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) 1991-2003, 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;
2872 if(SvROK(ret) && SvSMAGICAL(sv = SvRV(ret)))
2873 mg = mg_find(sv, PERL_MAGIC_qr);
2874 else if (SvSMAGICAL(ret)) {
2875 if (SvGMAGICAL(ret))
2876 sv_unmagic(ret, PERL_MAGIC_qr);
2878 mg = mg_find(ret, PERL_MAGIC_qr);
2882 re = (regexp *)mg->mg_obj;
2883 (void)ReREFCNT_inc(re);
2887 char *t = SvPV(ret, len);
2889 char *oprecomp = PL_regprecomp;
2890 I32 osize = PL_regsize;
2891 I32 onpar = PL_regnpar;
2894 if (DO_UTF8(ret)) pm.op_pmdynflags |= PMdf_DYN_UTF8;
2895 re = CALLREGCOMP(aTHX_ t, t + len, &pm);
2897 & (SVs_TEMP | SVs_PADTMP | SVf_READONLY
2899 sv_magic(ret,(SV*)ReREFCNT_inc(re),
2901 PL_regprecomp = oprecomp;
2906 PerlIO_printf(Perl_debug_log,
2907 "Entering embedded `%s%.60s%s%s'\n",
2911 (strlen(re->precomp) > 60 ? "..." : ""))
2914 state.prev = PL_reg_call_cc;
2915 state.cc = PL_regcc;
2916 state.re = PL_reg_re;
2920 cp = regcppush(0); /* Save *all* the positions. */
2923 state.ss = PL_savestack_ix;
2924 *PL_reglastparen = 0;
2925 *PL_reglastcloseparen = 0;
2926 PL_reg_call_cc = &state;
2927 PL_reginput = locinput;
2928 toggleutf = ((PL_reg_flags & RF_utf8) != 0) ^
2929 ((re->reganch & ROPT_UTF8) != 0);
2930 if (toggleutf) PL_reg_flags ^= RF_utf8;
2932 /* XXXX This is too dramatic a measure... */
2935 if (regmatch(re->program + 1)) {
2936 /* Even though we succeeded, we need to restore
2937 global variables, since we may be wrapped inside
2938 SUSPEND, thus the match may be not finished yet. */
2940 /* XXXX Do this only if SUSPENDed? */
2941 PL_reg_call_cc = state.prev;
2942 PL_regcc = state.cc;
2943 PL_reg_re = state.re;
2944 cache_re(PL_reg_re);
2945 if (toggleutf) PL_reg_flags ^= RF_utf8;
2947 /* XXXX This is too dramatic a measure... */
2950 /* These are needed even if not SUSPEND. */
2956 REGCP_UNWIND(lastcp);
2958 PL_reg_call_cc = state.prev;
2959 PL_regcc = state.cc;
2960 PL_reg_re = state.re;
2961 cache_re(PL_reg_re);
2962 if (toggleutf) PL_reg_flags ^= RF_utf8;
2964 /* XXXX This is too dramatic a measure... */
2974 sv_setsv(save_scalar(PL_replgv), ret);
2978 n = ARG(scan); /* which paren pair */
2979 PL_reg_start_tmp[n] = locinput;
2984 n = ARG(scan); /* which paren pair */
2985 PL_regstartp[n] = PL_reg_start_tmp[n] - PL_bostr;
2986 PL_regendp[n] = locinput - PL_bostr;
2987 if (n > (I32)*PL_reglastparen)
2988 *PL_reglastparen = n;
2989 *PL_reglastcloseparen = n;
2992 n = ARG(scan); /* which paren pair */
2993 sw = ((I32)*PL_reglastparen >= n && PL_regendp[n] != -1);
2996 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2998 next = NEXTOPER(NEXTOPER(scan));
3000 next = scan + ARG(scan);
3001 if (OP(next) == IFTHEN) /* Fake one. */
3002 next = NEXTOPER(NEXTOPER(next));
3006 logical = scan->flags;
3008 /*******************************************************************
3009 PL_regcc contains infoblock about the innermost (...)* loop, and
3010 a pointer to the next outer infoblock.
3012 Here is how Y(A)*Z is processed (if it is compiled into CURLYX/WHILEM):
3014 1) After matching X, regnode for CURLYX is processed;
3016 2) This regnode creates infoblock on the stack, and calls
3017 regmatch() recursively with the starting point at WHILEM node;
3019 3) Each hit of WHILEM node tries to match A and Z (in the order
3020 depending on the current iteration, min/max of {min,max} and
3021 greediness). The information about where are nodes for "A"
3022 and "Z" is read from the infoblock, as is info on how many times "A"
3023 was already matched, and greediness.
3025 4) After A matches, the same WHILEM node is hit again.
3027 5) Each time WHILEM is hit, PL_regcc is the infoblock created by CURLYX
3028 of the same pair. Thus when WHILEM tries to match Z, it temporarily
3029 resets PL_regcc, since this Y(A)*Z can be a part of some other loop:
3030 as in (Y(A)*Z)*. If Z matches, the automaton will hit the WHILEM node
3031 of the external loop.
3033 Currently present infoblocks form a tree with a stem formed by PL_curcc
3034 and whatever it mentions via ->next, and additional attached trees
3035 corresponding to temporarily unset infoblocks as in "5" above.
3037 In the following picture infoblocks for outer loop of
3038 (Y(A)*?Z)*?T are denoted O, for inner I. NULL starting block
3039 is denoted by x. The matched string is YAAZYAZT. Temporarily postponed
3040 infoblocks are drawn below the "reset" infoblock.
3042 In fact in the picture below we do not show failed matches for Z and T
3043 by WHILEM blocks. [We illustrate minimal matches, since for them it is
3044 more obvious *why* one needs to *temporary* unset infoblocks.]
3046 Matched REx position InfoBlocks Comment
3050 Y A)*?Z)*?T x <- O <- I
3051 YA )*?Z)*?T x <- O <- I
3052 YA A)*?Z)*?T x <- O <- I
3053 YAA )*?Z)*?T x <- O <- I
3054 YAA Z)*?T x <- O # Temporary unset I
3057 YAAZ Y(A)*?Z)*?T x <- O
3060 YAAZY (A)*?Z)*?T x <- O
3063 YAAZY A)*?Z)*?T x <- O <- I
3066 YAAZYA )*?Z)*?T x <- O <- I
3069 YAAZYA Z)*?T x <- O # Temporary unset I
3075 YAAZYAZ T x # Temporary unset O
3082 *******************************************************************/
3085 CHECKPOINT cp = PL_savestack_ix;
3086 /* No need to save/restore up to this paren */
3087 I32 parenfloor = scan->flags;
3089 if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */
3091 cc.oldcc = PL_regcc;
3093 /* XXXX Probably it is better to teach regpush to support
3094 parenfloor > PL_regsize... */
3095 if (parenfloor > (I32)*PL_reglastparen)
3096 parenfloor = *PL_reglastparen; /* Pessimization... */
3097 cc.parenfloor = parenfloor;
3099 cc.min = ARG1(scan);
3100 cc.max = ARG2(scan);
3101 cc.scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
3105 PL_reginput = locinput;
3106 n = regmatch(PREVOPER(next)); /* start on the WHILEM */
3108 PL_regcc = cc.oldcc;
3114 * This is really hard to understand, because after we match
3115 * what we're trying to match, we must make sure the rest of
3116 * the REx is going to match for sure, and to do that we have
3117 * to go back UP the parse tree by recursing ever deeper. And
3118 * if it fails, we have to reset our parent's current state
3119 * that we can try again after backing off.
3122 CHECKPOINT cp, lastcp;
3123 CURCUR* cc = PL_regcc;
3124 char *lastloc = cc->lastloc; /* Detection of 0-len. */
3126 n = cc->cur + 1; /* how many we know we matched */
3127 PL_reginput = locinput;
3130 PerlIO_printf(Perl_debug_log,
3131 "%*s %ld out of %ld..%ld cc=%"UVxf"\n",
3132 REPORT_CODE_OFF+PL_regindent*2, "",
3133 (long)n, (long)cc->min,
3134 (long)cc->max, PTR2UV(cc))
3137 /* If degenerate scan matches "", assume scan done. */
3139 if (locinput == cc->lastloc && n >= cc->min) {
3140 PL_regcc = cc->oldcc;
3144 PerlIO_printf(Perl_debug_log,
3145 "%*s empty match detected, try continuation...\n",
3146 REPORT_CODE_OFF+PL_regindent*2, "")
3148 if (regmatch(cc->next))
3156 /* First just match a string of min scans. */
3160 cc->lastloc = locinput;
3161 if (regmatch(cc->scan))
3164 cc->lastloc = lastloc;
3169 /* Check whether we already were at this position.
3170 Postpone detection until we know the match is not
3171 *that* much linear. */
3172 if (!PL_reg_maxiter) {
3173 PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4);
3174 PL_reg_leftiter = PL_reg_maxiter;
3176 if (PL_reg_leftiter-- == 0) {
3177 I32 size = (PL_reg_maxiter + 7)/8;
3178 if (PL_reg_poscache) {
3179 if ((I32)PL_reg_poscache_size < size) {
3180 Renew(PL_reg_poscache, size, char);
3181 PL_reg_poscache_size = size;
3183 Zero(PL_reg_poscache, size, char);
3186 PL_reg_poscache_size = size;
3187 Newz(29, PL_reg_poscache, size, char);
3190 PerlIO_printf(Perl_debug_log,
3191 "%sDetected a super-linear match, switching on caching%s...\n",
3192 PL_colors[4], PL_colors[5])
3195 if (PL_reg_leftiter < 0) {
3196 I32 o = locinput - PL_bostr, b;
3198 o = (scan->flags & 0xf) - 1 + o * (scan->flags>>4);
3201 if (PL_reg_poscache[o] & (1<<b)) {
3203 PerlIO_printf(Perl_debug_log,
3204 "%*s already tried at this position...\n",
3205 REPORT_CODE_OFF+PL_regindent*2, "")
3209 PL_reg_poscache[o] |= (1<<b);
3213 /* Prefer next over scan for minimal matching. */
3216 PL_regcc = cc->oldcc;
3219 cp = regcppush(cc->parenfloor);
3221 if (regmatch(cc->next)) {
3223 sayYES; /* All done. */
3225 REGCP_UNWIND(lastcp);
3231 if (n >= cc->max) { /* Maximum greed exceeded? */
3232 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3233 && !(PL_reg_flags & RF_warned)) {
3234 PL_reg_flags |= RF_warned;
3235 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s limit (%d) exceeded",
3236 "Complex regular subexpression recursion",
3243 PerlIO_printf(Perl_debug_log,
3244 "%*s trying longer...\n",
3245 REPORT_CODE_OFF+PL_regindent*2, "")
3247 /* Try scanning more and see if it helps. */
3248 PL_reginput = locinput;
3250 cc->lastloc = locinput;
3251 cp = regcppush(cc->parenfloor);
3253 if (regmatch(cc->scan)) {
3257 REGCP_UNWIND(lastcp);
3260 cc->lastloc = lastloc;
3264 /* Prefer scan over next for maximal matching. */
3266 if (n < cc->max) { /* More greed allowed? */
3267 cp = regcppush(cc->parenfloor);
3269 cc->lastloc = locinput;
3271 if (regmatch(cc->scan)) {
3275 REGCP_UNWIND(lastcp);
3276 regcppop(); /* Restore some previous $<digit>s? */
3277 PL_reginput = locinput;
3279 PerlIO_printf(Perl_debug_log,
3280 "%*s failed, try continuation...\n",
3281 REPORT_CODE_OFF+PL_regindent*2, "")
3284 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3285 && !(PL_reg_flags & RF_warned)) {
3286 PL_reg_flags |= RF_warned;
3287 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s limit (%d) exceeded",
3288 "Complex regular subexpression recursion",
3292 /* Failed deeper matches of scan, so see if this one works. */
3293 PL_regcc = cc->oldcc;
3296 if (regmatch(cc->next))
3302 cc->lastloc = lastloc;
3307 next = scan + ARG(scan);
3310 inner = NEXTOPER(NEXTOPER(scan));
3313 inner = NEXTOPER(scan);
3317 if (OP(next) != c1) /* No choice. */
3318 next = inner; /* Avoid recursion. */
3320 I32 lastparen = *PL_reglastparen;
3322 re_unwind_branch_t *uw;
3324 /* Put unwinding data on stack */
3325 unwind1 = SSNEWt(1,re_unwind_branch_t);
3326 uw = SSPTRt(unwind1,re_unwind_branch_t);
3329 uw->type = ((c1 == BRANCH)
3331 : RE_UNWIND_BRANCHJ);
3332 uw->lastparen = lastparen;
3334 uw->locinput = locinput;
3335 uw->nextchr = nextchr;
3337 uw->regindent = ++PL_regindent;
3340 REGCP_SET(uw->lastcp);
3342 /* Now go into the first branch */
3355 /* We suppose that the next guy does not need
3356 backtracking: in particular, it is of constant length,
3357 and has no parenths to influence future backrefs. */
3358 ln = ARG1(scan); /* min to match */
3359 n = ARG2(scan); /* max to match */
3360 paren = scan->flags;
3362 if (paren > PL_regsize)
3364 if (paren > (I32)*PL_reglastparen)
3365 *PL_reglastparen = paren;
3367 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3369 scan += NEXT_OFF(scan); /* Skip former OPEN. */
3370 PL_reginput = locinput;
3373 if (ln && regrepeat_hard(scan, ln, &l) < ln)
3375 /* if we matched something zero-length we don't need to
3376 backtrack - capturing parens are already defined, so
3377 the caveat in the maximal case doesn't apply
3379 XXXX if ln == 0, we can redo this check first time
3380 through the following loop
3383 n = ln; /* don't backtrack */
3384 locinput = PL_reginput;
3385 if (HAS_TEXT(next) || JUMPABLE(next)) {
3386 regnode *text_node = next;
3388 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3390 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3392 if (PL_regkind[(U8)OP(text_node)] == REF) {
3394 n = ARG(text_node); /* which paren pair */
3395 ln = PL_regstartp[n];
3396 /* assume yes if we haven't seen CLOSEn */
3398 (I32)*PL_reglastparen < n ||
3405 c1 = *(PL_bostr + ln);
3407 else { c1 = (U8)*STRING(text_node); }
3408 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3410 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3411 c2 = PL_fold_locale[c1];
3420 /* This may be improved if l == 0. */
3421 while (n >= ln || (n == REG_INFTY && ln > 0 && l)) { /* ln overflow ? */
3422 /* If it could work, try it. */
3424 UCHARAT(PL_reginput) == c1 ||
3425 UCHARAT(PL_reginput) == c2)
3429 PL_regstartp[paren] =
3430 HOPc(PL_reginput, -l) - PL_bostr;
3431 PL_regendp[paren] = PL_reginput - PL_bostr;
3434 PL_regendp[paren] = -1;
3438 REGCP_UNWIND(lastcp);
3440 /* Couldn't or didn't -- move forward. */
3441 PL_reginput = locinput;
3442 if (regrepeat_hard(scan, 1, &l)) {
3444 locinput = PL_reginput;
3451 n = regrepeat_hard(scan, n, &l);
3452 /* if we matched something zero-length we don't need to
3453 backtrack, unless the minimum count is zero and we
3454 are capturing the result - in that case the capture
3455 being defined or not may affect later execution
3457 if (n != 0 && l == 0 && !(paren && ln == 0))
3458 ln = n; /* don't backtrack */
3459 locinput = PL_reginput;
3461 PerlIO_printf(Perl_debug_log,
3462 "%*s matched %"IVdf" times, len=%"IVdf"...\n",
3463 (int)(REPORT_CODE_OFF+PL_regindent*2), "",
3467 if (HAS_TEXT(next) || JUMPABLE(next)) {
3468 regnode *text_node = next;
3470 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3472 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3474 if (PL_regkind[(U8)OP(text_node)] == REF) {
3476 n = ARG(text_node); /* which paren pair */
3477 ln = PL_regstartp[n];
3478 /* assume yes if we haven't seen CLOSEn */
3480 (I32)*PL_reglastparen < n ||
3487 c1 = *(PL_bostr + ln);
3489 else { c1 = (U8)*STRING(text_node); }
3491 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3493 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3494 c2 = PL_fold_locale[c1];
3505 /* If it could work, try it. */
3507 UCHARAT(PL_reginput) == c1 ||
3508 UCHARAT(PL_reginput) == c2)
3511 PerlIO_printf(Perl_debug_log,
3512 "%*s trying tail with n=%"IVdf"...\n",
3513 (int)(REPORT_CODE_OFF+PL_regindent*2), "", (IV)n)
3517 PL_regstartp[paren] = HOPc(PL_reginput, -l) - PL_bostr;
3518 PL_regendp[paren] = PL_reginput - PL_bostr;
3521 PL_regendp[paren] = -1;
3525 REGCP_UNWIND(lastcp);
3527 /* Couldn't or didn't -- back up. */
3529 locinput = HOPc(locinput, -l);
3530 PL_reginput = locinput;
3537 paren = scan->flags; /* Which paren to set */
3538 if (paren > PL_regsize)
3540 if (paren > (I32)*PL_reglastparen)
3541 *PL_reglastparen = paren;
3542 ln = ARG1(scan); /* min to match */
3543 n = ARG2(scan); /* max to match */
3544 scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE);
3548 ln = ARG1(scan); /* min to match */
3549 n = ARG2(scan); /* max to match */
3550 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3555 scan = NEXTOPER(scan);
3561 scan = NEXTOPER(scan);
3565 * Lookahead to avoid useless match attempts
3566 * when we know what character comes next.
3570 * Used to only do .*x and .*?x, but now it allows
3571 * for )'s, ('s and (?{ ... })'s to be in the way
3572 * of the quantifier and the EXACT-like node. -- japhy
3575 if (HAS_TEXT(next) || JUMPABLE(next)) {
3577 regnode *text_node = next;
3579 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3581 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3583 if (PL_regkind[(U8)OP(text_node)] == REF) {
3585 n = ARG(text_node); /* which paren pair */
3586 ln = PL_regstartp[n];
3587 /* assume yes if we haven't seen CLOSEn */
3589 (I32)*PL_reglastparen < n ||
3594 goto assume_ok_easy;
3596 s = (U8*)PL_bostr + ln;
3598 else { s = (U8*)STRING(text_node); }
3602 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3604 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3605 c2 = PL_fold_locale[c1];
3608 if (OP(text_node) == EXACTF || OP(text_node) == REFF) {
3609 STRLEN ulen1, ulen2;
3610 U8 tmpbuf1[UTF8_MAXLEN_UCLC+1];
3611 U8 tmpbuf2[UTF8_MAXLEN_UCLC+1];
3613 to_utf8_lower((U8*)s, tmpbuf1, &ulen1);
3614 to_utf8_upper((U8*)s, tmpbuf2, &ulen2);
3616 c1 = utf8n_to_uvuni(tmpbuf1, UTF8_MAXLEN, 0,
3618 0 : UTF8_ALLOW_ANY);
3619 c2 = utf8n_to_uvuni(tmpbuf2, UTF8_MAXLEN, 0,
3621 0 : UTF8_ALLOW_ANY);
3624 c2 = c1 = utf8n_to_uvchr(s, UTF8_MAXLEN, 0,
3626 0 : UTF8_ALLOW_ANY);
3634 PL_reginput = locinput;
3638 if (ln && regrepeat(scan, ln) < ln)
3640 locinput = PL_reginput;
3643 char *e; /* Should not check after this */
3644 char *old = locinput;
3647 if (n == REG_INFTY) {
3650 while (UTF8_IS_CONTINUATION(*(U8*)e))
3656 m >0 && e + UTF8SKIP(e) <= PL_regeol; m--)
3660 e = locinput + n - ln;
3665 /* Find place 'next' could work */
3668 while (locinput <= e &&
3669 UCHARAT(locinput) != c1)
3672 while (locinput <= e
3673 && UCHARAT(locinput) != c1
3674 && UCHARAT(locinput) != c2)
3677 count = locinput - old;
3682 /* count initialised to
3683 * utf8_distance(old, locinput) */
3684 while (locinput <= e &&
3685 utf8n_to_uvchr((U8*)locinput,
3688 0 : UTF8_ALLOW_ANY) != (UV)c1) {
3693 /* count initialised to
3694 * utf8_distance(old, locinput) */
3695 while (locinput <= e) {
3696 UV c = utf8n_to_uvchr((U8*)locinput,
3699 0 : UTF8_ALLOW_ANY);
3700 if (c == (UV)c1 || c == (UV)c2)
3709 /* PL_reginput == old now */
3710 if (locinput != old) {
3711 ln = 1; /* Did some */
3712 if (regrepeat(scan, count) < count)
3715 /* PL_reginput == locinput now */
3716 TRYPAREN(paren, ln, locinput);
3717 PL_reginput = locinput; /* Could be reset... */
3718 REGCP_UNWIND(lastcp);
3719 /* Couldn't or didn't -- move forward. */
3722 locinput += UTF8SKIP(locinput);
3729 while (n >= ln || (n == REG_INFTY && ln > 0)) { /* ln overflow ? */
3733 c = utf8n_to_uvchr((U8*)PL_reginput,
3736 0 : UTF8_ALLOW_ANY);
3738 c = UCHARAT(PL_reginput);
3739 /* If it could work, try it. */
3740 if (c == (UV)c1 || c == (UV)c2)
3742 TRYPAREN(paren, ln, PL_reginput);
3743 REGCP_UNWIND(lastcp);
3746 /* If it could work, try it. */
3747 else if (c1 == -1000)
3749 TRYPAREN(paren, ln, PL_reginput);
3750 REGCP_UNWIND(lastcp);
3752 /* Couldn't or didn't -- move forward. */
3753 PL_reginput = locinput;
3754 if (regrepeat(scan, 1)) {
3756 locinput = PL_reginput;
3764 n = regrepeat(scan, n);
3765 locinput = PL_reginput;
3766 if (ln < n && PL_regkind[(U8)OP(next)] == EOL &&
3767 ((!PL_multiline && OP(next) != MEOL) ||
3768 OP(next) == SEOL || OP(next) == EOS))
3770 ln = n; /* why back off? */
3771 /* ...because $ and \Z can match before *and* after
3772 newline at the end. Consider "\n\n" =~ /\n+\Z\n/.
3773 We should back off by one in this case. */
3774 if (UCHARAT(PL_reginput - 1) == '\n' && OP(next) != EOS)
3783 c = utf8n_to_uvchr((U8*)PL_reginput,
3786 0 : UTF8_ALLOW_ANY);
3788 c = UCHARAT(PL_reginput);
3790 /* If it could work, try it. */
3791 if (c1 == -1000 || c == (UV)c1 || c == (UV)c2)
3793 TRYPAREN(paren, n, PL_reginput);
3794 REGCP_UNWIND(lastcp);
3796 /* Couldn't or didn't -- back up. */
3798 PL_reginput = locinput = HOPc(locinput, -1);
3806 c = utf8n_to_uvchr((U8*)PL_reginput,
3809 0 : UTF8_ALLOW_ANY);
3811 c = UCHARAT(PL_reginput);
3813 /* If it could work, try it. */
3814 if (c1 == -1000 || c == (UV)c1 || c == (UV)c2)
3816 TRYPAREN(paren, n, PL_reginput);
3817 REGCP_UNWIND(lastcp);
3819 /* Couldn't or didn't -- back up. */
3821 PL_reginput = locinput = HOPc(locinput, -1);
3828 if (PL_reg_call_cc) {
3829 re_cc_state *cur_call_cc = PL_reg_call_cc;
3830 CURCUR *cctmp = PL_regcc;
3831 regexp *re = PL_reg_re;
3832 CHECKPOINT cp, lastcp;
3834 cp = regcppush(0); /* Save *all* the positions. */
3836 regcp_set_to(PL_reg_call_cc->ss); /* Restore parens of
3838 PL_reginput = locinput; /* Make position available to
3840 cache_re(PL_reg_call_cc->re);
3841 PL_regcc = PL_reg_call_cc->cc;
3842 PL_reg_call_cc = PL_reg_call_cc->prev;
3843 if (regmatch(cur_call_cc->node)) {
3844 PL_reg_call_cc = cur_call_cc;
3848 REGCP_UNWIND(lastcp);
3850 PL_reg_call_cc = cur_call_cc;
3856 PerlIO_printf(Perl_debug_log,
3857 "%*s continuation failed...\n",
3858 REPORT_CODE_OFF+PL_regindent*2, "")
3862 if (locinput < PL_regtill) {
3863 DEBUG_r(PerlIO_printf(Perl_debug_log,
3864 "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n",
3866 (long)(locinput - PL_reg_starttry),
3867 (long)(PL_regtill - PL_reg_starttry),
3869 sayNO_FINAL; /* Cannot match: too short. */
3871 PL_reginput = locinput; /* put where regtry can find it */
3872 sayYES_FINAL; /* Success! */
3874 PL_reginput = locinput; /* put where regtry can find it */
3875 sayYES_LOUD; /* Success! */
3878 PL_reginput = locinput;
3883 s = HOPBACKc(locinput, scan->flags);
3889 PL_reginput = locinput;
3894 s = HOPBACKc(locinput, scan->flags);
3900 PL_reginput = locinput;
3903 inner = NEXTOPER(NEXTOPER(scan));
3904 if (regmatch(inner) != n) {
3919 if (OP(scan) == SUSPEND) {
3920 locinput = PL_reginput;
3921 nextchr = UCHARAT(locinput);
3926 next = scan + ARG(scan);
3931 PerlIO_printf(Perl_error_log, "%"UVxf" %d\n",
3932 PTR2UV(scan), OP(scan));
3933 Perl_croak(aTHX_ "regexp memory corruption");
3940 * We get here only if there's trouble -- normally "case END" is
3941 * the terminating point.
3943 Perl_croak(aTHX_ "corrupted regexp pointers");
3949 PerlIO_printf(Perl_debug_log,
3950 "%*s %scould match...%s\n",
3951 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4],PL_colors[5])
3955 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n",
3956 PL_colors[4],PL_colors[5]));
3962 #if 0 /* Breaks $^R */
3970 PerlIO_printf(Perl_debug_log,
3971 "%*s %sfailed...%s\n",
3972 REPORT_CODE_OFF+PL_regindent*2, "",PL_colors[4],PL_colors[5])
3978 re_unwind_t *uw = SSPTRt(unwind,re_unwind_t);
3981 case RE_UNWIND_BRANCH:
3982 case RE_UNWIND_BRANCHJ:
3984 re_unwind_branch_t *uwb = &(uw->branch);
3985 I32 lastparen = uwb->lastparen;
3987 REGCP_UNWIND(uwb->lastcp);
3988 for (n = *PL_reglastparen; n > lastparen; n--)
3990 *PL_reglastparen = n;
3991 scan = next = uwb->next;
3993 OP(scan) != (uwb->type == RE_UNWIND_BRANCH
3994 ? BRANCH : BRANCHJ) ) { /* Failure */
4001 /* Have more choice yet. Reuse the same uwb. */
4003 if ((n = (uwb->type == RE_UNWIND_BRANCH
4004 ? NEXT_OFF(next) : ARG(next))))
4007 next = NULL; /* XXXX Needn't unwinding in this case... */
4009 next = NEXTOPER(scan);
4010 if (uwb->type == RE_UNWIND_BRANCHJ)
4011 next = NEXTOPER(next);
4012 locinput = uwb->locinput;
4013 nextchr = uwb->nextchr;
4015 PL_regindent = uwb->regindent;
4022 Perl_croak(aTHX_ "regexp unwind memory corruption");
4033 - regrepeat - repeatedly match something simple, report how many
4036 * [This routine now assumes that it will only match on things of length 1.
4037 * That was true before, but now we assume scan - reginput is the count,
4038 * rather than incrementing count on every character. [Er, except utf8.]]
4041 S_regrepeat(pTHX_ regnode *p, I32 max)
4043 register char *scan;
4045 register char *loceol = PL_regeol;
4046 register I32 hardcount = 0;
4047 register bool do_utf8 = PL_reg_match_utf8;
4050 if (max == REG_INFTY)
4052 else if (max < loceol - scan)
4053 loceol = scan + max;
4058 while (scan < loceol && hardcount < max && *scan != '\n') {
4059 scan += UTF8SKIP(scan);
4063 while (scan < loceol && *scan != '\n')
4070 while (scan < loceol && hardcount < max) {
4071 scan += UTF8SKIP(scan);
4081 case EXACT: /* length of string is 1 */
4083 while (scan < loceol && UCHARAT(scan) == c)
4086 case EXACTF: /* length of string is 1 */
4088 while (scan < loceol &&
4089 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold[c]))
4092 case EXACTFL: /* length of string is 1 */
4093 PL_reg_flags |= RF_tainted;
4095 while (scan < loceol &&
4096 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold_locale[c]))
4102 while (hardcount < max && scan < loceol &&
4103 reginclass(p, (U8*)scan, 0, do_utf8)) {
4104 scan += UTF8SKIP(scan);
4108 while (scan < loceol && REGINCLASS(p, (U8*)scan))
4115 LOAD_UTF8_CHARCLASS(alnum,"a");
4116 while (hardcount < max && scan < loceol &&
4117 swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
4118 scan += UTF8SKIP(scan);
4122 while (scan < loceol && isALNUM(*scan))
4127 PL_reg_flags |= RF_tainted;
4130 while (hardcount < max && scan < loceol &&
4131 isALNUM_LC_utf8((U8*)scan)) {
4132 scan += UTF8SKIP(scan);
4136 while (scan < loceol && isALNUM_LC(*scan))
4143 LOAD_UTF8_CHARCLASS(alnum,"a");
4144 while (hardcount < max && scan < loceol &&
4145 !swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
4146 scan += UTF8SKIP(scan);
4150 while (scan < loceol && !isALNUM(*scan))
4155 PL_reg_flags |= RF_tainted;
4158 while (hardcount < max && scan < loceol &&
4159 !isALNUM_LC_utf8((U8*)scan)) {
4160 scan += UTF8SKIP(scan);
4164 while (scan < loceol && !isALNUM_LC(*scan))
4171 LOAD_UTF8_CHARCLASS(space," ");
4172 while (hardcount < max && scan < loceol &&
4174 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
4175 scan += UTF8SKIP(scan);
4179 while (scan < loceol && isSPACE(*scan))
4184 PL_reg_flags |= RF_tainted;
4187 while (hardcount < max && scan < loceol &&
4188 (*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
4189 scan += UTF8SKIP(scan);
4193 while (scan < loceol && isSPACE_LC(*scan))
4200 LOAD_UTF8_CHARCLASS(space," ");
4201 while (hardcount < max && scan < loceol &&
4203 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
4204 scan += UTF8SKIP(scan);
4208 while (scan < loceol && !isSPACE(*scan))
4213 PL_reg_flags |= RF_tainted;
4216 while (hardcount < max && scan < loceol &&
4217 !(*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
4218 scan += UTF8SKIP(scan);
4222 while (scan < loceol && !isSPACE_LC(*scan))
4229 LOAD_UTF8_CHARCLASS(digit,"0");
4230 while (hardcount < max && scan < loceol &&
4231 swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
4232 scan += UTF8SKIP(scan);
4236 while (scan < loceol && isDIGIT(*scan))
4243 LOAD_UTF8_CHARCLASS(digit,"0");
4244 while (hardcount < max && scan < loceol &&
4245 !swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
4246 scan += UTF8SKIP(scan);
4250 while (scan < loceol && !isDIGIT(*scan))
4254 default: /* Called on something of 0 width. */
4255 break; /* So match right here or not at all. */
4261 c = scan - PL_reginput;
4266 SV *prop = sv_newmortal();
4269 PerlIO_printf(Perl_debug_log,
4270 "%*s %s can match %"IVdf" times out of %"IVdf"...\n",
4271 REPORT_CODE_OFF+1, "", SvPVX(prop),(IV)c,(IV)max);
4278 - regrepeat_hard - repeatedly match something, report total lenth and length
4280 * The repeater is supposed to have constant length.
4284 S_regrepeat_hard(pTHX_ regnode *p, I32 max, I32 *lp)
4286 register char *scan = Nullch;
4287 register char *start;
4288 register char *loceol = PL_regeol;
4290 I32 count = 0, res = 1;
4295 start = PL_reginput;
4296 if (PL_reg_match_utf8) {
4297 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
4300 while (start < PL_reginput) {
4302 start += UTF8SKIP(start);
4313 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
4315 *lp = l = PL_reginput - start;
4316 if (max != REG_INFTY && l*max < loceol - scan)
4317 loceol = scan + l*max;
4330 - regclass_swash - prepare the utf8 swash
4334 Perl_regclass_swash(pTHX_ register regnode* node, bool doinit, SV** listsvp, SV **altsvp)
4340 if (PL_regdata && PL_regdata->count) {
4343 if (PL_regdata->what[n] == 's') {
4344 SV *rv = (SV*)PL_regdata->data[n];
4345 AV *av = (AV*)SvRV((SV*)rv);
4346 SV **ary = AvARRAY(av);
4349 /* See the end of regcomp.c:S_reglass() for
4350 * documentation of these array elements. */
4353 a = SvTYPE(ary[1]) == SVt_RV ? &ary[1] : 0;
4354 b = SvTYPE(ary[2]) == SVt_PVAV ? &ary[2] : 0;
4358 else if (si && doinit) {
4359 sw = swash_init("utf8", "", si, 1, 0);
4360 (void)av_store(av, 1, sw);
4376 - reginclass - determine if a character falls into a character class
4378 The n is the ANYOF regnode, the p is the target string, lenp
4379 is pointer to the maximum length of how far to go in the p
4380 (if the lenp is zero, UTF8SKIP(p) is used),
4381 do_utf8 tells whether the target string is in UTF-8.
4386 S_reginclass(pTHX_ register regnode *n, register U8* p, STRLEN* lenp, register bool do_utf8)
4388 char flags = ANYOF_FLAGS(n);
4394 if (do_utf8 && !UTF8_IS_INVARIANT(c))
4395 c = utf8n_to_uvchr(p, UTF8_MAXLEN, &len,
4396 ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
4398 plen = lenp ? *lenp : UNISKIP(NATIVE_TO_UNI(c));
4399 if (do_utf8 || (flags & ANYOF_UNICODE)) {
4402 if (do_utf8 && !ANYOF_RUNTIME(n)) {
4403 if (len != (STRLEN)-1 && c < 256 && ANYOF_BITMAP_TEST(n, c))
4406 if (!match && do_utf8 && (flags & ANYOF_UNICODE_ALL) && c >= 256)
4410 SV *sw = regclass_swash(n, TRUE, 0, (SV**)&av);
4413 if (swash_fetch(sw, p, do_utf8))
4415 else if (flags & ANYOF_FOLD) {
4416 if (!match && lenp && av) {
4419 for (i = 0; i <= av_len(av); i++) {
4420 SV* sv = *av_fetch(av, i, FALSE);
4422 char *s = SvPV(sv, len);
4424 if (len <= plen && memEQ(s, (char*)p, len)) {
4432 U8 tmpbuf[UTF8_MAXLEN_FOLD+1];
4435 to_utf8_fold(p, tmpbuf, &tmplen);
4436 if (swash_fetch(sw, tmpbuf, do_utf8))
4442 if (match && lenp && *lenp == 0)
4443 *lenp = UNISKIP(NATIVE_TO_UNI(c));
4445 if (!match && c < 256) {
4446 if (ANYOF_BITMAP_TEST(n, c))
4448 else if (flags & ANYOF_FOLD) {
4451 if (flags & ANYOF_LOCALE) {
4452 PL_reg_flags |= RF_tainted;
4453 f = PL_fold_locale[c];
4457 if (f != c && ANYOF_BITMAP_TEST(n, f))
4461 if (!match && (flags & ANYOF_CLASS)) {
4462 PL_reg_flags |= RF_tainted;
4464 (ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) ||
4465 (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) ||
4466 (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) ||
4467 (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) ||
4468 (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) ||
4469 (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) ||
4470 (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) ||
4471 (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) ||
4472 (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) ||
4473 (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) ||
4474 (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII(c)) ||
4475 (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII(c)) ||
4476 (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) ||
4477 (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) ||
4478 (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) ||
4479 (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) ||
4480 (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) ||
4481 (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) ||
4482 (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) ||
4483 (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) ||
4484 (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) ||
4485 (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) ||
4486 (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) ||
4487 (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) ||
4488 (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) ||
4489 (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) ||
4490 (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) ||
4491 (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) ||
4492 (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK(c)) ||
4493 (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK(c))
4494 ) /* How's that for a conditional? */
4501 return (flags & ANYOF_INVERT) ? !match : match;
4505 S_reghop(pTHX_ U8 *s, I32 off)
4507 return S_reghop3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4511 S_reghop3(pTHX_ U8 *s, I32 off, U8* lim)
4514 while (off-- && s < lim) {
4515 /* XXX could check well-formedness here */
4523 if (UTF8_IS_CONTINUED(*s)) {
4524 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4527 /* XXX could check well-formedness here */
4535 S_reghopmaybe(pTHX_ U8 *s, I32 off)
4537 return S_reghopmaybe3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4541 S_reghopmaybe3(pTHX_ U8* s, I32 off, U8* lim)
4544 while (off-- && s < lim) {
4545 /* XXX could check well-formedness here */
4555 if (UTF8_IS_CONTINUED(*s)) {
4556 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4559 /* XXX could check well-formedness here */
4571 restore_pos(pTHX_ void *arg)
4573 if (PL_reg_eval_set) {
4574 if (PL_reg_oldsaved) {
4575 PL_reg_re->subbeg = PL_reg_oldsaved;
4576 PL_reg_re->sublen = PL_reg_oldsavedlen;
4577 #ifdef PERL_COPY_ON_WRITE
4578 PL_reg_re->saved_copy = PL_nrs;
4580 RX_MATCH_COPIED_on(PL_reg_re);
4582 PL_reg_magic->mg_len = PL_reg_oldpos;
4583 PL_reg_eval_set = 0;
4584 PL_curpm = PL_reg_oldcurpm;
4589 S_to_utf8_substr(pTHX_ register regexp *prog)
4592 if (prog->float_substr && !prog->float_utf8) {
4593 prog->float_utf8 = sv = NEWSV(117, 0);
4594 SvSetSV(sv, prog->float_substr);
4595 sv_utf8_upgrade(sv);
4596 if (SvTAIL(prog->float_substr))
4598 if (prog->float_substr == prog->check_substr)
4599 prog->check_utf8 = sv;
4601 if (prog->anchored_substr && !prog->anchored_utf8) {
4602 prog->anchored_utf8 = sv = NEWSV(118, 0);
4603 SvSetSV(sv, prog->anchored_substr);
4604 sv_utf8_upgrade(sv);
4605 if (SvTAIL(prog->anchored_substr))
4607 if (prog->anchored_substr == prog->check_substr)
4608 prog->check_utf8 = sv;
4613 S_to_byte_substr(pTHX_ register regexp *prog)
4616 if (prog->float_utf8 && !prog->float_substr) {
4617 prog->float_substr = sv = NEWSV(117, 0);
4618 SvSetSV(sv, prog->float_utf8);
4619 if (sv_utf8_downgrade(sv, TRUE)) {
4620 if (SvTAIL(prog->float_utf8))
4624 prog->float_substr = sv = &PL_sv_undef;
4626 if (prog->float_utf8 == prog->check_utf8)
4627 prog->check_substr = sv;
4629 if (prog->anchored_utf8 && !prog->anchored_substr) {
4630 prog->anchored_substr = sv = NEWSV(118, 0);
4631 SvSetSV(sv, prog->anchored_utf8);
4632 if (sv_utf8_downgrade(sv, TRUE)) {
4633 if (SvTAIL(prog->anchored_utf8))
4637 prog->anchored_substr = sv = &PL_sv_undef;
4639 if (prog->anchored_utf8 == prog->check_utf8)
4640 prog->check_substr = sv;