5 * "One Ring to rule them all, One Ring to find them..."
8 /* NOTE: this is derived from Henry Spencer's regexp code, and should not
9 * confused with the original package (see point 3 below). Thanks, Henry!
12 /* Additional note: this code is very heavily munged from Henry's version
13 * in places. In some spots I've traded clarity for efficiency, so don't
14 * blame Henry for some of the lack of readability.
17 /* The names of the functions have been changed from regcomp and
18 * regexec to pregcomp and pregexec in order to avoid conflicts
19 * with the POSIX routines of the same names.
22 #ifdef PERL_EXT_RE_BUILD
23 /* need to replace pregcomp et al, so enable that */
24 # ifndef PERL_IN_XSUB_RE
25 # define PERL_IN_XSUB_RE
27 /* need access to debugger hooks */
28 # if defined(PERL_EXT_RE_DEBUG) && !defined(DEBUGGING)
33 #ifdef PERL_IN_XSUB_RE
34 /* We *really* need to overwrite these symbols: */
35 # define Perl_regexec_flags my_regexec
36 # define Perl_regdump my_regdump
37 # define Perl_regprop my_regprop
38 # define Perl_re_intuit_start my_re_intuit_start
39 /* *These* symbols are masked to allow static link. */
40 # define Perl_pregexec my_pregexec
41 # define Perl_reginitcolors my_reginitcolors
42 # define Perl_regclass_swash my_regclass_swash
44 # define PERL_NO_GET_CONTEXT
49 * pregcomp and pregexec -- regsub and regerror are not used in perl
51 * Copyright (c) 1986 by University of Toronto.
52 * Written by Henry Spencer. Not derived from licensed software.
54 * Permission is granted to anyone to use this software for any
55 * purpose on any computer system, and to redistribute it freely,
56 * subject to the following restrictions:
58 * 1. The author is not responsible for the consequences of use of
59 * this software, no matter how awful, even if they arise
62 * 2. The origin of this software must not be misrepresented, either
63 * by explicit claim or by omission.
65 * 3. Altered versions must be plainly marked as such, and must not
66 * be misrepresented as being the original software.
68 **** Alterations to Henry's code are...
70 **** Copyright (c) 1991-2002, Larry Wall
72 **** You may distribute under the terms of either the GNU General Public
73 **** License or the Artistic License, as specified in the README file.
75 * Beware that some of this code is subtly aware of the way operator
76 * precedence is structured in regular expressions. Serious changes in
77 * regular-expression syntax might require a total rethink.
80 #define PERL_IN_REGEXEC_C
85 #define RF_tainted 1 /* tainted information used? */
86 #define RF_warned 2 /* warned about big count? */
87 #define RF_evaled 4 /* Did an EVAL with setting? */
88 #define RF_utf8 8 /* String contains multibyte chars? */
90 #define UTF ((PL_reg_flags & RF_utf8) != 0)
92 #define RS_init 1 /* eval environment created */
93 #define RS_set 2 /* replsv value is set */
99 #define REGINCLASS(p,c) (ANYOF_FLAGS(p) ? reginclass(p,c,0,0) : ANYOF_BITMAP_TEST(p,*(c)))
105 #define CHR_SVLEN(sv) (do_utf8 ? sv_len_utf8(sv) : SvCUR(sv))
106 #define CHR_DIST(a,b) (PL_reg_match_utf8 ? utf8_distance(a,b) : a - b)
108 #define reghop_c(pos,off) ((char*)reghop((U8*)pos, off))
109 #define reghopmaybe_c(pos,off) ((char*)reghopmaybe((U8*)pos, off))
110 #define HOP(pos,off) (PL_reg_match_utf8 ? reghop((U8*)pos, off) : (U8*)(pos + off))
111 #define HOPMAYBE(pos,off) (PL_reg_match_utf8 ? reghopmaybe((U8*)pos, off) : (U8*)(pos + off))
112 #define HOPc(pos,off) ((char*)HOP(pos,off))
113 #define HOPMAYBEc(pos,off) ((char*)HOPMAYBE(pos,off))
115 #define HOPBACK(pos, off) ( \
116 (PL_reg_match_utf8) \
117 ? reghopmaybe((U8*)pos, -off) \
118 : (pos - off >= PL_bostr) \
122 #define HOPBACKc(pos, off) (char*)HOPBACK(pos, off)
124 #define reghop3_c(pos,off,lim) ((char*)reghop3((U8*)pos, off, (U8*)lim))
125 #define reghopmaybe3_c(pos,off,lim) ((char*)reghopmaybe3((U8*)pos, off, (U8*)lim))
126 #define HOP3(pos,off,lim) (PL_reg_match_utf8 ? reghop3((U8*)pos, off, (U8*)lim) : (U8*)(pos + off))
127 #define HOPMAYBE3(pos,off,lim) (PL_reg_match_utf8 ? reghopmaybe3((U8*)pos, off, (U8*)lim) : (U8*)(pos + off))
128 #define HOP3c(pos,off,lim) ((char*)HOP3(pos,off,lim))
129 #define HOPMAYBE3c(pos,off,lim) ((char*)HOPMAYBE3(pos,off,lim))
131 #define LOAD_UTF8_CHARCLASS(a,b) STMT_START { if (!CAT2(PL_utf8_,a)) { ENTER; save_re_context(); (void)CAT2(is_utf8_, a)((U8*)b); LEAVE; } } STMT_END
133 /* for use after a quantifier and before an EXACT-like node -- japhy */
134 #define JUMPABLE(rn) ( \
135 OP(rn) == OPEN || OP(rn) == CLOSE || OP(rn) == EVAL || \
136 OP(rn) == SUSPEND || OP(rn) == IFMATCH || \
137 OP(rn) == PLUS || OP(rn) == MINMOD || \
138 (PL_regkind[(U8)OP(rn)] == CURLY && ARG1(rn) > 0) \
141 #define HAS_TEXT(rn) ( \
142 PL_regkind[(U8)OP(rn)] == EXACT || PL_regkind[(U8)OP(rn)] == REF \
146 Search for mandatory following text node; for lookahead, the text must
147 follow but for lookbehind (rn->flags != 0) we skip to the next step.
149 #define FIND_NEXT_IMPT(rn) STMT_START { \
150 while (JUMPABLE(rn)) \
151 if (OP(rn) == SUSPEND || PL_regkind[(U8)OP(rn)] == CURLY) \
152 rn = NEXTOPER(NEXTOPER(rn)); \
153 else if (OP(rn) == PLUS) \
155 else if (OP(rn) == IFMATCH) \
156 rn = (rn->flags == 0) ? NEXTOPER(NEXTOPER(rn)) : rn + ARG(rn); \
157 else rn += NEXT_OFF(rn); \
160 static void restore_pos(pTHX_ void *arg);
163 S_regcppush(pTHX_ I32 parenfloor)
165 int retval = PL_savestack_ix;
166 #define REGCP_PAREN_ELEMS 4
167 int paren_elems_to_push = (PL_regsize - parenfloor) * REGCP_PAREN_ELEMS;
170 if (paren_elems_to_push < 0)
171 Perl_croak(aTHX_ "panic: paren_elems_to_push < 0");
173 #define REGCP_OTHER_ELEMS 6
174 SSCHECK(paren_elems_to_push + REGCP_OTHER_ELEMS);
175 for (p = PL_regsize; p > parenfloor; p--) {
176 /* REGCP_PARENS_ELEMS are pushed per pairs of parentheses. */
177 SSPUSHINT(PL_regendp[p]);
178 SSPUSHINT(PL_regstartp[p]);
179 SSPUSHPTR(PL_reg_start_tmp[p]);
182 /* REGCP_OTHER_ELEMS are pushed in any case, parentheses or no. */
183 SSPUSHINT(PL_regsize);
184 SSPUSHINT(*PL_reglastparen);
185 SSPUSHINT(*PL_reglastcloseparen);
186 SSPUSHPTR(PL_reginput);
187 #define REGCP_FRAME_ELEMS 2
188 /* REGCP_FRAME_ELEMS are part of the REGCP_OTHER_ELEMS and
189 * are needed for the regexp context stack bookkeeping. */
190 SSPUSHINT(paren_elems_to_push + REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS);
191 SSPUSHINT(SAVEt_REGCONTEXT); /* Magic cookie. */
196 /* These are needed since we do not localize EVAL nodes: */
197 # define REGCP_SET(cp) DEBUG_r(PerlIO_printf(Perl_debug_log, \
198 " Setting an EVAL scope, savestack=%"IVdf"\n", \
199 (IV)PL_savestack_ix)); cp = PL_savestack_ix
201 # define REGCP_UNWIND(cp) DEBUG_r(cp != PL_savestack_ix ? \
202 PerlIO_printf(Perl_debug_log, \
203 " Clearing an EVAL scope, savestack=%"IVdf"..%"IVdf"\n", \
204 (IV)(cp), (IV)PL_savestack_ix) : 0); regcpblow(cp)
214 /* Pop REGCP_OTHER_ELEMS before the parentheses loop starts. */
216 assert(i == SAVEt_REGCONTEXT); /* Check that the magic cookie is there. */
217 i = SSPOPINT; /* Parentheses elements to pop. */
218 input = (char *) SSPOPPTR;
219 *PL_reglastcloseparen = SSPOPINT;
220 *PL_reglastparen = SSPOPINT;
221 PL_regsize = SSPOPINT;
223 /* Now restore the parentheses context. */
224 for (i -= (REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS);
225 i > 0; i -= REGCP_PAREN_ELEMS) {
226 paren = (U32)SSPOPINT;
227 PL_reg_start_tmp[paren] = (char *) SSPOPPTR;
228 PL_regstartp[paren] = SSPOPINT;
230 if (paren <= *PL_reglastparen)
231 PL_regendp[paren] = tmps;
233 PerlIO_printf(Perl_debug_log,
234 " restoring \\%"UVuf" to %"IVdf"(%"IVdf")..%"IVdf"%s\n",
235 (UV)paren, (IV)PL_regstartp[paren],
236 (IV)(PL_reg_start_tmp[paren] - PL_bostr),
237 (IV)PL_regendp[paren],
238 (paren > *PL_reglastparen ? "(no)" : ""));
242 if ((I32)(*PL_reglastparen + 1) <= PL_regnpar) {
243 PerlIO_printf(Perl_debug_log,
244 " restoring \\%"IVdf"..\\%"IVdf" to undef\n",
245 (IV)(*PL_reglastparen + 1), (IV)PL_regnpar);
249 /* It would seem that the similar code in regtry()
250 * already takes care of this, and in fact it is in
251 * a better location to since this code can #if 0-ed out
252 * but the code in regtry() is needed or otherwise tests
253 * requiring null fields (pat.t#187 and split.t#{13,14}
254 * (as of patchlevel 7877) will fail. Then again,
255 * this code seems to be necessary or otherwise
256 * building DynaLoader will fail:
257 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
259 for (paren = *PL_reglastparen + 1; (I32)paren <= PL_regnpar; paren++) {
260 if ((I32)paren > PL_regsize)
261 PL_regstartp[paren] = -1;
262 PL_regendp[paren] = -1;
269 S_regcp_set_to(pTHX_ I32 ss)
271 I32 tmp = PL_savestack_ix;
273 PL_savestack_ix = ss;
275 PL_savestack_ix = tmp;
279 typedef struct re_cc_state
283 struct re_cc_state *prev;
288 #define regcpblow(cp) LEAVE_SCOPE(cp) /* Ignores regcppush()ed data. */
290 #define TRYPAREN(paren, n, input) { \
293 PL_regstartp[paren] = HOPc(input, -1) - PL_bostr; \
294 PL_regendp[paren] = input - PL_bostr; \
297 PL_regendp[paren] = -1; \
299 if (regmatch(next)) \
302 PL_regendp[paren] = -1; \
307 * pregexec and friends
311 - pregexec - match a regexp against a string
314 Perl_pregexec(pTHX_ register regexp *prog, char *stringarg, register char *strend,
315 char *strbeg, I32 minend, SV *screamer, U32 nosave)
316 /* strend: pointer to null at end of string */
317 /* strbeg: real beginning of string */
318 /* minend: end of match must be >=minend after stringarg. */
319 /* nosave: For optimizations. */
322 regexec_flags(prog, stringarg, strend, strbeg, minend, screamer, NULL,
323 nosave ? 0 : REXEC_COPY_STR);
327 S_cache_re(pTHX_ regexp *prog)
329 PL_regprecomp = prog->precomp; /* Needed for FAIL. */
331 PL_regprogram = prog->program;
333 PL_regnpar = prog->nparens;
334 PL_regdata = prog->data;
339 * Need to implement the following flags for reg_anch:
341 * USE_INTUIT_NOML - Useful to call re_intuit_start() first
343 * INTUIT_AUTORITATIVE_NOML - Can trust a positive answer
344 * INTUIT_AUTORITATIVE_ML
345 * INTUIT_ONCE_NOML - Intuit can match in one location only.
348 * Another flag for this function: SECOND_TIME (so that float substrs
349 * with giant delta may be not rechecked).
352 /* Assumptions: if ANCH_GPOS, then strpos is anchored. XXXX Check GPOS logic */
354 /* If SCREAM, then SvPVX(sv) should be compatible with strpos and strend.
355 Otherwise, only SvCUR(sv) is used to get strbeg. */
357 /* XXXX We assume that strpos is strbeg unless sv. */
359 /* XXXX Some places assume that there is a fixed substring.
360 An update may be needed if optimizer marks as "INTUITable"
361 RExen without fixed substrings. Similarly, it is assumed that
362 lengths of all the strings are no more than minlen, thus they
363 cannot come from lookahead.
364 (Or minlen should take into account lookahead.) */
366 /* A failure to find a constant substring means that there is no need to make
367 an expensive call to REx engine, thus we celebrate a failure. Similarly,
368 finding a substring too deep into the string means that less calls to
369 regtry() should be needed.
371 REx compiler's optimizer found 4 possible hints:
372 a) Anchored substring;
374 c) Whether we are anchored (beginning-of-line or \G);
375 d) First node (of those at offset 0) which may distingush positions;
376 We use a)b)d) and multiline-part of c), and try to find a position in the
377 string which does not contradict any of them.
380 /* Most of decisions we do here should have been done at compile time.
381 The nodes of the REx which we used for the search should have been
382 deleted from the finite automaton. */
385 Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos,
386 char *strend, U32 flags, re_scream_pos_data *data)
388 register I32 start_shift = 0;
389 /* Should be nonnegative! */
390 register I32 end_shift = 0;
395 int do_utf8 = sv ? SvUTF8(sv) : 0; /* if no sv we have to assume bytes */
397 register char *other_last = Nullch; /* other substr checked before this */
398 char *check_at = Nullch; /* check substr found at this pos */
400 char *i_strpos = strpos;
401 SV *dsv = PERL_DEBUG_PAD_ZERO(0);
404 if (prog->reganch & ROPT_UTF8) {
405 DEBUG_r(PerlIO_printf(Perl_debug_log,
406 "UTF-8 regex...\n"));
407 PL_reg_flags |= RF_utf8;
411 char *s = PL_reg_match_utf8 ?
412 sv_uni_display(dsv, sv, 60, UNI_DISPLAY_REGEX) :
414 int len = PL_reg_match_utf8 ?
415 strlen(s) : strend - strpos;
418 if (PL_reg_match_utf8)
419 DEBUG_r(PerlIO_printf(Perl_debug_log,
420 "UTF-8 target...\n"));
421 PerlIO_printf(Perl_debug_log,
422 "%sGuessing start of match, REx%s `%s%.60s%s%s' against `%s%.*s%s%s'...\n",
423 PL_colors[4],PL_colors[5],PL_colors[0],
426 (strlen(prog->precomp) > 60 ? "..." : ""),
428 (int)(len > 60 ? 60 : len),
430 (len > 60 ? "..." : "")
434 /* CHR_DIST() would be more correct here but it makes things slow. */
435 if (prog->minlen > strend - strpos) {
436 DEBUG_r(PerlIO_printf(Perl_debug_log,
437 "String too short... [re_intuit_start]\n"));
440 strbeg = (sv && SvPOK(sv)) ? strend - SvCUR(sv) : strpos;
443 if (!prog->check_utf8 && prog->check_substr)
444 to_utf8_substr(prog);
445 check = prog->check_utf8;
447 if (!prog->check_substr && prog->check_utf8)
448 to_byte_substr(prog);
449 check = prog->check_substr;
451 if (check == &PL_sv_undef) {
452 DEBUG_r(PerlIO_printf(Perl_debug_log,
453 "Non-utf string cannot match utf check string\n"));
456 if (prog->reganch & ROPT_ANCH) { /* Match at beg-of-str or after \n */
457 ml_anch = !( (prog->reganch & ROPT_ANCH_SINGLE)
458 || ( (prog->reganch & ROPT_ANCH_BOL)
459 && !PL_multiline ) ); /* Check after \n? */
462 if ( !(prog->reganch & (ROPT_ANCH_GPOS /* Checked by the caller */
463 | ROPT_IMPLICIT)) /* not a real BOL */
464 /* SvCUR is not set on references: SvRV and SvPVX overlap */
466 && (strpos != strbeg)) {
467 DEBUG_r(PerlIO_printf(Perl_debug_log, "Not at start...\n"));
470 if (prog->check_offset_min == prog->check_offset_max &&
471 !(prog->reganch & ROPT_CANY_SEEN)) {
472 /* Substring at constant offset from beg-of-str... */
475 s = HOP3c(strpos, prog->check_offset_min, strend);
477 slen = SvCUR(check); /* >= 1 */
479 if ( strend - s > slen || strend - s < slen - 1
480 || (strend - s == slen && strend[-1] != '\n')) {
481 DEBUG_r(PerlIO_printf(Perl_debug_log, "String too long...\n"));
484 /* Now should match s[0..slen-2] */
486 if (slen && (*SvPVX(check) != *s
488 && memNE(SvPVX(check), s, slen)))) {
490 DEBUG_r(PerlIO_printf(Perl_debug_log, "String not equal...\n"));
494 else if (*SvPVX(check) != *s
495 || ((slen = SvCUR(check)) > 1
496 && memNE(SvPVX(check), s, slen)))
498 goto success_at_start;
501 /* Match is anchored, but substr is not anchored wrt beg-of-str. */
503 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
504 end_shift = prog->minlen - start_shift -
505 CHR_SVLEN(check) + (SvTAIL(check) != 0);
507 I32 end = prog->check_offset_max + CHR_SVLEN(check)
508 - (SvTAIL(check) != 0);
509 I32 eshift = CHR_DIST((U8*)strend, (U8*)s) - end;
511 if (end_shift < eshift)
515 else { /* Can match at random position */
518 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
519 /* Should be nonnegative! */
520 end_shift = prog->minlen - start_shift -
521 CHR_SVLEN(check) + (SvTAIL(check) != 0);
524 #ifdef DEBUGGING /* 7/99: reports of failure (with the older version) */
526 Perl_croak(aTHX_ "panic: end_shift");
530 /* Find a possible match in the region s..strend by looking for
531 the "check" substring in the region corrected by start/end_shift. */
532 if (flags & REXEC_SCREAM) {
533 I32 p = -1; /* Internal iterator of scream. */
534 I32 *pp = data ? data->scream_pos : &p;
536 if (PL_screamfirst[BmRARE(check)] >= 0
537 || ( BmRARE(check) == '\n'
538 && (BmPREVIOUS(check) == SvCUR(check) - 1)
540 s = screaminstr(sv, check,
541 start_shift + (s - strbeg), end_shift, pp, 0);
545 *data->scream_olds = s;
547 else if (prog->reganch & ROPT_CANY_SEEN)
548 s = fbm_instr((U8*)(s + start_shift),
549 (U8*)(strend - end_shift),
550 check, PL_multiline ? FBMrf_MULTILINE : 0);
552 s = fbm_instr(HOP3(s, start_shift, strend),
553 HOP3(strend, -end_shift, strbeg),
554 check, PL_multiline ? FBMrf_MULTILINE : 0);
556 /* Update the count-of-usability, remove useless subpatterns,
559 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s %s substr `%s%.*s%s'%s%s",
560 (s ? "Found" : "Did not find"),
561 (check == (do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) ? "anchored" : "floating"),
563 (int)(SvCUR(check) - (SvTAIL(check)!=0)),
565 PL_colors[1], (SvTAIL(check) ? "$" : ""),
566 (s ? " at offset " : "...\n") ) );
573 /* Finish the diagnostic message */
574 DEBUG_r(PerlIO_printf(Perl_debug_log, "%ld...\n", (long)(s - i_strpos)) );
576 /* Got a candidate. Check MBOL anchoring, and the *other* substr.
577 Start with the other substr.
578 XXXX no SCREAM optimization yet - and a very coarse implementation
579 XXXX /ttx+/ results in anchored=`ttx', floating=`x'. floating will
580 *always* match. Probably should be marked during compile...
581 Probably it is right to do no SCREAM here...
584 if (do_utf8 ? (prog->float_utf8 && prog->anchored_utf8) : (prog->float_substr && prog->anchored_substr)) {
585 /* Take into account the "other" substring. */
586 /* XXXX May be hopelessly wrong for UTF... */
589 if (check == (do_utf8 ? prog->float_utf8 : prog->float_substr)) {
592 char *last = HOP3c(s, -start_shift, strbeg), *last1, *last2;
596 t = s - prog->check_offset_max;
597 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
599 || ((t = reghopmaybe3_c(s, -(prog->check_offset_max), strpos))
604 t = HOP3c(t, prog->anchored_offset, strend);
605 if (t < other_last) /* These positions already checked */
607 last2 = last1 = HOP3c(strend, -prog->minlen, strbeg);
610 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
611 /* On end-of-str: see comment below. */
612 must = do_utf8 ? prog->anchored_utf8 : prog->anchored_substr;
613 if (must == &PL_sv_undef) {
615 DEBUG_r(must = prog->anchored_utf8); /* for debug */
620 HOP3(HOP3(last1, prog->anchored_offset, strend)
621 + SvCUR(must), -(SvTAIL(must)!=0), strbeg),
623 PL_multiline ? FBMrf_MULTILINE : 0
625 DEBUG_r(PerlIO_printf(Perl_debug_log,
626 "%s anchored substr `%s%.*s%s'%s",
627 (s ? "Found" : "Contradicts"),
630 - (SvTAIL(must)!=0)),
632 PL_colors[1], (SvTAIL(must) ? "$" : "")));
634 if (last1 >= last2) {
635 DEBUG_r(PerlIO_printf(Perl_debug_log,
636 ", giving up...\n"));
639 DEBUG_r(PerlIO_printf(Perl_debug_log,
640 ", trying floating at offset %ld...\n",
641 (long)(HOP3c(s1, 1, strend) - i_strpos)));
642 other_last = HOP3c(last1, prog->anchored_offset+1, strend);
643 s = HOP3c(last, 1, strend);
647 DEBUG_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
648 (long)(s - i_strpos)));
649 t = HOP3c(s, -prog->anchored_offset, strbeg);
650 other_last = HOP3c(s, 1, strend);
658 else { /* Take into account the floating substring. */
663 t = HOP3c(s, -start_shift, strbeg);
665 HOP3c(strend, -prog->minlen + prog->float_min_offset, strbeg);
666 if (CHR_DIST((U8*)last, (U8*)t) > prog->float_max_offset)
667 last = HOP3c(t, prog->float_max_offset, strend);
668 s = HOP3c(t, prog->float_min_offset, strend);
671 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
672 must = do_utf8 ? prog->float_utf8 : prog->float_substr;
673 /* fbm_instr() takes into account exact value of end-of-str
674 if the check is SvTAIL(ed). Since false positives are OK,
675 and end-of-str is not later than strend we are OK. */
676 if (must == &PL_sv_undef) {
678 DEBUG_r(must = prog->float_utf8); /* for debug message */
681 s = fbm_instr((unsigned char*)s,
682 (unsigned char*)last + SvCUR(must)
684 must, PL_multiline ? FBMrf_MULTILINE : 0);
685 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s floating substr `%s%.*s%s'%s",
686 (s ? "Found" : "Contradicts"),
688 (int)(SvCUR(must) - (SvTAIL(must)!=0)),
690 PL_colors[1], (SvTAIL(must) ? "$" : "")));
693 DEBUG_r(PerlIO_printf(Perl_debug_log,
694 ", giving up...\n"));
697 DEBUG_r(PerlIO_printf(Perl_debug_log,
698 ", trying anchored starting at offset %ld...\n",
699 (long)(s1 + 1 - i_strpos)));
701 s = HOP3c(t, 1, strend);
705 DEBUG_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
706 (long)(s - i_strpos)));
707 other_last = s; /* Fix this later. --Hugo */
716 t = s - prog->check_offset_max;
717 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
719 || ((t = reghopmaybe3_c(s, -prog->check_offset_max, strpos))
721 /* Fixed substring is found far enough so that the match
722 cannot start at strpos. */
724 if (ml_anch && t[-1] != '\n') {
725 /* Eventually fbm_*() should handle this, but often
726 anchored_offset is not 0, so this check will not be wasted. */
727 /* XXXX In the code below we prefer to look for "^" even in
728 presence of anchored substrings. And we search even
729 beyond the found float position. These pessimizations
730 are historical artefacts only. */
732 while (t < strend - prog->minlen) {
734 if (t < check_at - prog->check_offset_min) {
735 if (do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) {
736 /* Since we moved from the found position,
737 we definitely contradict the found anchored
738 substr. Due to the above check we do not
739 contradict "check" substr.
740 Thus we can arrive here only if check substr
741 is float. Redo checking for "other"=="fixed".
744 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld, rescanning for anchored from offset %ld...\n",
745 PL_colors[0],PL_colors[1], (long)(strpos - i_strpos), (long)(strpos - i_strpos + prog->anchored_offset)));
746 goto do_other_anchored;
748 /* We don't contradict the found floating substring. */
749 /* XXXX Why not check for STCLASS? */
751 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld...\n",
752 PL_colors[0],PL_colors[1], (long)(s - i_strpos)));
755 /* Position contradicts check-string */
756 /* XXXX probably better to look for check-string
757 than for "\n", so one should lower the limit for t? */
758 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m, restarting lookup for check-string at offset %ld...\n",
759 PL_colors[0],PL_colors[1], (long)(t + 1 - i_strpos)));
760 other_last = strpos = s = t + 1;
765 DEBUG_r(PerlIO_printf(Perl_debug_log, "Did not find /%s^%s/m...\n",
766 PL_colors[0],PL_colors[1]));
770 DEBUG_r(PerlIO_printf(Perl_debug_log, "Starting position does not contradict /%s^%s/m...\n",
771 PL_colors[0],PL_colors[1]));
775 ++BmUSEFUL(do_utf8 ? prog->check_utf8 : prog->check_substr); /* hooray/5 */
778 /* The found string does not prohibit matching at strpos,
779 - no optimization of calling REx engine can be performed,
780 unless it was an MBOL and we are not after MBOL,
781 or a future STCLASS check will fail this. */
783 /* Even in this situation we may use MBOL flag if strpos is offset
784 wrt the start of the string. */
785 if (ml_anch && sv && !SvROK(sv) /* See prev comment on SvROK */
786 && (strpos != strbeg) && strpos[-1] != '\n'
787 /* May be due to an implicit anchor of m{.*foo} */
788 && !(prog->reganch & ROPT_IMPLICIT))
793 DEBUG_r( if (ml_anch)
794 PerlIO_printf(Perl_debug_log, "Position at offset %ld does not contradict /%s^%s/m...\n",
795 (long)(strpos - i_strpos), PL_colors[0],PL_colors[1]);
798 if (!(prog->reganch & ROPT_NAUGHTY) /* XXXX If strpos moved? */
800 prog->check_utf8 /* Could be deleted already */
801 && --BmUSEFUL(prog->check_utf8) < 0
802 && (prog->check_utf8 == prog->float_utf8)
804 prog->check_substr /* Could be deleted already */
805 && --BmUSEFUL(prog->check_substr) < 0
806 && (prog->check_substr == prog->float_substr)
809 /* If flags & SOMETHING - do not do it many times on the same match */
810 DEBUG_r(PerlIO_printf(Perl_debug_log, "... Disabling check substring...\n"));
811 SvREFCNT_dec(do_utf8 ? prog->check_utf8 : prog->check_substr);
812 if (do_utf8 ? prog->check_substr : prog->check_utf8)
813 SvREFCNT_dec(do_utf8 ? prog->check_substr : prog->check_utf8);
814 prog->check_substr = prog->check_utf8 = Nullsv; /* disable */
815 prog->float_substr = prog->float_utf8 = Nullsv; /* clear */
816 check = Nullsv; /* abort */
818 /* XXXX This is a remnant of the old implementation. It
819 looks wasteful, since now INTUIT can use many
821 prog->reganch &= ~RE_USE_INTUIT;
828 /* XXXX BmUSEFUL already changed, maybe multiple change is meaningful... */
829 if (prog->regstclass) {
830 /* minlen == 0 is possible if regstclass is \b or \B,
831 and the fixed substr is ''$.
832 Since minlen is already taken into account, s+1 is before strend;
833 accidentally, minlen >= 1 guaranties no false positives at s + 1
834 even for \b or \B. But (minlen? 1 : 0) below assumes that
835 regstclass does not come from lookahead... */
836 /* If regstclass takes bytelength more than 1: If charlength==1, OK.
837 This leaves EXACTF only, which is dealt with in find_byclass(). */
838 U8* str = (U8*)STRING(prog->regstclass);
839 int cl_l = (PL_regkind[(U8)OP(prog->regstclass)] == EXACT
840 ? CHR_DIST(str+STR_LEN(prog->regstclass), str)
842 char *endpos = (prog->anchored_substr || prog->anchored_utf8 || ml_anch)
843 ? HOP3c(s, (prog->minlen ? cl_l : 0), strend)
844 : (prog->float_substr || prog->float_utf8
845 ? HOP3c(HOP3c(check_at, -start_shift, strbeg),
848 char *startpos = strbeg;
851 if (prog->reganch & ROPT_UTF8) {
852 PL_regdata = prog->data;
855 s = find_byclass(prog, prog->regstclass, s, endpos, startpos, 1);
860 if (endpos == strend) {
861 DEBUG_r( PerlIO_printf(Perl_debug_log,
862 "Could not match STCLASS...\n") );
865 DEBUG_r( PerlIO_printf(Perl_debug_log,
866 "This position contradicts STCLASS...\n") );
867 if ((prog->reganch & ROPT_ANCH) && !ml_anch)
869 /* Contradict one of substrings */
870 if (prog->anchored_substr || prog->anchored_utf8) {
871 if ((do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) == check) {
872 DEBUG_r( what = "anchored" );
874 s = HOP3c(t, 1, strend);
875 if (s + start_shift + end_shift > strend) {
876 /* XXXX Should be taken into account earlier? */
877 DEBUG_r( PerlIO_printf(Perl_debug_log,
878 "Could not match STCLASS...\n") );
883 DEBUG_r( PerlIO_printf(Perl_debug_log,
884 "Looking for %s substr starting at offset %ld...\n",
885 what, (long)(s + start_shift - i_strpos)) );
888 /* Have both, check_string is floating */
889 if (t + start_shift >= check_at) /* Contradicts floating=check */
890 goto retry_floating_check;
891 /* Recheck anchored substring, but not floating... */
895 DEBUG_r( PerlIO_printf(Perl_debug_log,
896 "Looking for anchored substr starting at offset %ld...\n",
897 (long)(other_last - i_strpos)) );
898 goto do_other_anchored;
900 /* Another way we could have checked stclass at the
901 current position only: */
906 DEBUG_r( PerlIO_printf(Perl_debug_log,
907 "Looking for /%s^%s/m starting at offset %ld...\n",
908 PL_colors[0],PL_colors[1], (long)(t - i_strpos)) );
911 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr)) /* Could have been deleted */
913 /* Check is floating subtring. */
914 retry_floating_check:
915 t = check_at - start_shift;
916 DEBUG_r( what = "floating" );
917 goto hop_and_restart;
920 DEBUG_r(PerlIO_printf(Perl_debug_log,
921 "By STCLASS: moving %ld --> %ld\n",
922 (long)(t - i_strpos), (long)(s - i_strpos))
926 DEBUG_r(PerlIO_printf(Perl_debug_log,
927 "Does not contradict STCLASS...\n");
932 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s%s:%s match at offset %ld\n",
933 PL_colors[4], (check ? "Guessed" : "Giving up"),
934 PL_colors[5], (long)(s - i_strpos)) );
937 fail_finish: /* Substring not found */
938 if (prog->check_substr || prog->check_utf8) /* could be removed already */
939 BmUSEFUL(do_utf8 ? prog->check_utf8 : prog->check_substr) += 5; /* hooray */
941 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch rejected by optimizer%s\n",
942 PL_colors[4],PL_colors[5]));
946 /* We know what class REx starts with. Try to find this position... */
948 S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, char *strend, char *startpos, I32 norun)
950 I32 doevery = (prog->reganch & ROPT_SKIP) == 0;
956 register I32 tmp = 1; /* Scratch variable? */
957 register bool do_utf8 = PL_reg_match_utf8;
959 /* We know what class it must start with. */
963 STRLEN skip = do_utf8 ? UTF8SKIP(s) : 1;
966 reginclass(c, (U8*)s, 0, do_utf8) :
967 REGINCLASS(c, (U8*)s) ||
968 (ANYOF_FOLD_SHARP_S(c, s, strend) &&
969 /* The assignment of 2 is intentional:
970 * for the sharp s, the skip is 2. */
971 (skip = SHARP_S_SKIP)
973 if (tmp && (norun || regtry(prog, s)))
985 if (tmp && (norun || regtry(prog, s)))
997 U8 tmpbuf1[UTF8_MAXLEN_UCLC+1];
998 U8 tmpbuf2[UTF8_MAXLEN_UCLC+1];
1000 to_utf8_lower((U8*)m, tmpbuf1, &ulen1);
1001 to_utf8_upper((U8*)m, tmpbuf2, &ulen2);
1003 c1 = utf8n_to_uvchr(tmpbuf1, UTF8_MAXLEN_UCLC,
1004 0, ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
1005 c2 = utf8n_to_uvchr(tmpbuf2, UTF8_MAXLEN_UCLC,
1006 0, ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
1017 c2 = PL_fold_locale[c1];
1019 e = HOP3c(strend, -(I32)ln, s);
1022 e = s; /* Due to minlen logic of intuit() */
1024 /* The idea in the EXACTF* cases is to first find the
1025 * first character of the EXACTF* node and then, if
1026 * necessary, case-insensitively compare the full
1027 * text of the node. The c1 and c2 are the first
1028 * characters (though in Unicode it gets a bit
1029 * more complicated because there are more cases
1030 * than just upper and lower: one needs to use
1031 * the so-called folding case for case-insensitive
1032 * matching (called "loose matching" in Unicode).
1033 * ibcmp_utf8() will do just that. */
1037 U8 tmpbuf [UTF8_MAXLEN+1];
1038 U8 foldbuf[UTF8_MAXLEN_FOLD+1];
1039 STRLEN len, foldlen;
1043 c = utf8n_to_uvchr((U8*)s, UTF8_MAXLEN, &len,
1045 0 : UTF8_ALLOW_ANY);
1048 ibcmp_utf8(s, (char **)0, 0, do_utf8,
1049 m, (char **)0, ln, (bool)UTF))
1050 && (norun || regtry(prog, s)) )
1053 uvchr_to_utf8(tmpbuf, c);
1054 f = to_utf8_fold(tmpbuf, foldbuf, &foldlen);
1056 && (f == c1 || f == c2)
1057 && (ln == foldlen ||
1058 !ibcmp_utf8((char *) foldbuf,
1059 (char **)0, foldlen, do_utf8,
1061 (char **)0, ln, (bool)UTF))
1062 && (norun || regtry(prog, s)) )
1070 c = utf8n_to_uvchr((U8*)s, UTF8_MAXLEN, &len,
1072 0 : UTF8_ALLOW_ANY);
1074 /* Handle some of the three Greek sigmas cases.
1075 * Note that not all the possible combinations
1076 * are handled here: some of them are handled
1077 * by the standard folding rules, and some of
1078 * them (the character class or ANYOF cases)
1079 * are handled during compiletime in
1080 * regexec.c:S_regclass(). */
1081 if (c == (UV)UNICODE_GREEK_CAPITAL_LETTER_SIGMA ||
1082 c == (UV)UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA)
1083 c = (UV)UNICODE_GREEK_SMALL_LETTER_SIGMA;
1085 if ( (c == c1 || c == c2)
1087 ibcmp_utf8(s, (char **)0, 0, do_utf8,
1088 m, (char **)0, ln, (bool)UTF))
1089 && (norun || regtry(prog, s)) )
1092 uvchr_to_utf8(tmpbuf, c);
1093 f = to_utf8_fold(tmpbuf, foldbuf, &foldlen);
1095 && (f == c1 || f == c2)
1096 && (ln == foldlen ||
1097 !ibcmp_utf8((char *) foldbuf,
1098 (char **)0, foldlen, do_utf8,
1100 (char **)0, ln, (bool)UTF))
1101 && (norun || regtry(prog, s)) )
1112 && (ln == 1 || !(OP(c) == EXACTF
1114 : ibcmp_locale(s, m, ln)))
1115 && (norun || regtry(prog, s)) )
1121 if ( (*(U8*)s == c1 || *(U8*)s == c2)
1122 && (ln == 1 || !(OP(c) == EXACTF
1124 : ibcmp_locale(s, m, ln)))
1125 && (norun || regtry(prog, s)) )
1132 PL_reg_flags |= RF_tainted;
1139 U8 *r = reghop3((U8*)s, -1, (U8*)PL_bostr);
1141 tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, 0);
1143 tmp = ((OP(c) == BOUND ?
1144 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1145 LOAD_UTF8_CHARCLASS(alnum,"a");
1146 while (s < strend) {
1147 if (tmp == !(OP(c) == BOUND ?
1148 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1149 isALNUM_LC_utf8((U8*)s)))
1152 if ((norun || regtry(prog, s)))
1159 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1160 tmp = ((OP(c) == BOUND ? isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1161 while (s < strend) {
1163 !(OP(c) == BOUND ? isALNUM(*s) : isALNUM_LC(*s))) {
1165 if ((norun || regtry(prog, s)))
1171 if ((!prog->minlen && tmp) && (norun || regtry(prog, s)))
1175 PL_reg_flags |= RF_tainted;
1182 U8 *r = reghop3((U8*)s, -1, (U8*)PL_bostr);
1184 tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, 0);
1186 tmp = ((OP(c) == NBOUND ?
1187 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1188 LOAD_UTF8_CHARCLASS(alnum,"a");
1189 while (s < strend) {
1190 if (tmp == !(OP(c) == NBOUND ?
1191 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1192 isALNUM_LC_utf8((U8*)s)))
1194 else if ((norun || regtry(prog, s)))
1200 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1201 tmp = ((OP(c) == NBOUND ?
1202 isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1203 while (s < strend) {
1205 !(OP(c) == NBOUND ? isALNUM(*s) : isALNUM_LC(*s)))
1207 else if ((norun || regtry(prog, s)))
1212 if ((!prog->minlen && !tmp) && (norun || regtry(prog, s)))
1217 LOAD_UTF8_CHARCLASS(alnum,"a");
1218 while (s < strend) {
1219 if (swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1220 if (tmp && (norun || regtry(prog, s)))
1231 while (s < strend) {
1233 if (tmp && (norun || regtry(prog, s)))
1245 PL_reg_flags |= RF_tainted;
1247 while (s < strend) {
1248 if (isALNUM_LC_utf8((U8*)s)) {
1249 if (tmp && (norun || regtry(prog, s)))
1260 while (s < strend) {
1261 if (isALNUM_LC(*s)) {
1262 if (tmp && (norun || regtry(prog, s)))
1275 LOAD_UTF8_CHARCLASS(alnum,"a");
1276 while (s < strend) {
1277 if (!swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1278 if (tmp && (norun || regtry(prog, s)))
1289 while (s < strend) {
1291 if (tmp && (norun || regtry(prog, s)))
1303 PL_reg_flags |= RF_tainted;
1305 while (s < strend) {
1306 if (!isALNUM_LC_utf8((U8*)s)) {
1307 if (tmp && (norun || regtry(prog, s)))
1318 while (s < strend) {
1319 if (!isALNUM_LC(*s)) {
1320 if (tmp && (norun || regtry(prog, s)))
1333 LOAD_UTF8_CHARCLASS(space," ");
1334 while (s < strend) {
1335 if (*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8)) {
1336 if (tmp && (norun || regtry(prog, s)))
1347 while (s < strend) {
1349 if (tmp && (norun || regtry(prog, s)))
1361 PL_reg_flags |= RF_tainted;
1363 while (s < strend) {
1364 if (*s == ' ' || isSPACE_LC_utf8((U8*)s)) {
1365 if (tmp && (norun || regtry(prog, s)))
1376 while (s < strend) {
1377 if (isSPACE_LC(*s)) {
1378 if (tmp && (norun || regtry(prog, s)))
1391 LOAD_UTF8_CHARCLASS(space," ");
1392 while (s < strend) {
1393 if (!(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8))) {
1394 if (tmp && (norun || regtry(prog, s)))
1405 while (s < strend) {
1407 if (tmp && (norun || regtry(prog, s)))
1419 PL_reg_flags |= RF_tainted;
1421 while (s < strend) {
1422 if (!(*s == ' ' || isSPACE_LC_utf8((U8*)s))) {
1423 if (tmp && (norun || regtry(prog, s)))
1434 while (s < strend) {
1435 if (!isSPACE_LC(*s)) {
1436 if (tmp && (norun || regtry(prog, s)))
1449 LOAD_UTF8_CHARCLASS(digit,"0");
1450 while (s < strend) {
1451 if (swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1452 if (tmp && (norun || regtry(prog, s)))
1463 while (s < strend) {
1465 if (tmp && (norun || regtry(prog, s)))
1477 PL_reg_flags |= RF_tainted;
1479 while (s < strend) {
1480 if (isDIGIT_LC_utf8((U8*)s)) {
1481 if (tmp && (norun || regtry(prog, s)))
1492 while (s < strend) {
1493 if (isDIGIT_LC(*s)) {
1494 if (tmp && (norun || regtry(prog, s)))
1507 LOAD_UTF8_CHARCLASS(digit,"0");
1508 while (s < strend) {
1509 if (!swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1510 if (tmp && (norun || regtry(prog, s)))
1521 while (s < strend) {
1523 if (tmp && (norun || regtry(prog, s)))
1535 PL_reg_flags |= RF_tainted;
1537 while (s < strend) {
1538 if (!isDIGIT_LC_utf8((U8*)s)) {
1539 if (tmp && (norun || regtry(prog, s)))
1550 while (s < strend) {
1551 if (!isDIGIT_LC(*s)) {
1552 if (tmp && (norun || regtry(prog, s)))
1564 Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c));
1573 - regexec_flags - match a regexp against a string
1576 Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char *strend,
1577 char *strbeg, I32 minend, SV *sv, void *data, U32 flags)
1578 /* strend: pointer to null at end of string */
1579 /* strbeg: real beginning of string */
1580 /* minend: end of match must be >=minend after stringarg. */
1581 /* data: May be used for some additional optimizations. */
1582 /* nosave: For optimizations. */
1585 register regnode *c;
1586 register char *startpos = stringarg;
1587 I32 minlen; /* must match at least this many chars */
1588 I32 dontbother = 0; /* how many characters not to try at end */
1589 /* I32 start_shift = 0; */ /* Offset of the start to find
1590 constant substr. */ /* CC */
1591 I32 end_shift = 0; /* Same for the end. */ /* CC */
1592 I32 scream_pos = -1; /* Internal iterator of scream. */
1594 SV* oreplsv = GvSV(PL_replgv);
1595 bool do_utf8 = DO_UTF8(sv);
1597 SV *dsv0 = PERL_DEBUG_PAD_ZERO(0);
1598 SV *dsv1 = PERL_DEBUG_PAD_ZERO(1);
1605 PL_regnarrate = DEBUG_r_TEST;
1608 /* Be paranoid... */
1609 if (prog == NULL || startpos == NULL) {
1610 Perl_croak(aTHX_ "NULL regexp parameter");
1614 minlen = prog->minlen;
1615 if (strend - startpos < minlen) {
1616 DEBUG_r(PerlIO_printf(Perl_debug_log,
1617 "String too short [regexec_flags]...\n"));
1621 /* Check validity of program. */
1622 if (UCHARAT(prog->program) != REG_MAGIC) {
1623 Perl_croak(aTHX_ "corrupted regexp program");
1627 PL_reg_eval_set = 0;
1630 if (prog->reganch & ROPT_UTF8)
1631 PL_reg_flags |= RF_utf8;
1633 /* Mark beginning of line for ^ and lookbehind. */
1634 PL_regbol = startpos;
1638 /* Mark end of line for $ (and such) */
1641 /* see how far we have to get to not match where we matched before */
1642 PL_regtill = startpos+minend;
1644 /* We start without call_cc context. */
1647 /* If there is a "must appear" string, look for it. */
1650 if (prog->reganch & ROPT_GPOS_SEEN) { /* Need to have PL_reg_ganch */
1653 if (flags & REXEC_IGNOREPOS) /* Means: check only at start */
1654 PL_reg_ganch = startpos;
1655 else if (sv && SvTYPE(sv) >= SVt_PVMG
1657 && (mg = mg_find(sv, PERL_MAGIC_regex_global))
1658 && mg->mg_len >= 0) {
1659 PL_reg_ganch = strbeg + mg->mg_len; /* Defined pos() */
1660 if (prog->reganch & ROPT_ANCH_GPOS) {
1661 if (s > PL_reg_ganch)
1666 else /* pos() not defined */
1667 PL_reg_ganch = strbeg;
1670 if (!(flags & REXEC_CHECKED) && (prog->check_substr != Nullsv || prog->check_utf8 != Nullsv)) {
1671 re_scream_pos_data d;
1673 d.scream_olds = &scream_olds;
1674 d.scream_pos = &scream_pos;
1675 s = re_intuit_start(prog, sv, s, strend, flags, &d);
1677 DEBUG_r(PerlIO_printf(Perl_debug_log, "Not present...\n"));
1678 goto phooey; /* not present */
1684 pv_uni_display(dsv0, (U8*)prog->precomp, prog->prelen, 60,
1685 UNI_DISPLAY_REGEX) :
1687 int len0 = UTF ? SvCUR(dsv0) : prog->prelen;
1688 char *s1 = do_utf8 ? sv_uni_display(dsv1, sv, 60,
1689 UNI_DISPLAY_REGEX) : startpos;
1690 int len1 = do_utf8 ? SvCUR(dsv1) : strend - startpos;
1693 PerlIO_printf(Perl_debug_log,
1694 "%sMatching REx%s `%s%*.*s%s%s' against `%s%.*s%s%s'\n",
1695 PL_colors[4],PL_colors[5],PL_colors[0],
1698 len0 > 60 ? "..." : "",
1700 (int)(len1 > 60 ? 60 : len1),
1702 (len1 > 60 ? "..." : "")
1706 /* Simplest case: anchored match need be tried only once. */
1707 /* [unless only anchor is BOL and multiline is set] */
1708 if (prog->reganch & (ROPT_ANCH & ~ROPT_ANCH_GPOS)) {
1709 if (s == startpos && regtry(prog, startpos))
1711 else if (PL_multiline || (prog->reganch & ROPT_IMPLICIT)
1712 || (prog->reganch & ROPT_ANCH_MBOL)) /* XXXX SBOL? */
1717 dontbother = minlen - 1;
1718 end = HOP3c(strend, -dontbother, strbeg) - 1;
1719 /* for multiline we only have to try after newlines */
1720 if (prog->check_substr || prog->check_utf8) {
1724 if (regtry(prog, s))
1729 if (prog->reganch & RE_USE_INTUIT) {
1730 s = re_intuit_start(prog, sv, s + 1, strend, flags, NULL);
1741 if (*s++ == '\n') { /* don't need PL_utf8skip here */
1742 if (regtry(prog, s))
1749 } else if (prog->reganch & ROPT_ANCH_GPOS) {
1750 if (regtry(prog, PL_reg_ganch))
1755 /* Messy cases: unanchored match. */
1756 if ((prog->anchored_substr || prog->anchored_utf8) && prog->reganch & ROPT_SKIP) {
1757 /* we have /x+whatever/ */
1758 /* it must be a one character string (XXXX Except UTF?) */
1763 if (!(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr))
1764 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1765 ch = SvPVX(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr)[0];
1768 while (s < strend) {
1770 DEBUG_r( did_match = 1 );
1771 if (regtry(prog, s)) goto got_it;
1773 while (s < strend && *s == ch)
1780 while (s < strend) {
1782 DEBUG_r( did_match = 1 );
1783 if (regtry(prog, s)) goto got_it;
1785 while (s < strend && *s == ch)
1791 DEBUG_r(if (!did_match)
1792 PerlIO_printf(Perl_debug_log,
1793 "Did not find anchored character...\n")
1797 else if (prog->anchored_substr != Nullsv
1798 || prog->anchored_utf8 != Nullsv
1799 || ((prog->float_substr != Nullsv || prog->float_utf8 != Nullsv)
1800 && prog->float_max_offset < strend - s)) {
1805 char *last1; /* Last position checked before */
1809 if (prog->anchored_substr || prog->anchored_utf8) {
1810 if (!(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr))
1811 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1812 must = do_utf8 ? prog->anchored_utf8 : prog->anchored_substr;
1813 back_max = back_min = prog->anchored_offset;
1815 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr))
1816 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1817 must = do_utf8 ? prog->float_utf8 : prog->float_substr;
1818 back_max = prog->float_max_offset;
1819 back_min = prog->float_min_offset;
1821 if (must == &PL_sv_undef)
1822 /* could not downgrade utf8 check substring, so must fail */
1825 last = HOP3c(strend, /* Cannot start after this */
1826 -(I32)(CHR_SVLEN(must)
1827 - (SvTAIL(must) != 0) + back_min), strbeg);
1830 last1 = HOPc(s, -1);
1832 last1 = s - 1; /* bogus */
1834 /* XXXX check_substr already used to find `s', can optimize if
1835 check_substr==must. */
1837 dontbother = end_shift;
1838 strend = HOPc(strend, -dontbother);
1839 while ( (s <= last) &&
1840 ((flags & REXEC_SCREAM)
1841 ? (s = screaminstr(sv, must, HOP3c(s, back_min, strend) - strbeg,
1842 end_shift, &scream_pos, 0))
1843 : (s = fbm_instr((unsigned char*)HOP3(s, back_min, strend),
1844 (unsigned char*)strend, must,
1845 PL_multiline ? FBMrf_MULTILINE : 0))) ) {
1846 DEBUG_r( did_match = 1 );
1847 if (HOPc(s, -back_max) > last1) {
1848 last1 = HOPc(s, -back_min);
1849 s = HOPc(s, -back_max);
1852 char *t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1;
1854 last1 = HOPc(s, -back_min);
1858 while (s <= last1) {
1859 if (regtry(prog, s))
1865 while (s <= last1) {
1866 if (regtry(prog, s))
1872 DEBUG_r(if (!did_match)
1873 PerlIO_printf(Perl_debug_log,
1874 "Did not find %s substr `%s%.*s%s'%s...\n",
1875 ((must == prog->anchored_substr || must == prog->anchored_utf8)
1876 ? "anchored" : "floating"),
1878 (int)(SvCUR(must) - (SvTAIL(must)!=0)),
1880 PL_colors[1], (SvTAIL(must) ? "$" : ""))
1884 else if ((c = prog->regstclass)) {
1886 I32 op = (U8)OP(prog->regstclass);
1887 /* don't bother with what can't match */
1888 if (PL_regkind[op] != EXACT && op != CANY)
1889 strend = HOPc(strend, -(minlen - 1));
1892 SV *prop = sv_newmortal();
1900 pv_uni_display(dsv0, (U8*)SvPVX(prop), SvCUR(prop), 60,
1901 UNI_DISPLAY_REGEX) :
1903 len0 = UTF ? SvCUR(dsv0) : SvCUR(prop);
1905 sv_uni_display(dsv1, sv, 60, UNI_DISPLAY_REGEX) : s;
1906 len1 = UTF ? SvCUR(dsv1) : strend - s;
1907 PerlIO_printf(Perl_debug_log,
1908 "Matching stclass `%*.*s' against `%*.*s'\n",
1912 if (find_byclass(prog, c, s, strend, startpos, 0))
1914 DEBUG_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass...\n"));
1918 if (prog->float_substr != Nullsv || prog->float_utf8 != Nullsv) {
1923 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr))
1924 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1925 float_real = do_utf8 ? prog->float_utf8 : prog->float_substr;
1927 if (flags & REXEC_SCREAM) {
1928 last = screaminstr(sv, float_real, s - strbeg,
1929 end_shift, &scream_pos, 1); /* last one */
1931 last = scream_olds; /* Only one occurrence. */
1935 char *little = SvPV(float_real, len);
1937 if (SvTAIL(float_real)) {
1938 if (memEQ(strend - len + 1, little, len - 1))
1939 last = strend - len + 1;
1940 else if (!PL_multiline)
1941 last = memEQ(strend - len, little, len)
1942 ? strend - len : Nullch;
1948 last = rninstr(s, strend, little, little + len);
1950 last = strend; /* matching `$' */
1954 DEBUG_r(PerlIO_printf(Perl_debug_log,
1955 "%sCan't trim the tail, match fails (should not happen)%s\n",
1956 PL_colors[4],PL_colors[5]));
1957 goto phooey; /* Should not happen! */
1959 dontbother = strend - last + prog->float_min_offset;
1961 if (minlen && (dontbother < minlen))
1962 dontbother = minlen - 1;
1963 strend -= dontbother; /* this one's always in bytes! */
1964 /* We don't know much -- general case. */
1967 if (regtry(prog, s))
1976 if (regtry(prog, s))
1978 } while (s++ < strend);
1986 RX_MATCH_TAINTED_set(prog, PL_reg_flags & RF_tainted);
1988 if (PL_reg_eval_set) {
1989 /* Preserve the current value of $^R */
1990 if (oreplsv != GvSV(PL_replgv))
1991 sv_setsv(oreplsv, GvSV(PL_replgv));/* So that when GvSV(replgv) is
1992 restored, the value remains
1994 restore_pos(aTHX_ 0);
1997 /* make sure $`, $&, $', and $digit will work later */
1998 if ( !(flags & REXEC_NOT_FIRST) ) {
1999 if (RX_MATCH_COPIED(prog)) {
2000 Safefree(prog->subbeg);
2001 RX_MATCH_COPIED_off(prog);
2003 if (flags & REXEC_COPY_STR) {
2004 I32 i = PL_regeol - startpos + (stringarg - strbeg);
2006 s = savepvn(strbeg, i);
2009 RX_MATCH_COPIED_on(prog);
2012 prog->subbeg = strbeg;
2013 prog->sublen = PL_regeol - strbeg; /* strend may have been modified */
2020 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n",
2021 PL_colors[4],PL_colors[5]));
2022 if (PL_reg_eval_set)
2023 restore_pos(aTHX_ 0);
2028 - regtry - try match at specific point
2030 STATIC I32 /* 0 failure, 1 success */
2031 S_regtry(pTHX_ regexp *prog, char *startpos)
2039 PL_regindent = 0; /* XXXX Not good when matches are reenterable... */
2041 if ((prog->reganch & ROPT_EVAL_SEEN) && !PL_reg_eval_set) {
2044 PL_reg_eval_set = RS_init;
2046 PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %"IVdf"\n",
2047 (IV)(PL_stack_sp - PL_stack_base));
2049 SAVEI32(cxstack[cxstack_ix].blk_oldsp);
2050 cxstack[cxstack_ix].blk_oldsp = PL_stack_sp - PL_stack_base;
2051 /* Otherwise OP_NEXTSTATE will free whatever on stack now. */
2053 /* Apparently this is not needed, judging by wantarray. */
2054 /* SAVEI8(cxstack[cxstack_ix].blk_gimme);
2055 cxstack[cxstack_ix].blk_gimme = G_SCALAR; */
2058 /* Make $_ available to executed code. */
2059 if (PL_reg_sv != DEFSV) {
2060 /* SAVE_DEFSV does *not* suffice here for USE_5005THREADS */
2065 if (!(SvTYPE(PL_reg_sv) >= SVt_PVMG && SvMAGIC(PL_reg_sv)
2066 && (mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global)))) {
2067 /* prepare for quick setting of pos */
2068 sv_magic(PL_reg_sv, (SV*)0,
2069 PERL_MAGIC_regex_global, Nullch, 0);
2070 mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global);
2074 PL_reg_oldpos = mg->mg_len;
2075 SAVEDESTRUCTOR_X(restore_pos, 0);
2077 if (!PL_reg_curpm) {
2078 Newz(22,PL_reg_curpm, 1, PMOP);
2081 SV* repointer = newSViv(0);
2082 /* so we know which PL_regex_padav element is PL_reg_curpm */
2083 SvFLAGS(repointer) |= SVf_BREAK;
2084 av_push(PL_regex_padav,repointer);
2085 PL_reg_curpm->op_pmoffset = av_len(PL_regex_padav);
2086 PL_regex_pad = AvARRAY(PL_regex_padav);
2090 PM_SETRE(PL_reg_curpm, prog);
2091 PL_reg_oldcurpm = PL_curpm;
2092 PL_curpm = PL_reg_curpm;
2093 if (RX_MATCH_COPIED(prog)) {
2094 /* Here is a serious problem: we cannot rewrite subbeg,
2095 since it may be needed if this match fails. Thus
2096 $` inside (?{}) could fail... */
2097 PL_reg_oldsaved = prog->subbeg;
2098 PL_reg_oldsavedlen = prog->sublen;
2099 RX_MATCH_COPIED_off(prog);
2102 PL_reg_oldsaved = Nullch;
2103 prog->subbeg = PL_bostr;
2104 prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */
2106 prog->startp[0] = startpos - PL_bostr;
2107 PL_reginput = startpos;
2108 PL_regstartp = prog->startp;
2109 PL_regendp = prog->endp;
2110 PL_reglastparen = &prog->lastparen;
2111 PL_reglastcloseparen = &prog->lastcloseparen;
2112 prog->lastparen = 0;
2113 prog->lastcloseparen = 0;
2115 DEBUG_r(PL_reg_starttry = startpos);
2116 if (PL_reg_start_tmpl <= prog->nparens) {
2117 PL_reg_start_tmpl = prog->nparens*3/2 + 3;
2118 if(PL_reg_start_tmp)
2119 Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
2121 New(22,PL_reg_start_tmp, PL_reg_start_tmpl, char*);
2124 /* XXXX What this code is doing here?!!! There should be no need
2125 to do this again and again, PL_reglastparen should take care of
2128 /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code.
2129 * Actually, the code in regcppop() (which Ilya may be meaning by
2130 * PL_reglastparen), is not needed at all by the test suite
2131 * (op/regexp, op/pat, op/split), but that code is needed, oddly
2132 * enough, for building DynaLoader, or otherwise this
2133 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
2134 * will happen. Meanwhile, this code *is* needed for the
2135 * above-mentioned test suite tests to succeed. The common theme
2136 * on those tests seems to be returning null fields from matches.
2141 if (prog->nparens) {
2142 for (i = prog->nparens; i > (I32)*PL_reglastparen; i--) {
2149 if (regmatch(prog->program + 1)) {
2150 prog->endp[0] = PL_reginput - PL_bostr;
2153 REGCP_UNWIND(lastcp);
2157 #define RE_UNWIND_BRANCH 1
2158 #define RE_UNWIND_BRANCHJ 2
2162 typedef struct { /* XX: makes sense to enlarge it... */
2166 } re_unwind_generic_t;
2179 } re_unwind_branch_t;
2181 typedef union re_unwind_t {
2183 re_unwind_generic_t generic;
2184 re_unwind_branch_t branch;
2187 #define sayYES goto yes
2188 #define sayNO goto no
2189 #define sayNO_ANYOF goto no_anyof
2190 #define sayYES_FINAL goto yes_final
2191 #define sayYES_LOUD goto yes_loud
2192 #define sayNO_FINAL goto no_final
2193 #define sayNO_SILENT goto do_no
2194 #define saySAME(x) if (x) goto yes; else goto no
2196 #define REPORT_CODE_OFF 24
2199 - regmatch - main matching routine
2201 * Conceptually the strategy is simple: check to see whether the current
2202 * node matches, call self recursively to see whether the rest matches,
2203 * and then act accordingly. In practice we make some effort to avoid
2204 * recursion, in particular by going through "ordinary" nodes (that don't
2205 * need to know whether the rest of the match failed) by a loop instead of
2208 /* [lwall] I've hoisted the register declarations to the outer block in order to
2209 * maybe save a little bit of pushing and popping on the stack. It also takes
2210 * advantage of machines that use a register save mask on subroutine entry.
2212 STATIC I32 /* 0 failure, 1 success */
2213 S_regmatch(pTHX_ regnode *prog)
2215 register regnode *scan; /* Current node. */
2216 regnode *next; /* Next node. */
2217 regnode *inner; /* Next node in internal branch. */
2218 register I32 nextchr; /* renamed nextchr - nextchar colides with
2219 function of same name */
2220 register I32 n; /* no or next */
2221 register I32 ln = 0; /* len or last */
2222 register char *s = Nullch; /* operand or save */
2223 register char *locinput = PL_reginput;
2224 register I32 c1 = 0, c2 = 0, paren; /* case fold search, parenth */
2225 int minmod = 0, sw = 0, logical = 0;
2228 I32 firstcp = PL_savestack_ix;
2230 register bool do_utf8 = PL_reg_match_utf8;
2232 SV *dsv0 = PERL_DEBUG_PAD_ZERO(0);
2233 SV *dsv1 = PERL_DEBUG_PAD_ZERO(1);
2234 SV *dsv2 = PERL_DEBUG_PAD_ZERO(2);
2241 /* Note that nextchr is a byte even in UTF */
2242 nextchr = UCHARAT(locinput);
2244 while (scan != NULL) {
2247 SV *prop = sv_newmortal();
2248 int docolor = *PL_colors[0];
2249 int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */
2250 int l = (PL_regeol - locinput) > taill ? taill : (PL_regeol - locinput);
2251 /* The part of the string before starttry has one color
2252 (pref0_len chars), between starttry and current
2253 position another one (pref_len - pref0_len chars),
2254 after the current position the third one.
2255 We assume that pref0_len <= pref_len, otherwise we
2256 decrease pref0_len. */
2257 int pref_len = (locinput - PL_bostr) > (5 + taill) - l
2258 ? (5 + taill) - l : locinput - PL_bostr;
2261 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len)))
2263 pref0_len = pref_len - (locinput - PL_reg_starttry);
2264 if (l + pref_len < (5 + taill) && l < PL_regeol - locinput)
2265 l = ( PL_regeol - locinput > (5 + taill) - pref_len
2266 ? (5 + taill) - pref_len : PL_regeol - locinput);
2267 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput + l)))
2271 if (pref0_len > pref_len)
2272 pref0_len = pref_len;
2273 regprop(prop, scan);
2276 do_utf8 && OP(scan) != CANY ?
2277 pv_uni_display(dsv0, (U8*)(locinput - pref_len),
2278 pref0_len, 60, UNI_DISPLAY_REGEX) :
2279 locinput - pref_len;
2280 int len0 = do_utf8 ? strlen(s0) : pref0_len;
2281 char *s1 = do_utf8 && OP(scan) != CANY ?
2282 pv_uni_display(dsv1, (U8*)(locinput - pref_len + pref0_len),
2283 pref_len - pref0_len, 60, UNI_DISPLAY_REGEX) :
2284 locinput - pref_len + pref0_len;
2285 int len1 = do_utf8 ? strlen(s1) : pref_len - pref0_len;
2286 char *s2 = do_utf8 && OP(scan) != CANY ?
2287 pv_uni_display(dsv2, (U8*)locinput,
2288 PL_regeol - locinput, 60, UNI_DISPLAY_REGEX) :
2290 int len2 = do_utf8 ? strlen(s2) : l;
2291 PerlIO_printf(Perl_debug_log,
2292 "%4"IVdf" <%s%.*s%s%s%.*s%s%s%s%.*s%s>%*s|%3"IVdf":%*s%s\n",
2293 (IV)(locinput - PL_bostr),
2300 (docolor ? "" : "> <"),
2304 15 - l - pref_len + 1,
2306 (IV)(scan - PL_regprogram), PL_regindent*2, "",
2311 next = scan + NEXT_OFF(scan);
2317 if (locinput == PL_bostr || (PL_multiline &&
2318 (nextchr || locinput < PL_regeol) && locinput[-1] == '\n') )
2320 /* regtill = regbol; */
2325 if (locinput == PL_bostr ||
2326 ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n'))
2332 if (locinput == PL_bostr)
2336 if (locinput == PL_reg_ganch)
2346 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2351 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2353 if (PL_regeol - locinput > 1)
2357 if (PL_regeol != locinput)
2361 if (!nextchr && locinput >= PL_regeol)
2364 locinput += PL_utf8skip[nextchr];
2365 if (locinput > PL_regeol)
2367 nextchr = UCHARAT(locinput);
2370 nextchr = UCHARAT(++locinput);
2373 if (!nextchr && locinput >= PL_regeol)
2375 nextchr = UCHARAT(++locinput);
2378 if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n')
2381 locinput += PL_utf8skip[nextchr];
2382 if (locinput > PL_regeol)
2384 nextchr = UCHARAT(locinput);
2387 nextchr = UCHARAT(++locinput);
2392 if (do_utf8 != UTF) {
2393 /* The target and the pattern have differing utf8ness. */
2399 /* The target is utf8, the pattern is not utf8. */
2403 if (NATIVE_TO_UNI(*(U8*)s) !=
2404 utf8n_to_uvuni((U8*)l, UTF8_MAXLEN, &ulen,
2406 0 : UTF8_ALLOW_ANY))
2413 /* The target is not utf8, the pattern is utf8. */
2417 if (NATIVE_TO_UNI(*((U8*)l)) !=
2418 utf8n_to_uvuni((U8*)s, UTF8_MAXLEN, &ulen,
2420 0 : UTF8_ALLOW_ANY))
2427 nextchr = UCHARAT(locinput);
2430 /* The target and the pattern have the same utf8ness. */
2431 /* Inline the first character, for speed. */
2432 if (UCHARAT(s) != nextchr)
2434 if (PL_regeol - locinput < ln)
2436 if (ln > 1 && memNE(s, locinput, ln))
2439 nextchr = UCHARAT(locinput);
2442 PL_reg_flags |= RF_tainted;
2448 if (do_utf8 || UTF) {
2449 /* Either target or the pattern are utf8. */
2451 char *e = PL_regeol;
2453 if (ibcmp_utf8(s, 0, ln, (bool)UTF,
2454 l, &e, 0, do_utf8)) {
2455 /* One more case for the sharp s:
2456 * pack("U0U*", 0xDF) =~ /ss/i,
2457 * the 0xC3 0x9F are the UTF-8
2458 * byte sequence for the U+00DF. */
2460 toLOWER(s[0]) == 's' &&
2462 toLOWER(s[1]) == 's' &&
2469 nextchr = UCHARAT(locinput);
2473 /* Neither the target and the pattern are utf8. */
2475 /* Inline the first character, for speed. */
2476 if (UCHARAT(s) != nextchr &&
2477 UCHARAT(s) != ((OP(scan) == EXACTF)
2478 ? PL_fold : PL_fold_locale)[nextchr])
2480 if (PL_regeol - locinput < ln)
2482 if (ln > 1 && (OP(scan) == EXACTF
2483 ? ibcmp(s, locinput, ln)
2484 : ibcmp_locale(s, locinput, ln)))
2487 nextchr = UCHARAT(locinput);
2491 STRLEN inclasslen = PL_regeol - locinput;
2493 if (!reginclass(scan, (U8*)locinput, &inclasslen, do_utf8))
2495 if (locinput >= PL_regeol)
2497 locinput += inclasslen ? inclasslen : UTF8SKIP(locinput);
2498 nextchr = UCHARAT(locinput);
2503 nextchr = UCHARAT(locinput);
2504 if (!REGINCLASS(scan, (U8*)locinput))
2506 if (!nextchr && locinput >= PL_regeol)
2508 nextchr = UCHARAT(++locinput);
2512 /* If we might have the case of the German sharp s
2513 * in a casefolding Unicode character class. */
2515 if (ANYOF_FOLD_SHARP_S(scan, locinput, PL_regeol)) {
2516 locinput += SHARP_S_SKIP;
2517 nextchr = UCHARAT(locinput);
2523 PL_reg_flags |= RF_tainted;
2529 LOAD_UTF8_CHARCLASS(alnum,"a");
2530 if (!(OP(scan) == ALNUM
2531 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2532 : isALNUM_LC_utf8((U8*)locinput)))
2536 locinput += PL_utf8skip[nextchr];
2537 nextchr = UCHARAT(locinput);
2540 if (!(OP(scan) == ALNUM
2541 ? isALNUM(nextchr) : isALNUM_LC(nextchr)))
2543 nextchr = UCHARAT(++locinput);
2546 PL_reg_flags |= RF_tainted;
2549 if (!nextchr && locinput >= PL_regeol)
2552 LOAD_UTF8_CHARCLASS(alnum,"a");
2553 if (OP(scan) == NALNUM
2554 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2555 : isALNUM_LC_utf8((U8*)locinput))
2559 locinput += PL_utf8skip[nextchr];
2560 nextchr = UCHARAT(locinput);
2563 if (OP(scan) == NALNUM
2564 ? isALNUM(nextchr) : isALNUM_LC(nextchr))
2566 nextchr = UCHARAT(++locinput);
2570 PL_reg_flags |= RF_tainted;
2574 /* was last char in word? */
2576 if (locinput == PL_bostr)
2579 U8 *r = reghop3((U8*)locinput, -1, (U8*)PL_bostr);
2581 ln = utf8n_to_uvchr(r, UTF8SKIP(r), 0, 0);
2583 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2584 ln = isALNUM_uni(ln);
2585 LOAD_UTF8_CHARCLASS(alnum,"a");
2586 n = swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8);
2589 ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(ln));
2590 n = isALNUM_LC_utf8((U8*)locinput);
2594 ln = (locinput != PL_bostr) ?
2595 UCHARAT(locinput - 1) : '\n';
2596 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2598 n = isALNUM(nextchr);
2601 ln = isALNUM_LC(ln);
2602 n = isALNUM_LC(nextchr);
2605 if (((!ln) == (!n)) == (OP(scan) == BOUND ||
2606 OP(scan) == BOUNDL))
2610 PL_reg_flags |= RF_tainted;
2616 if (UTF8_IS_CONTINUED(nextchr)) {
2617 LOAD_UTF8_CHARCLASS(space," ");
2618 if (!(OP(scan) == SPACE
2619 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2620 : isSPACE_LC_utf8((U8*)locinput)))
2624 locinput += PL_utf8skip[nextchr];
2625 nextchr = UCHARAT(locinput);
2628 if (!(OP(scan) == SPACE
2629 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2631 nextchr = UCHARAT(++locinput);
2634 if (!(OP(scan) == SPACE
2635 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2637 nextchr = UCHARAT(++locinput);
2641 PL_reg_flags |= RF_tainted;
2644 if (!nextchr && locinput >= PL_regeol)
2647 LOAD_UTF8_CHARCLASS(space," ");
2648 if (OP(scan) == NSPACE
2649 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2650 : isSPACE_LC_utf8((U8*)locinput))
2654 locinput += PL_utf8skip[nextchr];
2655 nextchr = UCHARAT(locinput);
2658 if (OP(scan) == NSPACE
2659 ? isSPACE(nextchr) : isSPACE_LC(nextchr))
2661 nextchr = UCHARAT(++locinput);
2664 PL_reg_flags |= RF_tainted;
2670 LOAD_UTF8_CHARCLASS(digit,"0");
2671 if (!(OP(scan) == DIGIT
2672 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2673 : isDIGIT_LC_utf8((U8*)locinput)))
2677 locinput += PL_utf8skip[nextchr];
2678 nextchr = UCHARAT(locinput);
2681 if (!(OP(scan) == DIGIT
2682 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr)))
2684 nextchr = UCHARAT(++locinput);
2687 PL_reg_flags |= RF_tainted;
2690 if (!nextchr && locinput >= PL_regeol)
2693 LOAD_UTF8_CHARCLASS(digit,"0");
2694 if (OP(scan) == NDIGIT
2695 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2696 : isDIGIT_LC_utf8((U8*)locinput))
2700 locinput += PL_utf8skip[nextchr];
2701 nextchr = UCHARAT(locinput);
2704 if (OP(scan) == NDIGIT
2705 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr))
2707 nextchr = UCHARAT(++locinput);
2710 if (locinput >= PL_regeol)
2713 LOAD_UTF8_CHARCLASS(mark,"~");
2714 if (swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2716 locinput += PL_utf8skip[nextchr];
2717 while (locinput < PL_regeol &&
2718 swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2719 locinput += UTF8SKIP(locinput);
2720 if (locinput > PL_regeol)
2725 nextchr = UCHARAT(locinput);
2728 PL_reg_flags |= RF_tainted;
2732 n = ARG(scan); /* which paren pair */
2733 ln = PL_regstartp[n];
2734 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2735 if ((I32)*PL_reglastparen < n || ln == -1)
2736 sayNO; /* Do not match unless seen CLOSEn. */
2737 if (ln == PL_regendp[n])
2741 if (do_utf8 && OP(scan) != REF) { /* REF can do byte comparison */
2743 char *e = PL_bostr + PL_regendp[n];
2745 * Note that we can't do the "other character" lookup trick as
2746 * in the 8-bit case (no pun intended) because in Unicode we
2747 * have to map both upper and title case to lower case.
2749 if (OP(scan) == REFF) {
2750 STRLEN ulen1, ulen2;
2751 U8 tmpbuf1[UTF8_MAXLEN_UCLC+1];
2752 U8 tmpbuf2[UTF8_MAXLEN_UCLC+1];
2756 toLOWER_utf8((U8*)s, tmpbuf1, &ulen1);
2757 toLOWER_utf8((U8*)l, tmpbuf2, &ulen2);
2758 if (ulen1 != ulen2 || memNE((char *)tmpbuf1, (char *)tmpbuf2, ulen1))
2765 nextchr = UCHARAT(locinput);
2769 /* Inline the first character, for speed. */
2770 if (UCHARAT(s) != nextchr &&
2772 (UCHARAT(s) != ((OP(scan) == REFF
2773 ? PL_fold : PL_fold_locale)[nextchr]))))
2775 ln = PL_regendp[n] - ln;
2776 if (locinput + ln > PL_regeol)
2778 if (ln > 1 && (OP(scan) == REF
2779 ? memNE(s, locinput, ln)
2781 ? ibcmp(s, locinput, ln)
2782 : ibcmp_locale(s, locinput, ln))))
2785 nextchr = UCHARAT(locinput);
2796 OP_4tree *oop = PL_op;
2797 COP *ocurcop = PL_curcop;
2802 PL_op = (OP_4tree*)PL_regdata->data[n];
2803 DEBUG_r( PerlIO_printf(Perl_debug_log, " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) );
2804 PAD_SAVE_LOCAL(old_comppad, (PAD*)PL_regdata->data[n + 2]);
2805 PL_regendp[0] = PL_reg_magic->mg_len = locinput - PL_bostr;
2809 CALLRUNOPS(aTHX); /* Scalar context. */
2812 ret = &PL_sv_undef; /* protect against empty (?{}) blocks. */
2820 PAD_RESTORE_LOCAL(old_comppad);
2821 PL_curcop = ocurcop;
2823 if (logical == 2) { /* Postponed subexpression. */
2825 MAGIC *mg = Null(MAGIC*);
2827 CHECKPOINT cp, lastcp;
2830 if(SvROK(ret) || SvRMAGICAL(ret)) {
2831 SV *sv = SvROK(ret) ? SvRV(ret) : ret;
2834 mg = mg_find(sv, PERL_MAGIC_qr);
2837 re = (regexp *)mg->mg_obj;
2838 (void)ReREFCNT_inc(re);
2842 char *t = SvPV(ret, len);
2844 char *oprecomp = PL_regprecomp;
2845 I32 osize = PL_regsize;
2846 I32 onpar = PL_regnpar;
2849 if (DO_UTF8(ret)) pm.op_pmdynflags |= PMdf_DYN_UTF8;
2850 re = CALLREGCOMP(aTHX_ t, t + len, &pm);
2852 & (SVs_TEMP | SVs_PADTMP | SVf_READONLY)))
2853 sv_magic(ret,(SV*)ReREFCNT_inc(re),
2855 PL_regprecomp = oprecomp;
2860 PerlIO_printf(Perl_debug_log,
2861 "Entering embedded `%s%.60s%s%s'\n",
2865 (strlen(re->precomp) > 60 ? "..." : ""))
2868 state.prev = PL_reg_call_cc;
2869 state.cc = PL_regcc;
2870 state.re = PL_reg_re;
2874 cp = regcppush(0); /* Save *all* the positions. */
2877 state.ss = PL_savestack_ix;
2878 *PL_reglastparen = 0;
2879 *PL_reglastcloseparen = 0;
2880 PL_reg_call_cc = &state;
2881 PL_reginput = locinput;
2882 toggleutf = ((PL_reg_flags & RF_utf8) != 0) ^
2883 ((re->reganch & ROPT_UTF8) != 0);
2884 if (toggleutf) PL_reg_flags ^= RF_utf8;
2886 /* XXXX This is too dramatic a measure... */
2889 if (regmatch(re->program + 1)) {
2890 /* Even though we succeeded, we need to restore
2891 global variables, since we may be wrapped inside
2892 SUSPEND, thus the match may be not finished yet. */
2894 /* XXXX Do this only if SUSPENDed? */
2895 PL_reg_call_cc = state.prev;
2896 PL_regcc = state.cc;
2897 PL_reg_re = state.re;
2898 cache_re(PL_reg_re);
2899 if (toggleutf) PL_reg_flags ^= RF_utf8;
2901 /* XXXX This is too dramatic a measure... */
2904 /* These are needed even if not SUSPEND. */
2910 REGCP_UNWIND(lastcp);
2912 PL_reg_call_cc = state.prev;
2913 PL_regcc = state.cc;
2914 PL_reg_re = state.re;
2915 cache_re(PL_reg_re);
2916 if (toggleutf) PL_reg_flags ^= RF_utf8;
2918 /* XXXX This is too dramatic a measure... */
2928 sv_setsv(save_scalar(PL_replgv), ret);
2932 n = ARG(scan); /* which paren pair */
2933 PL_reg_start_tmp[n] = locinput;
2938 n = ARG(scan); /* which paren pair */
2939 PL_regstartp[n] = PL_reg_start_tmp[n] - PL_bostr;
2940 PL_regendp[n] = locinput - PL_bostr;
2941 if (n > (I32)*PL_reglastparen)
2942 *PL_reglastparen = n;
2943 *PL_reglastcloseparen = n;
2946 n = ARG(scan); /* which paren pair */
2947 sw = ((I32)*PL_reglastparen >= n && PL_regendp[n] != -1);
2950 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2952 next = NEXTOPER(NEXTOPER(scan));
2954 next = scan + ARG(scan);
2955 if (OP(next) == IFTHEN) /* Fake one. */
2956 next = NEXTOPER(NEXTOPER(next));
2960 logical = scan->flags;
2962 /*******************************************************************
2963 PL_regcc contains infoblock about the innermost (...)* loop, and
2964 a pointer to the next outer infoblock.
2966 Here is how Y(A)*Z is processed (if it is compiled into CURLYX/WHILEM):
2968 1) After matching X, regnode for CURLYX is processed;
2970 2) This regnode creates infoblock on the stack, and calls
2971 regmatch() recursively with the starting point at WHILEM node;
2973 3) Each hit of WHILEM node tries to match A and Z (in the order
2974 depending on the current iteration, min/max of {min,max} and
2975 greediness). The information about where are nodes for "A"
2976 and "Z" is read from the infoblock, as is info on how many times "A"
2977 was already matched, and greediness.
2979 4) After A matches, the same WHILEM node is hit again.
2981 5) Each time WHILEM is hit, PL_regcc is the infoblock created by CURLYX
2982 of the same pair. Thus when WHILEM tries to match Z, it temporarily
2983 resets PL_regcc, since this Y(A)*Z can be a part of some other loop:
2984 as in (Y(A)*Z)*. If Z matches, the automaton will hit the WHILEM node
2985 of the external loop.
2987 Currently present infoblocks form a tree with a stem formed by PL_curcc
2988 and whatever it mentions via ->next, and additional attached trees
2989 corresponding to temporarily unset infoblocks as in "5" above.
2991 In the following picture infoblocks for outer loop of
2992 (Y(A)*?Z)*?T are denoted O, for inner I. NULL starting block
2993 is denoted by x. The matched string is YAAZYAZT. Temporarily postponed
2994 infoblocks are drawn below the "reset" infoblock.
2996 In fact in the picture below we do not show failed matches for Z and T
2997 by WHILEM blocks. [We illustrate minimal matches, since for them it is
2998 more obvious *why* one needs to *temporary* unset infoblocks.]
3000 Matched REx position InfoBlocks Comment
3004 Y A)*?Z)*?T x <- O <- I
3005 YA )*?Z)*?T x <- O <- I
3006 YA A)*?Z)*?T x <- O <- I
3007 YAA )*?Z)*?T x <- O <- I
3008 YAA Z)*?T x <- O # Temporary unset I
3011 YAAZ Y(A)*?Z)*?T x <- O
3014 YAAZY (A)*?Z)*?T x <- O
3017 YAAZY A)*?Z)*?T x <- O <- I
3020 YAAZYA )*?Z)*?T x <- O <- I
3023 YAAZYA Z)*?T x <- O # Temporary unset I
3029 YAAZYAZ T x # Temporary unset O
3036 *******************************************************************/
3039 CHECKPOINT cp = PL_savestack_ix;
3040 /* No need to save/restore up to this paren */
3041 I32 parenfloor = scan->flags;
3043 if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */
3045 cc.oldcc = PL_regcc;
3047 /* XXXX Probably it is better to teach regpush to support
3048 parenfloor > PL_regsize... */
3049 if (parenfloor > (I32)*PL_reglastparen)
3050 parenfloor = *PL_reglastparen; /* Pessimization... */
3051 cc.parenfloor = parenfloor;
3053 cc.min = ARG1(scan);
3054 cc.max = ARG2(scan);
3055 cc.scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
3059 PL_reginput = locinput;
3060 n = regmatch(PREVOPER(next)); /* start on the WHILEM */
3062 PL_regcc = cc.oldcc;
3068 * This is really hard to understand, because after we match
3069 * what we're trying to match, we must make sure the rest of
3070 * the REx is going to match for sure, and to do that we have
3071 * to go back UP the parse tree by recursing ever deeper. And
3072 * if it fails, we have to reset our parent's current state
3073 * that we can try again after backing off.
3076 CHECKPOINT cp, lastcp;
3077 CURCUR* cc = PL_regcc;
3078 char *lastloc = cc->lastloc; /* Detection of 0-len. */
3080 n = cc->cur + 1; /* how many we know we matched */
3081 PL_reginput = locinput;
3084 PerlIO_printf(Perl_debug_log,
3085 "%*s %ld out of %ld..%ld cc=%"UVxf"\n",
3086 REPORT_CODE_OFF+PL_regindent*2, "",
3087 (long)n, (long)cc->min,
3088 (long)cc->max, PTR2UV(cc))
3091 /* If degenerate scan matches "", assume scan done. */
3093 if (locinput == cc->lastloc && n >= cc->min) {
3094 PL_regcc = cc->oldcc;
3098 PerlIO_printf(Perl_debug_log,
3099 "%*s empty match detected, try continuation...\n",
3100 REPORT_CODE_OFF+PL_regindent*2, "")
3102 if (regmatch(cc->next))
3110 /* First just match a string of min scans. */
3114 cc->lastloc = locinput;
3115 if (regmatch(cc->scan))
3118 cc->lastloc = lastloc;
3123 /* Check whether we already were at this position.
3124 Postpone detection until we know the match is not
3125 *that* much linear. */
3126 if (!PL_reg_maxiter) {
3127 PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4);
3128 PL_reg_leftiter = PL_reg_maxiter;
3130 if (PL_reg_leftiter-- == 0) {
3131 I32 size = (PL_reg_maxiter + 7)/8;
3132 if (PL_reg_poscache) {
3133 if ((I32)PL_reg_poscache_size < size) {
3134 Renew(PL_reg_poscache, size, char);
3135 PL_reg_poscache_size = size;
3137 Zero(PL_reg_poscache, size, char);
3140 PL_reg_poscache_size = size;
3141 Newz(29, PL_reg_poscache, size, char);
3144 PerlIO_printf(Perl_debug_log,
3145 "%sDetected a super-linear match, switching on caching%s...\n",
3146 PL_colors[4], PL_colors[5])
3149 if (PL_reg_leftiter < 0) {
3150 I32 o = locinput - PL_bostr, b;
3152 o = (scan->flags & 0xf) - 1 + o * (scan->flags>>4);
3155 if (PL_reg_poscache[o] & (1<<b)) {
3157 PerlIO_printf(Perl_debug_log,
3158 "%*s already tried at this position...\n",
3159 REPORT_CODE_OFF+PL_regindent*2, "")
3163 PL_reg_poscache[o] |= (1<<b);
3167 /* Prefer next over scan for minimal matching. */
3170 PL_regcc = cc->oldcc;
3173 cp = regcppush(cc->parenfloor);
3175 if (regmatch(cc->next)) {
3177 sayYES; /* All done. */
3179 REGCP_UNWIND(lastcp);
3185 if (n >= cc->max) { /* Maximum greed exceeded? */
3186 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3187 && !(PL_reg_flags & RF_warned)) {
3188 PL_reg_flags |= RF_warned;
3189 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s limit (%d) exceeded",
3190 "Complex regular subexpression recursion",
3197 PerlIO_printf(Perl_debug_log,
3198 "%*s trying longer...\n",
3199 REPORT_CODE_OFF+PL_regindent*2, "")
3201 /* Try scanning more and see if it helps. */
3202 PL_reginput = locinput;
3204 cc->lastloc = locinput;
3205 cp = regcppush(cc->parenfloor);
3207 if (regmatch(cc->scan)) {
3211 REGCP_UNWIND(lastcp);
3214 cc->lastloc = lastloc;
3218 /* Prefer scan over next for maximal matching. */
3220 if (n < cc->max) { /* More greed allowed? */
3221 cp = regcppush(cc->parenfloor);
3223 cc->lastloc = locinput;
3225 if (regmatch(cc->scan)) {
3229 REGCP_UNWIND(lastcp);
3230 regcppop(); /* Restore some previous $<digit>s? */
3231 PL_reginput = locinput;
3233 PerlIO_printf(Perl_debug_log,
3234 "%*s failed, try continuation...\n",
3235 REPORT_CODE_OFF+PL_regindent*2, "")
3238 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3239 && !(PL_reg_flags & RF_warned)) {
3240 PL_reg_flags |= RF_warned;
3241 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s limit (%d) exceeded",
3242 "Complex regular subexpression recursion",
3246 /* Failed deeper matches of scan, so see if this one works. */
3247 PL_regcc = cc->oldcc;
3250 if (regmatch(cc->next))
3256 cc->lastloc = lastloc;
3261 next = scan + ARG(scan);
3264 inner = NEXTOPER(NEXTOPER(scan));
3267 inner = NEXTOPER(scan);
3271 if (OP(next) != c1) /* No choice. */
3272 next = inner; /* Avoid recursion. */
3274 I32 lastparen = *PL_reglastparen;
3276 re_unwind_branch_t *uw;
3278 /* Put unwinding data on stack */
3279 unwind1 = SSNEWt(1,re_unwind_branch_t);
3280 uw = SSPTRt(unwind1,re_unwind_branch_t);
3283 uw->type = ((c1 == BRANCH)
3285 : RE_UNWIND_BRANCHJ);
3286 uw->lastparen = lastparen;
3288 uw->locinput = locinput;
3289 uw->nextchr = nextchr;
3291 uw->regindent = ++PL_regindent;
3294 REGCP_SET(uw->lastcp);
3296 /* Now go into the first branch */
3309 /* We suppose that the next guy does not need
3310 backtracking: in particular, it is of constant length,
3311 and has no parenths to influence future backrefs. */
3312 ln = ARG1(scan); /* min to match */
3313 n = ARG2(scan); /* max to match */
3314 paren = scan->flags;
3316 if (paren > PL_regsize)
3318 if (paren > (I32)*PL_reglastparen)
3319 *PL_reglastparen = paren;
3321 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3323 scan += NEXT_OFF(scan); /* Skip former OPEN. */
3324 PL_reginput = locinput;
3327 if (ln && regrepeat_hard(scan, ln, &l) < ln)
3329 /* if we matched something zero-length we don't need to
3330 backtrack - capturing parens are already defined, so
3331 the caveat in the maximal case doesn't apply
3333 XXXX if ln == 0, we can redo this check first time
3334 through the following loop
3337 n = ln; /* don't backtrack */
3338 locinput = PL_reginput;
3339 if (HAS_TEXT(next) || JUMPABLE(next)) {
3340 regnode *text_node = next;
3342 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3344 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3346 if (PL_regkind[(U8)OP(text_node)] == REF) {
3348 n = ARG(text_node); /* which paren pair */
3349 ln = PL_regstartp[n];
3350 /* assume yes if we haven't seen CLOSEn */
3352 (I32)*PL_reglastparen < n ||
3359 c1 = *(PL_bostr + ln);
3361 else { c1 = (U8)*STRING(text_node); }
3362 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3364 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3365 c2 = PL_fold_locale[c1];
3374 /* This may be improved if l == 0. */
3375 while (n >= ln || (n == REG_INFTY && ln > 0 && l)) { /* ln overflow ? */
3376 /* If it could work, try it. */
3378 UCHARAT(PL_reginput) == c1 ||
3379 UCHARAT(PL_reginput) == c2)
3383 PL_regstartp[paren] =
3384 HOPc(PL_reginput, -l) - PL_bostr;
3385 PL_regendp[paren] = PL_reginput - PL_bostr;
3388 PL_regendp[paren] = -1;
3392 REGCP_UNWIND(lastcp);
3394 /* Couldn't or didn't -- move forward. */
3395 PL_reginput = locinput;
3396 if (regrepeat_hard(scan, 1, &l)) {
3398 locinput = PL_reginput;
3405 n = regrepeat_hard(scan, n, &l);
3406 /* if we matched something zero-length we don't need to
3407 backtrack, unless the minimum count is zero and we
3408 are capturing the result - in that case the capture
3409 being defined or not may affect later execution
3411 if (n != 0 && l == 0 && !(paren && ln == 0))
3412 ln = n; /* don't backtrack */
3413 locinput = PL_reginput;
3415 PerlIO_printf(Perl_debug_log,
3416 "%*s matched %"IVdf" times, len=%"IVdf"...\n",
3417 (int)(REPORT_CODE_OFF+PL_regindent*2), "",
3421 if (HAS_TEXT(next) || JUMPABLE(next)) {
3422 regnode *text_node = next;
3424 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3426 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3428 if (PL_regkind[(U8)OP(text_node)] == REF) {
3430 n = ARG(text_node); /* which paren pair */
3431 ln = PL_regstartp[n];
3432 /* assume yes if we haven't seen CLOSEn */
3434 (I32)*PL_reglastparen < n ||
3441 c1 = *(PL_bostr + ln);
3443 else { c1 = (U8)*STRING(text_node); }
3445 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3447 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3448 c2 = PL_fold_locale[c1];
3459 /* If it could work, try it. */
3461 UCHARAT(PL_reginput) == c1 ||
3462 UCHARAT(PL_reginput) == c2)
3465 PerlIO_printf(Perl_debug_log,
3466 "%*s trying tail with n=%"IVdf"...\n",
3467 (int)(REPORT_CODE_OFF+PL_regindent*2), "", (IV)n)
3471 PL_regstartp[paren] = HOPc(PL_reginput, -l) - PL_bostr;
3472 PL_regendp[paren] = PL_reginput - PL_bostr;
3475 PL_regendp[paren] = -1;
3479 REGCP_UNWIND(lastcp);
3481 /* Couldn't or didn't -- back up. */
3483 locinput = HOPc(locinput, -l);
3484 PL_reginput = locinput;
3491 paren = scan->flags; /* Which paren to set */
3492 if (paren > PL_regsize)
3494 if (paren > (I32)*PL_reglastparen)
3495 *PL_reglastparen = paren;
3496 ln = ARG1(scan); /* min to match */
3497 n = ARG2(scan); /* max to match */
3498 scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE);
3502 ln = ARG1(scan); /* min to match */
3503 n = ARG2(scan); /* max to match */
3504 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3509 scan = NEXTOPER(scan);
3515 scan = NEXTOPER(scan);
3519 * Lookahead to avoid useless match attempts
3520 * when we know what character comes next.
3524 * Used to only do .*x and .*?x, but now it allows
3525 * for )'s, ('s and (?{ ... })'s to be in the way
3526 * of the quantifier and the EXACT-like node. -- japhy
3529 if (HAS_TEXT(next) || JUMPABLE(next)) {
3531 regnode *text_node = next;
3533 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3535 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3537 if (PL_regkind[(U8)OP(text_node)] == REF) {
3539 n = ARG(text_node); /* which paren pair */
3540 ln = PL_regstartp[n];
3541 /* assume yes if we haven't seen CLOSEn */
3543 (I32)*PL_reglastparen < n ||
3548 goto assume_ok_easy;
3550 s = (U8*)PL_bostr + ln;
3552 else { s = (U8*)STRING(text_node); }
3556 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3558 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3559 c2 = PL_fold_locale[c1];
3562 if (OP(text_node) == EXACTF || OP(text_node) == REFF) {
3563 STRLEN ulen1, ulen2;
3564 U8 tmpbuf1[UTF8_MAXLEN_UCLC+1];
3565 U8 tmpbuf2[UTF8_MAXLEN_UCLC+1];
3567 to_utf8_lower((U8*)s, tmpbuf1, &ulen1);
3568 to_utf8_upper((U8*)s, tmpbuf2, &ulen2);
3570 c1 = utf8n_to_uvuni(tmpbuf1, UTF8_MAXLEN, 0,
3572 0 : UTF8_ALLOW_ANY);
3573 c2 = utf8n_to_uvuni(tmpbuf2, UTF8_MAXLEN, 0,
3575 0 : UTF8_ALLOW_ANY);
3578 c2 = c1 = utf8n_to_uvchr(s, UTF8_MAXLEN, 0,
3580 0 : UTF8_ALLOW_ANY);
3588 PL_reginput = locinput;
3592 if (ln && regrepeat(scan, ln) < ln)
3594 locinput = PL_reginput;
3597 char *e; /* Should not check after this */
3598 char *old = locinput;
3601 if (n == REG_INFTY) {
3604 while (UTF8_IS_CONTINUATION(*(U8*)e))
3610 m >0 && e + UTF8SKIP(e) <= PL_regeol; m--)
3614 e = locinput + n - ln;
3619 /* Find place 'next' could work */
3622 while (locinput <= e &&
3623 UCHARAT(locinput) != c1)
3626 while (locinput <= e
3627 && UCHARAT(locinput) != c1
3628 && UCHARAT(locinput) != c2)
3631 count = locinput - old;
3636 /* count initialised to
3637 * utf8_distance(old, locinput) */
3638 while (locinput <= e &&
3639 utf8n_to_uvchr((U8*)locinput,
3642 0 : UTF8_ALLOW_ANY) != (UV)c1) {
3647 /* count initialised to
3648 * utf8_distance(old, locinput) */
3649 while (locinput <= e) {
3650 UV c = utf8n_to_uvchr((U8*)locinput,
3653 0 : UTF8_ALLOW_ANY);
3654 if (c == (UV)c1 || c == (UV)c2)
3663 /* PL_reginput == old now */
3664 if (locinput != old) {
3665 ln = 1; /* Did some */
3666 if (regrepeat(scan, count) < count)
3669 /* PL_reginput == locinput now */
3670 TRYPAREN(paren, ln, locinput);
3671 PL_reginput = locinput; /* Could be reset... */
3672 REGCP_UNWIND(lastcp);
3673 /* Couldn't or didn't -- move forward. */
3676 locinput += UTF8SKIP(locinput);
3683 while (n >= ln || (n == REG_INFTY && ln > 0)) { /* ln overflow ? */
3687 c = utf8n_to_uvchr((U8*)PL_reginput,
3690 0 : UTF8_ALLOW_ANY);
3692 c = UCHARAT(PL_reginput);
3693 /* If it could work, try it. */
3694 if (c == (UV)c1 || c == (UV)c2)
3696 TRYPAREN(paren, ln, PL_reginput);
3697 REGCP_UNWIND(lastcp);
3700 /* If it could work, try it. */
3701 else if (c1 == -1000)
3703 TRYPAREN(paren, ln, PL_reginput);
3704 REGCP_UNWIND(lastcp);
3706 /* Couldn't or didn't -- move forward. */
3707 PL_reginput = locinput;
3708 if (regrepeat(scan, 1)) {
3710 locinput = PL_reginput;
3718 n = regrepeat(scan, n);
3719 locinput = PL_reginput;
3720 if (ln < n && PL_regkind[(U8)OP(next)] == EOL &&
3721 ((!PL_multiline && OP(next) != MEOL) ||
3722 OP(next) == SEOL || OP(next) == EOS))
3724 ln = n; /* why back off? */
3725 /* ...because $ and \Z can match before *and* after
3726 newline at the end. Consider "\n\n" =~ /\n+\Z\n/.
3727 We should back off by one in this case. */
3728 if (UCHARAT(PL_reginput - 1) == '\n' && OP(next) != EOS)
3737 c = utf8n_to_uvchr((U8*)PL_reginput,
3740 0 : UTF8_ALLOW_ANY);
3742 c = UCHARAT(PL_reginput);
3744 /* If it could work, try it. */
3745 if (c1 == -1000 || c == (UV)c1 || c == (UV)c2)
3747 TRYPAREN(paren, n, PL_reginput);
3748 REGCP_UNWIND(lastcp);
3750 /* Couldn't or didn't -- back up. */
3752 PL_reginput = locinput = HOPc(locinput, -1);
3760 c = utf8n_to_uvchr((U8*)PL_reginput,
3763 0 : UTF8_ALLOW_ANY);
3765 c = UCHARAT(PL_reginput);
3767 /* If it could work, try it. */
3768 if (c1 == -1000 || c == (UV)c1 || c == (UV)c2)
3770 TRYPAREN(paren, n, PL_reginput);
3771 REGCP_UNWIND(lastcp);
3773 /* Couldn't or didn't -- back up. */
3775 PL_reginput = locinput = HOPc(locinput, -1);
3782 if (PL_reg_call_cc) {
3783 re_cc_state *cur_call_cc = PL_reg_call_cc;
3784 CURCUR *cctmp = PL_regcc;
3785 regexp *re = PL_reg_re;
3786 CHECKPOINT cp, lastcp;
3788 cp = regcppush(0); /* Save *all* the positions. */
3790 regcp_set_to(PL_reg_call_cc->ss); /* Restore parens of
3792 PL_reginput = locinput; /* Make position available to
3794 cache_re(PL_reg_call_cc->re);
3795 PL_regcc = PL_reg_call_cc->cc;
3796 PL_reg_call_cc = PL_reg_call_cc->prev;
3797 if (regmatch(cur_call_cc->node)) {
3798 PL_reg_call_cc = cur_call_cc;
3802 REGCP_UNWIND(lastcp);
3804 PL_reg_call_cc = cur_call_cc;
3810 PerlIO_printf(Perl_debug_log,
3811 "%*s continuation failed...\n",
3812 REPORT_CODE_OFF+PL_regindent*2, "")
3816 if (locinput < PL_regtill) {
3817 DEBUG_r(PerlIO_printf(Perl_debug_log,
3818 "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n",
3820 (long)(locinput - PL_reg_starttry),
3821 (long)(PL_regtill - PL_reg_starttry),
3823 sayNO_FINAL; /* Cannot match: too short. */
3825 PL_reginput = locinput; /* put where regtry can find it */
3826 sayYES_FINAL; /* Success! */
3828 PL_reginput = locinput; /* put where regtry can find it */
3829 sayYES_LOUD; /* Success! */
3832 PL_reginput = locinput;
3837 s = HOPBACKc(locinput, scan->flags);
3843 PL_reginput = locinput;
3848 s = HOPBACKc(locinput, scan->flags);
3854 PL_reginput = locinput;
3857 inner = NEXTOPER(NEXTOPER(scan));
3858 if (regmatch(inner) != n) {
3873 if (OP(scan) == SUSPEND) {
3874 locinput = PL_reginput;
3875 nextchr = UCHARAT(locinput);
3880 next = scan + ARG(scan);
3885 PerlIO_printf(Perl_error_log, "%"UVxf" %d\n",
3886 PTR2UV(scan), OP(scan));
3887 Perl_croak(aTHX_ "regexp memory corruption");
3894 * We get here only if there's trouble -- normally "case END" is
3895 * the terminating point.
3897 Perl_croak(aTHX_ "corrupted regexp pointers");
3903 PerlIO_printf(Perl_debug_log,
3904 "%*s %scould match...%s\n",
3905 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4],PL_colors[5])
3909 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n",
3910 PL_colors[4],PL_colors[5]));
3916 #if 0 /* Breaks $^R */
3924 PerlIO_printf(Perl_debug_log,
3925 "%*s %sfailed...%s\n",
3926 REPORT_CODE_OFF+PL_regindent*2, "",PL_colors[4],PL_colors[5])
3932 re_unwind_t *uw = SSPTRt(unwind,re_unwind_t);
3935 case RE_UNWIND_BRANCH:
3936 case RE_UNWIND_BRANCHJ:
3938 re_unwind_branch_t *uwb = &(uw->branch);
3939 I32 lastparen = uwb->lastparen;
3941 REGCP_UNWIND(uwb->lastcp);
3942 for (n = *PL_reglastparen; n > lastparen; n--)
3944 *PL_reglastparen = n;
3945 scan = next = uwb->next;
3947 OP(scan) != (uwb->type == RE_UNWIND_BRANCH
3948 ? BRANCH : BRANCHJ) ) { /* Failure */
3955 /* Have more choice yet. Reuse the same uwb. */
3957 if ((n = (uwb->type == RE_UNWIND_BRANCH
3958 ? NEXT_OFF(next) : ARG(next))))
3961 next = NULL; /* XXXX Needn't unwinding in this case... */
3963 next = NEXTOPER(scan);
3964 if (uwb->type == RE_UNWIND_BRANCHJ)
3965 next = NEXTOPER(next);
3966 locinput = uwb->locinput;
3967 nextchr = uwb->nextchr;
3969 PL_regindent = uwb->regindent;
3976 Perl_croak(aTHX_ "regexp unwind memory corruption");
3987 - regrepeat - repeatedly match something simple, report how many
3990 * [This routine now assumes that it will only match on things of length 1.
3991 * That was true before, but now we assume scan - reginput is the count,
3992 * rather than incrementing count on every character. [Er, except utf8.]]
3995 S_regrepeat(pTHX_ regnode *p, I32 max)
3997 register char *scan;
3999 register char *loceol = PL_regeol;
4000 register I32 hardcount = 0;
4001 register bool do_utf8 = PL_reg_match_utf8;
4004 if (max == REG_INFTY)
4006 else if (max < loceol - scan)
4007 loceol = scan + max;
4012 while (scan < loceol && hardcount < max && *scan != '\n') {
4013 scan += UTF8SKIP(scan);
4017 while (scan < loceol && *scan != '\n')
4024 while (scan < loceol && hardcount < max) {
4025 scan += UTF8SKIP(scan);
4035 case EXACT: /* length of string is 1 */
4037 while (scan < loceol && UCHARAT(scan) == c)
4040 case EXACTF: /* length of string is 1 */
4042 while (scan < loceol &&
4043 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold[c]))
4046 case EXACTFL: /* length of string is 1 */
4047 PL_reg_flags |= RF_tainted;
4049 while (scan < loceol &&
4050 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold_locale[c]))
4056 while (hardcount < max && scan < loceol &&
4057 reginclass(p, (U8*)scan, 0, do_utf8)) {
4058 scan += UTF8SKIP(scan);
4062 while (scan < loceol && REGINCLASS(p, (U8*)scan))
4069 LOAD_UTF8_CHARCLASS(alnum,"a");
4070 while (hardcount < max && scan < loceol &&
4071 swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
4072 scan += UTF8SKIP(scan);
4076 while (scan < loceol && isALNUM(*scan))
4081 PL_reg_flags |= RF_tainted;
4084 while (hardcount < max && scan < loceol &&
4085 isALNUM_LC_utf8((U8*)scan)) {
4086 scan += UTF8SKIP(scan);
4090 while (scan < loceol && isALNUM_LC(*scan))
4097 LOAD_UTF8_CHARCLASS(alnum,"a");
4098 while (hardcount < max && scan < loceol &&
4099 !swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
4100 scan += UTF8SKIP(scan);
4104 while (scan < loceol && !isALNUM(*scan))
4109 PL_reg_flags |= RF_tainted;
4112 while (hardcount < max && scan < loceol &&
4113 !isALNUM_LC_utf8((U8*)scan)) {
4114 scan += UTF8SKIP(scan);
4118 while (scan < loceol && !isALNUM_LC(*scan))
4125 LOAD_UTF8_CHARCLASS(space," ");
4126 while (hardcount < max && scan < loceol &&
4128 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
4129 scan += UTF8SKIP(scan);
4133 while (scan < loceol && isSPACE(*scan))
4138 PL_reg_flags |= RF_tainted;
4141 while (hardcount < max && scan < loceol &&
4142 (*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
4143 scan += UTF8SKIP(scan);
4147 while (scan < loceol && isSPACE_LC(*scan))
4154 LOAD_UTF8_CHARCLASS(space," ");
4155 while (hardcount < max && scan < loceol &&
4157 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
4158 scan += UTF8SKIP(scan);
4162 while (scan < loceol && !isSPACE(*scan))
4167 PL_reg_flags |= RF_tainted;
4170 while (hardcount < max && scan < loceol &&
4171 !(*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
4172 scan += UTF8SKIP(scan);
4176 while (scan < loceol && !isSPACE_LC(*scan))
4183 LOAD_UTF8_CHARCLASS(digit,"0");
4184 while (hardcount < max && scan < loceol &&
4185 swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
4186 scan += UTF8SKIP(scan);
4190 while (scan < loceol && isDIGIT(*scan))
4197 LOAD_UTF8_CHARCLASS(digit,"0");
4198 while (hardcount < max && scan < loceol &&
4199 !swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
4200 scan += UTF8SKIP(scan);
4204 while (scan < loceol && !isDIGIT(*scan))
4208 default: /* Called on something of 0 width. */
4209 break; /* So match right here or not at all. */
4215 c = scan - PL_reginput;
4220 SV *prop = sv_newmortal();
4223 PerlIO_printf(Perl_debug_log,
4224 "%*s %s can match %"IVdf" times out of %"IVdf"...\n",
4225 REPORT_CODE_OFF+1, "", SvPVX(prop),(IV)c,(IV)max);
4232 - regrepeat_hard - repeatedly match something, report total lenth and length
4234 * The repeater is supposed to have constant length.
4238 S_regrepeat_hard(pTHX_ regnode *p, I32 max, I32 *lp)
4240 register char *scan = Nullch;
4241 register char *start;
4242 register char *loceol = PL_regeol;
4244 I32 count = 0, res = 1;
4249 start = PL_reginput;
4250 if (PL_reg_match_utf8) {
4251 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
4254 while (start < PL_reginput) {
4256 start += UTF8SKIP(start);
4267 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
4269 *lp = l = PL_reginput - start;
4270 if (max != REG_INFTY && l*max < loceol - scan)
4271 loceol = scan + l*max;
4284 - regclass_swash - prepare the utf8 swash
4288 Perl_regclass_swash(pTHX_ register regnode* node, bool doinit, SV** listsvp, SV **altsvp)
4294 if (PL_regdata && PL_regdata->count) {
4297 if (PL_regdata->what[n] == 's') {
4298 SV *rv = (SV*)PL_regdata->data[n];
4299 AV *av = (AV*)SvRV((SV*)rv);
4300 SV **ary = AvARRAY(av);
4303 /* See the end of regcomp.c:S_reglass() for
4304 * documentation of these array elements. */
4307 a = SvTYPE(ary[1]) == SVt_RV ? &ary[1] : 0;
4308 b = SvTYPE(ary[2]) == SVt_PVAV ? &ary[2] : 0;
4312 else if (si && doinit) {
4313 sw = swash_init("utf8", "", si, 1, 0);
4314 (void)av_store(av, 1, sw);
4330 - reginclass - determine if a character falls into a character class
4332 The n is the ANYOF regnode, the p is the target string, lenp
4333 is pointer to the maximum length of how far to go in the p
4334 (if the lenp is zero, UTF8SKIP(p) is used),
4335 do_utf8 tells whether the target string is in UTF-8.
4340 S_reginclass(pTHX_ register regnode *n, register U8* p, STRLEN* lenp, register bool do_utf8)
4342 char flags = ANYOF_FLAGS(n);
4348 if (do_utf8 && !UTF8_IS_INVARIANT(c))
4349 c = utf8n_to_uvchr(p, UTF8_MAXLEN, &len,
4350 ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
4352 plen = lenp ? *lenp : UNISKIP(NATIVE_TO_UNI(c));
4353 if (do_utf8 || (flags & ANYOF_UNICODE)) {
4356 if (do_utf8 && !ANYOF_RUNTIME(n)) {
4357 if (len != (STRLEN)-1 && c < 256 && ANYOF_BITMAP_TEST(n, c))
4360 if (!match && do_utf8 && (flags & ANYOF_UNICODE_ALL) && c >= 256)
4364 SV *sw = regclass_swash(n, TRUE, 0, (SV**)&av);
4367 if (swash_fetch(sw, p, do_utf8))
4369 else if (flags & ANYOF_FOLD) {
4370 if (!match && lenp && av) {
4373 for (i = 0; i <= av_len(av); i++) {
4374 SV* sv = *av_fetch(av, i, FALSE);
4376 char *s = SvPV(sv, len);
4378 if (len <= plen && memEQ(s, (char*)p, len)) {
4386 U8 tmpbuf[UTF8_MAXLEN_FOLD+1];
4389 to_utf8_fold(p, tmpbuf, &tmplen);
4390 if (swash_fetch(sw, tmpbuf, do_utf8))
4396 if (match && lenp && *lenp == 0)
4397 *lenp = UNISKIP(NATIVE_TO_UNI(c));
4399 if (!match && c < 256) {
4400 if (ANYOF_BITMAP_TEST(n, c))
4402 else if (flags & ANYOF_FOLD) {
4405 if (flags & ANYOF_LOCALE) {
4406 PL_reg_flags |= RF_tainted;
4407 f = PL_fold_locale[c];
4411 if (f != c && ANYOF_BITMAP_TEST(n, f))
4415 if (!match && (flags & ANYOF_CLASS)) {
4416 PL_reg_flags |= RF_tainted;
4418 (ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) ||
4419 (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) ||
4420 (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) ||
4421 (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) ||
4422 (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) ||
4423 (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) ||
4424 (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) ||
4425 (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) ||
4426 (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) ||
4427 (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) ||
4428 (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII(c)) ||
4429 (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII(c)) ||
4430 (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) ||
4431 (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) ||
4432 (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) ||
4433 (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) ||
4434 (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) ||
4435 (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) ||
4436 (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) ||
4437 (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) ||
4438 (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) ||
4439 (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) ||
4440 (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) ||
4441 (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) ||
4442 (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) ||
4443 (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) ||
4444 (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) ||
4445 (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) ||
4446 (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK(c)) ||
4447 (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK(c))
4448 ) /* How's that for a conditional? */
4455 return (flags & ANYOF_INVERT) ? !match : match;
4459 S_reghop(pTHX_ U8 *s, I32 off)
4461 return S_reghop3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4465 S_reghop3(pTHX_ U8 *s, I32 off, U8* lim)
4468 while (off-- && s < lim) {
4469 /* XXX could check well-formedness here */
4477 if (UTF8_IS_CONTINUED(*s)) {
4478 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4481 /* XXX could check well-formedness here */
4489 S_reghopmaybe(pTHX_ U8 *s, I32 off)
4491 return S_reghopmaybe3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4495 S_reghopmaybe3(pTHX_ U8* s, I32 off, U8* lim)
4498 while (off-- && s < lim) {
4499 /* XXX could check well-formedness here */
4509 if (UTF8_IS_CONTINUED(*s)) {
4510 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4513 /* XXX could check well-formedness here */
4525 restore_pos(pTHX_ void *arg)
4527 if (PL_reg_eval_set) {
4528 if (PL_reg_oldsaved) {
4529 PL_reg_re->subbeg = PL_reg_oldsaved;
4530 PL_reg_re->sublen = PL_reg_oldsavedlen;
4531 RX_MATCH_COPIED_on(PL_reg_re);
4533 PL_reg_magic->mg_len = PL_reg_oldpos;
4534 PL_reg_eval_set = 0;
4535 PL_curpm = PL_reg_oldcurpm;
4540 S_to_utf8_substr(pTHX_ register regexp *prog)
4543 if (prog->float_substr && !prog->float_utf8) {
4544 prog->float_utf8 = sv = NEWSV(117, 0);
4545 SvSetSV(sv, prog->float_substr);
4546 sv_utf8_upgrade(sv);
4547 if (SvTAIL(prog->float_substr))
4549 if (prog->float_substr == prog->check_substr)
4550 prog->check_utf8 = sv;
4552 if (prog->anchored_substr && !prog->anchored_utf8) {
4553 prog->anchored_utf8 = sv = NEWSV(118, 0);
4554 SvSetSV(sv, prog->anchored_substr);
4555 sv_utf8_upgrade(sv);
4556 if (SvTAIL(prog->anchored_substr))
4558 if (prog->anchored_substr == prog->check_substr)
4559 prog->check_utf8 = sv;
4564 S_to_byte_substr(pTHX_ register regexp *prog)
4567 if (prog->float_utf8 && !prog->float_substr) {
4568 prog->float_substr = sv = NEWSV(117, 0);
4569 SvSetSV(sv, prog->float_utf8);
4570 if (sv_utf8_downgrade(sv, TRUE)) {
4571 if (SvTAIL(prog->float_utf8))
4575 prog->float_substr = sv = &PL_sv_undef;
4577 if (prog->float_utf8 == prog->check_utf8)
4578 prog->check_substr = sv;
4580 if (prog->anchored_utf8 && !prog->anchored_substr) {
4581 prog->anchored_substr = sv = NEWSV(118, 0);
4582 SvSetSV(sv, prog->anchored_utf8);
4583 if (sv_utf8_downgrade(sv, TRUE)) {
4584 if (SvTAIL(prog->anchored_utf8))
4588 prog->anchored_substr = sv = &PL_sv_undef;
4590 if (prog->anchored_utf8 == prog->check_utf8)
4591 prog->check_substr = sv;