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
85 #define RF_tainted 1 /* tainted information used? */
86 #define RF_warned 2 /* warned about big count? */
87 #define RF_evaled 4 /* Did an EVAL with setting? */
88 #define RF_utf8 8 /* String contains multibyte chars? */
90 #define UTF (PL_reg_flags & RF_utf8)
92 #define RS_init 1 /* eval environment created */
93 #define RS_set 2 /* replsv value is set */
103 #define CHR_SVLEN(sv) (UTF ? sv_len_utf8(sv) : SvCUR(sv))
104 #define CHR_DIST(a,b) (PL_reg_match_utf8 ? utf8_distance(a,b) : a - b)
106 #define reghop_c(pos,off) ((char*)reghop((U8*)pos, off))
107 #define reghopmaybe_c(pos,off) ((char*)reghopmaybe((U8*)pos, off))
108 #define HOP(pos,off) (PL_reg_match_utf8 ? reghop((U8*)pos, off) : (U8*)(pos + off))
109 #define HOPMAYBE(pos,off) (PL_reg_match_utf8 ? reghopmaybe((U8*)pos, off) : (U8*)(pos + off))
110 #define HOPc(pos,off) ((char*)HOP(pos,off))
111 #define HOPMAYBEc(pos,off) ((char*)HOPMAYBE(pos,off))
113 #define HOPBACK(pos, off) ( \
114 (UTF && PL_reg_match_utf8) \
115 ? reghopmaybe((U8*)pos, -off) \
116 : (pos - off >= PL_bostr) \
120 #define HOPBACKc(pos, off) (char*)HOPBACK(pos, off)
122 #define reghop3_c(pos,off,lim) ((char*)reghop3((U8*)pos, off, (U8*)lim))
123 #define reghopmaybe3_c(pos,off,lim) ((char*)reghopmaybe3((U8*)pos, off, (U8*)lim))
124 #define HOP3(pos,off,lim) (PL_reg_match_utf8 ? reghop3((U8*)pos, off, (U8*)lim) : (U8*)(pos + off))
125 #define HOPMAYBE3(pos,off,lim) (PL_reg_match_utf8 ? reghopmaybe3((U8*)pos, off, (U8*)lim) : (U8*)(pos + off))
126 #define HOP3c(pos,off,lim) ((char*)HOP3(pos,off,lim))
127 #define HOPMAYBE3c(pos,off,lim) ((char*)HOPMAYBE3(pos,off,lim))
129 #define LOAD_UTF8_CHARCLASS(a,b) STMT_START { if (!CAT2(PL_utf8_,a)) (void)CAT2(is_utf8_, a)((U8*)b); } STMT_END
131 /* for use after a quantifier and before an EXACT-like node -- japhy */
132 #define JUMPABLE(rn) ( \
133 OP(rn) == OPEN || OP(rn) == CLOSE || OP(rn) == EVAL || \
134 OP(rn) == SUSPEND || OP(rn) == IFMATCH \
137 #define NEAR_EXACT(rn) (PL_regkind[(U8)OP(rn)] == EXACT || JUMPABLE(rn))
139 #define NEXT_IMPT(rn) STMT_START { \
140 while (JUMPABLE(rn)) \
141 if (OP(rn) == SUSPEND || OP(rn) == IFMATCH) \
142 rn = NEXTOPER(NEXTOPER(rn)); \
143 else rn += NEXT_OFF(rn); \
146 static void restore_pos(pTHX_ void *arg);
149 S_regcppush(pTHX_ I32 parenfloor)
151 int retval = PL_savestack_ix;
152 #define REGCP_PAREN_ELEMS 4
153 int paren_elems_to_push = (PL_regsize - parenfloor) * REGCP_PAREN_ELEMS;
156 if (paren_elems_to_push < 0)
157 Perl_croak(aTHX_ "panic: paren_elems_to_push < 0");
159 #define REGCP_OTHER_ELEMS 6
160 SSCHECK(paren_elems_to_push + REGCP_OTHER_ELEMS);
161 for (p = PL_regsize; p > parenfloor; p--) {
162 /* REGCP_PARENS_ELEMS are pushed per pairs of parentheses. */
163 SSPUSHINT(PL_regendp[p]);
164 SSPUSHINT(PL_regstartp[p]);
165 SSPUSHPTR(PL_reg_start_tmp[p]);
168 /* REGCP_OTHER_ELEMS are pushed in any case, parentheses or no. */
169 SSPUSHINT(PL_regsize);
170 SSPUSHINT(*PL_reglastparen);
171 SSPUSHINT(*PL_reglastcloseparen);
172 SSPUSHPTR(PL_reginput);
173 #define REGCP_FRAME_ELEMS 2
174 /* REGCP_FRAME_ELEMS are part of the REGCP_OTHER_ELEMS and
175 * are needed for the regexp context stack bookkeeping. */
176 SSPUSHINT(paren_elems_to_push + REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS);
177 SSPUSHINT(SAVEt_REGCONTEXT); /* Magic cookie. */
182 /* These are needed since we do not localize EVAL nodes: */
183 # define REGCP_SET(cp) DEBUG_r(PerlIO_printf(Perl_debug_log, \
184 " Setting an EVAL scope, savestack=%"IVdf"\n", \
185 (IV)PL_savestack_ix)); cp = PL_savestack_ix
187 # define REGCP_UNWIND(cp) DEBUG_r(cp != PL_savestack_ix ? \
188 PerlIO_printf(Perl_debug_log, \
189 " Clearing an EVAL scope, savestack=%"IVdf"..%"IVdf"\n", \
190 (IV)(cp), (IV)PL_savestack_ix) : 0); regcpblow(cp)
200 /* Pop REGCP_OTHER_ELEMS before the parentheses loop starts. */
202 assert(i == SAVEt_REGCONTEXT); /* Check that the magic cookie is there. */
203 i = SSPOPINT; /* Parentheses elements to pop. */
204 input = (char *) SSPOPPTR;
205 *PL_reglastcloseparen = SSPOPINT;
206 *PL_reglastparen = SSPOPINT;
207 PL_regsize = SSPOPINT;
209 /* Now restore the parentheses context. */
210 for (i -= (REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS);
211 i > 0; i -= REGCP_PAREN_ELEMS) {
212 paren = (U32)SSPOPINT;
213 PL_reg_start_tmp[paren] = (char *) SSPOPPTR;
214 PL_regstartp[paren] = SSPOPINT;
216 if (paren <= *PL_reglastparen)
217 PL_regendp[paren] = tmps;
219 PerlIO_printf(Perl_debug_log,
220 " restoring \\%"UVuf" to %"IVdf"(%"IVdf")..%"IVdf"%s\n",
221 (UV)paren, (IV)PL_regstartp[paren],
222 (IV)(PL_reg_start_tmp[paren] - PL_bostr),
223 (IV)PL_regendp[paren],
224 (paren > *PL_reglastparen ? "(no)" : ""));
228 if (*PL_reglastparen + 1 <= PL_regnpar) {
229 PerlIO_printf(Perl_debug_log,
230 " restoring \\%"IVdf"..\\%"IVdf" to undef\n",
231 (IV)(*PL_reglastparen + 1), (IV)PL_regnpar);
235 /* It would seem that the similar code in regtry()
236 * already takes care of this, and in fact it is in
237 * a better location to since this code can #if 0-ed out
238 * but the code in regtry() is needed or otherwise tests
239 * requiring null fields (pat.t#187 and split.t#{13,14}
240 * (as of patchlevel 7877) will fail. Then again,
241 * this code seems to be necessary or otherwise
242 * building DynaLoader will fail:
243 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
245 for (paren = *PL_reglastparen + 1; paren <= PL_regnpar; paren++) {
246 if (paren > PL_regsize)
247 PL_regstartp[paren] = -1;
248 PL_regendp[paren] = -1;
255 S_regcp_set_to(pTHX_ I32 ss)
257 I32 tmp = PL_savestack_ix;
259 PL_savestack_ix = ss;
261 PL_savestack_ix = tmp;
265 typedef struct re_cc_state
269 struct re_cc_state *prev;
274 #define regcpblow(cp) LEAVE_SCOPE(cp) /* Ignores regcppush()ed data. */
276 #define TRYPAREN(paren, n, input) { \
279 PL_regstartp[paren] = HOPc(input, -1) - PL_bostr; \
280 PL_regendp[paren] = input - PL_bostr; \
283 PL_regendp[paren] = -1; \
285 if (regmatch(next)) \
288 PL_regendp[paren] = -1; \
293 * pregexec and friends
297 - pregexec - match a regexp against a string
300 Perl_pregexec(pTHX_ register regexp *prog, char *stringarg, register char *strend,
301 char *strbeg, I32 minend, SV *screamer, U32 nosave)
302 /* strend: pointer to null at end of string */
303 /* strbeg: real beginning of string */
304 /* minend: end of match must be >=minend after stringarg. */
305 /* nosave: For optimizations. */
308 regexec_flags(prog, stringarg, strend, strbeg, minend, screamer, NULL,
309 nosave ? 0 : REXEC_COPY_STR);
313 S_cache_re(pTHX_ regexp *prog)
315 PL_regprecomp = prog->precomp; /* Needed for FAIL. */
317 PL_regprogram = prog->program;
319 PL_regnpar = prog->nparens;
320 PL_regdata = prog->data;
325 * Need to implement the following flags for reg_anch:
327 * USE_INTUIT_NOML - Useful to call re_intuit_start() first
329 * INTUIT_AUTORITATIVE_NOML - Can trust a positive answer
330 * INTUIT_AUTORITATIVE_ML
331 * INTUIT_ONCE_NOML - Intuit can match in one location only.
334 * Another flag for this function: SECOND_TIME (so that float substrs
335 * with giant delta may be not rechecked).
338 /* Assumptions: if ANCH_GPOS, then strpos is anchored. XXXX Check GPOS logic */
340 /* If SCREAM, then SvPVX(sv) should be compatible with strpos and strend.
341 Otherwise, only SvCUR(sv) is used to get strbeg. */
343 /* XXXX We assume that strpos is strbeg unless sv. */
345 /* XXXX Some places assume that there is a fixed substring.
346 An update may be needed if optimizer marks as "INTUITable"
347 RExen without fixed substrings. Similarly, it is assumed that
348 lengths of all the strings are no more than minlen, thus they
349 cannot come from lookahead.
350 (Or minlen should take into account lookahead.) */
352 /* A failure to find a constant substring means that there is no need to make
353 an expensive call to REx engine, thus we celebrate a failure. Similarly,
354 finding a substring too deep into the string means that less calls to
355 regtry() should be needed.
357 REx compiler's optimizer found 4 possible hints:
358 a) Anchored substring;
360 c) Whether we are anchored (beginning-of-line or \G);
361 d) First node (of those at offset 0) which may distingush positions;
362 We use a)b)d) and multiline-part of c), and try to find a position in the
363 string which does not contradict any of them.
366 /* Most of decisions we do here should have been done at compile time.
367 The nodes of the REx which we used for the search should have been
368 deleted from the finite automaton. */
371 Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos,
372 char *strend, U32 flags, re_scream_pos_data *data)
374 register I32 start_shift = 0;
375 /* Should be nonnegative! */
376 register I32 end_shift = 0;
382 register char *other_last = Nullch; /* other substr checked before this */
383 char *check_at = Nullch; /* check substr found at this pos */
385 char *i_strpos = strpos;
386 SV *dsv = sv_2mortal(newSVpvn("", 0));
390 char *s = UTF ? sv_uni_display(dsv, sv, 60, 0) : strpos;
391 STRLEN len = UTF ? strlen(s) : strend - strpos;
394 PerlIO_printf(Perl_debug_log,
395 "%sGuessing start of match, REx%s `%s%.60s%s%s' against `%s%.*s%s%s'...\n",
396 PL_colors[4],PL_colors[5],PL_colors[0],
399 (strlen(prog->precomp) > 60 ? "..." : ""),
401 (int)(len > 60 ? 60 : len),
403 (len > 60 ? "..." : "")
407 if (prog->reganch & ROPT_UTF8)
408 PL_reg_flags |= RF_utf8;
410 if (prog->minlen > CHR_DIST((U8*)strend, (U8*)strpos)) {
411 DEBUG_r(PerlIO_printf(Perl_debug_log, "String too short...\n"));
414 strbeg = (sv && SvPOK(sv)) ? strend - SvCUR(sv) : strpos;
416 check = prog->check_substr;
417 if (prog->reganch & ROPT_ANCH) { /* Match at beg-of-str or after \n */
418 ml_anch = !( (prog->reganch & ROPT_ANCH_SINGLE)
419 || ( (prog->reganch & ROPT_ANCH_BOL)
420 && !PL_multiline ) ); /* Check after \n? */
423 if ( !(prog->reganch & (ROPT_ANCH_GPOS /* Checked by the caller */
424 | ROPT_IMPLICIT)) /* not a real BOL */
425 /* SvCUR is not set on references: SvRV and SvPVX overlap */
427 && (strpos != strbeg)) {
428 DEBUG_r(PerlIO_printf(Perl_debug_log, "Not at start...\n"));
431 if (prog->check_offset_min == prog->check_offset_max &&
432 !(prog->reganch & ROPT_CANY_SEEN)) {
433 /* Substring at constant offset from beg-of-str... */
436 s = HOP3c(strpos, prog->check_offset_min, strend);
438 slen = SvCUR(check); /* >= 1 */
440 if ( strend - s > slen || strend - s < slen - 1
441 || (strend - s == slen && strend[-1] != '\n')) {
442 DEBUG_r(PerlIO_printf(Perl_debug_log, "String too long...\n"));
445 /* Now should match s[0..slen-2] */
447 if (slen && (*SvPVX(check) != *s
449 && memNE(SvPVX(check), s, slen)))) {
451 DEBUG_r(PerlIO_printf(Perl_debug_log, "String not equal...\n"));
455 else if (*SvPVX(check) != *s
456 || ((slen = SvCUR(check)) > 1
457 && memNE(SvPVX(check), s, slen)))
459 goto success_at_start;
462 /* Match is anchored, but substr is not anchored wrt beg-of-str. */
464 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
465 end_shift = prog->minlen - start_shift -
466 CHR_SVLEN(check) + (SvTAIL(check) != 0);
468 I32 end = prog->check_offset_max + CHR_SVLEN(check)
469 - (SvTAIL(check) != 0);
470 I32 eshift = CHR_DIST((U8*)strend, (U8*)s) - end;
472 if (end_shift < eshift)
476 else { /* Can match at random position */
479 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
480 /* Should be nonnegative! */
481 end_shift = prog->minlen - start_shift -
482 CHR_SVLEN(check) + (SvTAIL(check) != 0);
485 #ifdef DEBUGGING /* 7/99: reports of failure (with the older version) */
487 Perl_croak(aTHX_ "panic: end_shift");
491 /* Find a possible match in the region s..strend by looking for
492 the "check" substring in the region corrected by start/end_shift. */
493 if (flags & REXEC_SCREAM) {
494 I32 p = -1; /* Internal iterator of scream. */
495 I32 *pp = data ? data->scream_pos : &p;
497 if (PL_screamfirst[BmRARE(check)] >= 0
498 || ( BmRARE(check) == '\n'
499 && (BmPREVIOUS(check) == SvCUR(check) - 1)
501 s = screaminstr(sv, check,
502 start_shift + (s - strbeg), end_shift, pp, 0);
506 *data->scream_olds = s;
508 else if (prog->reganch & ROPT_CANY_SEEN)
509 s = fbm_instr((U8*)(s + start_shift),
510 (U8*)(strend - end_shift),
511 check, PL_multiline ? FBMrf_MULTILINE : 0);
513 s = fbm_instr(HOP3(s, start_shift, strend),
514 HOP3(strend, -end_shift, strbeg),
515 check, PL_multiline ? FBMrf_MULTILINE : 0);
517 /* Update the count-of-usability, remove useless subpatterns,
520 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s %s substr `%s%.*s%s'%s%s",
521 (s ? "Found" : "Did not find"),
522 ((check == prog->anchored_substr) ? "anchored" : "floating"),
524 (int)(SvCUR(check) - (SvTAIL(check)!=0)),
526 PL_colors[1], (SvTAIL(check) ? "$" : ""),
527 (s ? " at offset " : "...\n") ) );
534 /* Finish the diagnostic message */
535 DEBUG_r(PerlIO_printf(Perl_debug_log, "%ld...\n", (long)(s - i_strpos)) );
537 /* Got a candidate. Check MBOL anchoring, and the *other* substr.
538 Start with the other substr.
539 XXXX no SCREAM optimization yet - and a very coarse implementation
540 XXXX /ttx+/ results in anchored=`ttx', floating=`x'. floating will
541 *always* match. Probably should be marked during compile...
542 Probably it is right to do no SCREAM here...
545 if (prog->float_substr && prog->anchored_substr) {
546 /* Take into account the "other" substring. */
547 /* XXXX May be hopelessly wrong for UTF... */
550 if (check == prog->float_substr) {
553 char *last = HOP3c(s, -start_shift, strbeg), *last1, *last2;
556 t = s - prog->check_offset_max;
557 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
558 && (!(prog->reganch & ROPT_UTF8)
559 || ((t = reghopmaybe3_c(s, -(prog->check_offset_max), strpos))
564 t = HOP3c(t, prog->anchored_offset, strend);
565 if (t < other_last) /* These positions already checked */
567 last2 = last1 = HOP3c(strend, -prog->minlen, strbeg);
570 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
571 /* On end-of-str: see comment below. */
572 s = fbm_instr((unsigned char*)t,
573 HOP3(HOP3(last1, prog->anchored_offset, strend)
574 + SvCUR(prog->anchored_substr),
575 -(SvTAIL(prog->anchored_substr)!=0), strbeg),
576 prog->anchored_substr,
577 PL_multiline ? FBMrf_MULTILINE : 0);
578 DEBUG_r(PerlIO_printf(Perl_debug_log,
579 "%s anchored substr `%s%.*s%s'%s",
580 (s ? "Found" : "Contradicts"),
582 (int)(SvCUR(prog->anchored_substr)
583 - (SvTAIL(prog->anchored_substr)!=0)),
584 SvPVX(prog->anchored_substr),
585 PL_colors[1], (SvTAIL(prog->anchored_substr) ? "$" : "")));
587 if (last1 >= last2) {
588 DEBUG_r(PerlIO_printf(Perl_debug_log,
589 ", giving up...\n"));
592 DEBUG_r(PerlIO_printf(Perl_debug_log,
593 ", trying floating at offset %ld...\n",
594 (long)(HOP3c(s1, 1, strend) - i_strpos)));
595 other_last = HOP3c(last1, prog->anchored_offset+1, strend);
596 s = HOP3c(last, 1, strend);
600 DEBUG_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
601 (long)(s - i_strpos)));
602 t = HOP3c(s, -prog->anchored_offset, strbeg);
603 other_last = HOP3c(s, 1, strend);
611 else { /* Take into account the floating substring. */
615 t = HOP3c(s, -start_shift, strbeg);
617 HOP3c(strend, -prog->minlen + prog->float_min_offset, strbeg);
618 if (CHR_DIST((U8*)last, (U8*)t) > prog->float_max_offset)
619 last = HOP3c(t, prog->float_max_offset, strend);
620 s = HOP3c(t, prog->float_min_offset, strend);
623 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
624 /* fbm_instr() takes into account exact value of end-of-str
625 if the check is SvTAIL(ed). Since false positives are OK,
626 and end-of-str is not later than strend we are OK. */
627 s = fbm_instr((unsigned char*)s,
628 (unsigned char*)last + SvCUR(prog->float_substr)
629 - (SvTAIL(prog->float_substr)!=0),
630 prog->float_substr, PL_multiline ? FBMrf_MULTILINE : 0);
631 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s floating substr `%s%.*s%s'%s",
632 (s ? "Found" : "Contradicts"),
634 (int)(SvCUR(prog->float_substr)
635 - (SvTAIL(prog->float_substr)!=0)),
636 SvPVX(prog->float_substr),
637 PL_colors[1], (SvTAIL(prog->float_substr) ? "$" : "")));
640 DEBUG_r(PerlIO_printf(Perl_debug_log,
641 ", giving up...\n"));
644 DEBUG_r(PerlIO_printf(Perl_debug_log,
645 ", trying anchored starting at offset %ld...\n",
646 (long)(s1 + 1 - i_strpos)));
648 s = HOP3c(t, 1, strend);
652 DEBUG_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
653 (long)(s - i_strpos)));
654 other_last = s; /* Fix this later. --Hugo */
663 t = s - prog->check_offset_max;
664 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
665 && (!(prog->reganch & ROPT_UTF8)
666 || ((t = reghopmaybe3_c(s, -prog->check_offset_max, strpos))
668 /* Fixed substring is found far enough so that the match
669 cannot start at strpos. */
671 if (ml_anch && t[-1] != '\n') {
672 /* Eventually fbm_*() should handle this, but often
673 anchored_offset is not 0, so this check will not be wasted. */
674 /* XXXX In the code below we prefer to look for "^" even in
675 presence of anchored substrings. And we search even
676 beyond the found float position. These pessimizations
677 are historical artefacts only. */
679 while (t < strend - prog->minlen) {
681 if (t < check_at - prog->check_offset_min) {
682 if (prog->anchored_substr) {
683 /* Since we moved from the found position,
684 we definitely contradict the found anchored
685 substr. Due to the above check we do not
686 contradict "check" substr.
687 Thus we can arrive here only if check substr
688 is float. Redo checking for "other"=="fixed".
691 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld, rescanning for anchored from offset %ld...\n",
692 PL_colors[0],PL_colors[1], (long)(strpos - i_strpos), (long)(strpos - i_strpos + prog->anchored_offset)));
693 goto do_other_anchored;
695 /* We don't contradict the found floating substring. */
696 /* XXXX Why not check for STCLASS? */
698 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld...\n",
699 PL_colors[0],PL_colors[1], (long)(s - i_strpos)));
702 /* Position contradicts check-string */
703 /* XXXX probably better to look for check-string
704 than for "\n", so one should lower the limit for t? */
705 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m, restarting lookup for check-string at offset %ld...\n",
706 PL_colors[0],PL_colors[1], (long)(t + 1 - i_strpos)));
707 other_last = strpos = s = t + 1;
712 DEBUG_r(PerlIO_printf(Perl_debug_log, "Did not find /%s^%s/m...\n",
713 PL_colors[0],PL_colors[1]));
717 DEBUG_r(PerlIO_printf(Perl_debug_log, "Starting position does not contradict /%s^%s/m...\n",
718 PL_colors[0],PL_colors[1]));
722 ++BmUSEFUL(prog->check_substr); /* hooray/5 */
725 /* The found string does not prohibit matching at strpos,
726 - no optimization of calling REx engine can be performed,
727 unless it was an MBOL and we are not after MBOL,
728 or a future STCLASS check will fail this. */
730 /* Even in this situation we may use MBOL flag if strpos is offset
731 wrt the start of the string. */
732 if (ml_anch && sv && !SvROK(sv) /* See prev comment on SvROK */
733 && (strpos != strbeg) && strpos[-1] != '\n'
734 /* May be due to an implicit anchor of m{.*foo} */
735 && !(prog->reganch & ROPT_IMPLICIT))
740 DEBUG_r( if (ml_anch)
741 PerlIO_printf(Perl_debug_log, "Position at offset %ld does not contradict /%s^%s/m...\n",
742 (long)(strpos - i_strpos), PL_colors[0],PL_colors[1]);
745 if (!(prog->reganch & ROPT_NAUGHTY) /* XXXX If strpos moved? */
746 && prog->check_substr /* Could be deleted already */
747 && --BmUSEFUL(prog->check_substr) < 0
748 && prog->check_substr == prog->float_substr)
750 /* If flags & SOMETHING - do not do it many times on the same match */
751 DEBUG_r(PerlIO_printf(Perl_debug_log, "... Disabling check substring...\n"));
752 SvREFCNT_dec(prog->check_substr);
753 prog->check_substr = Nullsv; /* disable */
754 prog->float_substr = Nullsv; /* clear */
755 check = Nullsv; /* abort */
757 /* XXXX This is a remnant of the old implementation. It
758 looks wasteful, since now INTUIT can use many
760 prog->reganch &= ~RE_USE_INTUIT;
767 /* XXXX BmUSEFUL already changed, maybe multiple change is meaningful... */
768 if (prog->regstclass) {
769 /* minlen == 0 is possible if regstclass is \b or \B,
770 and the fixed substr is ''$.
771 Since minlen is already taken into account, s+1 is before strend;
772 accidentally, minlen >= 1 guaranties no false positives at s + 1
773 even for \b or \B. But (minlen? 1 : 0) below assumes that
774 regstclass does not come from lookahead... */
775 /* If regstclass takes bytelength more than 1: If charlength==1, OK.
776 This leaves EXACTF only, which is dealt with in find_byclass(). */
777 U8* str = (U8*)STRING(prog->regstclass);
778 int cl_l = (PL_regkind[(U8)OP(prog->regstclass)] == EXACT
779 ? CHR_DIST(str+STR_LEN(prog->regstclass), str)
781 char *endpos = (prog->anchored_substr || ml_anch)
782 ? HOP3c(s, (prog->minlen ? cl_l : 0), strend)
783 : (prog->float_substr
784 ? HOP3c(HOP3c(check_at, -start_shift, strbeg),
787 char *startpos = strbeg;
790 if (prog->reganch & ROPT_UTF8) {
791 PL_regdata = prog->data;
794 s = find_byclass(prog, prog->regstclass, s, endpos, startpos, 1);
799 if (endpos == strend) {
800 DEBUG_r( PerlIO_printf(Perl_debug_log,
801 "Could not match STCLASS...\n") );
804 DEBUG_r( PerlIO_printf(Perl_debug_log,
805 "This position contradicts STCLASS...\n") );
806 if ((prog->reganch & ROPT_ANCH) && !ml_anch)
808 /* Contradict one of substrings */
809 if (prog->anchored_substr) {
810 if (prog->anchored_substr == check) {
811 DEBUG_r( what = "anchored" );
813 s = HOP3c(t, 1, strend);
814 if (s + start_shift + end_shift > strend) {
815 /* XXXX Should be taken into account earlier? */
816 DEBUG_r( PerlIO_printf(Perl_debug_log,
817 "Could not match STCLASS...\n") );
822 DEBUG_r( PerlIO_printf(Perl_debug_log,
823 "Looking for %s substr starting at offset %ld...\n",
824 what, (long)(s + start_shift - i_strpos)) );
827 /* Have both, check_string is floating */
828 if (t + start_shift >= check_at) /* Contradicts floating=check */
829 goto retry_floating_check;
830 /* Recheck anchored substring, but not floating... */
834 DEBUG_r( PerlIO_printf(Perl_debug_log,
835 "Looking for anchored substr starting at offset %ld...\n",
836 (long)(other_last - i_strpos)) );
837 goto do_other_anchored;
839 /* Another way we could have checked stclass at the
840 current position only: */
845 DEBUG_r( PerlIO_printf(Perl_debug_log,
846 "Looking for /%s^%s/m starting at offset %ld...\n",
847 PL_colors[0],PL_colors[1], (long)(t - i_strpos)) );
850 if (!prog->float_substr) /* Could have been deleted */
852 /* Check is floating subtring. */
853 retry_floating_check:
854 t = check_at - start_shift;
855 DEBUG_r( what = "floating" );
856 goto hop_and_restart;
859 DEBUG_r(PerlIO_printf(Perl_debug_log,
860 "By STCLASS: moving %ld --> %ld\n",
861 (long)(t - i_strpos), (long)(s - i_strpos))
865 DEBUG_r(PerlIO_printf(Perl_debug_log,
866 "Does not contradict STCLASS...\n");
871 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s%s:%s match at offset %ld\n",
872 PL_colors[4], (check ? "Guessed" : "Giving up"),
873 PL_colors[5], (long)(s - i_strpos)) );
876 fail_finish: /* Substring not found */
877 if (prog->check_substr) /* could be removed already */
878 BmUSEFUL(prog->check_substr) += 5; /* hooray */
880 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch rejected by optimizer%s\n",
881 PL_colors[4],PL_colors[5]));
885 /* We know what class REx starts with. Try to find this position... */
887 S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, char *strend, char *startpos, I32 norun)
889 I32 doevery = (prog->reganch & ROPT_SKIP) == 0;
895 register I32 tmp = 1; /* Scratch variable? */
896 register bool do_utf8 = PL_reg_match_utf8;
898 /* We know what class it must start with. */
902 if (reginclass(c, (U8*)s, do_utf8)) {
903 if (tmp && (norun || regtry(prog, s)))
910 s += do_utf8 ? UTF8SKIP(s) : 1;
915 if (tmp && (norun || regtry(prog, s)))
927 U8 tmpbuf1[UTF8_MAXLEN*2+1];
928 U8 tmpbuf2[UTF8_MAXLEN*2+1];
930 to_utf8_lower((U8*)m, tmpbuf1, &ulen1);
931 to_utf8_upper((U8*)m, tmpbuf2, &ulen2);
933 c1 = utf8_to_uvuni(tmpbuf1, 0);
934 c2 = utf8_to_uvuni(tmpbuf2, 0);
945 c2 = PL_fold_locale[c1];
950 e = s; /* Due to minlen logic of intuit() */
956 if ( utf8_to_uvchr((U8*)s, &len) == c1
963 UV c = utf8_to_uvchr((U8*)s, &len);
964 if ( (c == c1 || c == c2) && regtry(prog, s) )
973 && (ln == 1 || !(OP(c) == EXACTF
975 : ibcmp_locale(s, m, ln)))
976 && (norun || regtry(prog, s)) )
982 if ( (*(U8*)s == c1 || *(U8*)s == c2)
983 && (ln == 1 || !(OP(c) == EXACTF
985 : ibcmp_locale(s, m, ln)))
986 && (norun || regtry(prog, s)) )
993 PL_reg_flags |= RF_tainted;
1000 U8 *r = reghop3((U8*)s, -1, (U8*)startpos);
1003 tmp = (I32)utf8n_to_uvchr(r, s - (char*)r, 0, 0);
1005 tmp = ((OP(c) == BOUND ?
1006 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1007 LOAD_UTF8_CHARCLASS(alnum,"a");
1008 while (s < strend) {
1009 if (tmp == !(OP(c) == BOUND ?
1010 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1011 isALNUM_LC_utf8((U8*)s)))
1014 if ((norun || regtry(prog, s)))
1021 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1022 tmp = ((OP(c) == BOUND ? isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1023 while (s < strend) {
1025 !(OP(c) == BOUND ? isALNUM(*s) : isALNUM_LC(*s))) {
1027 if ((norun || regtry(prog, s)))
1033 if ((!prog->minlen && tmp) && (norun || regtry(prog, s)))
1037 PL_reg_flags |= RF_tainted;
1044 U8 *r = reghop3((U8*)s, -1, (U8*)startpos);
1047 tmp = (I32)utf8n_to_uvchr(r, s - (char*)r, 0, 0);
1049 tmp = ((OP(c) == NBOUND ?
1050 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1051 LOAD_UTF8_CHARCLASS(alnum,"a");
1052 while (s < strend) {
1053 if (tmp == !(OP(c) == NBOUND ?
1054 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1055 isALNUM_LC_utf8((U8*)s)))
1057 else if ((norun || regtry(prog, s)))
1063 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1064 tmp = ((OP(c) == NBOUND ?
1065 isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1066 while (s < strend) {
1068 !(OP(c) == NBOUND ? isALNUM(*s) : isALNUM_LC(*s)))
1070 else if ((norun || regtry(prog, s)))
1075 if ((!prog->minlen && !tmp) && (norun || regtry(prog, s)))
1080 LOAD_UTF8_CHARCLASS(alnum,"a");
1081 while (s < strend) {
1082 if (swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1083 if (tmp && (norun || regtry(prog, s)))
1094 while (s < strend) {
1096 if (tmp && (norun || regtry(prog, s)))
1108 PL_reg_flags |= RF_tainted;
1110 while (s < strend) {
1111 if (isALNUM_LC_utf8((U8*)s)) {
1112 if (tmp && (norun || regtry(prog, s)))
1123 while (s < strend) {
1124 if (isALNUM_LC(*s)) {
1125 if (tmp && (norun || regtry(prog, s)))
1138 LOAD_UTF8_CHARCLASS(alnum,"a");
1139 while (s < strend) {
1140 if (!swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1141 if (tmp && (norun || regtry(prog, s)))
1152 while (s < strend) {
1154 if (tmp && (norun || regtry(prog, s)))
1166 PL_reg_flags |= RF_tainted;
1168 while (s < strend) {
1169 if (!isALNUM_LC_utf8((U8*)s)) {
1170 if (tmp && (norun || regtry(prog, s)))
1181 while (s < strend) {
1182 if (!isALNUM_LC(*s)) {
1183 if (tmp && (norun || regtry(prog, s)))
1196 LOAD_UTF8_CHARCLASS(space," ");
1197 while (s < strend) {
1198 if (*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8)) {
1199 if (tmp && (norun || regtry(prog, s)))
1210 while (s < strend) {
1212 if (tmp && (norun || regtry(prog, s)))
1224 PL_reg_flags |= RF_tainted;
1226 while (s < strend) {
1227 if (*s == ' ' || isSPACE_LC_utf8((U8*)s)) {
1228 if (tmp && (norun || regtry(prog, s)))
1239 while (s < strend) {
1240 if (isSPACE_LC(*s)) {
1241 if (tmp && (norun || regtry(prog, s)))
1254 LOAD_UTF8_CHARCLASS(space," ");
1255 while (s < strend) {
1256 if (!(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8))) {
1257 if (tmp && (norun || regtry(prog, s)))
1268 while (s < strend) {
1270 if (tmp && (norun || regtry(prog, s)))
1282 PL_reg_flags |= RF_tainted;
1284 while (s < strend) {
1285 if (!(*s == ' ' || isSPACE_LC_utf8((U8*)s))) {
1286 if (tmp && (norun || regtry(prog, s)))
1297 while (s < strend) {
1298 if (!isSPACE_LC(*s)) {
1299 if (tmp && (norun || regtry(prog, s)))
1312 LOAD_UTF8_CHARCLASS(digit,"0");
1313 while (s < strend) {
1314 if (swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1315 if (tmp && (norun || regtry(prog, s)))
1326 while (s < strend) {
1328 if (tmp && (norun || regtry(prog, s)))
1340 PL_reg_flags |= RF_tainted;
1342 while (s < strend) {
1343 if (isDIGIT_LC_utf8((U8*)s)) {
1344 if (tmp && (norun || regtry(prog, s)))
1355 while (s < strend) {
1356 if (isDIGIT_LC(*s)) {
1357 if (tmp && (norun || regtry(prog, s)))
1370 LOAD_UTF8_CHARCLASS(digit,"0");
1371 while (s < strend) {
1372 if (!swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1373 if (tmp && (norun || regtry(prog, s)))
1384 while (s < strend) {
1386 if (tmp && (norun || regtry(prog, s)))
1398 PL_reg_flags |= RF_tainted;
1400 while (s < strend) {
1401 if (!isDIGIT_LC_utf8((U8*)s)) {
1402 if (tmp && (norun || regtry(prog, s)))
1413 while (s < strend) {
1414 if (!isDIGIT_LC(*s)) {
1415 if (tmp && (norun || regtry(prog, s)))
1427 Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c));
1436 - regexec_flags - match a regexp against a string
1439 Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char *strend,
1440 char *strbeg, I32 minend, SV *sv, void *data, U32 flags)
1441 /* strend: pointer to null at end of string */
1442 /* strbeg: real beginning of string */
1443 /* minend: end of match must be >=minend after stringarg. */
1444 /* data: May be used for some additional optimizations. */
1445 /* nosave: For optimizations. */
1448 register regnode *c;
1449 register char *startpos = stringarg;
1450 I32 minlen; /* must match at least this many chars */
1451 I32 dontbother = 0; /* how many characters not to try at end */
1452 /* I32 start_shift = 0; */ /* Offset of the start to find
1453 constant substr. */ /* CC */
1454 I32 end_shift = 0; /* Same for the end. */ /* CC */
1455 I32 scream_pos = -1; /* Internal iterator of scream. */
1457 SV* oreplsv = GvSV(PL_replgv);
1458 bool do_utf8 = DO_UTF8(sv);
1460 SV *dsv = sv_2mortal(newSVpvn("", 0));
1467 PL_regnarrate = DEBUG_r_TEST;
1470 /* Be paranoid... */
1471 if (prog == NULL || startpos == NULL) {
1472 Perl_croak(aTHX_ "NULL regexp parameter");
1476 minlen = prog->minlen;
1477 if (do_utf8 && !(prog->reganch & ROPT_CANY_SEEN)) {
1478 if (utf8_distance((U8*)strend, (U8*)startpos) < minlen) goto phooey;
1481 if (strend - startpos < minlen) goto phooey;
1484 /* Check validity of program. */
1485 if (UCHARAT(prog->program) != REG_MAGIC) {
1486 Perl_croak(aTHX_ "corrupted regexp program");
1490 PL_reg_eval_set = 0;
1493 if (prog->reganch & ROPT_UTF8)
1494 PL_reg_flags |= RF_utf8;
1496 /* Mark beginning of line for ^ and lookbehind. */
1497 PL_regbol = startpos;
1501 /* Mark end of line for $ (and such) */
1504 /* see how far we have to get to not match where we matched before */
1505 PL_regtill = startpos+minend;
1507 /* We start without call_cc context. */
1510 /* If there is a "must appear" string, look for it. */
1513 if (prog->reganch & ROPT_GPOS_SEEN) { /* Need to have PL_reg_ganch */
1516 if (flags & REXEC_IGNOREPOS) /* Means: check only at start */
1517 PL_reg_ganch = startpos;
1518 else if (sv && SvTYPE(sv) >= SVt_PVMG
1520 && (mg = mg_find(sv, PERL_MAGIC_regex_global))
1521 && mg->mg_len >= 0) {
1522 PL_reg_ganch = strbeg + mg->mg_len; /* Defined pos() */
1523 if (prog->reganch & ROPT_ANCH_GPOS) {
1524 if (s > PL_reg_ganch)
1529 else /* pos() not defined */
1530 PL_reg_ganch = strbeg;
1533 if (do_utf8 == (UTF!=0) &&
1534 !(flags & REXEC_CHECKED) && prog->check_substr != Nullsv) {
1535 re_scream_pos_data d;
1537 d.scream_olds = &scream_olds;
1538 d.scream_pos = &scream_pos;
1539 s = re_intuit_start(prog, sv, s, strend, flags, &d);
1541 goto phooey; /* not present */
1545 char *s = UTF ? sv_uni_display(dsv, sv, 60, 0) : startpos;
1546 STRLEN len = UTF ? strlen(s) : strend - startpos;
1549 PerlIO_printf(Perl_debug_log,
1550 "%sMatching REx%s `%s%.60s%s%s' against `%s%.*s%s%s'\n",
1551 PL_colors[4],PL_colors[5],PL_colors[0],
1554 (strlen(prog->precomp) > 60 ? "..." : ""),
1556 (int)(len > 60 ? 60 : len),
1558 (len > 60 ? "..." : "")
1562 /* Simplest case: anchored match need be tried only once. */
1563 /* [unless only anchor is BOL and multiline is set] */
1564 if (prog->reganch & (ROPT_ANCH & ~ROPT_ANCH_GPOS)) {
1565 if (s == startpos && regtry(prog, startpos))
1567 else if (PL_multiline || (prog->reganch & ROPT_IMPLICIT)
1568 || (prog->reganch & ROPT_ANCH_MBOL)) /* XXXX SBOL? */
1573 dontbother = minlen - 1;
1574 end = HOP3c(strend, -dontbother, strbeg) - 1;
1575 /* for multiline we only have to try after newlines */
1576 if (prog->check_substr) {
1580 if (regtry(prog, s))
1585 if (prog->reganch & RE_USE_INTUIT) {
1586 s = re_intuit_start(prog, sv, s + 1, strend, flags, NULL);
1597 if (*s++ == '\n') { /* don't need PL_utf8skip here */
1598 if (regtry(prog, s))
1605 } else if (prog->reganch & ROPT_ANCH_GPOS) {
1606 if (regtry(prog, PL_reg_ganch))
1611 /* Messy cases: unanchored match. */
1612 if (prog->anchored_substr && prog->reganch & ROPT_SKIP) {
1613 /* we have /x+whatever/ */
1614 /* it must be a one character string (XXXX Except UTF?) */
1615 char ch = SvPVX(prog->anchored_substr)[0];
1621 while (s < strend) {
1623 DEBUG_r( did_match = 1 );
1624 if (regtry(prog, s)) goto got_it;
1626 while (s < strend && *s == ch)
1633 while (s < strend) {
1635 DEBUG_r( did_match = 1 );
1636 if (regtry(prog, s)) goto got_it;
1638 while (s < strend && *s == ch)
1644 DEBUG_r(if (!did_match)
1645 PerlIO_printf(Perl_debug_log,
1646 "Did not find anchored character...\n")
1650 else if (do_utf8 == (UTF!=0) &&
1651 (prog->anchored_substr != Nullsv
1652 || (prog->float_substr != Nullsv
1653 && prog->float_max_offset < strend - s))) {
1654 SV *must = prog->anchored_substr
1655 ? prog->anchored_substr : prog->float_substr;
1657 prog->anchored_substr ? prog->anchored_offset : prog->float_max_offset;
1659 prog->anchored_substr ? prog->anchored_offset : prog->float_min_offset;
1660 char *last = HOP3c(strend, /* Cannot start after this */
1661 -(I32)(CHR_SVLEN(must)
1662 - (SvTAIL(must) != 0) + back_min), strbeg);
1663 char *last1; /* Last position checked before */
1669 last1 = HOPc(s, -1);
1671 last1 = s - 1; /* bogus */
1673 /* XXXX check_substr already used to find `s', can optimize if
1674 check_substr==must. */
1676 dontbother = end_shift;
1677 strend = HOPc(strend, -dontbother);
1678 while ( (s <= last) &&
1679 ((flags & REXEC_SCREAM)
1680 ? (s = screaminstr(sv, must, HOP3c(s, back_min, strend) - strbeg,
1681 end_shift, &scream_pos, 0))
1682 : (s = fbm_instr((unsigned char*)HOP3(s, back_min, strend),
1683 (unsigned char*)strend, must,
1684 PL_multiline ? FBMrf_MULTILINE : 0))) ) {
1685 DEBUG_r( did_match = 1 );
1686 if (HOPc(s, -back_max) > last1) {
1687 last1 = HOPc(s, -back_min);
1688 s = HOPc(s, -back_max);
1691 char *t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1;
1693 last1 = HOPc(s, -back_min);
1697 while (s <= last1) {
1698 if (regtry(prog, s))
1704 while (s <= last1) {
1705 if (regtry(prog, s))
1711 DEBUG_r(if (!did_match)
1712 PerlIO_printf(Perl_debug_log,
1713 "Did not find %s substr `%s%.*s%s'%s...\n",
1714 ((must == prog->anchored_substr)
1715 ? "anchored" : "floating"),
1717 (int)(SvCUR(must) - (SvTAIL(must)!=0)),
1719 PL_colors[1], (SvTAIL(must) ? "$" : ""))
1723 else if ((c = prog->regstclass)) {
1724 if (minlen && PL_regkind[(U8)OP(prog->regstclass)] != EXACT)
1725 /* don't bother with what can't match */
1726 strend = HOPc(strend, -(minlen - 1));
1728 SV *prop = sv_newmortal();
1730 PerlIO_printf(Perl_debug_log, "Matching stclass `%s' against `%s'\n", SvPVX(prop), UTF ? sv_uni_display(dsv, sv, 60, 0) : s);
1732 if (find_byclass(prog, c, s, strend, startpos, 0))
1734 DEBUG_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass...\n"));
1738 if (prog->float_substr != Nullsv) { /* Trim the end. */
1741 if (flags & REXEC_SCREAM) {
1742 last = screaminstr(sv, prog->float_substr, s - strbeg,
1743 end_shift, &scream_pos, 1); /* last one */
1745 last = scream_olds; /* Only one occurrence. */
1749 char *little = SvPV(prog->float_substr, len);
1751 if (SvTAIL(prog->float_substr)) {
1752 if (memEQ(strend - len + 1, little, len - 1))
1753 last = strend - len + 1;
1754 else if (!PL_multiline)
1755 last = memEQ(strend - len, little, len)
1756 ? strend - len : Nullch;
1762 last = rninstr(s, strend, little, little + len);
1764 last = strend; /* matching `$' */
1768 DEBUG_r(PerlIO_printf(Perl_debug_log,
1769 "%sCan't trim the tail, match fails (should not happen)%s\n",
1770 PL_colors[4],PL_colors[5]));
1771 goto phooey; /* Should not happen! */
1773 dontbother = strend - last + prog->float_min_offset;
1775 if (minlen && (dontbother < minlen))
1776 dontbother = minlen - 1;
1777 strend -= dontbother; /* this one's always in bytes! */
1778 /* We don't know much -- general case. */
1781 if (regtry(prog, s))
1790 if (regtry(prog, s))
1792 } while (s++ < strend);
1800 RX_MATCH_TAINTED_set(prog, PL_reg_flags & RF_tainted);
1802 if (PL_reg_eval_set) {
1803 /* Preserve the current value of $^R */
1804 if (oreplsv != GvSV(PL_replgv))
1805 sv_setsv(oreplsv, GvSV(PL_replgv));/* So that when GvSV(replgv) is
1806 restored, the value remains
1808 restore_pos(aTHX_ 0);
1811 /* make sure $`, $&, $', and $digit will work later */
1812 if ( !(flags & REXEC_NOT_FIRST) ) {
1813 if (RX_MATCH_COPIED(prog)) {
1814 Safefree(prog->subbeg);
1815 RX_MATCH_COPIED_off(prog);
1817 if (flags & REXEC_COPY_STR) {
1818 I32 i = PL_regeol - startpos + (stringarg - strbeg);
1820 s = savepvn(strbeg, i);
1823 RX_MATCH_COPIED_on(prog);
1826 prog->subbeg = strbeg;
1827 prog->sublen = PL_regeol - strbeg; /* strend may have been modified */
1834 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n",
1835 PL_colors[4],PL_colors[5]));
1836 if (PL_reg_eval_set)
1837 restore_pos(aTHX_ 0);
1842 - regtry - try match at specific point
1844 STATIC I32 /* 0 failure, 1 success */
1845 S_regtry(pTHX_ regexp *prog, char *startpos)
1853 PL_regindent = 0; /* XXXX Not good when matches are reenterable... */
1855 if ((prog->reganch & ROPT_EVAL_SEEN) && !PL_reg_eval_set) {
1858 PL_reg_eval_set = RS_init;
1860 PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %"IVdf"\n",
1861 (IV)(PL_stack_sp - PL_stack_base));
1863 SAVEI32(cxstack[cxstack_ix].blk_oldsp);
1864 cxstack[cxstack_ix].blk_oldsp = PL_stack_sp - PL_stack_base;
1865 /* Otherwise OP_NEXTSTATE will free whatever on stack now. */
1867 /* Apparently this is not needed, judging by wantarray. */
1868 /* SAVEI8(cxstack[cxstack_ix].blk_gimme);
1869 cxstack[cxstack_ix].blk_gimme = G_SCALAR; */
1872 /* Make $_ available to executed code. */
1873 if (PL_reg_sv != DEFSV) {
1874 /* SAVE_DEFSV does *not* suffice here for USE_5005THREADS */
1879 if (!(SvTYPE(PL_reg_sv) >= SVt_PVMG && SvMAGIC(PL_reg_sv)
1880 && (mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global)))) {
1881 /* prepare for quick setting of pos */
1882 sv_magic(PL_reg_sv, (SV*)0,
1883 PERL_MAGIC_regex_global, Nullch, 0);
1884 mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global);
1888 PL_reg_oldpos = mg->mg_len;
1889 SAVEDESTRUCTOR_X(restore_pos, 0);
1891 if (!PL_reg_curpm) {
1892 Newz(22,PL_reg_curpm, 1, PMOP);
1895 SV* repointer = newSViv(0);
1896 /* so we know which PL_regex_padav element is PL_reg_curpm */
1897 SvFLAGS(repointer) |= SVf_BREAK;
1898 av_push(PL_regex_padav,repointer);
1899 PL_reg_curpm->op_pmoffset = av_len(PL_regex_padav);
1900 PL_regex_pad = AvARRAY(PL_regex_padav);
1904 PM_SETRE(PL_reg_curpm, prog);
1905 PL_reg_oldcurpm = PL_curpm;
1906 PL_curpm = PL_reg_curpm;
1907 if (RX_MATCH_COPIED(prog)) {
1908 /* Here is a serious problem: we cannot rewrite subbeg,
1909 since it may be needed if this match fails. Thus
1910 $` inside (?{}) could fail... */
1911 PL_reg_oldsaved = prog->subbeg;
1912 PL_reg_oldsavedlen = prog->sublen;
1913 RX_MATCH_COPIED_off(prog);
1916 PL_reg_oldsaved = Nullch;
1917 prog->subbeg = PL_bostr;
1918 prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */
1920 prog->startp[0] = startpos - PL_bostr;
1921 PL_reginput = startpos;
1922 PL_regstartp = prog->startp;
1923 PL_regendp = prog->endp;
1924 PL_reglastparen = &prog->lastparen;
1925 PL_reglastcloseparen = &prog->lastcloseparen;
1926 prog->lastparen = 0;
1928 DEBUG_r(PL_reg_starttry = startpos);
1929 if (PL_reg_start_tmpl <= prog->nparens) {
1930 PL_reg_start_tmpl = prog->nparens*3/2 + 3;
1931 if(PL_reg_start_tmp)
1932 Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
1934 New(22,PL_reg_start_tmp, PL_reg_start_tmpl, char*);
1937 /* XXXX What this code is doing here?!!! There should be no need
1938 to do this again and again, PL_reglastparen should take care of
1941 /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code.
1942 * Actually, the code in regcppop() (which Ilya may be meaning by
1943 * PL_reglastparen), is not needed at all by the test suite
1944 * (op/regexp, op/pat, op/split), but that code is needed, oddly
1945 * enough, for building DynaLoader, or otherwise this
1946 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
1947 * will happen. Meanwhile, this code *is* needed for the
1948 * above-mentioned test suite tests to succeed. The common theme
1949 * on those tests seems to be returning null fields from matches.
1954 if (prog->nparens) {
1955 for (i = prog->nparens; i > *PL_reglastparen; i--) {
1962 if (regmatch(prog->program + 1)) {
1963 prog->endp[0] = PL_reginput - PL_bostr;
1966 REGCP_UNWIND(lastcp);
1970 #define RE_UNWIND_BRANCH 1
1971 #define RE_UNWIND_BRANCHJ 2
1975 typedef struct { /* XX: makes sense to enlarge it... */
1979 } re_unwind_generic_t;
1992 } re_unwind_branch_t;
1994 typedef union re_unwind_t {
1996 re_unwind_generic_t generic;
1997 re_unwind_branch_t branch;
2000 #define sayYES goto yes
2001 #define sayNO goto no
2002 #define sayYES_FINAL goto yes_final
2003 #define sayYES_LOUD goto yes_loud
2004 #define sayNO_FINAL goto no_final
2005 #define sayNO_SILENT goto do_no
2006 #define saySAME(x) if (x) goto yes; else goto no
2008 #define REPORT_CODE_OFF 24
2011 - regmatch - main matching routine
2013 * Conceptually the strategy is simple: check to see whether the current
2014 * node matches, call self recursively to see whether the rest matches,
2015 * and then act accordingly. In practice we make some effort to avoid
2016 * recursion, in particular by going through "ordinary" nodes (that don't
2017 * need to know whether the rest of the match failed) by a loop instead of
2020 /* [lwall] I've hoisted the register declarations to the outer block in order to
2021 * maybe save a little bit of pushing and popping on the stack. It also takes
2022 * advantage of machines that use a register save mask on subroutine entry.
2024 STATIC I32 /* 0 failure, 1 success */
2025 S_regmatch(pTHX_ regnode *prog)
2027 register regnode *scan; /* Current node. */
2028 regnode *next; /* Next node. */
2029 regnode *inner; /* Next node in internal branch. */
2030 register I32 nextchr; /* renamed nextchr - nextchar colides with
2031 function of same name */
2032 register I32 n; /* no or next */
2033 register I32 ln = 0; /* len or last */
2034 register char *s = Nullch; /* operand or save */
2035 register char *locinput = PL_reginput;
2036 register I32 c1 = 0, c2 = 0, paren; /* case fold search, parenth */
2037 int minmod = 0, sw = 0, logical = 0;
2040 I32 firstcp = PL_savestack_ix;
2042 register bool do_utf8 = PL_reg_match_utf8;
2044 SV *dsv0 = sv_2mortal(newSVpvn("", 0));
2045 SV *dsv1 = sv_2mortal(newSVpvn("", 0));
2046 SV *dsv2 = sv_2mortal(newSVpvn("", 0));
2053 /* Note that nextchr is a byte even in UTF */
2054 nextchr = UCHARAT(locinput);
2056 while (scan != NULL) {
2059 SV *prop = sv_newmortal();
2060 int docolor = *PL_colors[0];
2061 int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */
2062 int l = (PL_regeol - locinput) > taill ? taill : (PL_regeol - locinput);
2063 /* The part of the string before starttry has one color
2064 (pref0_len chars), between starttry and current
2065 position another one (pref_len - pref0_len chars),
2066 after the current position the third one.
2067 We assume that pref0_len <= pref_len, otherwise we
2068 decrease pref0_len. */
2069 int pref_len = (locinput - PL_bostr) > (5 + taill) - l
2070 ? (5 + taill) - l : locinput - PL_bostr;
2073 while (UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len)))
2075 pref0_len = pref_len - (locinput - PL_reg_starttry);
2076 if (l + pref_len < (5 + taill) && l < PL_regeol - locinput)
2077 l = ( PL_regeol - locinput > (5 + taill) - pref_len
2078 ? (5 + taill) - pref_len : PL_regeol - locinput);
2079 while (UTF8_IS_CONTINUATION(*(U8*)(locinput + l)))
2083 if (pref0_len > pref_len)
2084 pref0_len = pref_len;
2085 regprop(prop, scan);
2089 pv_uni_display(dsv0, (U8*)(locinput - pref_len),
2091 locinput - pref_len;
2092 STRLEN len0 = UTF ? strlen(s0) : pref0_len;
2094 pv_uni_display(dsv1, (U8*)(locinput - pref_len + pref0_len),
2095 pref_len - pref0_len, 60, 0) :
2096 locinput - pref_len + pref0_len;
2097 STRLEN len1 = UTF ? strlen(s1) : pref_len - pref0_len;
2099 pv_uni_display(dsv2, (U8*)locinput,
2100 PL_regeol - locinput, 60, 0) :
2102 STRLEN len2 = UTF ? strlen(s2) : l;
2103 PerlIO_printf(Perl_debug_log,
2104 "%4"IVdf" <%s%.*s%s%s%.*s%s%s%s%.*s%s>%*s|%3"IVdf":%*s%s\n",
2105 (IV)(locinput - PL_bostr),
2112 (docolor ? "" : "> <"),
2116 15 - l - pref_len + 1,
2118 (IV)(scan - PL_regprogram), PL_regindent*2, "",
2123 next = scan + NEXT_OFF(scan);
2129 if (locinput == PL_bostr || (PL_multiline &&
2130 (nextchr || locinput < PL_regeol) && locinput[-1] == '\n') )
2132 /* regtill = regbol; */
2137 if (locinput == PL_bostr ||
2138 ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n'))
2144 if (locinput == PL_bostr)
2148 if (locinput == PL_reg_ganch)
2158 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2163 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2165 if (PL_regeol - locinput > 1)
2169 if (PL_regeol != locinput)
2173 if (!nextchr && locinput >= PL_regeol)
2176 locinput += PL_utf8skip[nextchr];
2177 if (locinput > PL_regeol)
2179 nextchr = UCHARAT(locinput);
2182 nextchr = UCHARAT(++locinput);
2185 if (!nextchr && locinput >= PL_regeol)
2187 nextchr = UCHARAT(++locinput);
2190 if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n')
2193 locinput += PL_utf8skip[nextchr];
2194 if (locinput > PL_regeol)
2196 nextchr = UCHARAT(locinput);
2199 nextchr = UCHARAT(++locinput);
2204 if (do_utf8 != (UTF!=0)) {
2212 if (*((U8*)s) != utf8_to_uvchr((U8*)l, &len))
2221 if (*((U8*)l) != utf8_to_uvchr((U8*)s, &len))
2227 nextchr = UCHARAT(locinput);
2230 /* Inline the first character, for speed. */
2231 if (UCHARAT(s) != nextchr)
2233 if (PL_regeol - locinput < ln)
2235 if (ln > 1 && memNE(s, locinput, ln))
2238 nextchr = UCHARAT(locinput);
2241 PL_reg_flags |= RF_tainted;
2251 U8 tmpbuf[UTF8_MAXLEN*2+1];
2256 toLOWER_utf8((U8*)l, tmpbuf, &ulen);
2257 if (memNE(s, tmpbuf, ulen))
2263 nextchr = UCHARAT(locinput);
2267 /* Inline the first character, for speed. */
2268 if (UCHARAT(s) != nextchr &&
2269 UCHARAT(s) != ((OP(scan) == EXACTF)
2270 ? PL_fold : PL_fold_locale)[nextchr])
2272 if (PL_regeol - locinput < ln)
2274 if (ln > 1 && (OP(scan) == EXACTF
2275 ? ibcmp(s, locinput, ln)
2276 : ibcmp_locale(s, locinput, ln)))
2279 nextchr = UCHARAT(locinput);
2283 if (!reginclass(scan, (U8*)locinput, do_utf8))
2285 if (locinput >= PL_regeol)
2287 locinput += PL_utf8skip[nextchr];
2288 nextchr = UCHARAT(locinput);
2292 nextchr = UCHARAT(locinput);
2293 if (!reginclass(scan, (U8*)locinput, do_utf8))
2295 if (!nextchr && locinput >= PL_regeol)
2297 nextchr = UCHARAT(++locinput);
2301 PL_reg_flags |= RF_tainted;
2307 LOAD_UTF8_CHARCLASS(alnum,"a");
2308 if (!(OP(scan) == ALNUM
2309 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2310 : isALNUM_LC_utf8((U8*)locinput)))
2314 locinput += PL_utf8skip[nextchr];
2315 nextchr = UCHARAT(locinput);
2318 if (!(OP(scan) == ALNUM
2319 ? isALNUM(nextchr) : isALNUM_LC(nextchr)))
2321 nextchr = UCHARAT(++locinput);
2324 PL_reg_flags |= RF_tainted;
2327 if (!nextchr && locinput >= PL_regeol)
2330 LOAD_UTF8_CHARCLASS(alnum,"a");
2331 if (OP(scan) == NALNUM
2332 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2333 : isALNUM_LC_utf8((U8*)locinput))
2337 locinput += PL_utf8skip[nextchr];
2338 nextchr = UCHARAT(locinput);
2341 if (OP(scan) == NALNUM
2342 ? isALNUM(nextchr) : isALNUM_LC(nextchr))
2344 nextchr = UCHARAT(++locinput);
2348 PL_reg_flags |= RF_tainted;
2352 /* was last char in word? */
2354 if (locinput == PL_bostr)
2357 U8 *r = reghop((U8*)locinput, -1);
2359 ln = utf8n_to_uvchr(r, s - (char*)r, 0, 0);
2361 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2362 ln = isALNUM_uni(ln);
2363 LOAD_UTF8_CHARCLASS(alnum,"a");
2364 n = swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8);
2367 ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(ln));
2368 n = isALNUM_LC_utf8((U8*)locinput);
2372 ln = (locinput != PL_bostr) ?
2373 UCHARAT(locinput - 1) : '\n';
2374 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2376 n = isALNUM(nextchr);
2379 ln = isALNUM_LC(ln);
2380 n = isALNUM_LC(nextchr);
2383 if (((!ln) == (!n)) == (OP(scan) == BOUND ||
2384 OP(scan) == BOUNDL))
2388 PL_reg_flags |= RF_tainted;
2394 if (UTF8_IS_CONTINUED(nextchr)) {
2395 LOAD_UTF8_CHARCLASS(space," ");
2396 if (!(OP(scan) == SPACE
2397 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2398 : isSPACE_LC_utf8((U8*)locinput)))
2402 locinput += PL_utf8skip[nextchr];
2403 nextchr = UCHARAT(locinput);
2406 if (!(OP(scan) == SPACE
2407 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2409 nextchr = UCHARAT(++locinput);
2412 if (!(OP(scan) == SPACE
2413 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2415 nextchr = UCHARAT(++locinput);
2419 PL_reg_flags |= RF_tainted;
2422 if (!nextchr && locinput >= PL_regeol)
2425 LOAD_UTF8_CHARCLASS(space," ");
2426 if (OP(scan) == NSPACE
2427 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2428 : isSPACE_LC_utf8((U8*)locinput))
2432 locinput += PL_utf8skip[nextchr];
2433 nextchr = UCHARAT(locinput);
2436 if (OP(scan) == NSPACE
2437 ? isSPACE(nextchr) : isSPACE_LC(nextchr))
2439 nextchr = UCHARAT(++locinput);
2442 PL_reg_flags |= RF_tainted;
2448 LOAD_UTF8_CHARCLASS(digit,"0");
2449 if (!(OP(scan) == DIGIT
2450 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2451 : isDIGIT_LC_utf8((U8*)locinput)))
2455 locinput += PL_utf8skip[nextchr];
2456 nextchr = UCHARAT(locinput);
2459 if (!(OP(scan) == DIGIT
2460 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr)))
2462 nextchr = UCHARAT(++locinput);
2465 PL_reg_flags |= RF_tainted;
2468 if (!nextchr && locinput >= PL_regeol)
2471 LOAD_UTF8_CHARCLASS(digit,"0");
2472 if (OP(scan) == NDIGIT
2473 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2474 : isDIGIT_LC_utf8((U8*)locinput))
2478 locinput += PL_utf8skip[nextchr];
2479 nextchr = UCHARAT(locinput);
2482 if (OP(scan) == NDIGIT
2483 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr))
2485 nextchr = UCHARAT(++locinput);
2488 LOAD_UTF8_CHARCLASS(mark,"~");
2489 if (locinput >= PL_regeol ||
2490 swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2492 locinput += PL_utf8skip[nextchr];
2493 while (locinput < PL_regeol &&
2494 swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2495 locinput += UTF8SKIP(locinput);
2496 if (locinput > PL_regeol)
2498 nextchr = UCHARAT(locinput);
2501 PL_reg_flags |= RF_tainted;
2505 n = ARG(scan); /* which paren pair */
2506 ln = PL_regstartp[n];
2507 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2508 if (*PL_reglastparen < n || ln == -1)
2509 sayNO; /* Do not match unless seen CLOSEn. */
2510 if (ln == PL_regendp[n])
2514 if (do_utf8 && OP(scan) != REF) { /* REF can do byte comparison */
2516 char *e = PL_bostr + PL_regendp[n];
2518 * Note that we can't do the "other character" lookup trick as
2519 * in the 8-bit case (no pun intended) because in Unicode we
2520 * have to map both upper and title case to lower case.
2522 if (OP(scan) == REFF) {
2523 STRLEN ulen1, ulen2;
2524 U8 tmpbuf1[UTF8_MAXLEN*2+1];
2525 U8 tmpbuf2[UTF8_MAXLEN*2+1];
2529 toLOWER_utf8((U8*)s, tmpbuf1, &ulen1);
2530 toLOWER_utf8((U8*)l, tmpbuf2, &ulen2);
2531 if (ulen1 != ulen2 || memNE(tmpbuf1, tmpbuf2, ulen1))
2538 nextchr = UCHARAT(locinput);
2542 /* Inline the first character, for speed. */
2543 if (UCHARAT(s) != nextchr &&
2545 (UCHARAT(s) != ((OP(scan) == REFF
2546 ? PL_fold : PL_fold_locale)[nextchr]))))
2548 ln = PL_regendp[n] - ln;
2549 if (locinput + ln > PL_regeol)
2551 if (ln > 1 && (OP(scan) == REF
2552 ? memNE(s, locinput, ln)
2554 ? ibcmp(s, locinput, ln)
2555 : ibcmp_locale(s, locinput, ln))))
2558 nextchr = UCHARAT(locinput);
2569 OP_4tree *oop = PL_op;
2570 COP *ocurcop = PL_curcop;
2571 SV **ocurpad = PL_curpad;
2575 PL_op = (OP_4tree*)PL_regdata->data[n];
2576 DEBUG_r( PerlIO_printf(Perl_debug_log, " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) );
2577 PL_curpad = AvARRAY((AV*)PL_regdata->data[n + 2]);
2578 PL_regendp[0] = PL_reg_magic->mg_len = locinput - PL_bostr;
2582 CALLRUNOPS(aTHX); /* Scalar context. */
2585 ret = Nullsv; /* protect against empty (?{}) blocks. */
2593 PL_curpad = ocurpad;
2594 PL_curcop = ocurcop;
2596 if (logical == 2) { /* Postponed subexpression. */
2598 MAGIC *mg = Null(MAGIC*);
2600 CHECKPOINT cp, lastcp;
2602 if(SvROK(ret) || SvRMAGICAL(ret)) {
2603 SV *sv = SvROK(ret) ? SvRV(ret) : ret;
2606 mg = mg_find(sv, PERL_MAGIC_qr);
2609 re = (regexp *)mg->mg_obj;
2610 (void)ReREFCNT_inc(re);
2614 char *t = SvPV(ret, len);
2616 char *oprecomp = PL_regprecomp;
2617 I32 osize = PL_regsize;
2618 I32 onpar = PL_regnpar;
2621 re = CALLREGCOMP(aTHX_ t, t + len, &pm);
2623 & (SVs_TEMP | SVs_PADTMP | SVf_READONLY)))
2624 sv_magic(ret,(SV*)ReREFCNT_inc(re),
2626 PL_regprecomp = oprecomp;
2631 PerlIO_printf(Perl_debug_log,
2632 "Entering embedded `%s%.60s%s%s'\n",
2636 (strlen(re->precomp) > 60 ? "..." : ""))
2639 state.prev = PL_reg_call_cc;
2640 state.cc = PL_regcc;
2641 state.re = PL_reg_re;
2645 cp = regcppush(0); /* Save *all* the positions. */
2648 state.ss = PL_savestack_ix;
2649 *PL_reglastparen = 0;
2650 *PL_reglastcloseparen = 0;
2651 PL_reg_call_cc = &state;
2652 PL_reginput = locinput;
2654 /* XXXX This is too dramatic a measure... */
2657 if (regmatch(re->program + 1)) {
2658 /* Even though we succeeded, we need to restore
2659 global variables, since we may be wrapped inside
2660 SUSPEND, thus the match may be not finished yet. */
2662 /* XXXX Do this only if SUSPENDed? */
2663 PL_reg_call_cc = state.prev;
2664 PL_regcc = state.cc;
2665 PL_reg_re = state.re;
2666 cache_re(PL_reg_re);
2668 /* XXXX This is too dramatic a measure... */
2671 /* These are needed even if not SUSPEND. */
2677 REGCP_UNWIND(lastcp);
2679 PL_reg_call_cc = state.prev;
2680 PL_regcc = state.cc;
2681 PL_reg_re = state.re;
2682 cache_re(PL_reg_re);
2684 /* XXXX This is too dramatic a measure... */
2694 sv_setsv(save_scalar(PL_replgv), ret);
2698 n = ARG(scan); /* which paren pair */
2699 PL_reg_start_tmp[n] = locinput;
2704 n = ARG(scan); /* which paren pair */
2705 PL_regstartp[n] = PL_reg_start_tmp[n] - PL_bostr;
2706 PL_regendp[n] = locinput - PL_bostr;
2707 if (n > *PL_reglastparen)
2708 *PL_reglastparen = n;
2709 *PL_reglastcloseparen = n;
2712 n = ARG(scan); /* which paren pair */
2713 sw = (*PL_reglastparen >= n && PL_regendp[n] != -1);
2716 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2718 next = NEXTOPER(NEXTOPER(scan));
2720 next = scan + ARG(scan);
2721 if (OP(next) == IFTHEN) /* Fake one. */
2722 next = NEXTOPER(NEXTOPER(next));
2726 logical = scan->flags;
2728 /*******************************************************************
2729 PL_regcc contains infoblock about the innermost (...)* loop, and
2730 a pointer to the next outer infoblock.
2732 Here is how Y(A)*Z is processed (if it is compiled into CURLYX/WHILEM):
2734 1) After matching X, regnode for CURLYX is processed;
2736 2) This regnode creates infoblock on the stack, and calls
2737 regmatch() recursively with the starting point at WHILEM node;
2739 3) Each hit of WHILEM node tries to match A and Z (in the order
2740 depending on the current iteration, min/max of {min,max} and
2741 greediness). The information about where are nodes for "A"
2742 and "Z" is read from the infoblock, as is info on how many times "A"
2743 was already matched, and greediness.
2745 4) After A matches, the same WHILEM node is hit again.
2747 5) Each time WHILEM is hit, PL_regcc is the infoblock created by CURLYX
2748 of the same pair. Thus when WHILEM tries to match Z, it temporarily
2749 resets PL_regcc, since this Y(A)*Z can be a part of some other loop:
2750 as in (Y(A)*Z)*. If Z matches, the automaton will hit the WHILEM node
2751 of the external loop.
2753 Currently present infoblocks form a tree with a stem formed by PL_curcc
2754 and whatever it mentions via ->next, and additional attached trees
2755 corresponding to temporarily unset infoblocks as in "5" above.
2757 In the following picture infoblocks for outer loop of
2758 (Y(A)*?Z)*?T are denoted O, for inner I. NULL starting block
2759 is denoted by x. The matched string is YAAZYAZT. Temporarily postponed
2760 infoblocks are drawn below the "reset" infoblock.
2762 In fact in the picture below we do not show failed matches for Z and T
2763 by WHILEM blocks. [We illustrate minimal matches, since for them it is
2764 more obvious *why* one needs to *temporary* unset infoblocks.]
2766 Matched REx position InfoBlocks Comment
2770 Y A)*?Z)*?T x <- O <- I
2771 YA )*?Z)*?T x <- O <- I
2772 YA A)*?Z)*?T x <- O <- I
2773 YAA )*?Z)*?T x <- O <- I
2774 YAA Z)*?T x <- O # Temporary unset I
2777 YAAZ Y(A)*?Z)*?T x <- O
2780 YAAZY (A)*?Z)*?T x <- O
2783 YAAZY A)*?Z)*?T x <- O <- I
2786 YAAZYA )*?Z)*?T x <- O <- I
2789 YAAZYA Z)*?T x <- O # Temporary unset I
2795 YAAZYAZ T x # Temporary unset O
2802 *******************************************************************/
2805 CHECKPOINT cp = PL_savestack_ix;
2806 /* No need to save/restore up to this paren */
2807 I32 parenfloor = scan->flags;
2809 if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */
2811 cc.oldcc = PL_regcc;
2813 /* XXXX Probably it is better to teach regpush to support
2814 parenfloor > PL_regsize... */
2815 if (parenfloor > *PL_reglastparen)
2816 parenfloor = *PL_reglastparen; /* Pessimization... */
2817 cc.parenfloor = parenfloor;
2819 cc.min = ARG1(scan);
2820 cc.max = ARG2(scan);
2821 cc.scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
2825 PL_reginput = locinput;
2826 n = regmatch(PREVOPER(next)); /* start on the WHILEM */
2828 PL_regcc = cc.oldcc;
2834 * This is really hard to understand, because after we match
2835 * what we're trying to match, we must make sure the rest of
2836 * the REx is going to match for sure, and to do that we have
2837 * to go back UP the parse tree by recursing ever deeper. And
2838 * if it fails, we have to reset our parent's current state
2839 * that we can try again after backing off.
2842 CHECKPOINT cp, lastcp;
2843 CURCUR* cc = PL_regcc;
2844 char *lastloc = cc->lastloc; /* Detection of 0-len. */
2846 n = cc->cur + 1; /* how many we know we matched */
2847 PL_reginput = locinput;
2850 PerlIO_printf(Perl_debug_log,
2851 "%*s %ld out of %ld..%ld cc=%lx\n",
2852 REPORT_CODE_OFF+PL_regindent*2, "",
2853 (long)n, (long)cc->min,
2854 (long)cc->max, (long)cc)
2857 /* If degenerate scan matches "", assume scan done. */
2859 if (locinput == cc->lastloc && n >= cc->min) {
2860 PL_regcc = cc->oldcc;
2864 PerlIO_printf(Perl_debug_log,
2865 "%*s empty match detected, try continuation...\n",
2866 REPORT_CODE_OFF+PL_regindent*2, "")
2868 if (regmatch(cc->next))
2876 /* First just match a string of min scans. */
2880 cc->lastloc = locinput;
2881 if (regmatch(cc->scan))
2884 cc->lastloc = lastloc;
2889 /* Check whether we already were at this position.
2890 Postpone detection until we know the match is not
2891 *that* much linear. */
2892 if (!PL_reg_maxiter) {
2893 PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4);
2894 PL_reg_leftiter = PL_reg_maxiter;
2896 if (PL_reg_leftiter-- == 0) {
2897 I32 size = (PL_reg_maxiter + 7)/8;
2898 if (PL_reg_poscache) {
2899 if (PL_reg_poscache_size < size) {
2900 Renew(PL_reg_poscache, size, char);
2901 PL_reg_poscache_size = size;
2903 Zero(PL_reg_poscache, size, char);
2906 PL_reg_poscache_size = size;
2907 Newz(29, PL_reg_poscache, size, char);
2910 PerlIO_printf(Perl_debug_log,
2911 "%sDetected a super-linear match, switching on caching%s...\n",
2912 PL_colors[4], PL_colors[5])
2915 if (PL_reg_leftiter < 0) {
2916 I32 o = locinput - PL_bostr, b;
2918 o = (scan->flags & 0xf) - 1 + o * (scan->flags>>4);
2921 if (PL_reg_poscache[o] & (1<<b)) {
2923 PerlIO_printf(Perl_debug_log,
2924 "%*s already tried at this position...\n",
2925 REPORT_CODE_OFF+PL_regindent*2, "")
2929 PL_reg_poscache[o] |= (1<<b);
2933 /* Prefer next over scan for minimal matching. */
2936 PL_regcc = cc->oldcc;
2939 cp = regcppush(cc->parenfloor);
2941 if (regmatch(cc->next)) {
2943 sayYES; /* All done. */
2945 REGCP_UNWIND(lastcp);
2951 if (n >= cc->max) { /* Maximum greed exceeded? */
2952 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
2953 && !(PL_reg_flags & RF_warned)) {
2954 PL_reg_flags |= RF_warned;
2955 Perl_warner(aTHX_ WARN_REGEXP, "%s limit (%d) exceeded",
2956 "Complex regular subexpression recursion",
2963 PerlIO_printf(Perl_debug_log,
2964 "%*s trying longer...\n",
2965 REPORT_CODE_OFF+PL_regindent*2, "")
2967 /* Try scanning more and see if it helps. */
2968 PL_reginput = locinput;
2970 cc->lastloc = locinput;
2971 cp = regcppush(cc->parenfloor);
2973 if (regmatch(cc->scan)) {
2977 REGCP_UNWIND(lastcp);
2980 cc->lastloc = lastloc;
2984 /* Prefer scan over next for maximal matching. */
2986 if (n < cc->max) { /* More greed allowed? */
2987 cp = regcppush(cc->parenfloor);
2989 cc->lastloc = locinput;
2991 if (regmatch(cc->scan)) {
2995 REGCP_UNWIND(lastcp);
2996 regcppop(); /* Restore some previous $<digit>s? */
2997 PL_reginput = locinput;
2999 PerlIO_printf(Perl_debug_log,
3000 "%*s failed, try continuation...\n",
3001 REPORT_CODE_OFF+PL_regindent*2, "")
3004 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3005 && !(PL_reg_flags & RF_warned)) {
3006 PL_reg_flags |= RF_warned;
3007 Perl_warner(aTHX_ WARN_REGEXP, "%s limit (%d) exceeded",
3008 "Complex regular subexpression recursion",
3012 /* Failed deeper matches of scan, so see if this one works. */
3013 PL_regcc = cc->oldcc;
3016 if (regmatch(cc->next))
3022 cc->lastloc = lastloc;
3027 next = scan + ARG(scan);
3030 inner = NEXTOPER(NEXTOPER(scan));
3033 inner = NEXTOPER(scan);
3037 if (OP(next) != c1) /* No choice. */
3038 next = inner; /* Avoid recursion. */
3040 I32 lastparen = *PL_reglastparen;
3042 re_unwind_branch_t *uw;
3044 /* Put unwinding data on stack */
3045 unwind1 = SSNEWt(1,re_unwind_branch_t);
3046 uw = SSPTRt(unwind1,re_unwind_branch_t);
3049 uw->type = ((c1 == BRANCH)
3051 : RE_UNWIND_BRANCHJ);
3052 uw->lastparen = lastparen;
3054 uw->locinput = locinput;
3055 uw->nextchr = nextchr;
3057 uw->regindent = ++PL_regindent;
3060 REGCP_SET(uw->lastcp);
3062 /* Now go into the first branch */
3075 /* We suppose that the next guy does not need
3076 backtracking: in particular, it is of constant length,
3077 and has no parenths to influence future backrefs. */
3078 ln = ARG1(scan); /* min to match */
3079 n = ARG2(scan); /* max to match */
3080 paren = scan->flags;
3082 if (paren > PL_regsize)
3084 if (paren > *PL_reglastparen)
3085 *PL_reglastparen = paren;
3087 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3089 scan += NEXT_OFF(scan); /* Skip former OPEN. */
3090 PL_reginput = locinput;
3093 if (ln && regrepeat_hard(scan, ln, &l) < ln)
3095 /* if we matched something zero-length we don't need to
3096 backtrack - capturing parens are already defined, so
3097 the caveat in the maximal case doesn't apply
3099 XXXX if ln == 0, we can redo this check first time
3100 through the following loop
3103 n = ln; /* don't backtrack */
3104 locinput = PL_reginput;
3105 if (NEAR_EXACT(next)) {
3106 regnode *text_node = next;
3108 if (PL_regkind[(U8)OP(next)] != EXACT)
3109 NEXT_IMPT(text_node);
3111 if (PL_regkind[(U8)OP(text_node)] != EXACT) {
3115 c1 = (U8)*STRING(text_node);
3116 if (OP(next) == EXACTF)
3118 else if (OP(text_node) == EXACTFL)
3119 c2 = PL_fold_locale[c1];
3127 /* This may be improved if l == 0. */
3128 while (n >= ln || (n == REG_INFTY && ln > 0 && l)) { /* ln overflow ? */
3129 /* If it could work, try it. */
3131 UCHARAT(PL_reginput) == c1 ||
3132 UCHARAT(PL_reginput) == c2)
3136 PL_regstartp[paren] =
3137 HOPc(PL_reginput, -l) - PL_bostr;
3138 PL_regendp[paren] = PL_reginput - PL_bostr;
3141 PL_regendp[paren] = -1;
3145 REGCP_UNWIND(lastcp);
3147 /* Couldn't or didn't -- move forward. */
3148 PL_reginput = locinput;
3149 if (regrepeat_hard(scan, 1, &l)) {
3151 locinput = PL_reginput;
3158 n = regrepeat_hard(scan, n, &l);
3159 /* if we matched something zero-length we don't need to
3160 backtrack, unless the minimum count is zero and we
3161 are capturing the result - in that case the capture
3162 being defined or not may affect later execution
3164 if (n != 0 && l == 0 && !(paren && ln == 0))
3165 ln = n; /* don't backtrack */
3166 locinput = PL_reginput;
3168 PerlIO_printf(Perl_debug_log,
3169 "%*s matched %"IVdf" times, len=%"IVdf"...\n",
3170 (int)(REPORT_CODE_OFF+PL_regindent*2), "",
3174 if (NEAR_EXACT(next)) {
3175 regnode *text_node = next;
3177 if (PL_regkind[(U8)OP(next)] != EXACT)
3178 NEXT_IMPT(text_node);
3180 if (PL_regkind[(U8)OP(text_node)] != EXACT) {
3184 c1 = (U8)*STRING(text_node);
3185 if (OP(text_node) == EXACTF)
3187 else if (OP(text_node) == EXACTFL)
3188 c2 = PL_fold_locale[c1];
3198 /* If it could work, try it. */
3200 UCHARAT(PL_reginput) == c1 ||
3201 UCHARAT(PL_reginput) == c2)
3204 PerlIO_printf(Perl_debug_log,
3205 "%*s trying tail with n=%"IVdf"...\n",
3206 (int)(REPORT_CODE_OFF+PL_regindent*2), "", (IV)n)
3210 PL_regstartp[paren] = HOPc(PL_reginput, -l) - PL_bostr;
3211 PL_regendp[paren] = PL_reginput - PL_bostr;
3214 PL_regendp[paren] = -1;
3218 REGCP_UNWIND(lastcp);
3220 /* Couldn't or didn't -- back up. */
3222 locinput = HOPc(locinput, -l);
3223 PL_reginput = locinput;
3230 paren = scan->flags; /* Which paren to set */
3231 if (paren > PL_regsize)
3233 if (paren > *PL_reglastparen)
3234 *PL_reglastparen = paren;
3235 ln = ARG1(scan); /* min to match */
3236 n = ARG2(scan); /* max to match */
3237 scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE);
3241 ln = ARG1(scan); /* min to match */
3242 n = ARG2(scan); /* max to match */
3243 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3248 scan = NEXTOPER(scan);
3254 scan = NEXTOPER(scan);
3258 * Lookahead to avoid useless match attempts
3259 * when we know what character comes next.
3263 * Used to only do .*x and .*?x, but now it allows
3264 * for )'s, ('s and (?{ ... })'s to be in the way
3265 * of the quantifier and the EXACT-like node. -- japhy
3268 if (NEAR_EXACT(next)) {
3270 regnode *text_node = next;
3272 if (PL_regkind[(U8)OP(next)] != EXACT)
3273 NEXT_IMPT(text_node);
3275 if (PL_regkind[(U8)OP(text_node)] != EXACT) {
3279 s = (U8*)STRING(text_node);
3283 if (OP(text_node) == EXACTF)
3285 else if (OP(text_node) == EXACTFL)
3286 c2 = PL_fold_locale[c1];
3289 if (OP(text_node) == EXACTF) {
3290 STRLEN ulen1, ulen2;
3291 U8 tmpbuf1[UTF8_MAXLEN*2+1];
3292 U8 tmpbuf2[UTF8_MAXLEN*2+1];
3294 to_utf8_lower((U8*)s, tmpbuf1, &ulen1);
3295 to_utf8_upper((U8*)s, tmpbuf2, &ulen2);
3297 c1 = utf8_to_uvuni(tmpbuf1, 0);
3298 c2 = utf8_to_uvuni(tmpbuf2, 0);
3301 c2 = c1 = utf8_to_uvchr(s, NULL);
3308 PL_reginput = locinput;
3312 if (ln && regrepeat(scan, ln) < ln)
3314 locinput = PL_reginput;
3317 char *e; /* Should not check after this */
3318 char *old = locinput;
3320 if (n == REG_INFTY) {
3323 while (UTF8_IS_CONTINUATION(*(U8*)e))
3329 m >0 && e + UTF8SKIP(e) <= PL_regeol; m--)
3333 e = locinput + n - ln;
3339 /* Find place 'next' could work */
3342 while (locinput <= e &&
3343 UCHARAT(locinput) != c1)
3346 while (locinput <= e
3347 && UCHARAT(locinput) != c1
3348 && UCHARAT(locinput) != c2)
3351 count = locinput - old;
3358 utf8_to_uvchr((U8*)locinput, &len) != c1;
3363 for (count = 0; locinput <= e; count++) {
3364 UV c = utf8_to_uvchr((U8*)locinput, &len);
3365 if (c == c1 || c == c2)
3373 /* PL_reginput == old now */
3374 if (locinput != old) {
3375 ln = 1; /* Did some */
3376 if (regrepeat(scan, count) < count)
3379 /* PL_reginput == locinput now */
3380 TRYPAREN(paren, ln, locinput);
3381 PL_reginput = locinput; /* Could be reset... */
3382 REGCP_UNWIND(lastcp);
3383 /* Couldn't or didn't -- move forward. */
3386 locinput += UTF8SKIP(locinput);
3392 while (n >= ln || (n == REG_INFTY && ln > 0)) { /* ln overflow ? */
3396 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3398 c = UCHARAT(PL_reginput);
3399 /* If it could work, try it. */
3400 if (c == c1 || c == c2)
3402 TRYPAREN(paren, n, PL_reginput);
3403 REGCP_UNWIND(lastcp);
3406 /* If it could work, try it. */
3407 else if (c1 == -1000)
3409 TRYPAREN(paren, n, PL_reginput);
3410 REGCP_UNWIND(lastcp);
3412 /* Couldn't or didn't -- move forward. */
3413 PL_reginput = locinput;
3414 if (regrepeat(scan, 1)) {
3416 locinput = PL_reginput;
3424 n = regrepeat(scan, n);
3425 locinput = PL_reginput;
3426 if (ln < n && PL_regkind[(U8)OP(next)] == EOL &&
3427 (!PL_multiline || OP(next) == SEOL || OP(next) == EOS)) {
3428 ln = n; /* why back off? */
3429 /* ...because $ and \Z can match before *and* after
3430 newline at the end. Consider "\n\n" =~ /\n+\Z\n/.
3431 We should back off by one in this case. */
3432 if (UCHARAT(PL_reginput - 1) == '\n' && OP(next) != EOS)
3441 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3443 c = UCHARAT(PL_reginput);
3445 /* If it could work, try it. */
3446 if (c1 == -1000 || c == c1 || c == c2)
3448 TRYPAREN(paren, n, PL_reginput);
3449 REGCP_UNWIND(lastcp);
3451 /* Couldn't or didn't -- back up. */
3453 PL_reginput = locinput = HOPc(locinput, -1);
3461 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3463 c = UCHARAT(PL_reginput);
3465 /* If it could work, try it. */
3466 if (c1 == -1000 || c == c1 || c == c2)
3468 TRYPAREN(paren, n, PL_reginput);
3469 REGCP_UNWIND(lastcp);
3471 /* Couldn't or didn't -- back up. */
3473 PL_reginput = locinput = HOPc(locinput, -1);
3480 if (PL_reg_call_cc) {
3481 re_cc_state *cur_call_cc = PL_reg_call_cc;
3482 CURCUR *cctmp = PL_regcc;
3483 regexp *re = PL_reg_re;
3484 CHECKPOINT cp, lastcp;
3486 cp = regcppush(0); /* Save *all* the positions. */
3488 regcp_set_to(PL_reg_call_cc->ss); /* Restore parens of
3490 PL_reginput = locinput; /* Make position available to
3492 cache_re(PL_reg_call_cc->re);
3493 PL_regcc = PL_reg_call_cc->cc;
3494 PL_reg_call_cc = PL_reg_call_cc->prev;
3495 if (regmatch(cur_call_cc->node)) {
3496 PL_reg_call_cc = cur_call_cc;
3500 REGCP_UNWIND(lastcp);
3502 PL_reg_call_cc = cur_call_cc;
3508 PerlIO_printf(Perl_debug_log,
3509 "%*s continuation failed...\n",
3510 REPORT_CODE_OFF+PL_regindent*2, "")
3514 if (locinput < PL_regtill) {
3515 DEBUG_r(PerlIO_printf(Perl_debug_log,
3516 "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n",
3518 (long)(locinput - PL_reg_starttry),
3519 (long)(PL_regtill - PL_reg_starttry),
3521 sayNO_FINAL; /* Cannot match: too short. */
3523 PL_reginput = locinput; /* put where regtry can find it */
3524 sayYES_FINAL; /* Success! */
3526 PL_reginput = locinput; /* put where regtry can find it */
3527 sayYES_LOUD; /* Success! */
3530 PL_reginput = locinput;
3535 s = HOPBACKc(locinput, scan->flags);
3541 PL_reginput = locinput;
3546 s = HOPBACKc(locinput, scan->flags);
3552 PL_reginput = locinput;
3555 inner = NEXTOPER(NEXTOPER(scan));
3556 if (regmatch(inner) != n) {
3571 if (OP(scan) == SUSPEND) {
3572 locinput = PL_reginput;
3573 nextchr = UCHARAT(locinput);
3578 next = scan + ARG(scan);
3583 PerlIO_printf(Perl_error_log, "%"UVxf" %d\n",
3584 PTR2UV(scan), OP(scan));
3585 Perl_croak(aTHX_ "regexp memory corruption");
3592 * We get here only if there's trouble -- normally "case END" is
3593 * the terminating point.
3595 Perl_croak(aTHX_ "corrupted regexp pointers");
3601 PerlIO_printf(Perl_debug_log,
3602 "%*s %scould match...%s\n",
3603 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4],PL_colors[5])
3607 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n",
3608 PL_colors[4],PL_colors[5]));
3614 #if 0 /* Breaks $^R */
3622 PerlIO_printf(Perl_debug_log,
3623 "%*s %sfailed...%s\n",
3624 REPORT_CODE_OFF+PL_regindent*2, "",PL_colors[4],PL_colors[5])
3630 re_unwind_t *uw = SSPTRt(unwind,re_unwind_t);
3633 case RE_UNWIND_BRANCH:
3634 case RE_UNWIND_BRANCHJ:
3636 re_unwind_branch_t *uwb = &(uw->branch);
3637 I32 lastparen = uwb->lastparen;
3639 REGCP_UNWIND(uwb->lastcp);
3640 for (n = *PL_reglastparen; n > lastparen; n--)
3642 *PL_reglastparen = n;
3643 scan = next = uwb->next;
3645 OP(scan) != (uwb->type == RE_UNWIND_BRANCH
3646 ? BRANCH : BRANCHJ) ) { /* Failure */
3653 /* Have more choice yet. Reuse the same uwb. */
3655 if ((n = (uwb->type == RE_UNWIND_BRANCH
3656 ? NEXT_OFF(next) : ARG(next))))
3659 next = NULL; /* XXXX Needn't unwinding in this case... */
3661 next = NEXTOPER(scan);
3662 if (uwb->type == RE_UNWIND_BRANCHJ)
3663 next = NEXTOPER(next);
3664 locinput = uwb->locinput;
3665 nextchr = uwb->nextchr;
3667 PL_regindent = uwb->regindent;
3674 Perl_croak(aTHX_ "regexp unwind memory corruption");
3685 - regrepeat - repeatedly match something simple, report how many
3688 * [This routine now assumes that it will only match on things of length 1.
3689 * That was true before, but now we assume scan - reginput is the count,
3690 * rather than incrementing count on every character. [Er, except utf8.]]
3693 S_regrepeat(pTHX_ regnode *p, I32 max)
3695 register char *scan;
3697 register char *loceol = PL_regeol;
3698 register I32 hardcount = 0;
3699 register bool do_utf8 = PL_reg_match_utf8;
3702 if (max != REG_INFTY && max < loceol - scan)
3703 loceol = scan + max;
3708 while (scan < loceol && hardcount < max && *scan != '\n') {
3709 scan += UTF8SKIP(scan);
3713 while (scan < loceol && *scan != '\n')
3723 case EXACT: /* length of string is 1 */
3725 while (scan < loceol && UCHARAT(scan) == c)
3728 case EXACTF: /* length of string is 1 */
3730 while (scan < loceol &&
3731 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold[c]))
3734 case EXACTFL: /* length of string is 1 */
3735 PL_reg_flags |= RF_tainted;
3737 while (scan < loceol &&
3738 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold_locale[c]))
3744 while (hardcount < max && scan < loceol &&
3745 reginclass(p, (U8*)scan, do_utf8)) {
3746 scan += UTF8SKIP(scan);
3750 while (scan < loceol && reginclass(p, (U8*)scan, do_utf8))
3757 LOAD_UTF8_CHARCLASS(alnum,"a");
3758 while (hardcount < max && scan < loceol &&
3759 swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
3760 scan += UTF8SKIP(scan);
3764 while (scan < loceol && isALNUM(*scan))
3769 PL_reg_flags |= RF_tainted;
3772 while (hardcount < max && scan < loceol &&
3773 isALNUM_LC_utf8((U8*)scan)) {
3774 scan += UTF8SKIP(scan);
3778 while (scan < loceol && isALNUM_LC(*scan))
3785 LOAD_UTF8_CHARCLASS(alnum,"a");
3786 while (hardcount < max && scan < loceol &&
3787 !swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
3788 scan += UTF8SKIP(scan);
3792 while (scan < loceol && !isALNUM(*scan))
3797 PL_reg_flags |= RF_tainted;
3800 while (hardcount < max && scan < loceol &&
3801 !isALNUM_LC_utf8((U8*)scan)) {
3802 scan += UTF8SKIP(scan);
3806 while (scan < loceol && !isALNUM_LC(*scan))
3813 LOAD_UTF8_CHARCLASS(space," ");
3814 while (hardcount < max && scan < loceol &&
3816 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
3817 scan += UTF8SKIP(scan);
3821 while (scan < loceol && isSPACE(*scan))
3826 PL_reg_flags |= RF_tainted;
3829 while (hardcount < max && scan < loceol &&
3830 (*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
3831 scan += UTF8SKIP(scan);
3835 while (scan < loceol && isSPACE_LC(*scan))
3842 LOAD_UTF8_CHARCLASS(space," ");
3843 while (hardcount < max && scan < loceol &&
3845 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
3846 scan += UTF8SKIP(scan);
3850 while (scan < loceol && !isSPACE(*scan))
3855 PL_reg_flags |= RF_tainted;
3858 while (hardcount < max && scan < loceol &&
3859 !(*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
3860 scan += UTF8SKIP(scan);
3864 while (scan < loceol && !isSPACE_LC(*scan))
3871 LOAD_UTF8_CHARCLASS(digit,"0");
3872 while (hardcount < max && scan < loceol &&
3873 swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
3874 scan += UTF8SKIP(scan);
3878 while (scan < loceol && isDIGIT(*scan))
3885 LOAD_UTF8_CHARCLASS(digit,"0");
3886 while (hardcount < max && scan < loceol &&
3887 !swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
3888 scan += UTF8SKIP(scan);
3892 while (scan < loceol && !isDIGIT(*scan))
3896 default: /* Called on something of 0 width. */
3897 break; /* So match right here or not at all. */
3903 c = scan - PL_reginput;
3908 SV *prop = sv_newmortal();
3911 PerlIO_printf(Perl_debug_log,
3912 "%*s %s can match %"IVdf" times out of %"IVdf"...\n",
3913 REPORT_CODE_OFF+1, "", SvPVX(prop),(IV)c,(IV)max);
3920 - regrepeat_hard - repeatedly match something, report total lenth and length
3922 * The repeater is supposed to have constant length.
3926 S_regrepeat_hard(pTHX_ regnode *p, I32 max, I32 *lp)
3928 register char *scan = Nullch;
3929 register char *start;
3930 register char *loceol = PL_regeol;
3932 I32 count = 0, res = 1;
3937 start = PL_reginput;
3938 if (PL_reg_match_utf8) {
3939 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
3942 while (start < PL_reginput) {
3944 start += UTF8SKIP(start);
3955 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
3957 *lp = l = PL_reginput - start;
3958 if (max != REG_INFTY && l*max < loceol - scan)
3959 loceol = scan + l*max;
3972 - regclass_swash - prepare the utf8 swash
3976 Perl_regclass_swash(pTHX_ register regnode* node, bool doinit, SV** initsvp)
3981 if (PL_regdata && PL_regdata->count) {
3984 if (PL_regdata->what[n] == 's') {
3985 SV *rv = (SV*)PL_regdata->data[n];
3986 AV *av = (AV*)SvRV((SV*)rv);
3989 si = *av_fetch(av, 0, FALSE);
3990 a = av_fetch(av, 1, FALSE);
3994 else if (si && doinit) {
3995 sw = swash_init("utf8", "", si, 1, 0);
3996 (void)av_store(av, 1, sw);
4008 - reginclass - determine if a character falls into a character class
4012 S_reginclass(pTHX_ register regnode *n, register U8* p, register bool do_utf8)
4014 char flags = ANYOF_FLAGS(n);
4019 c = do_utf8 ? utf8_to_uvchr(p, &len) : *p;
4021 if (do_utf8 || (flags & ANYOF_UNICODE)) {
4022 if (do_utf8 && !ANYOF_RUNTIME(n)) {
4023 if (len != (STRLEN)-1 && c < 256 && ANYOF_BITMAP_TEST(n, c))
4026 if (!match && do_utf8 && (flags & ANYOF_UNICODE_ALL) && c >= 256)
4029 SV *sw = regclass_swash(n, TRUE, 0);
4032 if (swash_fetch(sw, p, do_utf8))
4034 else if (flags & ANYOF_FOLD) {
4036 U8 tmpbuf[UTF8_MAXLEN*2+1];
4038 toLOWER_utf8(p, tmpbuf, &ulen);
4039 if (swash_fetch(sw, tmpbuf, do_utf8))
4045 if (!match && c < 256) {
4046 if (ANYOF_BITMAP_TEST(n, c))
4048 else if (flags & ANYOF_FOLD) {
4051 if (flags & ANYOF_LOCALE) {
4052 PL_reg_flags |= RF_tainted;
4053 f = PL_fold_locale[c];
4057 if (f != c && ANYOF_BITMAP_TEST(n, f))
4061 if (!match && (flags & ANYOF_CLASS)) {
4062 PL_reg_flags |= RF_tainted;
4064 (ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) ||
4065 (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) ||
4066 (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) ||
4067 (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) ||
4068 (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) ||
4069 (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) ||
4070 (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) ||
4071 (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) ||
4072 (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) ||
4073 (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) ||
4074 (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII(c)) ||
4075 (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII(c)) ||
4076 (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) ||
4077 (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) ||
4078 (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) ||
4079 (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) ||
4080 (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) ||
4081 (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) ||
4082 (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) ||
4083 (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) ||
4084 (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) ||
4085 (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) ||
4086 (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) ||
4087 (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) ||
4088 (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) ||
4089 (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) ||
4090 (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) ||
4091 (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) ||
4092 (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK(c)) ||
4093 (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK(c))
4094 ) /* How's that for a conditional? */
4101 return (flags & ANYOF_INVERT) ? !match : match;
4105 S_reghop(pTHX_ U8 *s, I32 off)
4107 return S_reghop3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4111 S_reghop3(pTHX_ U8 *s, I32 off, U8* lim)
4114 while (off-- && s < lim) {
4115 /* XXX could check well-formedness here */
4123 if (UTF8_IS_CONTINUED(*s)) {
4124 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4127 /* XXX could check well-formedness here */
4135 S_reghopmaybe(pTHX_ U8 *s, I32 off)
4137 return S_reghopmaybe3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4141 S_reghopmaybe3(pTHX_ U8* s, I32 off, U8* lim)
4144 while (off-- && s < lim) {
4145 /* XXX could check well-formedness here */
4155 if (UTF8_IS_CONTINUED(*s)) {
4156 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4159 /* XXX could check well-formedness here */
4171 restore_pos(pTHX_ void *arg)
4173 if (PL_reg_eval_set) {
4174 if (PL_reg_oldsaved) {
4175 PL_reg_re->subbeg = PL_reg_oldsaved;
4176 PL_reg_re->sublen = PL_reg_oldsavedlen;
4177 RX_MATCH_COPIED_on(PL_reg_re);
4179 PL_reg_magic->mg_len = PL_reg_oldpos;
4180 PL_reg_eval_set = 0;
4181 PL_curpm = PL_reg_oldcurpm;