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, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
71 **** 2000, 2001, 2002, 2003, by Larry Wall and others
73 **** You may distribute under the terms of either the GNU General Public
74 **** License or the Artistic License, as specified in the README file.
76 * Beware that some of this code is subtly aware of the way operator
77 * precedence is structured in regular expressions. Serious changes in
78 * regular-expression syntax might require a total rethink.
81 #define PERL_IN_REGEXEC_C
86 #define RF_tainted 1 /* tainted information used? */
87 #define RF_warned 2 /* warned about big count? */
88 #define RF_evaled 4 /* Did an EVAL with setting? */
89 #define RF_utf8 8 /* String contains multibyte chars? */
91 #define UTF ((PL_reg_flags & RF_utf8) != 0)
93 #define RS_init 1 /* eval environment created */
94 #define RS_set 2 /* replsv value is set */
100 #define REGINCLASS(p,c) (ANYOF_FLAGS(p) ? reginclass(p,c,0,0) : ANYOF_BITMAP_TEST(p,*(c)))
106 #define CHR_SVLEN(sv) (do_utf8 ? sv_len_utf8(sv) : SvCUR(sv))
107 #define CHR_DIST(a,b) (PL_reg_match_utf8 ? utf8_distance(a,b) : a - b)
109 #define reghop_c(pos,off) ((char*)reghop((U8*)pos, off))
110 #define reghopmaybe_c(pos,off) ((char*)reghopmaybe((U8*)pos, off))
111 #define HOP(pos,off) (PL_reg_match_utf8 ? reghop((U8*)pos, off) : (U8*)(pos + off))
112 #define HOPMAYBE(pos,off) (PL_reg_match_utf8 ? reghopmaybe((U8*)pos, off) : (U8*)(pos + off))
113 #define HOPc(pos,off) ((char*)HOP(pos,off))
114 #define HOPMAYBEc(pos,off) ((char*)HOPMAYBE(pos,off))
116 #define HOPBACK(pos, off) ( \
117 (PL_reg_match_utf8) \
118 ? reghopmaybe((U8*)pos, -off) \
119 : (pos - off >= PL_bostr) \
123 #define HOPBACKc(pos, off) (char*)HOPBACK(pos, off)
125 #define reghop3_c(pos,off,lim) ((char*)reghop3((U8*)pos, off, (U8*)lim))
126 #define reghopmaybe3_c(pos,off,lim) ((char*)reghopmaybe3((U8*)pos, off, (U8*)lim))
127 #define HOP3(pos,off,lim) (PL_reg_match_utf8 ? reghop3((U8*)pos, off, (U8*)lim) : (U8*)(pos + off))
128 #define HOPMAYBE3(pos,off,lim) (PL_reg_match_utf8 ? reghopmaybe3((U8*)pos, off, (U8*)lim) : (U8*)(pos + off))
129 #define HOP3c(pos,off,lim) ((char*)HOP3(pos,off,lim))
130 #define HOPMAYBE3c(pos,off,lim) ((char*)HOPMAYBE3(pos,off,lim))
132 #define LOAD_UTF8_CHARCLASS(a,b) STMT_START { if (!CAT2(PL_utf8_,a)) { ENTER; save_re_context(); (void)CAT2(is_utf8_, a)((U8*)b); LEAVE; } } STMT_END
134 /* for use after a quantifier and before an EXACT-like node -- japhy */
135 #define JUMPABLE(rn) ( \
136 OP(rn) == OPEN || OP(rn) == CLOSE || OP(rn) == EVAL || \
137 OP(rn) == SUSPEND || OP(rn) == IFMATCH || \
138 OP(rn) == PLUS || OP(rn) == MINMOD || \
139 (PL_regkind[(U8)OP(rn)] == CURLY && ARG1(rn) > 0) \
142 #define HAS_TEXT(rn) ( \
143 PL_regkind[(U8)OP(rn)] == EXACT || PL_regkind[(U8)OP(rn)] == REF \
147 Search for mandatory following text node; for lookahead, the text must
148 follow but for lookbehind (rn->flags != 0) we skip to the next step.
150 #define FIND_NEXT_IMPT(rn) STMT_START { \
151 while (JUMPABLE(rn)) \
152 if (OP(rn) == SUSPEND || PL_regkind[(U8)OP(rn)] == CURLY) \
153 rn = NEXTOPER(NEXTOPER(rn)); \
154 else if (OP(rn) == PLUS) \
156 else if (OP(rn) == IFMATCH) \
157 rn = (rn->flags == 0) ? NEXTOPER(NEXTOPER(rn)) : rn + ARG(rn); \
158 else rn += NEXT_OFF(rn); \
161 static void restore_pos(pTHX_ void *arg);
164 S_regcppush(pTHX_ I32 parenfloor)
166 int retval = PL_savestack_ix;
167 #define REGCP_PAREN_ELEMS 4
168 int paren_elems_to_push = (PL_regsize - parenfloor) * REGCP_PAREN_ELEMS;
171 if (paren_elems_to_push < 0)
172 Perl_croak(aTHX_ "panic: paren_elems_to_push < 0");
174 #define REGCP_OTHER_ELEMS 6
175 SSGROW(paren_elems_to_push + REGCP_OTHER_ELEMS);
176 for (p = PL_regsize; p > parenfloor; p--) {
177 /* REGCP_PARENS_ELEMS are pushed per pairs of parentheses. */
178 SSPUSHINT(PL_regendp[p]);
179 SSPUSHINT(PL_regstartp[p]);
180 SSPUSHPTR(PL_reg_start_tmp[p]);
183 /* REGCP_OTHER_ELEMS are pushed in any case, parentheses or no. */
184 SSPUSHINT(PL_regsize);
185 SSPUSHINT(*PL_reglastparen);
186 SSPUSHINT(*PL_reglastcloseparen);
187 SSPUSHPTR(PL_reginput);
188 #define REGCP_FRAME_ELEMS 2
189 /* REGCP_FRAME_ELEMS are part of the REGCP_OTHER_ELEMS and
190 * are needed for the regexp context stack bookkeeping. */
191 SSPUSHINT(paren_elems_to_push + REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS);
192 SSPUSHINT(SAVEt_REGCONTEXT); /* Magic cookie. */
197 /* These are needed since we do not localize EVAL nodes: */
198 # define REGCP_SET(cp) DEBUG_r(PerlIO_printf(Perl_debug_log, \
199 " Setting an EVAL scope, savestack=%"IVdf"\n", \
200 (IV)PL_savestack_ix)); cp = PL_savestack_ix
202 # define REGCP_UNWIND(cp) DEBUG_r(cp != PL_savestack_ix ? \
203 PerlIO_printf(Perl_debug_log, \
204 " Clearing an EVAL scope, savestack=%"IVdf"..%"IVdf"\n", \
205 (IV)(cp), (IV)PL_savestack_ix) : 0); regcpblow(cp)
215 /* Pop REGCP_OTHER_ELEMS before the parentheses loop starts. */
217 assert(i == SAVEt_REGCONTEXT); /* Check that the magic cookie is there. */
218 i = SSPOPINT; /* Parentheses elements to pop. */
219 input = (char *) SSPOPPTR;
220 *PL_reglastcloseparen = SSPOPINT;
221 *PL_reglastparen = SSPOPINT;
222 PL_regsize = SSPOPINT;
224 /* Now restore the parentheses context. */
225 for (i -= (REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS);
226 i > 0; i -= REGCP_PAREN_ELEMS) {
227 paren = (U32)SSPOPINT;
228 PL_reg_start_tmp[paren] = (char *) SSPOPPTR;
229 PL_regstartp[paren] = SSPOPINT;
231 if (paren <= *PL_reglastparen)
232 PL_regendp[paren] = tmps;
234 PerlIO_printf(Perl_debug_log,
235 " restoring \\%"UVuf" to %"IVdf"(%"IVdf")..%"IVdf"%s\n",
236 (UV)paren, (IV)PL_regstartp[paren],
237 (IV)(PL_reg_start_tmp[paren] - PL_bostr),
238 (IV)PL_regendp[paren],
239 (paren > *PL_reglastparen ? "(no)" : ""));
243 if ((I32)(*PL_reglastparen + 1) <= PL_regnpar) {
244 PerlIO_printf(Perl_debug_log,
245 " restoring \\%"IVdf"..\\%"IVdf" to undef\n",
246 (IV)(*PL_reglastparen + 1), (IV)PL_regnpar);
250 /* It would seem that the similar code in regtry()
251 * already takes care of this, and in fact it is in
252 * a better location to since this code can #if 0-ed out
253 * but the code in regtry() is needed or otherwise tests
254 * requiring null fields (pat.t#187 and split.t#{13,14}
255 * (as of patchlevel 7877) will fail. Then again,
256 * this code seems to be necessary or otherwise
257 * building DynaLoader will fail:
258 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
260 for (paren = *PL_reglastparen + 1; (I32)paren <= PL_regnpar; paren++) {
261 if ((I32)paren > PL_regsize)
262 PL_regstartp[paren] = -1;
263 PL_regendp[paren] = -1;
270 S_regcp_set_to(pTHX_ I32 ss)
272 I32 tmp = PL_savestack_ix;
274 PL_savestack_ix = ss;
276 PL_savestack_ix = tmp;
280 typedef struct re_cc_state
284 struct re_cc_state *prev;
289 #define regcpblow(cp) LEAVE_SCOPE(cp) /* Ignores regcppush()ed data. */
291 #define TRYPAREN(paren, n, input) { \
294 PL_regstartp[paren] = HOPc(input, -1) - PL_bostr; \
295 PL_regendp[paren] = input - PL_bostr; \
298 PL_regendp[paren] = -1; \
300 if (regmatch(next)) \
303 PL_regendp[paren] = -1; \
308 * pregexec and friends
312 - pregexec - match a regexp against a string
315 Perl_pregexec(pTHX_ register regexp *prog, char *stringarg, register char *strend,
316 char *strbeg, I32 minend, SV *screamer, U32 nosave)
317 /* strend: pointer to null at end of string */
318 /* strbeg: real beginning of string */
319 /* minend: end of match must be >=minend after stringarg. */
320 /* nosave: For optimizations. */
323 regexec_flags(prog, stringarg, strend, strbeg, minend, screamer, NULL,
324 nosave ? 0 : REXEC_COPY_STR);
328 S_cache_re(pTHX_ regexp *prog)
330 PL_regprecomp = prog->precomp; /* Needed for FAIL. */
332 PL_regprogram = prog->program;
334 PL_regnpar = prog->nparens;
335 PL_regdata = prog->data;
340 * Need to implement the following flags for reg_anch:
342 * USE_INTUIT_NOML - Useful to call re_intuit_start() first
344 * INTUIT_AUTORITATIVE_NOML - Can trust a positive answer
345 * INTUIT_AUTORITATIVE_ML
346 * INTUIT_ONCE_NOML - Intuit can match in one location only.
349 * Another flag for this function: SECOND_TIME (so that float substrs
350 * with giant delta may be not rechecked).
353 /* Assumptions: if ANCH_GPOS, then strpos is anchored. XXXX Check GPOS logic */
355 /* If SCREAM, then SvPVX(sv) should be compatible with strpos and strend.
356 Otherwise, only SvCUR(sv) is used to get strbeg. */
358 /* XXXX We assume that strpos is strbeg unless sv. */
360 /* XXXX Some places assume that there is a fixed substring.
361 An update may be needed if optimizer marks as "INTUITable"
362 RExen without fixed substrings. Similarly, it is assumed that
363 lengths of all the strings are no more than minlen, thus they
364 cannot come from lookahead.
365 (Or minlen should take into account lookahead.) */
367 /* A failure to find a constant substring means that there is no need to make
368 an expensive call to REx engine, thus we celebrate a failure. Similarly,
369 finding a substring too deep into the string means that less calls to
370 regtry() should be needed.
372 REx compiler's optimizer found 4 possible hints:
373 a) Anchored substring;
375 c) Whether we are anchored (beginning-of-line or \G);
376 d) First node (of those at offset 0) which may distingush positions;
377 We use a)b)d) and multiline-part of c), and try to find a position in the
378 string which does not contradict any of them.
381 /* Most of decisions we do here should have been done at compile time.
382 The nodes of the REx which we used for the search should have been
383 deleted from the finite automaton. */
386 Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos,
387 char *strend, U32 flags, re_scream_pos_data *data)
389 register I32 start_shift = 0;
390 /* Should be nonnegative! */
391 register I32 end_shift = 0;
396 int do_utf8 = sv ? SvUTF8(sv) : 0; /* if no sv we have to assume bytes */
398 register char *other_last = Nullch; /* other substr checked before this */
399 char *check_at = Nullch; /* check substr found at this pos */
401 char *i_strpos = strpos;
402 SV *dsv = PERL_DEBUG_PAD_ZERO(0);
404 RX_MATCH_UTF8_set(prog,do_utf8);
406 if (prog->reganch & ROPT_UTF8) {
407 DEBUG_r(PerlIO_printf(Perl_debug_log,
408 "UTF-8 regex...\n"));
409 PL_reg_flags |= RF_utf8;
413 char *s = PL_reg_match_utf8 ?
414 sv_uni_display(dsv, sv, 60, UNI_DISPLAY_REGEX) :
416 int len = PL_reg_match_utf8 ?
417 strlen(s) : strend - strpos;
420 if (PL_reg_match_utf8)
421 DEBUG_r(PerlIO_printf(Perl_debug_log,
422 "UTF-8 target...\n"));
423 PerlIO_printf(Perl_debug_log,
424 "%sGuessing start of match, REx%s `%s%.60s%s%s' against `%s%.*s%s%s'...\n",
425 PL_colors[4],PL_colors[5],PL_colors[0],
428 (strlen(prog->precomp) > 60 ? "..." : ""),
430 (int)(len > 60 ? 60 : len),
432 (len > 60 ? "..." : "")
436 /* CHR_DIST() would be more correct here but it makes things slow. */
437 if (prog->minlen > strend - strpos) {
438 DEBUG_r(PerlIO_printf(Perl_debug_log,
439 "String too short... [re_intuit_start]\n"));
442 strbeg = (sv && SvPOK(sv)) ? strend - SvCUR(sv) : strpos;
445 if (!prog->check_utf8 && prog->check_substr)
446 to_utf8_substr(prog);
447 check = prog->check_utf8;
449 if (!prog->check_substr && prog->check_utf8)
450 to_byte_substr(prog);
451 check = prog->check_substr;
453 if (check == &PL_sv_undef) {
454 DEBUG_r(PerlIO_printf(Perl_debug_log,
455 "Non-utf string cannot match utf check string\n"));
458 if (prog->reganch & ROPT_ANCH) { /* Match at beg-of-str or after \n */
459 ml_anch = !( (prog->reganch & ROPT_ANCH_SINGLE)
460 || ( (prog->reganch & ROPT_ANCH_BOL)
461 && !PL_multiline ) ); /* Check after \n? */
464 if ( !(prog->reganch & (ROPT_ANCH_GPOS /* Checked by the caller */
465 | ROPT_IMPLICIT)) /* not a real BOL */
466 /* SvCUR is not set on references: SvRV and SvPVX overlap */
468 && (strpos != strbeg)) {
469 DEBUG_r(PerlIO_printf(Perl_debug_log, "Not at start...\n"));
472 if (prog->check_offset_min == prog->check_offset_max &&
473 !(prog->reganch & ROPT_CANY_SEEN)) {
474 /* Substring at constant offset from beg-of-str... */
477 s = HOP3c(strpos, prog->check_offset_min, strend);
479 slen = SvCUR(check); /* >= 1 */
481 if ( strend - s > slen || strend - s < slen - 1
482 || (strend - s == slen && strend[-1] != '\n')) {
483 DEBUG_r(PerlIO_printf(Perl_debug_log, "String too long...\n"));
486 /* Now should match s[0..slen-2] */
488 if (slen && (*SvPVX(check) != *s
490 && memNE(SvPVX(check), s, slen)))) {
492 DEBUG_r(PerlIO_printf(Perl_debug_log, "String not equal...\n"));
496 else if (*SvPVX(check) != *s
497 || ((slen = SvCUR(check)) > 1
498 && memNE(SvPVX(check), s, slen)))
500 goto success_at_start;
503 /* Match is anchored, but substr is not anchored wrt beg-of-str. */
505 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
506 end_shift = prog->minlen - start_shift -
507 CHR_SVLEN(check) + (SvTAIL(check) != 0);
509 I32 end = prog->check_offset_max + CHR_SVLEN(check)
510 - (SvTAIL(check) != 0);
511 I32 eshift = CHR_DIST((U8*)strend, (U8*)s) - end;
513 if (end_shift < eshift)
517 else { /* Can match at random position */
520 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
521 /* Should be nonnegative! */
522 end_shift = prog->minlen - start_shift -
523 CHR_SVLEN(check) + (SvTAIL(check) != 0);
526 #ifdef DEBUGGING /* 7/99: reports of failure (with the older version) */
528 Perl_croak(aTHX_ "panic: end_shift");
532 /* Find a possible match in the region s..strend by looking for
533 the "check" substring in the region corrected by start/end_shift. */
534 if (flags & REXEC_SCREAM) {
535 I32 p = -1; /* Internal iterator of scream. */
536 I32 *pp = data ? data->scream_pos : &p;
538 if (PL_screamfirst[BmRARE(check)] >= 0
539 || ( BmRARE(check) == '\n'
540 && (BmPREVIOUS(check) == SvCUR(check) - 1)
542 s = screaminstr(sv, check,
543 start_shift + (s - strbeg), end_shift, pp, 0);
546 /* we may be pointing at the wrong string */
547 if (s && RX_MATCH_COPIED(prog))
548 s = strbeg + (s - SvPVX(sv));
550 *data->scream_olds = s;
552 else if (prog->reganch & ROPT_CANY_SEEN)
553 s = fbm_instr((U8*)(s + start_shift),
554 (U8*)(strend - end_shift),
555 check, PL_multiline ? FBMrf_MULTILINE : 0);
557 s = fbm_instr(HOP3(s, start_shift, strend),
558 HOP3(strend, -end_shift, strbeg),
559 check, PL_multiline ? FBMrf_MULTILINE : 0);
561 /* Update the count-of-usability, remove useless subpatterns,
564 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s %s substr `%s%.*s%s'%s%s",
565 (s ? "Found" : "Did not find"),
566 (check == (do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) ? "anchored" : "floating"),
568 (int)(SvCUR(check) - (SvTAIL(check)!=0)),
570 PL_colors[1], (SvTAIL(check) ? "$" : ""),
571 (s ? " at offset " : "...\n") ) );
578 /* Finish the diagnostic message */
579 DEBUG_r(PerlIO_printf(Perl_debug_log, "%ld...\n", (long)(s - i_strpos)) );
581 /* Got a candidate. Check MBOL anchoring, and the *other* substr.
582 Start with the other substr.
583 XXXX no SCREAM optimization yet - and a very coarse implementation
584 XXXX /ttx+/ results in anchored=`ttx', floating=`x'. floating will
585 *always* match. Probably should be marked during compile...
586 Probably it is right to do no SCREAM here...
589 if (do_utf8 ? (prog->float_utf8 && prog->anchored_utf8) : (prog->float_substr && prog->anchored_substr)) {
590 /* Take into account the "other" substring. */
591 /* XXXX May be hopelessly wrong for UTF... */
594 if (check == (do_utf8 ? prog->float_utf8 : prog->float_substr)) {
597 char *last = HOP3c(s, -start_shift, strbeg), *last1, *last2;
601 t = s - prog->check_offset_max;
602 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
604 || ((t = reghopmaybe3_c(s, -(prog->check_offset_max), strpos))
609 t = HOP3c(t, prog->anchored_offset, strend);
610 if (t < other_last) /* These positions already checked */
612 last2 = last1 = HOP3c(strend, -prog->minlen, strbeg);
615 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
616 /* On end-of-str: see comment below. */
617 must = do_utf8 ? prog->anchored_utf8 : prog->anchored_substr;
618 if (must == &PL_sv_undef) {
620 DEBUG_r(must = prog->anchored_utf8); /* for debug */
625 HOP3(HOP3(last1, prog->anchored_offset, strend)
626 + SvCUR(must), -(SvTAIL(must)!=0), strbeg),
628 PL_multiline ? FBMrf_MULTILINE : 0
630 DEBUG_r(PerlIO_printf(Perl_debug_log,
631 "%s anchored substr `%s%.*s%s'%s",
632 (s ? "Found" : "Contradicts"),
635 - (SvTAIL(must)!=0)),
637 PL_colors[1], (SvTAIL(must) ? "$" : "")));
639 if (last1 >= last2) {
640 DEBUG_r(PerlIO_printf(Perl_debug_log,
641 ", giving up...\n"));
644 DEBUG_r(PerlIO_printf(Perl_debug_log,
645 ", trying floating at offset %ld...\n",
646 (long)(HOP3c(s1, 1, strend) - i_strpos)));
647 other_last = HOP3c(last1, prog->anchored_offset+1, strend);
648 s = HOP3c(last, 1, strend);
652 DEBUG_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
653 (long)(s - i_strpos)));
654 t = HOP3c(s, -prog->anchored_offset, strbeg);
655 other_last = HOP3c(s, 1, strend);
663 else { /* Take into account the floating substring. */
668 t = HOP3c(s, -start_shift, strbeg);
670 HOP3c(strend, -prog->minlen + prog->float_min_offset, strbeg);
671 if (CHR_DIST((U8*)last, (U8*)t) > prog->float_max_offset)
672 last = HOP3c(t, prog->float_max_offset, strend);
673 s = HOP3c(t, prog->float_min_offset, strend);
676 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
677 must = do_utf8 ? prog->float_utf8 : prog->float_substr;
678 /* fbm_instr() takes into account exact value of end-of-str
679 if the check is SvTAIL(ed). Since false positives are OK,
680 and end-of-str is not later than strend we are OK. */
681 if (must == &PL_sv_undef) {
683 DEBUG_r(must = prog->float_utf8); /* for debug message */
686 s = fbm_instr((unsigned char*)s,
687 (unsigned char*)last + SvCUR(must)
689 must, PL_multiline ? FBMrf_MULTILINE : 0);
690 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s floating substr `%s%.*s%s'%s",
691 (s ? "Found" : "Contradicts"),
693 (int)(SvCUR(must) - (SvTAIL(must)!=0)),
695 PL_colors[1], (SvTAIL(must) ? "$" : "")));
698 DEBUG_r(PerlIO_printf(Perl_debug_log,
699 ", giving up...\n"));
702 DEBUG_r(PerlIO_printf(Perl_debug_log,
703 ", trying anchored starting at offset %ld...\n",
704 (long)(s1 + 1 - i_strpos)));
706 s = HOP3c(t, 1, strend);
710 DEBUG_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
711 (long)(s - i_strpos)));
712 other_last = s; /* Fix this later. --Hugo */
721 t = s - prog->check_offset_max;
722 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
724 || ((t = reghopmaybe3_c(s, -prog->check_offset_max, strpos))
726 /* Fixed substring is found far enough so that the match
727 cannot start at strpos. */
729 if (ml_anch && t[-1] != '\n') {
730 /* Eventually fbm_*() should handle this, but often
731 anchored_offset is not 0, so this check will not be wasted. */
732 /* XXXX In the code below we prefer to look for "^" even in
733 presence of anchored substrings. And we search even
734 beyond the found float position. These pessimizations
735 are historical artefacts only. */
737 while (t < strend - prog->minlen) {
739 if (t < check_at - prog->check_offset_min) {
740 if (do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) {
741 /* Since we moved from the found position,
742 we definitely contradict the found anchored
743 substr. Due to the above check we do not
744 contradict "check" substr.
745 Thus we can arrive here only if check substr
746 is float. Redo checking for "other"=="fixed".
749 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld, rescanning for anchored from offset %ld...\n",
750 PL_colors[0],PL_colors[1], (long)(strpos - i_strpos), (long)(strpos - i_strpos + prog->anchored_offset)));
751 goto do_other_anchored;
753 /* We don't contradict the found floating substring. */
754 /* XXXX Why not check for STCLASS? */
756 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld...\n",
757 PL_colors[0],PL_colors[1], (long)(s - i_strpos)));
760 /* Position contradicts check-string */
761 /* XXXX probably better to look for check-string
762 than for "\n", so one should lower the limit for t? */
763 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m, restarting lookup for check-string at offset %ld...\n",
764 PL_colors[0],PL_colors[1], (long)(t + 1 - i_strpos)));
765 other_last = strpos = s = t + 1;
770 DEBUG_r(PerlIO_printf(Perl_debug_log, "Did not find /%s^%s/m...\n",
771 PL_colors[0],PL_colors[1]));
775 DEBUG_r(PerlIO_printf(Perl_debug_log, "Starting position does not contradict /%s^%s/m...\n",
776 PL_colors[0],PL_colors[1]));
780 ++BmUSEFUL(do_utf8 ? prog->check_utf8 : prog->check_substr); /* hooray/5 */
783 /* The found string does not prohibit matching at strpos,
784 - no optimization of calling REx engine can be performed,
785 unless it was an MBOL and we are not after MBOL,
786 or a future STCLASS check will fail this. */
788 /* Even in this situation we may use MBOL flag if strpos is offset
789 wrt the start of the string. */
790 if (ml_anch && sv && !SvROK(sv) /* See prev comment on SvROK */
791 && (strpos != strbeg) && strpos[-1] != '\n'
792 /* May be due to an implicit anchor of m{.*foo} */
793 && !(prog->reganch & ROPT_IMPLICIT))
798 DEBUG_r( if (ml_anch)
799 PerlIO_printf(Perl_debug_log, "Position at offset %ld does not contradict /%s^%s/m...\n",
800 (long)(strpos - i_strpos), PL_colors[0],PL_colors[1]);
803 if (!(prog->reganch & ROPT_NAUGHTY) /* XXXX If strpos moved? */
805 prog->check_utf8 /* Could be deleted already */
806 && --BmUSEFUL(prog->check_utf8) < 0
807 && (prog->check_utf8 == prog->float_utf8)
809 prog->check_substr /* Could be deleted already */
810 && --BmUSEFUL(prog->check_substr) < 0
811 && (prog->check_substr == prog->float_substr)
814 /* If flags & SOMETHING - do not do it many times on the same match */
815 DEBUG_r(PerlIO_printf(Perl_debug_log, "... Disabling check substring...\n"));
816 SvREFCNT_dec(do_utf8 ? prog->check_utf8 : prog->check_substr);
817 if (do_utf8 ? prog->check_substr : prog->check_utf8)
818 SvREFCNT_dec(do_utf8 ? prog->check_substr : prog->check_utf8);
819 prog->check_substr = prog->check_utf8 = Nullsv; /* disable */
820 prog->float_substr = prog->float_utf8 = Nullsv; /* clear */
821 check = Nullsv; /* abort */
823 /* XXXX This is a remnant of the old implementation. It
824 looks wasteful, since now INTUIT can use many
826 prog->reganch &= ~RE_USE_INTUIT;
833 /* XXXX BmUSEFUL already changed, maybe multiple change is meaningful... */
834 if (prog->regstclass) {
835 /* minlen == 0 is possible if regstclass is \b or \B,
836 and the fixed substr is ''$.
837 Since minlen is already taken into account, s+1 is before strend;
838 accidentally, minlen >= 1 guaranties no false positives at s + 1
839 even for \b or \B. But (minlen? 1 : 0) below assumes that
840 regstclass does not come from lookahead... */
841 /* If regstclass takes bytelength more than 1: If charlength==1, OK.
842 This leaves EXACTF only, which is dealt with in find_byclass(). */
843 U8* str = (U8*)STRING(prog->regstclass);
844 int cl_l = (PL_regkind[(U8)OP(prog->regstclass)] == EXACT
845 ? CHR_DIST(str+STR_LEN(prog->regstclass), str)
847 char *endpos = (prog->anchored_substr || prog->anchored_utf8 || ml_anch)
848 ? HOP3c(s, (prog->minlen ? cl_l : 0), strend)
849 : (prog->float_substr || prog->float_utf8
850 ? HOP3c(HOP3c(check_at, -start_shift, strbeg),
853 char *startpos = strbeg;
857 s = find_byclass(prog, prog->regstclass, s, endpos, startpos, 1);
862 if (endpos == strend) {
863 DEBUG_r( PerlIO_printf(Perl_debug_log,
864 "Could not match STCLASS...\n") );
867 DEBUG_r( PerlIO_printf(Perl_debug_log,
868 "This position contradicts STCLASS...\n") );
869 if ((prog->reganch & ROPT_ANCH) && !ml_anch)
871 /* Contradict one of substrings */
872 if (prog->anchored_substr || prog->anchored_utf8) {
873 if ((do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) == check) {
874 DEBUG_r( what = "anchored" );
876 s = HOP3c(t, 1, strend);
877 if (s + start_shift + end_shift > strend) {
878 /* XXXX Should be taken into account earlier? */
879 DEBUG_r( PerlIO_printf(Perl_debug_log,
880 "Could not match STCLASS...\n") );
885 DEBUG_r( PerlIO_printf(Perl_debug_log,
886 "Looking for %s substr starting at offset %ld...\n",
887 what, (long)(s + start_shift - i_strpos)) );
890 /* Have both, check_string is floating */
891 if (t + start_shift >= check_at) /* Contradicts floating=check */
892 goto retry_floating_check;
893 /* Recheck anchored substring, but not floating... */
897 DEBUG_r( PerlIO_printf(Perl_debug_log,
898 "Looking for anchored substr starting at offset %ld...\n",
899 (long)(other_last - i_strpos)) );
900 goto do_other_anchored;
902 /* Another way we could have checked stclass at the
903 current position only: */
908 DEBUG_r( PerlIO_printf(Perl_debug_log,
909 "Looking for /%s^%s/m starting at offset %ld...\n",
910 PL_colors[0],PL_colors[1], (long)(t - i_strpos)) );
913 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr)) /* Could have been deleted */
915 /* Check is floating subtring. */
916 retry_floating_check:
917 t = check_at - start_shift;
918 DEBUG_r( what = "floating" );
919 goto hop_and_restart;
922 DEBUG_r(PerlIO_printf(Perl_debug_log,
923 "By STCLASS: moving %ld --> %ld\n",
924 (long)(t - i_strpos), (long)(s - i_strpos))
928 DEBUG_r(PerlIO_printf(Perl_debug_log,
929 "Does not contradict STCLASS...\n");
934 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s%s:%s match at offset %ld\n",
935 PL_colors[4], (check ? "Guessed" : "Giving up"),
936 PL_colors[5], (long)(s - i_strpos)) );
939 fail_finish: /* Substring not found */
940 if (prog->check_substr || prog->check_utf8) /* could be removed already */
941 BmUSEFUL(do_utf8 ? prog->check_utf8 : prog->check_substr) += 5; /* hooray */
943 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch rejected by optimizer%s\n",
944 PL_colors[4],PL_colors[5]));
948 /* We know what class REx starts with. Try to find this position... */
950 S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, char *strend, char *startpos, I32 norun)
952 I32 doevery = (prog->reganch & ROPT_SKIP) == 0;
958 register I32 tmp = 1; /* Scratch variable? */
959 register bool do_utf8 = PL_reg_match_utf8;
961 /* We know what class it must start with. */
966 if ((ANYOF_FLAGS(c) & ANYOF_UNICODE) ||
967 !UTF8_IS_INVARIANT((U8)s[0]) ?
968 reginclass(c, (U8*)s, 0, do_utf8) :
969 REGINCLASS(c, (U8*)s)) {
970 if (tmp && (norun || regtry(prog, s)))
984 if (REGINCLASS(c, (U8*)s) ||
985 (ANYOF_FOLD_SHARP_S(c, s, strend) &&
986 /* The assignment of 2 is intentional:
987 * for the folded sharp s, the skip is 2. */
988 (skip = SHARP_S_SKIP))) {
989 if (tmp && (norun || regtry(prog, s)))
1001 while (s < strend) {
1002 if (tmp && (norun || regtry(prog, s)))
1013 STRLEN ulen1, ulen2;
1014 U8 tmpbuf1[UTF8_MAXLEN_UCLC+1];
1015 U8 tmpbuf2[UTF8_MAXLEN_UCLC+1];
1017 to_utf8_lower((U8*)m, tmpbuf1, &ulen1);
1018 to_utf8_upper((U8*)m, tmpbuf2, &ulen2);
1020 c1 = utf8n_to_uvchr(tmpbuf1, UTF8_MAXLEN_UCLC,
1021 0, ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
1022 c2 = utf8n_to_uvchr(tmpbuf2, UTF8_MAXLEN_UCLC,
1023 0, ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
1034 c2 = PL_fold_locale[c1];
1036 e = HOP3c(strend, -(I32)ln, s);
1039 e = s; /* Due to minlen logic of intuit() */
1041 /* The idea in the EXACTF* cases is to first find the
1042 * first character of the EXACTF* node and then, if
1043 * necessary, case-insensitively compare the full
1044 * text of the node. The c1 and c2 are the first
1045 * characters (though in Unicode it gets a bit
1046 * more complicated because there are more cases
1047 * than just upper and lower: one needs to use
1048 * the so-called folding case for case-insensitive
1049 * matching (called "loose matching" in Unicode).
1050 * ibcmp_utf8() will do just that. */
1054 U8 tmpbuf [UTF8_MAXLEN+1];
1055 U8 foldbuf[UTF8_MAXLEN_FOLD+1];
1056 STRLEN len, foldlen;
1060 c = utf8n_to_uvchr((U8*)s, UTF8_MAXLEN, &len,
1062 0 : UTF8_ALLOW_ANY);
1065 ibcmp_utf8(s, (char **)0, 0, do_utf8,
1066 m, (char **)0, ln, (bool)UTF))
1067 && (norun || regtry(prog, s)) )
1070 uvchr_to_utf8(tmpbuf, c);
1071 f = to_utf8_fold(tmpbuf, foldbuf, &foldlen);
1073 && (f == c1 || f == c2)
1074 && (ln == foldlen ||
1075 !ibcmp_utf8((char *) foldbuf,
1076 (char **)0, foldlen, do_utf8,
1078 (char **)0, ln, (bool)UTF))
1079 && (norun || regtry(prog, s)) )
1087 c = utf8n_to_uvchr((U8*)s, UTF8_MAXLEN, &len,
1089 0 : UTF8_ALLOW_ANY);
1091 /* Handle some of the three Greek sigmas cases.
1092 * Note that not all the possible combinations
1093 * are handled here: some of them are handled
1094 * by the standard folding rules, and some of
1095 * them (the character class or ANYOF cases)
1096 * are handled during compiletime in
1097 * regexec.c:S_regclass(). */
1098 if (c == (UV)UNICODE_GREEK_CAPITAL_LETTER_SIGMA ||
1099 c == (UV)UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA)
1100 c = (UV)UNICODE_GREEK_SMALL_LETTER_SIGMA;
1102 if ( (c == c1 || c == c2)
1104 ibcmp_utf8(s, (char **)0, 0, do_utf8,
1105 m, (char **)0, ln, (bool)UTF))
1106 && (norun || regtry(prog, s)) )
1109 uvchr_to_utf8(tmpbuf, c);
1110 f = to_utf8_fold(tmpbuf, foldbuf, &foldlen);
1112 && (f == c1 || f == c2)
1113 && (ln == foldlen ||
1114 !ibcmp_utf8((char *) foldbuf,
1115 (char **)0, foldlen, do_utf8,
1117 (char **)0, ln, (bool)UTF))
1118 && (norun || regtry(prog, s)) )
1129 && (ln == 1 || !(OP(c) == EXACTF
1131 : ibcmp_locale(s, m, ln)))
1132 && (norun || regtry(prog, s)) )
1138 if ( (*(U8*)s == c1 || *(U8*)s == c2)
1139 && (ln == 1 || !(OP(c) == EXACTF
1141 : ibcmp_locale(s, m, ln)))
1142 && (norun || regtry(prog, s)) )
1149 PL_reg_flags |= RF_tainted;
1156 U8 *r = reghop3((U8*)s, -1, (U8*)PL_bostr);
1158 tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, 0);
1160 tmp = ((OP(c) == BOUND ?
1161 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1162 LOAD_UTF8_CHARCLASS(alnum,"a");
1163 while (s < strend) {
1164 if (tmp == !(OP(c) == BOUND ?
1165 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1166 isALNUM_LC_utf8((U8*)s)))
1169 if ((norun || regtry(prog, s)))
1176 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1177 tmp = ((OP(c) == BOUND ? isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1178 while (s < strend) {
1180 !(OP(c) == BOUND ? isALNUM(*s) : isALNUM_LC(*s))) {
1182 if ((norun || regtry(prog, s)))
1188 if ((!prog->minlen && tmp) && (norun || regtry(prog, s)))
1192 PL_reg_flags |= RF_tainted;
1199 U8 *r = reghop3((U8*)s, -1, (U8*)PL_bostr);
1201 tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, 0);
1203 tmp = ((OP(c) == NBOUND ?
1204 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1205 LOAD_UTF8_CHARCLASS(alnum,"a");
1206 while (s < strend) {
1207 if (tmp == !(OP(c) == NBOUND ?
1208 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1209 isALNUM_LC_utf8((U8*)s)))
1211 else if ((norun || regtry(prog, s)))
1217 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1218 tmp = ((OP(c) == NBOUND ?
1219 isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1220 while (s < strend) {
1222 !(OP(c) == NBOUND ? isALNUM(*s) : isALNUM_LC(*s)))
1224 else if ((norun || regtry(prog, s)))
1229 if ((!prog->minlen && !tmp) && (norun || regtry(prog, s)))
1234 LOAD_UTF8_CHARCLASS(alnum,"a");
1235 while (s < strend) {
1236 if (swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1237 if (tmp && (norun || regtry(prog, s)))
1248 while (s < strend) {
1250 if (tmp && (norun || regtry(prog, s)))
1262 PL_reg_flags |= RF_tainted;
1264 while (s < strend) {
1265 if (isALNUM_LC_utf8((U8*)s)) {
1266 if (tmp && (norun || regtry(prog, s)))
1277 while (s < strend) {
1278 if (isALNUM_LC(*s)) {
1279 if (tmp && (norun || regtry(prog, s)))
1292 LOAD_UTF8_CHARCLASS(alnum,"a");
1293 while (s < strend) {
1294 if (!swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1295 if (tmp && (norun || regtry(prog, s)))
1306 while (s < strend) {
1308 if (tmp && (norun || regtry(prog, s)))
1320 PL_reg_flags |= RF_tainted;
1322 while (s < strend) {
1323 if (!isALNUM_LC_utf8((U8*)s)) {
1324 if (tmp && (norun || regtry(prog, s)))
1335 while (s < strend) {
1336 if (!isALNUM_LC(*s)) {
1337 if (tmp && (norun || regtry(prog, s)))
1350 LOAD_UTF8_CHARCLASS(space," ");
1351 while (s < strend) {
1352 if (*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8)) {
1353 if (tmp && (norun || regtry(prog, s)))
1364 while (s < strend) {
1366 if (tmp && (norun || regtry(prog, s)))
1378 PL_reg_flags |= RF_tainted;
1380 while (s < strend) {
1381 if (*s == ' ' || isSPACE_LC_utf8((U8*)s)) {
1382 if (tmp && (norun || regtry(prog, s)))
1393 while (s < strend) {
1394 if (isSPACE_LC(*s)) {
1395 if (tmp && (norun || regtry(prog, s)))
1408 LOAD_UTF8_CHARCLASS(space," ");
1409 while (s < strend) {
1410 if (!(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8))) {
1411 if (tmp && (norun || regtry(prog, s)))
1422 while (s < strend) {
1424 if (tmp && (norun || regtry(prog, s)))
1436 PL_reg_flags |= RF_tainted;
1438 while (s < strend) {
1439 if (!(*s == ' ' || isSPACE_LC_utf8((U8*)s))) {
1440 if (tmp && (norun || regtry(prog, s)))
1451 while (s < strend) {
1452 if (!isSPACE_LC(*s)) {
1453 if (tmp && (norun || regtry(prog, s)))
1466 LOAD_UTF8_CHARCLASS(digit,"0");
1467 while (s < strend) {
1468 if (swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1469 if (tmp && (norun || regtry(prog, s)))
1480 while (s < strend) {
1482 if (tmp && (norun || regtry(prog, s)))
1494 PL_reg_flags |= RF_tainted;
1496 while (s < strend) {
1497 if (isDIGIT_LC_utf8((U8*)s)) {
1498 if (tmp && (norun || regtry(prog, s)))
1509 while (s < strend) {
1510 if (isDIGIT_LC(*s)) {
1511 if (tmp && (norun || regtry(prog, s)))
1524 LOAD_UTF8_CHARCLASS(digit,"0");
1525 while (s < strend) {
1526 if (!swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1527 if (tmp && (norun || regtry(prog, s)))
1538 while (s < strend) {
1540 if (tmp && (norun || regtry(prog, s)))
1552 PL_reg_flags |= RF_tainted;
1554 while (s < strend) {
1555 if (!isDIGIT_LC_utf8((U8*)s)) {
1556 if (tmp && (norun || regtry(prog, s)))
1567 while (s < strend) {
1568 if (!isDIGIT_LC(*s)) {
1569 if (tmp && (norun || regtry(prog, s)))
1581 Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c));
1590 - regexec_flags - match a regexp against a string
1593 Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char *strend,
1594 char *strbeg, I32 minend, SV *sv, void *data, U32 flags)
1595 /* strend: pointer to null at end of string */
1596 /* strbeg: real beginning of string */
1597 /* minend: end of match must be >=minend after stringarg. */
1598 /* data: May be used for some additional optimizations. */
1599 /* nosave: For optimizations. */
1602 register regnode *c;
1603 register char *startpos = stringarg;
1604 I32 minlen; /* must match at least this many chars */
1605 I32 dontbother = 0; /* how many characters not to try at end */
1606 /* I32 start_shift = 0; */ /* Offset of the start to find
1607 constant substr. */ /* CC */
1608 I32 end_shift = 0; /* Same for the end. */ /* CC */
1609 I32 scream_pos = -1; /* Internal iterator of scream. */
1611 SV* oreplsv = GvSV(PL_replgv);
1612 bool do_utf8 = DO_UTF8(sv);
1614 SV *dsv0 = PERL_DEBUG_PAD_ZERO(0);
1615 SV *dsv1 = PERL_DEBUG_PAD_ZERO(1);
1617 RX_MATCH_UTF8_set(prog,do_utf8);
1623 PL_regnarrate = DEBUG_r_TEST;
1626 /* Be paranoid... */
1627 if (prog == NULL || startpos == NULL) {
1628 Perl_croak(aTHX_ "NULL regexp parameter");
1632 minlen = prog->minlen;
1633 if (strend - startpos < minlen) {
1634 DEBUG_r(PerlIO_printf(Perl_debug_log,
1635 "String too short [regexec_flags]...\n"));
1639 /* Check validity of program. */
1640 if (UCHARAT(prog->program) != REG_MAGIC) {
1641 Perl_croak(aTHX_ "corrupted regexp program");
1645 PL_reg_eval_set = 0;
1648 if (prog->reganch & ROPT_UTF8)
1649 PL_reg_flags |= RF_utf8;
1651 /* Mark beginning of line for ^ and lookbehind. */
1652 PL_regbol = startpos;
1656 /* Mark end of line for $ (and such) */
1659 /* see how far we have to get to not match where we matched before */
1660 PL_regtill = startpos+minend;
1662 /* We start without call_cc context. */
1665 /* If there is a "must appear" string, look for it. */
1668 if (prog->reganch & ROPT_GPOS_SEEN) { /* Need to have PL_reg_ganch */
1671 if (flags & REXEC_IGNOREPOS) /* Means: check only at start */
1672 PL_reg_ganch = startpos;
1673 else if (sv && SvTYPE(sv) >= SVt_PVMG
1675 && (mg = mg_find(sv, PERL_MAGIC_regex_global))
1676 && mg->mg_len >= 0) {
1677 PL_reg_ganch = strbeg + mg->mg_len; /* Defined pos() */
1678 if (prog->reganch & ROPT_ANCH_GPOS) {
1679 if (s > PL_reg_ganch)
1684 else /* pos() not defined */
1685 PL_reg_ganch = strbeg;
1688 if (!(flags & REXEC_CHECKED) && (prog->check_substr != Nullsv || prog->check_utf8 != Nullsv)) {
1689 re_scream_pos_data d;
1691 d.scream_olds = &scream_olds;
1692 d.scream_pos = &scream_pos;
1693 s = re_intuit_start(prog, sv, s, strend, flags, &d);
1695 DEBUG_r(PerlIO_printf(Perl_debug_log, "Not present...\n"));
1696 goto phooey; /* not present */
1702 pv_uni_display(dsv0, (U8*)prog->precomp, prog->prelen, 60,
1703 UNI_DISPLAY_REGEX) :
1705 int len0 = UTF ? SvCUR(dsv0) : prog->prelen;
1706 char *s1 = do_utf8 ? sv_uni_display(dsv1, sv, 60,
1707 UNI_DISPLAY_REGEX) : startpos;
1708 int len1 = do_utf8 ? SvCUR(dsv1) : strend - startpos;
1711 PerlIO_printf(Perl_debug_log,
1712 "%sMatching REx%s `%s%*.*s%s%s' against `%s%.*s%s%s'\n",
1713 PL_colors[4],PL_colors[5],PL_colors[0],
1716 len0 > 60 ? "..." : "",
1718 (int)(len1 > 60 ? 60 : len1),
1720 (len1 > 60 ? "..." : "")
1724 /* Simplest case: anchored match need be tried only once. */
1725 /* [unless only anchor is BOL and multiline is set] */
1726 if (prog->reganch & (ROPT_ANCH & ~ROPT_ANCH_GPOS)) {
1727 if (s == startpos && regtry(prog, startpos))
1729 else if (PL_multiline || (prog->reganch & ROPT_IMPLICIT)
1730 || (prog->reganch & ROPT_ANCH_MBOL)) /* XXXX SBOL? */
1735 dontbother = minlen - 1;
1736 end = HOP3c(strend, -dontbother, strbeg) - 1;
1737 /* for multiline we only have to try after newlines */
1738 if (prog->check_substr || prog->check_utf8) {
1742 if (regtry(prog, s))
1747 if (prog->reganch & RE_USE_INTUIT) {
1748 s = re_intuit_start(prog, sv, s + 1, strend, flags, NULL);
1759 if (*s++ == '\n') { /* don't need PL_utf8skip here */
1760 if (regtry(prog, s))
1767 } else if (prog->reganch & ROPT_ANCH_GPOS) {
1768 if (regtry(prog, PL_reg_ganch))
1773 /* Messy cases: unanchored match. */
1774 if ((prog->anchored_substr || prog->anchored_utf8) && prog->reganch & ROPT_SKIP) {
1775 /* we have /x+whatever/ */
1776 /* it must be a one character string (XXXX Except UTF?) */
1781 if (!(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr))
1782 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1783 ch = SvPVX(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr)[0];
1786 while (s < strend) {
1788 DEBUG_r( did_match = 1 );
1789 if (regtry(prog, s)) goto got_it;
1791 while (s < strend && *s == ch)
1798 while (s < strend) {
1800 DEBUG_r( did_match = 1 );
1801 if (regtry(prog, s)) goto got_it;
1803 while (s < strend && *s == ch)
1809 DEBUG_r(if (!did_match)
1810 PerlIO_printf(Perl_debug_log,
1811 "Did not find anchored character...\n")
1815 else if (prog->anchored_substr != Nullsv
1816 || prog->anchored_utf8 != Nullsv
1817 || ((prog->float_substr != Nullsv || prog->float_utf8 != Nullsv)
1818 && prog->float_max_offset < strend - s)) {
1823 char *last1; /* Last position checked before */
1827 if (prog->anchored_substr || prog->anchored_utf8) {
1828 if (!(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr))
1829 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1830 must = do_utf8 ? prog->anchored_utf8 : prog->anchored_substr;
1831 back_max = back_min = prog->anchored_offset;
1833 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr))
1834 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1835 must = do_utf8 ? prog->float_utf8 : prog->float_substr;
1836 back_max = prog->float_max_offset;
1837 back_min = prog->float_min_offset;
1839 if (must == &PL_sv_undef)
1840 /* could not downgrade utf8 check substring, so must fail */
1843 last = HOP3c(strend, /* Cannot start after this */
1844 -(I32)(CHR_SVLEN(must)
1845 - (SvTAIL(must) != 0) + back_min), strbeg);
1848 last1 = HOPc(s, -1);
1850 last1 = s - 1; /* bogus */
1852 /* XXXX check_substr already used to find `s', can optimize if
1853 check_substr==must. */
1855 dontbother = end_shift;
1856 strend = HOPc(strend, -dontbother);
1857 while ( (s <= last) &&
1858 ((flags & REXEC_SCREAM)
1859 ? (s = screaminstr(sv, must, HOP3c(s, back_min, strend) - strbeg,
1860 end_shift, &scream_pos, 0))
1861 : (s = fbm_instr((unsigned char*)HOP3(s, back_min, strend),
1862 (unsigned char*)strend, must,
1863 PL_multiline ? FBMrf_MULTILINE : 0))) ) {
1864 /* we may be pointing at the wrong string */
1865 if ((flags & REXEC_SCREAM) && RX_MATCH_COPIED(prog))
1866 s = strbeg + (s - SvPVX(sv));
1867 DEBUG_r( did_match = 1 );
1868 if (HOPc(s, -back_max) > last1) {
1869 last1 = HOPc(s, -back_min);
1870 s = HOPc(s, -back_max);
1873 char *t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1;
1875 last1 = HOPc(s, -back_min);
1879 while (s <= last1) {
1880 if (regtry(prog, s))
1886 while (s <= last1) {
1887 if (regtry(prog, s))
1893 DEBUG_r(if (!did_match)
1894 PerlIO_printf(Perl_debug_log,
1895 "Did not find %s substr `%s%.*s%s'%s...\n",
1896 ((must == prog->anchored_substr || must == prog->anchored_utf8)
1897 ? "anchored" : "floating"),
1899 (int)(SvCUR(must) - (SvTAIL(must)!=0)),
1901 PL_colors[1], (SvTAIL(must) ? "$" : ""))
1905 else if ((c = prog->regstclass)) {
1907 I32 op = (U8)OP(prog->regstclass);
1908 /* don't bother with what can't match */
1909 if (PL_regkind[op] != EXACT && op != CANY)
1910 strend = HOPc(strend, -(minlen - 1));
1913 SV *prop = sv_newmortal();
1921 pv_uni_display(dsv0, (U8*)SvPVX(prop), SvCUR(prop), 60,
1922 UNI_DISPLAY_REGEX) :
1924 len0 = UTF ? SvCUR(dsv0) : SvCUR(prop);
1926 sv_uni_display(dsv1, sv, 60, UNI_DISPLAY_REGEX) : s;
1927 len1 = UTF ? SvCUR(dsv1) : strend - s;
1928 PerlIO_printf(Perl_debug_log,
1929 "Matching stclass `%*.*s' against `%*.*s'\n",
1933 if (find_byclass(prog, c, s, strend, startpos, 0))
1935 DEBUG_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass...\n"));
1939 if (prog->float_substr != Nullsv || prog->float_utf8 != Nullsv) {
1944 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr))
1945 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1946 float_real = do_utf8 ? prog->float_utf8 : prog->float_substr;
1948 if (flags & REXEC_SCREAM) {
1949 last = screaminstr(sv, float_real, s - strbeg,
1950 end_shift, &scream_pos, 1); /* last one */
1952 last = scream_olds; /* Only one occurrence. */
1953 /* we may be pointing at the wrong string */
1954 else if (RX_MATCH_COPIED(prog))
1955 s = strbeg + (s - SvPVX(sv));
1959 char *little = SvPV(float_real, len);
1961 if (SvTAIL(float_real)) {
1962 if (memEQ(strend - len + 1, little, len - 1))
1963 last = strend - len + 1;
1964 else if (!PL_multiline)
1965 last = memEQ(strend - len, little, len)
1966 ? strend - len : Nullch;
1972 last = rninstr(s, strend, little, little + len);
1974 last = strend; /* matching `$' */
1978 DEBUG_r(PerlIO_printf(Perl_debug_log,
1979 "%sCan't trim the tail, match fails (should not happen)%s\n",
1980 PL_colors[4],PL_colors[5]));
1981 goto phooey; /* Should not happen! */
1983 dontbother = strend - last + prog->float_min_offset;
1985 if (minlen && (dontbother < minlen))
1986 dontbother = minlen - 1;
1987 strend -= dontbother; /* this one's always in bytes! */
1988 /* We don't know much -- general case. */
1991 if (regtry(prog, s))
2000 if (regtry(prog, s))
2002 } while (s++ < strend);
2010 RX_MATCH_TAINTED_set(prog, PL_reg_flags & RF_tainted);
2012 if (PL_reg_eval_set) {
2013 /* Preserve the current value of $^R */
2014 if (oreplsv != GvSV(PL_replgv))
2015 sv_setsv(oreplsv, GvSV(PL_replgv));/* So that when GvSV(replgv) is
2016 restored, the value remains
2018 restore_pos(aTHX_ 0);
2021 /* make sure $`, $&, $', and $digit will work later */
2022 if ( !(flags & REXEC_NOT_FIRST) ) {
2023 if (RX_MATCH_COPIED(prog)) {
2024 Safefree(prog->subbeg);
2025 RX_MATCH_COPIED_off(prog);
2027 if (flags & REXEC_COPY_STR) {
2028 I32 i = PL_regeol - startpos + (stringarg - strbeg);
2030 s = savepvn(strbeg, i);
2033 RX_MATCH_COPIED_on(prog);
2036 prog->subbeg = strbeg;
2037 prog->sublen = PL_regeol - strbeg; /* strend may have been modified */
2044 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n",
2045 PL_colors[4],PL_colors[5]));
2046 if (PL_reg_eval_set)
2047 restore_pos(aTHX_ 0);
2052 - regtry - try match at specific point
2054 STATIC I32 /* 0 failure, 1 success */
2055 S_regtry(pTHX_ regexp *prog, char *startpos)
2063 PL_regindent = 0; /* XXXX Not good when matches are reenterable... */
2065 if ((prog->reganch & ROPT_EVAL_SEEN) && !PL_reg_eval_set) {
2068 PL_reg_eval_set = RS_init;
2070 PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %"IVdf"\n",
2071 (IV)(PL_stack_sp - PL_stack_base));
2073 SAVEI32(cxstack[cxstack_ix].blk_oldsp);
2074 cxstack[cxstack_ix].blk_oldsp = PL_stack_sp - PL_stack_base;
2075 /* Otherwise OP_NEXTSTATE will free whatever on stack now. */
2077 /* Apparently this is not needed, judging by wantarray. */
2078 /* SAVEI8(cxstack[cxstack_ix].blk_gimme);
2079 cxstack[cxstack_ix].blk_gimme = G_SCALAR; */
2082 /* Make $_ available to executed code. */
2083 if (PL_reg_sv != DEFSV) {
2084 /* SAVE_DEFSV does *not* suffice here for USE_5005THREADS */
2089 if (!(SvTYPE(PL_reg_sv) >= SVt_PVMG && SvMAGIC(PL_reg_sv)
2090 && (mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global)))) {
2091 /* prepare for quick setting of pos */
2092 sv_magic(PL_reg_sv, (SV*)0,
2093 PERL_MAGIC_regex_global, Nullch, 0);
2094 mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global);
2098 PL_reg_oldpos = mg->mg_len;
2099 SAVEDESTRUCTOR_X(restore_pos, 0);
2101 if (!PL_reg_curpm) {
2102 Newz(22,PL_reg_curpm, 1, PMOP);
2105 SV* repointer = newSViv(0);
2106 /* so we know which PL_regex_padav element is PL_reg_curpm */
2107 SvFLAGS(repointer) |= SVf_BREAK;
2108 av_push(PL_regex_padav,repointer);
2109 PL_reg_curpm->op_pmoffset = av_len(PL_regex_padav);
2110 PL_regex_pad = AvARRAY(PL_regex_padav);
2114 PM_SETRE(PL_reg_curpm, prog);
2115 PL_reg_oldcurpm = PL_curpm;
2116 PL_curpm = PL_reg_curpm;
2117 if (RX_MATCH_COPIED(prog)) {
2118 /* Here is a serious problem: we cannot rewrite subbeg,
2119 since it may be needed if this match fails. Thus
2120 $` inside (?{}) could fail... */
2121 PL_reg_oldsaved = prog->subbeg;
2122 PL_reg_oldsavedlen = prog->sublen;
2123 RX_MATCH_COPIED_off(prog);
2126 PL_reg_oldsaved = Nullch;
2127 prog->subbeg = PL_bostr;
2128 prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */
2130 prog->startp[0] = startpos - PL_bostr;
2131 PL_reginput = startpos;
2132 PL_regstartp = prog->startp;
2133 PL_regendp = prog->endp;
2134 PL_reglastparen = &prog->lastparen;
2135 PL_reglastcloseparen = &prog->lastcloseparen;
2136 prog->lastparen = 0;
2137 prog->lastcloseparen = 0;
2139 DEBUG_r(PL_reg_starttry = startpos);
2140 if (PL_reg_start_tmpl <= prog->nparens) {
2141 PL_reg_start_tmpl = prog->nparens*3/2 + 3;
2142 if(PL_reg_start_tmp)
2143 Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
2145 New(22,PL_reg_start_tmp, PL_reg_start_tmpl, char*);
2148 /* XXXX What this code is doing here?!!! There should be no need
2149 to do this again and again, PL_reglastparen should take care of
2152 /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code.
2153 * Actually, the code in regcppop() (which Ilya may be meaning by
2154 * PL_reglastparen), is not needed at all by the test suite
2155 * (op/regexp, op/pat, op/split), but that code is needed, oddly
2156 * enough, for building DynaLoader, or otherwise this
2157 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
2158 * will happen. Meanwhile, this code *is* needed for the
2159 * above-mentioned test suite tests to succeed. The common theme
2160 * on those tests seems to be returning null fields from matches.
2165 if (prog->nparens) {
2166 for (i = prog->nparens; i > (I32)*PL_reglastparen; i--) {
2173 if (regmatch(prog->program + 1)) {
2174 prog->endp[0] = PL_reginput - PL_bostr;
2177 REGCP_UNWIND(lastcp);
2181 #define RE_UNWIND_BRANCH 1
2182 #define RE_UNWIND_BRANCHJ 2
2186 typedef struct { /* XX: makes sense to enlarge it... */
2190 } re_unwind_generic_t;
2203 } re_unwind_branch_t;
2205 typedef union re_unwind_t {
2207 re_unwind_generic_t generic;
2208 re_unwind_branch_t branch;
2211 #define sayYES goto yes
2212 #define sayNO goto no
2213 #define sayNO_ANYOF goto no_anyof
2214 #define sayYES_FINAL goto yes_final
2215 #define sayYES_LOUD goto yes_loud
2216 #define sayNO_FINAL goto no_final
2217 #define sayNO_SILENT goto do_no
2218 #define saySAME(x) if (x) goto yes; else goto no
2220 #define REPORT_CODE_OFF 24
2223 - regmatch - main matching routine
2225 * Conceptually the strategy is simple: check to see whether the current
2226 * node matches, call self recursively to see whether the rest matches,
2227 * and then act accordingly. In practice we make some effort to avoid
2228 * recursion, in particular by going through "ordinary" nodes (that don't
2229 * need to know whether the rest of the match failed) by a loop instead of
2232 /* [lwall] I've hoisted the register declarations to the outer block in order to
2233 * maybe save a little bit of pushing and popping on the stack. It also takes
2234 * advantage of machines that use a register save mask on subroutine entry.
2236 STATIC I32 /* 0 failure, 1 success */
2237 S_regmatch(pTHX_ regnode *prog)
2239 register regnode *scan; /* Current node. */
2240 regnode *next; /* Next node. */
2241 regnode *inner; /* Next node in internal branch. */
2242 register I32 nextchr; /* renamed nextchr - nextchar colides with
2243 function of same name */
2244 register I32 n; /* no or next */
2245 register I32 ln = 0; /* len or last */
2246 register char *s = Nullch; /* operand or save */
2247 register char *locinput = PL_reginput;
2248 register I32 c1 = 0, c2 = 0, paren; /* case fold search, parenth */
2249 int minmod = 0, sw = 0, logical = 0;
2252 I32 firstcp = PL_savestack_ix;
2254 register bool do_utf8 = PL_reg_match_utf8;
2256 SV *dsv0 = PERL_DEBUG_PAD_ZERO(0);
2257 SV *dsv1 = PERL_DEBUG_PAD_ZERO(1);
2258 SV *dsv2 = PERL_DEBUG_PAD_ZERO(2);
2265 /* Note that nextchr is a byte even in UTF */
2266 nextchr = UCHARAT(locinput);
2268 while (scan != NULL) {
2271 SV *prop = sv_newmortal();
2272 int docolor = *PL_colors[0];
2273 int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */
2274 int l = (PL_regeol - locinput) > taill ? taill : (PL_regeol - locinput);
2275 /* The part of the string before starttry has one color
2276 (pref0_len chars), between starttry and current
2277 position another one (pref_len - pref0_len chars),
2278 after the current position the third one.
2279 We assume that pref0_len <= pref_len, otherwise we
2280 decrease pref0_len. */
2281 int pref_len = (locinput - PL_bostr) > (5 + taill) - l
2282 ? (5 + taill) - l : locinput - PL_bostr;
2285 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len)))
2287 pref0_len = pref_len - (locinput - PL_reg_starttry);
2288 if (l + pref_len < (5 + taill) && l < PL_regeol - locinput)
2289 l = ( PL_regeol - locinput > (5 + taill) - pref_len
2290 ? (5 + taill) - pref_len : PL_regeol - locinput);
2291 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput + l)))
2295 if (pref0_len > pref_len)
2296 pref0_len = pref_len;
2297 regprop(prop, scan);
2300 do_utf8 && OP(scan) != CANY ?
2301 pv_uni_display(dsv0, (U8*)(locinput - pref_len),
2302 pref0_len, 60, UNI_DISPLAY_REGEX) :
2303 locinput - pref_len;
2304 int len0 = do_utf8 ? strlen(s0) : pref0_len;
2305 char *s1 = do_utf8 && OP(scan) != CANY ?
2306 pv_uni_display(dsv1, (U8*)(locinput - pref_len + pref0_len),
2307 pref_len - pref0_len, 60, UNI_DISPLAY_REGEX) :
2308 locinput - pref_len + pref0_len;
2309 int len1 = do_utf8 ? strlen(s1) : pref_len - pref0_len;
2310 char *s2 = do_utf8 && OP(scan) != CANY ?
2311 pv_uni_display(dsv2, (U8*)locinput,
2312 PL_regeol - locinput, 60, UNI_DISPLAY_REGEX) :
2314 int len2 = do_utf8 ? strlen(s2) : l;
2315 PerlIO_printf(Perl_debug_log,
2316 "%4"IVdf" <%s%.*s%s%s%.*s%s%s%s%.*s%s>%*s|%3"IVdf":%*s%s\n",
2317 (IV)(locinput - PL_bostr),
2324 (docolor ? "" : "> <"),
2328 15 - l - pref_len + 1,
2330 (IV)(scan - PL_regprogram), PL_regindent*2, "",
2335 next = scan + NEXT_OFF(scan);
2341 if (locinput == PL_bostr || (PL_multiline &&
2342 (nextchr || locinput < PL_regeol) && locinput[-1] == '\n') )
2344 /* regtill = regbol; */
2349 if (locinput == PL_bostr ||
2350 ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n'))
2356 if (locinput == PL_bostr)
2360 if (locinput == PL_reg_ganch)
2370 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2375 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2377 if (PL_regeol - locinput > 1)
2381 if (PL_regeol != locinput)
2385 if (!nextchr && locinput >= PL_regeol)
2388 locinput += PL_utf8skip[nextchr];
2389 if (locinput > PL_regeol)
2391 nextchr = UCHARAT(locinput);
2394 nextchr = UCHARAT(++locinput);
2397 if (!nextchr && locinput >= PL_regeol)
2399 nextchr = UCHARAT(++locinput);
2402 if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n')
2405 locinput += PL_utf8skip[nextchr];
2406 if (locinput > PL_regeol)
2408 nextchr = UCHARAT(locinput);
2411 nextchr = UCHARAT(++locinput);
2416 if (do_utf8 != UTF) {
2417 /* The target and the pattern have differing utf8ness. */
2423 /* The target is utf8, the pattern is not utf8. */
2427 if (NATIVE_TO_UNI(*(U8*)s) !=
2428 utf8n_to_uvuni((U8*)l, UTF8_MAXLEN, &ulen,
2430 0 : UTF8_ALLOW_ANY))
2437 /* The target is not utf8, the pattern is utf8. */
2441 if (NATIVE_TO_UNI(*((U8*)l)) !=
2442 utf8n_to_uvuni((U8*)s, UTF8_MAXLEN, &ulen,
2444 0 : UTF8_ALLOW_ANY))
2451 nextchr = UCHARAT(locinput);
2454 /* The target and the pattern have the same utf8ness. */
2455 /* Inline the first character, for speed. */
2456 if (UCHARAT(s) != nextchr)
2458 if (PL_regeol - locinput < ln)
2460 if (ln > 1 && memNE(s, locinput, ln))
2463 nextchr = UCHARAT(locinput);
2466 PL_reg_flags |= RF_tainted;
2472 if (do_utf8 || UTF) {
2473 /* Either target or the pattern are utf8. */
2475 char *e = PL_regeol;
2477 if (ibcmp_utf8(s, 0, ln, (bool)UTF,
2478 l, &e, 0, do_utf8)) {
2479 /* One more case for the sharp s:
2480 * pack("U0U*", 0xDF) =~ /ss/i,
2481 * the 0xC3 0x9F are the UTF-8
2482 * byte sequence for the U+00DF. */
2484 toLOWER(s[0]) == 's' &&
2486 toLOWER(s[1]) == 's' &&
2493 nextchr = UCHARAT(locinput);
2497 /* Neither the target and the pattern are utf8. */
2499 /* Inline the first character, for speed. */
2500 if (UCHARAT(s) != nextchr &&
2501 UCHARAT(s) != ((OP(scan) == EXACTF)
2502 ? PL_fold : PL_fold_locale)[nextchr])
2504 if (PL_regeol - locinput < ln)
2506 if (ln > 1 && (OP(scan) == EXACTF
2507 ? ibcmp(s, locinput, ln)
2508 : ibcmp_locale(s, locinput, ln)))
2511 nextchr = UCHARAT(locinput);
2515 STRLEN inclasslen = PL_regeol - locinput;
2517 if (!reginclass(scan, (U8*)locinput, &inclasslen, do_utf8))
2519 if (locinput >= PL_regeol)
2521 locinput += inclasslen ? inclasslen : UTF8SKIP(locinput);
2522 nextchr = UCHARAT(locinput);
2527 nextchr = UCHARAT(locinput);
2528 if (!REGINCLASS(scan, (U8*)locinput))
2530 if (!nextchr && locinput >= PL_regeol)
2532 nextchr = UCHARAT(++locinput);
2536 /* If we might have the case of the German sharp s
2537 * in a casefolding Unicode character class. */
2539 if (ANYOF_FOLD_SHARP_S(scan, locinput, PL_regeol)) {
2540 locinput += SHARP_S_SKIP;
2541 nextchr = UCHARAT(locinput);
2547 PL_reg_flags |= RF_tainted;
2553 LOAD_UTF8_CHARCLASS(alnum,"a");
2554 if (!(OP(scan) == ALNUM
2555 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2556 : isALNUM_LC_utf8((U8*)locinput)))
2560 locinput += PL_utf8skip[nextchr];
2561 nextchr = UCHARAT(locinput);
2564 if (!(OP(scan) == ALNUM
2565 ? isALNUM(nextchr) : isALNUM_LC(nextchr)))
2567 nextchr = UCHARAT(++locinput);
2570 PL_reg_flags |= RF_tainted;
2573 if (!nextchr && locinput >= PL_regeol)
2576 LOAD_UTF8_CHARCLASS(alnum,"a");
2577 if (OP(scan) == NALNUM
2578 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2579 : isALNUM_LC_utf8((U8*)locinput))
2583 locinput += PL_utf8skip[nextchr];
2584 nextchr = UCHARAT(locinput);
2587 if (OP(scan) == NALNUM
2588 ? isALNUM(nextchr) : isALNUM_LC(nextchr))
2590 nextchr = UCHARAT(++locinput);
2594 PL_reg_flags |= RF_tainted;
2598 /* was last char in word? */
2600 if (locinput == PL_bostr)
2603 U8 *r = reghop3((U8*)locinput, -1, (U8*)PL_bostr);
2605 ln = utf8n_to_uvchr(r, UTF8SKIP(r), 0, 0);
2607 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2608 ln = isALNUM_uni(ln);
2609 LOAD_UTF8_CHARCLASS(alnum,"a");
2610 n = swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8);
2613 ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(ln));
2614 n = isALNUM_LC_utf8((U8*)locinput);
2618 ln = (locinput != PL_bostr) ?
2619 UCHARAT(locinput - 1) : '\n';
2620 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2622 n = isALNUM(nextchr);
2625 ln = isALNUM_LC(ln);
2626 n = isALNUM_LC(nextchr);
2629 if (((!ln) == (!n)) == (OP(scan) == BOUND ||
2630 OP(scan) == BOUNDL))
2634 PL_reg_flags |= RF_tainted;
2640 if (UTF8_IS_CONTINUED(nextchr)) {
2641 LOAD_UTF8_CHARCLASS(space," ");
2642 if (!(OP(scan) == SPACE
2643 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2644 : isSPACE_LC_utf8((U8*)locinput)))
2648 locinput += PL_utf8skip[nextchr];
2649 nextchr = UCHARAT(locinput);
2652 if (!(OP(scan) == SPACE
2653 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2655 nextchr = UCHARAT(++locinput);
2658 if (!(OP(scan) == SPACE
2659 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2661 nextchr = UCHARAT(++locinput);
2665 PL_reg_flags |= RF_tainted;
2668 if (!nextchr && locinput >= PL_regeol)
2671 LOAD_UTF8_CHARCLASS(space," ");
2672 if (OP(scan) == NSPACE
2673 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2674 : isSPACE_LC_utf8((U8*)locinput))
2678 locinput += PL_utf8skip[nextchr];
2679 nextchr = UCHARAT(locinput);
2682 if (OP(scan) == NSPACE
2683 ? isSPACE(nextchr) : isSPACE_LC(nextchr))
2685 nextchr = UCHARAT(++locinput);
2688 PL_reg_flags |= RF_tainted;
2694 LOAD_UTF8_CHARCLASS(digit,"0");
2695 if (!(OP(scan) == DIGIT
2696 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2697 : isDIGIT_LC_utf8((U8*)locinput)))
2701 locinput += PL_utf8skip[nextchr];
2702 nextchr = UCHARAT(locinput);
2705 if (!(OP(scan) == DIGIT
2706 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr)))
2708 nextchr = UCHARAT(++locinput);
2711 PL_reg_flags |= RF_tainted;
2714 if (!nextchr && locinput >= PL_regeol)
2717 LOAD_UTF8_CHARCLASS(digit,"0");
2718 if (OP(scan) == NDIGIT
2719 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2720 : isDIGIT_LC_utf8((U8*)locinput))
2724 locinput += PL_utf8skip[nextchr];
2725 nextchr = UCHARAT(locinput);
2728 if (OP(scan) == NDIGIT
2729 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr))
2731 nextchr = UCHARAT(++locinput);
2734 if (locinput >= PL_regeol)
2737 LOAD_UTF8_CHARCLASS(mark,"~");
2738 if (swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2740 locinput += PL_utf8skip[nextchr];
2741 while (locinput < PL_regeol &&
2742 swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2743 locinput += UTF8SKIP(locinput);
2744 if (locinput > PL_regeol)
2749 nextchr = UCHARAT(locinput);
2752 PL_reg_flags |= RF_tainted;
2756 n = ARG(scan); /* which paren pair */
2757 ln = PL_regstartp[n];
2758 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2759 if ((I32)*PL_reglastparen < n || ln == -1)
2760 sayNO; /* Do not match unless seen CLOSEn. */
2761 if (ln == PL_regendp[n])
2765 if (do_utf8 && OP(scan) != REF) { /* REF can do byte comparison */
2767 char *e = PL_bostr + PL_regendp[n];
2769 * Note that we can't do the "other character" lookup trick as
2770 * in the 8-bit case (no pun intended) because in Unicode we
2771 * have to map both upper and title case to lower case.
2773 if (OP(scan) == REFF) {
2774 STRLEN ulen1, ulen2;
2775 U8 tmpbuf1[UTF8_MAXLEN_UCLC+1];
2776 U8 tmpbuf2[UTF8_MAXLEN_UCLC+1];
2780 toLOWER_utf8((U8*)s, tmpbuf1, &ulen1);
2781 toLOWER_utf8((U8*)l, tmpbuf2, &ulen2);
2782 if (ulen1 != ulen2 || memNE((char *)tmpbuf1, (char *)tmpbuf2, ulen1))
2789 nextchr = UCHARAT(locinput);
2793 /* Inline the first character, for speed. */
2794 if (UCHARAT(s) != nextchr &&
2796 (UCHARAT(s) != ((OP(scan) == REFF
2797 ? PL_fold : PL_fold_locale)[nextchr]))))
2799 ln = PL_regendp[n] - ln;
2800 if (locinput + ln > PL_regeol)
2802 if (ln > 1 && (OP(scan) == REF
2803 ? memNE(s, locinput, ln)
2805 ? ibcmp(s, locinput, ln)
2806 : ibcmp_locale(s, locinput, ln))))
2809 nextchr = UCHARAT(locinput);
2820 OP_4tree *oop = PL_op;
2821 COP *ocurcop = PL_curcop;
2824 struct regexp *oreg = PL_reg_re;
2827 PL_op = (OP_4tree*)PL_regdata->data[n];
2828 DEBUG_r( PerlIO_printf(Perl_debug_log, " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) );
2829 PAD_SAVE_LOCAL(old_comppad, (PAD*)PL_regdata->data[n + 2]);
2830 PL_regendp[0] = PL_reg_magic->mg_len = locinput - PL_bostr;
2834 CALLRUNOPS(aTHX); /* Scalar context. */
2837 ret = &PL_sv_undef; /* protect against empty (?{}) blocks. */
2845 PAD_RESTORE_LOCAL(old_comppad);
2846 PL_curcop = ocurcop;
2848 if (logical == 2) { /* Postponed subexpression. */
2850 MAGIC *mg = Null(MAGIC*);
2852 CHECKPOINT cp, lastcp;
2856 if(SvROK(ret) && SvSMAGICAL(sv = SvRV(ret)))
2857 mg = mg_find(sv, PERL_MAGIC_qr);
2858 else if (SvSMAGICAL(ret)) {
2859 if (SvGMAGICAL(ret))
2860 sv_unmagic(ret, PERL_MAGIC_qr);
2862 mg = mg_find(ret, PERL_MAGIC_qr);
2866 re = (regexp *)mg->mg_obj;
2867 (void)ReREFCNT_inc(re);
2871 char *t = SvPV(ret, len);
2873 char *oprecomp = PL_regprecomp;
2874 I32 osize = PL_regsize;
2875 I32 onpar = PL_regnpar;
2878 if (DO_UTF8(ret)) pm.op_pmdynflags |= PMdf_DYN_UTF8;
2879 re = CALLREGCOMP(aTHX_ t, t + len, &pm);
2881 & (SVs_TEMP | SVs_PADTMP | SVf_READONLY
2883 sv_magic(ret,(SV*)ReREFCNT_inc(re),
2885 PL_regprecomp = oprecomp;
2890 PerlIO_printf(Perl_debug_log,
2891 "Entering embedded `%s%.60s%s%s'\n",
2895 (strlen(re->precomp) > 60 ? "..." : ""))
2898 state.prev = PL_reg_call_cc;
2899 state.cc = PL_regcc;
2900 state.re = PL_reg_re;
2904 cp = regcppush(0); /* Save *all* the positions. */
2907 state.ss = PL_savestack_ix;
2908 *PL_reglastparen = 0;
2909 *PL_reglastcloseparen = 0;
2910 PL_reg_call_cc = &state;
2911 PL_reginput = locinput;
2912 toggleutf = ((PL_reg_flags & RF_utf8) != 0) ^
2913 ((re->reganch & ROPT_UTF8) != 0);
2914 if (toggleutf) PL_reg_flags ^= RF_utf8;
2916 /* XXXX This is too dramatic a measure... */
2919 if (regmatch(re->program + 1)) {
2920 /* Even though we succeeded, we need to restore
2921 global variables, since we may be wrapped inside
2922 SUSPEND, thus the match may be not finished yet. */
2924 /* XXXX Do this only if SUSPENDed? */
2925 PL_reg_call_cc = state.prev;
2926 PL_regcc = state.cc;
2927 PL_reg_re = state.re;
2928 cache_re(PL_reg_re);
2929 if (toggleutf) PL_reg_flags ^= RF_utf8;
2931 /* XXXX This is too dramatic a measure... */
2934 /* These are needed even if not SUSPEND. */
2940 REGCP_UNWIND(lastcp);
2942 PL_reg_call_cc = state.prev;
2943 PL_regcc = state.cc;
2944 PL_reg_re = state.re;
2945 cache_re(PL_reg_re);
2946 if (toggleutf) PL_reg_flags ^= RF_utf8;
2948 /* XXXX This is too dramatic a measure... */
2958 sv_setsv(save_scalar(PL_replgv), ret);
2964 n = ARG(scan); /* which paren pair */
2965 PL_reg_start_tmp[n] = locinput;
2970 n = ARG(scan); /* which paren pair */
2971 PL_regstartp[n] = PL_reg_start_tmp[n] - PL_bostr;
2972 PL_regendp[n] = locinput - PL_bostr;
2973 if (n > (I32)*PL_reglastparen)
2974 *PL_reglastparen = n;
2975 *PL_reglastcloseparen = n;
2978 n = ARG(scan); /* which paren pair */
2979 sw = ((I32)*PL_reglastparen >= n && PL_regendp[n] != -1);
2982 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2984 next = NEXTOPER(NEXTOPER(scan));
2986 next = scan + ARG(scan);
2987 if (OP(next) == IFTHEN) /* Fake one. */
2988 next = NEXTOPER(NEXTOPER(next));
2992 logical = scan->flags;
2994 /*******************************************************************
2995 PL_regcc contains infoblock about the innermost (...)* loop, and
2996 a pointer to the next outer infoblock.
2998 Here is how Y(A)*Z is processed (if it is compiled into CURLYX/WHILEM):
3000 1) After matching X, regnode for CURLYX is processed;
3002 2) This regnode creates infoblock on the stack, and calls
3003 regmatch() recursively with the starting point at WHILEM node;
3005 3) Each hit of WHILEM node tries to match A and Z (in the order
3006 depending on the current iteration, min/max of {min,max} and
3007 greediness). The information about where are nodes for "A"
3008 and "Z" is read from the infoblock, as is info on how many times "A"
3009 was already matched, and greediness.
3011 4) After A matches, the same WHILEM node is hit again.
3013 5) Each time WHILEM is hit, PL_regcc is the infoblock created by CURLYX
3014 of the same pair. Thus when WHILEM tries to match Z, it temporarily
3015 resets PL_regcc, since this Y(A)*Z can be a part of some other loop:
3016 as in (Y(A)*Z)*. If Z matches, the automaton will hit the WHILEM node
3017 of the external loop.
3019 Currently present infoblocks form a tree with a stem formed by PL_curcc
3020 and whatever it mentions via ->next, and additional attached trees
3021 corresponding to temporarily unset infoblocks as in "5" above.
3023 In the following picture infoblocks for outer loop of
3024 (Y(A)*?Z)*?T are denoted O, for inner I. NULL starting block
3025 is denoted by x. The matched string is YAAZYAZT. Temporarily postponed
3026 infoblocks are drawn below the "reset" infoblock.
3028 In fact in the picture below we do not show failed matches for Z and T
3029 by WHILEM blocks. [We illustrate minimal matches, since for them it is
3030 more obvious *why* one needs to *temporary* unset infoblocks.]
3032 Matched REx position InfoBlocks Comment
3036 Y A)*?Z)*?T x <- O <- I
3037 YA )*?Z)*?T x <- O <- I
3038 YA A)*?Z)*?T x <- O <- I
3039 YAA )*?Z)*?T x <- O <- I
3040 YAA Z)*?T x <- O # Temporary unset I
3043 YAAZ Y(A)*?Z)*?T x <- O
3046 YAAZY (A)*?Z)*?T x <- O
3049 YAAZY A)*?Z)*?T x <- O <- I
3052 YAAZYA )*?Z)*?T x <- O <- I
3055 YAAZYA Z)*?T x <- O # Temporary unset I
3061 YAAZYAZ T x # Temporary unset O
3068 *******************************************************************/
3071 CHECKPOINT cp = PL_savestack_ix;
3072 /* No need to save/restore up to this paren */
3073 I32 parenfloor = scan->flags;
3075 if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */
3077 cc.oldcc = PL_regcc;
3079 /* XXXX Probably it is better to teach regpush to support
3080 parenfloor > PL_regsize... */
3081 if (parenfloor > (I32)*PL_reglastparen)
3082 parenfloor = *PL_reglastparen; /* Pessimization... */
3083 cc.parenfloor = parenfloor;
3085 cc.min = ARG1(scan);
3086 cc.max = ARG2(scan);
3087 cc.scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
3091 PL_reginput = locinput;
3092 n = regmatch(PREVOPER(next)); /* start on the WHILEM */
3094 PL_regcc = cc.oldcc;
3100 * This is really hard to understand, because after we match
3101 * what we're trying to match, we must make sure the rest of
3102 * the REx is going to match for sure, and to do that we have
3103 * to go back UP the parse tree by recursing ever deeper. And
3104 * if it fails, we have to reset our parent's current state
3105 * that we can try again after backing off.
3108 CHECKPOINT cp, lastcp;
3109 CURCUR* cc = PL_regcc;
3110 char *lastloc = cc->lastloc; /* Detection of 0-len. */
3112 n = cc->cur + 1; /* how many we know we matched */
3113 PL_reginput = locinput;
3116 PerlIO_printf(Perl_debug_log,
3117 "%*s %ld out of %ld..%ld cc=%"UVxf"\n",
3118 REPORT_CODE_OFF+PL_regindent*2, "",
3119 (long)n, (long)cc->min,
3120 (long)cc->max, PTR2UV(cc))
3123 /* If degenerate scan matches "", assume scan done. */
3125 if (locinput == cc->lastloc && n >= cc->min) {
3126 PL_regcc = cc->oldcc;
3130 PerlIO_printf(Perl_debug_log,
3131 "%*s empty match detected, try continuation...\n",
3132 REPORT_CODE_OFF+PL_regindent*2, "")
3134 if (regmatch(cc->next))
3142 /* First just match a string of min scans. */
3146 cc->lastloc = locinput;
3147 if (regmatch(cc->scan))
3150 cc->lastloc = lastloc;
3155 /* Check whether we already were at this position.
3156 Postpone detection until we know the match is not
3157 *that* much linear. */
3158 if (!PL_reg_maxiter) {
3159 PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4);
3160 PL_reg_leftiter = PL_reg_maxiter;
3162 if (PL_reg_leftiter-- == 0) {
3163 I32 size = (PL_reg_maxiter + 7)/8;
3164 if (PL_reg_poscache) {
3165 if ((I32)PL_reg_poscache_size < size) {
3166 Renew(PL_reg_poscache, size, char);
3167 PL_reg_poscache_size = size;
3169 Zero(PL_reg_poscache, size, char);
3172 PL_reg_poscache_size = size;
3173 Newz(29, PL_reg_poscache, size, char);
3176 PerlIO_printf(Perl_debug_log,
3177 "%sDetected a super-linear match, switching on caching%s...\n",
3178 PL_colors[4], PL_colors[5])
3181 if (PL_reg_leftiter < 0) {
3182 I32 o = locinput - PL_bostr, b;
3184 o = (scan->flags & 0xf) - 1 + o * (scan->flags>>4);
3187 if (PL_reg_poscache[o] & (1<<b)) {
3189 PerlIO_printf(Perl_debug_log,
3190 "%*s already tried at this position...\n",
3191 REPORT_CODE_OFF+PL_regindent*2, "")
3195 PL_reg_poscache[o] |= (1<<b);
3199 /* Prefer next over scan for minimal matching. */
3202 PL_regcc = cc->oldcc;
3205 cp = regcppush(cc->parenfloor);
3207 if (regmatch(cc->next)) {
3209 sayYES; /* All done. */
3211 REGCP_UNWIND(lastcp);
3217 if (n >= cc->max) { /* Maximum greed exceeded? */
3218 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3219 && !(PL_reg_flags & RF_warned)) {
3220 PL_reg_flags |= RF_warned;
3221 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s limit (%d) exceeded",
3222 "Complex regular subexpression recursion",
3229 PerlIO_printf(Perl_debug_log,
3230 "%*s trying longer...\n",
3231 REPORT_CODE_OFF+PL_regindent*2, "")
3233 /* Try scanning more and see if it helps. */
3234 PL_reginput = locinput;
3236 cc->lastloc = locinput;
3237 cp = regcppush(cc->parenfloor);
3239 if (regmatch(cc->scan)) {
3243 REGCP_UNWIND(lastcp);
3246 cc->lastloc = lastloc;
3250 /* Prefer scan over next for maximal matching. */
3252 if (n < cc->max) { /* More greed allowed? */
3253 cp = regcppush(cc->parenfloor);
3255 cc->lastloc = locinput;
3257 if (regmatch(cc->scan)) {
3261 REGCP_UNWIND(lastcp);
3262 regcppop(); /* Restore some previous $<digit>s? */
3263 PL_reginput = locinput;
3265 PerlIO_printf(Perl_debug_log,
3266 "%*s failed, try continuation...\n",
3267 REPORT_CODE_OFF+PL_regindent*2, "")
3270 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3271 && !(PL_reg_flags & RF_warned)) {
3272 PL_reg_flags |= RF_warned;
3273 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s limit (%d) exceeded",
3274 "Complex regular subexpression recursion",
3278 /* Failed deeper matches of scan, so see if this one works. */
3279 PL_regcc = cc->oldcc;
3282 if (regmatch(cc->next))
3288 cc->lastloc = lastloc;
3293 next = scan + ARG(scan);
3296 inner = NEXTOPER(NEXTOPER(scan));
3299 inner = NEXTOPER(scan);
3303 if (OP(next) != c1) /* No choice. */
3304 next = inner; /* Avoid recursion. */
3306 I32 lastparen = *PL_reglastparen;
3308 re_unwind_branch_t *uw;
3310 /* Put unwinding data on stack */
3311 unwind1 = SSNEWt(1,re_unwind_branch_t);
3312 uw = SSPTRt(unwind1,re_unwind_branch_t);
3315 uw->type = ((c1 == BRANCH)
3317 : RE_UNWIND_BRANCHJ);
3318 uw->lastparen = lastparen;
3320 uw->locinput = locinput;
3321 uw->nextchr = nextchr;
3323 uw->regindent = ++PL_regindent;
3326 REGCP_SET(uw->lastcp);
3328 /* Now go into the first branch */
3341 /* We suppose that the next guy does not need
3342 backtracking: in particular, it is of constant length,
3343 and has no parenths to influence future backrefs. */
3344 ln = ARG1(scan); /* min to match */
3345 n = ARG2(scan); /* max to match */
3346 paren = scan->flags;
3348 if (paren > PL_regsize)
3350 if (paren > (I32)*PL_reglastparen)
3351 *PL_reglastparen = paren;
3353 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3355 scan += NEXT_OFF(scan); /* Skip former OPEN. */
3356 PL_reginput = locinput;
3359 if (ln && regrepeat_hard(scan, ln, &l) < ln)
3361 /* if we matched something zero-length we don't need to
3362 backtrack - capturing parens are already defined, so
3363 the caveat in the maximal case doesn't apply
3365 XXXX if ln == 0, we can redo this check first time
3366 through the following loop
3369 n = ln; /* don't backtrack */
3370 locinput = PL_reginput;
3371 if (HAS_TEXT(next) || JUMPABLE(next)) {
3372 regnode *text_node = next;
3374 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3376 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3378 if (PL_regkind[(U8)OP(text_node)] == REF) {
3380 n = ARG(text_node); /* which paren pair */
3381 ln = PL_regstartp[n];
3382 /* assume yes if we haven't seen CLOSEn */
3384 (I32)*PL_reglastparen < n ||
3391 c1 = *(PL_bostr + ln);
3393 else { c1 = (U8)*STRING(text_node); }
3394 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3396 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3397 c2 = PL_fold_locale[c1];
3406 /* This may be improved if l == 0. */
3407 while (n >= ln || (n == REG_INFTY && ln > 0 && l)) { /* ln overflow ? */
3408 /* If it could work, try it. */
3410 UCHARAT(PL_reginput) == c1 ||
3411 UCHARAT(PL_reginput) == c2)
3415 PL_regstartp[paren] =
3416 HOPc(PL_reginput, -l) - PL_bostr;
3417 PL_regendp[paren] = PL_reginput - PL_bostr;
3420 PL_regendp[paren] = -1;
3424 REGCP_UNWIND(lastcp);
3426 /* Couldn't or didn't -- move forward. */
3427 PL_reginput = locinput;
3428 if (regrepeat_hard(scan, 1, &l)) {
3430 locinput = PL_reginput;
3437 n = regrepeat_hard(scan, n, &l);
3438 /* if we matched something zero-length we don't need to
3439 backtrack, unless the minimum count is zero and we
3440 are capturing the result - in that case the capture
3441 being defined or not may affect later execution
3443 if (n != 0 && l == 0 && !(paren && ln == 0))
3444 ln = n; /* don't backtrack */
3445 locinput = PL_reginput;
3447 PerlIO_printf(Perl_debug_log,
3448 "%*s matched %"IVdf" times, len=%"IVdf"...\n",
3449 (int)(REPORT_CODE_OFF+PL_regindent*2), "",
3453 if (HAS_TEXT(next) || JUMPABLE(next)) {
3454 regnode *text_node = next;
3456 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3458 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3460 if (PL_regkind[(U8)OP(text_node)] == REF) {
3462 n = ARG(text_node); /* which paren pair */
3463 ln = PL_regstartp[n];
3464 /* assume yes if we haven't seen CLOSEn */
3466 (I32)*PL_reglastparen < n ||
3473 c1 = *(PL_bostr + ln);
3475 else { c1 = (U8)*STRING(text_node); }
3477 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3479 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3480 c2 = PL_fold_locale[c1];
3491 /* If it could work, try it. */
3493 UCHARAT(PL_reginput) == c1 ||
3494 UCHARAT(PL_reginput) == c2)
3497 PerlIO_printf(Perl_debug_log,
3498 "%*s trying tail with n=%"IVdf"...\n",
3499 (int)(REPORT_CODE_OFF+PL_regindent*2), "", (IV)n)
3503 PL_regstartp[paren] = HOPc(PL_reginput, -l) - PL_bostr;
3504 PL_regendp[paren] = PL_reginput - PL_bostr;
3507 PL_regendp[paren] = -1;
3511 REGCP_UNWIND(lastcp);
3513 /* Couldn't or didn't -- back up. */
3515 locinput = HOPc(locinput, -l);
3516 PL_reginput = locinput;
3523 paren = scan->flags; /* Which paren to set */
3524 if (paren > PL_regsize)
3526 if (paren > (I32)*PL_reglastparen)
3527 *PL_reglastparen = paren;
3528 ln = ARG1(scan); /* min to match */
3529 n = ARG2(scan); /* max to match */
3530 scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE);
3534 ln = ARG1(scan); /* min to match */
3535 n = ARG2(scan); /* max to match */
3536 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3541 scan = NEXTOPER(scan);
3547 scan = NEXTOPER(scan);
3551 * Lookahead to avoid useless match attempts
3552 * when we know what character comes next.
3556 * Used to only do .*x and .*?x, but now it allows
3557 * for )'s, ('s and (?{ ... })'s to be in the way
3558 * of the quantifier and the EXACT-like node. -- japhy
3561 if (HAS_TEXT(next) || JUMPABLE(next)) {
3563 regnode *text_node = next;
3565 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3567 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3569 if (PL_regkind[(U8)OP(text_node)] == REF) {
3571 n = ARG(text_node); /* which paren pair */
3572 ln = PL_regstartp[n];
3573 /* assume yes if we haven't seen CLOSEn */
3575 (I32)*PL_reglastparen < n ||
3580 goto assume_ok_easy;
3582 s = (U8*)PL_bostr + ln;
3584 else { s = (U8*)STRING(text_node); }
3588 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3590 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3591 c2 = PL_fold_locale[c1];
3594 if (OP(text_node) == EXACTF || OP(text_node) == REFF) {
3595 STRLEN ulen1, ulen2;
3596 U8 tmpbuf1[UTF8_MAXLEN_UCLC+1];
3597 U8 tmpbuf2[UTF8_MAXLEN_UCLC+1];
3599 to_utf8_lower((U8*)s, tmpbuf1, &ulen1);
3600 to_utf8_upper((U8*)s, tmpbuf2, &ulen2);
3602 c1 = utf8n_to_uvuni(tmpbuf1, UTF8_MAXLEN, 0,
3604 0 : UTF8_ALLOW_ANY);
3605 c2 = utf8n_to_uvuni(tmpbuf2, UTF8_MAXLEN, 0,
3607 0 : UTF8_ALLOW_ANY);
3610 c2 = c1 = utf8n_to_uvchr(s, UTF8_MAXLEN, 0,
3612 0 : UTF8_ALLOW_ANY);
3620 PL_reginput = locinput;
3624 if (ln && regrepeat(scan, ln) < ln)
3626 locinput = PL_reginput;
3629 char *e; /* Should not check after this */
3630 char *old = locinput;
3633 if (n == REG_INFTY) {
3636 while (UTF8_IS_CONTINUATION(*(U8*)e))
3642 m >0 && e + UTF8SKIP(e) <= PL_regeol; m--)
3646 e = locinput + n - ln;
3651 /* Find place 'next' could work */
3654 while (locinput <= e &&
3655 UCHARAT(locinput) != c1)
3658 while (locinput <= e
3659 && UCHARAT(locinput) != c1
3660 && UCHARAT(locinput) != c2)
3663 count = locinput - old;
3668 /* count initialised to
3669 * utf8_distance(old, locinput) */
3670 while (locinput <= e &&
3671 utf8n_to_uvchr((U8*)locinput,
3674 0 : UTF8_ALLOW_ANY) != (UV)c1) {
3679 /* count initialised to
3680 * utf8_distance(old, locinput) */
3681 while (locinput <= e) {
3682 UV c = utf8n_to_uvchr((U8*)locinput,
3685 0 : UTF8_ALLOW_ANY);
3686 if (c == (UV)c1 || c == (UV)c2)
3695 /* PL_reginput == old now */
3696 if (locinput != old) {
3697 ln = 1; /* Did some */
3698 if (regrepeat(scan, count) < count)
3701 /* PL_reginput == locinput now */
3702 TRYPAREN(paren, ln, locinput);
3703 PL_reginput = locinput; /* Could be reset... */
3704 REGCP_UNWIND(lastcp);
3705 /* Couldn't or didn't -- move forward. */
3708 locinput += UTF8SKIP(locinput);
3715 while (n >= ln || (n == REG_INFTY && ln > 0)) { /* ln overflow ? */
3719 c = utf8n_to_uvchr((U8*)PL_reginput,
3722 0 : UTF8_ALLOW_ANY);
3724 c = UCHARAT(PL_reginput);
3725 /* If it could work, try it. */
3726 if (c == (UV)c1 || c == (UV)c2)
3728 TRYPAREN(paren, ln, PL_reginput);
3729 REGCP_UNWIND(lastcp);
3732 /* If it could work, try it. */
3733 else if (c1 == -1000)
3735 TRYPAREN(paren, ln, PL_reginput);
3736 REGCP_UNWIND(lastcp);
3738 /* Couldn't or didn't -- move forward. */
3739 PL_reginput = locinput;
3740 if (regrepeat(scan, 1)) {
3742 locinput = PL_reginput;
3750 n = regrepeat(scan, n);
3751 locinput = PL_reginput;
3752 if (ln < n && PL_regkind[(U8)OP(next)] == EOL &&
3753 ((!PL_multiline && OP(next) != MEOL) ||
3754 OP(next) == SEOL || OP(next) == EOS))
3756 ln = n; /* why back off? */
3757 /* ...because $ and \Z can match before *and* after
3758 newline at the end. Consider "\n\n" =~ /\n+\Z\n/.
3759 We should back off by one in this case. */
3760 if (UCHARAT(PL_reginput - 1) == '\n' && OP(next) != EOS)
3769 c = utf8n_to_uvchr((U8*)PL_reginput,
3772 0 : UTF8_ALLOW_ANY);
3774 c = UCHARAT(PL_reginput);
3776 /* If it could work, try it. */
3777 if (c1 == -1000 || c == (UV)c1 || c == (UV)c2)
3779 TRYPAREN(paren, n, PL_reginput);
3780 REGCP_UNWIND(lastcp);
3782 /* Couldn't or didn't -- back up. */
3784 PL_reginput = locinput = HOPc(locinput, -1);
3792 c = utf8n_to_uvchr((U8*)PL_reginput,
3795 0 : UTF8_ALLOW_ANY);
3797 c = UCHARAT(PL_reginput);
3799 /* If it could work, try it. */
3800 if (c1 == -1000 || c == (UV)c1 || c == (UV)c2)
3802 TRYPAREN(paren, n, PL_reginput);
3803 REGCP_UNWIND(lastcp);
3805 /* Couldn't or didn't -- back up. */
3807 PL_reginput = locinput = HOPc(locinput, -1);
3814 if (PL_reg_call_cc) {
3815 re_cc_state *cur_call_cc = PL_reg_call_cc;
3816 CURCUR *cctmp = PL_regcc;
3817 regexp *re = PL_reg_re;
3818 CHECKPOINT cp, lastcp;
3820 cp = regcppush(0); /* Save *all* the positions. */
3822 regcp_set_to(PL_reg_call_cc->ss); /* Restore parens of
3824 PL_reginput = locinput; /* Make position available to
3826 cache_re(PL_reg_call_cc->re);
3827 PL_regcc = PL_reg_call_cc->cc;
3828 PL_reg_call_cc = PL_reg_call_cc->prev;
3829 if (regmatch(cur_call_cc->node)) {
3830 PL_reg_call_cc = cur_call_cc;
3834 REGCP_UNWIND(lastcp);
3836 PL_reg_call_cc = cur_call_cc;
3842 PerlIO_printf(Perl_debug_log,
3843 "%*s continuation failed...\n",
3844 REPORT_CODE_OFF+PL_regindent*2, "")
3848 if (locinput < PL_regtill) {
3849 DEBUG_r(PerlIO_printf(Perl_debug_log,
3850 "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n",
3852 (long)(locinput - PL_reg_starttry),
3853 (long)(PL_regtill - PL_reg_starttry),
3855 sayNO_FINAL; /* Cannot match: too short. */
3857 PL_reginput = locinput; /* put where regtry can find it */
3858 sayYES_FINAL; /* Success! */
3860 PL_reginput = locinput; /* put where regtry can find it */
3861 sayYES_LOUD; /* Success! */
3864 PL_reginput = locinput;
3869 s = HOPBACKc(locinput, scan->flags);
3875 PL_reginput = locinput;
3880 s = HOPBACKc(locinput, scan->flags);
3886 PL_reginput = locinput;
3889 inner = NEXTOPER(NEXTOPER(scan));
3890 if (regmatch(inner) != n) {
3905 if (OP(scan) == SUSPEND) {
3906 locinput = PL_reginput;
3907 nextchr = UCHARAT(locinput);
3912 next = scan + ARG(scan);
3917 PerlIO_printf(Perl_error_log, "%"UVxf" %d\n",
3918 PTR2UV(scan), OP(scan));
3919 Perl_croak(aTHX_ "regexp memory corruption");
3926 * We get here only if there's trouble -- normally "case END" is
3927 * the terminating point.
3929 Perl_croak(aTHX_ "corrupted regexp pointers");
3935 PerlIO_printf(Perl_debug_log,
3936 "%*s %scould match...%s\n",
3937 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4],PL_colors[5])
3941 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n",
3942 PL_colors[4],PL_colors[5]));
3948 #if 0 /* Breaks $^R */
3956 PerlIO_printf(Perl_debug_log,
3957 "%*s %sfailed...%s\n",
3958 REPORT_CODE_OFF+PL_regindent*2, "",PL_colors[4],PL_colors[5])
3964 re_unwind_t *uw = SSPTRt(unwind,re_unwind_t);
3967 case RE_UNWIND_BRANCH:
3968 case RE_UNWIND_BRANCHJ:
3970 re_unwind_branch_t *uwb = &(uw->branch);
3971 I32 lastparen = uwb->lastparen;
3973 REGCP_UNWIND(uwb->lastcp);
3974 for (n = *PL_reglastparen; n > lastparen; n--)
3976 *PL_reglastparen = n;
3977 scan = next = uwb->next;
3979 OP(scan) != (uwb->type == RE_UNWIND_BRANCH
3980 ? BRANCH : BRANCHJ) ) { /* Failure */
3987 /* Have more choice yet. Reuse the same uwb. */
3989 if ((n = (uwb->type == RE_UNWIND_BRANCH
3990 ? NEXT_OFF(next) : ARG(next))))
3993 next = NULL; /* XXXX Needn't unwinding in this case... */
3995 next = NEXTOPER(scan);
3996 if (uwb->type == RE_UNWIND_BRANCHJ)
3997 next = NEXTOPER(next);
3998 locinput = uwb->locinput;
3999 nextchr = uwb->nextchr;
4001 PL_regindent = uwb->regindent;
4008 Perl_croak(aTHX_ "regexp unwind memory corruption");
4019 - regrepeat - repeatedly match something simple, report how many
4022 * [This routine now assumes that it will only match on things of length 1.
4023 * That was true before, but now we assume scan - reginput is the count,
4024 * rather than incrementing count on every character. [Er, except utf8.]]
4027 S_regrepeat(pTHX_ regnode *p, I32 max)
4029 register char *scan;
4031 register char *loceol = PL_regeol;
4032 register I32 hardcount = 0;
4033 register bool do_utf8 = PL_reg_match_utf8;
4036 if (max == REG_INFTY)
4038 else if (max < loceol - scan)
4039 loceol = scan + max;
4044 while (scan < loceol && hardcount < max && *scan != '\n') {
4045 scan += UTF8SKIP(scan);
4049 while (scan < loceol && *scan != '\n')
4056 while (scan < loceol && hardcount < max) {
4057 scan += UTF8SKIP(scan);
4067 case EXACT: /* length of string is 1 */
4069 while (scan < loceol && UCHARAT(scan) == c)
4072 case EXACTF: /* length of string is 1 */
4074 while (scan < loceol &&
4075 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold[c]))
4078 case EXACTFL: /* length of string is 1 */
4079 PL_reg_flags |= RF_tainted;
4081 while (scan < loceol &&
4082 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold_locale[c]))
4088 while (hardcount < max && scan < loceol &&
4089 reginclass(p, (U8*)scan, 0, do_utf8)) {
4090 scan += UTF8SKIP(scan);
4094 while (scan < loceol && REGINCLASS(p, (U8*)scan))
4101 LOAD_UTF8_CHARCLASS(alnum,"a");
4102 while (hardcount < max && scan < loceol &&
4103 swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
4104 scan += UTF8SKIP(scan);
4108 while (scan < loceol && isALNUM(*scan))
4113 PL_reg_flags |= RF_tainted;
4116 while (hardcount < max && scan < loceol &&
4117 isALNUM_LC_utf8((U8*)scan)) {
4118 scan += UTF8SKIP(scan);
4122 while (scan < loceol && isALNUM_LC(*scan))
4129 LOAD_UTF8_CHARCLASS(alnum,"a");
4130 while (hardcount < max && scan < loceol &&
4131 !swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
4132 scan += UTF8SKIP(scan);
4136 while (scan < loceol && !isALNUM(*scan))
4141 PL_reg_flags |= RF_tainted;
4144 while (hardcount < max && scan < loceol &&
4145 !isALNUM_LC_utf8((U8*)scan)) {
4146 scan += UTF8SKIP(scan);
4150 while (scan < loceol && !isALNUM_LC(*scan))
4157 LOAD_UTF8_CHARCLASS(space," ");
4158 while (hardcount < max && scan < loceol &&
4160 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
4161 scan += UTF8SKIP(scan);
4165 while (scan < loceol && isSPACE(*scan))
4170 PL_reg_flags |= RF_tainted;
4173 while (hardcount < max && scan < loceol &&
4174 (*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
4175 scan += UTF8SKIP(scan);
4179 while (scan < loceol && isSPACE_LC(*scan))
4186 LOAD_UTF8_CHARCLASS(space," ");
4187 while (hardcount < max && scan < loceol &&
4189 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
4190 scan += UTF8SKIP(scan);
4194 while (scan < loceol && !isSPACE(*scan))
4199 PL_reg_flags |= RF_tainted;
4202 while (hardcount < max && scan < loceol &&
4203 !(*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
4204 scan += UTF8SKIP(scan);
4208 while (scan < loceol && !isSPACE_LC(*scan))
4215 LOAD_UTF8_CHARCLASS(digit,"0");
4216 while (hardcount < max && scan < loceol &&
4217 swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
4218 scan += UTF8SKIP(scan);
4222 while (scan < loceol && isDIGIT(*scan))
4229 LOAD_UTF8_CHARCLASS(digit,"0");
4230 while (hardcount < max && scan < loceol &&
4231 !swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
4232 scan += UTF8SKIP(scan);
4236 while (scan < loceol && !isDIGIT(*scan))
4240 default: /* Called on something of 0 width. */
4241 break; /* So match right here or not at all. */
4247 c = scan - PL_reginput;
4252 SV *prop = sv_newmortal();
4255 PerlIO_printf(Perl_debug_log,
4256 "%*s %s can match %"IVdf" times out of %"IVdf"...\n",
4257 REPORT_CODE_OFF+1, "", SvPVX(prop),(IV)c,(IV)max);
4264 - regrepeat_hard - repeatedly match something, report total lenth and length
4266 * The repeater is supposed to have constant length.
4270 S_regrepeat_hard(pTHX_ regnode *p, I32 max, I32 *lp)
4272 register char *scan = Nullch;
4273 register char *start;
4274 register char *loceol = PL_regeol;
4276 I32 count = 0, res = 1;
4281 start = PL_reginput;
4282 if (PL_reg_match_utf8) {
4283 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
4286 while (start < PL_reginput) {
4288 start += UTF8SKIP(start);
4299 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
4301 *lp = l = PL_reginput - start;
4302 if (max != REG_INFTY && l*max < loceol - scan)
4303 loceol = scan + l*max;
4316 - regclass_swash - prepare the utf8 swash
4320 Perl_regclass_swash(pTHX_ register regnode* node, bool doinit, SV** listsvp, SV **altsvp)
4326 if (PL_regdata && PL_regdata->count) {
4329 if (PL_regdata->what[n] == 's') {
4330 SV *rv = (SV*)PL_regdata->data[n];
4331 AV *av = (AV*)SvRV((SV*)rv);
4332 SV **ary = AvARRAY(av);
4335 /* See the end of regcomp.c:S_reglass() for
4336 * documentation of these array elements. */
4339 a = SvTYPE(ary[1]) == SVt_RV ? &ary[1] : 0;
4340 b = SvTYPE(ary[2]) == SVt_PVAV ? &ary[2] : 0;
4344 else if (si && doinit) {
4345 sw = swash_init("utf8", "", si, 1, 0);
4346 (void)av_store(av, 1, sw);
4362 - reginclass - determine if a character falls into a character class
4364 The n is the ANYOF regnode, the p is the target string, lenp
4365 is pointer to the maximum length of how far to go in the p
4366 (if the lenp is zero, UTF8SKIP(p) is used),
4367 do_utf8 tells whether the target string is in UTF-8.
4372 S_reginclass(pTHX_ register regnode *n, register U8* p, STRLEN* lenp, register bool do_utf8)
4374 char flags = ANYOF_FLAGS(n);
4380 if (do_utf8 && !UTF8_IS_INVARIANT(c))
4381 c = utf8n_to_uvchr(p, UTF8_MAXLEN, &len,
4382 ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
4384 plen = lenp ? *lenp : UNISKIP(NATIVE_TO_UNI(c));
4385 if (do_utf8 || (flags & ANYOF_UNICODE)) {
4388 if (do_utf8 && !ANYOF_RUNTIME(n)) {
4389 if (len != (STRLEN)-1 && c < 256 && ANYOF_BITMAP_TEST(n, c))
4392 if (!match && do_utf8 && (flags & ANYOF_UNICODE_ALL) && c >= 256)
4396 SV *sw = regclass_swash(n, TRUE, 0, (SV**)&av);
4399 if (swash_fetch(sw, p, do_utf8))
4401 else if (flags & ANYOF_FOLD) {
4402 if (!match && lenp && av) {
4405 for (i = 0; i <= av_len(av); i++) {
4406 SV* sv = *av_fetch(av, i, FALSE);
4408 char *s = SvPV(sv, len);
4410 if (len <= plen && memEQ(s, (char*)p, len)) {
4418 U8 tmpbuf[UTF8_MAXLEN_FOLD+1];
4421 to_utf8_fold(p, tmpbuf, &tmplen);
4422 if (swash_fetch(sw, tmpbuf, do_utf8))
4428 if (match && lenp && *lenp == 0)
4429 *lenp = UNISKIP(NATIVE_TO_UNI(c));
4431 if (!match && c < 256) {
4432 if (ANYOF_BITMAP_TEST(n, c))
4434 else if (flags & ANYOF_FOLD) {
4437 if (flags & ANYOF_LOCALE) {
4438 PL_reg_flags |= RF_tainted;
4439 f = PL_fold_locale[c];
4443 if (f != c && ANYOF_BITMAP_TEST(n, f))
4447 if (!match && (flags & ANYOF_CLASS)) {
4448 PL_reg_flags |= RF_tainted;
4450 (ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) ||
4451 (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) ||
4452 (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) ||
4453 (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) ||
4454 (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) ||
4455 (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) ||
4456 (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) ||
4457 (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) ||
4458 (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) ||
4459 (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) ||
4460 (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII(c)) ||
4461 (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII(c)) ||
4462 (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) ||
4463 (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) ||
4464 (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) ||
4465 (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) ||
4466 (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) ||
4467 (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) ||
4468 (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) ||
4469 (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) ||
4470 (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) ||
4471 (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) ||
4472 (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) ||
4473 (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) ||
4474 (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) ||
4475 (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) ||
4476 (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) ||
4477 (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) ||
4478 (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK(c)) ||
4479 (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK(c))
4480 ) /* How's that for a conditional? */
4487 return (flags & ANYOF_INVERT) ? !match : match;
4491 S_reghop(pTHX_ U8 *s, I32 off)
4493 return S_reghop3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4497 S_reghop3(pTHX_ U8 *s, I32 off, U8* lim)
4500 while (off-- && s < lim) {
4501 /* XXX could check well-formedness here */
4509 if (UTF8_IS_CONTINUED(*s)) {
4510 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4513 /* XXX could check well-formedness here */
4521 S_reghopmaybe(pTHX_ U8 *s, I32 off)
4523 return S_reghopmaybe3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4527 S_reghopmaybe3(pTHX_ U8* s, I32 off, U8* lim)
4530 while (off-- && s < lim) {
4531 /* XXX could check well-formedness here */
4541 if (UTF8_IS_CONTINUED(*s)) {
4542 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4545 /* XXX could check well-formedness here */
4557 restore_pos(pTHX_ void *arg)
4559 if (PL_reg_eval_set) {
4560 if (PL_reg_oldsaved) {
4561 PL_reg_re->subbeg = PL_reg_oldsaved;
4562 PL_reg_re->sublen = PL_reg_oldsavedlen;
4563 RX_MATCH_COPIED_on(PL_reg_re);
4565 PL_reg_magic->mg_len = PL_reg_oldpos;
4566 PL_reg_eval_set = 0;
4567 PL_curpm = PL_reg_oldcurpm;
4572 S_to_utf8_substr(pTHX_ register regexp *prog)
4575 if (prog->float_substr && !prog->float_utf8) {
4576 prog->float_utf8 = sv = NEWSV(117, 0);
4577 SvSetSV(sv, prog->float_substr);
4578 sv_utf8_upgrade(sv);
4579 if (SvTAIL(prog->float_substr))
4581 if (prog->float_substr == prog->check_substr)
4582 prog->check_utf8 = sv;
4584 if (prog->anchored_substr && !prog->anchored_utf8) {
4585 prog->anchored_utf8 = sv = NEWSV(118, 0);
4586 SvSetSV(sv, prog->anchored_substr);
4587 sv_utf8_upgrade(sv);
4588 if (SvTAIL(prog->anchored_substr))
4590 if (prog->anchored_substr == prog->check_substr)
4591 prog->check_utf8 = sv;
4596 S_to_byte_substr(pTHX_ register regexp *prog)
4599 if (prog->float_utf8 && !prog->float_substr) {
4600 prog->float_substr = sv = NEWSV(117, 0);
4601 SvSetSV(sv, prog->float_utf8);
4602 if (sv_utf8_downgrade(sv, TRUE)) {
4603 if (SvTAIL(prog->float_utf8))
4607 prog->float_substr = sv = &PL_sv_undef;
4609 if (prog->float_utf8 == prog->check_utf8)
4610 prog->check_substr = sv;
4612 if (prog->anchored_utf8 && !prog->anchored_substr) {
4613 prog->anchored_substr = sv = NEWSV(118, 0);
4614 SvSetSV(sv, prog->anchored_utf8);
4615 if (sv_utf8_downgrade(sv, TRUE)) {
4616 if (SvTAIL(prog->anchored_utf8))
4620 prog->anchored_substr = sv = &PL_sv_undef;
4622 if (prog->anchored_utf8 == prog->check_utf8)
4623 prog->check_substr = sv;