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) 1991-2002, Larry Wall
72 **** You may distribute under the terms of either the GNU General Public
73 **** License or the Artistic License, as specified in the README file.
75 * Beware that some of this code is subtly aware of the way operator
76 * precedence is structured in regular expressions. Serious changes in
77 * regular-expression syntax might require a total rethink.
80 #define PERL_IN_REGEXEC_C
85 #define RF_tainted 1 /* tainted information used? */
86 #define RF_warned 2 /* warned about big count? */
87 #define RF_evaled 4 /* Did an EVAL with setting? */
88 #define RF_utf8 8 /* String contains multibyte chars? */
90 #define UTF ((PL_reg_flags & RF_utf8) != 0)
92 #define RS_init 1 /* eval environment created */
93 #define RS_set 2 /* replsv value is set */
99 #define REGINCLASS(p,c) (ANYOF_FLAGS(p) ? reginclass(p,c,0,0) : ANYOF_BITMAP_TEST(p,*(c)))
105 #define CHR_SVLEN(sv) (do_utf8 ? sv_len_utf8(sv) : SvCUR(sv))
106 #define CHR_DIST(a,b) (PL_reg_match_utf8 ? utf8_distance(a,b) : a - b)
108 #define reghop_c(pos,off) ((char*)reghop((U8*)pos, off))
109 #define reghopmaybe_c(pos,off) ((char*)reghopmaybe((U8*)pos, off))
110 #define HOP(pos,off) (PL_reg_match_utf8 ? reghop((U8*)pos, off) : (U8*)(pos + off))
111 #define HOPMAYBE(pos,off) (PL_reg_match_utf8 ? reghopmaybe((U8*)pos, off) : (U8*)(pos + off))
112 #define HOPc(pos,off) ((char*)HOP(pos,off))
113 #define HOPMAYBEc(pos,off) ((char*)HOPMAYBE(pos,off))
115 #define HOPBACK(pos, off) ( \
116 (PL_reg_match_utf8) \
117 ? reghopmaybe((U8*)pos, -off) \
118 : (pos - off >= PL_bostr) \
122 #define HOPBACKc(pos, off) (char*)HOPBACK(pos, off)
124 #define reghop3_c(pos,off,lim) ((char*)reghop3((U8*)pos, off, (U8*)lim))
125 #define reghopmaybe3_c(pos,off,lim) ((char*)reghopmaybe3((U8*)pos, off, (U8*)lim))
126 #define HOP3(pos,off,lim) (PL_reg_match_utf8 ? reghop3((U8*)pos, off, (U8*)lim) : (U8*)(pos + off))
127 #define HOPMAYBE3(pos,off,lim) (PL_reg_match_utf8 ? reghopmaybe3((U8*)pos, off, (U8*)lim) : (U8*)(pos + off))
128 #define HOP3c(pos,off,lim) ((char*)HOP3(pos,off,lim))
129 #define HOPMAYBE3c(pos,off,lim) ((char*)HOPMAYBE3(pos,off,lim))
131 #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
133 /* for use after a quantifier and before an EXACT-like node -- japhy */
134 #define JUMPABLE(rn) ( \
135 OP(rn) == OPEN || OP(rn) == CLOSE || OP(rn) == EVAL || \
136 OP(rn) == SUSPEND || OP(rn) == IFMATCH || \
137 OP(rn) == PLUS || OP(rn) == MINMOD || \
138 (PL_regkind[(U8)OP(rn)] == CURLY && ARG1(rn) > 0) \
141 #define HAS_TEXT(rn) ( \
142 PL_regkind[(U8)OP(rn)] == EXACT || PL_regkind[(U8)OP(rn)] == REF \
146 Search for mandatory following text node; for lookahead, the text must
147 follow but for lookbehind (rn->flags != 0) we skip to the next step.
149 #define FIND_NEXT_IMPT(rn) STMT_START { \
150 while (JUMPABLE(rn)) \
151 if (OP(rn) == SUSPEND || PL_regkind[(U8)OP(rn)] == CURLY) \
152 rn = NEXTOPER(NEXTOPER(rn)); \
153 else if (OP(rn) == PLUS) \
155 else if (OP(rn) == IFMATCH) \
156 rn = (rn->flags == 0) ? NEXTOPER(NEXTOPER(rn)) : rn + ARG(rn); \
157 else rn += NEXT_OFF(rn); \
160 static void restore_pos(pTHX_ void *arg);
163 S_regcppush(pTHX_ I32 parenfloor)
165 int retval = PL_savestack_ix;
166 #define REGCP_PAREN_ELEMS 4
167 int paren_elems_to_push = (PL_regsize - parenfloor) * REGCP_PAREN_ELEMS;
170 if (paren_elems_to_push < 0)
171 Perl_croak(aTHX_ "panic: paren_elems_to_push < 0");
173 #define REGCP_OTHER_ELEMS 6
174 SSCHECK(paren_elems_to_push + REGCP_OTHER_ELEMS);
175 for (p = PL_regsize; p > parenfloor; p--) {
176 /* REGCP_PARENS_ELEMS are pushed per pairs of parentheses. */
177 SSPUSHINT(PL_regendp[p]);
178 SSPUSHINT(PL_regstartp[p]);
179 SSPUSHPTR(PL_reg_start_tmp[p]);
182 /* REGCP_OTHER_ELEMS are pushed in any case, parentheses or no. */
183 SSPUSHINT(PL_regsize);
184 SSPUSHINT(*PL_reglastparen);
185 SSPUSHINT(*PL_reglastcloseparen);
186 SSPUSHPTR(PL_reginput);
187 #define REGCP_FRAME_ELEMS 2
188 /* REGCP_FRAME_ELEMS are part of the REGCP_OTHER_ELEMS and
189 * are needed for the regexp context stack bookkeeping. */
190 SSPUSHINT(paren_elems_to_push + REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS);
191 SSPUSHINT(SAVEt_REGCONTEXT); /* Magic cookie. */
196 /* These are needed since we do not localize EVAL nodes: */
197 # define REGCP_SET(cp) DEBUG_r(PerlIO_printf(Perl_debug_log, \
198 " Setting an EVAL scope, savestack=%"IVdf"\n", \
199 (IV)PL_savestack_ix)); cp = PL_savestack_ix
201 # define REGCP_UNWIND(cp) DEBUG_r(cp != PL_savestack_ix ? \
202 PerlIO_printf(Perl_debug_log, \
203 " Clearing an EVAL scope, savestack=%"IVdf"..%"IVdf"\n", \
204 (IV)(cp), (IV)PL_savestack_ix) : 0); regcpblow(cp)
214 /* Pop REGCP_OTHER_ELEMS before the parentheses loop starts. */
216 assert(i == SAVEt_REGCONTEXT); /* Check that the magic cookie is there. */
217 i = SSPOPINT; /* Parentheses elements to pop. */
218 input = (char *) SSPOPPTR;
219 *PL_reglastcloseparen = SSPOPINT;
220 *PL_reglastparen = SSPOPINT;
221 PL_regsize = SSPOPINT;
223 /* Now restore the parentheses context. */
224 for (i -= (REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS);
225 i > 0; i -= REGCP_PAREN_ELEMS) {
226 paren = (U32)SSPOPINT;
227 PL_reg_start_tmp[paren] = (char *) SSPOPPTR;
228 PL_regstartp[paren] = SSPOPINT;
230 if (paren <= *PL_reglastparen)
231 PL_regendp[paren] = tmps;
233 PerlIO_printf(Perl_debug_log,
234 " restoring \\%"UVuf" to %"IVdf"(%"IVdf")..%"IVdf"%s\n",
235 (UV)paren, (IV)PL_regstartp[paren],
236 (IV)(PL_reg_start_tmp[paren] - PL_bostr),
237 (IV)PL_regendp[paren],
238 (paren > *PL_reglastparen ? "(no)" : ""));
242 if ((I32)(*PL_reglastparen + 1) <= PL_regnpar) {
243 PerlIO_printf(Perl_debug_log,
244 " restoring \\%"IVdf"..\\%"IVdf" to undef\n",
245 (IV)(*PL_reglastparen + 1), (IV)PL_regnpar);
249 /* It would seem that the similar code in regtry()
250 * already takes care of this, and in fact it is in
251 * a better location to since this code can #if 0-ed out
252 * but the code in regtry() is needed or otherwise tests
253 * requiring null fields (pat.t#187 and split.t#{13,14}
254 * (as of patchlevel 7877) will fail. Then again,
255 * this code seems to be necessary or otherwise
256 * building DynaLoader will fail:
257 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
259 for (paren = *PL_reglastparen + 1; (I32)paren <= PL_regnpar; paren++) {
260 if ((I32)paren > PL_regsize)
261 PL_regstartp[paren] = -1;
262 PL_regendp[paren] = -1;
269 S_regcp_set_to(pTHX_ I32 ss)
271 I32 tmp = PL_savestack_ix;
273 PL_savestack_ix = ss;
275 PL_savestack_ix = tmp;
279 typedef struct re_cc_state
283 struct re_cc_state *prev;
288 #define regcpblow(cp) LEAVE_SCOPE(cp) /* Ignores regcppush()ed data. */
290 #define TRYPAREN(paren, n, input) { \
293 PL_regstartp[paren] = HOPc(input, -1) - PL_bostr; \
294 PL_regendp[paren] = input - PL_bostr; \
297 PL_regendp[paren] = -1; \
299 if (regmatch(next)) \
302 PL_regendp[paren] = -1; \
307 * pregexec and friends
311 - pregexec - match a regexp against a string
314 Perl_pregexec(pTHX_ register regexp *prog, char *stringarg, register char *strend,
315 char *strbeg, I32 minend, SV *screamer, U32 nosave)
316 /* strend: pointer to null at end of string */
317 /* strbeg: real beginning of string */
318 /* minend: end of match must be >=minend after stringarg. */
319 /* nosave: For optimizations. */
322 regexec_flags(prog, stringarg, strend, strbeg, minend, screamer, NULL,
323 nosave ? 0 : REXEC_COPY_STR);
327 S_cache_re(pTHX_ regexp *prog)
329 PL_regprecomp = prog->precomp; /* Needed for FAIL. */
331 PL_regprogram = prog->program;
333 PL_regnpar = prog->nparens;
334 PL_regdata = prog->data;
339 * Need to implement the following flags for reg_anch:
341 * USE_INTUIT_NOML - Useful to call re_intuit_start() first
343 * INTUIT_AUTORITATIVE_NOML - Can trust a positive answer
344 * INTUIT_AUTORITATIVE_ML
345 * INTUIT_ONCE_NOML - Intuit can match in one location only.
348 * Another flag for this function: SECOND_TIME (so that float substrs
349 * with giant delta may be not rechecked).
352 /* Assumptions: if ANCH_GPOS, then strpos is anchored. XXXX Check GPOS logic */
354 /* If SCREAM, then SvPVX(sv) should be compatible with strpos and strend.
355 Otherwise, only SvCUR(sv) is used to get strbeg. */
357 /* XXXX We assume that strpos is strbeg unless sv. */
359 /* XXXX Some places assume that there is a fixed substring.
360 An update may be needed if optimizer marks as "INTUITable"
361 RExen without fixed substrings. Similarly, it is assumed that
362 lengths of all the strings are no more than minlen, thus they
363 cannot come from lookahead.
364 (Or minlen should take into account lookahead.) */
366 /* A failure to find a constant substring means that there is no need to make
367 an expensive call to REx engine, thus we celebrate a failure. Similarly,
368 finding a substring too deep into the string means that less calls to
369 regtry() should be needed.
371 REx compiler's optimizer found 4 possible hints:
372 a) Anchored substring;
374 c) Whether we are anchored (beginning-of-line or \G);
375 d) First node (of those at offset 0) which may distingush positions;
376 We use a)b)d) and multiline-part of c), and try to find a position in the
377 string which does not contradict any of them.
380 /* Most of decisions we do here should have been done at compile time.
381 The nodes of the REx which we used for the search should have been
382 deleted from the finite automaton. */
385 Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos,
386 char *strend, U32 flags, re_scream_pos_data *data)
388 register I32 start_shift = 0;
389 /* Should be nonnegative! */
390 register I32 end_shift = 0;
395 int do_utf8 = sv ? SvUTF8(sv) : 0; /* if no sv we have to assume bytes */
397 register char *other_last = Nullch; /* other substr checked before this */
398 char *check_at = Nullch; /* check substr found at this pos */
400 char *i_strpos = strpos;
401 SV *dsv = PERL_DEBUG_PAD_ZERO(0);
403 RX_MATCH_UTF8_set(prog,do_utf8);
405 if (prog->reganch & ROPT_UTF8) {
406 DEBUG_r(PerlIO_printf(Perl_debug_log,
407 "UTF-8 regex...\n"));
408 PL_reg_flags |= RF_utf8;
412 char *s = PL_reg_match_utf8 ?
413 sv_uni_display(dsv, sv, 60, UNI_DISPLAY_REGEX) :
415 int len = PL_reg_match_utf8 ?
416 strlen(s) : strend - strpos;
419 if (PL_reg_match_utf8)
420 DEBUG_r(PerlIO_printf(Perl_debug_log,
421 "UTF-8 target...\n"));
422 PerlIO_printf(Perl_debug_log,
423 "%sGuessing start of match, REx%s `%s%.60s%s%s' against `%s%.*s%s%s'...\n",
424 PL_colors[4],PL_colors[5],PL_colors[0],
427 (strlen(prog->precomp) > 60 ? "..." : ""),
429 (int)(len > 60 ? 60 : len),
431 (len > 60 ? "..." : "")
435 /* CHR_DIST() would be more correct here but it makes things slow. */
436 if (prog->minlen > strend - strpos) {
437 DEBUG_r(PerlIO_printf(Perl_debug_log,
438 "String too short... [re_intuit_start]\n"));
441 strbeg = (sv && SvPOK(sv)) ? strend - SvCUR(sv) : strpos;
444 if (!prog->check_utf8 && prog->check_substr)
445 to_utf8_substr(prog);
446 check = prog->check_utf8;
448 if (!prog->check_substr && prog->check_utf8)
449 to_byte_substr(prog);
450 check = prog->check_substr;
452 if (check == &PL_sv_undef) {
453 DEBUG_r(PerlIO_printf(Perl_debug_log,
454 "Non-utf string cannot match utf check string\n"));
457 if (prog->reganch & ROPT_ANCH) { /* Match at beg-of-str or after \n */
458 ml_anch = !( (prog->reganch & ROPT_ANCH_SINGLE)
459 || ( (prog->reganch & ROPT_ANCH_BOL)
460 && !PL_multiline ) ); /* Check after \n? */
463 if ( !(prog->reganch & (ROPT_ANCH_GPOS /* Checked by the caller */
464 | ROPT_IMPLICIT)) /* not a real BOL */
465 /* SvCUR is not set on references: SvRV and SvPVX overlap */
467 && (strpos != strbeg)) {
468 DEBUG_r(PerlIO_printf(Perl_debug_log, "Not at start...\n"));
471 if (prog->check_offset_min == prog->check_offset_max &&
472 !(prog->reganch & ROPT_CANY_SEEN)) {
473 /* Substring at constant offset from beg-of-str... */
476 s = HOP3c(strpos, prog->check_offset_min, strend);
478 slen = SvCUR(check); /* >= 1 */
480 if ( strend - s > slen || strend - s < slen - 1
481 || (strend - s == slen && strend[-1] != '\n')) {
482 DEBUG_r(PerlIO_printf(Perl_debug_log, "String too long...\n"));
485 /* Now should match s[0..slen-2] */
487 if (slen && (*SvPVX(check) != *s
489 && memNE(SvPVX(check), s, slen)))) {
491 DEBUG_r(PerlIO_printf(Perl_debug_log, "String not equal...\n"));
495 else if (*SvPVX(check) != *s
496 || ((slen = SvCUR(check)) > 1
497 && memNE(SvPVX(check), s, slen)))
499 goto success_at_start;
502 /* Match is anchored, but substr is not anchored wrt beg-of-str. */
504 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
505 end_shift = prog->minlen - start_shift -
506 CHR_SVLEN(check) + (SvTAIL(check) != 0);
508 I32 end = prog->check_offset_max + CHR_SVLEN(check)
509 - (SvTAIL(check) != 0);
510 I32 eshift = CHR_DIST((U8*)strend, (U8*)s) - end;
512 if (end_shift < eshift)
516 else { /* Can match at random position */
519 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
520 /* Should be nonnegative! */
521 end_shift = prog->minlen - start_shift -
522 CHR_SVLEN(check) + (SvTAIL(check) != 0);
525 #ifdef DEBUGGING /* 7/99: reports of failure (with the older version) */
527 Perl_croak(aTHX_ "panic: end_shift");
531 /* Find a possible match in the region s..strend by looking for
532 the "check" substring in the region corrected by start/end_shift. */
533 if (flags & REXEC_SCREAM) {
534 I32 p = -1; /* Internal iterator of scream. */
535 I32 *pp = data ? data->scream_pos : &p;
537 if (PL_screamfirst[BmRARE(check)] >= 0
538 || ( BmRARE(check) == '\n'
539 && (BmPREVIOUS(check) == SvCUR(check) - 1)
541 s = screaminstr(sv, check,
542 start_shift + (s - strbeg), end_shift, pp, 0);
545 /* we may be pointing at the wrong string */
546 if (s && RX_MATCH_COPIED(prog))
547 s = prog->subbeg + (s - SvPVX(sv));
549 *data->scream_olds = s;
551 else if (prog->reganch & ROPT_CANY_SEEN)
552 s = fbm_instr((U8*)(s + start_shift),
553 (U8*)(strend - end_shift),
554 check, PL_multiline ? FBMrf_MULTILINE : 0);
556 s = fbm_instr(HOP3(s, start_shift, strend),
557 HOP3(strend, -end_shift, strbeg),
558 check, PL_multiline ? FBMrf_MULTILINE : 0);
560 /* Update the count-of-usability, remove useless subpatterns,
563 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s %s substr `%s%.*s%s'%s%s",
564 (s ? "Found" : "Did not find"),
565 (check == (do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) ? "anchored" : "floating"),
567 (int)(SvCUR(check) - (SvTAIL(check)!=0)),
569 PL_colors[1], (SvTAIL(check) ? "$" : ""),
570 (s ? " at offset " : "...\n") ) );
577 /* Finish the diagnostic message */
578 DEBUG_r(PerlIO_printf(Perl_debug_log, "%ld...\n", (long)(s - i_strpos)) );
580 /* Got a candidate. Check MBOL anchoring, and the *other* substr.
581 Start with the other substr.
582 XXXX no SCREAM optimization yet - and a very coarse implementation
583 XXXX /ttx+/ results in anchored=`ttx', floating=`x'. floating will
584 *always* match. Probably should be marked during compile...
585 Probably it is right to do no SCREAM here...
588 if (do_utf8 ? (prog->float_utf8 && prog->anchored_utf8) : (prog->float_substr && prog->anchored_substr)) {
589 /* Take into account the "other" substring. */
590 /* XXXX May be hopelessly wrong for UTF... */
593 if (check == (do_utf8 ? prog->float_utf8 : prog->float_substr)) {
596 char *last = HOP3c(s, -start_shift, strbeg), *last1, *last2;
600 t = s - prog->check_offset_max;
601 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
603 || ((t = reghopmaybe3_c(s, -(prog->check_offset_max), strpos))
608 t = HOP3c(t, prog->anchored_offset, strend);
609 if (t < other_last) /* These positions already checked */
611 last2 = last1 = HOP3c(strend, -prog->minlen, strbeg);
614 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
615 /* On end-of-str: see comment below. */
616 must = do_utf8 ? prog->anchored_utf8 : prog->anchored_substr;
617 if (must == &PL_sv_undef) {
619 DEBUG_r(must = prog->anchored_utf8); /* for debug */
624 HOP3(HOP3(last1, prog->anchored_offset, strend)
625 + SvCUR(must), -(SvTAIL(must)!=0), strbeg),
627 PL_multiline ? FBMrf_MULTILINE : 0
629 DEBUG_r(PerlIO_printf(Perl_debug_log,
630 "%s anchored substr `%s%.*s%s'%s",
631 (s ? "Found" : "Contradicts"),
634 - (SvTAIL(must)!=0)),
636 PL_colors[1], (SvTAIL(must) ? "$" : "")));
638 if (last1 >= last2) {
639 DEBUG_r(PerlIO_printf(Perl_debug_log,
640 ", giving up...\n"));
643 DEBUG_r(PerlIO_printf(Perl_debug_log,
644 ", trying floating at offset %ld...\n",
645 (long)(HOP3c(s1, 1, strend) - i_strpos)));
646 other_last = HOP3c(last1, prog->anchored_offset+1, strend);
647 s = HOP3c(last, 1, strend);
651 DEBUG_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
652 (long)(s - i_strpos)));
653 t = HOP3c(s, -prog->anchored_offset, strbeg);
654 other_last = HOP3c(s, 1, strend);
662 else { /* Take into account the floating substring. */
667 t = HOP3c(s, -start_shift, strbeg);
669 HOP3c(strend, -prog->minlen + prog->float_min_offset, strbeg);
670 if (CHR_DIST((U8*)last, (U8*)t) > prog->float_max_offset)
671 last = HOP3c(t, prog->float_max_offset, strend);
672 s = HOP3c(t, prog->float_min_offset, strend);
675 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
676 must = do_utf8 ? prog->float_utf8 : prog->float_substr;
677 /* fbm_instr() takes into account exact value of end-of-str
678 if the check is SvTAIL(ed). Since false positives are OK,
679 and end-of-str is not later than strend we are OK. */
680 if (must == &PL_sv_undef) {
682 DEBUG_r(must = prog->float_utf8); /* for debug message */
685 s = fbm_instr((unsigned char*)s,
686 (unsigned char*)last + SvCUR(must)
688 must, PL_multiline ? FBMrf_MULTILINE : 0);
689 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s floating substr `%s%.*s%s'%s",
690 (s ? "Found" : "Contradicts"),
692 (int)(SvCUR(must) - (SvTAIL(must)!=0)),
694 PL_colors[1], (SvTAIL(must) ? "$" : "")));
697 DEBUG_r(PerlIO_printf(Perl_debug_log,
698 ", giving up...\n"));
701 DEBUG_r(PerlIO_printf(Perl_debug_log,
702 ", trying anchored starting at offset %ld...\n",
703 (long)(s1 + 1 - i_strpos)));
705 s = HOP3c(t, 1, strend);
709 DEBUG_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
710 (long)(s - i_strpos)));
711 other_last = s; /* Fix this later. --Hugo */
720 t = s - prog->check_offset_max;
721 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
723 || ((t = reghopmaybe3_c(s, -prog->check_offset_max, strpos))
725 /* Fixed substring is found far enough so that the match
726 cannot start at strpos. */
728 if (ml_anch && t[-1] != '\n') {
729 /* Eventually fbm_*() should handle this, but often
730 anchored_offset is not 0, so this check will not be wasted. */
731 /* XXXX In the code below we prefer to look for "^" even in
732 presence of anchored substrings. And we search even
733 beyond the found float position. These pessimizations
734 are historical artefacts only. */
736 while (t < strend - prog->minlen) {
738 if (t < check_at - prog->check_offset_min) {
739 if (do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) {
740 /* Since we moved from the found position,
741 we definitely contradict the found anchored
742 substr. Due to the above check we do not
743 contradict "check" substr.
744 Thus we can arrive here only if check substr
745 is float. Redo checking for "other"=="fixed".
748 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld, rescanning for anchored from offset %ld...\n",
749 PL_colors[0],PL_colors[1], (long)(strpos - i_strpos), (long)(strpos - i_strpos + prog->anchored_offset)));
750 goto do_other_anchored;
752 /* We don't contradict the found floating substring. */
753 /* XXXX Why not check for STCLASS? */
755 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld...\n",
756 PL_colors[0],PL_colors[1], (long)(s - i_strpos)));
759 /* Position contradicts check-string */
760 /* XXXX probably better to look for check-string
761 than for "\n", so one should lower the limit for t? */
762 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m, restarting lookup for check-string at offset %ld...\n",
763 PL_colors[0],PL_colors[1], (long)(t + 1 - i_strpos)));
764 other_last = strpos = s = t + 1;
769 DEBUG_r(PerlIO_printf(Perl_debug_log, "Did not find /%s^%s/m...\n",
770 PL_colors[0],PL_colors[1]));
774 DEBUG_r(PerlIO_printf(Perl_debug_log, "Starting position does not contradict /%s^%s/m...\n",
775 PL_colors[0],PL_colors[1]));
779 ++BmUSEFUL(do_utf8 ? prog->check_utf8 : prog->check_substr); /* hooray/5 */
782 /* The found string does not prohibit matching at strpos,
783 - no optimization of calling REx engine can be performed,
784 unless it was an MBOL and we are not after MBOL,
785 or a future STCLASS check will fail this. */
787 /* Even in this situation we may use MBOL flag if strpos is offset
788 wrt the start of the string. */
789 if (ml_anch && sv && !SvROK(sv) /* See prev comment on SvROK */
790 && (strpos != strbeg) && strpos[-1] != '\n'
791 /* May be due to an implicit anchor of m{.*foo} */
792 && !(prog->reganch & ROPT_IMPLICIT))
797 DEBUG_r( if (ml_anch)
798 PerlIO_printf(Perl_debug_log, "Position at offset %ld does not contradict /%s^%s/m...\n",
799 (long)(strpos - i_strpos), PL_colors[0],PL_colors[1]);
802 if (!(prog->reganch & ROPT_NAUGHTY) /* XXXX If strpos moved? */
804 prog->check_utf8 /* Could be deleted already */
805 && --BmUSEFUL(prog->check_utf8) < 0
806 && (prog->check_utf8 == prog->float_utf8)
808 prog->check_substr /* Could be deleted already */
809 && --BmUSEFUL(prog->check_substr) < 0
810 && (prog->check_substr == prog->float_substr)
813 /* If flags & SOMETHING - do not do it many times on the same match */
814 DEBUG_r(PerlIO_printf(Perl_debug_log, "... Disabling check substring...\n"));
815 SvREFCNT_dec(do_utf8 ? prog->check_utf8 : prog->check_substr);
816 if (do_utf8 ? prog->check_substr : prog->check_utf8)
817 SvREFCNT_dec(do_utf8 ? prog->check_substr : prog->check_utf8);
818 prog->check_substr = prog->check_utf8 = Nullsv; /* disable */
819 prog->float_substr = prog->float_utf8 = Nullsv; /* clear */
820 check = Nullsv; /* abort */
822 /* XXXX This is a remnant of the old implementation. It
823 looks wasteful, since now INTUIT can use many
825 prog->reganch &= ~RE_USE_INTUIT;
832 /* XXXX BmUSEFUL already changed, maybe multiple change is meaningful... */
833 if (prog->regstclass) {
834 /* minlen == 0 is possible if regstclass is \b or \B,
835 and the fixed substr is ''$.
836 Since minlen is already taken into account, s+1 is before strend;
837 accidentally, minlen >= 1 guaranties no false positives at s + 1
838 even for \b or \B. But (minlen? 1 : 0) below assumes that
839 regstclass does not come from lookahead... */
840 /* If regstclass takes bytelength more than 1: If charlength==1, OK.
841 This leaves EXACTF only, which is dealt with in find_byclass(). */
842 U8* str = (U8*)STRING(prog->regstclass);
843 int cl_l = (PL_regkind[(U8)OP(prog->regstclass)] == EXACT
844 ? CHR_DIST(str+STR_LEN(prog->regstclass), str)
846 char *endpos = (prog->anchored_substr || prog->anchored_utf8 || ml_anch)
847 ? HOP3c(s, (prog->minlen ? cl_l : 0), strend)
848 : (prog->float_substr || prog->float_utf8
849 ? HOP3c(HOP3c(check_at, -start_shift, strbeg),
852 char *startpos = strbeg;
855 if (prog->reganch & ROPT_UTF8) {
856 PL_regdata = prog->data;
860 s = find_byclass(prog, prog->regstclass, s, endpos, startpos, 1);
865 if (endpos == strend) {
866 DEBUG_r( PerlIO_printf(Perl_debug_log,
867 "Could not match STCLASS...\n") );
870 DEBUG_r( PerlIO_printf(Perl_debug_log,
871 "This position contradicts STCLASS...\n") );
872 if ((prog->reganch & ROPT_ANCH) && !ml_anch)
874 /* Contradict one of substrings */
875 if (prog->anchored_substr || prog->anchored_utf8) {
876 if ((do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) == check) {
877 DEBUG_r( what = "anchored" );
879 s = HOP3c(t, 1, strend);
880 if (s + start_shift + end_shift > strend) {
881 /* XXXX Should be taken into account earlier? */
882 DEBUG_r( PerlIO_printf(Perl_debug_log,
883 "Could not match STCLASS...\n") );
888 DEBUG_r( PerlIO_printf(Perl_debug_log,
889 "Looking for %s substr starting at offset %ld...\n",
890 what, (long)(s + start_shift - i_strpos)) );
893 /* Have both, check_string is floating */
894 if (t + start_shift >= check_at) /* Contradicts floating=check */
895 goto retry_floating_check;
896 /* Recheck anchored substring, but not floating... */
900 DEBUG_r( PerlIO_printf(Perl_debug_log,
901 "Looking for anchored substr starting at offset %ld...\n",
902 (long)(other_last - i_strpos)) );
903 goto do_other_anchored;
905 /* Another way we could have checked stclass at the
906 current position only: */
911 DEBUG_r( PerlIO_printf(Perl_debug_log,
912 "Looking for /%s^%s/m starting at offset %ld...\n",
913 PL_colors[0],PL_colors[1], (long)(t - i_strpos)) );
916 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr)) /* Could have been deleted */
918 /* Check is floating subtring. */
919 retry_floating_check:
920 t = check_at - start_shift;
921 DEBUG_r( what = "floating" );
922 goto hop_and_restart;
925 DEBUG_r(PerlIO_printf(Perl_debug_log,
926 "By STCLASS: moving %ld --> %ld\n",
927 (long)(t - i_strpos), (long)(s - i_strpos))
931 DEBUG_r(PerlIO_printf(Perl_debug_log,
932 "Does not contradict STCLASS...\n");
937 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s%s:%s match at offset %ld\n",
938 PL_colors[4], (check ? "Guessed" : "Giving up"),
939 PL_colors[5], (long)(s - i_strpos)) );
942 fail_finish: /* Substring not found */
943 if (prog->check_substr || prog->check_utf8) /* could be removed already */
944 BmUSEFUL(do_utf8 ? prog->check_utf8 : prog->check_substr) += 5; /* hooray */
946 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch rejected by optimizer%s\n",
947 PL_colors[4],PL_colors[5]));
951 /* We know what class REx starts with. Try to find this position... */
953 S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, char *strend, char *startpos, I32 norun)
955 I32 doevery = (prog->reganch & ROPT_SKIP) == 0;
961 register I32 tmp = 1; /* Scratch variable? */
962 register bool do_utf8 = PL_reg_match_utf8;
964 /* We know what class it must start with. */
969 if ((ANYOF_FLAGS(c) & ANYOF_UNICODE) ||
970 !UTF8_IS_INVARIANT((U8)s[0]) ?
971 reginclass(c, (U8*)s, 0, do_utf8) :
972 REGINCLASS(c, (U8*)s)) {
973 if (tmp && (norun || regtry(prog, s)))
987 if (REGINCLASS(c, (U8*)s) ||
988 (ANYOF_FOLD_SHARP_S(c, s, strend) &&
989 /* The assignment of 2 is intentional:
990 * for the folded sharp s, the skip is 2. */
991 (skip = SHARP_S_SKIP))) {
992 if (tmp && (norun || regtry(prog, s)))
1004 while (s < strend) {
1005 if (tmp && (norun || regtry(prog, s)))
1016 STRLEN ulen1, ulen2;
1017 U8 tmpbuf1[UTF8_MAXLEN_UCLC+1];
1018 U8 tmpbuf2[UTF8_MAXLEN_UCLC+1];
1020 to_utf8_lower((U8*)m, tmpbuf1, &ulen1);
1021 to_utf8_upper((U8*)m, tmpbuf2, &ulen2);
1023 c1 = utf8n_to_uvchr(tmpbuf1, UTF8_MAXLEN_UCLC,
1024 0, ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
1025 c2 = utf8n_to_uvchr(tmpbuf2, UTF8_MAXLEN_UCLC,
1026 0, ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
1037 c2 = PL_fold_locale[c1];
1039 e = HOP3c(strend, -(I32)ln, s);
1042 e = s; /* Due to minlen logic of intuit() */
1044 /* The idea in the EXACTF* cases is to first find the
1045 * first character of the EXACTF* node and then, if
1046 * necessary, case-insensitively compare the full
1047 * text of the node. The c1 and c2 are the first
1048 * characters (though in Unicode it gets a bit
1049 * more complicated because there are more cases
1050 * than just upper and lower: one needs to use
1051 * the so-called folding case for case-insensitive
1052 * matching (called "loose matching" in Unicode).
1053 * ibcmp_utf8() will do just that. */
1057 U8 tmpbuf [UTF8_MAXLEN+1];
1058 U8 foldbuf[UTF8_MAXLEN_FOLD+1];
1059 STRLEN len, foldlen;
1063 c = utf8n_to_uvchr((U8*)s, UTF8_MAXLEN, &len,
1065 0 : UTF8_ALLOW_ANY);
1068 ibcmp_utf8(s, (char **)0, 0, do_utf8,
1069 m, (char **)0, ln, (bool)UTF))
1070 && (norun || regtry(prog, s)) )
1073 uvchr_to_utf8(tmpbuf, c);
1074 f = to_utf8_fold(tmpbuf, foldbuf, &foldlen);
1076 && (f == c1 || f == c2)
1077 && (ln == foldlen ||
1078 !ibcmp_utf8((char *) foldbuf,
1079 (char **)0, foldlen, do_utf8,
1081 (char **)0, ln, (bool)UTF))
1082 && (norun || regtry(prog, s)) )
1090 c = utf8n_to_uvchr((U8*)s, UTF8_MAXLEN, &len,
1092 0 : UTF8_ALLOW_ANY);
1094 /* Handle some of the three Greek sigmas cases.
1095 * Note that not all the possible combinations
1096 * are handled here: some of them are handled
1097 * by the standard folding rules, and some of
1098 * them (the character class or ANYOF cases)
1099 * are handled during compiletime in
1100 * regexec.c:S_regclass(). */
1101 if (c == (UV)UNICODE_GREEK_CAPITAL_LETTER_SIGMA ||
1102 c == (UV)UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA)
1103 c = (UV)UNICODE_GREEK_SMALL_LETTER_SIGMA;
1105 if ( (c == c1 || c == c2)
1107 ibcmp_utf8(s, (char **)0, 0, do_utf8,
1108 m, (char **)0, ln, (bool)UTF))
1109 && (norun || regtry(prog, s)) )
1112 uvchr_to_utf8(tmpbuf, c);
1113 f = to_utf8_fold(tmpbuf, foldbuf, &foldlen);
1115 && (f == c1 || f == c2)
1116 && (ln == foldlen ||
1117 !ibcmp_utf8((char *) foldbuf,
1118 (char **)0, foldlen, do_utf8,
1120 (char **)0, ln, (bool)UTF))
1121 && (norun || regtry(prog, s)) )
1132 && (ln == 1 || !(OP(c) == EXACTF
1134 : ibcmp_locale(s, m, ln)))
1135 && (norun || regtry(prog, s)) )
1141 if ( (*(U8*)s == c1 || *(U8*)s == c2)
1142 && (ln == 1 || !(OP(c) == EXACTF
1144 : ibcmp_locale(s, m, ln)))
1145 && (norun || regtry(prog, s)) )
1152 PL_reg_flags |= RF_tainted;
1159 U8 *r = reghop3((U8*)s, -1, (U8*)PL_bostr);
1161 tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, 0);
1163 tmp = ((OP(c) == BOUND ?
1164 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1165 LOAD_UTF8_CHARCLASS(alnum,"a");
1166 while (s < strend) {
1167 if (tmp == !(OP(c) == BOUND ?
1168 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1169 isALNUM_LC_utf8((U8*)s)))
1172 if ((norun || regtry(prog, s)))
1179 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1180 tmp = ((OP(c) == BOUND ? isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1181 while (s < strend) {
1183 !(OP(c) == BOUND ? isALNUM(*s) : isALNUM_LC(*s))) {
1185 if ((norun || regtry(prog, s)))
1191 if ((!prog->minlen && tmp) && (norun || regtry(prog, s)))
1195 PL_reg_flags |= RF_tainted;
1202 U8 *r = reghop3((U8*)s, -1, (U8*)PL_bostr);
1204 tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, 0);
1206 tmp = ((OP(c) == NBOUND ?
1207 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1208 LOAD_UTF8_CHARCLASS(alnum,"a");
1209 while (s < strend) {
1210 if (tmp == !(OP(c) == NBOUND ?
1211 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1212 isALNUM_LC_utf8((U8*)s)))
1214 else if ((norun || regtry(prog, s)))
1220 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1221 tmp = ((OP(c) == NBOUND ?
1222 isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1223 while (s < strend) {
1225 !(OP(c) == NBOUND ? isALNUM(*s) : isALNUM_LC(*s)))
1227 else if ((norun || regtry(prog, s)))
1232 if ((!prog->minlen && !tmp) && (norun || regtry(prog, s)))
1237 LOAD_UTF8_CHARCLASS(alnum,"a");
1238 while (s < strend) {
1239 if (swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1240 if (tmp && (norun || regtry(prog, s)))
1251 while (s < strend) {
1253 if (tmp && (norun || regtry(prog, s)))
1265 PL_reg_flags |= RF_tainted;
1267 while (s < strend) {
1268 if (isALNUM_LC_utf8((U8*)s)) {
1269 if (tmp && (norun || regtry(prog, s)))
1280 while (s < strend) {
1281 if (isALNUM_LC(*s)) {
1282 if (tmp && (norun || regtry(prog, s)))
1295 LOAD_UTF8_CHARCLASS(alnum,"a");
1296 while (s < strend) {
1297 if (!swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1298 if (tmp && (norun || regtry(prog, s)))
1309 while (s < strend) {
1311 if (tmp && (norun || regtry(prog, s)))
1323 PL_reg_flags |= RF_tainted;
1325 while (s < strend) {
1326 if (!isALNUM_LC_utf8((U8*)s)) {
1327 if (tmp && (norun || regtry(prog, s)))
1338 while (s < strend) {
1339 if (!isALNUM_LC(*s)) {
1340 if (tmp && (norun || regtry(prog, s)))
1353 LOAD_UTF8_CHARCLASS(space," ");
1354 while (s < strend) {
1355 if (*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8)) {
1356 if (tmp && (norun || regtry(prog, s)))
1367 while (s < strend) {
1369 if (tmp && (norun || regtry(prog, s)))
1381 PL_reg_flags |= RF_tainted;
1383 while (s < strend) {
1384 if (*s == ' ' || isSPACE_LC_utf8((U8*)s)) {
1385 if (tmp && (norun || regtry(prog, s)))
1396 while (s < strend) {
1397 if (isSPACE_LC(*s)) {
1398 if (tmp && (norun || regtry(prog, s)))
1411 LOAD_UTF8_CHARCLASS(space," ");
1412 while (s < strend) {
1413 if (!(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8))) {
1414 if (tmp && (norun || regtry(prog, s)))
1425 while (s < strend) {
1427 if (tmp && (norun || regtry(prog, s)))
1439 PL_reg_flags |= RF_tainted;
1441 while (s < strend) {
1442 if (!(*s == ' ' || isSPACE_LC_utf8((U8*)s))) {
1443 if (tmp && (norun || regtry(prog, s)))
1454 while (s < strend) {
1455 if (!isSPACE_LC(*s)) {
1456 if (tmp && (norun || regtry(prog, s)))
1469 LOAD_UTF8_CHARCLASS(digit,"0");
1470 while (s < strend) {
1471 if (swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1472 if (tmp && (norun || regtry(prog, s)))
1483 while (s < strend) {
1485 if (tmp && (norun || regtry(prog, s)))
1497 PL_reg_flags |= RF_tainted;
1499 while (s < strend) {
1500 if (isDIGIT_LC_utf8((U8*)s)) {
1501 if (tmp && (norun || regtry(prog, s)))
1512 while (s < strend) {
1513 if (isDIGIT_LC(*s)) {
1514 if (tmp && (norun || regtry(prog, s)))
1527 LOAD_UTF8_CHARCLASS(digit,"0");
1528 while (s < strend) {
1529 if (!swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1530 if (tmp && (norun || regtry(prog, s)))
1541 while (s < strend) {
1543 if (tmp && (norun || regtry(prog, s)))
1555 PL_reg_flags |= RF_tainted;
1557 while (s < strend) {
1558 if (!isDIGIT_LC_utf8((U8*)s)) {
1559 if (tmp && (norun || regtry(prog, s)))
1570 while (s < strend) {
1571 if (!isDIGIT_LC(*s)) {
1572 if (tmp && (norun || regtry(prog, s)))
1584 Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c));
1593 - regexec_flags - match a regexp against a string
1596 Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char *strend,
1597 char *strbeg, I32 minend, SV *sv, void *data, U32 flags)
1598 /* strend: pointer to null at end of string */
1599 /* strbeg: real beginning of string */
1600 /* minend: end of match must be >=minend after stringarg. */
1601 /* data: May be used for some additional optimizations. */
1602 /* nosave: For optimizations. */
1605 register regnode *c;
1606 register char *startpos = stringarg;
1607 I32 minlen; /* must match at least this many chars */
1608 I32 dontbother = 0; /* how many characters not to try at end */
1609 /* I32 start_shift = 0; */ /* Offset of the start to find
1610 constant substr. */ /* CC */
1611 I32 end_shift = 0; /* Same for the end. */ /* CC */
1612 I32 scream_pos = -1; /* Internal iterator of scream. */
1614 SV* oreplsv = GvSV(PL_replgv);
1615 bool do_utf8 = DO_UTF8(sv);
1617 SV *dsv0 = PERL_DEBUG_PAD_ZERO(0);
1618 SV *dsv1 = PERL_DEBUG_PAD_ZERO(1);
1620 RX_MATCH_UTF8_set(prog,do_utf8);
1626 PL_regnarrate = DEBUG_r_TEST;
1629 /* Be paranoid... */
1630 if (prog == NULL || startpos == NULL) {
1631 Perl_croak(aTHX_ "NULL regexp parameter");
1635 minlen = prog->minlen;
1636 if (strend - startpos < minlen) {
1637 DEBUG_r(PerlIO_printf(Perl_debug_log,
1638 "String too short [regexec_flags]...\n"));
1642 /* Check validity of program. */
1643 if (UCHARAT(prog->program) != REG_MAGIC) {
1644 Perl_croak(aTHX_ "corrupted regexp program");
1648 PL_reg_eval_set = 0;
1651 if (prog->reganch & ROPT_UTF8)
1652 PL_reg_flags |= RF_utf8;
1654 /* Mark beginning of line for ^ and lookbehind. */
1655 PL_regbol = startpos;
1659 /* Mark end of line for $ (and such) */
1662 /* see how far we have to get to not match where we matched before */
1663 PL_regtill = startpos+minend;
1665 /* We start without call_cc context. */
1668 /* If there is a "must appear" string, look for it. */
1671 if (prog->reganch & ROPT_GPOS_SEEN) { /* Need to have PL_reg_ganch */
1674 if (flags & REXEC_IGNOREPOS) /* Means: check only at start */
1675 PL_reg_ganch = startpos;
1676 else if (sv && SvTYPE(sv) >= SVt_PVMG
1678 && (mg = mg_find(sv, PERL_MAGIC_regex_global))
1679 && mg->mg_len >= 0) {
1680 PL_reg_ganch = strbeg + mg->mg_len; /* Defined pos() */
1681 if (prog->reganch & ROPT_ANCH_GPOS) {
1682 if (s > PL_reg_ganch)
1687 else /* pos() not defined */
1688 PL_reg_ganch = strbeg;
1691 if (!(flags & REXEC_CHECKED) && (prog->check_substr != Nullsv || prog->check_utf8 != Nullsv)) {
1692 re_scream_pos_data d;
1694 d.scream_olds = &scream_olds;
1695 d.scream_pos = &scream_pos;
1696 s = re_intuit_start(prog, sv, s, strend, flags, &d);
1698 DEBUG_r(PerlIO_printf(Perl_debug_log, "Not present...\n"));
1699 goto phooey; /* not present */
1705 pv_uni_display(dsv0, (U8*)prog->precomp, prog->prelen, 60,
1706 UNI_DISPLAY_REGEX) :
1708 int len0 = UTF ? SvCUR(dsv0) : prog->prelen;
1709 char *s1 = do_utf8 ? sv_uni_display(dsv1, sv, 60,
1710 UNI_DISPLAY_REGEX) : startpos;
1711 int len1 = do_utf8 ? SvCUR(dsv1) : strend - startpos;
1714 PerlIO_printf(Perl_debug_log,
1715 "%sMatching REx%s `%s%*.*s%s%s' against `%s%.*s%s%s'\n",
1716 PL_colors[4],PL_colors[5],PL_colors[0],
1719 len0 > 60 ? "..." : "",
1721 (int)(len1 > 60 ? 60 : len1),
1723 (len1 > 60 ? "..." : "")
1727 /* Simplest case: anchored match need be tried only once. */
1728 /* [unless only anchor is BOL and multiline is set] */
1729 if (prog->reganch & (ROPT_ANCH & ~ROPT_ANCH_GPOS)) {
1730 if (s == startpos && regtry(prog, startpos))
1732 else if (PL_multiline || (prog->reganch & ROPT_IMPLICIT)
1733 || (prog->reganch & ROPT_ANCH_MBOL)) /* XXXX SBOL? */
1738 dontbother = minlen - 1;
1739 end = HOP3c(strend, -dontbother, strbeg) - 1;
1740 /* for multiline we only have to try after newlines */
1741 if (prog->check_substr || prog->check_utf8) {
1745 if (regtry(prog, s))
1750 if (prog->reganch & RE_USE_INTUIT) {
1751 s = re_intuit_start(prog, sv, s + 1, strend, flags, NULL);
1762 if (*s++ == '\n') { /* don't need PL_utf8skip here */
1763 if (regtry(prog, s))
1770 } else if (prog->reganch & ROPT_ANCH_GPOS) {
1771 if (regtry(prog, PL_reg_ganch))
1776 /* Messy cases: unanchored match. */
1777 if ((prog->anchored_substr || prog->anchored_utf8) && prog->reganch & ROPT_SKIP) {
1778 /* we have /x+whatever/ */
1779 /* it must be a one character string (XXXX Except UTF?) */
1784 if (!(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr))
1785 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1786 ch = SvPVX(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr)[0];
1789 while (s < strend) {
1791 DEBUG_r( did_match = 1 );
1792 if (regtry(prog, s)) goto got_it;
1794 while (s < strend && *s == ch)
1801 while (s < strend) {
1803 DEBUG_r( did_match = 1 );
1804 if (regtry(prog, s)) goto got_it;
1806 while (s < strend && *s == ch)
1812 DEBUG_r(if (!did_match)
1813 PerlIO_printf(Perl_debug_log,
1814 "Did not find anchored character...\n")
1818 else if (prog->anchored_substr != Nullsv
1819 || prog->anchored_utf8 != Nullsv
1820 || ((prog->float_substr != Nullsv || prog->float_utf8 != Nullsv)
1821 && prog->float_max_offset < strend - s)) {
1826 char *last1; /* Last position checked before */
1830 if (prog->anchored_substr || prog->anchored_utf8) {
1831 if (!(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr))
1832 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1833 must = do_utf8 ? prog->anchored_utf8 : prog->anchored_substr;
1834 back_max = back_min = prog->anchored_offset;
1836 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr))
1837 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1838 must = do_utf8 ? prog->float_utf8 : prog->float_substr;
1839 back_max = prog->float_max_offset;
1840 back_min = prog->float_min_offset;
1842 if (must == &PL_sv_undef)
1843 /* could not downgrade utf8 check substring, so must fail */
1846 last = HOP3c(strend, /* Cannot start after this */
1847 -(I32)(CHR_SVLEN(must)
1848 - (SvTAIL(must) != 0) + back_min), strbeg);
1851 last1 = HOPc(s, -1);
1853 last1 = s - 1; /* bogus */
1855 /* XXXX check_substr already used to find `s', can optimize if
1856 check_substr==must. */
1858 dontbother = end_shift;
1859 strend = HOPc(strend, -dontbother);
1860 while ( (s <= last) &&
1861 ((flags & REXEC_SCREAM)
1862 ? (s = screaminstr(sv, must, HOP3c(s, back_min, strend) - strbeg,
1863 end_shift, &scream_pos, 0))
1864 : (s = fbm_instr((unsigned char*)HOP3(s, back_min, strend),
1865 (unsigned char*)strend, must,
1866 PL_multiline ? FBMrf_MULTILINE : 0))) ) {
1867 /* we may be pointing at the wrong string */
1868 if ((flags & REXEC_SCREAM) && RX_MATCH_COPIED(prog))
1869 s = prog->subbeg + (s - SvPVX(sv));
1870 DEBUG_r( did_match = 1 );
1871 if (HOPc(s, -back_max) > last1) {
1872 last1 = HOPc(s, -back_min);
1873 s = HOPc(s, -back_max);
1876 char *t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1;
1878 last1 = HOPc(s, -back_min);
1882 while (s <= last1) {
1883 if (regtry(prog, s))
1889 while (s <= last1) {
1890 if (regtry(prog, s))
1896 DEBUG_r(if (!did_match)
1897 PerlIO_printf(Perl_debug_log,
1898 "Did not find %s substr `%s%.*s%s'%s...\n",
1899 ((must == prog->anchored_substr || must == prog->anchored_utf8)
1900 ? "anchored" : "floating"),
1902 (int)(SvCUR(must) - (SvTAIL(must)!=0)),
1904 PL_colors[1], (SvTAIL(must) ? "$" : ""))
1908 else if ((c = prog->regstclass)) {
1910 I32 op = (U8)OP(prog->regstclass);
1911 /* don't bother with what can't match */
1912 if (PL_regkind[op] != EXACT && op != CANY)
1913 strend = HOPc(strend, -(minlen - 1));
1916 SV *prop = sv_newmortal();
1924 pv_uni_display(dsv0, (U8*)SvPVX(prop), SvCUR(prop), 60,
1925 UNI_DISPLAY_REGEX) :
1927 len0 = UTF ? SvCUR(dsv0) : SvCUR(prop);
1929 sv_uni_display(dsv1, sv, 60, UNI_DISPLAY_REGEX) : s;
1930 len1 = UTF ? SvCUR(dsv1) : strend - s;
1931 PerlIO_printf(Perl_debug_log,
1932 "Matching stclass `%*.*s' against `%*.*s'\n",
1936 if (find_byclass(prog, c, s, strend, startpos, 0))
1938 DEBUG_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass...\n"));
1942 if (prog->float_substr != Nullsv || prog->float_utf8 != Nullsv) {
1947 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr))
1948 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1949 float_real = do_utf8 ? prog->float_utf8 : prog->float_substr;
1951 if (flags & REXEC_SCREAM) {
1952 last = screaminstr(sv, float_real, s - strbeg,
1953 end_shift, &scream_pos, 1); /* last one */
1955 last = scream_olds; /* Only one occurrence. */
1956 /* we may be pointing at the wrong string */
1957 else if (RX_MATCH_COPIED(prog))
1958 s = prog->subbeg + (s - SvPVX(sv));
1962 char *little = SvPV(float_real, len);
1964 if (SvTAIL(float_real)) {
1965 if (memEQ(strend - len + 1, little, len - 1))
1966 last = strend - len + 1;
1967 else if (!PL_multiline)
1968 last = memEQ(strend - len, little, len)
1969 ? strend - len : Nullch;
1975 last = rninstr(s, strend, little, little + len);
1977 last = strend; /* matching `$' */
1981 DEBUG_r(PerlIO_printf(Perl_debug_log,
1982 "%sCan't trim the tail, match fails (should not happen)%s\n",
1983 PL_colors[4],PL_colors[5]));
1984 goto phooey; /* Should not happen! */
1986 dontbother = strend - last + prog->float_min_offset;
1988 if (minlen && (dontbother < minlen))
1989 dontbother = minlen - 1;
1990 strend -= dontbother; /* this one's always in bytes! */
1991 /* We don't know much -- general case. */
1994 if (regtry(prog, s))
2003 if (regtry(prog, s))
2005 } while (s++ < strend);
2013 RX_MATCH_TAINTED_set(prog, PL_reg_flags & RF_tainted);
2015 if (PL_reg_eval_set) {
2016 /* Preserve the current value of $^R */
2017 if (oreplsv != GvSV(PL_replgv))
2018 sv_setsv(oreplsv, GvSV(PL_replgv));/* So that when GvSV(replgv) is
2019 restored, the value remains
2021 restore_pos(aTHX_ 0);
2024 /* make sure $`, $&, $', and $digit will work later */
2025 if ( !(flags & REXEC_NOT_FIRST) ) {
2026 RX_MATCH_COPY_FREE(prog);
2027 if (flags & REXEC_COPY_STR) {
2028 I32 i = PL_regeol - startpos + (stringarg - strbeg);
2029 #ifdef PERL_COPY_ON_WRITE
2031 || (SvFLAGS(sv) & CAN_COW_MASK) == CAN_COW_FLAGS)) {
2033 PerlIO_printf(Perl_debug_log,
2034 "Copy on write: regexp capture, type %d\n",
2037 prog->saved_copy = sv_setsv_cow(prog->saved_copy, sv);
2038 prog->subbeg = SvPVX(prog->saved_copy);
2039 assert (SvPOKp(prog->saved_copy));
2043 RX_MATCH_COPIED_on(prog);
2044 s = savepvn(strbeg, i);
2050 prog->subbeg = strbeg;
2051 prog->sublen = PL_regeol - strbeg; /* strend may have been modified */
2058 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n",
2059 PL_colors[4],PL_colors[5]));
2060 if (PL_reg_eval_set)
2061 restore_pos(aTHX_ 0);
2066 - regtry - try match at specific point
2068 STATIC I32 /* 0 failure, 1 success */
2069 S_regtry(pTHX_ regexp *prog, char *startpos)
2077 PL_regindent = 0; /* XXXX Not good when matches are reenterable... */
2079 if ((prog->reganch & ROPT_EVAL_SEEN) && !PL_reg_eval_set) {
2082 PL_reg_eval_set = RS_init;
2084 PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %"IVdf"\n",
2085 (IV)(PL_stack_sp - PL_stack_base));
2087 SAVEI32(cxstack[cxstack_ix].blk_oldsp);
2088 cxstack[cxstack_ix].blk_oldsp = PL_stack_sp - PL_stack_base;
2089 /* Otherwise OP_NEXTSTATE will free whatever on stack now. */
2091 /* Apparently this is not needed, judging by wantarray. */
2092 /* SAVEI8(cxstack[cxstack_ix].blk_gimme);
2093 cxstack[cxstack_ix].blk_gimme = G_SCALAR; */
2096 /* Make $_ available to executed code. */
2097 if (PL_reg_sv != DEFSV) {
2098 /* SAVE_DEFSV does *not* suffice here for USE_5005THREADS */
2103 if (!(SvTYPE(PL_reg_sv) >= SVt_PVMG && SvMAGIC(PL_reg_sv)
2104 && (mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global)))) {
2105 /* prepare for quick setting of pos */
2106 sv_magic(PL_reg_sv, (SV*)0,
2107 PERL_MAGIC_regex_global, Nullch, 0);
2108 mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global);
2112 PL_reg_oldpos = mg->mg_len;
2113 SAVEDESTRUCTOR_X(restore_pos, 0);
2115 if (!PL_reg_curpm) {
2116 Newz(22,PL_reg_curpm, 1, PMOP);
2119 SV* repointer = newSViv(0);
2120 /* so we know which PL_regex_padav element is PL_reg_curpm */
2121 SvFLAGS(repointer) |= SVf_BREAK;
2122 av_push(PL_regex_padav,repointer);
2123 PL_reg_curpm->op_pmoffset = av_len(PL_regex_padav);
2124 PL_regex_pad = AvARRAY(PL_regex_padav);
2128 PM_SETRE(PL_reg_curpm, prog);
2129 PL_reg_oldcurpm = PL_curpm;
2130 PL_curpm = PL_reg_curpm;
2131 if (RX_MATCH_COPIED(prog)) {
2132 /* Here is a serious problem: we cannot rewrite subbeg,
2133 since it may be needed if this match fails. Thus
2134 $` inside (?{}) could fail... */
2135 PL_reg_oldsaved = prog->subbeg;
2136 PL_reg_oldsavedlen = prog->sublen;
2137 #ifdef PERL_COPY_ON_WRITE
2138 PL_nrs = prog->saved_copy;
2140 RX_MATCH_COPIED_off(prog);
2143 PL_reg_oldsaved = Nullch;
2144 prog->subbeg = PL_bostr;
2145 prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */
2147 prog->startp[0] = startpos - PL_bostr;
2148 PL_reginput = startpos;
2149 PL_regstartp = prog->startp;
2150 PL_regendp = prog->endp;
2151 PL_reglastparen = &prog->lastparen;
2152 PL_reglastcloseparen = &prog->lastcloseparen;
2153 prog->lastparen = 0;
2154 prog->lastcloseparen = 0;
2156 DEBUG_r(PL_reg_starttry = startpos);
2157 if (PL_reg_start_tmpl <= prog->nparens) {
2158 PL_reg_start_tmpl = prog->nparens*3/2 + 3;
2159 if(PL_reg_start_tmp)
2160 Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
2162 New(22,PL_reg_start_tmp, PL_reg_start_tmpl, char*);
2165 /* XXXX What this code is doing here?!!! There should be no need
2166 to do this again and again, PL_reglastparen should take care of
2169 /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code.
2170 * Actually, the code in regcppop() (which Ilya may be meaning by
2171 * PL_reglastparen), is not needed at all by the test suite
2172 * (op/regexp, op/pat, op/split), but that code is needed, oddly
2173 * enough, for building DynaLoader, or otherwise this
2174 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
2175 * will happen. Meanwhile, this code *is* needed for the
2176 * above-mentioned test suite tests to succeed. The common theme
2177 * on those tests seems to be returning null fields from matches.
2182 if (prog->nparens) {
2183 for (i = prog->nparens; i > (I32)*PL_reglastparen; i--) {
2190 if (regmatch(prog->program + 1)) {
2191 prog->endp[0] = PL_reginput - PL_bostr;
2194 REGCP_UNWIND(lastcp);
2198 #define RE_UNWIND_BRANCH 1
2199 #define RE_UNWIND_BRANCHJ 2
2203 typedef struct { /* XX: makes sense to enlarge it... */
2207 } re_unwind_generic_t;
2220 } re_unwind_branch_t;
2222 typedef union re_unwind_t {
2224 re_unwind_generic_t generic;
2225 re_unwind_branch_t branch;
2228 #define sayYES goto yes
2229 #define sayNO goto no
2230 #define sayNO_ANYOF goto no_anyof
2231 #define sayYES_FINAL goto yes_final
2232 #define sayYES_LOUD goto yes_loud
2233 #define sayNO_FINAL goto no_final
2234 #define sayNO_SILENT goto do_no
2235 #define saySAME(x) if (x) goto yes; else goto no
2237 #define REPORT_CODE_OFF 24
2240 - regmatch - main matching routine
2242 * Conceptually the strategy is simple: check to see whether the current
2243 * node matches, call self recursively to see whether the rest matches,
2244 * and then act accordingly. In practice we make some effort to avoid
2245 * recursion, in particular by going through "ordinary" nodes (that don't
2246 * need to know whether the rest of the match failed) by a loop instead of
2249 /* [lwall] I've hoisted the register declarations to the outer block in order to
2250 * maybe save a little bit of pushing and popping on the stack. It also takes
2251 * advantage of machines that use a register save mask on subroutine entry.
2253 STATIC I32 /* 0 failure, 1 success */
2254 S_regmatch(pTHX_ regnode *prog)
2256 register regnode *scan; /* Current node. */
2257 regnode *next; /* Next node. */
2258 regnode *inner; /* Next node in internal branch. */
2259 register I32 nextchr; /* renamed nextchr - nextchar colides with
2260 function of same name */
2261 register I32 n; /* no or next */
2262 register I32 ln = 0; /* len or last */
2263 register char *s = Nullch; /* operand or save */
2264 register char *locinput = PL_reginput;
2265 register I32 c1 = 0, c2 = 0, paren; /* case fold search, parenth */
2266 int minmod = 0, sw = 0, logical = 0;
2269 I32 firstcp = PL_savestack_ix;
2271 register bool do_utf8 = PL_reg_match_utf8;
2273 SV *dsv0 = PERL_DEBUG_PAD_ZERO(0);
2274 SV *dsv1 = PERL_DEBUG_PAD_ZERO(1);
2275 SV *dsv2 = PERL_DEBUG_PAD_ZERO(2);
2282 /* Note that nextchr is a byte even in UTF */
2283 nextchr = UCHARAT(locinput);
2285 while (scan != NULL) {
2288 SV *prop = sv_newmortal();
2289 int docolor = *PL_colors[0];
2290 int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */
2291 int l = (PL_regeol - locinput) > taill ? taill : (PL_regeol - locinput);
2292 /* The part of the string before starttry has one color
2293 (pref0_len chars), between starttry and current
2294 position another one (pref_len - pref0_len chars),
2295 after the current position the third one.
2296 We assume that pref0_len <= pref_len, otherwise we
2297 decrease pref0_len. */
2298 int pref_len = (locinput - PL_bostr) > (5 + taill) - l
2299 ? (5 + taill) - l : locinput - PL_bostr;
2302 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len)))
2304 pref0_len = pref_len - (locinput - PL_reg_starttry);
2305 if (l + pref_len < (5 + taill) && l < PL_regeol - locinput)
2306 l = ( PL_regeol - locinput > (5 + taill) - pref_len
2307 ? (5 + taill) - pref_len : PL_regeol - locinput);
2308 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput + l)))
2312 if (pref0_len > pref_len)
2313 pref0_len = pref_len;
2314 regprop(prop, scan);
2317 do_utf8 && OP(scan) != CANY ?
2318 pv_uni_display(dsv0, (U8*)(locinput - pref_len),
2319 pref0_len, 60, UNI_DISPLAY_REGEX) :
2320 locinput - pref_len;
2321 int len0 = do_utf8 ? strlen(s0) : pref0_len;
2322 char *s1 = do_utf8 && OP(scan) != CANY ?
2323 pv_uni_display(dsv1, (U8*)(locinput - pref_len + pref0_len),
2324 pref_len - pref0_len, 60, UNI_DISPLAY_REGEX) :
2325 locinput - pref_len + pref0_len;
2326 int len1 = do_utf8 ? strlen(s1) : pref_len - pref0_len;
2327 char *s2 = do_utf8 && OP(scan) != CANY ?
2328 pv_uni_display(dsv2, (U8*)locinput,
2329 PL_regeol - locinput, 60, UNI_DISPLAY_REGEX) :
2331 int len2 = do_utf8 ? strlen(s2) : l;
2332 PerlIO_printf(Perl_debug_log,
2333 "%4"IVdf" <%s%.*s%s%s%.*s%s%s%s%.*s%s>%*s|%3"IVdf":%*s%s\n",
2334 (IV)(locinput - PL_bostr),
2341 (docolor ? "" : "> <"),
2345 15 - l - pref_len + 1,
2347 (IV)(scan - PL_regprogram), PL_regindent*2, "",
2352 next = scan + NEXT_OFF(scan);
2358 if (locinput == PL_bostr || (PL_multiline &&
2359 (nextchr || locinput < PL_regeol) && locinput[-1] == '\n') )
2361 /* regtill = regbol; */
2366 if (locinput == PL_bostr ||
2367 ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n'))
2373 if (locinput == PL_bostr)
2377 if (locinput == PL_reg_ganch)
2387 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2392 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2394 if (PL_regeol - locinput > 1)
2398 if (PL_regeol != locinput)
2402 if (!nextchr && locinput >= PL_regeol)
2405 locinput += PL_utf8skip[nextchr];
2406 if (locinput > PL_regeol)
2408 nextchr = UCHARAT(locinput);
2411 nextchr = UCHARAT(++locinput);
2414 if (!nextchr && locinput >= PL_regeol)
2416 nextchr = UCHARAT(++locinput);
2419 if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n')
2422 locinput += PL_utf8skip[nextchr];
2423 if (locinput > PL_regeol)
2425 nextchr = UCHARAT(locinput);
2428 nextchr = UCHARAT(++locinput);
2433 if (do_utf8 != UTF) {
2434 /* The target and the pattern have differing utf8ness. */
2440 /* The target is utf8, the pattern is not utf8. */
2444 if (NATIVE_TO_UNI(*(U8*)s) !=
2445 utf8n_to_uvuni((U8*)l, UTF8_MAXLEN, &ulen,
2447 0 : UTF8_ALLOW_ANY))
2454 /* The target is not utf8, the pattern is utf8. */
2458 if (NATIVE_TO_UNI(*((U8*)l)) !=
2459 utf8n_to_uvuni((U8*)s, UTF8_MAXLEN, &ulen,
2461 0 : UTF8_ALLOW_ANY))
2468 nextchr = UCHARAT(locinput);
2471 /* The target and the pattern have the same utf8ness. */
2472 /* Inline the first character, for speed. */
2473 if (UCHARAT(s) != nextchr)
2475 if (PL_regeol - locinput < ln)
2477 if (ln > 1 && memNE(s, locinput, ln))
2480 nextchr = UCHARAT(locinput);
2483 PL_reg_flags |= RF_tainted;
2489 if (do_utf8 || UTF) {
2490 /* Either target or the pattern are utf8. */
2492 char *e = PL_regeol;
2494 if (ibcmp_utf8(s, 0, ln, (bool)UTF,
2495 l, &e, 0, do_utf8)) {
2496 /* One more case for the sharp s:
2497 * pack("U0U*", 0xDF) =~ /ss/i,
2498 * the 0xC3 0x9F are the UTF-8
2499 * byte sequence for the U+00DF. */
2501 toLOWER(s[0]) == 's' &&
2503 toLOWER(s[1]) == 's' &&
2510 nextchr = UCHARAT(locinput);
2514 /* Neither the target and the pattern are utf8. */
2516 /* Inline the first character, for speed. */
2517 if (UCHARAT(s) != nextchr &&
2518 UCHARAT(s) != ((OP(scan) == EXACTF)
2519 ? PL_fold : PL_fold_locale)[nextchr])
2521 if (PL_regeol - locinput < ln)
2523 if (ln > 1 && (OP(scan) == EXACTF
2524 ? ibcmp(s, locinput, ln)
2525 : ibcmp_locale(s, locinput, ln)))
2528 nextchr = UCHARAT(locinput);
2532 STRLEN inclasslen = PL_regeol - locinput;
2534 if (!reginclass(scan, (U8*)locinput, &inclasslen, do_utf8))
2536 if (locinput >= PL_regeol)
2538 locinput += inclasslen ? inclasslen : UTF8SKIP(locinput);
2539 nextchr = UCHARAT(locinput);
2544 nextchr = UCHARAT(locinput);
2545 if (!REGINCLASS(scan, (U8*)locinput))
2547 if (!nextchr && locinput >= PL_regeol)
2549 nextchr = UCHARAT(++locinput);
2553 /* If we might have the case of the German sharp s
2554 * in a casefolding Unicode character class. */
2556 if (ANYOF_FOLD_SHARP_S(scan, locinput, PL_regeol)) {
2557 locinput += SHARP_S_SKIP;
2558 nextchr = UCHARAT(locinput);
2564 PL_reg_flags |= RF_tainted;
2570 LOAD_UTF8_CHARCLASS(alnum,"a");
2571 if (!(OP(scan) == ALNUM
2572 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2573 : isALNUM_LC_utf8((U8*)locinput)))
2577 locinput += PL_utf8skip[nextchr];
2578 nextchr = UCHARAT(locinput);
2581 if (!(OP(scan) == ALNUM
2582 ? isALNUM(nextchr) : isALNUM_LC(nextchr)))
2584 nextchr = UCHARAT(++locinput);
2587 PL_reg_flags |= RF_tainted;
2590 if (!nextchr && locinput >= PL_regeol)
2593 LOAD_UTF8_CHARCLASS(alnum,"a");
2594 if (OP(scan) == NALNUM
2595 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2596 : isALNUM_LC_utf8((U8*)locinput))
2600 locinput += PL_utf8skip[nextchr];
2601 nextchr = UCHARAT(locinput);
2604 if (OP(scan) == NALNUM
2605 ? isALNUM(nextchr) : isALNUM_LC(nextchr))
2607 nextchr = UCHARAT(++locinput);
2611 PL_reg_flags |= RF_tainted;
2615 /* was last char in word? */
2617 if (locinput == PL_bostr)
2620 U8 *r = reghop3((U8*)locinput, -1, (U8*)PL_bostr);
2622 ln = utf8n_to_uvchr(r, UTF8SKIP(r), 0, 0);
2624 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2625 ln = isALNUM_uni(ln);
2626 LOAD_UTF8_CHARCLASS(alnum,"a");
2627 n = swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8);
2630 ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(ln));
2631 n = isALNUM_LC_utf8((U8*)locinput);
2635 ln = (locinput != PL_bostr) ?
2636 UCHARAT(locinput - 1) : '\n';
2637 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2639 n = isALNUM(nextchr);
2642 ln = isALNUM_LC(ln);
2643 n = isALNUM_LC(nextchr);
2646 if (((!ln) == (!n)) == (OP(scan) == BOUND ||
2647 OP(scan) == BOUNDL))
2651 PL_reg_flags |= RF_tainted;
2657 if (UTF8_IS_CONTINUED(nextchr)) {
2658 LOAD_UTF8_CHARCLASS(space," ");
2659 if (!(OP(scan) == SPACE
2660 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2661 : isSPACE_LC_utf8((U8*)locinput)))
2665 locinput += PL_utf8skip[nextchr];
2666 nextchr = UCHARAT(locinput);
2669 if (!(OP(scan) == SPACE
2670 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2672 nextchr = UCHARAT(++locinput);
2675 if (!(OP(scan) == SPACE
2676 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2678 nextchr = UCHARAT(++locinput);
2682 PL_reg_flags |= RF_tainted;
2685 if (!nextchr && locinput >= PL_regeol)
2688 LOAD_UTF8_CHARCLASS(space," ");
2689 if (OP(scan) == NSPACE
2690 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2691 : isSPACE_LC_utf8((U8*)locinput))
2695 locinput += PL_utf8skip[nextchr];
2696 nextchr = UCHARAT(locinput);
2699 if (OP(scan) == NSPACE
2700 ? isSPACE(nextchr) : isSPACE_LC(nextchr))
2702 nextchr = UCHARAT(++locinput);
2705 PL_reg_flags |= RF_tainted;
2711 LOAD_UTF8_CHARCLASS(digit,"0");
2712 if (!(OP(scan) == DIGIT
2713 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2714 : isDIGIT_LC_utf8((U8*)locinput)))
2718 locinput += PL_utf8skip[nextchr];
2719 nextchr = UCHARAT(locinput);
2722 if (!(OP(scan) == DIGIT
2723 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr)))
2725 nextchr = UCHARAT(++locinput);
2728 PL_reg_flags |= RF_tainted;
2731 if (!nextchr && locinput >= PL_regeol)
2734 LOAD_UTF8_CHARCLASS(digit,"0");
2735 if (OP(scan) == NDIGIT
2736 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2737 : isDIGIT_LC_utf8((U8*)locinput))
2741 locinput += PL_utf8skip[nextchr];
2742 nextchr = UCHARAT(locinput);
2745 if (OP(scan) == NDIGIT
2746 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr))
2748 nextchr = UCHARAT(++locinput);
2751 if (locinput >= PL_regeol)
2754 LOAD_UTF8_CHARCLASS(mark,"~");
2755 if (swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2757 locinput += PL_utf8skip[nextchr];
2758 while (locinput < PL_regeol &&
2759 swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2760 locinput += UTF8SKIP(locinput);
2761 if (locinput > PL_regeol)
2766 nextchr = UCHARAT(locinput);
2769 PL_reg_flags |= RF_tainted;
2773 n = ARG(scan); /* which paren pair */
2774 ln = PL_regstartp[n];
2775 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2776 if ((I32)*PL_reglastparen < n || ln == -1)
2777 sayNO; /* Do not match unless seen CLOSEn. */
2778 if (ln == PL_regendp[n])
2782 if (do_utf8 && OP(scan) != REF) { /* REF can do byte comparison */
2784 char *e = PL_bostr + PL_regendp[n];
2786 * Note that we can't do the "other character" lookup trick as
2787 * in the 8-bit case (no pun intended) because in Unicode we
2788 * have to map both upper and title case to lower case.
2790 if (OP(scan) == REFF) {
2791 STRLEN ulen1, ulen2;
2792 U8 tmpbuf1[UTF8_MAXLEN_UCLC+1];
2793 U8 tmpbuf2[UTF8_MAXLEN_UCLC+1];
2797 toLOWER_utf8((U8*)s, tmpbuf1, &ulen1);
2798 toLOWER_utf8((U8*)l, tmpbuf2, &ulen2);
2799 if (ulen1 != ulen2 || memNE((char *)tmpbuf1, (char *)tmpbuf2, ulen1))
2806 nextchr = UCHARAT(locinput);
2810 /* Inline the first character, for speed. */
2811 if (UCHARAT(s) != nextchr &&
2813 (UCHARAT(s) != ((OP(scan) == REFF
2814 ? PL_fold : PL_fold_locale)[nextchr]))))
2816 ln = PL_regendp[n] - ln;
2817 if (locinput + ln > PL_regeol)
2819 if (ln > 1 && (OP(scan) == REF
2820 ? memNE(s, locinput, ln)
2822 ? ibcmp(s, locinput, ln)
2823 : ibcmp_locale(s, locinput, ln))))
2826 nextchr = UCHARAT(locinput);
2837 OP_4tree *oop = PL_op;
2838 COP *ocurcop = PL_curcop;
2843 PL_op = (OP_4tree*)PL_regdata->data[n];
2844 DEBUG_r( PerlIO_printf(Perl_debug_log, " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) );
2845 PAD_SAVE_LOCAL(old_comppad, (PAD*)PL_regdata->data[n + 2]);
2846 PL_regendp[0] = PL_reg_magic->mg_len = locinput - PL_bostr;
2850 CALLRUNOPS(aTHX); /* Scalar context. */
2853 ret = &PL_sv_undef; /* protect against empty (?{}) blocks. */
2861 PAD_RESTORE_LOCAL(old_comppad);
2862 PL_curcop = ocurcop;
2864 if (logical == 2) { /* Postponed subexpression. */
2866 MAGIC *mg = Null(MAGIC*);
2868 CHECKPOINT cp, lastcp;
2871 if(SvROK(ret) || SvRMAGICAL(ret)) {
2872 SV *sv = SvROK(ret) ? SvRV(ret) : ret;
2875 mg = mg_find(sv, PERL_MAGIC_qr);
2878 re = (regexp *)mg->mg_obj;
2879 (void)ReREFCNT_inc(re);
2883 char *t = SvPV(ret, len);
2885 char *oprecomp = PL_regprecomp;
2886 I32 osize = PL_regsize;
2887 I32 onpar = PL_regnpar;
2890 if (DO_UTF8(ret)) pm.op_pmdynflags |= PMdf_DYN_UTF8;
2891 re = CALLREGCOMP(aTHX_ t, t + len, &pm);
2893 & (SVs_TEMP | SVs_PADTMP | SVf_READONLY)))
2894 sv_magic(ret,(SV*)ReREFCNT_inc(re),
2896 PL_regprecomp = oprecomp;
2901 PerlIO_printf(Perl_debug_log,
2902 "Entering embedded `%s%.60s%s%s'\n",
2906 (strlen(re->precomp) > 60 ? "..." : ""))
2909 state.prev = PL_reg_call_cc;
2910 state.cc = PL_regcc;
2911 state.re = PL_reg_re;
2915 cp = regcppush(0); /* Save *all* the positions. */
2918 state.ss = PL_savestack_ix;
2919 *PL_reglastparen = 0;
2920 *PL_reglastcloseparen = 0;
2921 PL_reg_call_cc = &state;
2922 PL_reginput = locinput;
2923 toggleutf = ((PL_reg_flags & RF_utf8) != 0) ^
2924 ((re->reganch & ROPT_UTF8) != 0);
2925 if (toggleutf) PL_reg_flags ^= RF_utf8;
2927 /* XXXX This is too dramatic a measure... */
2930 if (regmatch(re->program + 1)) {
2931 /* Even though we succeeded, we need to restore
2932 global variables, since we may be wrapped inside
2933 SUSPEND, thus the match may be not finished yet. */
2935 /* XXXX Do this only if SUSPENDed? */
2936 PL_reg_call_cc = state.prev;
2937 PL_regcc = state.cc;
2938 PL_reg_re = state.re;
2939 cache_re(PL_reg_re);
2940 if (toggleutf) PL_reg_flags ^= RF_utf8;
2942 /* XXXX This is too dramatic a measure... */
2945 /* These are needed even if not SUSPEND. */
2951 REGCP_UNWIND(lastcp);
2953 PL_reg_call_cc = state.prev;
2954 PL_regcc = state.cc;
2955 PL_reg_re = state.re;
2956 cache_re(PL_reg_re);
2957 if (toggleutf) PL_reg_flags ^= RF_utf8;
2959 /* XXXX This is too dramatic a measure... */
2969 sv_setsv(save_scalar(PL_replgv), ret);
2973 n = ARG(scan); /* which paren pair */
2974 PL_reg_start_tmp[n] = locinput;
2979 n = ARG(scan); /* which paren pair */
2980 PL_regstartp[n] = PL_reg_start_tmp[n] - PL_bostr;
2981 PL_regendp[n] = locinput - PL_bostr;
2982 if (n > (I32)*PL_reglastparen)
2983 *PL_reglastparen = n;
2984 *PL_reglastcloseparen = n;
2987 n = ARG(scan); /* which paren pair */
2988 sw = ((I32)*PL_reglastparen >= n && PL_regendp[n] != -1);
2991 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2993 next = NEXTOPER(NEXTOPER(scan));
2995 next = scan + ARG(scan);
2996 if (OP(next) == IFTHEN) /* Fake one. */
2997 next = NEXTOPER(NEXTOPER(next));
3001 logical = scan->flags;
3003 /*******************************************************************
3004 PL_regcc contains infoblock about the innermost (...)* loop, and
3005 a pointer to the next outer infoblock.
3007 Here is how Y(A)*Z is processed (if it is compiled into CURLYX/WHILEM):
3009 1) After matching X, regnode for CURLYX is processed;
3011 2) This regnode creates infoblock on the stack, and calls
3012 regmatch() recursively with the starting point at WHILEM node;
3014 3) Each hit of WHILEM node tries to match A and Z (in the order
3015 depending on the current iteration, min/max of {min,max} and
3016 greediness). The information about where are nodes for "A"
3017 and "Z" is read from the infoblock, as is info on how many times "A"
3018 was already matched, and greediness.
3020 4) After A matches, the same WHILEM node is hit again.
3022 5) Each time WHILEM is hit, PL_regcc is the infoblock created by CURLYX
3023 of the same pair. Thus when WHILEM tries to match Z, it temporarily
3024 resets PL_regcc, since this Y(A)*Z can be a part of some other loop:
3025 as in (Y(A)*Z)*. If Z matches, the automaton will hit the WHILEM node
3026 of the external loop.
3028 Currently present infoblocks form a tree with a stem formed by PL_curcc
3029 and whatever it mentions via ->next, and additional attached trees
3030 corresponding to temporarily unset infoblocks as in "5" above.
3032 In the following picture infoblocks for outer loop of
3033 (Y(A)*?Z)*?T are denoted O, for inner I. NULL starting block
3034 is denoted by x. The matched string is YAAZYAZT. Temporarily postponed
3035 infoblocks are drawn below the "reset" infoblock.
3037 In fact in the picture below we do not show failed matches for Z and T
3038 by WHILEM blocks. [We illustrate minimal matches, since for them it is
3039 more obvious *why* one needs to *temporary* unset infoblocks.]
3041 Matched REx position InfoBlocks Comment
3045 Y A)*?Z)*?T x <- O <- I
3046 YA )*?Z)*?T x <- O <- I
3047 YA A)*?Z)*?T x <- O <- I
3048 YAA )*?Z)*?T x <- O <- I
3049 YAA Z)*?T x <- O # Temporary unset I
3052 YAAZ Y(A)*?Z)*?T x <- O
3055 YAAZY (A)*?Z)*?T x <- O
3058 YAAZY A)*?Z)*?T x <- O <- I
3061 YAAZYA )*?Z)*?T x <- O <- I
3064 YAAZYA Z)*?T x <- O # Temporary unset I
3070 YAAZYAZ T x # Temporary unset O
3077 *******************************************************************/
3080 CHECKPOINT cp = PL_savestack_ix;
3081 /* No need to save/restore up to this paren */
3082 I32 parenfloor = scan->flags;
3084 if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */
3086 cc.oldcc = PL_regcc;
3088 /* XXXX Probably it is better to teach regpush to support
3089 parenfloor > PL_regsize... */
3090 if (parenfloor > (I32)*PL_reglastparen)
3091 parenfloor = *PL_reglastparen; /* Pessimization... */
3092 cc.parenfloor = parenfloor;
3094 cc.min = ARG1(scan);
3095 cc.max = ARG2(scan);
3096 cc.scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
3100 PL_reginput = locinput;
3101 n = regmatch(PREVOPER(next)); /* start on the WHILEM */
3103 PL_regcc = cc.oldcc;
3109 * This is really hard to understand, because after we match
3110 * what we're trying to match, we must make sure the rest of
3111 * the REx is going to match for sure, and to do that we have
3112 * to go back UP the parse tree by recursing ever deeper. And
3113 * if it fails, we have to reset our parent's current state
3114 * that we can try again after backing off.
3117 CHECKPOINT cp, lastcp;
3118 CURCUR* cc = PL_regcc;
3119 char *lastloc = cc->lastloc; /* Detection of 0-len. */
3121 n = cc->cur + 1; /* how many we know we matched */
3122 PL_reginput = locinput;
3125 PerlIO_printf(Perl_debug_log,
3126 "%*s %ld out of %ld..%ld cc=%"UVxf"\n",
3127 REPORT_CODE_OFF+PL_regindent*2, "",
3128 (long)n, (long)cc->min,
3129 (long)cc->max, PTR2UV(cc))
3132 /* If degenerate scan matches "", assume scan done. */
3134 if (locinput == cc->lastloc && n >= cc->min) {
3135 PL_regcc = cc->oldcc;
3139 PerlIO_printf(Perl_debug_log,
3140 "%*s empty match detected, try continuation...\n",
3141 REPORT_CODE_OFF+PL_regindent*2, "")
3143 if (regmatch(cc->next))
3151 /* First just match a string of min scans. */
3155 cc->lastloc = locinput;
3156 if (regmatch(cc->scan))
3159 cc->lastloc = lastloc;
3164 /* Check whether we already were at this position.
3165 Postpone detection until we know the match is not
3166 *that* much linear. */
3167 if (!PL_reg_maxiter) {
3168 PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4);
3169 PL_reg_leftiter = PL_reg_maxiter;
3171 if (PL_reg_leftiter-- == 0) {
3172 I32 size = (PL_reg_maxiter + 7)/8;
3173 if (PL_reg_poscache) {
3174 if ((I32)PL_reg_poscache_size < size) {
3175 Renew(PL_reg_poscache, size, char);
3176 PL_reg_poscache_size = size;
3178 Zero(PL_reg_poscache, size, char);
3181 PL_reg_poscache_size = size;
3182 Newz(29, PL_reg_poscache, size, char);
3185 PerlIO_printf(Perl_debug_log,
3186 "%sDetected a super-linear match, switching on caching%s...\n",
3187 PL_colors[4], PL_colors[5])
3190 if (PL_reg_leftiter < 0) {
3191 I32 o = locinput - PL_bostr, b;
3193 o = (scan->flags & 0xf) - 1 + o * (scan->flags>>4);
3196 if (PL_reg_poscache[o] & (1<<b)) {
3198 PerlIO_printf(Perl_debug_log,
3199 "%*s already tried at this position...\n",
3200 REPORT_CODE_OFF+PL_regindent*2, "")
3204 PL_reg_poscache[o] |= (1<<b);
3208 /* Prefer next over scan for minimal matching. */
3211 PL_regcc = cc->oldcc;
3214 cp = regcppush(cc->parenfloor);
3216 if (regmatch(cc->next)) {
3218 sayYES; /* All done. */
3220 REGCP_UNWIND(lastcp);
3226 if (n >= cc->max) { /* Maximum greed exceeded? */
3227 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3228 && !(PL_reg_flags & RF_warned)) {
3229 PL_reg_flags |= RF_warned;
3230 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s limit (%d) exceeded",
3231 "Complex regular subexpression recursion",
3238 PerlIO_printf(Perl_debug_log,
3239 "%*s trying longer...\n",
3240 REPORT_CODE_OFF+PL_regindent*2, "")
3242 /* Try scanning more and see if it helps. */
3243 PL_reginput = locinput;
3245 cc->lastloc = locinput;
3246 cp = regcppush(cc->parenfloor);
3248 if (regmatch(cc->scan)) {
3252 REGCP_UNWIND(lastcp);
3255 cc->lastloc = lastloc;
3259 /* Prefer scan over next for maximal matching. */
3261 if (n < cc->max) { /* More greed allowed? */
3262 cp = regcppush(cc->parenfloor);
3264 cc->lastloc = locinput;
3266 if (regmatch(cc->scan)) {
3270 REGCP_UNWIND(lastcp);
3271 regcppop(); /* Restore some previous $<digit>s? */
3272 PL_reginput = locinput;
3274 PerlIO_printf(Perl_debug_log,
3275 "%*s failed, try continuation...\n",
3276 REPORT_CODE_OFF+PL_regindent*2, "")
3279 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3280 && !(PL_reg_flags & RF_warned)) {
3281 PL_reg_flags |= RF_warned;
3282 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s limit (%d) exceeded",
3283 "Complex regular subexpression recursion",
3287 /* Failed deeper matches of scan, so see if this one works. */
3288 PL_regcc = cc->oldcc;
3291 if (regmatch(cc->next))
3297 cc->lastloc = lastloc;
3302 next = scan + ARG(scan);
3305 inner = NEXTOPER(NEXTOPER(scan));
3308 inner = NEXTOPER(scan);
3312 if (OP(next) != c1) /* No choice. */
3313 next = inner; /* Avoid recursion. */
3315 I32 lastparen = *PL_reglastparen;
3317 re_unwind_branch_t *uw;
3319 /* Put unwinding data on stack */
3320 unwind1 = SSNEWt(1,re_unwind_branch_t);
3321 uw = SSPTRt(unwind1,re_unwind_branch_t);
3324 uw->type = ((c1 == BRANCH)
3326 : RE_UNWIND_BRANCHJ);
3327 uw->lastparen = lastparen;
3329 uw->locinput = locinput;
3330 uw->nextchr = nextchr;
3332 uw->regindent = ++PL_regindent;
3335 REGCP_SET(uw->lastcp);
3337 /* Now go into the first branch */
3350 /* We suppose that the next guy does not need
3351 backtracking: in particular, it is of constant length,
3352 and has no parenths to influence future backrefs. */
3353 ln = ARG1(scan); /* min to match */
3354 n = ARG2(scan); /* max to match */
3355 paren = scan->flags;
3357 if (paren > PL_regsize)
3359 if (paren > (I32)*PL_reglastparen)
3360 *PL_reglastparen = paren;
3362 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3364 scan += NEXT_OFF(scan); /* Skip former OPEN. */
3365 PL_reginput = locinput;
3368 if (ln && regrepeat_hard(scan, ln, &l) < ln)
3370 /* if we matched something zero-length we don't need to
3371 backtrack - capturing parens are already defined, so
3372 the caveat in the maximal case doesn't apply
3374 XXXX if ln == 0, we can redo this check first time
3375 through the following loop
3378 n = ln; /* don't backtrack */
3379 locinput = PL_reginput;
3380 if (HAS_TEXT(next) || JUMPABLE(next)) {
3381 regnode *text_node = next;
3383 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3385 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3387 if (PL_regkind[(U8)OP(text_node)] == REF) {
3389 n = ARG(text_node); /* which paren pair */
3390 ln = PL_regstartp[n];
3391 /* assume yes if we haven't seen CLOSEn */
3393 (I32)*PL_reglastparen < n ||
3400 c1 = *(PL_bostr + ln);
3402 else { c1 = (U8)*STRING(text_node); }
3403 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3405 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3406 c2 = PL_fold_locale[c1];
3415 /* This may be improved if l == 0. */
3416 while (n >= ln || (n == REG_INFTY && ln > 0 && l)) { /* ln overflow ? */
3417 /* If it could work, try it. */
3419 UCHARAT(PL_reginput) == c1 ||
3420 UCHARAT(PL_reginput) == c2)
3424 PL_regstartp[paren] =
3425 HOPc(PL_reginput, -l) - PL_bostr;
3426 PL_regendp[paren] = PL_reginput - PL_bostr;
3429 PL_regendp[paren] = -1;
3433 REGCP_UNWIND(lastcp);
3435 /* Couldn't or didn't -- move forward. */
3436 PL_reginput = locinput;
3437 if (regrepeat_hard(scan, 1, &l)) {
3439 locinput = PL_reginput;
3446 n = regrepeat_hard(scan, n, &l);
3447 /* if we matched something zero-length we don't need to
3448 backtrack, unless the minimum count is zero and we
3449 are capturing the result - in that case the capture
3450 being defined or not may affect later execution
3452 if (n != 0 && l == 0 && !(paren && ln == 0))
3453 ln = n; /* don't backtrack */
3454 locinput = PL_reginput;
3456 PerlIO_printf(Perl_debug_log,
3457 "%*s matched %"IVdf" times, len=%"IVdf"...\n",
3458 (int)(REPORT_CODE_OFF+PL_regindent*2), "",
3462 if (HAS_TEXT(next) || JUMPABLE(next)) {
3463 regnode *text_node = next;
3465 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3467 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3469 if (PL_regkind[(U8)OP(text_node)] == REF) {
3471 n = ARG(text_node); /* which paren pair */
3472 ln = PL_regstartp[n];
3473 /* assume yes if we haven't seen CLOSEn */
3475 (I32)*PL_reglastparen < n ||
3482 c1 = *(PL_bostr + ln);
3484 else { c1 = (U8)*STRING(text_node); }
3486 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3488 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3489 c2 = PL_fold_locale[c1];
3500 /* If it could work, try it. */
3502 UCHARAT(PL_reginput) == c1 ||
3503 UCHARAT(PL_reginput) == c2)
3506 PerlIO_printf(Perl_debug_log,
3507 "%*s trying tail with n=%"IVdf"...\n",
3508 (int)(REPORT_CODE_OFF+PL_regindent*2), "", (IV)n)
3512 PL_regstartp[paren] = HOPc(PL_reginput, -l) - PL_bostr;
3513 PL_regendp[paren] = PL_reginput - PL_bostr;
3516 PL_regendp[paren] = -1;
3520 REGCP_UNWIND(lastcp);
3522 /* Couldn't or didn't -- back up. */
3524 locinput = HOPc(locinput, -l);
3525 PL_reginput = locinput;
3532 paren = scan->flags; /* Which paren to set */
3533 if (paren > PL_regsize)
3535 if (paren > (I32)*PL_reglastparen)
3536 *PL_reglastparen = paren;
3537 ln = ARG1(scan); /* min to match */
3538 n = ARG2(scan); /* max to match */
3539 scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE);
3543 ln = ARG1(scan); /* min to match */
3544 n = ARG2(scan); /* max to match */
3545 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3550 scan = NEXTOPER(scan);
3556 scan = NEXTOPER(scan);
3560 * Lookahead to avoid useless match attempts
3561 * when we know what character comes next.
3565 * Used to only do .*x and .*?x, but now it allows
3566 * for )'s, ('s and (?{ ... })'s to be in the way
3567 * of the quantifier and the EXACT-like node. -- japhy
3570 if (HAS_TEXT(next) || JUMPABLE(next)) {
3572 regnode *text_node = next;
3574 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3576 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3578 if (PL_regkind[(U8)OP(text_node)] == REF) {
3580 n = ARG(text_node); /* which paren pair */
3581 ln = PL_regstartp[n];
3582 /* assume yes if we haven't seen CLOSEn */
3584 (I32)*PL_reglastparen < n ||
3589 goto assume_ok_easy;
3591 s = (U8*)PL_bostr + ln;
3593 else { s = (U8*)STRING(text_node); }
3597 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3599 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3600 c2 = PL_fold_locale[c1];
3603 if (OP(text_node) == EXACTF || OP(text_node) == REFF) {
3604 STRLEN ulen1, ulen2;
3605 U8 tmpbuf1[UTF8_MAXLEN_UCLC+1];
3606 U8 tmpbuf2[UTF8_MAXLEN_UCLC+1];
3608 to_utf8_lower((U8*)s, tmpbuf1, &ulen1);
3609 to_utf8_upper((U8*)s, tmpbuf2, &ulen2);
3611 c1 = utf8n_to_uvuni(tmpbuf1, UTF8_MAXLEN, 0,
3613 0 : UTF8_ALLOW_ANY);
3614 c2 = utf8n_to_uvuni(tmpbuf2, UTF8_MAXLEN, 0,
3616 0 : UTF8_ALLOW_ANY);
3619 c2 = c1 = utf8n_to_uvchr(s, UTF8_MAXLEN, 0,
3621 0 : UTF8_ALLOW_ANY);
3629 PL_reginput = locinput;
3633 if (ln && regrepeat(scan, ln) < ln)
3635 locinput = PL_reginput;
3638 char *e; /* Should not check after this */
3639 char *old = locinput;
3642 if (n == REG_INFTY) {
3645 while (UTF8_IS_CONTINUATION(*(U8*)e))
3651 m >0 && e + UTF8SKIP(e) <= PL_regeol; m--)
3655 e = locinput + n - ln;
3660 /* Find place 'next' could work */
3663 while (locinput <= e &&
3664 UCHARAT(locinput) != c1)
3667 while (locinput <= e
3668 && UCHARAT(locinput) != c1
3669 && UCHARAT(locinput) != c2)
3672 count = locinput - old;
3677 /* count initialised to
3678 * utf8_distance(old, locinput) */
3679 while (locinput <= e &&
3680 utf8n_to_uvchr((U8*)locinput,
3683 0 : UTF8_ALLOW_ANY) != (UV)c1) {
3688 /* count initialised to
3689 * utf8_distance(old, locinput) */
3690 while (locinput <= e) {
3691 UV c = utf8n_to_uvchr((U8*)locinput,
3694 0 : UTF8_ALLOW_ANY);
3695 if (c == (UV)c1 || c == (UV)c2)
3704 /* PL_reginput == old now */
3705 if (locinput != old) {
3706 ln = 1; /* Did some */
3707 if (regrepeat(scan, count) < count)
3710 /* PL_reginput == locinput now */
3711 TRYPAREN(paren, ln, locinput);
3712 PL_reginput = locinput; /* Could be reset... */
3713 REGCP_UNWIND(lastcp);
3714 /* Couldn't or didn't -- move forward. */
3717 locinput += UTF8SKIP(locinput);
3724 while (n >= ln || (n == REG_INFTY && ln > 0)) { /* ln overflow ? */
3728 c = utf8n_to_uvchr((U8*)PL_reginput,
3731 0 : UTF8_ALLOW_ANY);
3733 c = UCHARAT(PL_reginput);
3734 /* If it could work, try it. */
3735 if (c == (UV)c1 || c == (UV)c2)
3737 TRYPAREN(paren, ln, PL_reginput);
3738 REGCP_UNWIND(lastcp);
3741 /* If it could work, try it. */
3742 else if (c1 == -1000)
3744 TRYPAREN(paren, ln, PL_reginput);
3745 REGCP_UNWIND(lastcp);
3747 /* Couldn't or didn't -- move forward. */
3748 PL_reginput = locinput;
3749 if (regrepeat(scan, 1)) {
3751 locinput = PL_reginput;
3759 n = regrepeat(scan, n);
3760 locinput = PL_reginput;
3761 if (ln < n && PL_regkind[(U8)OP(next)] == EOL &&
3762 ((!PL_multiline && OP(next) != MEOL) ||
3763 OP(next) == SEOL || OP(next) == EOS))
3765 ln = n; /* why back off? */
3766 /* ...because $ and \Z can match before *and* after
3767 newline at the end. Consider "\n\n" =~ /\n+\Z\n/.
3768 We should back off by one in this case. */
3769 if (UCHARAT(PL_reginput - 1) == '\n' && OP(next) != EOS)
3778 c = utf8n_to_uvchr((U8*)PL_reginput,
3781 0 : UTF8_ALLOW_ANY);
3783 c = UCHARAT(PL_reginput);
3785 /* If it could work, try it. */
3786 if (c1 == -1000 || c == (UV)c1 || c == (UV)c2)
3788 TRYPAREN(paren, n, PL_reginput);
3789 REGCP_UNWIND(lastcp);
3791 /* Couldn't or didn't -- back up. */
3793 PL_reginput = locinput = HOPc(locinput, -1);
3801 c = utf8n_to_uvchr((U8*)PL_reginput,
3804 0 : UTF8_ALLOW_ANY);
3806 c = UCHARAT(PL_reginput);
3808 /* If it could work, try it. */
3809 if (c1 == -1000 || c == (UV)c1 || c == (UV)c2)
3811 TRYPAREN(paren, n, PL_reginput);
3812 REGCP_UNWIND(lastcp);
3814 /* Couldn't or didn't -- back up. */
3816 PL_reginput = locinput = HOPc(locinput, -1);
3823 if (PL_reg_call_cc) {
3824 re_cc_state *cur_call_cc = PL_reg_call_cc;
3825 CURCUR *cctmp = PL_regcc;
3826 regexp *re = PL_reg_re;
3827 CHECKPOINT cp, lastcp;
3829 cp = regcppush(0); /* Save *all* the positions. */
3831 regcp_set_to(PL_reg_call_cc->ss); /* Restore parens of
3833 PL_reginput = locinput; /* Make position available to
3835 cache_re(PL_reg_call_cc->re);
3836 PL_regcc = PL_reg_call_cc->cc;
3837 PL_reg_call_cc = PL_reg_call_cc->prev;
3838 if (regmatch(cur_call_cc->node)) {
3839 PL_reg_call_cc = cur_call_cc;
3843 REGCP_UNWIND(lastcp);
3845 PL_reg_call_cc = cur_call_cc;
3851 PerlIO_printf(Perl_debug_log,
3852 "%*s continuation failed...\n",
3853 REPORT_CODE_OFF+PL_regindent*2, "")
3857 if (locinput < PL_regtill) {
3858 DEBUG_r(PerlIO_printf(Perl_debug_log,
3859 "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n",
3861 (long)(locinput - PL_reg_starttry),
3862 (long)(PL_regtill - PL_reg_starttry),
3864 sayNO_FINAL; /* Cannot match: too short. */
3866 PL_reginput = locinput; /* put where regtry can find it */
3867 sayYES_FINAL; /* Success! */
3869 PL_reginput = locinput; /* put where regtry can find it */
3870 sayYES_LOUD; /* Success! */
3873 PL_reginput = locinput;
3878 s = HOPBACKc(locinput, scan->flags);
3884 PL_reginput = locinput;
3889 s = HOPBACKc(locinput, scan->flags);
3895 PL_reginput = locinput;
3898 inner = NEXTOPER(NEXTOPER(scan));
3899 if (regmatch(inner) != n) {
3914 if (OP(scan) == SUSPEND) {
3915 locinput = PL_reginput;
3916 nextchr = UCHARAT(locinput);
3921 next = scan + ARG(scan);
3926 PerlIO_printf(Perl_error_log, "%"UVxf" %d\n",
3927 PTR2UV(scan), OP(scan));
3928 Perl_croak(aTHX_ "regexp memory corruption");
3935 * We get here only if there's trouble -- normally "case END" is
3936 * the terminating point.
3938 Perl_croak(aTHX_ "corrupted regexp pointers");
3944 PerlIO_printf(Perl_debug_log,
3945 "%*s %scould match...%s\n",
3946 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4],PL_colors[5])
3950 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n",
3951 PL_colors[4],PL_colors[5]));
3957 #if 0 /* Breaks $^R */
3965 PerlIO_printf(Perl_debug_log,
3966 "%*s %sfailed...%s\n",
3967 REPORT_CODE_OFF+PL_regindent*2, "",PL_colors[4],PL_colors[5])
3973 re_unwind_t *uw = SSPTRt(unwind,re_unwind_t);
3976 case RE_UNWIND_BRANCH:
3977 case RE_UNWIND_BRANCHJ:
3979 re_unwind_branch_t *uwb = &(uw->branch);
3980 I32 lastparen = uwb->lastparen;
3982 REGCP_UNWIND(uwb->lastcp);
3983 for (n = *PL_reglastparen; n > lastparen; n--)
3985 *PL_reglastparen = n;
3986 scan = next = uwb->next;
3988 OP(scan) != (uwb->type == RE_UNWIND_BRANCH
3989 ? BRANCH : BRANCHJ) ) { /* Failure */
3996 /* Have more choice yet. Reuse the same uwb. */
3998 if ((n = (uwb->type == RE_UNWIND_BRANCH
3999 ? NEXT_OFF(next) : ARG(next))))
4002 next = NULL; /* XXXX Needn't unwinding in this case... */
4004 next = NEXTOPER(scan);
4005 if (uwb->type == RE_UNWIND_BRANCHJ)
4006 next = NEXTOPER(next);
4007 locinput = uwb->locinput;
4008 nextchr = uwb->nextchr;
4010 PL_regindent = uwb->regindent;
4017 Perl_croak(aTHX_ "regexp unwind memory corruption");
4028 - regrepeat - repeatedly match something simple, report how many
4031 * [This routine now assumes that it will only match on things of length 1.
4032 * That was true before, but now we assume scan - reginput is the count,
4033 * rather than incrementing count on every character. [Er, except utf8.]]
4036 S_regrepeat(pTHX_ regnode *p, I32 max)
4038 register char *scan;
4040 register char *loceol = PL_regeol;
4041 register I32 hardcount = 0;
4042 register bool do_utf8 = PL_reg_match_utf8;
4045 if (max == REG_INFTY)
4047 else if (max < loceol - scan)
4048 loceol = scan + max;
4053 while (scan < loceol && hardcount < max && *scan != '\n') {
4054 scan += UTF8SKIP(scan);
4058 while (scan < loceol && *scan != '\n')
4065 while (scan < loceol && hardcount < max) {
4066 scan += UTF8SKIP(scan);
4076 case EXACT: /* length of string is 1 */
4078 while (scan < loceol && UCHARAT(scan) == c)
4081 case EXACTF: /* length of string is 1 */
4083 while (scan < loceol &&
4084 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold[c]))
4087 case EXACTFL: /* length of string is 1 */
4088 PL_reg_flags |= RF_tainted;
4090 while (scan < loceol &&
4091 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold_locale[c]))
4097 while (hardcount < max && scan < loceol &&
4098 reginclass(p, (U8*)scan, 0, do_utf8)) {
4099 scan += UTF8SKIP(scan);
4103 while (scan < loceol && REGINCLASS(p, (U8*)scan))
4110 LOAD_UTF8_CHARCLASS(alnum,"a");
4111 while (hardcount < max && scan < loceol &&
4112 swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
4113 scan += UTF8SKIP(scan);
4117 while (scan < loceol && isALNUM(*scan))
4122 PL_reg_flags |= RF_tainted;
4125 while (hardcount < max && scan < loceol &&
4126 isALNUM_LC_utf8((U8*)scan)) {
4127 scan += UTF8SKIP(scan);
4131 while (scan < loceol && isALNUM_LC(*scan))
4138 LOAD_UTF8_CHARCLASS(alnum,"a");
4139 while (hardcount < max && scan < loceol &&
4140 !swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
4141 scan += UTF8SKIP(scan);
4145 while (scan < loceol && !isALNUM(*scan))
4150 PL_reg_flags |= RF_tainted;
4153 while (hardcount < max && scan < loceol &&
4154 !isALNUM_LC_utf8((U8*)scan)) {
4155 scan += UTF8SKIP(scan);
4159 while (scan < loceol && !isALNUM_LC(*scan))
4166 LOAD_UTF8_CHARCLASS(space," ");
4167 while (hardcount < max && scan < loceol &&
4169 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
4170 scan += UTF8SKIP(scan);
4174 while (scan < loceol && isSPACE(*scan))
4179 PL_reg_flags |= RF_tainted;
4182 while (hardcount < max && scan < loceol &&
4183 (*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
4184 scan += UTF8SKIP(scan);
4188 while (scan < loceol && isSPACE_LC(*scan))
4195 LOAD_UTF8_CHARCLASS(space," ");
4196 while (hardcount < max && scan < loceol &&
4198 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
4199 scan += UTF8SKIP(scan);
4203 while (scan < loceol && !isSPACE(*scan))
4208 PL_reg_flags |= RF_tainted;
4211 while (hardcount < max && scan < loceol &&
4212 !(*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
4213 scan += UTF8SKIP(scan);
4217 while (scan < loceol && !isSPACE_LC(*scan))
4224 LOAD_UTF8_CHARCLASS(digit,"0");
4225 while (hardcount < max && scan < loceol &&
4226 swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
4227 scan += UTF8SKIP(scan);
4231 while (scan < loceol && isDIGIT(*scan))
4238 LOAD_UTF8_CHARCLASS(digit,"0");
4239 while (hardcount < max && scan < loceol &&
4240 !swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
4241 scan += UTF8SKIP(scan);
4245 while (scan < loceol && !isDIGIT(*scan))
4249 default: /* Called on something of 0 width. */
4250 break; /* So match right here or not at all. */
4256 c = scan - PL_reginput;
4261 SV *prop = sv_newmortal();
4264 PerlIO_printf(Perl_debug_log,
4265 "%*s %s can match %"IVdf" times out of %"IVdf"...\n",
4266 REPORT_CODE_OFF+1, "", SvPVX(prop),(IV)c,(IV)max);
4273 - regrepeat_hard - repeatedly match something, report total lenth and length
4275 * The repeater is supposed to have constant length.
4279 S_regrepeat_hard(pTHX_ regnode *p, I32 max, I32 *lp)
4281 register char *scan = Nullch;
4282 register char *start;
4283 register char *loceol = PL_regeol;
4285 I32 count = 0, res = 1;
4290 start = PL_reginput;
4291 if (PL_reg_match_utf8) {
4292 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
4295 while (start < PL_reginput) {
4297 start += UTF8SKIP(start);
4308 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
4310 *lp = l = PL_reginput - start;
4311 if (max != REG_INFTY && l*max < loceol - scan)
4312 loceol = scan + l*max;
4325 - regclass_swash - prepare the utf8 swash
4329 Perl_regclass_swash(pTHX_ register regnode* node, bool doinit, SV** listsvp, SV **altsvp)
4335 if (PL_regdata && PL_regdata->count) {
4338 if (PL_regdata->what[n] == 's') {
4339 SV *rv = (SV*)PL_regdata->data[n];
4340 AV *av = (AV*)SvRV((SV*)rv);
4341 SV **ary = AvARRAY(av);
4344 /* See the end of regcomp.c:S_reglass() for
4345 * documentation of these array elements. */
4348 a = SvTYPE(ary[1]) == SVt_RV ? &ary[1] : 0;
4349 b = SvTYPE(ary[2]) == SVt_PVAV ? &ary[2] : 0;
4353 else if (si && doinit) {
4354 sw = swash_init("utf8", "", si, 1, 0);
4355 (void)av_store(av, 1, sw);
4371 - reginclass - determine if a character falls into a character class
4373 The n is the ANYOF regnode, the p is the target string, lenp
4374 is pointer to the maximum length of how far to go in the p
4375 (if the lenp is zero, UTF8SKIP(p) is used),
4376 do_utf8 tells whether the target string is in UTF-8.
4381 S_reginclass(pTHX_ register regnode *n, register U8* p, STRLEN* lenp, register bool do_utf8)
4383 char flags = ANYOF_FLAGS(n);
4389 if (do_utf8 && !UTF8_IS_INVARIANT(c))
4390 c = utf8n_to_uvchr(p, UTF8_MAXLEN, &len,
4391 ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
4393 plen = lenp ? *lenp : UNISKIP(NATIVE_TO_UNI(c));
4394 if (do_utf8 || (flags & ANYOF_UNICODE)) {
4397 if (do_utf8 && !ANYOF_RUNTIME(n)) {
4398 if (len != (STRLEN)-1 && c < 256 && ANYOF_BITMAP_TEST(n, c))
4401 if (!match && do_utf8 && (flags & ANYOF_UNICODE_ALL) && c >= 256)
4405 SV *sw = regclass_swash(n, TRUE, 0, (SV**)&av);
4408 if (swash_fetch(sw, p, do_utf8))
4410 else if (flags & ANYOF_FOLD) {
4411 if (!match && lenp && av) {
4414 for (i = 0; i <= av_len(av); i++) {
4415 SV* sv = *av_fetch(av, i, FALSE);
4417 char *s = SvPV(sv, len);
4419 if (len <= plen && memEQ(s, (char*)p, len)) {
4427 U8 tmpbuf[UTF8_MAXLEN_FOLD+1];
4430 to_utf8_fold(p, tmpbuf, &tmplen);
4431 if (swash_fetch(sw, tmpbuf, do_utf8))
4437 if (match && lenp && *lenp == 0)
4438 *lenp = UNISKIP(NATIVE_TO_UNI(c));
4440 if (!match && c < 256) {
4441 if (ANYOF_BITMAP_TEST(n, c))
4443 else if (flags & ANYOF_FOLD) {
4446 if (flags & ANYOF_LOCALE) {
4447 PL_reg_flags |= RF_tainted;
4448 f = PL_fold_locale[c];
4452 if (f != c && ANYOF_BITMAP_TEST(n, f))
4456 if (!match && (flags & ANYOF_CLASS)) {
4457 PL_reg_flags |= RF_tainted;
4459 (ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) ||
4460 (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) ||
4461 (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) ||
4462 (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) ||
4463 (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) ||
4464 (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) ||
4465 (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) ||
4466 (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) ||
4467 (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) ||
4468 (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) ||
4469 (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII(c)) ||
4470 (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII(c)) ||
4471 (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) ||
4472 (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) ||
4473 (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) ||
4474 (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) ||
4475 (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) ||
4476 (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) ||
4477 (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) ||
4478 (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) ||
4479 (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) ||
4480 (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) ||
4481 (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) ||
4482 (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) ||
4483 (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) ||
4484 (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) ||
4485 (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) ||
4486 (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) ||
4487 (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK(c)) ||
4488 (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK(c))
4489 ) /* How's that for a conditional? */
4496 return (flags & ANYOF_INVERT) ? !match : match;
4500 S_reghop(pTHX_ U8 *s, I32 off)
4502 return S_reghop3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4506 S_reghop3(pTHX_ U8 *s, I32 off, U8* lim)
4509 while (off-- && s < lim) {
4510 /* XXX could check well-formedness here */
4518 if (UTF8_IS_CONTINUED(*s)) {
4519 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4522 /* XXX could check well-formedness here */
4530 S_reghopmaybe(pTHX_ U8 *s, I32 off)
4532 return S_reghopmaybe3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4536 S_reghopmaybe3(pTHX_ U8* s, I32 off, U8* lim)
4539 while (off-- && s < lim) {
4540 /* XXX could check well-formedness here */
4550 if (UTF8_IS_CONTINUED(*s)) {
4551 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4554 /* XXX could check well-formedness here */
4566 restore_pos(pTHX_ void *arg)
4568 if (PL_reg_eval_set) {
4569 if (PL_reg_oldsaved) {
4570 PL_reg_re->subbeg = PL_reg_oldsaved;
4571 PL_reg_re->sublen = PL_reg_oldsavedlen;
4572 #ifdef PERL_COPY_ON_WRITE
4573 PL_reg_re->saved_copy = PL_nrs;
4575 RX_MATCH_COPIED_on(PL_reg_re);
4577 PL_reg_magic->mg_len = PL_reg_oldpos;
4578 PL_reg_eval_set = 0;
4579 PL_curpm = PL_reg_oldcurpm;
4584 S_to_utf8_substr(pTHX_ register regexp *prog)
4587 if (prog->float_substr && !prog->float_utf8) {
4588 prog->float_utf8 = sv = NEWSV(117, 0);
4589 SvSetSV(sv, prog->float_substr);
4590 sv_utf8_upgrade(sv);
4591 if (SvTAIL(prog->float_substr))
4593 if (prog->float_substr == prog->check_substr)
4594 prog->check_utf8 = sv;
4596 if (prog->anchored_substr && !prog->anchored_utf8) {
4597 prog->anchored_utf8 = sv = NEWSV(118, 0);
4598 SvSetSV(sv, prog->anchored_substr);
4599 sv_utf8_upgrade(sv);
4600 if (SvTAIL(prog->anchored_substr))
4602 if (prog->anchored_substr == prog->check_substr)
4603 prog->check_utf8 = sv;
4608 S_to_byte_substr(pTHX_ register regexp *prog)
4611 if (prog->float_utf8 && !prog->float_substr) {
4612 prog->float_substr = sv = NEWSV(117, 0);
4613 SvSetSV(sv, prog->float_utf8);
4614 if (sv_utf8_downgrade(sv, TRUE)) {
4615 if (SvTAIL(prog->float_utf8))
4619 prog->float_substr = sv = &PL_sv_undef;
4621 if (prog->float_utf8 == prog->check_utf8)
4622 prog->check_substr = sv;
4624 if (prog->anchored_utf8 && !prog->anchored_substr) {
4625 prog->anchored_substr = sv = NEWSV(118, 0);
4626 SvSetSV(sv, prog->anchored_utf8);
4627 if (sv_utf8_downgrade(sv, TRUE)) {
4628 if (SvTAIL(prog->anchored_utf8))
4632 prog->anchored_substr = sv = &PL_sv_undef;
4634 if (prog->anchored_utf8 == prog->check_utf8)
4635 prog->check_substr = sv;