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 || \
135 OP(rn) == PLUS || OP(rn) == MINMOD || \
136 (PL_regkind[(U8)OP(rn)] == CURLY && ARG1(rn) > 0) \
139 #define HAS_TEXT(rn) ( \
140 PL_regkind[(U8)OP(rn)] == EXACT || PL_regkind[(U8)OP(rn)] == REF \
143 #define FIND_NEXT_IMPT(rn) STMT_START { \
144 while (JUMPABLE(rn)) \
145 if (OP(rn) == SUSPEND || OP(rn) == IFMATCH || \
146 PL_regkind[(U8)OP(rn)] == CURLY) \
147 rn = NEXTOPER(NEXTOPER(rn)); \
148 else if (OP(rn) == PLUS) \
150 else rn += NEXT_OFF(rn); \
153 static void restore_pos(pTHX_ void *arg);
156 S_regcppush(pTHX_ I32 parenfloor)
158 int retval = PL_savestack_ix;
159 #define REGCP_PAREN_ELEMS 4
160 int paren_elems_to_push = (PL_regsize - parenfloor) * REGCP_PAREN_ELEMS;
163 if (paren_elems_to_push < 0)
164 Perl_croak(aTHX_ "panic: paren_elems_to_push < 0");
166 #define REGCP_OTHER_ELEMS 6
167 SSCHECK(paren_elems_to_push + REGCP_OTHER_ELEMS);
168 for (p = PL_regsize; p > parenfloor; p--) {
169 /* REGCP_PARENS_ELEMS are pushed per pairs of parentheses. */
170 SSPUSHINT(PL_regendp[p]);
171 SSPUSHINT(PL_regstartp[p]);
172 SSPUSHPTR(PL_reg_start_tmp[p]);
175 /* REGCP_OTHER_ELEMS are pushed in any case, parentheses or no. */
176 SSPUSHINT(PL_regsize);
177 SSPUSHINT(*PL_reglastparen);
178 SSPUSHINT(*PL_reglastcloseparen);
179 SSPUSHPTR(PL_reginput);
180 #define REGCP_FRAME_ELEMS 2
181 /* REGCP_FRAME_ELEMS are part of the REGCP_OTHER_ELEMS and
182 * are needed for the regexp context stack bookkeeping. */
183 SSPUSHINT(paren_elems_to_push + REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS);
184 SSPUSHINT(SAVEt_REGCONTEXT); /* Magic cookie. */
189 /* These are needed since we do not localize EVAL nodes: */
190 # define REGCP_SET(cp) DEBUG_r(PerlIO_printf(Perl_debug_log, \
191 " Setting an EVAL scope, savestack=%"IVdf"\n", \
192 (IV)PL_savestack_ix)); cp = PL_savestack_ix
194 # define REGCP_UNWIND(cp) DEBUG_r(cp != PL_savestack_ix ? \
195 PerlIO_printf(Perl_debug_log, \
196 " Clearing an EVAL scope, savestack=%"IVdf"..%"IVdf"\n", \
197 (IV)(cp), (IV)PL_savestack_ix) : 0); regcpblow(cp)
207 /* Pop REGCP_OTHER_ELEMS before the parentheses loop starts. */
209 assert(i == SAVEt_REGCONTEXT); /* Check that the magic cookie is there. */
210 i = SSPOPINT; /* Parentheses elements to pop. */
211 input = (char *) SSPOPPTR;
212 *PL_reglastcloseparen = SSPOPINT;
213 *PL_reglastparen = SSPOPINT;
214 PL_regsize = SSPOPINT;
216 /* Now restore the parentheses context. */
217 for (i -= (REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS);
218 i > 0; i -= REGCP_PAREN_ELEMS) {
219 paren = (U32)SSPOPINT;
220 PL_reg_start_tmp[paren] = (char *) SSPOPPTR;
221 PL_regstartp[paren] = SSPOPINT;
223 if (paren <= *PL_reglastparen)
224 PL_regendp[paren] = tmps;
226 PerlIO_printf(Perl_debug_log,
227 " restoring \\%"UVuf" to %"IVdf"(%"IVdf")..%"IVdf"%s\n",
228 (UV)paren, (IV)PL_regstartp[paren],
229 (IV)(PL_reg_start_tmp[paren] - PL_bostr),
230 (IV)PL_regendp[paren],
231 (paren > *PL_reglastparen ? "(no)" : ""));
235 if (*PL_reglastparen + 1 <= PL_regnpar) {
236 PerlIO_printf(Perl_debug_log,
237 " restoring \\%"IVdf"..\\%"IVdf" to undef\n",
238 (IV)(*PL_reglastparen + 1), (IV)PL_regnpar);
242 /* It would seem that the similar code in regtry()
243 * already takes care of this, and in fact it is in
244 * a better location to since this code can #if 0-ed out
245 * but the code in regtry() is needed or otherwise tests
246 * requiring null fields (pat.t#187 and split.t#{13,14}
247 * (as of patchlevel 7877) will fail. Then again,
248 * this code seems to be necessary or otherwise
249 * building DynaLoader will fail:
250 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
252 for (paren = *PL_reglastparen + 1; paren <= PL_regnpar; paren++) {
253 if (paren > PL_regsize)
254 PL_regstartp[paren] = -1;
255 PL_regendp[paren] = -1;
262 S_regcp_set_to(pTHX_ I32 ss)
264 I32 tmp = PL_savestack_ix;
266 PL_savestack_ix = ss;
268 PL_savestack_ix = tmp;
272 typedef struct re_cc_state
276 struct re_cc_state *prev;
281 #define regcpblow(cp) LEAVE_SCOPE(cp) /* Ignores regcppush()ed data. */
283 #define TRYPAREN(paren, n, input) { \
286 PL_regstartp[paren] = HOPc(input, -1) - PL_bostr; \
287 PL_regendp[paren] = input - PL_bostr; \
290 PL_regendp[paren] = -1; \
292 if (regmatch(next)) \
295 PL_regendp[paren] = -1; \
300 * pregexec and friends
304 - pregexec - match a regexp against a string
307 Perl_pregexec(pTHX_ register regexp *prog, char *stringarg, register char *strend,
308 char *strbeg, I32 minend, SV *screamer, U32 nosave)
309 /* strend: pointer to null at end of string */
310 /* strbeg: real beginning of string */
311 /* minend: end of match must be >=minend after stringarg. */
312 /* nosave: For optimizations. */
315 regexec_flags(prog, stringarg, strend, strbeg, minend, screamer, NULL,
316 nosave ? 0 : REXEC_COPY_STR);
320 S_cache_re(pTHX_ regexp *prog)
322 PL_regprecomp = prog->precomp; /* Needed for FAIL. */
324 PL_regprogram = prog->program;
326 PL_regnpar = prog->nparens;
327 PL_regdata = prog->data;
332 * Need to implement the following flags for reg_anch:
334 * USE_INTUIT_NOML - Useful to call re_intuit_start() first
336 * INTUIT_AUTORITATIVE_NOML - Can trust a positive answer
337 * INTUIT_AUTORITATIVE_ML
338 * INTUIT_ONCE_NOML - Intuit can match in one location only.
341 * Another flag for this function: SECOND_TIME (so that float substrs
342 * with giant delta may be not rechecked).
345 /* Assumptions: if ANCH_GPOS, then strpos is anchored. XXXX Check GPOS logic */
347 /* If SCREAM, then SvPVX(sv) should be compatible with strpos and strend.
348 Otherwise, only SvCUR(sv) is used to get strbeg. */
350 /* XXXX We assume that strpos is strbeg unless sv. */
352 /* XXXX Some places assume that there is a fixed substring.
353 An update may be needed if optimizer marks as "INTUITable"
354 RExen without fixed substrings. Similarly, it is assumed that
355 lengths of all the strings are no more than minlen, thus they
356 cannot come from lookahead.
357 (Or minlen should take into account lookahead.) */
359 /* A failure to find a constant substring means that there is no need to make
360 an expensive call to REx engine, thus we celebrate a failure. Similarly,
361 finding a substring too deep into the string means that less calls to
362 regtry() should be needed.
364 REx compiler's optimizer found 4 possible hints:
365 a) Anchored substring;
367 c) Whether we are anchored (beginning-of-line or \G);
368 d) First node (of those at offset 0) which may distingush positions;
369 We use a)b)d) and multiline-part of c), and try to find a position in the
370 string which does not contradict any of them.
373 /* Most of decisions we do here should have been done at compile time.
374 The nodes of the REx which we used for the search should have been
375 deleted from the finite automaton. */
378 Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos,
379 char *strend, U32 flags, re_scream_pos_data *data)
381 register I32 start_shift = 0;
382 /* Should be nonnegative! */
383 register I32 end_shift = 0;
389 register char *other_last = Nullch; /* other substr checked before this */
390 char *check_at = Nullch; /* check substr found at this pos */
392 char *i_strpos = strpos;
393 SV *dsv = PERL_DEBUG_PAD_ZERO(0);
397 char*s = UTF ? sv_uni_display(dsv, sv, 60, 0) : strpos;
398 int len = UTF ? strlen(s) : strend - strpos;
401 PerlIO_printf(Perl_debug_log,
402 "%sGuessing start of match, REx%s `%s%.60s%s%s' against `%s%.*s%s%s'...\n",
403 PL_colors[4],PL_colors[5],PL_colors[0],
406 (strlen(prog->precomp) > 60 ? "..." : ""),
408 (int)(len > 60 ? 60 : len),
410 (len > 60 ? "..." : "")
414 if (prog->reganch & ROPT_UTF8)
415 PL_reg_flags |= RF_utf8;
417 if (prog->minlen > CHR_DIST((U8*)strend, (U8*)strpos)) {
418 DEBUG_r(PerlIO_printf(Perl_debug_log,
419 "String too short... [re_intuit_start]\n"));
422 strbeg = (sv && SvPOK(sv)) ? strend - SvCUR(sv) : strpos;
424 check = prog->check_substr;
425 if (prog->reganch & ROPT_ANCH) { /* Match at beg-of-str or after \n */
426 ml_anch = !( (prog->reganch & ROPT_ANCH_SINGLE)
427 || ( (prog->reganch & ROPT_ANCH_BOL)
428 && !PL_multiline ) ); /* Check after \n? */
431 if ( !(prog->reganch & (ROPT_ANCH_GPOS /* Checked by the caller */
432 | ROPT_IMPLICIT)) /* not a real BOL */
433 /* SvCUR is not set on references: SvRV and SvPVX overlap */
435 && (strpos != strbeg)) {
436 DEBUG_r(PerlIO_printf(Perl_debug_log, "Not at start...\n"));
439 if (prog->check_offset_min == prog->check_offset_max &&
440 !(prog->reganch & ROPT_CANY_SEEN)) {
441 /* Substring at constant offset from beg-of-str... */
444 s = HOP3c(strpos, prog->check_offset_min, strend);
446 slen = SvCUR(check); /* >= 1 */
448 if ( strend - s > slen || strend - s < slen - 1
449 || (strend - s == slen && strend[-1] != '\n')) {
450 DEBUG_r(PerlIO_printf(Perl_debug_log, "String too long...\n"));
453 /* Now should match s[0..slen-2] */
455 if (slen && (*SvPVX(check) != *s
457 && memNE(SvPVX(check), s, slen)))) {
459 DEBUG_r(PerlIO_printf(Perl_debug_log, "String not equal...\n"));
463 else if (*SvPVX(check) != *s
464 || ((slen = SvCUR(check)) > 1
465 && memNE(SvPVX(check), s, slen)))
467 goto success_at_start;
470 /* Match is anchored, but substr is not anchored wrt beg-of-str. */
472 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
473 end_shift = prog->minlen - start_shift -
474 CHR_SVLEN(check) + (SvTAIL(check) != 0);
476 I32 end = prog->check_offset_max + CHR_SVLEN(check)
477 - (SvTAIL(check) != 0);
478 I32 eshift = CHR_DIST((U8*)strend, (U8*)s) - end;
480 if (end_shift < eshift)
484 else { /* Can match at random position */
487 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
488 /* Should be nonnegative! */
489 end_shift = prog->minlen - start_shift -
490 CHR_SVLEN(check) + (SvTAIL(check) != 0);
493 #ifdef DEBUGGING /* 7/99: reports of failure (with the older version) */
495 Perl_croak(aTHX_ "panic: end_shift");
499 /* Find a possible match in the region s..strend by looking for
500 the "check" substring in the region corrected by start/end_shift. */
501 if (flags & REXEC_SCREAM) {
502 I32 p = -1; /* Internal iterator of scream. */
503 I32 *pp = data ? data->scream_pos : &p;
505 if (PL_screamfirst[BmRARE(check)] >= 0
506 || ( BmRARE(check) == '\n'
507 && (BmPREVIOUS(check) == SvCUR(check) - 1)
509 s = screaminstr(sv, check,
510 start_shift + (s - strbeg), end_shift, pp, 0);
514 *data->scream_olds = s;
516 else if (prog->reganch & ROPT_CANY_SEEN)
517 s = fbm_instr((U8*)(s + start_shift),
518 (U8*)(strend - end_shift),
519 check, PL_multiline ? FBMrf_MULTILINE : 0);
521 s = fbm_instr(HOP3(s, start_shift, strend),
522 HOP3(strend, -end_shift, strbeg),
523 check, PL_multiline ? FBMrf_MULTILINE : 0);
525 /* Update the count-of-usability, remove useless subpatterns,
528 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s %s substr `%s%.*s%s'%s%s",
529 (s ? "Found" : "Did not find"),
530 ((check == prog->anchored_substr) ? "anchored" : "floating"),
532 (int)(SvCUR(check) - (SvTAIL(check)!=0)),
534 PL_colors[1], (SvTAIL(check) ? "$" : ""),
535 (s ? " at offset " : "...\n") ) );
542 /* Finish the diagnostic message */
543 DEBUG_r(PerlIO_printf(Perl_debug_log, "%ld...\n", (long)(s - i_strpos)) );
545 /* Got a candidate. Check MBOL anchoring, and the *other* substr.
546 Start with the other substr.
547 XXXX no SCREAM optimization yet - and a very coarse implementation
548 XXXX /ttx+/ results in anchored=`ttx', floating=`x'. floating will
549 *always* match. Probably should be marked during compile...
550 Probably it is right to do no SCREAM here...
553 if (prog->float_substr && prog->anchored_substr) {
554 /* Take into account the "other" substring. */
555 /* XXXX May be hopelessly wrong for UTF... */
558 if (check == prog->float_substr) {
561 char *last = HOP3c(s, -start_shift, strbeg), *last1, *last2;
564 t = s - prog->check_offset_max;
565 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
566 && (!(prog->reganch & ROPT_UTF8)
567 || ((t = reghopmaybe3_c(s, -(prog->check_offset_max), strpos))
572 t = HOP3c(t, prog->anchored_offset, strend);
573 if (t < other_last) /* These positions already checked */
575 last2 = last1 = HOP3c(strend, -prog->minlen, strbeg);
578 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
579 /* On end-of-str: see comment below. */
580 s = fbm_instr((unsigned char*)t,
581 HOP3(HOP3(last1, prog->anchored_offset, strend)
582 + SvCUR(prog->anchored_substr),
583 -(SvTAIL(prog->anchored_substr)!=0), strbeg),
584 prog->anchored_substr,
585 PL_multiline ? FBMrf_MULTILINE : 0);
586 DEBUG_r(PerlIO_printf(Perl_debug_log,
587 "%s anchored substr `%s%.*s%s'%s",
588 (s ? "Found" : "Contradicts"),
590 (int)(SvCUR(prog->anchored_substr)
591 - (SvTAIL(prog->anchored_substr)!=0)),
592 SvPVX(prog->anchored_substr),
593 PL_colors[1], (SvTAIL(prog->anchored_substr) ? "$" : "")));
595 if (last1 >= last2) {
596 DEBUG_r(PerlIO_printf(Perl_debug_log,
597 ", giving up...\n"));
600 DEBUG_r(PerlIO_printf(Perl_debug_log,
601 ", trying floating at offset %ld...\n",
602 (long)(HOP3c(s1, 1, strend) - i_strpos)));
603 other_last = HOP3c(last1, prog->anchored_offset+1, strend);
604 s = HOP3c(last, 1, strend);
608 DEBUG_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
609 (long)(s - i_strpos)));
610 t = HOP3c(s, -prog->anchored_offset, strbeg);
611 other_last = HOP3c(s, 1, strend);
619 else { /* Take into account the floating substring. */
623 t = HOP3c(s, -start_shift, strbeg);
625 HOP3c(strend, -prog->minlen + prog->float_min_offset, strbeg);
626 if (CHR_DIST((U8*)last, (U8*)t) > prog->float_max_offset)
627 last = HOP3c(t, prog->float_max_offset, strend);
628 s = HOP3c(t, prog->float_min_offset, strend);
631 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
632 /* fbm_instr() takes into account exact value of end-of-str
633 if the check is SvTAIL(ed). Since false positives are OK,
634 and end-of-str is not later than strend we are OK. */
635 s = fbm_instr((unsigned char*)s,
636 (unsigned char*)last + SvCUR(prog->float_substr)
637 - (SvTAIL(prog->float_substr)!=0),
638 prog->float_substr, PL_multiline ? FBMrf_MULTILINE : 0);
639 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s floating substr `%s%.*s%s'%s",
640 (s ? "Found" : "Contradicts"),
642 (int)(SvCUR(prog->float_substr)
643 - (SvTAIL(prog->float_substr)!=0)),
644 SvPVX(prog->float_substr),
645 PL_colors[1], (SvTAIL(prog->float_substr) ? "$" : "")));
648 DEBUG_r(PerlIO_printf(Perl_debug_log,
649 ", giving up...\n"));
652 DEBUG_r(PerlIO_printf(Perl_debug_log,
653 ", trying anchored starting at offset %ld...\n",
654 (long)(s1 + 1 - i_strpos)));
656 s = HOP3c(t, 1, strend);
660 DEBUG_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
661 (long)(s - i_strpos)));
662 other_last = s; /* Fix this later. --Hugo */
671 t = s - prog->check_offset_max;
672 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
673 && (!(prog->reganch & ROPT_UTF8)
674 || ((t = reghopmaybe3_c(s, -prog->check_offset_max, strpos))
676 /* Fixed substring is found far enough so that the match
677 cannot start at strpos. */
679 if (ml_anch && t[-1] != '\n') {
680 /* Eventually fbm_*() should handle this, but often
681 anchored_offset is not 0, so this check will not be wasted. */
682 /* XXXX In the code below we prefer to look for "^" even in
683 presence of anchored substrings. And we search even
684 beyond the found float position. These pessimizations
685 are historical artefacts only. */
687 while (t < strend - prog->minlen) {
689 if (t < check_at - prog->check_offset_min) {
690 if (prog->anchored_substr) {
691 /* Since we moved from the found position,
692 we definitely contradict the found anchored
693 substr. Due to the above check we do not
694 contradict "check" substr.
695 Thus we can arrive here only if check substr
696 is float. Redo checking for "other"=="fixed".
699 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld, rescanning for anchored from offset %ld...\n",
700 PL_colors[0],PL_colors[1], (long)(strpos - i_strpos), (long)(strpos - i_strpos + prog->anchored_offset)));
701 goto do_other_anchored;
703 /* We don't contradict the found floating substring. */
704 /* XXXX Why not check for STCLASS? */
706 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld...\n",
707 PL_colors[0],PL_colors[1], (long)(s - i_strpos)));
710 /* Position contradicts check-string */
711 /* XXXX probably better to look for check-string
712 than for "\n", so one should lower the limit for t? */
713 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m, restarting lookup for check-string at offset %ld...\n",
714 PL_colors[0],PL_colors[1], (long)(t + 1 - i_strpos)));
715 other_last = strpos = s = t + 1;
720 DEBUG_r(PerlIO_printf(Perl_debug_log, "Did not find /%s^%s/m...\n",
721 PL_colors[0],PL_colors[1]));
725 DEBUG_r(PerlIO_printf(Perl_debug_log, "Starting position does not contradict /%s^%s/m...\n",
726 PL_colors[0],PL_colors[1]));
730 ++BmUSEFUL(prog->check_substr); /* hooray/5 */
733 /* The found string does not prohibit matching at strpos,
734 - no optimization of calling REx engine can be performed,
735 unless it was an MBOL and we are not after MBOL,
736 or a future STCLASS check will fail this. */
738 /* Even in this situation we may use MBOL flag if strpos is offset
739 wrt the start of the string. */
740 if (ml_anch && sv && !SvROK(sv) /* See prev comment on SvROK */
741 && (strpos != strbeg) && strpos[-1] != '\n'
742 /* May be due to an implicit anchor of m{.*foo} */
743 && !(prog->reganch & ROPT_IMPLICIT))
748 DEBUG_r( if (ml_anch)
749 PerlIO_printf(Perl_debug_log, "Position at offset %ld does not contradict /%s^%s/m...\n",
750 (long)(strpos - i_strpos), PL_colors[0],PL_colors[1]);
753 if (!(prog->reganch & ROPT_NAUGHTY) /* XXXX If strpos moved? */
754 && prog->check_substr /* Could be deleted already */
755 && --BmUSEFUL(prog->check_substr) < 0
756 && prog->check_substr == prog->float_substr)
758 /* If flags & SOMETHING - do not do it many times on the same match */
759 DEBUG_r(PerlIO_printf(Perl_debug_log, "... Disabling check substring...\n"));
760 SvREFCNT_dec(prog->check_substr);
761 prog->check_substr = Nullsv; /* disable */
762 prog->float_substr = Nullsv; /* clear */
763 check = Nullsv; /* abort */
765 /* XXXX This is a remnant of the old implementation. It
766 looks wasteful, since now INTUIT can use many
768 prog->reganch &= ~RE_USE_INTUIT;
775 /* XXXX BmUSEFUL already changed, maybe multiple change is meaningful... */
776 if (prog->regstclass) {
777 /* minlen == 0 is possible if regstclass is \b or \B,
778 and the fixed substr is ''$.
779 Since minlen is already taken into account, s+1 is before strend;
780 accidentally, minlen >= 1 guaranties no false positives at s + 1
781 even for \b or \B. But (minlen? 1 : 0) below assumes that
782 regstclass does not come from lookahead... */
783 /* If regstclass takes bytelength more than 1: If charlength==1, OK.
784 This leaves EXACTF only, which is dealt with in find_byclass(). */
785 U8* str = (U8*)STRING(prog->regstclass);
786 int cl_l = (PL_regkind[(U8)OP(prog->regstclass)] == EXACT
787 ? CHR_DIST(str+STR_LEN(prog->regstclass), str)
789 char *endpos = (prog->anchored_substr || ml_anch)
790 ? HOP3c(s, (prog->minlen ? cl_l : 0), strend)
791 : (prog->float_substr
792 ? HOP3c(HOP3c(check_at, -start_shift, strbeg),
795 char *startpos = strbeg;
798 if (prog->reganch & ROPT_UTF8) {
799 PL_regdata = prog->data;
802 s = find_byclass(prog, prog->regstclass, s, endpos, startpos, 1);
807 if (endpos == strend) {
808 DEBUG_r( PerlIO_printf(Perl_debug_log,
809 "Could not match STCLASS...\n") );
812 DEBUG_r( PerlIO_printf(Perl_debug_log,
813 "This position contradicts STCLASS...\n") );
814 if ((prog->reganch & ROPT_ANCH) && !ml_anch)
816 /* Contradict one of substrings */
817 if (prog->anchored_substr) {
818 if (prog->anchored_substr == check) {
819 DEBUG_r( what = "anchored" );
821 s = HOP3c(t, 1, strend);
822 if (s + start_shift + end_shift > strend) {
823 /* XXXX Should be taken into account earlier? */
824 DEBUG_r( PerlIO_printf(Perl_debug_log,
825 "Could not match STCLASS...\n") );
830 DEBUG_r( PerlIO_printf(Perl_debug_log,
831 "Looking for %s substr starting at offset %ld...\n",
832 what, (long)(s + start_shift - i_strpos)) );
835 /* Have both, check_string is floating */
836 if (t + start_shift >= check_at) /* Contradicts floating=check */
837 goto retry_floating_check;
838 /* Recheck anchored substring, but not floating... */
842 DEBUG_r( PerlIO_printf(Perl_debug_log,
843 "Looking for anchored substr starting at offset %ld...\n",
844 (long)(other_last - i_strpos)) );
845 goto do_other_anchored;
847 /* Another way we could have checked stclass at the
848 current position only: */
853 DEBUG_r( PerlIO_printf(Perl_debug_log,
854 "Looking for /%s^%s/m starting at offset %ld...\n",
855 PL_colors[0],PL_colors[1], (long)(t - i_strpos)) );
858 if (!prog->float_substr) /* Could have been deleted */
860 /* Check is floating subtring. */
861 retry_floating_check:
862 t = check_at - start_shift;
863 DEBUG_r( what = "floating" );
864 goto hop_and_restart;
867 DEBUG_r(PerlIO_printf(Perl_debug_log,
868 "By STCLASS: moving %ld --> %ld\n",
869 (long)(t - i_strpos), (long)(s - i_strpos))
873 DEBUG_r(PerlIO_printf(Perl_debug_log,
874 "Does not contradict STCLASS...\n");
879 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s%s:%s match at offset %ld\n",
880 PL_colors[4], (check ? "Guessed" : "Giving up"),
881 PL_colors[5], (long)(s - i_strpos)) );
884 fail_finish: /* Substring not found */
885 if (prog->check_substr) /* could be removed already */
886 BmUSEFUL(prog->check_substr) += 5; /* hooray */
888 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch rejected by optimizer%s\n",
889 PL_colors[4],PL_colors[5]));
893 /* We know what class REx starts with. Try to find this position... */
895 S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, char *strend, char *startpos, I32 norun)
897 I32 doevery = (prog->reganch & ROPT_SKIP) == 0;
903 register I32 tmp = 1; /* Scratch variable? */
904 register bool do_utf8 = PL_reg_match_utf8;
906 /* We know what class it must start with. */
910 if (reginclass(c, (U8*)s, do_utf8)) {
911 if (tmp && (norun || regtry(prog, s)))
918 s += do_utf8 ? UTF8SKIP(s) : 1;
923 if (tmp && (norun || regtry(prog, s)))
935 U8 tmpbuf1[UTF8_MAXLEN_UCLC+1];
936 U8 tmpbuf2[UTF8_MAXLEN_UCLC+1];
938 to_utf8_lower((U8*)m, tmpbuf1, &ulen1);
939 to_utf8_upper((U8*)m, tmpbuf2, &ulen2);
941 c1 = utf8_to_uvuni(tmpbuf1, 0);
942 c2 = utf8_to_uvuni(tmpbuf2, 0);
953 c2 = PL_fold_locale[c1];
958 e = s; /* Due to minlen logic of intuit() */
962 /* The ibcmp_utf8() uses to_uni_fold() which is more
963 * correct folding for Unicode than using lowercase.
964 * However, it doesn't work quite fully since the folding
965 * is a one-to-many mapping and the regex optimizer is
966 * unaware of this, so it may throw out good matches.
967 * Fortunately, not getting this right is allowed
968 * for Unicode Regular Expression Support level 1,
969 * only one-to-one matching is required. --jhi */
972 if ( utf8_to_uvchr((U8*)s, &len) == c1
974 ibcmp_utf8(s, do_utf8, strend - s,
976 && (norun || regtry(prog, s)) )
982 UV c = utf8_to_uvchr((U8*)s, &len);
983 if ( (c == c1 || c == c2)
985 ibcmp_utf8(s, do_utf8, strend - s,
987 && (norun || regtry(prog, s)) )
996 && (ln == 1 || !(OP(c) == EXACTF
998 : ibcmp_locale(s, m, ln)))
999 && (norun || regtry(prog, s)) )
1005 if ( (*(U8*)s == c1 || *(U8*)s == c2)
1006 && (ln == 1 || !(OP(c) == EXACTF
1008 : ibcmp_locale(s, m, ln)))
1009 && (norun || regtry(prog, s)) )
1016 PL_reg_flags |= RF_tainted;
1023 U8 *r = reghop3((U8*)s, -1, (U8*)startpos);
1026 tmp = (I32)utf8n_to_uvchr(r, s - (char*)r, 0, 0);
1028 tmp = ((OP(c) == BOUND ?
1029 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1030 LOAD_UTF8_CHARCLASS(alnum,"a");
1031 while (s < strend) {
1032 if (tmp == !(OP(c) == BOUND ?
1033 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1034 isALNUM_LC_utf8((U8*)s)))
1037 if ((norun || regtry(prog, s)))
1044 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1045 tmp = ((OP(c) == BOUND ? isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1046 while (s < strend) {
1048 !(OP(c) == BOUND ? isALNUM(*s) : isALNUM_LC(*s))) {
1050 if ((norun || regtry(prog, s)))
1056 if ((!prog->minlen && tmp) && (norun || regtry(prog, s)))
1060 PL_reg_flags |= RF_tainted;
1067 U8 *r = reghop3((U8*)s, -1, (U8*)startpos);
1070 tmp = (I32)utf8n_to_uvchr(r, s - (char*)r, 0, 0);
1072 tmp = ((OP(c) == NBOUND ?
1073 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1074 LOAD_UTF8_CHARCLASS(alnum,"a");
1075 while (s < strend) {
1076 if (tmp == !(OP(c) == NBOUND ?
1077 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1078 isALNUM_LC_utf8((U8*)s)))
1080 else if ((norun || regtry(prog, s)))
1086 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1087 tmp = ((OP(c) == NBOUND ?
1088 isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1089 while (s < strend) {
1091 !(OP(c) == NBOUND ? isALNUM(*s) : isALNUM_LC(*s)))
1093 else if ((norun || regtry(prog, s)))
1098 if ((!prog->minlen && !tmp) && (norun || regtry(prog, s)))
1103 LOAD_UTF8_CHARCLASS(alnum,"a");
1104 while (s < strend) {
1105 if (swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1106 if (tmp && (norun || regtry(prog, s)))
1117 while (s < strend) {
1119 if (tmp && (norun || regtry(prog, s)))
1131 PL_reg_flags |= RF_tainted;
1133 while (s < strend) {
1134 if (isALNUM_LC_utf8((U8*)s)) {
1135 if (tmp && (norun || regtry(prog, s)))
1146 while (s < strend) {
1147 if (isALNUM_LC(*s)) {
1148 if (tmp && (norun || regtry(prog, s)))
1161 LOAD_UTF8_CHARCLASS(alnum,"a");
1162 while (s < strend) {
1163 if (!swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1164 if (tmp && (norun || regtry(prog, s)))
1175 while (s < strend) {
1177 if (tmp && (norun || regtry(prog, s)))
1189 PL_reg_flags |= RF_tainted;
1191 while (s < strend) {
1192 if (!isALNUM_LC_utf8((U8*)s)) {
1193 if (tmp && (norun || regtry(prog, s)))
1204 while (s < strend) {
1205 if (!isALNUM_LC(*s)) {
1206 if (tmp && (norun || regtry(prog, s)))
1219 LOAD_UTF8_CHARCLASS(space," ");
1220 while (s < strend) {
1221 if (*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8)) {
1222 if (tmp && (norun || regtry(prog, s)))
1233 while (s < strend) {
1235 if (tmp && (norun || regtry(prog, s)))
1247 PL_reg_flags |= RF_tainted;
1249 while (s < strend) {
1250 if (*s == ' ' || isSPACE_LC_utf8((U8*)s)) {
1251 if (tmp && (norun || regtry(prog, s)))
1262 while (s < strend) {
1263 if (isSPACE_LC(*s)) {
1264 if (tmp && (norun || regtry(prog, s)))
1277 LOAD_UTF8_CHARCLASS(space," ");
1278 while (s < strend) {
1279 if (!(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8))) {
1280 if (tmp && (norun || regtry(prog, s)))
1291 while (s < strend) {
1293 if (tmp && (norun || regtry(prog, s)))
1305 PL_reg_flags |= RF_tainted;
1307 while (s < strend) {
1308 if (!(*s == ' ' || isSPACE_LC_utf8((U8*)s))) {
1309 if (tmp && (norun || regtry(prog, s)))
1320 while (s < strend) {
1321 if (!isSPACE_LC(*s)) {
1322 if (tmp && (norun || regtry(prog, s)))
1335 LOAD_UTF8_CHARCLASS(digit,"0");
1336 while (s < strend) {
1337 if (swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1338 if (tmp && (norun || regtry(prog, s)))
1349 while (s < strend) {
1351 if (tmp && (norun || regtry(prog, s)))
1363 PL_reg_flags |= RF_tainted;
1365 while (s < strend) {
1366 if (isDIGIT_LC_utf8((U8*)s)) {
1367 if (tmp && (norun || regtry(prog, s)))
1378 while (s < strend) {
1379 if (isDIGIT_LC(*s)) {
1380 if (tmp && (norun || regtry(prog, s)))
1393 LOAD_UTF8_CHARCLASS(digit,"0");
1394 while (s < strend) {
1395 if (!swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1396 if (tmp && (norun || regtry(prog, s)))
1407 while (s < strend) {
1409 if (tmp && (norun || regtry(prog, s)))
1421 PL_reg_flags |= RF_tainted;
1423 while (s < strend) {
1424 if (!isDIGIT_LC_utf8((U8*)s)) {
1425 if (tmp && (norun || regtry(prog, s)))
1436 while (s < strend) {
1437 if (!isDIGIT_LC(*s)) {
1438 if (tmp && (norun || regtry(prog, s)))
1450 Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c));
1459 - regexec_flags - match a regexp against a string
1462 Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char *strend,
1463 char *strbeg, I32 minend, SV *sv, void *data, U32 flags)
1464 /* strend: pointer to null at end of string */
1465 /* strbeg: real beginning of string */
1466 /* minend: end of match must be >=minend after stringarg. */
1467 /* data: May be used for some additional optimizations. */
1468 /* nosave: For optimizations. */
1471 register regnode *c;
1472 register char *startpos = stringarg;
1473 I32 minlen; /* must match at least this many chars */
1474 I32 dontbother = 0; /* how many characters not to try at end */
1475 /* I32 start_shift = 0; */ /* Offset of the start to find
1476 constant substr. */ /* CC */
1477 I32 end_shift = 0; /* Same for the end. */ /* CC */
1478 I32 scream_pos = -1; /* Internal iterator of scream. */
1480 SV* oreplsv = GvSV(PL_replgv);
1481 bool do_utf8 = DO_UTF8(sv);
1483 SV *dsv = PERL_DEBUG_PAD_ZERO(0);
1490 PL_regnarrate = DEBUG_r_TEST;
1493 /* Be paranoid... */
1494 if (prog == NULL || startpos == NULL) {
1495 Perl_croak(aTHX_ "NULL regexp parameter");
1499 minlen = prog->minlen;
1500 if (strend - startpos < minlen) {
1501 DEBUG_r(PerlIO_printf(Perl_debug_log,
1502 "String too short [regexec_flags]...\n"));
1506 /* Check validity of program. */
1507 if (UCHARAT(prog->program) != REG_MAGIC) {
1508 Perl_croak(aTHX_ "corrupted regexp program");
1512 PL_reg_eval_set = 0;
1515 if (prog->reganch & ROPT_UTF8)
1516 PL_reg_flags |= RF_utf8;
1518 /* Mark beginning of line for ^ and lookbehind. */
1519 PL_regbol = startpos;
1523 /* Mark end of line for $ (and such) */
1526 /* see how far we have to get to not match where we matched before */
1527 PL_regtill = startpos+minend;
1529 /* We start without call_cc context. */
1532 /* If there is a "must appear" string, look for it. */
1535 if (prog->reganch & ROPT_GPOS_SEEN) { /* Need to have PL_reg_ganch */
1538 if (flags & REXEC_IGNOREPOS) /* Means: check only at start */
1539 PL_reg_ganch = startpos;
1540 else if (sv && SvTYPE(sv) >= SVt_PVMG
1542 && (mg = mg_find(sv, PERL_MAGIC_regex_global))
1543 && mg->mg_len >= 0) {
1544 PL_reg_ganch = strbeg + mg->mg_len; /* Defined pos() */
1545 if (prog->reganch & ROPT_ANCH_GPOS) {
1546 if (s > PL_reg_ganch)
1551 else /* pos() not defined */
1552 PL_reg_ganch = strbeg;
1555 if (do_utf8 == (UTF!=0) &&
1556 !(flags & REXEC_CHECKED) && prog->check_substr != Nullsv) {
1557 re_scream_pos_data d;
1559 d.scream_olds = &scream_olds;
1560 d.scream_pos = &scream_pos;
1561 s = re_intuit_start(prog, sv, s, strend, flags, &d);
1563 DEBUG_r(PerlIO_printf(Perl_debug_log, "Not present...\n"));
1564 goto phooey; /* not present */
1569 char *s = do_utf8 ? sv_uni_display(dsv, sv, 60, 0) : startpos;
1570 int len = do_utf8 ? strlen(s) : strend - startpos;
1573 PerlIO_printf(Perl_debug_log,
1574 "%sMatching REx%s `%s%.60s%s%s' against `%s%.*s%s%s'\n",
1575 PL_colors[4],PL_colors[5],PL_colors[0],
1578 (strlen(prog->precomp) > 60 ? "..." : ""),
1580 (int)(len > 60 ? 60 : len),
1582 (len > 60 ? "..." : "")
1586 /* Simplest case: anchored match need be tried only once. */
1587 /* [unless only anchor is BOL and multiline is set] */
1588 if (prog->reganch & (ROPT_ANCH & ~ROPT_ANCH_GPOS)) {
1589 if (s == startpos && regtry(prog, startpos))
1591 else if (PL_multiline || (prog->reganch & ROPT_IMPLICIT)
1592 || (prog->reganch & ROPT_ANCH_MBOL)) /* XXXX SBOL? */
1597 dontbother = minlen - 1;
1598 end = HOP3c(strend, -dontbother, strbeg) - 1;
1599 /* for multiline we only have to try after newlines */
1600 if (prog->check_substr) {
1604 if (regtry(prog, s))
1609 if (prog->reganch & RE_USE_INTUIT) {
1610 s = re_intuit_start(prog, sv, s + 1, strend, flags, NULL);
1621 if (*s++ == '\n') { /* don't need PL_utf8skip here */
1622 if (regtry(prog, s))
1629 } else if (prog->reganch & ROPT_ANCH_GPOS) {
1630 if (regtry(prog, PL_reg_ganch))
1635 /* Messy cases: unanchored match. */
1636 if (prog->anchored_substr && prog->reganch & ROPT_SKIP) {
1637 /* we have /x+whatever/ */
1638 /* it must be a one character string (XXXX Except UTF?) */
1639 char ch = SvPVX(prog->anchored_substr)[0];
1645 while (s < strend) {
1647 DEBUG_r( did_match = 1 );
1648 if (regtry(prog, s)) goto got_it;
1650 while (s < strend && *s == ch)
1657 while (s < strend) {
1659 DEBUG_r( did_match = 1 );
1660 if (regtry(prog, s)) goto got_it;
1662 while (s < strend && *s == ch)
1668 DEBUG_r(if (!did_match)
1669 PerlIO_printf(Perl_debug_log,
1670 "Did not find anchored character...\n")
1674 else if (do_utf8 == (UTF!=0) &&
1675 (prog->anchored_substr != Nullsv
1676 || (prog->float_substr != Nullsv
1677 && prog->float_max_offset < strend - s))) {
1678 SV *must = prog->anchored_substr
1679 ? prog->anchored_substr : prog->float_substr;
1681 prog->anchored_substr ? prog->anchored_offset : prog->float_max_offset;
1683 prog->anchored_substr ? prog->anchored_offset : prog->float_min_offset;
1684 char *last = HOP3c(strend, /* Cannot start after this */
1685 -(I32)(CHR_SVLEN(must)
1686 - (SvTAIL(must) != 0) + back_min), strbeg);
1687 char *last1; /* Last position checked before */
1693 last1 = HOPc(s, -1);
1695 last1 = s - 1; /* bogus */
1697 /* XXXX check_substr already used to find `s', can optimize if
1698 check_substr==must. */
1700 dontbother = end_shift;
1701 strend = HOPc(strend, -dontbother);
1702 while ( (s <= last) &&
1703 ((flags & REXEC_SCREAM)
1704 ? (s = screaminstr(sv, must, HOP3c(s, back_min, strend) - strbeg,
1705 end_shift, &scream_pos, 0))
1706 : (s = fbm_instr((unsigned char*)HOP3(s, back_min, strend),
1707 (unsigned char*)strend, must,
1708 PL_multiline ? FBMrf_MULTILINE : 0))) ) {
1709 DEBUG_r( did_match = 1 );
1710 if (HOPc(s, -back_max) > last1) {
1711 last1 = HOPc(s, -back_min);
1712 s = HOPc(s, -back_max);
1715 char *t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1;
1717 last1 = HOPc(s, -back_min);
1721 while (s <= last1) {
1722 if (regtry(prog, s))
1728 while (s <= last1) {
1729 if (regtry(prog, s))
1735 DEBUG_r(if (!did_match)
1736 PerlIO_printf(Perl_debug_log,
1737 "Did not find %s substr `%s%.*s%s'%s...\n",
1738 ((must == prog->anchored_substr)
1739 ? "anchored" : "floating"),
1741 (int)(SvCUR(must) - (SvTAIL(must)!=0)),
1743 PL_colors[1], (SvTAIL(must) ? "$" : ""))
1747 else if ((c = prog->regstclass)) {
1748 if (minlen && PL_regkind[(U8)OP(prog->regstclass)] != EXACT)
1749 /* don't bother with what can't match */
1750 strend = HOPc(strend, -(minlen - 1));
1752 SV *prop = sv_newmortal();
1754 PerlIO_printf(Perl_debug_log, "Matching stclass `%s' against `%s'\n", SvPVX(prop), UTF ? sv_uni_display(dsv, sv, 60, 0) : s);
1756 if (find_byclass(prog, c, s, strend, startpos, 0))
1758 DEBUG_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass...\n"));
1762 if (prog->float_substr != Nullsv) { /* Trim the end. */
1765 if (flags & REXEC_SCREAM) {
1766 last = screaminstr(sv, prog->float_substr, s - strbeg,
1767 end_shift, &scream_pos, 1); /* last one */
1769 last = scream_olds; /* Only one occurrence. */
1773 char *little = SvPV(prog->float_substr, len);
1775 if (SvTAIL(prog->float_substr)) {
1776 if (memEQ(strend - len + 1, little, len - 1))
1777 last = strend - len + 1;
1778 else if (!PL_multiline)
1779 last = memEQ(strend - len, little, len)
1780 ? strend - len : Nullch;
1786 last = rninstr(s, strend, little, little + len);
1788 last = strend; /* matching `$' */
1792 DEBUG_r(PerlIO_printf(Perl_debug_log,
1793 "%sCan't trim the tail, match fails (should not happen)%s\n",
1794 PL_colors[4],PL_colors[5]));
1795 goto phooey; /* Should not happen! */
1797 dontbother = strend - last + prog->float_min_offset;
1799 if (minlen && (dontbother < minlen))
1800 dontbother = minlen - 1;
1801 strend -= dontbother; /* this one's always in bytes! */
1802 /* We don't know much -- general case. */
1805 if (regtry(prog, s))
1814 if (regtry(prog, s))
1816 } while (s++ < strend);
1824 RX_MATCH_TAINTED_set(prog, PL_reg_flags & RF_tainted);
1826 if (PL_reg_eval_set) {
1827 /* Preserve the current value of $^R */
1828 if (oreplsv != GvSV(PL_replgv))
1829 sv_setsv(oreplsv, GvSV(PL_replgv));/* So that when GvSV(replgv) is
1830 restored, the value remains
1832 restore_pos(aTHX_ 0);
1835 /* make sure $`, $&, $', and $digit will work later */
1836 if ( !(flags & REXEC_NOT_FIRST) ) {
1837 if (RX_MATCH_COPIED(prog)) {
1838 Safefree(prog->subbeg);
1839 RX_MATCH_COPIED_off(prog);
1841 if (flags & REXEC_COPY_STR) {
1842 I32 i = PL_regeol - startpos + (stringarg - strbeg);
1844 s = savepvn(strbeg, i);
1847 RX_MATCH_COPIED_on(prog);
1850 prog->subbeg = strbeg;
1851 prog->sublen = PL_regeol - strbeg; /* strend may have been modified */
1858 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n",
1859 PL_colors[4],PL_colors[5]));
1860 if (PL_reg_eval_set)
1861 restore_pos(aTHX_ 0);
1866 - regtry - try match at specific point
1868 STATIC I32 /* 0 failure, 1 success */
1869 S_regtry(pTHX_ regexp *prog, char *startpos)
1877 PL_regindent = 0; /* XXXX Not good when matches are reenterable... */
1879 if ((prog->reganch & ROPT_EVAL_SEEN) && !PL_reg_eval_set) {
1882 PL_reg_eval_set = RS_init;
1884 PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %"IVdf"\n",
1885 (IV)(PL_stack_sp - PL_stack_base));
1887 SAVEI32(cxstack[cxstack_ix].blk_oldsp);
1888 cxstack[cxstack_ix].blk_oldsp = PL_stack_sp - PL_stack_base;
1889 /* Otherwise OP_NEXTSTATE will free whatever on stack now. */
1891 /* Apparently this is not needed, judging by wantarray. */
1892 /* SAVEI8(cxstack[cxstack_ix].blk_gimme);
1893 cxstack[cxstack_ix].blk_gimme = G_SCALAR; */
1896 /* Make $_ available to executed code. */
1897 if (PL_reg_sv != DEFSV) {
1898 /* SAVE_DEFSV does *not* suffice here for USE_5005THREADS */
1903 if (!(SvTYPE(PL_reg_sv) >= SVt_PVMG && SvMAGIC(PL_reg_sv)
1904 && (mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global)))) {
1905 /* prepare for quick setting of pos */
1906 sv_magic(PL_reg_sv, (SV*)0,
1907 PERL_MAGIC_regex_global, Nullch, 0);
1908 mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global);
1912 PL_reg_oldpos = mg->mg_len;
1913 SAVEDESTRUCTOR_X(restore_pos, 0);
1915 if (!PL_reg_curpm) {
1916 Newz(22,PL_reg_curpm, 1, PMOP);
1919 SV* repointer = newSViv(0);
1920 /* so we know which PL_regex_padav element is PL_reg_curpm */
1921 SvFLAGS(repointer) |= SVf_BREAK;
1922 av_push(PL_regex_padav,repointer);
1923 PL_reg_curpm->op_pmoffset = av_len(PL_regex_padav);
1924 PL_regex_pad = AvARRAY(PL_regex_padav);
1928 PM_SETRE(PL_reg_curpm, prog);
1929 PL_reg_oldcurpm = PL_curpm;
1930 PL_curpm = PL_reg_curpm;
1931 if (RX_MATCH_COPIED(prog)) {
1932 /* Here is a serious problem: we cannot rewrite subbeg,
1933 since it may be needed if this match fails. Thus
1934 $` inside (?{}) could fail... */
1935 PL_reg_oldsaved = prog->subbeg;
1936 PL_reg_oldsavedlen = prog->sublen;
1937 RX_MATCH_COPIED_off(prog);
1940 PL_reg_oldsaved = Nullch;
1941 prog->subbeg = PL_bostr;
1942 prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */
1944 prog->startp[0] = startpos - PL_bostr;
1945 PL_reginput = startpos;
1946 PL_regstartp = prog->startp;
1947 PL_regendp = prog->endp;
1948 PL_reglastparen = &prog->lastparen;
1949 PL_reglastcloseparen = &prog->lastcloseparen;
1950 prog->lastparen = 0;
1952 DEBUG_r(PL_reg_starttry = startpos);
1953 if (PL_reg_start_tmpl <= prog->nparens) {
1954 PL_reg_start_tmpl = prog->nparens*3/2 + 3;
1955 if(PL_reg_start_tmp)
1956 Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
1958 New(22,PL_reg_start_tmp, PL_reg_start_tmpl, char*);
1962 sv_setpvn(PERL_DEBUG_PAD(0), "", 0);
1963 sv_setpvn(PERL_DEBUG_PAD(1), "", 0);
1964 sv_setpvn(PERL_DEBUG_PAD(2), "", 0);
1967 /* XXXX What this code is doing here?!!! There should be no need
1968 to do this again and again, PL_reglastparen should take care of
1971 /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code.
1972 * Actually, the code in regcppop() (which Ilya may be meaning by
1973 * PL_reglastparen), is not needed at all by the test suite
1974 * (op/regexp, op/pat, op/split), but that code is needed, oddly
1975 * enough, for building DynaLoader, or otherwise this
1976 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
1977 * will happen. Meanwhile, this code *is* needed for the
1978 * above-mentioned test suite tests to succeed. The common theme
1979 * on those tests seems to be returning null fields from matches.
1984 if (prog->nparens) {
1985 for (i = prog->nparens; i > *PL_reglastparen; i--) {
1992 if (regmatch(prog->program + 1)) {
1993 prog->endp[0] = PL_reginput - PL_bostr;
1996 REGCP_UNWIND(lastcp);
2000 #define RE_UNWIND_BRANCH 1
2001 #define RE_UNWIND_BRANCHJ 2
2005 typedef struct { /* XX: makes sense to enlarge it... */
2009 } re_unwind_generic_t;
2022 } re_unwind_branch_t;
2024 typedef union re_unwind_t {
2026 re_unwind_generic_t generic;
2027 re_unwind_branch_t branch;
2030 #define sayYES goto yes
2031 #define sayNO goto no
2032 #define sayYES_FINAL goto yes_final
2033 #define sayYES_LOUD goto yes_loud
2034 #define sayNO_FINAL goto no_final
2035 #define sayNO_SILENT goto do_no
2036 #define saySAME(x) if (x) goto yes; else goto no
2038 #define REPORT_CODE_OFF 24
2041 - regmatch - main matching routine
2043 * Conceptually the strategy is simple: check to see whether the current
2044 * node matches, call self recursively to see whether the rest matches,
2045 * and then act accordingly. In practice we make some effort to avoid
2046 * recursion, in particular by going through "ordinary" nodes (that don't
2047 * need to know whether the rest of the match failed) by a loop instead of
2050 /* [lwall] I've hoisted the register declarations to the outer block in order to
2051 * maybe save a little bit of pushing and popping on the stack. It also takes
2052 * advantage of machines that use a register save mask on subroutine entry.
2054 STATIC I32 /* 0 failure, 1 success */
2055 S_regmatch(pTHX_ regnode *prog)
2057 register regnode *scan; /* Current node. */
2058 regnode *next; /* Next node. */
2059 regnode *inner; /* Next node in internal branch. */
2060 register I32 nextchr; /* renamed nextchr - nextchar colides with
2061 function of same name */
2062 register I32 n; /* no or next */
2063 register I32 ln = 0; /* len or last */
2064 register char *s = Nullch; /* operand or save */
2065 register char *locinput = PL_reginput;
2066 register I32 c1 = 0, c2 = 0, paren; /* case fold search, parenth */
2067 int minmod = 0, sw = 0, logical = 0;
2070 I32 firstcp = PL_savestack_ix;
2072 register bool do_utf8 = PL_reg_match_utf8;
2074 SV *dsv0 = PERL_DEBUG_PAD_ZERO(0);
2075 SV *dsv1 = PERL_DEBUG_PAD_ZERO(1);
2076 SV *dsv2 = PERL_DEBUG_PAD_ZERO(2);
2083 /* Note that nextchr is a byte even in UTF */
2084 nextchr = UCHARAT(locinput);
2086 while (scan != NULL) {
2089 SV *prop = sv_newmortal();
2090 int docolor = *PL_colors[0];
2091 int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */
2092 int l = (PL_regeol - locinput) > taill ? taill : (PL_regeol - locinput);
2093 /* The part of the string before starttry has one color
2094 (pref0_len chars), between starttry and current
2095 position another one (pref_len - pref0_len chars),
2096 after the current position the third one.
2097 We assume that pref0_len <= pref_len, otherwise we
2098 decrease pref0_len. */
2099 int pref_len = (locinput - PL_bostr) > (5 + taill) - l
2100 ? (5 + taill) - l : locinput - PL_bostr;
2103 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len)))
2105 pref0_len = pref_len - (locinput - PL_reg_starttry);
2106 if (l + pref_len < (5 + taill) && l < PL_regeol - locinput)
2107 l = ( PL_regeol - locinput > (5 + taill) - pref_len
2108 ? (5 + taill) - pref_len : PL_regeol - locinput);
2109 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput + l)))
2113 if (pref0_len > pref_len)
2114 pref0_len = pref_len;
2115 regprop(prop, scan);
2119 pv_uni_display(dsv0, (U8*)(locinput - pref_len),
2121 locinput - pref_len;
2122 int len0 = do_utf8 ? strlen(s0) : pref0_len;
2123 char *s1 = do_utf8 ?
2124 pv_uni_display(dsv1, (U8*)(locinput - pref_len + pref0_len),
2125 pref_len - pref0_len, 60, 0) :
2126 locinput - pref_len + pref0_len;
2127 int len1 = do_utf8 ? strlen(s1) : pref_len - pref0_len;
2128 char *s2 = do_utf8 ?
2129 pv_uni_display(dsv2, (U8*)locinput,
2130 PL_regeol - locinput, 60, 0) :
2132 int len2 = do_utf8 ? strlen(s2) : l;
2133 PerlIO_printf(Perl_debug_log,
2134 "%4"IVdf" <%s%.*s%s%s%.*s%s%s%s%.*s%s>%*s|%3"IVdf":%*s%s\n",
2135 (IV)(locinput - PL_bostr),
2142 (docolor ? "" : "> <"),
2146 15 - l - pref_len + 1,
2148 (IV)(scan - PL_regprogram), PL_regindent*2, "",
2153 next = scan + NEXT_OFF(scan);
2159 if (locinput == PL_bostr || (PL_multiline &&
2160 (nextchr || locinput < PL_regeol) && locinput[-1] == '\n') )
2162 /* regtill = regbol; */
2167 if (locinput == PL_bostr ||
2168 ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n'))
2174 if (locinput == PL_bostr)
2178 if (locinput == PL_reg_ganch)
2188 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2193 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2195 if (PL_regeol - locinput > 1)
2199 if (PL_regeol != locinput)
2203 if (!nextchr && locinput >= PL_regeol)
2206 locinput += PL_utf8skip[nextchr];
2207 if (locinput > PL_regeol)
2209 nextchr = UCHARAT(locinput);
2212 nextchr = UCHARAT(++locinput);
2215 if (!nextchr && locinput >= PL_regeol)
2217 nextchr = UCHARAT(++locinput);
2220 if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n')
2223 locinput += PL_utf8skip[nextchr];
2224 if (locinput > PL_regeol)
2226 nextchr = UCHARAT(locinput);
2229 nextchr = UCHARAT(++locinput);
2234 if (do_utf8 != (UTF!=0)) {
2235 /* The target and the pattern have differing utf8ness. */
2241 /* The target is utf8, the pattern is not utf8. */
2245 if (NATIVE_TO_UNI(*(U8*)s) !=
2246 utf8_to_uvchr((U8*)l, &ulen))
2253 /* The target is not utf8, the pattern is utf8. */
2257 if (NATIVE_TO_UNI(*((U8*)l)) !=
2258 utf8_to_uvchr((U8*)s, &ulen))
2265 nextchr = UCHARAT(locinput);
2268 /* The target and the pattern have the same utf8ness. */
2269 /* Inline the first character, for speed. */
2270 if (UCHARAT(s) != nextchr)
2272 if (PL_regeol - locinput < ln)
2274 if (ln > 1 && memNE(s, locinput, ln))
2277 nextchr = UCHARAT(locinput);
2280 PL_reg_flags |= RF_tainted;
2289 U8 tmpbuf[UTF8_MAXLEN_FOLD+1];
2291 if (do_utf8 != (UTF!=0)) {
2292 /* The target and the pattern have differing utf8ness. */
2293 STRLEN ulen1, ulen2;
2297 /* The target is utf8, the pattern is not utf8. */
2302 cs = to_uni_fold(NATIVE_TO_UNI(*(U8*)s),
2304 cl = utf8_to_uvchr((U8*)l, &ulen2);
2307 cl = to_uni_fold(cl, (U8*)l, &ulen2);
2308 if (ulen1 != ulen2 || cs != cl)
2316 /* The target is not utf8, the pattern is utf8. */
2321 cs = utf8_to_uvchr((U8*)s, &ulen1);
2323 cl = to_uni_fold(NATIVE_TO_UNI(*(U8*)l),
2327 cs = to_uni_fold(cs, (U8*)s, &ulen1);
2328 if (ulen1 != ulen2 || cs != cl)
2336 nextchr = UCHARAT(locinput);
2340 if (do_utf8 && UTF) {
2341 /* Both the target and the pattern are utf8. */
2347 if (UTF8SKIP(s) != UTF8SKIP(l) ||
2348 memNE(s, (char*)l, UTF8SKIP(s))) {
2349 to_utf8_fold((U8*)l, tmpbuf, &ulen);
2350 if (UTF8SKIP(s) != ulen ||
2351 memNE(s, (char*)tmpbuf, ulen))
2358 nextchr = UCHARAT(locinput);
2363 /* Neither the target and the pattern are utf8. */
2365 /* Inline the first character, for speed. */
2366 if (UCHARAT(s) != nextchr &&
2367 UCHARAT(s) != ((OP(scan) == EXACTF)
2368 ? PL_fold : PL_fold_locale)[nextchr])
2370 if (PL_regeol - locinput < ln)
2372 if (ln > 1 && (OP(scan) == EXACTF
2373 ? ibcmp(s, locinput, ln)
2374 : ibcmp_locale(s, locinput, ln)))
2377 nextchr = UCHARAT(locinput);
2381 if (!reginclass(scan, (U8*)locinput, do_utf8))
2383 if (locinput >= PL_regeol)
2385 locinput += PL_utf8skip[nextchr];
2386 nextchr = UCHARAT(locinput);
2390 nextchr = UCHARAT(locinput);
2391 if (!reginclass(scan, (U8*)locinput, do_utf8))
2393 if (!nextchr && locinput >= PL_regeol)
2395 nextchr = UCHARAT(++locinput);
2399 PL_reg_flags |= RF_tainted;
2405 LOAD_UTF8_CHARCLASS(alnum,"a");
2406 if (!(OP(scan) == ALNUM
2407 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2408 : isALNUM_LC_utf8((U8*)locinput)))
2412 locinput += PL_utf8skip[nextchr];
2413 nextchr = UCHARAT(locinput);
2416 if (!(OP(scan) == ALNUM
2417 ? isALNUM(nextchr) : isALNUM_LC(nextchr)))
2419 nextchr = UCHARAT(++locinput);
2422 PL_reg_flags |= RF_tainted;
2425 if (!nextchr && locinput >= PL_regeol)
2428 LOAD_UTF8_CHARCLASS(alnum,"a");
2429 if (OP(scan) == NALNUM
2430 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2431 : isALNUM_LC_utf8((U8*)locinput))
2435 locinput += PL_utf8skip[nextchr];
2436 nextchr = UCHARAT(locinput);
2439 if (OP(scan) == NALNUM
2440 ? isALNUM(nextchr) : isALNUM_LC(nextchr))
2442 nextchr = UCHARAT(++locinput);
2446 PL_reg_flags |= RF_tainted;
2450 /* was last char in word? */
2452 if (locinput == PL_bostr)
2455 U8 *r = reghop((U8*)locinput, -1);
2457 ln = utf8n_to_uvchr(r, s - (char*)r, 0, 0);
2459 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2460 ln = isALNUM_uni(ln);
2461 LOAD_UTF8_CHARCLASS(alnum,"a");
2462 n = swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8);
2465 ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(ln));
2466 n = isALNUM_LC_utf8((U8*)locinput);
2470 ln = (locinput != PL_bostr) ?
2471 UCHARAT(locinput - 1) : '\n';
2472 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2474 n = isALNUM(nextchr);
2477 ln = isALNUM_LC(ln);
2478 n = isALNUM_LC(nextchr);
2481 if (((!ln) == (!n)) == (OP(scan) == BOUND ||
2482 OP(scan) == BOUNDL))
2486 PL_reg_flags |= RF_tainted;
2492 if (UTF8_IS_CONTINUED(nextchr)) {
2493 LOAD_UTF8_CHARCLASS(space," ");
2494 if (!(OP(scan) == SPACE
2495 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2496 : isSPACE_LC_utf8((U8*)locinput)))
2500 locinput += PL_utf8skip[nextchr];
2501 nextchr = UCHARAT(locinput);
2504 if (!(OP(scan) == SPACE
2505 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2507 nextchr = UCHARAT(++locinput);
2510 if (!(OP(scan) == SPACE
2511 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2513 nextchr = UCHARAT(++locinput);
2517 PL_reg_flags |= RF_tainted;
2520 if (!nextchr && locinput >= PL_regeol)
2523 LOAD_UTF8_CHARCLASS(space," ");
2524 if (OP(scan) == NSPACE
2525 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2526 : isSPACE_LC_utf8((U8*)locinput))
2530 locinput += PL_utf8skip[nextchr];
2531 nextchr = UCHARAT(locinput);
2534 if (OP(scan) == NSPACE
2535 ? isSPACE(nextchr) : isSPACE_LC(nextchr))
2537 nextchr = UCHARAT(++locinput);
2540 PL_reg_flags |= RF_tainted;
2546 LOAD_UTF8_CHARCLASS(digit,"0");
2547 if (!(OP(scan) == DIGIT
2548 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2549 : isDIGIT_LC_utf8((U8*)locinput)))
2553 locinput += PL_utf8skip[nextchr];
2554 nextchr = UCHARAT(locinput);
2557 if (!(OP(scan) == DIGIT
2558 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr)))
2560 nextchr = UCHARAT(++locinput);
2563 PL_reg_flags |= RF_tainted;
2566 if (!nextchr && locinput >= PL_regeol)
2569 LOAD_UTF8_CHARCLASS(digit,"0");
2570 if (OP(scan) == NDIGIT
2571 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2572 : isDIGIT_LC_utf8((U8*)locinput))
2576 locinput += PL_utf8skip[nextchr];
2577 nextchr = UCHARAT(locinput);
2580 if (OP(scan) == NDIGIT
2581 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr))
2583 nextchr = UCHARAT(++locinput);
2586 LOAD_UTF8_CHARCLASS(mark,"~");
2587 if (locinput >= PL_regeol ||
2588 swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2590 locinput += PL_utf8skip[nextchr];
2591 while (locinput < PL_regeol &&
2592 swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2593 locinput += UTF8SKIP(locinput);
2594 if (locinput > PL_regeol)
2596 nextchr = UCHARAT(locinput);
2599 PL_reg_flags |= RF_tainted;
2603 n = ARG(scan); /* which paren pair */
2604 ln = PL_regstartp[n];
2605 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2606 if (*PL_reglastparen < n || ln == -1)
2607 sayNO; /* Do not match unless seen CLOSEn. */
2608 if (ln == PL_regendp[n])
2612 if (do_utf8 && OP(scan) != REF) { /* REF can do byte comparison */
2614 char *e = PL_bostr + PL_regendp[n];
2616 * Note that we can't do the "other character" lookup trick as
2617 * in the 8-bit case (no pun intended) because in Unicode we
2618 * have to map both upper and title case to lower case.
2620 if (OP(scan) == REFF) {
2621 STRLEN ulen1, ulen2;
2622 U8 tmpbuf1[UTF8_MAXLEN_UCLC+1];
2623 U8 tmpbuf2[UTF8_MAXLEN_UCLC+1];
2627 toLOWER_utf8((U8*)s, tmpbuf1, &ulen1);
2628 toLOWER_utf8((U8*)l, tmpbuf2, &ulen2);
2629 if (ulen1 != ulen2 || memNE((char *)tmpbuf1, (char *)tmpbuf2, ulen1))
2636 nextchr = UCHARAT(locinput);
2640 /* Inline the first character, for speed. */
2641 if (UCHARAT(s) != nextchr &&
2643 (UCHARAT(s) != ((OP(scan) == REFF
2644 ? PL_fold : PL_fold_locale)[nextchr]))))
2646 ln = PL_regendp[n] - ln;
2647 if (locinput + ln > PL_regeol)
2649 if (ln > 1 && (OP(scan) == REF
2650 ? memNE(s, locinput, ln)
2652 ? ibcmp(s, locinput, ln)
2653 : ibcmp_locale(s, locinput, ln))))
2656 nextchr = UCHARAT(locinput);
2667 OP_4tree *oop = PL_op;
2668 COP *ocurcop = PL_curcop;
2669 SV **ocurpad = PL_curpad;
2673 PL_op = (OP_4tree*)PL_regdata->data[n];
2674 DEBUG_r( PerlIO_printf(Perl_debug_log, " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) );
2675 PL_curpad = AvARRAY((AV*)PL_regdata->data[n + 2]);
2676 PL_regendp[0] = PL_reg_magic->mg_len = locinput - PL_bostr;
2680 CALLRUNOPS(aTHX); /* Scalar context. */
2683 ret = Nullsv; /* protect against empty (?{}) blocks. */
2691 PL_curpad = ocurpad;
2692 PL_curcop = ocurcop;
2694 if (logical == 2) { /* Postponed subexpression. */
2696 MAGIC *mg = Null(MAGIC*);
2698 CHECKPOINT cp, lastcp;
2700 if(SvROK(ret) || SvRMAGICAL(ret)) {
2701 SV *sv = SvROK(ret) ? SvRV(ret) : ret;
2704 mg = mg_find(sv, PERL_MAGIC_qr);
2707 re = (regexp *)mg->mg_obj;
2708 (void)ReREFCNT_inc(re);
2712 char *t = SvPV(ret, len);
2714 char *oprecomp = PL_regprecomp;
2715 I32 osize = PL_regsize;
2716 I32 onpar = PL_regnpar;
2719 re = CALLREGCOMP(aTHX_ t, t + len, &pm);
2721 & (SVs_TEMP | SVs_PADTMP | SVf_READONLY)))
2722 sv_magic(ret,(SV*)ReREFCNT_inc(re),
2724 PL_regprecomp = oprecomp;
2729 PerlIO_printf(Perl_debug_log,
2730 "Entering embedded `%s%.60s%s%s'\n",
2734 (strlen(re->precomp) > 60 ? "..." : ""))
2737 state.prev = PL_reg_call_cc;
2738 state.cc = PL_regcc;
2739 state.re = PL_reg_re;
2743 cp = regcppush(0); /* Save *all* the positions. */
2746 state.ss = PL_savestack_ix;
2747 *PL_reglastparen = 0;
2748 *PL_reglastcloseparen = 0;
2749 PL_reg_call_cc = &state;
2750 PL_reginput = locinput;
2752 /* XXXX This is too dramatic a measure... */
2755 if (regmatch(re->program + 1)) {
2756 /* Even though we succeeded, we need to restore
2757 global variables, since we may be wrapped inside
2758 SUSPEND, thus the match may be not finished yet. */
2760 /* XXXX Do this only if SUSPENDed? */
2761 PL_reg_call_cc = state.prev;
2762 PL_regcc = state.cc;
2763 PL_reg_re = state.re;
2764 cache_re(PL_reg_re);
2766 /* XXXX This is too dramatic a measure... */
2769 /* These are needed even if not SUSPEND. */
2775 REGCP_UNWIND(lastcp);
2777 PL_reg_call_cc = state.prev;
2778 PL_regcc = state.cc;
2779 PL_reg_re = state.re;
2780 cache_re(PL_reg_re);
2782 /* XXXX This is too dramatic a measure... */
2792 sv_setsv(save_scalar(PL_replgv), ret);
2796 n = ARG(scan); /* which paren pair */
2797 PL_reg_start_tmp[n] = locinput;
2802 n = ARG(scan); /* which paren pair */
2803 PL_regstartp[n] = PL_reg_start_tmp[n] - PL_bostr;
2804 PL_regendp[n] = locinput - PL_bostr;
2805 if (n > *PL_reglastparen)
2806 *PL_reglastparen = n;
2807 *PL_reglastcloseparen = n;
2810 n = ARG(scan); /* which paren pair */
2811 sw = (*PL_reglastparen >= n && PL_regendp[n] != -1);
2814 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2816 next = NEXTOPER(NEXTOPER(scan));
2818 next = scan + ARG(scan);
2819 if (OP(next) == IFTHEN) /* Fake one. */
2820 next = NEXTOPER(NEXTOPER(next));
2824 logical = scan->flags;
2826 /*******************************************************************
2827 PL_regcc contains infoblock about the innermost (...)* loop, and
2828 a pointer to the next outer infoblock.
2830 Here is how Y(A)*Z is processed (if it is compiled into CURLYX/WHILEM):
2832 1) After matching X, regnode for CURLYX is processed;
2834 2) This regnode creates infoblock on the stack, and calls
2835 regmatch() recursively with the starting point at WHILEM node;
2837 3) Each hit of WHILEM node tries to match A and Z (in the order
2838 depending on the current iteration, min/max of {min,max} and
2839 greediness). The information about where are nodes for "A"
2840 and "Z" is read from the infoblock, as is info on how many times "A"
2841 was already matched, and greediness.
2843 4) After A matches, the same WHILEM node is hit again.
2845 5) Each time WHILEM is hit, PL_regcc is the infoblock created by CURLYX
2846 of the same pair. Thus when WHILEM tries to match Z, it temporarily
2847 resets PL_regcc, since this Y(A)*Z can be a part of some other loop:
2848 as in (Y(A)*Z)*. If Z matches, the automaton will hit the WHILEM node
2849 of the external loop.
2851 Currently present infoblocks form a tree with a stem formed by PL_curcc
2852 and whatever it mentions via ->next, and additional attached trees
2853 corresponding to temporarily unset infoblocks as in "5" above.
2855 In the following picture infoblocks for outer loop of
2856 (Y(A)*?Z)*?T are denoted O, for inner I. NULL starting block
2857 is denoted by x. The matched string is YAAZYAZT. Temporarily postponed
2858 infoblocks are drawn below the "reset" infoblock.
2860 In fact in the picture below we do not show failed matches for Z and T
2861 by WHILEM blocks. [We illustrate minimal matches, since for them it is
2862 more obvious *why* one needs to *temporary* unset infoblocks.]
2864 Matched REx position InfoBlocks Comment
2868 Y A)*?Z)*?T x <- O <- I
2869 YA )*?Z)*?T x <- O <- I
2870 YA A)*?Z)*?T x <- O <- I
2871 YAA )*?Z)*?T x <- O <- I
2872 YAA Z)*?T x <- O # Temporary unset I
2875 YAAZ Y(A)*?Z)*?T x <- O
2878 YAAZY (A)*?Z)*?T x <- O
2881 YAAZY A)*?Z)*?T x <- O <- I
2884 YAAZYA )*?Z)*?T x <- O <- I
2887 YAAZYA Z)*?T x <- O # Temporary unset I
2893 YAAZYAZ T x # Temporary unset O
2900 *******************************************************************/
2903 CHECKPOINT cp = PL_savestack_ix;
2904 /* No need to save/restore up to this paren */
2905 I32 parenfloor = scan->flags;
2907 if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */
2909 cc.oldcc = PL_regcc;
2911 /* XXXX Probably it is better to teach regpush to support
2912 parenfloor > PL_regsize... */
2913 if (parenfloor > *PL_reglastparen)
2914 parenfloor = *PL_reglastparen; /* Pessimization... */
2915 cc.parenfloor = parenfloor;
2917 cc.min = ARG1(scan);
2918 cc.max = ARG2(scan);
2919 cc.scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
2923 PL_reginput = locinput;
2924 n = regmatch(PREVOPER(next)); /* start on the WHILEM */
2926 PL_regcc = cc.oldcc;
2932 * This is really hard to understand, because after we match
2933 * what we're trying to match, we must make sure the rest of
2934 * the REx is going to match for sure, and to do that we have
2935 * to go back UP the parse tree by recursing ever deeper. And
2936 * if it fails, we have to reset our parent's current state
2937 * that we can try again after backing off.
2940 CHECKPOINT cp, lastcp;
2941 CURCUR* cc = PL_regcc;
2942 char *lastloc = cc->lastloc; /* Detection of 0-len. */
2944 n = cc->cur + 1; /* how many we know we matched */
2945 PL_reginput = locinput;
2948 PerlIO_printf(Perl_debug_log,
2949 "%*s %ld out of %ld..%ld cc=%lx\n",
2950 REPORT_CODE_OFF+PL_regindent*2, "",
2951 (long)n, (long)cc->min,
2952 (long)cc->max, (long)cc)
2955 /* If degenerate scan matches "", assume scan done. */
2957 if (locinput == cc->lastloc && n >= cc->min) {
2958 PL_regcc = cc->oldcc;
2962 PerlIO_printf(Perl_debug_log,
2963 "%*s empty match detected, try continuation...\n",
2964 REPORT_CODE_OFF+PL_regindent*2, "")
2966 if (regmatch(cc->next))
2974 /* First just match a string of min scans. */
2978 cc->lastloc = locinput;
2979 if (regmatch(cc->scan))
2982 cc->lastloc = lastloc;
2987 /* Check whether we already were at this position.
2988 Postpone detection until we know the match is not
2989 *that* much linear. */
2990 if (!PL_reg_maxiter) {
2991 PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4);
2992 PL_reg_leftiter = PL_reg_maxiter;
2994 if (PL_reg_leftiter-- == 0) {
2995 I32 size = (PL_reg_maxiter + 7)/8;
2996 if (PL_reg_poscache) {
2997 if (PL_reg_poscache_size < size) {
2998 Renew(PL_reg_poscache, size, char);
2999 PL_reg_poscache_size = size;
3001 Zero(PL_reg_poscache, size, char);
3004 PL_reg_poscache_size = size;
3005 Newz(29, PL_reg_poscache, size, char);
3008 PerlIO_printf(Perl_debug_log,
3009 "%sDetected a super-linear match, switching on caching%s...\n",
3010 PL_colors[4], PL_colors[5])
3013 if (PL_reg_leftiter < 0) {
3014 I32 o = locinput - PL_bostr, b;
3016 o = (scan->flags & 0xf) - 1 + o * (scan->flags>>4);
3019 if (PL_reg_poscache[o] & (1<<b)) {
3021 PerlIO_printf(Perl_debug_log,
3022 "%*s already tried at this position...\n",
3023 REPORT_CODE_OFF+PL_regindent*2, "")
3027 PL_reg_poscache[o] |= (1<<b);
3031 /* Prefer next over scan for minimal matching. */
3034 PL_regcc = cc->oldcc;
3037 cp = regcppush(cc->parenfloor);
3039 if (regmatch(cc->next)) {
3041 sayYES; /* All done. */
3043 REGCP_UNWIND(lastcp);
3049 if (n >= cc->max) { /* Maximum greed exceeded? */
3050 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3051 && !(PL_reg_flags & RF_warned)) {
3052 PL_reg_flags |= RF_warned;
3053 Perl_warner(aTHX_ WARN_REGEXP, "%s limit (%d) exceeded",
3054 "Complex regular subexpression recursion",
3061 PerlIO_printf(Perl_debug_log,
3062 "%*s trying longer...\n",
3063 REPORT_CODE_OFF+PL_regindent*2, "")
3065 /* Try scanning more and see if it helps. */
3066 PL_reginput = locinput;
3068 cc->lastloc = locinput;
3069 cp = regcppush(cc->parenfloor);
3071 if (regmatch(cc->scan)) {
3075 REGCP_UNWIND(lastcp);
3078 cc->lastloc = lastloc;
3082 /* Prefer scan over next for maximal matching. */
3084 if (n < cc->max) { /* More greed allowed? */
3085 cp = regcppush(cc->parenfloor);
3087 cc->lastloc = locinput;
3089 if (regmatch(cc->scan)) {
3093 REGCP_UNWIND(lastcp);
3094 regcppop(); /* Restore some previous $<digit>s? */
3095 PL_reginput = locinput;
3097 PerlIO_printf(Perl_debug_log,
3098 "%*s failed, try continuation...\n",
3099 REPORT_CODE_OFF+PL_regindent*2, "")
3102 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3103 && !(PL_reg_flags & RF_warned)) {
3104 PL_reg_flags |= RF_warned;
3105 Perl_warner(aTHX_ WARN_REGEXP, "%s limit (%d) exceeded",
3106 "Complex regular subexpression recursion",
3110 /* Failed deeper matches of scan, so see if this one works. */
3111 PL_regcc = cc->oldcc;
3114 if (regmatch(cc->next))
3120 cc->lastloc = lastloc;
3125 next = scan + ARG(scan);
3128 inner = NEXTOPER(NEXTOPER(scan));
3131 inner = NEXTOPER(scan);
3135 if (OP(next) != c1) /* No choice. */
3136 next = inner; /* Avoid recursion. */
3138 I32 lastparen = *PL_reglastparen;
3140 re_unwind_branch_t *uw;
3142 /* Put unwinding data on stack */
3143 unwind1 = SSNEWt(1,re_unwind_branch_t);
3144 uw = SSPTRt(unwind1,re_unwind_branch_t);
3147 uw->type = ((c1 == BRANCH)
3149 : RE_UNWIND_BRANCHJ);
3150 uw->lastparen = lastparen;
3152 uw->locinput = locinput;
3153 uw->nextchr = nextchr;
3155 uw->regindent = ++PL_regindent;
3158 REGCP_SET(uw->lastcp);
3160 /* Now go into the first branch */
3173 /* We suppose that the next guy does not need
3174 backtracking: in particular, it is of constant length,
3175 and has no parenths to influence future backrefs. */
3176 ln = ARG1(scan); /* min to match */
3177 n = ARG2(scan); /* max to match */
3178 paren = scan->flags;
3180 if (paren > PL_regsize)
3182 if (paren > *PL_reglastparen)
3183 *PL_reglastparen = paren;
3185 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3187 scan += NEXT_OFF(scan); /* Skip former OPEN. */
3188 PL_reginput = locinput;
3191 if (ln && regrepeat_hard(scan, ln, &l) < ln)
3193 /* if we matched something zero-length we don't need to
3194 backtrack - capturing parens are already defined, so
3195 the caveat in the maximal case doesn't apply
3197 XXXX if ln == 0, we can redo this check first time
3198 through the following loop
3201 n = ln; /* don't backtrack */
3202 locinput = PL_reginput;
3203 if (HAS_TEXT(next) || JUMPABLE(next)) {
3204 regnode *text_node = next;
3206 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3208 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3210 if (PL_regkind[(U8)OP(text_node)] == REF) {
3212 n = ARG(text_node); /* which paren pair */
3213 ln = PL_regstartp[n];
3214 /* assume yes if we haven't seen CLOSEn */
3216 *PL_reglastparen < n ||
3223 c1 = *(PL_bostr + ln);
3225 else { c1 = (U8)*STRING(text_node); }
3226 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3228 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3229 c2 = PL_fold_locale[c1];
3238 /* This may be improved if l == 0. */
3239 while (n >= ln || (n == REG_INFTY && ln > 0 && l)) { /* ln overflow ? */
3240 /* If it could work, try it. */
3242 UCHARAT(PL_reginput) == c1 ||
3243 UCHARAT(PL_reginput) == c2)
3247 PL_regstartp[paren] =
3248 HOPc(PL_reginput, -l) - PL_bostr;
3249 PL_regendp[paren] = PL_reginput - PL_bostr;
3252 PL_regendp[paren] = -1;
3256 REGCP_UNWIND(lastcp);
3258 /* Couldn't or didn't -- move forward. */
3259 PL_reginput = locinput;
3260 if (regrepeat_hard(scan, 1, &l)) {
3262 locinput = PL_reginput;
3269 n = regrepeat_hard(scan, n, &l);
3270 /* if we matched something zero-length we don't need to
3271 backtrack, unless the minimum count is zero and we
3272 are capturing the result - in that case the capture
3273 being defined or not may affect later execution
3275 if (n != 0 && l == 0 && !(paren && ln == 0))
3276 ln = n; /* don't backtrack */
3277 locinput = PL_reginput;
3279 PerlIO_printf(Perl_debug_log,
3280 "%*s matched %"IVdf" times, len=%"IVdf"...\n",
3281 (int)(REPORT_CODE_OFF+PL_regindent*2), "",
3285 if (HAS_TEXT(next) || JUMPABLE(next)) {
3286 regnode *text_node = next;
3288 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3290 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3292 if (PL_regkind[(U8)OP(text_node)] == REF) {
3294 n = ARG(text_node); /* which paren pair */
3295 ln = PL_regstartp[n];
3296 /* assume yes if we haven't seen CLOSEn */
3298 *PL_reglastparen < n ||
3305 c1 = *(PL_bostr + ln);
3307 else { c1 = (U8)*STRING(text_node); }
3309 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3311 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3312 c2 = PL_fold_locale[c1];
3323 /* If it could work, try it. */
3325 UCHARAT(PL_reginput) == c1 ||
3326 UCHARAT(PL_reginput) == c2)
3329 PerlIO_printf(Perl_debug_log,
3330 "%*s trying tail with n=%"IVdf"...\n",
3331 (int)(REPORT_CODE_OFF+PL_regindent*2), "", (IV)n)
3335 PL_regstartp[paren] = HOPc(PL_reginput, -l) - PL_bostr;
3336 PL_regendp[paren] = PL_reginput - PL_bostr;
3339 PL_regendp[paren] = -1;
3343 REGCP_UNWIND(lastcp);
3345 /* Couldn't or didn't -- back up. */
3347 locinput = HOPc(locinput, -l);
3348 PL_reginput = locinput;
3355 paren = scan->flags; /* Which paren to set */
3356 if (paren > PL_regsize)
3358 if (paren > *PL_reglastparen)
3359 *PL_reglastparen = paren;
3360 ln = ARG1(scan); /* min to match */
3361 n = ARG2(scan); /* max to match */
3362 scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE);
3366 ln = ARG1(scan); /* min to match */
3367 n = ARG2(scan); /* max to match */
3368 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3373 scan = NEXTOPER(scan);
3379 scan = NEXTOPER(scan);
3383 * Lookahead to avoid useless match attempts
3384 * when we know what character comes next.
3388 * Used to only do .*x and .*?x, but now it allows
3389 * for )'s, ('s and (?{ ... })'s to be in the way
3390 * of the quantifier and the EXACT-like node. -- japhy
3393 if (HAS_TEXT(next) || JUMPABLE(next)) {
3395 regnode *text_node = next;
3397 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3399 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3401 if (PL_regkind[(U8)OP(text_node)] == REF) {
3403 n = ARG(text_node); /* which paren pair */
3404 ln = PL_regstartp[n];
3405 /* assume yes if we haven't seen CLOSEn */
3407 *PL_reglastparen < n ||
3412 goto assume_ok_easy;
3414 s = (U8*)PL_bostr + ln;
3416 else { s = (U8*)STRING(text_node); }
3420 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3422 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3423 c2 = PL_fold_locale[c1];
3426 if (OP(text_node) == EXACTF || OP(text_node) == REFF) {
3427 STRLEN ulen1, ulen2;
3428 U8 tmpbuf1[UTF8_MAXLEN_UCLC+1];
3429 U8 tmpbuf2[UTF8_MAXLEN_UCLC+1];
3431 to_utf8_lower((U8*)s, tmpbuf1, &ulen1);
3432 to_utf8_upper((U8*)s, tmpbuf2, &ulen2);
3434 c1 = utf8_to_uvuni(tmpbuf1, 0);
3435 c2 = utf8_to_uvuni(tmpbuf2, 0);
3438 c2 = c1 = utf8_to_uvchr(s, NULL);
3446 PL_reginput = locinput;
3450 if (ln && regrepeat(scan, ln) < ln)
3452 locinput = PL_reginput;
3455 char *e; /* Should not check after this */
3456 char *old = locinput;
3458 if (n == REG_INFTY) {
3461 while (UTF8_IS_CONTINUATION(*(U8*)e))
3467 m >0 && e + UTF8SKIP(e) <= PL_regeol; m--)
3471 e = locinput + n - ln;
3477 /* Find place 'next' could work */
3480 while (locinput <= e &&
3481 UCHARAT(locinput) != c1)
3484 while (locinput <= e
3485 && UCHARAT(locinput) != c1
3486 && UCHARAT(locinput) != c2)
3489 count = locinput - old;
3496 utf8_to_uvchr((U8*)locinput, &len) != c1;
3501 for (count = 0; locinput <= e; count++) {
3502 UV c = utf8_to_uvchr((U8*)locinput, &len);
3503 if (c == c1 || c == c2)
3511 /* PL_reginput == old now */
3512 if (locinput != old) {
3513 ln = 1; /* Did some */
3514 if (regrepeat(scan, count) < count)
3517 /* PL_reginput == locinput now */
3518 TRYPAREN(paren, ln, locinput);
3519 PL_reginput = locinput; /* Could be reset... */
3520 REGCP_UNWIND(lastcp);
3521 /* Couldn't or didn't -- move forward. */
3524 locinput += UTF8SKIP(locinput);
3530 while (n >= ln || (n == REG_INFTY && ln > 0)) { /* ln overflow ? */
3534 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3536 c = UCHARAT(PL_reginput);
3537 /* If it could work, try it. */
3538 if (c == c1 || c == c2)
3540 TRYPAREN(paren, n, PL_reginput);
3541 REGCP_UNWIND(lastcp);
3544 /* If it could work, try it. */
3545 else if (c1 == -1000)
3547 TRYPAREN(paren, n, PL_reginput);
3548 REGCP_UNWIND(lastcp);
3550 /* Couldn't or didn't -- move forward. */
3551 PL_reginput = locinput;
3552 if (regrepeat(scan, 1)) {
3554 locinput = PL_reginput;
3562 n = regrepeat(scan, n);
3563 locinput = PL_reginput;
3564 if (ln < n && PL_regkind[(U8)OP(next)] == EOL &&
3565 (!PL_multiline || OP(next) == SEOL || OP(next) == EOS)) {
3566 ln = n; /* why back off? */
3567 /* ...because $ and \Z can match before *and* after
3568 newline at the end. Consider "\n\n" =~ /\n+\Z\n/.
3569 We should back off by one in this case. */
3570 if (UCHARAT(PL_reginput - 1) == '\n' && OP(next) != EOS)
3579 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3581 c = UCHARAT(PL_reginput);
3583 /* If it could work, try it. */
3584 if (c1 == -1000 || c == c1 || c == c2)
3586 TRYPAREN(paren, n, PL_reginput);
3587 REGCP_UNWIND(lastcp);
3589 /* Couldn't or didn't -- back up. */
3591 PL_reginput = locinput = HOPc(locinput, -1);
3599 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3601 c = UCHARAT(PL_reginput);
3603 /* If it could work, try it. */
3604 if (c1 == -1000 || c == c1 || c == c2)
3606 TRYPAREN(paren, n, PL_reginput);
3607 REGCP_UNWIND(lastcp);
3609 /* Couldn't or didn't -- back up. */
3611 PL_reginput = locinput = HOPc(locinput, -1);
3618 if (PL_reg_call_cc) {
3619 re_cc_state *cur_call_cc = PL_reg_call_cc;
3620 CURCUR *cctmp = PL_regcc;
3621 regexp *re = PL_reg_re;
3622 CHECKPOINT cp, lastcp;
3624 cp = regcppush(0); /* Save *all* the positions. */
3626 regcp_set_to(PL_reg_call_cc->ss); /* Restore parens of
3628 PL_reginput = locinput; /* Make position available to
3630 cache_re(PL_reg_call_cc->re);
3631 PL_regcc = PL_reg_call_cc->cc;
3632 PL_reg_call_cc = PL_reg_call_cc->prev;
3633 if (regmatch(cur_call_cc->node)) {
3634 PL_reg_call_cc = cur_call_cc;
3638 REGCP_UNWIND(lastcp);
3640 PL_reg_call_cc = cur_call_cc;
3646 PerlIO_printf(Perl_debug_log,
3647 "%*s continuation failed...\n",
3648 REPORT_CODE_OFF+PL_regindent*2, "")
3652 if (locinput < PL_regtill) {
3653 DEBUG_r(PerlIO_printf(Perl_debug_log,
3654 "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n",
3656 (long)(locinput - PL_reg_starttry),
3657 (long)(PL_regtill - PL_reg_starttry),
3659 sayNO_FINAL; /* Cannot match: too short. */
3661 PL_reginput = locinput; /* put where regtry can find it */
3662 sayYES_FINAL; /* Success! */
3664 PL_reginput = locinput; /* put where regtry can find it */
3665 sayYES_LOUD; /* Success! */
3668 PL_reginput = locinput;
3673 s = HOPBACKc(locinput, scan->flags);
3679 PL_reginput = locinput;
3684 s = HOPBACKc(locinput, scan->flags);
3690 PL_reginput = locinput;
3693 inner = NEXTOPER(NEXTOPER(scan));
3694 if (regmatch(inner) != n) {
3709 if (OP(scan) == SUSPEND) {
3710 locinput = PL_reginput;
3711 nextchr = UCHARAT(locinput);
3716 next = scan + ARG(scan);
3721 PerlIO_printf(Perl_error_log, "%"UVxf" %d\n",
3722 PTR2UV(scan), OP(scan));
3723 Perl_croak(aTHX_ "regexp memory corruption");
3730 * We get here only if there's trouble -- normally "case END" is
3731 * the terminating point.
3733 Perl_croak(aTHX_ "corrupted regexp pointers");
3739 PerlIO_printf(Perl_debug_log,
3740 "%*s %scould match...%s\n",
3741 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4],PL_colors[5])
3745 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n",
3746 PL_colors[4],PL_colors[5]));
3752 #if 0 /* Breaks $^R */
3760 PerlIO_printf(Perl_debug_log,
3761 "%*s %sfailed...%s\n",
3762 REPORT_CODE_OFF+PL_regindent*2, "",PL_colors[4],PL_colors[5])
3768 re_unwind_t *uw = SSPTRt(unwind,re_unwind_t);
3771 case RE_UNWIND_BRANCH:
3772 case RE_UNWIND_BRANCHJ:
3774 re_unwind_branch_t *uwb = &(uw->branch);
3775 I32 lastparen = uwb->lastparen;
3777 REGCP_UNWIND(uwb->lastcp);
3778 for (n = *PL_reglastparen; n > lastparen; n--)
3780 *PL_reglastparen = n;
3781 scan = next = uwb->next;
3783 OP(scan) != (uwb->type == RE_UNWIND_BRANCH
3784 ? BRANCH : BRANCHJ) ) { /* Failure */
3791 /* Have more choice yet. Reuse the same uwb. */
3793 if ((n = (uwb->type == RE_UNWIND_BRANCH
3794 ? NEXT_OFF(next) : ARG(next))))
3797 next = NULL; /* XXXX Needn't unwinding in this case... */
3799 next = NEXTOPER(scan);
3800 if (uwb->type == RE_UNWIND_BRANCHJ)
3801 next = NEXTOPER(next);
3802 locinput = uwb->locinput;
3803 nextchr = uwb->nextchr;
3805 PL_regindent = uwb->regindent;
3812 Perl_croak(aTHX_ "regexp unwind memory corruption");
3823 - regrepeat - repeatedly match something simple, report how many
3826 * [This routine now assumes that it will only match on things of length 1.
3827 * That was true before, but now we assume scan - reginput is the count,
3828 * rather than incrementing count on every character. [Er, except utf8.]]
3831 S_regrepeat(pTHX_ regnode *p, I32 max)
3833 register char *scan;
3835 register char *loceol = PL_regeol;
3836 register I32 hardcount = 0;
3837 register bool do_utf8 = PL_reg_match_utf8;
3840 if (max != REG_INFTY && max < loceol - scan)
3841 loceol = scan + max;
3846 while (scan < loceol && hardcount < max && *scan != '\n') {
3847 scan += UTF8SKIP(scan);
3851 while (scan < loceol && *scan != '\n')
3861 case EXACT: /* length of string is 1 */
3863 while (scan < loceol && UCHARAT(scan) == c)
3866 case EXACTF: /* length of string is 1 */
3868 while (scan < loceol &&
3869 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold[c]))
3872 case EXACTFL: /* length of string is 1 */
3873 PL_reg_flags |= RF_tainted;
3875 while (scan < loceol &&
3876 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold_locale[c]))
3882 while (hardcount < max && scan < loceol &&
3883 reginclass(p, (U8*)scan, do_utf8)) {
3884 scan += UTF8SKIP(scan);
3888 while (scan < loceol && reginclass(p, (U8*)scan, do_utf8))
3895 LOAD_UTF8_CHARCLASS(alnum,"a");
3896 while (hardcount < max && scan < loceol &&
3897 swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
3898 scan += UTF8SKIP(scan);
3902 while (scan < loceol && isALNUM(*scan))
3907 PL_reg_flags |= RF_tainted;
3910 while (hardcount < max && scan < loceol &&
3911 isALNUM_LC_utf8((U8*)scan)) {
3912 scan += UTF8SKIP(scan);
3916 while (scan < loceol && isALNUM_LC(*scan))
3923 LOAD_UTF8_CHARCLASS(alnum,"a");
3924 while (hardcount < max && scan < loceol &&
3925 !swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
3926 scan += UTF8SKIP(scan);
3930 while (scan < loceol && !isALNUM(*scan))
3935 PL_reg_flags |= RF_tainted;
3938 while (hardcount < max && scan < loceol &&
3939 !isALNUM_LC_utf8((U8*)scan)) {
3940 scan += UTF8SKIP(scan);
3944 while (scan < loceol && !isALNUM_LC(*scan))
3951 LOAD_UTF8_CHARCLASS(space," ");
3952 while (hardcount < max && scan < loceol &&
3954 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
3955 scan += UTF8SKIP(scan);
3959 while (scan < loceol && isSPACE(*scan))
3964 PL_reg_flags |= RF_tainted;
3967 while (hardcount < max && scan < loceol &&
3968 (*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
3969 scan += UTF8SKIP(scan);
3973 while (scan < loceol && isSPACE_LC(*scan))
3980 LOAD_UTF8_CHARCLASS(space," ");
3981 while (hardcount < max && scan < loceol &&
3983 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
3984 scan += UTF8SKIP(scan);
3988 while (scan < loceol && !isSPACE(*scan))
3993 PL_reg_flags |= RF_tainted;
3996 while (hardcount < max && scan < loceol &&
3997 !(*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
3998 scan += UTF8SKIP(scan);
4002 while (scan < loceol && !isSPACE_LC(*scan))
4009 LOAD_UTF8_CHARCLASS(digit,"0");
4010 while (hardcount < max && scan < loceol &&
4011 swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
4012 scan += UTF8SKIP(scan);
4016 while (scan < loceol && isDIGIT(*scan))
4023 LOAD_UTF8_CHARCLASS(digit,"0");
4024 while (hardcount < max && scan < loceol &&
4025 !swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
4026 scan += UTF8SKIP(scan);
4030 while (scan < loceol && !isDIGIT(*scan))
4034 default: /* Called on something of 0 width. */
4035 break; /* So match right here or not at all. */
4041 c = scan - PL_reginput;
4046 SV *prop = sv_newmortal();
4049 PerlIO_printf(Perl_debug_log,
4050 "%*s %s can match %"IVdf" times out of %"IVdf"...\n",
4051 REPORT_CODE_OFF+1, "", SvPVX(prop),(IV)c,(IV)max);
4058 - regrepeat_hard - repeatedly match something, report total lenth and length
4060 * The repeater is supposed to have constant length.
4064 S_regrepeat_hard(pTHX_ regnode *p, I32 max, I32 *lp)
4066 register char *scan = Nullch;
4067 register char *start;
4068 register char *loceol = PL_regeol;
4070 I32 count = 0, res = 1;
4075 start = PL_reginput;
4076 if (PL_reg_match_utf8) {
4077 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
4080 while (start < PL_reginput) {
4082 start += UTF8SKIP(start);
4093 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
4095 *lp = l = PL_reginput - start;
4096 if (max != REG_INFTY && l*max < loceol - scan)
4097 loceol = scan + l*max;
4110 - regclass_swash - prepare the utf8 swash
4114 Perl_regclass_swash(pTHX_ register regnode* node, bool doinit, SV** initsvp)
4119 if (PL_regdata && PL_regdata->count) {
4122 if (PL_regdata->what[n] == 's') {
4123 SV *rv = (SV*)PL_regdata->data[n];
4124 AV *av = (AV*)SvRV((SV*)rv);
4127 si = *av_fetch(av, 0, FALSE);
4128 a = av_fetch(av, 1, FALSE);
4132 else if (si && doinit) {
4133 sw = swash_init("utf8", "", si, 1, 0);
4134 (void)av_store(av, 1, sw);
4146 - reginclass - determine if a character falls into a character class
4150 S_reginclass(pTHX_ register regnode *n, register U8* p, register bool do_utf8)
4152 char flags = ANYOF_FLAGS(n);
4157 c = do_utf8 ? utf8_to_uvchr(p, &len) : *p;
4159 if (do_utf8 || (flags & ANYOF_UNICODE)) {
4160 if (do_utf8 && !ANYOF_RUNTIME(n)) {
4161 if (len != (STRLEN)-1 && c < 256 && ANYOF_BITMAP_TEST(n, c))
4164 if (!match && do_utf8 && (flags & ANYOF_UNICODE_ALL) && c >= 256)
4167 SV *sw = regclass_swash(n, TRUE, 0);
4170 if (swash_fetch(sw, p, do_utf8))
4172 else if (flags & ANYOF_FOLD) {
4174 U8 tmpbuf[UTF8_MAXLEN_FOLD+1];
4176 to_utf8_fold(p, tmpbuf, &ulen);
4177 if (swash_fetch(sw, tmpbuf, do_utf8))
4179 to_utf8_upper(p, tmpbuf, &ulen);
4180 if (swash_fetch(sw, tmpbuf, do_utf8))
4186 if (!match && c < 256) {
4187 if (ANYOF_BITMAP_TEST(n, c))
4189 else if (flags & ANYOF_FOLD) {
4192 if (flags & ANYOF_LOCALE) {
4193 PL_reg_flags |= RF_tainted;
4194 f = PL_fold_locale[c];
4198 if (f != c && ANYOF_BITMAP_TEST(n, f))
4202 if (!match && (flags & ANYOF_CLASS)) {
4203 PL_reg_flags |= RF_tainted;
4205 (ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) ||
4206 (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) ||
4207 (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) ||
4208 (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) ||
4209 (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) ||
4210 (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) ||
4211 (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) ||
4212 (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) ||
4213 (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) ||
4214 (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) ||
4215 (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII(c)) ||
4216 (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII(c)) ||
4217 (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) ||
4218 (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) ||
4219 (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) ||
4220 (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) ||
4221 (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) ||
4222 (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) ||
4223 (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) ||
4224 (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) ||
4225 (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) ||
4226 (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) ||
4227 (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) ||
4228 (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) ||
4229 (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) ||
4230 (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) ||
4231 (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) ||
4232 (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) ||
4233 (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK(c)) ||
4234 (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK(c))
4235 ) /* How's that for a conditional? */
4242 return (flags & ANYOF_INVERT) ? !match : match;
4246 S_reghop(pTHX_ U8 *s, I32 off)
4248 return S_reghop3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4252 S_reghop3(pTHX_ U8 *s, I32 off, U8* lim)
4255 while (off-- && s < lim) {
4256 /* XXX could check well-formedness here */
4264 if (UTF8_IS_CONTINUED(*s)) {
4265 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4268 /* XXX could check well-formedness here */
4276 S_reghopmaybe(pTHX_ U8 *s, I32 off)
4278 return S_reghopmaybe3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4282 S_reghopmaybe3(pTHX_ U8* s, I32 off, U8* lim)
4285 while (off-- && s < lim) {
4286 /* XXX could check well-formedness here */
4296 if (UTF8_IS_CONTINUED(*s)) {
4297 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4300 /* XXX could check well-formedness here */
4312 restore_pos(pTHX_ void *arg)
4314 if (PL_reg_eval_set) {
4315 if (PL_reg_oldsaved) {
4316 PL_reg_re->subbeg = PL_reg_oldsaved;
4317 PL_reg_re->sublen = PL_reg_oldsavedlen;
4318 RX_MATCH_COPIED_on(PL_reg_re);
4320 PL_reg_magic->mg_len = PL_reg_oldpos;
4321 PL_reg_eval_set = 0;
4322 PL_curpm = PL_reg_oldcurpm;