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? */
90 #define RF_false 16 /* odd number of nested negatives */
92 #define UTF ((PL_reg_flags & RF_utf8) != 0)
94 #define RS_init 1 /* eval environment created */
95 #define RS_set 2 /* replsv value is set */
101 #define REGINCLASS(p,c) (ANYOF_FLAGS(p) ? reginclass(p,c,0,0) : ANYOF_BITMAP_TEST(p,*(c)))
107 #define CHR_SVLEN(sv) (do_utf8 ? sv_len_utf8(sv) : SvCUR(sv))
108 #define CHR_DIST(a,b) (PL_reg_match_utf8 ? utf8_distance(a,b) : a - b)
110 #define reghop_c(pos,off) ((char*)reghop((U8*)pos, off))
111 #define reghopmaybe_c(pos,off) ((char*)reghopmaybe((U8*)pos, off))
112 #define HOP(pos,off) (PL_reg_match_utf8 ? reghop((U8*)pos, off) : (U8*)(pos + off))
113 #define HOPMAYBE(pos,off) (PL_reg_match_utf8 ? reghopmaybe((U8*)pos, off) : (U8*)(pos + off))
114 #define HOPc(pos,off) ((char*)HOP(pos,off))
115 #define HOPMAYBEc(pos,off) ((char*)HOPMAYBE(pos,off))
117 #define HOPBACK(pos, off) ( \
118 (PL_reg_match_utf8) \
119 ? reghopmaybe((U8*)pos, -off) \
120 : (pos - off >= PL_bostr) \
124 #define HOPBACKc(pos, off) (char*)HOPBACK(pos, off)
126 #define reghop3_c(pos,off,lim) ((char*)reghop3((U8*)pos, off, (U8*)lim))
127 #define reghopmaybe3_c(pos,off,lim) ((char*)reghopmaybe3((U8*)pos, off, (U8*)lim))
128 #define HOP3(pos,off,lim) (PL_reg_match_utf8 ? reghop3((U8*)pos, off, (U8*)lim) : (U8*)(pos + off))
129 #define HOPMAYBE3(pos,off,lim) (PL_reg_match_utf8 ? reghopmaybe3((U8*)pos, off, (U8*)lim) : (U8*)(pos + off))
130 #define HOP3c(pos,off,lim) ((char*)HOP3(pos,off,lim))
131 #define HOPMAYBE3c(pos,off,lim) ((char*)HOPMAYBE3(pos,off,lim))
133 #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
135 /* for use after a quantifier and before an EXACT-like node -- japhy */
136 #define JUMPABLE(rn) ( \
137 OP(rn) == OPEN || OP(rn) == CLOSE || OP(rn) == EVAL || \
138 OP(rn) == SUSPEND || OP(rn) == IFMATCH || \
139 OP(rn) == PLUS || OP(rn) == MINMOD || \
140 (PL_regkind[(U8)OP(rn)] == CURLY && ARG1(rn) > 0) \
143 #define HAS_TEXT(rn) ( \
144 PL_regkind[(U8)OP(rn)] == EXACT || PL_regkind[(U8)OP(rn)] == REF \
148 Search for mandatory following text node; for lookahead, the text must
149 follow but for lookbehind (rn->flags != 0) we skip to the next step.
151 #define FIND_NEXT_IMPT(rn) STMT_START { \
152 while (JUMPABLE(rn)) \
153 if (OP(rn) == SUSPEND || PL_regkind[(U8)OP(rn)] == CURLY) \
154 rn = NEXTOPER(NEXTOPER(rn)); \
155 else if (OP(rn) == PLUS) \
157 else if (OP(rn) == IFMATCH) \
158 rn = (rn->flags == 0) ? NEXTOPER(NEXTOPER(rn)) : rn + ARG(rn); \
159 else rn += NEXT_OFF(rn); \
162 static void restore_pos(pTHX_ void *arg);
165 S_regcppush(pTHX_ I32 parenfloor)
167 int retval = PL_savestack_ix;
168 #define REGCP_PAREN_ELEMS 4
169 int paren_elems_to_push = (PL_regsize - parenfloor) * REGCP_PAREN_ELEMS;
172 if (paren_elems_to_push < 0)
173 Perl_croak(aTHX_ "panic: paren_elems_to_push < 0");
175 #define REGCP_OTHER_ELEMS 6
176 SSGROW(paren_elems_to_push + REGCP_OTHER_ELEMS);
177 for (p = PL_regsize; p > parenfloor; p--) {
178 /* REGCP_PARENS_ELEMS are pushed per pairs of parentheses. */
179 SSPUSHINT(PL_regendp[p]);
180 SSPUSHINT(PL_regstartp[p]);
181 SSPUSHPTR(PL_reg_start_tmp[p]);
184 /* REGCP_OTHER_ELEMS are pushed in any case, parentheses or no. */
185 SSPUSHINT(PL_regsize);
186 SSPUSHINT(*PL_reglastparen);
187 SSPUSHINT(*PL_reglastcloseparen);
188 SSPUSHPTR(PL_reginput);
189 #define REGCP_FRAME_ELEMS 2
190 /* REGCP_FRAME_ELEMS are part of the REGCP_OTHER_ELEMS and
191 * are needed for the regexp context stack bookkeeping. */
192 SSPUSHINT(paren_elems_to_push + REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS);
193 SSPUSHINT(SAVEt_REGCONTEXT); /* Magic cookie. */
198 /* These are needed since we do not localize EVAL nodes: */
199 # define REGCP_SET(cp) DEBUG_r(PerlIO_printf(Perl_debug_log, \
200 " Setting an EVAL scope, savestack=%"IVdf"\n", \
201 (IV)PL_savestack_ix)); cp = PL_savestack_ix
203 # define REGCP_UNWIND(cp) DEBUG_r(cp != PL_savestack_ix ? \
204 PerlIO_printf(Perl_debug_log, \
205 " Clearing an EVAL scope, savestack=%"IVdf"..%"IVdf"\n", \
206 (IV)(cp), (IV)PL_savestack_ix) : 0); regcpblow(cp)
216 /* Pop REGCP_OTHER_ELEMS before the parentheses loop starts. */
218 assert(i == SAVEt_REGCONTEXT); /* Check that the magic cookie is there. */
219 i = SSPOPINT; /* Parentheses elements to pop. */
220 input = (char *) SSPOPPTR;
221 *PL_reglastcloseparen = SSPOPINT;
222 *PL_reglastparen = SSPOPINT;
223 PL_regsize = SSPOPINT;
225 /* Now restore the parentheses context. */
226 for (i -= (REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS);
227 i > 0; i -= REGCP_PAREN_ELEMS) {
228 paren = (U32)SSPOPINT;
229 PL_reg_start_tmp[paren] = (char *) SSPOPPTR;
230 PL_regstartp[paren] = SSPOPINT;
232 if (paren <= *PL_reglastparen)
233 PL_regendp[paren] = tmps;
235 PerlIO_printf(Perl_debug_log,
236 " restoring \\%"UVuf" to %"IVdf"(%"IVdf")..%"IVdf"%s\n",
237 (UV)paren, (IV)PL_regstartp[paren],
238 (IV)(PL_reg_start_tmp[paren] - PL_bostr),
239 (IV)PL_regendp[paren],
240 (paren > *PL_reglastparen ? "(no)" : ""));
244 if ((I32)(*PL_reglastparen + 1) <= PL_regnpar) {
245 PerlIO_printf(Perl_debug_log,
246 " restoring \\%"IVdf"..\\%"IVdf" to undef\n",
247 (IV)(*PL_reglastparen + 1), (IV)PL_regnpar);
251 /* It would seem that the similar code in regtry()
252 * already takes care of this, and in fact it is in
253 * a better location to since this code can #if 0-ed out
254 * but the code in regtry() is needed or otherwise tests
255 * requiring null fields (pat.t#187 and split.t#{13,14}
256 * (as of patchlevel 7877) will fail. Then again,
257 * this code seems to be necessary or otherwise
258 * building DynaLoader will fail:
259 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
261 for (paren = *PL_reglastparen + 1; (I32)paren <= PL_regnpar; paren++) {
262 if ((I32)paren > PL_regsize)
263 PL_regstartp[paren] = -1;
264 PL_regendp[paren] = -1;
271 S_regcp_set_to(pTHX_ I32 ss)
273 I32 tmp = PL_savestack_ix;
275 PL_savestack_ix = ss;
277 PL_savestack_ix = tmp;
281 typedef struct re_cc_state
285 struct re_cc_state *prev;
290 #define regcpblow(cp) LEAVE_SCOPE(cp) /* Ignores regcppush()ed data. */
292 #define TRYPAREN(paren, n, input) { \
295 PL_regstartp[paren] = HOPc(input, -1) - PL_bostr; \
296 PL_regendp[paren] = input - PL_bostr; \
299 PL_regendp[paren] = -1; \
301 if (regmatch(next)) \
304 PL_regendp[paren] = -1; \
309 * pregexec and friends
313 - pregexec - match a regexp against a string
316 Perl_pregexec(pTHX_ register regexp *prog, char *stringarg, register char *strend,
317 char *strbeg, I32 minend, SV *screamer, U32 nosave)
318 /* strend: pointer to null at end of string */
319 /* strbeg: real beginning of string */
320 /* minend: end of match must be >=minend after stringarg. */
321 /* nosave: For optimizations. */
324 regexec_flags(prog, stringarg, strend, strbeg, minend, screamer, NULL,
325 nosave ? 0 : REXEC_COPY_STR);
329 S_cache_re(pTHX_ regexp *prog)
331 PL_regprecomp = prog->precomp; /* Needed for FAIL. */
333 PL_regprogram = prog->program;
335 PL_regnpar = prog->nparens;
336 PL_regdata = prog->data;
341 * Need to implement the following flags for reg_anch:
343 * USE_INTUIT_NOML - Useful to call re_intuit_start() first
345 * INTUIT_AUTORITATIVE_NOML - Can trust a positive answer
346 * INTUIT_AUTORITATIVE_ML
347 * INTUIT_ONCE_NOML - Intuit can match in one location only.
350 * Another flag for this function: SECOND_TIME (so that float substrs
351 * with giant delta may be not rechecked).
354 /* Assumptions: if ANCH_GPOS, then strpos is anchored. XXXX Check GPOS logic */
356 /* If SCREAM, then SvPVX(sv) should be compatible with strpos and strend.
357 Otherwise, only SvCUR(sv) is used to get strbeg. */
359 /* XXXX We assume that strpos is strbeg unless sv. */
361 /* XXXX Some places assume that there is a fixed substring.
362 An update may be needed if optimizer marks as "INTUITable"
363 RExen without fixed substrings. Similarly, it is assumed that
364 lengths of all the strings are no more than minlen, thus they
365 cannot come from lookahead.
366 (Or minlen should take into account lookahead.) */
368 /* A failure to find a constant substring means that there is no need to make
369 an expensive call to REx engine, thus we celebrate a failure. Similarly,
370 finding a substring too deep into the string means that less calls to
371 regtry() should be needed.
373 REx compiler's optimizer found 4 possible hints:
374 a) Anchored substring;
376 c) Whether we are anchored (beginning-of-line or \G);
377 d) First node (of those at offset 0) which may distingush positions;
378 We use a)b)d) and multiline-part of c), and try to find a position in the
379 string which does not contradict any of them.
382 /* Most of decisions we do here should have been done at compile time.
383 The nodes of the REx which we used for the search should have been
384 deleted from the finite automaton. */
387 Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos,
388 char *strend, U32 flags, re_scream_pos_data *data)
390 register I32 start_shift = 0;
391 /* Should be nonnegative! */
392 register I32 end_shift = 0;
397 int do_utf8 = sv ? SvUTF8(sv) : 0; /* if no sv we have to assume bytes */
399 register char *other_last = Nullch; /* other substr checked before this */
400 char *check_at = Nullch; /* check substr found at this pos */
402 char *i_strpos = strpos;
403 SV *dsv = PERL_DEBUG_PAD_ZERO(0);
405 RX_MATCH_UTF8_set(prog,do_utf8);
407 if (prog->reganch & ROPT_UTF8) {
408 DEBUG_r(PerlIO_printf(Perl_debug_log,
409 "UTF-8 regex...\n"));
410 PL_reg_flags |= RF_utf8;
414 char *s = PL_reg_match_utf8 ?
415 sv_uni_display(dsv, sv, 60, UNI_DISPLAY_REGEX) :
417 int len = PL_reg_match_utf8 ?
418 strlen(s) : strend - strpos;
421 if (PL_reg_match_utf8)
422 DEBUG_r(PerlIO_printf(Perl_debug_log,
423 "UTF-8 target...\n"));
424 PerlIO_printf(Perl_debug_log,
425 "%sGuessing start of match, REx%s `%s%.60s%s%s' against `%s%.*s%s%s'...\n",
426 PL_colors[4],PL_colors[5],PL_colors[0],
429 (strlen(prog->precomp) > 60 ? "..." : ""),
431 (int)(len > 60 ? 60 : len),
433 (len > 60 ? "..." : "")
437 /* CHR_DIST() would be more correct here but it makes things slow. */
438 if (prog->minlen > strend - strpos) {
439 DEBUG_r(PerlIO_printf(Perl_debug_log,
440 "String too short... [re_intuit_start]\n"));
443 strbeg = (sv && SvPOK(sv)) ? strend - SvCUR(sv) : strpos;
446 if (!prog->check_utf8 && prog->check_substr)
447 to_utf8_substr(prog);
448 check = prog->check_utf8;
450 if (!prog->check_substr && prog->check_utf8)
451 to_byte_substr(prog);
452 check = prog->check_substr;
454 if (check == &PL_sv_undef) {
455 DEBUG_r(PerlIO_printf(Perl_debug_log,
456 "Non-utf string cannot match utf check string\n"));
459 if (prog->reganch & ROPT_ANCH) { /* Match at beg-of-str or after \n */
460 ml_anch = !( (prog->reganch & ROPT_ANCH_SINGLE)
461 || ( (prog->reganch & ROPT_ANCH_BOL)
462 && !PL_multiline ) ); /* Check after \n? */
465 if ( !(prog->reganch & (ROPT_ANCH_GPOS /* Checked by the caller */
466 | ROPT_IMPLICIT)) /* not a real BOL */
467 /* SvCUR is not set on references: SvRV and SvPVX overlap */
469 && (strpos != strbeg)) {
470 DEBUG_r(PerlIO_printf(Perl_debug_log, "Not at start...\n"));
473 if (prog->check_offset_min == prog->check_offset_max &&
474 !(prog->reganch & ROPT_CANY_SEEN)) {
475 /* Substring at constant offset from beg-of-str... */
478 s = HOP3c(strpos, prog->check_offset_min, strend);
480 slen = SvCUR(check); /* >= 1 */
482 if ( strend - s > slen || strend - s < slen - 1
483 || (strend - s == slen && strend[-1] != '\n')) {
484 DEBUG_r(PerlIO_printf(Perl_debug_log, "String too long...\n"));
487 /* Now should match s[0..slen-2] */
489 if (slen && (*SvPVX(check) != *s
491 && memNE(SvPVX(check), s, slen)))) {
493 DEBUG_r(PerlIO_printf(Perl_debug_log, "String not equal...\n"));
497 else if (*SvPVX(check) != *s
498 || ((slen = SvCUR(check)) > 1
499 && memNE(SvPVX(check), s, slen)))
501 goto success_at_start;
504 /* Match is anchored, but substr is not anchored wrt beg-of-str. */
506 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
507 end_shift = prog->minlen - start_shift -
508 CHR_SVLEN(check) + (SvTAIL(check) != 0);
510 I32 end = prog->check_offset_max + CHR_SVLEN(check)
511 - (SvTAIL(check) != 0);
512 I32 eshift = CHR_DIST((U8*)strend, (U8*)s) - end;
514 if (end_shift < eshift)
518 else { /* Can match at random position */
521 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
522 /* Should be nonnegative! */
523 end_shift = prog->minlen - start_shift -
524 CHR_SVLEN(check) + (SvTAIL(check) != 0);
527 #ifdef DEBUGGING /* 7/99: reports of failure (with the older version) */
529 Perl_croak(aTHX_ "panic: end_shift");
533 /* Find a possible match in the region s..strend by looking for
534 the "check" substring in the region corrected by start/end_shift. */
535 if (flags & REXEC_SCREAM) {
536 I32 p = -1; /* Internal iterator of scream. */
537 I32 *pp = data ? data->scream_pos : &p;
539 if (PL_screamfirst[BmRARE(check)] >= 0
540 || ( BmRARE(check) == '\n'
541 && (BmPREVIOUS(check) == SvCUR(check) - 1)
543 s = screaminstr(sv, check,
544 start_shift + (s - strbeg), end_shift, pp, 0);
547 /* we may be pointing at the wrong string */
548 if (s && RX_MATCH_COPIED(prog))
549 s = strbeg + (s - SvPVX(sv));
551 *data->scream_olds = s;
553 else if (prog->reganch & ROPT_CANY_SEEN)
554 s = fbm_instr((U8*)(s + start_shift),
555 (U8*)(strend - end_shift),
556 check, PL_multiline ? FBMrf_MULTILINE : 0);
558 s = fbm_instr(HOP3(s, start_shift, strend),
559 HOP3(strend, -end_shift, strbeg),
560 check, PL_multiline ? FBMrf_MULTILINE : 0);
562 /* Update the count-of-usability, remove useless subpatterns,
565 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s %s substr `%s%.*s%s'%s%s",
566 (s ? "Found" : "Did not find"),
567 (check == (do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) ? "anchored" : "floating"),
569 (int)(SvCUR(check) - (SvTAIL(check)!=0)),
571 PL_colors[1], (SvTAIL(check) ? "$" : ""),
572 (s ? " at offset " : "...\n") ) );
579 /* Finish the diagnostic message */
580 DEBUG_r(PerlIO_printf(Perl_debug_log, "%ld...\n", (long)(s - i_strpos)) );
582 /* Got a candidate. Check MBOL anchoring, and the *other* substr.
583 Start with the other substr.
584 XXXX no SCREAM optimization yet - and a very coarse implementation
585 XXXX /ttx+/ results in anchored=`ttx', floating=`x'. floating will
586 *always* match. Probably should be marked during compile...
587 Probably it is right to do no SCREAM here...
590 if (do_utf8 ? (prog->float_utf8 && prog->anchored_utf8) : (prog->float_substr && prog->anchored_substr)) {
591 /* Take into account the "other" substring. */
592 /* XXXX May be hopelessly wrong for UTF... */
595 if (check == (do_utf8 ? prog->float_utf8 : prog->float_substr)) {
598 char *last = HOP3c(s, -start_shift, strbeg), *last1, *last2;
602 t = s - prog->check_offset_max;
603 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
605 || ((t = reghopmaybe3_c(s, -(prog->check_offset_max), strpos))
610 t = HOP3c(t, prog->anchored_offset, strend);
611 if (t < other_last) /* These positions already checked */
613 last2 = last1 = HOP3c(strend, -prog->minlen, strbeg);
616 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
617 /* On end-of-str: see comment below. */
618 must = do_utf8 ? prog->anchored_utf8 : prog->anchored_substr;
619 if (must == &PL_sv_undef) {
621 DEBUG_r(must = prog->anchored_utf8); /* for debug */
626 HOP3(HOP3(last1, prog->anchored_offset, strend)
627 + SvCUR(must), -(SvTAIL(must)!=0), strbeg),
629 PL_multiline ? FBMrf_MULTILINE : 0
631 DEBUG_r(PerlIO_printf(Perl_debug_log,
632 "%s anchored substr `%s%.*s%s'%s",
633 (s ? "Found" : "Contradicts"),
636 - (SvTAIL(must)!=0)),
638 PL_colors[1], (SvTAIL(must) ? "$" : "")));
640 if (last1 >= last2) {
641 DEBUG_r(PerlIO_printf(Perl_debug_log,
642 ", giving up...\n"));
645 DEBUG_r(PerlIO_printf(Perl_debug_log,
646 ", trying floating at offset %ld...\n",
647 (long)(HOP3c(s1, 1, strend) - i_strpos)));
648 other_last = HOP3c(last1, prog->anchored_offset+1, strend);
649 s = HOP3c(last, 1, strend);
653 DEBUG_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
654 (long)(s - i_strpos)));
655 t = HOP3c(s, -prog->anchored_offset, strbeg);
656 other_last = HOP3c(s, 1, strend);
664 else { /* Take into account the floating substring. */
669 t = HOP3c(s, -start_shift, strbeg);
671 HOP3c(strend, -prog->minlen + prog->float_min_offset, strbeg);
672 if (CHR_DIST((U8*)last, (U8*)t) > prog->float_max_offset)
673 last = HOP3c(t, prog->float_max_offset, strend);
674 s = HOP3c(t, prog->float_min_offset, strend);
677 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
678 must = do_utf8 ? prog->float_utf8 : prog->float_substr;
679 /* fbm_instr() takes into account exact value of end-of-str
680 if the check is SvTAIL(ed). Since false positives are OK,
681 and end-of-str is not later than strend we are OK. */
682 if (must == &PL_sv_undef) {
684 DEBUG_r(must = prog->float_utf8); /* for debug message */
687 s = fbm_instr((unsigned char*)s,
688 (unsigned char*)last + SvCUR(must)
690 must, PL_multiline ? FBMrf_MULTILINE : 0);
691 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s floating substr `%s%.*s%s'%s",
692 (s ? "Found" : "Contradicts"),
694 (int)(SvCUR(must) - (SvTAIL(must)!=0)),
696 PL_colors[1], (SvTAIL(must) ? "$" : "")));
699 DEBUG_r(PerlIO_printf(Perl_debug_log,
700 ", giving up...\n"));
703 DEBUG_r(PerlIO_printf(Perl_debug_log,
704 ", trying anchored starting at offset %ld...\n",
705 (long)(s1 + 1 - i_strpos)));
707 s = HOP3c(t, 1, strend);
711 DEBUG_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
712 (long)(s - i_strpos)));
713 other_last = s; /* Fix this later. --Hugo */
722 t = s - prog->check_offset_max;
723 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
725 || ((t = reghopmaybe3_c(s, -prog->check_offset_max, strpos))
727 /* Fixed substring is found far enough so that the match
728 cannot start at strpos. */
730 if (ml_anch && t[-1] != '\n') {
731 /* Eventually fbm_*() should handle this, but often
732 anchored_offset is not 0, so this check will not be wasted. */
733 /* XXXX In the code below we prefer to look for "^" even in
734 presence of anchored substrings. And we search even
735 beyond the found float position. These pessimizations
736 are historical artefacts only. */
738 while (t < strend - prog->minlen) {
740 if (t < check_at - prog->check_offset_min) {
741 if (do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) {
742 /* Since we moved from the found position,
743 we definitely contradict the found anchored
744 substr. Due to the above check we do not
745 contradict "check" substr.
746 Thus we can arrive here only if check substr
747 is float. Redo checking for "other"=="fixed".
750 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld, rescanning for anchored from offset %ld...\n",
751 PL_colors[0],PL_colors[1], (long)(strpos - i_strpos), (long)(strpos - i_strpos + prog->anchored_offset)));
752 goto do_other_anchored;
754 /* We don't contradict the found floating substring. */
755 /* XXXX Why not check for STCLASS? */
757 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld...\n",
758 PL_colors[0],PL_colors[1], (long)(s - i_strpos)));
761 /* Position contradicts check-string */
762 /* XXXX probably better to look for check-string
763 than for "\n", so one should lower the limit for t? */
764 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m, restarting lookup for check-string at offset %ld...\n",
765 PL_colors[0],PL_colors[1], (long)(t + 1 - i_strpos)));
766 other_last = strpos = s = t + 1;
771 DEBUG_r(PerlIO_printf(Perl_debug_log, "Did not find /%s^%s/m...\n",
772 PL_colors[0],PL_colors[1]));
776 DEBUG_r(PerlIO_printf(Perl_debug_log, "Starting position does not contradict /%s^%s/m...\n",
777 PL_colors[0],PL_colors[1]));
781 ++BmUSEFUL(do_utf8 ? prog->check_utf8 : prog->check_substr); /* hooray/5 */
784 /* The found string does not prohibit matching at strpos,
785 - no optimization of calling REx engine can be performed,
786 unless it was an MBOL and we are not after MBOL,
787 or a future STCLASS check will fail this. */
789 /* Even in this situation we may use MBOL flag if strpos is offset
790 wrt the start of the string. */
791 if (ml_anch && sv && !SvROK(sv) /* See prev comment on SvROK */
792 && (strpos != strbeg) && strpos[-1] != '\n'
793 /* May be due to an implicit anchor of m{.*foo} */
794 && !(prog->reganch & ROPT_IMPLICIT))
799 DEBUG_r( if (ml_anch)
800 PerlIO_printf(Perl_debug_log, "Position at offset %ld does not contradict /%s^%s/m...\n",
801 (long)(strpos - i_strpos), PL_colors[0],PL_colors[1]);
804 if (!(prog->reganch & ROPT_NAUGHTY) /* XXXX If strpos moved? */
806 prog->check_utf8 /* Could be deleted already */
807 && --BmUSEFUL(prog->check_utf8) < 0
808 && (prog->check_utf8 == prog->float_utf8)
810 prog->check_substr /* Could be deleted already */
811 && --BmUSEFUL(prog->check_substr) < 0
812 && (prog->check_substr == prog->float_substr)
815 /* If flags & SOMETHING - do not do it many times on the same match */
816 DEBUG_r(PerlIO_printf(Perl_debug_log, "... Disabling check substring...\n"));
817 SvREFCNT_dec(do_utf8 ? prog->check_utf8 : prog->check_substr);
818 if (do_utf8 ? prog->check_substr : prog->check_utf8)
819 SvREFCNT_dec(do_utf8 ? prog->check_substr : prog->check_utf8);
820 prog->check_substr = prog->check_utf8 = Nullsv; /* disable */
821 prog->float_substr = prog->float_utf8 = Nullsv; /* clear */
822 check = Nullsv; /* abort */
824 /* XXXX This is a remnant of the old implementation. It
825 looks wasteful, since now INTUIT can use many
827 prog->reganch &= ~RE_USE_INTUIT;
834 /* XXXX BmUSEFUL already changed, maybe multiple change is meaningful... */
835 if (prog->regstclass) {
836 /* minlen == 0 is possible if regstclass is \b or \B,
837 and the fixed substr is ''$.
838 Since minlen is already taken into account, s+1 is before strend;
839 accidentally, minlen >= 1 guaranties no false positives at s + 1
840 even for \b or \B. But (minlen? 1 : 0) below assumes that
841 regstclass does not come from lookahead... */
842 /* If regstclass takes bytelength more than 1: If charlength==1, OK.
843 This leaves EXACTF only, which is dealt with in find_byclass(). */
844 U8* str = (U8*)STRING(prog->regstclass);
845 int cl_l = (PL_regkind[(U8)OP(prog->regstclass)] == EXACT
846 ? CHR_DIST(str+STR_LEN(prog->regstclass), str)
848 char *endpos = (prog->anchored_substr || prog->anchored_utf8 || ml_anch)
849 ? HOP3c(s, (prog->minlen ? cl_l : 0), strend)
850 : (prog->float_substr || prog->float_utf8
851 ? HOP3c(HOP3c(check_at, -start_shift, strbeg),
854 char *startpos = strbeg;
858 s = find_byclass(prog, prog->regstclass, s, endpos, startpos, 1);
863 if (endpos == strend) {
864 DEBUG_r( PerlIO_printf(Perl_debug_log,
865 "Could not match STCLASS...\n") );
868 DEBUG_r( PerlIO_printf(Perl_debug_log,
869 "This position contradicts STCLASS...\n") );
870 if ((prog->reganch & ROPT_ANCH) && !ml_anch)
872 /* Contradict one of substrings */
873 if (prog->anchored_substr || prog->anchored_utf8) {
874 if ((do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) == check) {
875 DEBUG_r( what = "anchored" );
877 s = HOP3c(t, 1, strend);
878 if (s + start_shift + end_shift > strend) {
879 /* XXXX Should be taken into account earlier? */
880 DEBUG_r( PerlIO_printf(Perl_debug_log,
881 "Could not match STCLASS...\n") );
886 DEBUG_r( PerlIO_printf(Perl_debug_log,
887 "Looking for %s substr starting at offset %ld...\n",
888 what, (long)(s + start_shift - i_strpos)) );
891 /* Have both, check_string is floating */
892 if (t + start_shift >= check_at) /* Contradicts floating=check */
893 goto retry_floating_check;
894 /* Recheck anchored substring, but not floating... */
898 DEBUG_r( PerlIO_printf(Perl_debug_log,
899 "Looking for anchored substr starting at offset %ld...\n",
900 (long)(other_last - i_strpos)) );
901 goto do_other_anchored;
903 /* Another way we could have checked stclass at the
904 current position only: */
909 DEBUG_r( PerlIO_printf(Perl_debug_log,
910 "Looking for /%s^%s/m starting at offset %ld...\n",
911 PL_colors[0],PL_colors[1], (long)(t - i_strpos)) );
914 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr)) /* Could have been deleted */
916 /* Check is floating subtring. */
917 retry_floating_check:
918 t = check_at - start_shift;
919 DEBUG_r( what = "floating" );
920 goto hop_and_restart;
923 DEBUG_r(PerlIO_printf(Perl_debug_log,
924 "By STCLASS: moving %ld --> %ld\n",
925 (long)(t - i_strpos), (long)(s - i_strpos))
929 DEBUG_r(PerlIO_printf(Perl_debug_log,
930 "Does not contradict STCLASS...\n");
935 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s%s:%s match at offset %ld\n",
936 PL_colors[4], (check ? "Guessed" : "Giving up"),
937 PL_colors[5], (long)(s - i_strpos)) );
940 fail_finish: /* Substring not found */
941 if (prog->check_substr || prog->check_utf8) /* could be removed already */
942 BmUSEFUL(do_utf8 ? prog->check_utf8 : prog->check_substr) += 5; /* hooray */
944 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch rejected by optimizer%s\n",
945 PL_colors[4],PL_colors[5]));
949 /* We know what class REx starts with. Try to find this position... */
951 S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, char *strend, char *startpos, I32 norun)
953 I32 doevery = (prog->reganch & ROPT_SKIP) == 0;
959 register I32 tmp = 1; /* Scratch variable? */
960 register bool do_utf8 = PL_reg_match_utf8;
962 /* We know what class it must start with. */
967 if ((ANYOF_FLAGS(c) & ANYOF_UNICODE) ||
968 !UTF8_IS_INVARIANT((U8)s[0]) ?
969 reginclass(c, (U8*)s, 0, do_utf8) :
970 REGINCLASS(c, (U8*)s)) {
971 if (tmp && (norun || regtry(prog, s)))
985 if (REGINCLASS(c, (U8*)s) ||
986 (ANYOF_FOLD_SHARP_S(c, s, strend) &&
987 /* The assignment of 2 is intentional:
988 * for the folded sharp s, the skip is 2. */
989 (skip = SHARP_S_SKIP))) {
990 if (tmp && (norun || regtry(prog, s)))
1002 while (s < strend) {
1003 if (tmp && (norun || regtry(prog, s)))
1014 STRLEN ulen1, ulen2;
1015 U8 tmpbuf1[UTF8_MAXLEN_UCLC+1];
1016 U8 tmpbuf2[UTF8_MAXLEN_UCLC+1];
1018 to_utf8_lower((U8*)m, tmpbuf1, &ulen1);
1019 to_utf8_upper((U8*)m, tmpbuf2, &ulen2);
1021 c1 = utf8n_to_uvchr(tmpbuf1, UTF8_MAXLEN_UCLC,
1022 0, ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
1023 c2 = utf8n_to_uvchr(tmpbuf2, UTF8_MAXLEN_UCLC,
1024 0, ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
1035 c2 = PL_fold_locale[c1];
1037 e = HOP3c(strend, -(I32)ln, s);
1040 e = s; /* Due to minlen logic of intuit() */
1042 /* The idea in the EXACTF* cases is to first find the
1043 * first character of the EXACTF* node and then, if
1044 * necessary, case-insensitively compare the full
1045 * text of the node. The c1 and c2 are the first
1046 * characters (though in Unicode it gets a bit
1047 * more complicated because there are more cases
1048 * than just upper and lower: one needs to use
1049 * the so-called folding case for case-insensitive
1050 * matching (called "loose matching" in Unicode).
1051 * ibcmp_utf8() will do just that. */
1055 U8 tmpbuf [UTF8_MAXLEN+1];
1056 U8 foldbuf[UTF8_MAXLEN_FOLD+1];
1057 STRLEN len, foldlen;
1061 c = utf8n_to_uvchr((U8*)s, UTF8_MAXLEN, &len,
1063 0 : UTF8_ALLOW_ANY);
1066 ibcmp_utf8(s, (char **)0, 0, do_utf8,
1067 m, (char **)0, ln, (bool)UTF))
1068 && (norun || regtry(prog, s)) )
1071 uvchr_to_utf8(tmpbuf, c);
1072 f = to_utf8_fold(tmpbuf, foldbuf, &foldlen);
1074 && (f == c1 || f == c2)
1075 && (ln == foldlen ||
1076 !ibcmp_utf8((char *) foldbuf,
1077 (char **)0, foldlen, do_utf8,
1079 (char **)0, ln, (bool)UTF))
1080 && (norun || regtry(prog, s)) )
1088 c = utf8n_to_uvchr((U8*)s, UTF8_MAXLEN, &len,
1090 0 : UTF8_ALLOW_ANY);
1092 /* Handle some of the three Greek sigmas cases.
1093 * Note that not all the possible combinations
1094 * are handled here: some of them are handled
1095 * by the standard folding rules, and some of
1096 * them (the character class or ANYOF cases)
1097 * are handled during compiletime in
1098 * regexec.c:S_regclass(). */
1099 if (c == (UV)UNICODE_GREEK_CAPITAL_LETTER_SIGMA ||
1100 c == (UV)UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA)
1101 c = (UV)UNICODE_GREEK_SMALL_LETTER_SIGMA;
1103 if ( (c == c1 || c == c2)
1105 ibcmp_utf8(s, (char **)0, 0, do_utf8,
1106 m, (char **)0, ln, (bool)UTF))
1107 && (norun || regtry(prog, s)) )
1110 uvchr_to_utf8(tmpbuf, c);
1111 f = to_utf8_fold(tmpbuf, foldbuf, &foldlen);
1113 && (f == c1 || f == c2)
1114 && (ln == foldlen ||
1115 !ibcmp_utf8((char *) foldbuf,
1116 (char **)0, foldlen, do_utf8,
1118 (char **)0, ln, (bool)UTF))
1119 && (norun || regtry(prog, s)) )
1130 && (ln == 1 || !(OP(c) == EXACTF
1132 : ibcmp_locale(s, m, ln)))
1133 && (norun || regtry(prog, s)) )
1139 if ( (*(U8*)s == c1 || *(U8*)s == c2)
1140 && (ln == 1 || !(OP(c) == EXACTF
1142 : ibcmp_locale(s, m, ln)))
1143 && (norun || regtry(prog, s)) )
1150 PL_reg_flags |= RF_tainted;
1157 U8 *r = reghop3((U8*)s, -1, (U8*)PL_bostr);
1159 tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, 0);
1161 tmp = ((OP(c) == BOUND ?
1162 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1163 LOAD_UTF8_CHARCLASS(alnum,"a");
1164 while (s < strend) {
1165 if (tmp == !(OP(c) == BOUND ?
1166 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1167 isALNUM_LC_utf8((U8*)s)))
1170 if ((norun || regtry(prog, s)))
1177 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1178 tmp = ((OP(c) == BOUND ? isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1179 while (s < strend) {
1181 !(OP(c) == BOUND ? isALNUM(*s) : isALNUM_LC(*s))) {
1183 if ((norun || regtry(prog, s)))
1189 if ((!prog->minlen && tmp) && (norun || regtry(prog, s)))
1193 PL_reg_flags |= RF_tainted;
1200 U8 *r = reghop3((U8*)s, -1, (U8*)PL_bostr);
1202 tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, 0);
1204 tmp = ((OP(c) == NBOUND ?
1205 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1206 LOAD_UTF8_CHARCLASS(alnum,"a");
1207 while (s < strend) {
1208 if (tmp == !(OP(c) == NBOUND ?
1209 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1210 isALNUM_LC_utf8((U8*)s)))
1212 else if ((norun || regtry(prog, s)))
1218 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1219 tmp = ((OP(c) == NBOUND ?
1220 isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1221 while (s < strend) {
1223 !(OP(c) == NBOUND ? isALNUM(*s) : isALNUM_LC(*s)))
1225 else if ((norun || regtry(prog, s)))
1230 if ((!prog->minlen && !tmp) && (norun || regtry(prog, s)))
1235 LOAD_UTF8_CHARCLASS(alnum,"a");
1236 while (s < strend) {
1237 if (swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1238 if (tmp && (norun || regtry(prog, s)))
1249 while (s < strend) {
1251 if (tmp && (norun || regtry(prog, s)))
1263 PL_reg_flags |= RF_tainted;
1265 while (s < strend) {
1266 if (isALNUM_LC_utf8((U8*)s)) {
1267 if (tmp && (norun || regtry(prog, s)))
1278 while (s < strend) {
1279 if (isALNUM_LC(*s)) {
1280 if (tmp && (norun || regtry(prog, s)))
1293 LOAD_UTF8_CHARCLASS(alnum,"a");
1294 while (s < strend) {
1295 if (!swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1296 if (tmp && (norun || regtry(prog, s)))
1307 while (s < strend) {
1309 if (tmp && (norun || regtry(prog, s)))
1321 PL_reg_flags |= RF_tainted;
1323 while (s < strend) {
1324 if (!isALNUM_LC_utf8((U8*)s)) {
1325 if (tmp && (norun || regtry(prog, s)))
1336 while (s < strend) {
1337 if (!isALNUM_LC(*s)) {
1338 if (tmp && (norun || regtry(prog, s)))
1351 LOAD_UTF8_CHARCLASS(space," ");
1352 while (s < strend) {
1353 if (*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8)) {
1354 if (tmp && (norun || regtry(prog, s)))
1365 while (s < strend) {
1367 if (tmp && (norun || regtry(prog, s)))
1379 PL_reg_flags |= RF_tainted;
1381 while (s < strend) {
1382 if (*s == ' ' || isSPACE_LC_utf8((U8*)s)) {
1383 if (tmp && (norun || regtry(prog, s)))
1394 while (s < strend) {
1395 if (isSPACE_LC(*s)) {
1396 if (tmp && (norun || regtry(prog, s)))
1409 LOAD_UTF8_CHARCLASS(space," ");
1410 while (s < strend) {
1411 if (!(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8))) {
1412 if (tmp && (norun || regtry(prog, s)))
1423 while (s < strend) {
1425 if (tmp && (norun || regtry(prog, s)))
1437 PL_reg_flags |= RF_tainted;
1439 while (s < strend) {
1440 if (!(*s == ' ' || isSPACE_LC_utf8((U8*)s))) {
1441 if (tmp && (norun || regtry(prog, s)))
1452 while (s < strend) {
1453 if (!isSPACE_LC(*s)) {
1454 if (tmp && (norun || regtry(prog, s)))
1467 LOAD_UTF8_CHARCLASS(digit,"0");
1468 while (s < strend) {
1469 if (swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1470 if (tmp && (norun || regtry(prog, s)))
1481 while (s < strend) {
1483 if (tmp && (norun || regtry(prog, s)))
1495 PL_reg_flags |= RF_tainted;
1497 while (s < strend) {
1498 if (isDIGIT_LC_utf8((U8*)s)) {
1499 if (tmp && (norun || regtry(prog, s)))
1510 while (s < strend) {
1511 if (isDIGIT_LC(*s)) {
1512 if (tmp && (norun || regtry(prog, s)))
1525 LOAD_UTF8_CHARCLASS(digit,"0");
1526 while (s < strend) {
1527 if (!swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1528 if (tmp && (norun || regtry(prog, s)))
1539 while (s < strend) {
1541 if (tmp && (norun || regtry(prog, s)))
1553 PL_reg_flags |= RF_tainted;
1555 while (s < strend) {
1556 if (!isDIGIT_LC_utf8((U8*)s)) {
1557 if (tmp && (norun || regtry(prog, s)))
1568 while (s < strend) {
1569 if (!isDIGIT_LC(*s)) {
1570 if (tmp && (norun || regtry(prog, s)))
1582 Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c));
1591 - regexec_flags - match a regexp against a string
1594 Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char *strend,
1595 char *strbeg, I32 minend, SV *sv, void *data, U32 flags)
1596 /* strend: pointer to null at end of string */
1597 /* strbeg: real beginning of string */
1598 /* minend: end of match must be >=minend after stringarg. */
1599 /* data: May be used for some additional optimizations. */
1600 /* nosave: For optimizations. */
1603 register regnode *c;
1604 register char *startpos = stringarg;
1605 I32 minlen; /* must match at least this many chars */
1606 I32 dontbother = 0; /* how many characters not to try at end */
1607 /* I32 start_shift = 0; */ /* Offset of the start to find
1608 constant substr. */ /* CC */
1609 I32 end_shift = 0; /* Same for the end. */ /* CC */
1610 I32 scream_pos = -1; /* Internal iterator of scream. */
1612 SV* oreplsv = GvSV(PL_replgv);
1613 bool do_utf8 = DO_UTF8(sv);
1615 SV *dsv0 = PERL_DEBUG_PAD_ZERO(0);
1616 SV *dsv1 = PERL_DEBUG_PAD_ZERO(1);
1618 RX_MATCH_UTF8_set(prog,do_utf8);
1624 PL_regnarrate = DEBUG_r_TEST;
1627 /* Be paranoid... */
1628 if (prog == NULL || startpos == NULL) {
1629 Perl_croak(aTHX_ "NULL regexp parameter");
1633 minlen = prog->minlen;
1634 if (strend - startpos < minlen) {
1635 DEBUG_r(PerlIO_printf(Perl_debug_log,
1636 "String too short [regexec_flags]...\n"));
1640 /* Check validity of program. */
1641 if (UCHARAT(prog->program) != REG_MAGIC) {
1642 Perl_croak(aTHX_ "corrupted regexp program");
1646 PL_reg_eval_set = 0;
1649 if (prog->reganch & ROPT_UTF8)
1650 PL_reg_flags |= RF_utf8;
1652 /* Mark beginning of line for ^ and lookbehind. */
1653 PL_regbol = startpos;
1657 /* Mark end of line for $ (and such) */
1660 /* see how far we have to get to not match where we matched before */
1661 PL_regtill = startpos+minend;
1663 /* We start without call_cc context. */
1666 /* If there is a "must appear" string, look for it. */
1669 if (prog->reganch & ROPT_GPOS_SEEN) { /* Need to have PL_reg_ganch */
1672 if (flags & REXEC_IGNOREPOS) /* Means: check only at start */
1673 PL_reg_ganch = startpos;
1674 else if (sv && SvTYPE(sv) >= SVt_PVMG
1676 && (mg = mg_find(sv, PERL_MAGIC_regex_global))
1677 && mg->mg_len >= 0) {
1678 PL_reg_ganch = strbeg + mg->mg_len; /* Defined pos() */
1679 if (prog->reganch & ROPT_ANCH_GPOS) {
1680 if (s > PL_reg_ganch)
1685 else /* pos() not defined */
1686 PL_reg_ganch = strbeg;
1689 if (!(flags & REXEC_CHECKED) && (prog->check_substr != Nullsv || prog->check_utf8 != Nullsv)) {
1690 re_scream_pos_data d;
1692 d.scream_olds = &scream_olds;
1693 d.scream_pos = &scream_pos;
1694 s = re_intuit_start(prog, sv, s, strend, flags, &d);
1696 DEBUG_r(PerlIO_printf(Perl_debug_log, "Not present...\n"));
1697 goto phooey; /* not present */
1703 pv_uni_display(dsv0, (U8*)prog->precomp, prog->prelen, 60,
1704 UNI_DISPLAY_REGEX) :
1706 int len0 = UTF ? SvCUR(dsv0) : prog->prelen;
1707 char *s1 = do_utf8 ? sv_uni_display(dsv1, sv, 60,
1708 UNI_DISPLAY_REGEX) : startpos;
1709 int len1 = do_utf8 ? SvCUR(dsv1) : strend - startpos;
1712 PerlIO_printf(Perl_debug_log,
1713 "%sMatching REx%s `%s%*.*s%s%s' against `%s%.*s%s%s'\n",
1714 PL_colors[4],PL_colors[5],PL_colors[0],
1717 len0 > 60 ? "..." : "",
1719 (int)(len1 > 60 ? 60 : len1),
1721 (len1 > 60 ? "..." : "")
1725 /* Simplest case: anchored match need be tried only once. */
1726 /* [unless only anchor is BOL and multiline is set] */
1727 if (prog->reganch & (ROPT_ANCH & ~ROPT_ANCH_GPOS)) {
1728 if (s == startpos && regtry(prog, startpos))
1730 else if (PL_multiline || (prog->reganch & ROPT_IMPLICIT)
1731 || (prog->reganch & ROPT_ANCH_MBOL)) /* XXXX SBOL? */
1736 dontbother = minlen - 1;
1737 end = HOP3c(strend, -dontbother, strbeg) - 1;
1738 /* for multiline we only have to try after newlines */
1739 if (prog->check_substr || prog->check_utf8) {
1743 if (regtry(prog, s))
1748 if (prog->reganch & RE_USE_INTUIT) {
1749 s = re_intuit_start(prog, sv, s + 1, strend, flags, NULL);
1760 if (*s++ == '\n') { /* don't need PL_utf8skip here */
1761 if (regtry(prog, s))
1768 } else if (prog->reganch & ROPT_ANCH_GPOS) {
1769 if (regtry(prog, PL_reg_ganch))
1774 /* Messy cases: unanchored match. */
1775 if ((prog->anchored_substr || prog->anchored_utf8) && prog->reganch & ROPT_SKIP) {
1776 /* we have /x+whatever/ */
1777 /* it must be a one character string (XXXX Except UTF?) */
1782 if (!(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr))
1783 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1784 ch = SvPVX(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr)[0];
1787 while (s < strend) {
1789 DEBUG_r( did_match = 1 );
1790 if (regtry(prog, s)) goto got_it;
1792 while (s < strend && *s == ch)
1799 while (s < strend) {
1801 DEBUG_r( did_match = 1 );
1802 if (regtry(prog, s)) goto got_it;
1804 while (s < strend && *s == ch)
1810 DEBUG_r(if (!did_match)
1811 PerlIO_printf(Perl_debug_log,
1812 "Did not find anchored character...\n")
1816 else if (prog->anchored_substr != Nullsv
1817 || prog->anchored_utf8 != Nullsv
1818 || ((prog->float_substr != Nullsv || prog->float_utf8 != Nullsv)
1819 && prog->float_max_offset < strend - s)) {
1824 char *last1; /* Last position checked before */
1828 if (prog->anchored_substr || prog->anchored_utf8) {
1829 if (!(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr))
1830 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1831 must = do_utf8 ? prog->anchored_utf8 : prog->anchored_substr;
1832 back_max = back_min = prog->anchored_offset;
1834 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr))
1835 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1836 must = do_utf8 ? prog->float_utf8 : prog->float_substr;
1837 back_max = prog->float_max_offset;
1838 back_min = prog->float_min_offset;
1840 if (must == &PL_sv_undef)
1841 /* could not downgrade utf8 check substring, so must fail */
1844 last = HOP3c(strend, /* Cannot start after this */
1845 -(I32)(CHR_SVLEN(must)
1846 - (SvTAIL(must) != 0) + back_min), strbeg);
1849 last1 = HOPc(s, -1);
1851 last1 = s - 1; /* bogus */
1853 /* XXXX check_substr already used to find `s', can optimize if
1854 check_substr==must. */
1856 dontbother = end_shift;
1857 strend = HOPc(strend, -dontbother);
1858 while ( (s <= last) &&
1859 ((flags & REXEC_SCREAM)
1860 ? (s = screaminstr(sv, must, HOP3c(s, back_min, strend) - strbeg,
1861 end_shift, &scream_pos, 0))
1862 : (s = fbm_instr((unsigned char*)HOP3(s, back_min, strend),
1863 (unsigned char*)strend, must,
1864 PL_multiline ? FBMrf_MULTILINE : 0))) ) {
1865 /* we may be pointing at the wrong string */
1866 if ((flags & REXEC_SCREAM) && RX_MATCH_COPIED(prog))
1867 s = strbeg + (s - SvPVX(sv));
1868 DEBUG_r( did_match = 1 );
1869 if (HOPc(s, -back_max) > last1) {
1870 last1 = HOPc(s, -back_min);
1871 s = HOPc(s, -back_max);
1874 char *t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1;
1876 last1 = HOPc(s, -back_min);
1880 while (s <= last1) {
1881 if (regtry(prog, s))
1887 while (s <= last1) {
1888 if (regtry(prog, s))
1894 DEBUG_r(if (!did_match)
1895 PerlIO_printf(Perl_debug_log,
1896 "Did not find %s substr `%s%.*s%s'%s...\n",
1897 ((must == prog->anchored_substr || must == prog->anchored_utf8)
1898 ? "anchored" : "floating"),
1900 (int)(SvCUR(must) - (SvTAIL(must)!=0)),
1902 PL_colors[1], (SvTAIL(must) ? "$" : ""))
1906 else if ((c = prog->regstclass)) {
1908 I32 op = (U8)OP(prog->regstclass);
1909 /* don't bother with what can't match */
1910 if (PL_regkind[op] != EXACT && op != CANY)
1911 strend = HOPc(strend, -(minlen - 1));
1914 SV *prop = sv_newmortal();
1922 pv_uni_display(dsv0, (U8*)SvPVX(prop), SvCUR(prop), 60,
1923 UNI_DISPLAY_REGEX) :
1925 len0 = UTF ? SvCUR(dsv0) : SvCUR(prop);
1927 sv_uni_display(dsv1, sv, 60, UNI_DISPLAY_REGEX) : s;
1928 len1 = UTF ? SvCUR(dsv1) : strend - s;
1929 PerlIO_printf(Perl_debug_log,
1930 "Matching stclass `%*.*s' against `%*.*s'\n",
1934 if (find_byclass(prog, c, s, strend, startpos, 0))
1936 DEBUG_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass...\n"));
1940 if (prog->float_substr != Nullsv || prog->float_utf8 != Nullsv) {
1945 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr))
1946 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1947 float_real = do_utf8 ? prog->float_utf8 : prog->float_substr;
1949 if (flags & REXEC_SCREAM) {
1950 last = screaminstr(sv, float_real, s - strbeg,
1951 end_shift, &scream_pos, 1); /* last one */
1953 last = scream_olds; /* Only one occurrence. */
1954 /* we may be pointing at the wrong string */
1955 else if (RX_MATCH_COPIED(prog))
1956 s = strbeg + (s - SvPVX(sv));
1960 char *little = SvPV(float_real, len);
1962 if (SvTAIL(float_real)) {
1963 if (memEQ(strend - len + 1, little, len - 1))
1964 last = strend - len + 1;
1965 else if (!PL_multiline)
1966 last = memEQ(strend - len, little, len)
1967 ? strend - len : Nullch;
1973 last = rninstr(s, strend, little, little + len);
1975 last = strend; /* matching `$' */
1979 DEBUG_r(PerlIO_printf(Perl_debug_log,
1980 "%sCan't trim the tail, match fails (should not happen)%s\n",
1981 PL_colors[4],PL_colors[5]));
1982 goto phooey; /* Should not happen! */
1984 dontbother = strend - last + prog->float_min_offset;
1986 if (minlen && (dontbother < minlen))
1987 dontbother = minlen - 1;
1988 strend -= dontbother; /* this one's always in bytes! */
1989 /* We don't know much -- general case. */
1992 if (regtry(prog, s))
2001 if (regtry(prog, s))
2003 } while (s++ < strend);
2011 RX_MATCH_TAINTED_set(prog, PL_reg_flags & RF_tainted);
2013 if (PL_reg_eval_set) {
2014 /* Preserve the current value of $^R */
2015 if (oreplsv != GvSV(PL_replgv))
2016 sv_setsv(oreplsv, GvSV(PL_replgv));/* So that when GvSV(replgv) is
2017 restored, the value remains
2019 restore_pos(aTHX_ 0);
2022 /* make sure $`, $&, $', and $digit will work later */
2023 if ( !(flags & REXEC_NOT_FIRST) ) {
2024 RX_MATCH_COPY_FREE(prog);
2025 if (flags & REXEC_COPY_STR) {
2026 I32 i = PL_regeol - startpos + (stringarg - strbeg);
2027 #ifdef PERL_COPY_ON_WRITE
2029 || (SvFLAGS(sv) & CAN_COW_MASK) == CAN_COW_FLAGS)) {
2031 PerlIO_printf(Perl_debug_log,
2032 "Copy on write: regexp capture, type %d\n",
2035 prog->saved_copy = sv_setsv_cow(prog->saved_copy, sv);
2036 prog->subbeg = SvPVX(prog->saved_copy);
2037 assert (SvPOKp(prog->saved_copy));
2041 RX_MATCH_COPIED_on(prog);
2042 s = savepvn(strbeg, i);
2048 prog->subbeg = strbeg;
2049 prog->sublen = PL_regeol - strbeg; /* strend may have been modified */
2056 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n",
2057 PL_colors[4],PL_colors[5]));
2058 if (PL_reg_eval_set)
2059 restore_pos(aTHX_ 0);
2064 - regtry - try match at specific point
2066 STATIC I32 /* 0 failure, 1 success */
2067 S_regtry(pTHX_ regexp *prog, char *startpos)
2075 PL_regindent = 0; /* XXXX Not good when matches are reenterable... */
2077 if ((prog->reganch & ROPT_EVAL_SEEN) && !PL_reg_eval_set) {
2080 PL_reg_eval_set = RS_init;
2082 PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %"IVdf"\n",
2083 (IV)(PL_stack_sp - PL_stack_base));
2085 SAVEI32(cxstack[cxstack_ix].blk_oldsp);
2086 cxstack[cxstack_ix].blk_oldsp = PL_stack_sp - PL_stack_base;
2087 /* Otherwise OP_NEXTSTATE will free whatever on stack now. */
2089 /* Apparently this is not needed, judging by wantarray. */
2090 /* SAVEI8(cxstack[cxstack_ix].blk_gimme);
2091 cxstack[cxstack_ix].blk_gimme = G_SCALAR; */
2094 /* Make $_ available to executed code. */
2095 if (PL_reg_sv != DEFSV) {
2096 /* SAVE_DEFSV does *not* suffice here for USE_5005THREADS */
2101 if (!(SvTYPE(PL_reg_sv) >= SVt_PVMG && SvMAGIC(PL_reg_sv)
2102 && (mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global)))) {
2103 /* prepare for quick setting of pos */
2104 sv_magic(PL_reg_sv, (SV*)0,
2105 PERL_MAGIC_regex_global, Nullch, 0);
2106 mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global);
2110 PL_reg_oldpos = mg->mg_len;
2111 SAVEDESTRUCTOR_X(restore_pos, 0);
2113 if (!PL_reg_curpm) {
2114 Newz(22,PL_reg_curpm, 1, PMOP);
2117 SV* repointer = newSViv(0);
2118 /* so we know which PL_regex_padav element is PL_reg_curpm */
2119 SvFLAGS(repointer) |= SVf_BREAK;
2120 av_push(PL_regex_padav,repointer);
2121 PL_reg_curpm->op_pmoffset = av_len(PL_regex_padav);
2122 PL_regex_pad = AvARRAY(PL_regex_padav);
2126 PM_SETRE(PL_reg_curpm, prog);
2127 PL_reg_oldcurpm = PL_curpm;
2128 PL_curpm = PL_reg_curpm;
2129 if (RX_MATCH_COPIED(prog)) {
2130 /* Here is a serious problem: we cannot rewrite subbeg,
2131 since it may be needed if this match fails. Thus
2132 $` inside (?{}) could fail... */
2133 PL_reg_oldsaved = prog->subbeg;
2134 PL_reg_oldsavedlen = prog->sublen;
2135 #ifdef PERL_COPY_ON_WRITE
2136 PL_nrs = prog->saved_copy;
2138 RX_MATCH_COPIED_off(prog);
2141 PL_reg_oldsaved = Nullch;
2142 prog->subbeg = PL_bostr;
2143 prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */
2145 prog->startp[0] = startpos - PL_bostr;
2146 PL_reginput = startpos;
2147 PL_regstartp = prog->startp;
2148 PL_regendp = prog->endp;
2149 PL_reglastparen = &prog->lastparen;
2150 PL_reglastcloseparen = &prog->lastcloseparen;
2151 prog->lastparen = 0;
2152 prog->lastcloseparen = 0;
2154 DEBUG_r(PL_reg_starttry = startpos);
2155 if (PL_reg_start_tmpl <= prog->nparens) {
2156 PL_reg_start_tmpl = prog->nparens*3/2 + 3;
2157 if(PL_reg_start_tmp)
2158 Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
2160 New(22,PL_reg_start_tmp, PL_reg_start_tmpl, char*);
2163 /* XXXX What this code is doing here?!!! There should be no need
2164 to do this again and again, PL_reglastparen should take care of
2167 /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code.
2168 * Actually, the code in regcppop() (which Ilya may be meaning by
2169 * PL_reglastparen), is not needed at all by the test suite
2170 * (op/regexp, op/pat, op/split), but that code is needed, oddly
2171 * enough, for building DynaLoader, or otherwise this
2172 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
2173 * will happen. Meanwhile, this code *is* needed for the
2174 * above-mentioned test suite tests to succeed. The common theme
2175 * on those tests seems to be returning null fields from matches.
2180 if (prog->nparens) {
2181 for (i = prog->nparens; i > (I32)*PL_reglastparen; i--) {
2188 if (regmatch(prog->program + 1)) {
2189 prog->endp[0] = PL_reginput - PL_bostr;
2192 REGCP_UNWIND(lastcp);
2196 #define RE_UNWIND_BRANCH 1
2197 #define RE_UNWIND_BRANCHJ 2
2201 typedef struct { /* XX: makes sense to enlarge it... */
2205 } re_unwind_generic_t;
2218 } re_unwind_branch_t;
2220 typedef union re_unwind_t {
2222 re_unwind_generic_t generic;
2223 re_unwind_branch_t branch;
2226 #define sayYES goto yes
2227 #define sayNO goto no
2228 #define sayNO_ANYOF goto no_anyof
2229 #define sayYES_FINAL goto yes_final
2230 #define sayYES_LOUD goto yes_loud
2231 #define sayNO_FINAL goto no_final
2232 #define sayNO_SILENT goto do_no
2233 #define saySAME(x) if (x) goto yes; else goto no
2235 #define REPORT_CODE_OFF 24
2238 - regmatch - main matching routine
2240 * Conceptually the strategy is simple: check to see whether the current
2241 * node matches, call self recursively to see whether the rest matches,
2242 * and then act accordingly. In practice we make some effort to avoid
2243 * recursion, in particular by going through "ordinary" nodes (that don't
2244 * need to know whether the rest of the match failed) by a loop instead of
2247 /* [lwall] I've hoisted the register declarations to the outer block in order to
2248 * maybe save a little bit of pushing and popping on the stack. It also takes
2249 * advantage of machines that use a register save mask on subroutine entry.
2251 STATIC I32 /* 0 failure, 1 success */
2252 S_regmatch(pTHX_ regnode *prog)
2254 register regnode *scan; /* Current node. */
2255 regnode *next; /* Next node. */
2256 regnode *inner; /* Next node in internal branch. */
2257 register I32 nextchr; /* renamed nextchr - nextchar colides with
2258 function of same name */
2259 register I32 n; /* no or next */
2260 register I32 ln = 0; /* len or last */
2261 register char *s = Nullch; /* operand or save */
2262 register char *locinput = PL_reginput;
2263 register I32 c1 = 0, c2 = 0, paren; /* case fold search, parenth */
2264 int minmod = 0, sw = 0, logical = 0;
2267 I32 firstcp = PL_savestack_ix;
2269 register bool do_utf8 = PL_reg_match_utf8;
2271 SV *dsv0 = PERL_DEBUG_PAD_ZERO(0);
2272 SV *dsv1 = PERL_DEBUG_PAD_ZERO(1);
2273 SV *dsv2 = PERL_DEBUG_PAD_ZERO(2);
2280 /* Note that nextchr is a byte even in UTF */
2281 nextchr = UCHARAT(locinput);
2283 while (scan != NULL) {
2286 SV *prop = sv_newmortal();
2287 int docolor = *PL_colors[0];
2288 int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */
2289 int l = (PL_regeol - locinput) > taill ? taill : (PL_regeol - locinput);
2290 /* The part of the string before starttry has one color
2291 (pref0_len chars), between starttry and current
2292 position another one (pref_len - pref0_len chars),
2293 after the current position the third one.
2294 We assume that pref0_len <= pref_len, otherwise we
2295 decrease pref0_len. */
2296 int pref_len = (locinput - PL_bostr) > (5 + taill) - l
2297 ? (5 + taill) - l : locinput - PL_bostr;
2300 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len)))
2302 pref0_len = pref_len - (locinput - PL_reg_starttry);
2303 if (l + pref_len < (5 + taill) && l < PL_regeol - locinput)
2304 l = ( PL_regeol - locinput > (5 + taill) - pref_len
2305 ? (5 + taill) - pref_len : PL_regeol - locinput);
2306 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput + l)))
2310 if (pref0_len > pref_len)
2311 pref0_len = pref_len;
2312 regprop(prop, scan);
2315 do_utf8 && OP(scan) != CANY ?
2316 pv_uni_display(dsv0, (U8*)(locinput - pref_len),
2317 pref0_len, 60, UNI_DISPLAY_REGEX) :
2318 locinput - pref_len;
2319 int len0 = do_utf8 ? strlen(s0) : pref0_len;
2320 char *s1 = do_utf8 && OP(scan) != CANY ?
2321 pv_uni_display(dsv1, (U8*)(locinput - pref_len + pref0_len),
2322 pref_len - pref0_len, 60, UNI_DISPLAY_REGEX) :
2323 locinput - pref_len + pref0_len;
2324 int len1 = do_utf8 ? strlen(s1) : pref_len - pref0_len;
2325 char *s2 = do_utf8 && OP(scan) != CANY ?
2326 pv_uni_display(dsv2, (U8*)locinput,
2327 PL_regeol - locinput, 60, UNI_DISPLAY_REGEX) :
2329 int len2 = do_utf8 ? strlen(s2) : l;
2330 PerlIO_printf(Perl_debug_log,
2331 "%4"IVdf" <%s%.*s%s%s%.*s%s%s%s%.*s%s>%*s|%3"IVdf":%*s%s\n",
2332 (IV)(locinput - PL_bostr),
2339 (docolor ? "" : "> <"),
2343 15 - l - pref_len + 1,
2345 (IV)(scan - PL_regprogram), PL_regindent*2, "",
2350 next = scan + NEXT_OFF(scan);
2356 if (locinput == PL_bostr || (PL_multiline &&
2357 (nextchr || locinput < PL_regeol) && locinput[-1] == '\n') )
2359 /* regtill = regbol; */
2364 if (locinput == PL_bostr ||
2365 ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n'))
2371 if (locinput == PL_bostr)
2375 if (locinput == PL_reg_ganch)
2385 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2390 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2392 if (PL_regeol - locinput > 1)
2396 if (PL_regeol != locinput)
2400 if (!nextchr && locinput >= PL_regeol)
2403 locinput += PL_utf8skip[nextchr];
2404 if (locinput > PL_regeol)
2406 nextchr = UCHARAT(locinput);
2409 nextchr = UCHARAT(++locinput);
2412 if (!nextchr && locinput >= PL_regeol)
2414 nextchr = UCHARAT(++locinput);
2417 if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n')
2420 locinput += PL_utf8skip[nextchr];
2421 if (locinput > PL_regeol)
2423 nextchr = UCHARAT(locinput);
2426 nextchr = UCHARAT(++locinput);
2431 if (do_utf8 != UTF) {
2432 /* The target and the pattern have differing utf8ness. */
2438 /* The target is utf8, the pattern is not utf8. */
2442 if (NATIVE_TO_UNI(*(U8*)s) !=
2443 utf8n_to_uvuni((U8*)l, UTF8_MAXLEN, &ulen,
2445 0 : UTF8_ALLOW_ANY))
2452 /* The target is not utf8, the pattern is utf8. */
2456 if (NATIVE_TO_UNI(*((U8*)l)) !=
2457 utf8n_to_uvuni((U8*)s, UTF8_MAXLEN, &ulen,
2459 0 : UTF8_ALLOW_ANY))
2466 nextchr = UCHARAT(locinput);
2469 /* The target and the pattern have the same utf8ness. */
2470 /* Inline the first character, for speed. */
2471 if (UCHARAT(s) != nextchr)
2473 if (PL_regeol - locinput < ln)
2475 if (ln > 1 && memNE(s, locinput, ln))
2478 nextchr = UCHARAT(locinput);
2481 PL_reg_flags |= RF_tainted;
2487 if (do_utf8 || UTF) {
2488 /* Either target or the pattern are utf8. */
2490 char *e = PL_regeol;
2492 if (ibcmp_utf8(s, 0, ln, (bool)UTF,
2493 l, &e, 0, do_utf8)) {
2494 /* One more case for the sharp s:
2495 * pack("U0U*", 0xDF) =~ /ss/i,
2496 * the 0xC3 0x9F are the UTF-8
2497 * byte sequence for the U+00DF. */
2499 toLOWER(s[0]) == 's' &&
2501 toLOWER(s[1]) == 's' &&
2508 nextchr = UCHARAT(locinput);
2512 /* Neither the target and the pattern are utf8. */
2514 /* Inline the first character, for speed. */
2515 if (UCHARAT(s) != nextchr &&
2516 UCHARAT(s) != ((OP(scan) == EXACTF)
2517 ? PL_fold : PL_fold_locale)[nextchr])
2519 if (PL_regeol - locinput < ln)
2521 if (ln > 1 && (OP(scan) == EXACTF
2522 ? ibcmp(s, locinput, ln)
2523 : ibcmp_locale(s, locinput, ln)))
2526 nextchr = UCHARAT(locinput);
2530 STRLEN inclasslen = PL_regeol - locinput;
2532 if (!reginclass(scan, (U8*)locinput, &inclasslen, do_utf8))
2534 if (locinput >= PL_regeol)
2536 locinput += inclasslen ? inclasslen : UTF8SKIP(locinput);
2537 nextchr = UCHARAT(locinput);
2542 nextchr = UCHARAT(locinput);
2543 if (!REGINCLASS(scan, (U8*)locinput))
2545 if (!nextchr && locinput >= PL_regeol)
2547 nextchr = UCHARAT(++locinput);
2551 /* If we might have the case of the German sharp s
2552 * in a casefolding Unicode character class. */
2554 if (ANYOF_FOLD_SHARP_S(scan, locinput, PL_regeol)) {
2555 locinput += SHARP_S_SKIP;
2556 nextchr = UCHARAT(locinput);
2562 PL_reg_flags |= RF_tainted;
2568 LOAD_UTF8_CHARCLASS(alnum,"a");
2569 if (!(OP(scan) == ALNUM
2570 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2571 : isALNUM_LC_utf8((U8*)locinput)))
2575 locinput += PL_utf8skip[nextchr];
2576 nextchr = UCHARAT(locinput);
2579 if (!(OP(scan) == ALNUM
2580 ? isALNUM(nextchr) : isALNUM_LC(nextchr)))
2582 nextchr = UCHARAT(++locinput);
2585 PL_reg_flags |= RF_tainted;
2588 if (!nextchr && locinput >= PL_regeol)
2591 LOAD_UTF8_CHARCLASS(alnum,"a");
2592 if (OP(scan) == NALNUM
2593 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2594 : isALNUM_LC_utf8((U8*)locinput))
2598 locinput += PL_utf8skip[nextchr];
2599 nextchr = UCHARAT(locinput);
2602 if (OP(scan) == NALNUM
2603 ? isALNUM(nextchr) : isALNUM_LC(nextchr))
2605 nextchr = UCHARAT(++locinput);
2609 PL_reg_flags |= RF_tainted;
2613 /* was last char in word? */
2615 if (locinput == PL_bostr)
2618 U8 *r = reghop3((U8*)locinput, -1, (U8*)PL_bostr);
2620 ln = utf8n_to_uvchr(r, UTF8SKIP(r), 0, 0);
2622 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2623 ln = isALNUM_uni(ln);
2624 LOAD_UTF8_CHARCLASS(alnum,"a");
2625 n = swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8);
2628 ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(ln));
2629 n = isALNUM_LC_utf8((U8*)locinput);
2633 ln = (locinput != PL_bostr) ?
2634 UCHARAT(locinput - 1) : '\n';
2635 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2637 n = isALNUM(nextchr);
2640 ln = isALNUM_LC(ln);
2641 n = isALNUM_LC(nextchr);
2644 if (((!ln) == (!n)) == (OP(scan) == BOUND ||
2645 OP(scan) == BOUNDL))
2649 PL_reg_flags |= RF_tainted;
2655 if (UTF8_IS_CONTINUED(nextchr)) {
2656 LOAD_UTF8_CHARCLASS(space," ");
2657 if (!(OP(scan) == SPACE
2658 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2659 : isSPACE_LC_utf8((U8*)locinput)))
2663 locinput += PL_utf8skip[nextchr];
2664 nextchr = UCHARAT(locinput);
2667 if (!(OP(scan) == SPACE
2668 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2670 nextchr = UCHARAT(++locinput);
2673 if (!(OP(scan) == SPACE
2674 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2676 nextchr = UCHARAT(++locinput);
2680 PL_reg_flags |= RF_tainted;
2683 if (!nextchr && locinput >= PL_regeol)
2686 LOAD_UTF8_CHARCLASS(space," ");
2687 if (OP(scan) == NSPACE
2688 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2689 : isSPACE_LC_utf8((U8*)locinput))
2693 locinput += PL_utf8skip[nextchr];
2694 nextchr = UCHARAT(locinput);
2697 if (OP(scan) == NSPACE
2698 ? isSPACE(nextchr) : isSPACE_LC(nextchr))
2700 nextchr = UCHARAT(++locinput);
2703 PL_reg_flags |= RF_tainted;
2709 LOAD_UTF8_CHARCLASS(digit,"0");
2710 if (!(OP(scan) == DIGIT
2711 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2712 : isDIGIT_LC_utf8((U8*)locinput)))
2716 locinput += PL_utf8skip[nextchr];
2717 nextchr = UCHARAT(locinput);
2720 if (!(OP(scan) == DIGIT
2721 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr)))
2723 nextchr = UCHARAT(++locinput);
2726 PL_reg_flags |= RF_tainted;
2729 if (!nextchr && locinput >= PL_regeol)
2732 LOAD_UTF8_CHARCLASS(digit,"0");
2733 if (OP(scan) == NDIGIT
2734 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2735 : isDIGIT_LC_utf8((U8*)locinput))
2739 locinput += PL_utf8skip[nextchr];
2740 nextchr = UCHARAT(locinput);
2743 if (OP(scan) == NDIGIT
2744 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr))
2746 nextchr = UCHARAT(++locinput);
2749 if (locinput >= PL_regeol)
2752 LOAD_UTF8_CHARCLASS(mark,"~");
2753 if (swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2755 locinput += PL_utf8skip[nextchr];
2756 while (locinput < PL_regeol &&
2757 swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2758 locinput += UTF8SKIP(locinput);
2759 if (locinput > PL_regeol)
2764 nextchr = UCHARAT(locinput);
2767 PL_reg_flags |= RF_tainted;
2771 n = ARG(scan); /* which paren pair */
2772 ln = PL_regstartp[n];
2773 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2774 if ((I32)*PL_reglastparen < n || ln == -1)
2775 sayNO; /* Do not match unless seen CLOSEn. */
2776 if (ln == PL_regendp[n])
2780 if (do_utf8 && OP(scan) != REF) { /* REF can do byte comparison */
2782 char *e = PL_bostr + PL_regendp[n];
2784 * Note that we can't do the "other character" lookup trick as
2785 * in the 8-bit case (no pun intended) because in Unicode we
2786 * have to map both upper and title case to lower case.
2788 if (OP(scan) == REFF) {
2789 STRLEN ulen1, ulen2;
2790 U8 tmpbuf1[UTF8_MAXLEN_UCLC+1];
2791 U8 tmpbuf2[UTF8_MAXLEN_UCLC+1];
2795 toLOWER_utf8((U8*)s, tmpbuf1, &ulen1);
2796 toLOWER_utf8((U8*)l, tmpbuf2, &ulen2);
2797 if (ulen1 != ulen2 || memNE((char *)tmpbuf1, (char *)tmpbuf2, ulen1))
2804 nextchr = UCHARAT(locinput);
2808 /* Inline the first character, for speed. */
2809 if (UCHARAT(s) != nextchr &&
2811 (UCHARAT(s) != ((OP(scan) == REFF
2812 ? PL_fold : PL_fold_locale)[nextchr]))))
2814 ln = PL_regendp[n] - ln;
2815 if (locinput + ln > PL_regeol)
2817 if (ln > 1 && (OP(scan) == REF
2818 ? memNE(s, locinput, ln)
2820 ? ibcmp(s, locinput, ln)
2821 : ibcmp_locale(s, locinput, ln))))
2824 nextchr = UCHARAT(locinput);
2835 OP_4tree *oop = PL_op;
2836 COP *ocurcop = PL_curcop;
2839 struct regexp *oreg = PL_reg_re;
2842 PL_op = (OP_4tree*)PL_regdata->data[n];
2843 DEBUG_r( PerlIO_printf(Perl_debug_log, " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) );
2844 PAD_SAVE_LOCAL(old_comppad, (PAD*)PL_regdata->data[n + 2]);
2845 PL_regendp[0] = PL_reg_magic->mg_len = locinput - PL_bostr;
2849 CALLRUNOPS(aTHX); /* Scalar context. */
2852 ret = &PL_sv_undef; /* protect against empty (?{}) blocks. */
2860 PAD_RESTORE_LOCAL(old_comppad);
2861 PL_curcop = ocurcop;
2863 if (logical == 2) { /* Postponed subexpression. */
2865 MAGIC *mg = Null(MAGIC*);
2867 CHECKPOINT cp, lastcp;
2871 if(SvROK(ret) && SvSMAGICAL(sv = SvRV(ret)))
2872 mg = mg_find(sv, PERL_MAGIC_qr);
2873 else if (SvSMAGICAL(ret)) {
2874 if (SvGMAGICAL(ret))
2875 sv_unmagic(ret, PERL_MAGIC_qr);
2877 mg = mg_find(ret, PERL_MAGIC_qr);
2881 re = (regexp *)mg->mg_obj;
2882 (void)ReREFCNT_inc(re);
2886 char *t = SvPV(ret, len);
2888 char *oprecomp = PL_regprecomp;
2889 I32 osize = PL_regsize;
2890 I32 onpar = PL_regnpar;
2893 if (DO_UTF8(ret)) pm.op_pmdynflags |= PMdf_DYN_UTF8;
2894 re = CALLREGCOMP(aTHX_ t, t + len, &pm);
2896 & (SVs_TEMP | SVs_PADTMP | SVf_READONLY
2898 sv_magic(ret,(SV*)ReREFCNT_inc(re),
2900 PL_regprecomp = oprecomp;
2905 PerlIO_printf(Perl_debug_log,
2906 "Entering embedded `%s%.60s%s%s'\n",
2910 (strlen(re->precomp) > 60 ? "..." : ""))
2913 state.prev = PL_reg_call_cc;
2914 state.cc = PL_regcc;
2915 state.re = PL_reg_re;
2919 cp = regcppush(0); /* Save *all* the positions. */
2922 state.ss = PL_savestack_ix;
2923 *PL_reglastparen = 0;
2924 *PL_reglastcloseparen = 0;
2925 PL_reg_call_cc = &state;
2926 PL_reginput = locinput;
2927 toggleutf = ((PL_reg_flags & RF_utf8) != 0) ^
2928 ((re->reganch & ROPT_UTF8) != 0);
2929 if (toggleutf) PL_reg_flags ^= RF_utf8;
2931 /* XXXX This is too dramatic a measure... */
2934 if (regmatch(re->program + 1)) {
2935 /* Even though we succeeded, we need to restore
2936 global variables, since we may be wrapped inside
2937 SUSPEND, thus the match may be not finished yet. */
2939 /* XXXX Do this only if SUSPENDed? */
2940 PL_reg_call_cc = state.prev;
2941 PL_regcc = state.cc;
2942 PL_reg_re = state.re;
2943 cache_re(PL_reg_re);
2944 if (toggleutf) PL_reg_flags ^= RF_utf8;
2946 /* XXXX This is too dramatic a measure... */
2949 /* These are needed even if not SUSPEND. */
2955 REGCP_UNWIND(lastcp);
2957 PL_reg_call_cc = state.prev;
2958 PL_regcc = state.cc;
2959 PL_reg_re = state.re;
2960 cache_re(PL_reg_re);
2961 if (toggleutf) PL_reg_flags ^= RF_utf8;
2963 /* XXXX This is too dramatic a measure... */
2973 sv_setsv(save_scalar(PL_replgv), ret);
2979 n = ARG(scan); /* which paren pair */
2980 PL_reg_start_tmp[n] = locinput;
2985 n = ARG(scan); /* which paren pair */
2986 PL_regstartp[n] = PL_reg_start_tmp[n] - PL_bostr;
2987 PL_regendp[n] = locinput - PL_bostr;
2988 if (n > (I32)*PL_reglastparen)
2989 *PL_reglastparen = n;
2990 *PL_reglastcloseparen = n;
2993 n = ARG(scan); /* which paren pair */
2994 sw = ((I32)*PL_reglastparen >= n && PL_regendp[n] != -1);
2997 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2999 next = NEXTOPER(NEXTOPER(scan));
3001 next = scan + ARG(scan);
3002 if (OP(next) == IFTHEN) /* Fake one. */
3003 next = NEXTOPER(NEXTOPER(next));
3007 logical = scan->flags;
3009 /*******************************************************************
3010 PL_regcc contains infoblock about the innermost (...)* loop, and
3011 a pointer to the next outer infoblock.
3013 Here is how Y(A)*Z is processed (if it is compiled into CURLYX/WHILEM):
3015 1) After matching X, regnode for CURLYX is processed;
3017 2) This regnode creates infoblock on the stack, and calls
3018 regmatch() recursively with the starting point at WHILEM node;
3020 3) Each hit of WHILEM node tries to match A and Z (in the order
3021 depending on the current iteration, min/max of {min,max} and
3022 greediness). The information about where are nodes for "A"
3023 and "Z" is read from the infoblock, as is info on how many times "A"
3024 was already matched, and greediness.
3026 4) After A matches, the same WHILEM node is hit again.
3028 5) Each time WHILEM is hit, PL_regcc is the infoblock created by CURLYX
3029 of the same pair. Thus when WHILEM tries to match Z, it temporarily
3030 resets PL_regcc, since this Y(A)*Z can be a part of some other loop:
3031 as in (Y(A)*Z)*. If Z matches, the automaton will hit the WHILEM node
3032 of the external loop.
3034 Currently present infoblocks form a tree with a stem formed by PL_curcc
3035 and whatever it mentions via ->next, and additional attached trees
3036 corresponding to temporarily unset infoblocks as in "5" above.
3038 In the following picture infoblocks for outer loop of
3039 (Y(A)*?Z)*?T are denoted O, for inner I. NULL starting block
3040 is denoted by x. The matched string is YAAZYAZT. Temporarily postponed
3041 infoblocks are drawn below the "reset" infoblock.
3043 In fact in the picture below we do not show failed matches for Z and T
3044 by WHILEM blocks. [We illustrate minimal matches, since for them it is
3045 more obvious *why* one needs to *temporary* unset infoblocks.]
3047 Matched REx position InfoBlocks Comment
3051 Y A)*?Z)*?T x <- O <- I
3052 YA )*?Z)*?T x <- O <- I
3053 YA A)*?Z)*?T x <- O <- I
3054 YAA )*?Z)*?T x <- O <- I
3055 YAA Z)*?T x <- O # Temporary unset I
3058 YAAZ Y(A)*?Z)*?T x <- O
3061 YAAZY (A)*?Z)*?T x <- O
3064 YAAZY A)*?Z)*?T x <- O <- I
3067 YAAZYA )*?Z)*?T x <- O <- I
3070 YAAZYA Z)*?T x <- O # Temporary unset I
3076 YAAZYAZ T x # Temporary unset O
3083 *******************************************************************/
3086 CHECKPOINT cp = PL_savestack_ix;
3087 /* No need to save/restore up to this paren */
3088 I32 parenfloor = scan->flags;
3090 if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */
3092 cc.oldcc = PL_regcc;
3094 /* XXXX Probably it is better to teach regpush to support
3095 parenfloor > PL_regsize... */
3096 if (parenfloor > (I32)*PL_reglastparen)
3097 parenfloor = *PL_reglastparen; /* Pessimization... */
3098 cc.parenfloor = parenfloor;
3100 cc.min = ARG1(scan);
3101 cc.max = ARG2(scan);
3102 cc.scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
3106 PL_reginput = locinput;
3107 n = regmatch(PREVOPER(next)); /* start on the WHILEM */
3109 PL_regcc = cc.oldcc;
3115 * This is really hard to understand, because after we match
3116 * what we're trying to match, we must make sure the rest of
3117 * the REx is going to match for sure, and to do that we have
3118 * to go back UP the parse tree by recursing ever deeper. And
3119 * if it fails, we have to reset our parent's current state
3120 * that we can try again after backing off.
3123 CHECKPOINT cp, lastcp;
3124 CURCUR* cc = PL_regcc;
3125 char *lastloc = cc->lastloc; /* Detection of 0-len. */
3127 n = cc->cur + 1; /* how many we know we matched */
3128 PL_reginput = locinput;
3131 PerlIO_printf(Perl_debug_log,
3132 "%*s %ld out of %ld..%ld cc=%"UVxf"\n",
3133 REPORT_CODE_OFF+PL_regindent*2, "",
3134 (long)n, (long)cc->min,
3135 (long)cc->max, PTR2UV(cc))
3138 /* If degenerate scan matches "", assume scan done. */
3140 if (locinput == cc->lastloc && n >= cc->min) {
3141 PL_regcc = cc->oldcc;
3145 PerlIO_printf(Perl_debug_log,
3146 "%*s empty match detected, try continuation...\n",
3147 REPORT_CODE_OFF+PL_regindent*2, "")
3149 if (regmatch(cc->next))
3157 /* First just match a string of min scans. */
3161 cc->lastloc = locinput;
3162 if (regmatch(cc->scan))
3165 cc->lastloc = lastloc;
3170 /* Check whether we already were at this position.
3171 Postpone detection until we know the match is not
3172 *that* much linear. */
3173 if (!PL_reg_maxiter) {
3174 PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4);
3175 PL_reg_leftiter = PL_reg_maxiter;
3177 if (PL_reg_leftiter-- == 0) {
3178 I32 size = (PL_reg_maxiter + 7)/8;
3179 if (PL_reg_poscache) {
3180 if ((I32)PL_reg_poscache_size < size) {
3181 Renew(PL_reg_poscache, size, char);
3182 PL_reg_poscache_size = size;
3184 Zero(PL_reg_poscache, size, char);
3187 PL_reg_poscache_size = size;
3188 Newz(29, PL_reg_poscache, size, char);
3191 PerlIO_printf(Perl_debug_log,
3192 "%sDetected a super-linear match, switching on caching%s...\n",
3193 PL_colors[4], PL_colors[5])
3196 if (PL_reg_leftiter < 0) {
3197 I32 o = locinput - PL_bostr, b;
3199 o = (scan->flags & 0xf) - 1 + o * (scan->flags>>4);
3202 if (PL_reg_poscache[o] & (1<<b)) {
3204 PerlIO_printf(Perl_debug_log,
3205 "%*s already tried at this position...\n",
3206 REPORT_CODE_OFF+PL_regindent*2, "")
3208 if (PL_reg_flags & RF_false)
3213 PL_reg_poscache[o] |= (1<<b);
3217 /* Prefer next over scan for minimal matching. */
3220 PL_regcc = cc->oldcc;
3223 cp = regcppush(cc->parenfloor);
3225 if (regmatch(cc->next)) {
3227 sayYES; /* All done. */
3229 REGCP_UNWIND(lastcp);
3235 if (n >= cc->max) { /* Maximum greed exceeded? */
3236 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3237 && !(PL_reg_flags & RF_warned)) {
3238 PL_reg_flags |= RF_warned;
3239 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s limit (%d) exceeded",
3240 "Complex regular subexpression recursion",
3247 PerlIO_printf(Perl_debug_log,
3248 "%*s trying longer...\n",
3249 REPORT_CODE_OFF+PL_regindent*2, "")
3251 /* Try scanning more and see if it helps. */
3252 PL_reginput = locinput;
3254 cc->lastloc = locinput;
3255 cp = regcppush(cc->parenfloor);
3257 if (regmatch(cc->scan)) {
3261 REGCP_UNWIND(lastcp);
3264 cc->lastloc = lastloc;
3268 /* Prefer scan over next for maximal matching. */
3270 if (n < cc->max) { /* More greed allowed? */
3271 cp = regcppush(cc->parenfloor);
3273 cc->lastloc = locinput;
3275 if (regmatch(cc->scan)) {
3279 REGCP_UNWIND(lastcp);
3280 regcppop(); /* Restore some previous $<digit>s? */
3281 PL_reginput = locinput;
3283 PerlIO_printf(Perl_debug_log,
3284 "%*s failed, try continuation...\n",
3285 REPORT_CODE_OFF+PL_regindent*2, "")
3288 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3289 && !(PL_reg_flags & RF_warned)) {
3290 PL_reg_flags |= RF_warned;
3291 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s limit (%d) exceeded",
3292 "Complex regular subexpression recursion",
3296 /* Failed deeper matches of scan, so see if this one works. */
3297 PL_regcc = cc->oldcc;
3300 if (regmatch(cc->next))
3306 cc->lastloc = lastloc;
3311 next = scan + ARG(scan);
3314 inner = NEXTOPER(NEXTOPER(scan));
3317 inner = NEXTOPER(scan);
3321 if (OP(next) != c1) /* No choice. */
3322 next = inner; /* Avoid recursion. */
3324 I32 lastparen = *PL_reglastparen;
3326 re_unwind_branch_t *uw;
3328 /* Put unwinding data on stack */
3329 unwind1 = SSNEWt(1,re_unwind_branch_t);
3330 uw = SSPTRt(unwind1,re_unwind_branch_t);
3333 uw->type = ((c1 == BRANCH)
3335 : RE_UNWIND_BRANCHJ);
3336 uw->lastparen = lastparen;
3338 uw->locinput = locinput;
3339 uw->nextchr = nextchr;
3341 uw->regindent = ++PL_regindent;
3344 REGCP_SET(uw->lastcp);
3346 /* Now go into the first branch */
3359 /* We suppose that the next guy does not need
3360 backtracking: in particular, it is of constant length,
3361 and has no parenths to influence future backrefs. */
3362 ln = ARG1(scan); /* min to match */
3363 n = ARG2(scan); /* max to match */
3364 paren = scan->flags;
3366 if (paren > PL_regsize)
3368 if (paren > (I32)*PL_reglastparen)
3369 *PL_reglastparen = paren;
3371 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3373 scan += NEXT_OFF(scan); /* Skip former OPEN. */
3374 PL_reginput = locinput;
3377 if (ln && regrepeat_hard(scan, ln, &l) < ln)
3379 /* if we matched something zero-length we don't need to
3380 backtrack - capturing parens are already defined, so
3381 the caveat in the maximal case doesn't apply
3383 XXXX if ln == 0, we can redo this check first time
3384 through the following loop
3387 n = ln; /* don't backtrack */
3388 locinput = PL_reginput;
3389 if (HAS_TEXT(next) || JUMPABLE(next)) {
3390 regnode *text_node = next;
3392 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3394 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3396 if (PL_regkind[(U8)OP(text_node)] == REF) {
3400 else { c1 = (U8)*STRING(text_node); }
3401 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3403 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3404 c2 = PL_fold_locale[c1];
3413 /* This may be improved if l == 0. */
3414 while (n >= ln || (n == REG_INFTY && ln > 0 && l)) { /* ln overflow ? */
3415 /* If it could work, try it. */
3417 UCHARAT(PL_reginput) == c1 ||
3418 UCHARAT(PL_reginput) == c2)
3422 PL_regstartp[paren] =
3423 HOPc(PL_reginput, -l) - PL_bostr;
3424 PL_regendp[paren] = PL_reginput - PL_bostr;
3427 PL_regendp[paren] = -1;
3431 REGCP_UNWIND(lastcp);
3433 /* Couldn't or didn't -- move forward. */
3434 PL_reginput = locinput;
3435 if (regrepeat_hard(scan, 1, &l)) {
3437 locinput = PL_reginput;
3444 n = regrepeat_hard(scan, n, &l);
3445 /* if we matched something zero-length we don't need to
3446 backtrack, unless the minimum count is zero and we
3447 are capturing the result - in that case the capture
3448 being defined or not may affect later execution
3450 if (n != 0 && l == 0 && !(paren && ln == 0))
3451 ln = n; /* don't backtrack */
3452 locinput = PL_reginput;
3454 PerlIO_printf(Perl_debug_log,
3455 "%*s matched %"IVdf" times, len=%"IVdf"...\n",
3456 (int)(REPORT_CODE_OFF+PL_regindent*2), "",
3460 if (HAS_TEXT(next) || JUMPABLE(next)) {
3461 regnode *text_node = next;
3463 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3465 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3467 if (PL_regkind[(U8)OP(text_node)] == REF) {
3471 else { c1 = (U8)*STRING(text_node); }
3473 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3475 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3476 c2 = PL_fold_locale[c1];
3487 /* If it could work, try it. */
3489 UCHARAT(PL_reginput) == c1 ||
3490 UCHARAT(PL_reginput) == c2)
3493 PerlIO_printf(Perl_debug_log,
3494 "%*s trying tail with n=%"IVdf"...\n",
3495 (int)(REPORT_CODE_OFF+PL_regindent*2), "", (IV)n)
3499 PL_regstartp[paren] = HOPc(PL_reginput, -l) - PL_bostr;
3500 PL_regendp[paren] = PL_reginput - PL_bostr;
3503 PL_regendp[paren] = -1;
3507 REGCP_UNWIND(lastcp);
3509 /* Couldn't or didn't -- back up. */
3511 locinput = HOPc(locinput, -l);
3512 PL_reginput = locinput;
3519 paren = scan->flags; /* Which paren to set */
3520 if (paren > PL_regsize)
3522 if (paren > (I32)*PL_reglastparen)
3523 *PL_reglastparen = paren;
3524 ln = ARG1(scan); /* min to match */
3525 n = ARG2(scan); /* max to match */
3526 scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE);
3530 ln = ARG1(scan); /* min to match */
3531 n = ARG2(scan); /* max to match */
3532 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3537 scan = NEXTOPER(scan);
3543 scan = NEXTOPER(scan);
3547 * Lookahead to avoid useless match attempts
3548 * when we know what character comes next.
3552 * Used to only do .*x and .*?x, but now it allows
3553 * for )'s, ('s and (?{ ... })'s to be in the way
3554 * of the quantifier and the EXACT-like node. -- japhy
3557 if (HAS_TEXT(next) || JUMPABLE(next)) {
3559 regnode *text_node = next;
3561 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3563 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3565 if (PL_regkind[(U8)OP(text_node)] == REF) {
3567 goto assume_ok_easy;
3569 else { s = (U8*)STRING(text_node); }
3573 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3575 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3576 c2 = PL_fold_locale[c1];
3579 if (OP(text_node) == EXACTF || OP(text_node) == REFF) {
3580 STRLEN ulen1, ulen2;
3581 U8 tmpbuf1[UTF8_MAXLEN_UCLC+1];
3582 U8 tmpbuf2[UTF8_MAXLEN_UCLC+1];
3584 to_utf8_lower((U8*)s, tmpbuf1, &ulen1);
3585 to_utf8_upper((U8*)s, tmpbuf2, &ulen2);
3587 c1 = utf8n_to_uvuni(tmpbuf1, UTF8_MAXLEN, 0,
3589 0 : UTF8_ALLOW_ANY);
3590 c2 = utf8n_to_uvuni(tmpbuf2, UTF8_MAXLEN, 0,
3592 0 : UTF8_ALLOW_ANY);
3595 c2 = c1 = utf8n_to_uvchr(s, UTF8_MAXLEN, 0,
3597 0 : UTF8_ALLOW_ANY);
3605 PL_reginput = locinput;
3609 if (ln && regrepeat(scan, ln) < ln)
3611 locinput = PL_reginput;
3614 char *e; /* Should not check after this */
3615 char *old = locinput;
3618 if (n == REG_INFTY) {
3621 while (UTF8_IS_CONTINUATION(*(U8*)e))
3627 m >0 && e + UTF8SKIP(e) <= PL_regeol; m--)
3631 e = locinput + n - ln;
3636 /* Find place 'next' could work */
3639 while (locinput <= e &&
3640 UCHARAT(locinput) != c1)
3643 while (locinput <= e
3644 && UCHARAT(locinput) != c1
3645 && UCHARAT(locinput) != c2)
3648 count = locinput - old;
3653 /* count initialised to
3654 * utf8_distance(old, locinput) */
3655 while (locinput <= e &&
3656 utf8n_to_uvchr((U8*)locinput,
3659 0 : UTF8_ALLOW_ANY) != (UV)c1) {
3664 /* count initialised to
3665 * utf8_distance(old, locinput) */
3666 while (locinput <= e) {
3667 UV c = utf8n_to_uvchr((U8*)locinput,
3670 0 : UTF8_ALLOW_ANY);
3671 if (c == (UV)c1 || c == (UV)c2)
3680 /* PL_reginput == old now */
3681 if (locinput != old) {
3682 ln = 1; /* Did some */
3683 if (regrepeat(scan, count) < count)
3686 /* PL_reginput == locinput now */
3687 TRYPAREN(paren, ln, locinput);
3688 PL_reginput = locinput; /* Could be reset... */
3689 REGCP_UNWIND(lastcp);
3690 /* Couldn't or didn't -- move forward. */
3693 locinput += UTF8SKIP(locinput);
3700 while (n >= ln || (n == REG_INFTY && ln > 0)) { /* ln overflow ? */
3704 c = utf8n_to_uvchr((U8*)PL_reginput,
3707 0 : UTF8_ALLOW_ANY);
3709 c = UCHARAT(PL_reginput);
3710 /* If it could work, try it. */
3711 if (c == (UV)c1 || c == (UV)c2)
3713 TRYPAREN(paren, ln, PL_reginput);
3714 REGCP_UNWIND(lastcp);
3717 /* If it could work, try it. */
3718 else if (c1 == -1000)
3720 TRYPAREN(paren, ln, PL_reginput);
3721 REGCP_UNWIND(lastcp);
3723 /* Couldn't or didn't -- move forward. */
3724 PL_reginput = locinput;
3725 if (regrepeat(scan, 1)) {
3727 locinput = PL_reginput;
3735 n = regrepeat(scan, n);
3736 locinput = PL_reginput;
3737 if (ln < n && PL_regkind[(U8)OP(next)] == EOL &&
3738 ((!PL_multiline && OP(next) != MEOL) ||
3739 OP(next) == SEOL || OP(next) == EOS))
3741 ln = n; /* why back off? */
3742 /* ...because $ and \Z can match before *and* after
3743 newline at the end. Consider "\n\n" =~ /\n+\Z\n/.
3744 We should back off by one in this case. */
3745 if (UCHARAT(PL_reginput - 1) == '\n' && OP(next) != EOS)
3754 c = utf8n_to_uvchr((U8*)PL_reginput,
3757 0 : UTF8_ALLOW_ANY);
3759 c = UCHARAT(PL_reginput);
3761 /* If it could work, try it. */
3762 if (c1 == -1000 || c == (UV)c1 || c == (UV)c2)
3764 TRYPAREN(paren, n, PL_reginput);
3765 REGCP_UNWIND(lastcp);
3767 /* Couldn't or didn't -- back up. */
3769 PL_reginput = locinput = HOPc(locinput, -1);
3777 c = utf8n_to_uvchr((U8*)PL_reginput,
3780 0 : UTF8_ALLOW_ANY);
3782 c = UCHARAT(PL_reginput);
3784 /* If it could work, try it. */
3785 if (c1 == -1000 || c == (UV)c1 || c == (UV)c2)
3787 TRYPAREN(paren, n, PL_reginput);
3788 REGCP_UNWIND(lastcp);
3790 /* Couldn't or didn't -- back up. */
3792 PL_reginput = locinput = HOPc(locinput, -1);
3799 if (PL_reg_call_cc) {
3800 re_cc_state *cur_call_cc = PL_reg_call_cc;
3801 CURCUR *cctmp = PL_regcc;
3802 regexp *re = PL_reg_re;
3803 CHECKPOINT cp, lastcp;
3805 cp = regcppush(0); /* Save *all* the positions. */
3807 regcp_set_to(PL_reg_call_cc->ss); /* Restore parens of
3809 PL_reginput = locinput; /* Make position available to
3811 cache_re(PL_reg_call_cc->re);
3812 PL_regcc = PL_reg_call_cc->cc;
3813 PL_reg_call_cc = PL_reg_call_cc->prev;
3814 if (regmatch(cur_call_cc->node)) {
3815 PL_reg_call_cc = cur_call_cc;
3819 REGCP_UNWIND(lastcp);
3821 PL_reg_call_cc = cur_call_cc;
3827 PerlIO_printf(Perl_debug_log,
3828 "%*s continuation failed...\n",
3829 REPORT_CODE_OFF+PL_regindent*2, "")
3833 if (locinput < PL_regtill) {
3834 DEBUG_r(PerlIO_printf(Perl_debug_log,
3835 "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n",
3837 (long)(locinput - PL_reg_starttry),
3838 (long)(PL_regtill - PL_reg_starttry),
3840 sayNO_FINAL; /* Cannot match: too short. */
3842 PL_reginput = locinput; /* put where regtry can find it */
3843 sayYES_FINAL; /* Success! */
3845 PL_reginput = locinput; /* put where regtry can find it */
3846 sayYES_LOUD; /* Success! */
3849 PL_reginput = locinput;
3854 s = HOPBACKc(locinput, scan->flags);
3860 PL_reginput = locinput;
3861 PL_reg_flags ^= RF_false;
3866 s = HOPBACKc(locinput, scan->flags);
3872 PL_reginput = locinput;
3875 inner = NEXTOPER(NEXTOPER(scan));
3876 if (regmatch(inner) != n) {
3878 PL_reg_flags ^= RF_false;
3889 PL_reg_flags ^= RF_false;
3895 if (OP(scan) == SUSPEND) {
3896 locinput = PL_reginput;
3897 nextchr = UCHARAT(locinput);
3902 next = scan + ARG(scan);
3907 PerlIO_printf(Perl_error_log, "%"UVxf" %d\n",
3908 PTR2UV(scan), OP(scan));
3909 Perl_croak(aTHX_ "regexp memory corruption");
3916 * We get here only if there's trouble -- normally "case END" is
3917 * the terminating point.
3919 Perl_croak(aTHX_ "corrupted regexp pointers");
3925 PerlIO_printf(Perl_debug_log,
3926 "%*s %scould match...%s\n",
3927 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4],PL_colors[5])
3931 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n",
3932 PL_colors[4],PL_colors[5]));
3938 #if 0 /* Breaks $^R */
3946 PerlIO_printf(Perl_debug_log,
3947 "%*s %sfailed...%s\n",
3948 REPORT_CODE_OFF+PL_regindent*2, "",PL_colors[4],PL_colors[5])
3954 re_unwind_t *uw = SSPTRt(unwind,re_unwind_t);
3957 case RE_UNWIND_BRANCH:
3958 case RE_UNWIND_BRANCHJ:
3960 re_unwind_branch_t *uwb = &(uw->branch);
3961 I32 lastparen = uwb->lastparen;
3963 REGCP_UNWIND(uwb->lastcp);
3964 for (n = *PL_reglastparen; n > lastparen; n--)
3966 *PL_reglastparen = n;
3967 scan = next = uwb->next;
3969 OP(scan) != (uwb->type == RE_UNWIND_BRANCH
3970 ? BRANCH : BRANCHJ) ) { /* Failure */
3977 /* Have more choice yet. Reuse the same uwb. */
3979 if ((n = (uwb->type == RE_UNWIND_BRANCH
3980 ? NEXT_OFF(next) : ARG(next))))
3983 next = NULL; /* XXXX Needn't unwinding in this case... */
3985 next = NEXTOPER(scan);
3986 if (uwb->type == RE_UNWIND_BRANCHJ)
3987 next = NEXTOPER(next);
3988 locinput = uwb->locinput;
3989 nextchr = uwb->nextchr;
3991 PL_regindent = uwb->regindent;
3998 Perl_croak(aTHX_ "regexp unwind memory corruption");
4009 - regrepeat - repeatedly match something simple, report how many
4012 * [This routine now assumes that it will only match on things of length 1.
4013 * That was true before, but now we assume scan - reginput is the count,
4014 * rather than incrementing count on every character. [Er, except utf8.]]
4017 S_regrepeat(pTHX_ regnode *p, I32 max)
4019 register char *scan;
4021 register char *loceol = PL_regeol;
4022 register I32 hardcount = 0;
4023 register bool do_utf8 = PL_reg_match_utf8;
4026 if (max == REG_INFTY)
4028 else if (max < loceol - scan)
4029 loceol = scan + max;
4034 while (scan < loceol && hardcount < max && *scan != '\n') {
4035 scan += UTF8SKIP(scan);
4039 while (scan < loceol && *scan != '\n')
4046 while (scan < loceol && hardcount < max) {
4047 scan += UTF8SKIP(scan);
4057 case EXACT: /* length of string is 1 */
4059 while (scan < loceol && UCHARAT(scan) == c)
4062 case EXACTF: /* length of string is 1 */
4064 while (scan < loceol &&
4065 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold[c]))
4068 case EXACTFL: /* length of string is 1 */
4069 PL_reg_flags |= RF_tainted;
4071 while (scan < loceol &&
4072 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold_locale[c]))
4078 while (hardcount < max && scan < loceol &&
4079 reginclass(p, (U8*)scan, 0, do_utf8)) {
4080 scan += UTF8SKIP(scan);
4084 while (scan < loceol && REGINCLASS(p, (U8*)scan))
4091 LOAD_UTF8_CHARCLASS(alnum,"a");
4092 while (hardcount < max && scan < loceol &&
4093 swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
4094 scan += UTF8SKIP(scan);
4098 while (scan < loceol && isALNUM(*scan))
4103 PL_reg_flags |= RF_tainted;
4106 while (hardcount < max && scan < loceol &&
4107 isALNUM_LC_utf8((U8*)scan)) {
4108 scan += UTF8SKIP(scan);
4112 while (scan < loceol && isALNUM_LC(*scan))
4119 LOAD_UTF8_CHARCLASS(alnum,"a");
4120 while (hardcount < max && scan < loceol &&
4121 !swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
4122 scan += UTF8SKIP(scan);
4126 while (scan < loceol && !isALNUM(*scan))
4131 PL_reg_flags |= RF_tainted;
4134 while (hardcount < max && scan < loceol &&
4135 !isALNUM_LC_utf8((U8*)scan)) {
4136 scan += UTF8SKIP(scan);
4140 while (scan < loceol && !isALNUM_LC(*scan))
4147 LOAD_UTF8_CHARCLASS(space," ");
4148 while (hardcount < max && scan < loceol &&
4150 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
4151 scan += UTF8SKIP(scan);
4155 while (scan < loceol && isSPACE(*scan))
4160 PL_reg_flags |= RF_tainted;
4163 while (hardcount < max && scan < loceol &&
4164 (*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
4165 scan += UTF8SKIP(scan);
4169 while (scan < loceol && isSPACE_LC(*scan))
4176 LOAD_UTF8_CHARCLASS(space," ");
4177 while (hardcount < max && scan < loceol &&
4179 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
4180 scan += UTF8SKIP(scan);
4184 while (scan < loceol && !isSPACE(*scan))
4189 PL_reg_flags |= RF_tainted;
4192 while (hardcount < max && scan < loceol &&
4193 !(*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
4194 scan += UTF8SKIP(scan);
4198 while (scan < loceol && !isSPACE_LC(*scan))
4205 LOAD_UTF8_CHARCLASS(digit,"0");
4206 while (hardcount < max && scan < loceol &&
4207 swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
4208 scan += UTF8SKIP(scan);
4212 while (scan < loceol && isDIGIT(*scan))
4219 LOAD_UTF8_CHARCLASS(digit,"0");
4220 while (hardcount < max && scan < loceol &&
4221 !swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
4222 scan += UTF8SKIP(scan);
4226 while (scan < loceol && !isDIGIT(*scan))
4230 default: /* Called on something of 0 width. */
4231 break; /* So match right here or not at all. */
4237 c = scan - PL_reginput;
4242 SV *prop = sv_newmortal();
4245 PerlIO_printf(Perl_debug_log,
4246 "%*s %s can match %"IVdf" times out of %"IVdf"...\n",
4247 REPORT_CODE_OFF+1, "", SvPVX(prop),(IV)c,(IV)max);
4254 - regrepeat_hard - repeatedly match something, report total lenth and length
4256 * The repeater is supposed to have constant length.
4260 S_regrepeat_hard(pTHX_ regnode *p, I32 max, I32 *lp)
4262 register char *scan = Nullch;
4263 register char *start;
4264 register char *loceol = PL_regeol;
4266 I32 count = 0, res = 1;
4271 start = PL_reginput;
4272 if (PL_reg_match_utf8) {
4273 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
4276 while (start < PL_reginput) {
4278 start += UTF8SKIP(start);
4289 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
4291 *lp = l = PL_reginput - start;
4292 if (max != REG_INFTY && l*max < loceol - scan)
4293 loceol = scan + l*max;
4306 - regclass_swash - prepare the utf8 swash
4310 Perl_regclass_swash(pTHX_ register regnode* node, bool doinit, SV** listsvp, SV **altsvp)
4316 if (PL_regdata && PL_regdata->count) {
4319 if (PL_regdata->what[n] == 's') {
4320 SV *rv = (SV*)PL_regdata->data[n];
4321 AV *av = (AV*)SvRV((SV*)rv);
4322 SV **ary = AvARRAY(av);
4325 /* See the end of regcomp.c:S_reglass() for
4326 * documentation of these array elements. */
4329 a = SvTYPE(ary[1]) == SVt_RV ? &ary[1] : 0;
4330 b = SvTYPE(ary[2]) == SVt_PVAV ? &ary[2] : 0;
4334 else if (si && doinit) {
4335 sw = swash_init("utf8", "", si, 1, 0);
4336 (void)av_store(av, 1, sw);
4352 - reginclass - determine if a character falls into a character class
4354 The n is the ANYOF regnode, the p is the target string, lenp
4355 is pointer to the maximum length of how far to go in the p
4356 (if the lenp is zero, UTF8SKIP(p) is used),
4357 do_utf8 tells whether the target string is in UTF-8.
4362 S_reginclass(pTHX_ register regnode *n, register U8* p, STRLEN* lenp, register bool do_utf8)
4364 char flags = ANYOF_FLAGS(n);
4370 if (do_utf8 && !UTF8_IS_INVARIANT(c))
4371 c = utf8n_to_uvchr(p, UTF8_MAXLEN, &len,
4372 ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
4374 plen = lenp ? *lenp : UNISKIP(NATIVE_TO_UNI(c));
4375 if (do_utf8 || (flags & ANYOF_UNICODE)) {
4378 if (do_utf8 && !ANYOF_RUNTIME(n)) {
4379 if (len != (STRLEN)-1 && c < 256 && ANYOF_BITMAP_TEST(n, c))
4382 if (!match && do_utf8 && (flags & ANYOF_UNICODE_ALL) && c >= 256)
4386 SV *sw = regclass_swash(n, TRUE, 0, (SV**)&av);
4389 if (swash_fetch(sw, p, do_utf8))
4391 else if (flags & ANYOF_FOLD) {
4392 if (!match && lenp && av) {
4395 for (i = 0; i <= av_len(av); i++) {
4396 SV* sv = *av_fetch(av, i, FALSE);
4398 char *s = SvPV(sv, len);
4400 if (len <= plen && memEQ(s, (char*)p, len)) {
4408 U8 tmpbuf[UTF8_MAXLEN_FOLD+1];
4411 to_utf8_fold(p, tmpbuf, &tmplen);
4412 if (swash_fetch(sw, tmpbuf, do_utf8))
4418 if (match && lenp && *lenp == 0)
4419 *lenp = UNISKIP(NATIVE_TO_UNI(c));
4421 if (!match && c < 256) {
4422 if (ANYOF_BITMAP_TEST(n, c))
4424 else if (flags & ANYOF_FOLD) {
4427 if (flags & ANYOF_LOCALE) {
4428 PL_reg_flags |= RF_tainted;
4429 f = PL_fold_locale[c];
4433 if (f != c && ANYOF_BITMAP_TEST(n, f))
4437 if (!match && (flags & ANYOF_CLASS)) {
4438 PL_reg_flags |= RF_tainted;
4440 (ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) ||
4441 (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) ||
4442 (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) ||
4443 (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) ||
4444 (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) ||
4445 (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) ||
4446 (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) ||
4447 (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) ||
4448 (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) ||
4449 (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) ||
4450 (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII(c)) ||
4451 (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII(c)) ||
4452 (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) ||
4453 (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) ||
4454 (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) ||
4455 (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) ||
4456 (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) ||
4457 (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) ||
4458 (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) ||
4459 (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) ||
4460 (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) ||
4461 (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) ||
4462 (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) ||
4463 (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) ||
4464 (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) ||
4465 (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) ||
4466 (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) ||
4467 (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) ||
4468 (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK(c)) ||
4469 (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK(c))
4470 ) /* How's that for a conditional? */
4477 return (flags & ANYOF_INVERT) ? !match : match;
4481 S_reghop(pTHX_ U8 *s, I32 off)
4483 return S_reghop3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4487 S_reghop3(pTHX_ U8 *s, I32 off, U8* lim)
4490 while (off-- && s < lim) {
4491 /* XXX could check well-formedness here */
4499 if (UTF8_IS_CONTINUED(*s)) {
4500 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4503 /* XXX could check well-formedness here */
4511 S_reghopmaybe(pTHX_ U8 *s, I32 off)
4513 return S_reghopmaybe3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4517 S_reghopmaybe3(pTHX_ U8* s, I32 off, U8* lim)
4520 while (off-- && s < lim) {
4521 /* XXX could check well-formedness here */
4531 if (UTF8_IS_CONTINUED(*s)) {
4532 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4535 /* XXX could check well-formedness here */
4547 restore_pos(pTHX_ void *arg)
4549 if (PL_reg_eval_set) {
4550 if (PL_reg_oldsaved) {
4551 PL_reg_re->subbeg = PL_reg_oldsaved;
4552 PL_reg_re->sublen = PL_reg_oldsavedlen;
4553 #ifdef PERL_COPY_ON_WRITE
4554 PL_reg_re->saved_copy = PL_nrs;
4556 RX_MATCH_COPIED_on(PL_reg_re);
4558 PL_reg_magic->mg_len = PL_reg_oldpos;
4559 PL_reg_eval_set = 0;
4560 PL_curpm = PL_reg_oldcurpm;
4565 S_to_utf8_substr(pTHX_ register regexp *prog)
4568 if (prog->float_substr && !prog->float_utf8) {
4569 prog->float_utf8 = sv = NEWSV(117, 0);
4570 SvSetSV(sv, prog->float_substr);
4571 sv_utf8_upgrade(sv);
4572 if (SvTAIL(prog->float_substr))
4574 if (prog->float_substr == prog->check_substr)
4575 prog->check_utf8 = sv;
4577 if (prog->anchored_substr && !prog->anchored_utf8) {
4578 prog->anchored_utf8 = sv = NEWSV(118, 0);
4579 SvSetSV(sv, prog->anchored_substr);
4580 sv_utf8_upgrade(sv);
4581 if (SvTAIL(prog->anchored_substr))
4583 if (prog->anchored_substr == prog->check_substr)
4584 prog->check_utf8 = sv;
4589 S_to_byte_substr(pTHX_ register regexp *prog)
4592 if (prog->float_utf8 && !prog->float_substr) {
4593 prog->float_substr = sv = NEWSV(117, 0);
4594 SvSetSV(sv, prog->float_utf8);
4595 if (sv_utf8_downgrade(sv, TRUE)) {
4596 if (SvTAIL(prog->float_utf8))
4600 prog->float_substr = sv = &PL_sv_undef;
4602 if (prog->float_utf8 == prog->check_utf8)
4603 prog->check_substr = sv;
4605 if (prog->anchored_utf8 && !prog->anchored_substr) {
4606 prog->anchored_substr = sv = NEWSV(118, 0);
4607 SvSetSV(sv, prog->anchored_utf8);
4608 if (sv_utf8_downgrade(sv, TRUE)) {
4609 if (SvTAIL(prog->anchored_utf8))
4613 prog->anchored_substr = sv = &PL_sv_undef;
4615 if (prog->anchored_utf8 == prog->check_utf8)
4616 prog->check_substr = sv;