scan over next for maximal matching. */
2888 if (n < cc->max) { /* More greed allowed? */
2889 cp = regcppush(cc->parenfloor);
2891 cc->lastloc = locinput;
2893 if (regmatch(cc->scan)) {
2897 REGCP_UNWIND(lastcp);
2898 regcppop(); /* Restore some previous $<digit>s? */
2899 PL_reginput = locinput;
2901 PerlIO_printf(Perl_debug_log,
2902 "%*s failed, try continuation...\n",
2903 REPORT_CODE_OFF+PL_regindent*2, "")
2906 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
2907 && !(PL_reg_flags & RF_warned)) {
2908 PL_reg_flags |= RF_warned;
2909 Perl_warner(aTHX_ WARN_REGEXP, "%s limit (%d) exceeded",
2910 "Complex regular subexpression recursion",
2914 /* Failed deeper matches of scan, so see if this one works. */
2915 PL_regcc = cc->oldcc;
2918 if (regmatch(cc->next))
2924 cc->lastloc = lastloc;
2929 next = scan + ARG(scan);
2932 inner = NEXTOPER(NEXTOPER(scan));
2935 inner = NEXTOPER(scan);
2940 if (OP(next) != c1) /* No choice. */
2941 next = inner; /* Avoid recursion. */
2943 I32 lastparen = *PL_reglastparen;
2945 re_unwind_branch_t *uw;
2947 /* Put unwinding data on stack */
2948 unwind1 = SSNEWt(1,re_unwind_branch_t);
2949 uw = SSPTRt(unwind1,re_unwind_branch_t);
2952 uw->type = ((c1 == BRANCH)
2954 : RE_UNWIND_BRANCHJ);
2955 uw->lastparen = lastparen;
2957 uw->locinput = locinput;
2958 uw->nextchr = nextchr;
2960 uw->regindent = ++PL_regindent;
2963 REGCP_SET(uw->lastcp);
2965 /* Now go into the first branch */
2978 /* We suppose that the next guy does not need
2979 backtracking: in particular, it is of constant length,
2980 and has no parenths to influence future backrefs. */
2981 ln = ARG1(scan); /* min to match */
2982 n = ARG2(scan); /* max to match */
2983 paren = scan->flags;
2985 if (paren > PL_regsize)
2987 if (paren > *PL_reglastparen)
2988 *PL_reglastparen = paren;
2990 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
2992 scan += NEXT_OFF(scan); /* Skip former OPEN. */
2993 PL_reginput = locinput;
2996 if (ln && regrepeat_hard(scan, ln, &l) < ln)
2998 if (ln && l == 0 && n >= ln
2999 /* In fact, this is tricky. If paren, then the
3000 fact that we did/didnot match may influence
3001 future execution. */
3002 && !(paren && ln == 0))
3004 locinput = PL_reginput;
3005 if (PL_regkind[(U8)OP(next)] == EXACT) {
3006 c1 = (U8)*STRING(next);
3007 if (OP(next) == EXACTF)
3009 else if (OP(next) == EXACTFL)
3010 c2 = PL_fold_locale[c1];
3017 /* This may be improved if l == 0. */
3018 while (n >= ln || (n == REG_INFTY && ln > 0 && l)) { /* ln overflow ? */
3019 /* If it could work, try it. */
3021 UCHARAT(PL_reginput) == c1 ||
3022 UCHARAT(PL_reginput) == c2)
3026 PL_regstartp[paren] =
3027 HOPc(PL_reginput, -l) - PL_bostr;
3028 PL_regendp[paren] = PL_reginput - PL_bostr;
3031 PL_regendp[paren] = -1;
3035 REGCP_UNWIND(lastcp);
3037 /* Couldn't or didn't -- move forward. */
3038 PL_reginput = locinput;
3039 if (regrepeat_hard(scan, 1, &l)) {
3041 locinput = PL_reginput;
3048 n = regrepeat_hard(scan, n, &l);
3049 if (n != 0 && l == 0
3050 /* In fact, this is tricky. If paren, then the
3051 fact that we did/didnot match may influence
3052 future execution. */
3053 && !(paren && ln == 0))
3055 locinput = PL_reginput;
3057 PerlIO_printf(Perl_debug_log,
3058 "%*s matched %"IVdf" times, len=%"IVdf"...\n",
3059 (int)(REPORT_CODE_OFF+PL_regindent*2), "",
3063 if (PL_regkind[(U8)OP(next)] == EXACT) {
3064 c1 = (U8)*STRING(next);
3065 if (OP(next) == EXACTF)
3067 else if (OP(next) == EXACTFL)
3068 c2 = PL_fold_locale[c1];
3077 /* If it could work, try it. */
3079 UCHARAT(PL_reginput) == c1 ||
3080 UCHARAT(PL_reginput) == c2)
3083 PerlIO_printf(Perl_debug_log,
3084 "%*s trying tail with n=%"IVdf"...\n",
3085 (int)(REPORT_CODE_OFF+PL_regindent*2), "", (IV)n)
3089 PL_regstartp[paren] = HOPc(PL_reginput, -l) - PL_bostr;
3090 PL_regendp[paren] = PL_reginput - PL_bostr;
3093 PL_regendp[paren] = -1;
3097 REGCP_UNWIND(lastcp);
3099 /* Couldn't or didn't -- back up. */
3101 locinput = HOPc(locinput, -l);
3102 PL_reginput = locinput;
3109 paren = scan->flags; /* Which paren to set */
3110 if (paren > PL_regsize)
3112 if (paren > *PL_reglastparen)
3113 *PL_reglastparen = paren;
3114 ln = ARG1(scan); /* min to match */
3115 n = ARG2(scan); /* max to match */
3116 scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE);
3120 ln = ARG1(scan); /* min to match */
3121 n = ARG2(scan); /* max to match */
3122 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3127 scan = NEXTOPER(scan);
3133 scan = NEXTOPER(scan);
3137 * Lookahead to avoid useless match attempts
3138 * when we know what character comes next.
3140 if (PL_regkind[(U8)OP(next)] == EXACT) {
3141 U8 *s = (U8*)STRING(next);
3144 if (OP(next) == EXACTF)
3146 else if (OP(next) == EXACTFL)
3147 c2 = PL_fold_locale[c1];
3150 if (OP(next) == EXACTF) {
3151 c1 = to_utf8_lower(s);
3152 c2 = to_utf8_upper(s);
3155 c2 = c1 = utf8_to_uvchr(s, NULL);
3161 PL_reginput = locinput;
3165 if (ln && regrepeat(scan, ln) < ln)
3167 locinput = PL_reginput;
3170 char *e; /* Should not check after this */
3171 char *old = locinput;
3173 if (n == REG_INFTY) {
3176 while (UTF8_IS_CONTINUATION(*(U8*)e))
3182 m >0 && e + UTF8SKIP(e) <= PL_regeol; m--)
3186 e = locinput + n - ln;
3192 /* Find place 'next' could work */
3195 while (locinput <= e && *locinput != c1)
3198 while (locinput <= e
3203 count = locinput - old;
3210 utf8_to_uvchr((U8*)locinput, &len) != c1;
3215 for (count = 0; locinput <= e; count++) {
3216 UV c = utf8_to_uvchr((U8*)locinput, &len);
3217 if (c == c1 || c == c2)
3225 /* PL_reginput == old now */
3226 if (locinput != old) {
3227 ln = 1; /* Did some */
3228 if (regrepeat(scan, count) < count)
3231 /* PL_reginput == locinput now */
3232 TRYPAREN(paren, ln, locinput);
3233 PL_reginput = locinput; /* Could be reset... */
3234 REGCP_UNWIND(lastcp);
3235 /* Couldn't or didn't -- move forward. */
3238 locinput += UTF8SKIP(locinput);
3244 while (n >= ln || (n == REG_INFTY && ln > 0)) { /* ln overflow ? */
3248 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3250 c = UCHARAT(PL_reginput);
3251 /* If it could work, try it. */
3252 if (c == c1 || c == c2)
3254 TRYPAREN(paren, n, PL_reginput);
3255 REGCP_UNWIND(lastcp);
3258 /* If it could work, try it. */
3259 else if (c1 == -1000)
3261 TRYPAREN(paren, n, PL_reginput);
3262 REGCP_UNWIND(lastcp);
3264 /* Couldn't or didn't -- move forward. */
3265 PL_reginput = locinput;
3266 if (regrepeat(scan, 1)) {
3268 locinput = PL_reginput;
3276 n = regrepeat(scan, n);
3277 locinput = PL_reginput;
3278 if (ln < n && PL_regkind[(U8)OP(next)] == EOL &&
3279 (!PL_multiline || OP(next) == SEOL || OP(next) == EOS)) {
3280 ln = n; /* why back off? */
3281 /* ...because $ and \Z can match before *and* after
3282 newline at the end. Consider "\n\n" =~ /\n+\Z\n/.
3283 We should back off by one in this case. */
3284 if (UCHARAT(PL_reginput - 1) == '\n' && OP(next) != EOS)
3293 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3295 c = UCHARAT(PL_reginput);
3297 /* If it could work, try it. */
3298 if (c1 == -1000 || c == c1 || c == c2)
3300 TRYPAREN(paren, n, PL_reginput);
3301 REGCP_UNWIND(lastcp);
3303 /* Couldn't or didn't -- back up. */
3305 PL_reginput = locinput = HOPc(locinput, -1);
3313 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3315 c = UCHARAT(PL_reginput);
3317 /* If it could work, try it. */
3318 if (c1 == -1000 || c == c1 || c == c2)
3320 TRYPAREN(paren, n, PL_reginput);
3321 REGCP_UNWIND(lastcp);
3323 /* Couldn't or didn't -- back up. */
3325 PL_reginput = locinput = HOPc(locinput, -1);
3332 if (PL_reg_call_cc) {
3333 re_cc_state *cur_call_cc = PL_reg_call_cc;
3334 CURCUR *cctmp = PL_regcc;
3335 regexp *re = PL_reg_re;
3336 CHECKPOINT cp, lastcp;
3338 cp = regcppush(0); /* Save *all* the positions. */
3340 regcp_set_to(PL_reg_call_cc->ss); /* Restore parens of
3342 PL_reginput = locinput; /* Make position available to
3344 cache_re(PL_reg_call_cc->re);
3345 PL_regcc = PL_reg_call_cc->cc;
3346 PL_reg_call_cc = PL_reg_call_cc->prev;
3347 if (regmatch(cur_call_cc->node)) {
3348 PL_reg_call_cc = cur_call_cc;
3352 REGCP_UNWIND(lastcp);
3354 PL_reg_call_cc = cur_call_cc;
3360 PerlIO_printf(Perl_debug_log,
3361 "%*s continuation failed...\n",
3362 REPORT_CODE_OFF+PL_regindent*2, "")
3366 if (locinput < PL_regtill) {
3367 DEBUG_r(PerlIO_printf(Perl_debug_log,
3368 "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n",
3370 (long)(locinput - PL_reg_starttry),
3371 (long)(PL_regtill - PL_reg_starttry),
3373 sayNO_FINAL; /* Cannot match: too short. */
3375 PL_reginput = locinput; /* put where regtry can find it */
3376 sayYES_FINAL; /* Success! */
3378 PL_reginput = locinput; /* put where regtry can find it */
3379 sayYES_LOUD; /* Success! */
3382 PL_reginput = locinput;
3387 if (UTF) { /* XXXX This is absolutely
3388 broken, we read before
3390 s = HOPMAYBEc(locinput, -scan->flags);
3396 if (locinput < PL_bostr + scan->flags)
3398 PL_reginput = locinput - scan->flags;
3403 PL_reginput = locinput;
3408 if (UTF) { /* XXXX This is absolutely
3409 broken, we read before
3411 s = HOPMAYBEc(locinput, -scan->flags);
3412 if (!s || s < PL_bostr)
3417 if (locinput < PL_bostr + scan->flags)
3419 PL_reginput = locinput - scan->flags;
3424 PL_reginput = locinput;
3427 inner = NEXTOPER(NEXTOPER(scan));
3428 if (regmatch(inner) != n) {
3443 if (OP(scan) == SUSPEND) {
3444 locinput = PL_reginput;
3445 nextchr = UCHARAT(locinput);
3450 next = scan + ARG(scan);
3455 PerlIO_printf(Perl_error_log, "%"UVxf" %d\n",
3456 PTR2UV(scan), OP(scan));
3457 Perl_croak(aTHX_ "regexp memory corruption");
3464 * We get here only if there's trouble -- normally "case END" is
3465 * the terminating point.
3467 Perl_croak(aTHX_ "corrupted regexp pointers");
3473 PerlIO_printf(Perl_debug_log,
3474 "%*s %scould match...%s\n",
3475 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4],PL_colors[5])
3479 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n",
3480 PL_colors[4],PL_colors[5]));
3486 #if 0 /* Breaks $^R */
3494 PerlIO_printf(Perl_debug_log,
3495 "%*s %sfailed...%s\n",
3496 REPORT_CODE_OFF+PL_regindent*2, "",PL_colors[4],PL_colors[5])
3502 re_unwind_t *uw = SSPTRt(unwind,re_unwind_t);
3505 case RE_UNWIND_BRANCH:
3506 case RE_UNWIND_BRANCHJ:
3508 re_unwind_branch_t *uwb = &(uw->branch);
3509 I32 lastparen = uwb->lastparen;
3511 REGCP_UNWIND(uwb->lastcp);
3512 for (n = *PL_reglastparen; n > lastparen; n--)
3514 *PL_reglastparen = n;
3515 scan = next = uwb->next;
3517 OP(scan) != (uwb->type == RE_UNWIND_BRANCH
3518 ? BRANCH : BRANCHJ) ) { /* Failure */
3525 /* Have more choice yet. Reuse the same uwb. */
3527 if ((n = (uwb->type == RE_UNWIND_BRANCH
3528 ? NEXT_OFF(next) : ARG(next))))
3531 next = NULL; /* XXXX Needn't unwinding in this case... */
3533 next = NEXTOPER(scan);
3534 if (uwb->type == RE_UNWIND_BRANCHJ)
3535 next = NEXTOPER(next);
3536 locinput = uwb->locinput;
3537 nextchr = uwb->nextchr;
3539 PL_regindent = uwb->regindent;
3546 Perl_croak(aTHX_ "regexp unwind memory corruption");
3557 - regrepeat - repeatedly match something simple, report how many
3560 * [This routine now assumes that it will only match on things of length 1.
3561 * That was true before, but now we assume scan - reginput is the count,
3562 * rather than incrementing count on every character. [Er, except utf8.]]
3565 S_regrepeat(pTHX_ regnode *p, I32 max)
3567 register char *scan;
3569 register char *loceol = PL_regeol;
3570 register I32 hardcount = 0;
3571 register bool do_utf8 = DO_UTF8(PL_reg_sv);
3574 if (max != REG_INFTY && max < loceol - scan)
3575 loceol = scan + max;
3580 while (scan < loceol && hardcount < max && *scan != '\n') {
3581 scan += UTF8SKIP(scan);
3585 while (scan < loceol && *scan != '\n')
3592 case EXACT: /* length of string is 1 */
3594 while (scan < loceol && UCHARAT(scan) == c)
3597 case EXACTF: /* length of string is 1 */
3599 while (scan < loceol &&
3600 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold[c]))
3603 case EXACTFL: /* length of string is 1 */
3604 PL_reg_flags |= RF_tainted;
3606 while (scan < loceol &&
3607 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold_locale[c]))
3613 while (hardcount < max && scan < loceol &&
3614 reginclass(p, (U8*)scan, do_utf8)) {
3615 scan += UTF8SKIP(scan);
3619 while (scan < loceol && reginclass(p, (U8*)scan, do_utf8))
3626 LOAD_UTF8_CHARCLASS(alnum,"a");
3627 while (hardcount < max && scan < loceol &&
3628 swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
3629 scan += UTF8SKIP(scan);
3633 while (scan < loceol && isALNUM(*scan))
3638 PL_reg_flags |= RF_tainted;
3641 while (hardcount < max && scan < loceol &&
3642 isALNUM_LC_utf8((U8*)scan)) {
3643 scan += UTF8SKIP(scan);
3647 while (scan < loceol && isALNUM_LC(*scan))
3654 LOAD_UTF8_CHARCLASS(alnum,"a");
3655 while (hardcount < max && scan < loceol &&
3656 !swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
3657 scan += UTF8SKIP(scan);
3661 while (scan < loceol && !isALNUM(*scan))
3666 PL_reg_flags |= RF_tainted;
3669 while (hardcount < max && scan < loceol &&
3670 !isALNUM_LC_utf8((U8*)scan)) {
3671 scan += UTF8SKIP(scan);
3675 while (scan < loceol && !isALNUM_LC(*scan))
3682 LOAD_UTF8_CHARCLASS(space," ");
3683 while (hardcount < max && scan < loceol &&
3685 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
3686 scan += UTF8SKIP(scan);
3690 while (scan < loceol && isSPACE(*scan))
3695 PL_reg_flags |= RF_tainted;
3698 while (hardcount < max && scan < loceol &&
3699 (*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
3700 scan += UTF8SKIP(scan);
3704 while (scan < loceol && isSPACE_LC(*scan))
3711 LOAD_UTF8_CHARCLASS(space," ");
3712 while (hardcount < max && scan < loceol &&
3714 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
3715 scan += UTF8SKIP(scan);
3719 while (scan < loceol && !isSPACE(*scan))
3724 PL_reg_flags |= RF_tainted;
3727 while (hardcount < max && scan < loceol &&
3728 !(*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
3729 scan += UTF8SKIP(scan);
3733 while (scan < loceol && !isSPACE_LC(*scan))
3740 LOAD_UTF8_CHARCLASS(digit,"0");
3741 while (hardcount < max && scan < loceol &&
3742 swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
3743 scan += UTF8SKIP(scan);
3747 while (scan < loceol && isDIGIT(*scan))
3754 LOAD_UTF8_CHARCLASS(digit,"0");
3755 while (hardcount < max && scan < loceol &&
3756 !swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
3757 scan += UTF8SKIP(scan);
3761 while (scan < loceol && !isDIGIT(*scan))
3765 default: /* Called on something of 0 width. */
3766 break; /* So match right here or not at all. */
3772 c = scan - PL_reginput;
3777 SV *prop = sv_newmortal();
3780 PerlIO_printf(Perl_debug_log,
3781 "%*s %s can match %"IVdf" times out of %"IVdf"...\n",
3782 REPORT_CODE_OFF+1, "", SvPVX(prop),(IV)c,(IV)max);
3789 - regrepeat_hard - repeatedly match something, report total lenth and length
3791 * The repeater is supposed to have constant length.
3795 S_regrepeat_hard(pTHX_ regnode *p, I32 max, I32 *lp)
3797 register char *scan;
3798 register char *start;
3799 register char *loceol = PL_regeol;
3801 I32 count = 0, res = 1;
3806 start = PL_reginput;
3807 if (DO_UTF8(PL_reg_sv)) {
3808 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
3811 while (start < PL_reginput) {
3813 start += UTF8SKIP(start);
3824 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
3826 *lp = l = PL_reginput - start;
3827 if (max != REG_INFTY && l*max < loceol - scan)
3828 loceol = scan + l*max;
3841 - regclass_swash - prepare the utf8 swash
3845 Perl_regclass_swash(pTHX_ register regnode* node, bool doinit, SV** initsvp)
3850 if (PL_regdata && PL_regdata->count) {
3853 if (PL_regdata->what[n] == 's') {
3854 SV *rv = (SV*)PL_regdata->data[n];
3855 AV *av = (AV*)SvRV((SV*)rv);
3858 si = *av_fetch(av, 0, FALSE);
3859 a = av_fetch(av, 1, FALSE);
3863 else if (si && doinit) {
3864 sw = swash_init("utf8", "", si, 1, 0);
3865 (void)av_store(av, 1, sw);
3877 - reginclass - determine if a character falls into a character class
3881 S_reginclass(pTHX_ register regnode *n, register U8* p, register bool do_utf8)
3883 char flags = ANYOF_FLAGS(n);
3888 c = do_utf8 ? utf8_to_uvchr(p, &len) : *p;
3890 if (do_utf8 || (flags & ANYOF_UNICODE)) {
3891 if (do_utf8 && !ANYOF_RUNTIME(n)) {
3892 if (len != (STRLEN)-1 && c < 256 && ANYOF_BITMAP_TEST(n, c))
3895 if (!match && do_utf8 && (flags & ANYOF_UNICODE_ALL) && c >= 256)
3898 SV *sw = regclass_swash(n, TRUE, 0);
3901 if (swash_fetch(sw, p, do_utf8))
3903 else if (flags & ANYOF_FOLD) {
3904 U8 tmpbuf[UTF8_MAXLEN+1];
3906 if (flags & ANYOF_LOCALE) {
3907 PL_reg_flags |= RF_tainted;
3908 uvchr_to_utf8(tmpbuf, toLOWER_LC_utf8(p));
3911 uvchr_to_utf8(tmpbuf, toLOWER_utf8(p));
3912 if (swash_fetch(sw, tmpbuf, do_utf8))
3918 if (!match && c < 256) {
3919 if (ANYOF_BITMAP_TEST(n, c))
3921 else if (flags & ANYOF_FOLD) {
3924 if (flags & ANYOF_LOCALE) {
3925 PL_reg_flags |= RF_tainted;
3926 f = PL_fold_locale[c];
3930 if (f != c && ANYOF_BITMAP_TEST(n, f))
3934 if (!match && (flags & ANYOF_CLASS)) {
3935 PL_reg_flags |= RF_tainted;
3937 (ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) ||
3938 (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) ||
3939 (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) ||
3940 (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) ||
3941 (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) ||
3942 (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) ||
3943 (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) ||
3944 (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) ||
3945 (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) ||
3946 (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) ||
3947 (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII(c)) ||
3948 (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII(c)) ||
3949 (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) ||
3950 (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) ||
3951 (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) ||
3952 (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) ||
3953 (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) ||
3954 (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) ||
3955 (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) ||
3956 (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) ||
3957 (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) ||
3958 (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) ||
3959 (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) ||
3960 (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) ||
3961 (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) ||
3962 (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) ||
3963 (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) ||
3964 (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) ||
3965 (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK(c)) ||
3966 (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK(c))
3967 ) /* How's that for a conditional? */
3974 return (flags & ANYOF_INVERT) ? !match : match;
3978 S_reghop(pTHX_ U8 *s, I32 off)
3980 return S_reghop3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
3984 S_reghop3(pTHX_ U8 *s, I32 off, U8* lim)
3987 while (off-- && s < lim) {
3988 /* XXX could check well-formedness here */
3996 if (UTF8_IS_CONTINUED(*s)) {
3997 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4000 /* XXX could check well-formedness here */
4008 S_reghopmaybe(pTHX_ U8 *s, I32 off)
4010 return S_reghopmaybe3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4014 S_reghopmaybe3(pTHX_ U8* s, I32 off, U8* lim)
4017 while (off-- && s < lim) {
4018 /* XXX could check well-formedness here */
4028 if (UTF8_IS_CONTINUED(*s)) {
4029 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4032 /* XXX could check well-formedness here */
4048 restore_pos(pTHXo_ void *arg)
4050 if (PL_reg_eval_set) {
4051 if (PL_reg_oldsaved) {
4052 PL_reg_re->subbeg = PL_reg_oldsaved;
4053 PL_reg_re->sublen = PL_reg_oldsavedlen;
4054 RX_MATCH_COPIED_on(PL_reg_re);
4056 PL_reg_magic->mg_len = PL_reg_oldpos;
4057 PL_reg_eval_set = 0;
4058 PL_curpm = PL_reg_oldcurpm;