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);
544 /* we may be pointing at the wrong string */
545 if (s && RX_MATCH_COPIED(prog))
546 s = prog->subbeg + (s - SvPVX(sv));
548 *data->scream_olds = s;
550 else if (prog->reganch & ROPT_CANY_SEEN)
551 s = fbm_instr((U8*)(s + start_shift),
552 (U8*)(strend - end_shift),
553 check, PL_multiline ? FBMrf_MULTILINE : 0);
555 s = fbm_instr(HOP3(s, start_shift, strend),
556 HOP3(strend, -end_shift, strbeg),
557 check, PL_multiline ? FBMrf_MULTILINE : 0);
559 /* Update the count-of-usability, remove useless subpatterns,
562 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s %s substr `%s%.*s%s'%s%s",
563 (s ? "Found" : "Did not find"),
564 (check == (do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) ? "anchored" : "floating"),
566 (int)(SvCUR(check) - (SvTAIL(check)!=0)),
568 PL_colors[1], (SvTAIL(check) ? "$" : ""),
569 (s ? " at offset " : "...\n") ) );
576 /* Finish the diagnostic message */
577 DEBUG_r(PerlIO_printf(Perl_debug_log, "%ld...\n", (long)(s - i_strpos)) );
579 /* Got a candidate. Check MBOL anchoring, and the *other* substr.
580 Start with the other substr.
581 XXXX no SCREAM optimization yet - and a very coarse implementation
582 XXXX /ttx+/ results in anchored=`ttx', floating=`x'. floating will
583 *always* match. Probably should be marked during compile...
584 Probably it is right to do no SCREAM here...
587 if (do_utf8 ? (prog->float_utf8 && prog->anchored_utf8) : (prog->float_substr && prog->anchored_substr)) {
588 /* Take into account the "other" substring. */
589 /* XXXX May be hopelessly wrong for UTF... */
592 if (check == (do_utf8 ? prog->float_utf8 : prog->float_substr)) {
595 char *last = HOP3c(s, -start_shift, strbeg), *last1, *last2;
599 t = s - prog->check_offset_max;
600 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
602 || ((t = reghopmaybe3_c(s, -(prog->check_offset_max), strpos))
607 t = HOP3c(t, prog->anchored_offset, strend);
608 if (t < other_last) /* These positions already checked */
610 last2 = last1 = HOP3c(strend, -prog->minlen, strbeg);
613 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
614 /* On end-of-str: see comment below. */
615 must = do_utf8 ? prog->anchored_utf8 : prog->anchored_substr;
616 if (must == &PL_sv_undef) {
618 DEBUG_r(must = prog->anchored_utf8); /* for debug */
623 HOP3(HOP3(last1, prog->anchored_offset, strend)
624 + SvCUR(must), -(SvTAIL(must)!=0), strbeg),
626 PL_multiline ? FBMrf_MULTILINE : 0
628 DEBUG_r(PerlIO_printf(Perl_debug_log,
629 "%s anchored substr `%s%.*s%s'%s",
630 (s ? "Found" : "Contradicts"),
633 - (SvTAIL(must)!=0)),
635 PL_colors[1], (SvTAIL(must) ? "$" : "")));
637 if (last1 >= last2) {
638 DEBUG_r(PerlIO_printf(Perl_debug_log,
639 ", giving up...\n"));
642 DEBUG_r(PerlIO_printf(Perl_debug_log,
643 ", trying floating at offset %ld...\n",
644 (long)(HOP3c(s1, 1, strend) - i_strpos)));
645 other_last = HOP3c(last1, prog->anchored_offset+1, strend);
646 s = HOP3c(last, 1, strend);
650 DEBUG_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
651 (long)(s - i_strpos)));
652 t = HOP3c(s, -prog->anchored_offset, strbeg);
653 other_last = HOP3c(s, 1, strend);
661 else { /* Take into account the floating substring. */
666 t = HOP3c(s, -start_shift, strbeg);
668 HOP3c(strend, -prog->minlen + prog->float_min_offset, strbeg);
669 if (CHR_DIST((U8*)last, (U8*)t) > prog->float_max_offset)
670 last = HOP3c(t, prog->float_max_offset, strend);
671 s = HOP3c(t, prog->float_min_offset, strend);
674 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
675 must = do_utf8 ? prog->float_utf8 : prog->float_substr;
676 /* fbm_instr() takes into account exact value of end-of-str
677 if the check is SvTAIL(ed). Since false positives are OK,
678 and end-of-str is not later than strend we are OK. */
679 if (must == &PL_sv_undef) {
681 DEBUG_r(must = prog->float_utf8); /* for debug message */
684 s = fbm_instr((unsigned char*)s,
685 (unsigned char*)last + SvCUR(must)
687 must, PL_multiline ? FBMrf_MULTILINE : 0);
688 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s floating substr `%s%.*s%s'%s",
689 (s ? "Found" : "Contradicts"),
691 (int)(SvCUR(must) - (SvTAIL(must)!=0)),
693 PL_colors[1], (SvTAIL(must) ? "$" : "")));
696 DEBUG_r(PerlIO_printf(Perl_debug_log,
697 ", giving up...\n"));
700 DEBUG_r(PerlIO_printf(Perl_debug_log,
701 ", trying anchored starting at offset %ld...\n",
702 (long)(s1 + 1 - i_strpos)));
704 s = HOP3c(t, 1, strend);
708 DEBUG_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
709 (long)(s - i_strpos)));
710 other_last = s; /* Fix this later. --Hugo */
719 t = s - prog->check_offset_max;
720 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
722 || ((t = reghopmaybe3_c(s, -prog->check_offset_max, strpos))
724 /* Fixed substring is found far enough so that the match
725 cannot start at strpos. */
727 if (ml_anch && t[-1] != '\n') {
728 /* Eventually fbm_*() should handle this, but often
729 anchored_offset is not 0, so this check will not be wasted. */
730 /* XXXX In the code below we prefer to look for "^" even in
731 presence of anchored substrings. And we search even
732 beyond the found float position. These pessimizations
733 are historical artefacts only. */
735 while (t < strend - prog->minlen) {
737 if (t < check_at - prog->check_offset_min) {
738 if (do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) {
739 /* Since we moved from the found position,
740 we definitely contradict the found anchored
741 substr. Due to the above check we do not
742 contradict "check" substr.
743 Thus we can arrive here only if check substr
744 is float. Redo checking for "other"=="fixed".
747 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld, rescanning for anchored from offset %ld...\n",
748 PL_colors[0],PL_colors[1], (long)(strpos - i_strpos), (long)(strpos - i_strpos + prog->anchored_offset)));
749 goto do_other_anchored;
751 /* We don't contradict the found floating substring. */
752 /* XXXX Why not check for STCLASS? */
754 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld...\n",
755 PL_colors[0],PL_colors[1], (long)(s - i_strpos)));
758 /* Position contradicts check-string */
759 /* XXXX probably better to look for check-string
760 than for "\n", so one should lower the limit for t? */
761 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m, restarting lookup for check-string at offset %ld...\n",
762 PL_colors[0],PL_colors[1], (long)(t + 1 - i_strpos)));
763 other_last = strpos = s = t + 1;
768 DEBUG_r(PerlIO_printf(Perl_debug_log, "Did not find /%s^%s/m...\n",
769 PL_colors[0],PL_colors[1]));
773 DEBUG_r(PerlIO_printf(Perl_debug_log, "Starting position does not contradict /%s^%s/m...\n",
774 PL_colors[0],PL_colors[1]));
778 ++BmUSEFUL(do_utf8 ? prog->check_utf8 : prog->check_substr); /* hooray/5 */
781 /* The found string does not prohibit matching at strpos,
782 - no optimization of calling REx engine can be performed,
783 unless it was an MBOL and we are not after MBOL,
784 or a future STCLASS check will fail this. */
786 /* Even in this situation we may use MBOL flag if strpos is offset
787 wrt the start of the string. */
788 if (ml_anch && sv && !SvROK(sv) /* See prev comment on SvROK */
789 && (strpos != strbeg) && strpos[-1] != '\n'
790 /* May be due to an implicit anchor of m{.*foo} */
791 && !(prog->reganch & ROPT_IMPLICIT))
796 DEBUG_r( if (ml_anch)
797 PerlIO_printf(Perl_debug_log, "Position at offset %ld does not contradict /%s^%s/m...\n",
798 (long)(strpos - i_strpos), PL_colors[0],PL_colors[1]);
801 if (!(prog->reganch & ROPT_NAUGHTY) /* XXXX If strpos moved? */
803 prog->check_utf8 /* Could be deleted already */
804 && --BmUSEFUL(prog->check_utf8) < 0
805 && (prog->check_utf8 == prog->float_utf8)
807 prog->check_substr /* Could be deleted already */
808 && --BmUSEFUL(prog->check_substr) < 0
809 && (prog->check_substr == prog->float_substr)
812 /* If flags & SOMETHING - do not do it many times on the same match */
813 DEBUG_r(PerlIO_printf(Perl_debug_log, "... Disabling check substring...\n"));
814 SvREFCNT_dec(do_utf8 ? prog->check_utf8 : prog->check_substr);
815 if (do_utf8 ? prog->check_substr : prog->check_utf8)
816 SvREFCNT_dec(do_utf8 ? prog->check_substr : prog->check_utf8);
817 prog->check_substr = prog->check_utf8 = Nullsv; /* disable */
818 prog->float_substr = prog->float_utf8 = Nullsv; /* clear */
819 check = Nullsv; /* abort */
821 /* XXXX This is a remnant of the old implementation. It
822 looks wasteful, since now INTUIT can use many
824 prog->reganch &= ~RE_USE_INTUIT;
831 /* XXXX BmUSEFUL already changed, maybe multiple change is meaningful... */
832 if (prog->regstclass) {
833 /* minlen == 0 is possible if regstclass is \b or \B,
834 and the fixed substr is ''$.
835 Since minlen is already taken into account, s+1 is before strend;
836 accidentally, minlen >= 1 guaranties no false positives at s + 1
837 even for \b or \B. But (minlen? 1 : 0) below assumes that
838 regstclass does not come from lookahead... */
839 /* If regstclass takes bytelength more than 1: If charlength==1, OK.
840 This leaves EXACTF only, which is dealt with in find_byclass(). */
841 U8* str = (U8*)STRING(prog->regstclass);
842 int cl_l = (PL_regkind[(U8)OP(prog->regstclass)] == EXACT
843 ? CHR_DIST(str+STR_LEN(prog->regstclass), str)
845 char *endpos = (prog->anchored_substr || prog->anchored_utf8 || ml_anch)
846 ? HOP3c(s, (prog->minlen ? cl_l : 0), strend)
847 : (prog->float_substr || prog->float_utf8
848 ? HOP3c(HOP3c(check_at, -start_shift, strbeg),
851 char *startpos = strbeg;
854 if (prog->reganch & ROPT_UTF8) {
855 PL_regdata = prog->data;
858 s = find_byclass(prog, prog->regstclass, s, endpos, startpos, 1);
863 if (endpos == strend) {
864 DEBUG_r( PerlIO_printf(Perl_debug_log,
865 "Could not match STCLASS...\n") );
868 DEBUG_r( PerlIO_printf(Perl_debug_log,
869 "This position contradicts STCLASS...\n") );
870 if ((prog->reganch & ROPT_ANCH) && !ml_anch)
872 /* Contradict one of substrings */
873 if (prog->anchored_substr || prog->anchored_utf8) {
874 if ((do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) == check) {
875 DEBUG_r( what = "anchored" );
877 s = HOP3c(t, 1, strend);
878 if (s + start_shift + end_shift > strend) {
879 /* XXXX Should be taken into account earlier? */
880 DEBUG_r( PerlIO_printf(Perl_debug_log,
881 "Could not match STCLASS...\n") );
886 DEBUG_r( PerlIO_printf(Perl_debug_log,
887 "Looking for %s substr starting at offset %ld...\n",
888 what, (long)(s + start_shift - i_strpos)) );
891 /* Have both, check_string is floating */
892 if (t + start_shift >= check_at) /* Contradicts floating=check */
893 goto retry_floating_check;
894 /* Recheck anchored substring, but not floating... */
898 DEBUG_r( PerlIO_printf(Perl_debug_log,
899 "Looking for anchored substr starting at offset %ld...\n",
900 (long)(other_last - i_strpos)) );
901 goto do_other_anchored;
903 /* Another way we could have checked stclass at the
904 current position only: */
909 DEBUG_r( PerlIO_printf(Perl_debug_log,
910 "Looking for /%s^%s/m starting at offset %ld...\n",
911 PL_colors[0],PL_colors[1], (long)(t - i_strpos)) );
914 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr)) /* Could have been deleted */
916 /* Check is floating subtring. */
917 retry_floating_check:
918 t = check_at - start_shift;
919 DEBUG_r( what = "floating" );
920 goto hop_and_restart;
923 DEBUG_r(PerlIO_printf(Perl_debug_log,
924 "By STCLASS: moving %ld --> %ld\n",
925 (long)(t - i_strpos), (long)(s - i_strpos))
929 DEBUG_r(PerlIO_printf(Perl_debug_log,
930 "Does not contradict STCLASS...\n");
935 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s%s:%s match at offset %ld\n",
936 PL_colors[4], (check ? "Guessed" : "Giving up"),
937 PL_colors[5], (long)(s - i_strpos)) );
940 fail_finish: /* Substring not found */
941 if (prog->check_substr || prog->check_utf8) /* could be removed already */
942 BmUSEFUL(do_utf8 ? prog->check_utf8 : prog->check_substr) += 5; /* hooray */
944 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch rejected by optimizer%s\n",
945 PL_colors[4],PL_colors[5]));
949 /* We know what class REx starts with. Try to find this position... */
951 S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, char *strend, char *startpos, I32 norun)
953 I32 doevery = (prog->reganch & ROPT_SKIP) == 0;
959 register I32 tmp = 1; /* Scratch variable? */
960 register bool do_utf8 = PL_reg_match_utf8;
962 /* We know what class it must start with. */
967 if ((ANYOF_FLAGS(c) & ANYOF_UNICODE) ||
968 !UTF8_IS_INVARIANT((U8)s[0]) ?
969 reginclass(c, (U8*)s, 0, do_utf8) :
970 REGINCLASS(c, (U8*)s)) {
971 if (tmp && (norun || regtry(prog, s)))
985 if (REGINCLASS(c, (U8*)s) ||
986 (ANYOF_FOLD_SHARP_S(c, s, strend) &&
987 /* The assignment of 2 is intentional:
988 * for the folded sharp s, the skip is 2. */
989 (skip = SHARP_S_SKIP))) {
990 if (tmp && (norun || regtry(prog, s)))
1002 while (s < strend) {
1003 if (tmp && (norun || regtry(prog, s)))
1014 STRLEN ulen1, ulen2;
1015 U8 tmpbuf1[UTF8_MAXLEN_UCLC+1];
1016 U8 tmpbuf2[UTF8_MAXLEN_UCLC+1];
1018 to_utf8_lower((U8*)m, tmpbuf1, &ulen1);
1019 to_utf8_upper((U8*)m, tmpbuf2, &ulen2);
1021 c1 = utf8n_to_uvchr(tmpbuf1, UTF8_MAXLEN_UCLC,
1022 0, ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
1023 c2 = utf8n_to_uvchr(tmpbuf2, UTF8_MAXLEN_UCLC,
1024 0, ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
1035 c2 = PL_fold_locale[c1];
1037 e = HOP3c(strend, -(I32)ln, s);
1040 e = s; /* Due to minlen logic of intuit() */
1042 /* The idea in the EXACTF* cases is to first find the
1043 * first character of the EXACTF* node and then, if
1044 * necessary, case-insensitively compare the full
1045 * text of the node. The c1 and c2 are the first
1046 * characters (though in Unicode it gets a bit
1047 * more complicated because there are more cases
1048 * than just upper and lower: one needs to use
1049 * the so-called folding case for case-insensitive
1050 * matching (called "loose matching" in Unicode).
1051 * ibcmp_utf8() will do just that. */
1055 U8 tmpbuf [UTF8_MAXLEN+1];
1056 U8 foldbuf[UTF8_MAXLEN_FOLD+1];
1057 STRLEN len, foldlen;
1061 c = utf8n_to_uvchr((U8*)s, UTF8_MAXLEN, &len,
1063 0 : UTF8_ALLOW_ANY);
1066 ibcmp_utf8(s, (char **)0, 0, do_utf8,
1067 m, (char **)0, ln, (bool)UTF))
1068 && (norun || regtry(prog, s)) )
1071 uvchr_to_utf8(tmpbuf, c);
1072 f = to_utf8_fold(tmpbuf, foldbuf, &foldlen);
1074 && (f == c1 || f == c2)
1075 && (ln == foldlen ||
1076 !ibcmp_utf8((char *) foldbuf,
1077 (char **)0, foldlen, do_utf8,
1079 (char **)0, ln, (bool)UTF))
1080 && (norun || regtry(prog, s)) )
1088 c = utf8n_to_uvchr((U8*)s, UTF8_MAXLEN, &len,
1090 0 : UTF8_ALLOW_ANY);
1092 /* Handle some of the three Greek sigmas cases.
1093 * Note that not all the possible combinations
1094 * are handled here: some of them are handled
1095 * by the standard folding rules, and some of
1096 * them (the character class or ANYOF cases)
1097 * are handled during compiletime in
1098 * regexec.c:S_regclass(). */
1099 if (c == (UV)UNICODE_GREEK_CAPITAL_LETTER_SIGMA ||
1100 c == (UV)UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA)
1101 c = (UV)UNICODE_GREEK_SMALL_LETTER_SIGMA;
1103 if ( (c == c1 || c == c2)
1105 ibcmp_utf8(s, (char **)0, 0, do_utf8,
1106 m, (char **)0, ln, (bool)UTF))
1107 && (norun || regtry(prog, s)) )
1110 uvchr_to_utf8(tmpbuf, c);
1111 f = to_utf8_fold(tmpbuf, foldbuf, &foldlen);
1113 && (f == c1 || f == c2)
1114 && (ln == foldlen ||
1115 !ibcmp_utf8((char *) foldbuf,
1116 (char **)0, foldlen, do_utf8,
1118 (char **)0, ln, (bool)UTF))
1119 && (norun || regtry(prog, s)) )
1130 && (ln == 1 || !(OP(c) == EXACTF
1132 : ibcmp_locale(s, m, ln)))
1133 && (norun || regtry(prog, s)) )
1139 if ( (*(U8*)s == c1 || *(U8*)s == c2)
1140 && (ln == 1 || !(OP(c) == EXACTF
1142 : ibcmp_locale(s, m, ln)))
1143 && (norun || regtry(prog, s)) )
1150 PL_reg_flags |= RF_tainted;
1157 U8 *r = reghop3((U8*)s, -1, (U8*)PL_bostr);
1159 tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, 0);
1161 tmp = ((OP(c) == BOUND ?
1162 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1163 LOAD_UTF8_CHARCLASS(alnum,"a");
1164 while (s < strend) {
1165 if (tmp == !(OP(c) == BOUND ?
1166 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1167 isALNUM_LC_utf8((U8*)s)))
1170 if ((norun || regtry(prog, s)))
1177 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1178 tmp = ((OP(c) == BOUND ? isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1179 while (s < strend) {
1181 !(OP(c) == BOUND ? isALNUM(*s) : isALNUM_LC(*s))) {
1183 if ((norun || regtry(prog, s)))
1189 if ((!prog->minlen && tmp) && (norun || regtry(prog, s)))
1193 PL_reg_flags |= RF_tainted;
1200 U8 *r = reghop3((U8*)s, -1, (U8*)PL_bostr);
1202 tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, 0);
1204 tmp = ((OP(c) == NBOUND ?
1205 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1206 LOAD_UTF8_CHARCLASS(alnum,"a");
1207 while (s < strend) {
1208 if (tmp == !(OP(c) == NBOUND ?
1209 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1210 isALNUM_LC_utf8((U8*)s)))
1212 else if ((norun || regtry(prog, s)))
1218 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1219 tmp = ((OP(c) == NBOUND ?
1220 isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1221 while (s < strend) {
1223 !(OP(c) == NBOUND ? isALNUM(*s) : isALNUM_LC(*s)))
1225 else if ((norun || regtry(prog, s)))
1230 if ((!prog->minlen && !tmp) && (norun || regtry(prog, s)))
1235 LOAD_UTF8_CHARCLASS(alnum,"a");
1236 while (s < strend) {
1237 if (swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1238 if (tmp && (norun || regtry(prog, s)))
1249 while (s < strend) {
1251 if (tmp && (norun || regtry(prog, s)))
1263 PL_reg_flags |= RF_tainted;
1265 while (s < strend) {
1266 if (isALNUM_LC_utf8((U8*)s)) {
1267 if (tmp && (norun || regtry(prog, s)))
1278 while (s < strend) {
1279 if (isALNUM_LC(*s)) {
1280 if (tmp && (norun || regtry(prog, s)))
1293 LOAD_UTF8_CHARCLASS(alnum,"a");
1294 while (s < strend) {
1295 if (!swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1296 if (tmp && (norun || regtry(prog, s)))
1307 while (s < strend) {
1309 if (tmp && (norun || regtry(prog, s)))
1321 PL_reg_flags |= RF_tainted;
1323 while (s < strend) {
1324 if (!isALNUM_LC_utf8((U8*)s)) {
1325 if (tmp && (norun || regtry(prog, s)))
1336 while (s < strend) {
1337 if (!isALNUM_LC(*s)) {
1338 if (tmp && (norun || regtry(prog, s)))
1351 LOAD_UTF8_CHARCLASS(space," ");
1352 while (s < strend) {
1353 if (*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8)) {
1354 if (tmp && (norun || regtry(prog, s)))
1365 while (s < strend) {
1367 if (tmp && (norun || regtry(prog, s)))
1379 PL_reg_flags |= RF_tainted;
1381 while (s < strend) {
1382 if (*s == ' ' || isSPACE_LC_utf8((U8*)s)) {
1383 if (tmp && (norun || regtry(prog, s)))
1394 while (s < strend) {
1395 if (isSPACE_LC(*s)) {
1396 if (tmp && (norun || regtry(prog, s)))
1409 LOAD_UTF8_CHARCLASS(space," ");
1410 while (s < strend) {
1411 if (!(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8))) {
1412 if (tmp && (norun || regtry(prog, s)))
1423 while (s < strend) {
1425 if (tmp && (norun || regtry(prog, s)))
1437 PL_reg_flags |= RF_tainted;
1439 while (s < strend) {
1440 if (!(*s == ' ' || isSPACE_LC_utf8((U8*)s))) {
1441 if (tmp && (norun || regtry(prog, s)))
1452 while (s < strend) {
1453 if (!isSPACE_LC(*s)) {
1454 if (tmp && (norun || regtry(prog, s)))
1467 LOAD_UTF8_CHARCLASS(digit,"0");
1468 while (s < strend) {
1469 if (swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1470 if (tmp && (norun || regtry(prog, s)))
1481 while (s < strend) {
1483 if (tmp && (norun || regtry(prog, s)))
1495 PL_reg_flags |= RF_tainted;
1497 while (s < strend) {
1498 if (isDIGIT_LC_utf8((U8*)s)) {
1499 if (tmp && (norun || regtry(prog, s)))
1510 while (s < strend) {
1511 if (isDIGIT_LC(*s)) {
1512 if (tmp && (norun || regtry(prog, s)))
1525 LOAD_UTF8_CHARCLASS(digit,"0");
1526 while (s < strend) {
1527 if (!swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1528 if (tmp && (norun || regtry(prog, s)))
1539 while (s < strend) {
1541 if (tmp && (norun || regtry(prog, s)))
1553 PL_reg_flags |= RF_tainted;
1555 while (s < strend) {
1556 if (!isDIGIT_LC_utf8((U8*)s)) {
1557 if (tmp && (norun || regtry(prog, s)))
1568 while (s < strend) {
1569 if (!isDIGIT_LC(*s)) {
1570 if (tmp && (norun || regtry(prog, s)))
1582 Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c));
1591 - regexec_flags - match a regexp against a string
1594 Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char *strend,
1595 char *strbeg, I32 minend, SV *sv, void *data, U32 flags)
1596 /* strend: pointer to null at end of string */
1597 /* strbeg: real beginning of string */
1598 /* minend: end of match must be >=minend after stringarg. */
1599 /* data: May be used for some additional optimizations. */
1600 /* nosave: For optimizations. */
1603 register regnode *c;
1604 register char *startpos = stringarg;
1605 I32 minlen; /* must match at least this many chars */
1606 I32 dontbother = 0; /* how many characters not to try at end */
1607 /* I32 start_shift = 0; */ /* Offset of the start to find
1608 constant substr. */ /* CC */
1609 I32 end_shift = 0; /* Same for the end. */ /* CC */
1610 I32 scream_pos = -1; /* Internal iterator of scream. */
1612 SV* oreplsv = GvSV(PL_replgv);
1613 bool do_utf8 = DO_UTF8(sv);
1615 SV *dsv0 = PERL_DEBUG_PAD_ZERO(0);
1616 SV *dsv1 = PERL_DEBUG_PAD_ZERO(1);
1623 PL_regnarrate = DEBUG_r_TEST;
1626 /* Be paranoid... */
1627 if (prog == NULL || startpos == NULL) {
1628 Perl_croak(aTHX_ "NULL regexp parameter");
1632 minlen = prog->minlen;
1633 if (strend - startpos < minlen) {
1634 DEBUG_r(PerlIO_printf(Perl_debug_log,
1635 "String too short [regexec_flags]...\n"));
1639 /* Check validity of program. */
1640 if (UCHARAT(prog->program) != REG_MAGIC) {
1641 Perl_croak(aTHX_ "corrupted regexp program");
1645 PL_reg_eval_set = 0;
1648 if (prog->reganch & ROPT_UTF8)
1649 PL_reg_flags |= RF_utf8;
1651 /* Mark beginning of line for ^ and lookbehind. */
1652 PL_regbol = startpos;
1656 /* Mark end of line for $ (and such) */
1659 /* see how far we have to get to not match where we matched before */
1660 PL_regtill = startpos+minend;
1662 /* We start without call_cc context. */
1665 /* If there is a "must appear" string, look for it. */
1668 if (prog->reganch & ROPT_GPOS_SEEN) { /* Need to have PL_reg_ganch */
1671 if (flags & REXEC_IGNOREPOS) /* Means: check only at start */
1672 PL_reg_ganch = startpos;
1673 else if (sv && SvTYPE(sv) >= SVt_PVMG
1675 && (mg = mg_find(sv, PERL_MAGIC_regex_global))
1676 && mg->mg_len >= 0) {
1677 PL_reg_ganch = strbeg + mg->mg_len; /* Defined pos() */
1678 if (prog->reganch & ROPT_ANCH_GPOS) {
1679 if (s > PL_reg_ganch)
1684 else /* pos() not defined */
1685 PL_reg_ganch = strbeg;
1688 if (!(flags & REXEC_CHECKED) && (prog->check_substr != Nullsv || prog->check_utf8 != Nullsv)) {
1689 re_scream_pos_data d;
1691 d.scream_olds = &scream_olds;
1692 d.scream_pos = &scream_pos;
1693 s = re_intuit_start(prog, sv, s, strend, flags, &d);
1695 DEBUG_r(PerlIO_printf(Perl_debug_log, "Not present...\n"));
1696 goto phooey; /* not present */
1702 pv_uni_display(dsv0, (U8*)prog->precomp, prog->prelen, 60,
1703 UNI_DISPLAY_REGEX) :
1705 int len0 = UTF ? SvCUR(dsv0) : prog->prelen;
1706 char *s1 = do_utf8 ? sv_uni_display(dsv1, sv, 60,
1707 UNI_DISPLAY_REGEX) : startpos;
1708 int len1 = do_utf8 ? SvCUR(dsv1) : strend - startpos;
1711 PerlIO_printf(Perl_debug_log,
1712 "%sMatching REx%s `%s%*.*s%s%s' against `%s%.*s%s%s'\n",
1713 PL_colors[4],PL_colors[5],PL_colors[0],
1716 len0 > 60 ? "..." : "",
1718 (int)(len1 > 60 ? 60 : len1),
1720 (len1 > 60 ? "..." : "")
1724 /* Simplest case: anchored match need be tried only once. */
1725 /* [unless only anchor is BOL and multiline is set] */
1726 if (prog->reganch & (ROPT_ANCH & ~ROPT_ANCH_GPOS)) {
1727 if (s == startpos && regtry(prog, startpos))
1729 else if (PL_multiline || (prog->reganch & ROPT_IMPLICIT)
1730 || (prog->reganch & ROPT_ANCH_MBOL)) /* XXXX SBOL? */
1735 dontbother = minlen - 1;
1736 end = HOP3c(strend, -dontbother, strbeg) - 1;
1737 /* for multiline we only have to try after newlines */
1738 if (prog->check_substr || prog->check_utf8) {
1742 if (regtry(prog, s))
1747 if (prog->reganch & RE_USE_INTUIT) {
1748 s = re_intuit_start(prog, sv, s + 1, strend, flags, NULL);
1759 if (*s++ == '\n') { /* don't need PL_utf8skip here */
1760 if (regtry(prog, s))
1767 } else if (prog->reganch & ROPT_ANCH_GPOS) {
1768 if (regtry(prog, PL_reg_ganch))
1773 /* Messy cases: unanchored match. */
1774 if ((prog->anchored_substr || prog->anchored_utf8) && prog->reganch & ROPT_SKIP) {
1775 /* we have /x+whatever/ */
1776 /* it must be a one character string (XXXX Except UTF?) */
1781 if (!(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr))
1782 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1783 ch = SvPVX(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr)[0];
1786 while (s < strend) {
1788 DEBUG_r( did_match = 1 );
1789 if (regtry(prog, s)) goto got_it;
1791 while (s < strend && *s == ch)
1798 while (s < strend) {
1800 DEBUG_r( did_match = 1 );
1801 if (regtry(prog, s)) goto got_it;
1803 while (s < strend && *s == ch)
1809 DEBUG_r(if (!did_match)
1810 PerlIO_printf(Perl_debug_log,
1811 "Did not find anchored character...\n")
1815 else if (prog->anchored_substr != Nullsv
1816 || prog->anchored_utf8 != Nullsv
1817 || ((prog->float_substr != Nullsv || prog->float_utf8 != Nullsv)
1818 && prog->float_max_offset < strend - s)) {
1823 char *last1; /* Last position checked before */
1827 if (prog->anchored_substr || prog->anchored_utf8) {
1828 if (!(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr))
1829 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1830 must = do_utf8 ? prog->anchored_utf8 : prog->anchored_substr;
1831 back_max = back_min = prog->anchored_offset;
1833 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr))
1834 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1835 must = do_utf8 ? prog->float_utf8 : prog->float_substr;
1836 back_max = prog->float_max_offset;
1837 back_min = prog->float_min_offset;
1839 if (must == &PL_sv_undef)
1840 /* could not downgrade utf8 check substring, so must fail */
1843 last = HOP3c(strend, /* Cannot start after this */
1844 -(I32)(CHR_SVLEN(must)
1845 - (SvTAIL(must) != 0) + back_min), strbeg);
1848 last1 = HOPc(s, -1);
1850 last1 = s - 1; /* bogus */
1852 /* XXXX check_substr already used to find `s', can optimize if
1853 check_substr==must. */
1855 dontbother = end_shift;
1856 strend = HOPc(strend, -dontbother);
1857 while ( (s <= last) &&
1858 ((flags & REXEC_SCREAM)
1859 ? (s = screaminstr(sv, must, HOP3c(s, back_min, strend) - strbeg,
1860 end_shift, &scream_pos, 0))
1861 : (s = fbm_instr((unsigned char*)HOP3(s, back_min, strend),
1862 (unsigned char*)strend, must,
1863 PL_multiline ? FBMrf_MULTILINE : 0))) ) {
1864 /* we may be pointing at the wrong string */
1865 if ((flags & REXEC_SCREAM) && RX_MATCH_COPIED(prog))
1866 s = prog->subbeg + (s - SvPVX(sv));
1867 DEBUG_r( did_match = 1 );
1868 if (HOPc(s, -back_max) > last1) {
1869 last1 = HOPc(s, -back_min);
1870 s = HOPc(s, -back_max);
1873 char *t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1;
1875 last1 = HOPc(s, -back_min);
1879 while (s <= last1) {
1880 if (regtry(prog, s))
1886 while (s <= last1) {
1887 if (regtry(prog, s))
1893 DEBUG_r(if (!did_match)
1894 PerlIO_printf(Perl_debug_log,
1895 "Did not find %s substr `%s%.*s%s'%s...\n",
1896 ((must == prog->anchored_substr || must == prog->anchored_utf8)
1897 ? "anchored" : "floating"),
1899 (int)(SvCUR(must) - (SvTAIL(must)!=0)),
1901 PL_colors[1], (SvTAIL(must) ? "$" : ""))
1905 else if ((c = prog->regstclass)) {
1907 I32 op = (U8)OP(prog->regstclass);
1908 /* don't bother with what can't match */
1909 if (PL_regkind[op] != EXACT && op != CANY)
1910 strend = HOPc(strend, -(minlen - 1));
1913 SV *prop = sv_newmortal();
1921 pv_uni_display(dsv0, (U8*)SvPVX(prop), SvCUR(prop), 60,
1922 UNI_DISPLAY_REGEX) :
1924 len0 = UTF ? SvCUR(dsv0) : SvCUR(prop);
1926 sv_uni_display(dsv1, sv, 60, UNI_DISPLAY_REGEX) : s;
1927 len1 = UTF ? SvCUR(dsv1) : strend - s;
1928 PerlIO_printf(Perl_debug_log,
1929 "Matching stclass `%*.*s' against `%*.*s'\n",
1933 if (find_byclass(prog, c, s, strend, startpos, 0))
1935 DEBUG_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass...\n"));
1939 if (prog->float_substr != Nullsv || prog->float_utf8 != Nullsv) {
1944 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr))
1945 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1946 float_real = do_utf8 ? prog->float_utf8 : prog->float_substr;
1948 if (flags & REXEC_SCREAM) {
1949 last = screaminstr(sv, float_real, s - strbeg,
1950 end_shift, &scream_pos, 1); /* last one */
1952 last = scream_olds; /* Only one occurrence. */
1953 /* we may be pointing at the wrong string */
1954 else if (RX_MATCH_COPIED(prog))
1955 s = prog->subbeg + (s - SvPVX(sv));
1959 char *little = SvPV(float_real, len);
1961 if (SvTAIL(float_real)) {
1962 if (memEQ(strend - len + 1, little, len - 1))
1963 last = strend - len + 1;
1964 else if (!PL_multiline)
1965 last = memEQ(strend - len, little, len)
1966 ? strend - len : Nullch;
1972 last = rninstr(s, strend, little, little + len);
1974 last = strend; /* matching `$' */
1978 DEBUG_r(PerlIO_printf(Perl_debug_log,
1979 "%sCan't trim the tail, match fails (should not happen)%s\n",
1980 PL_colors[4],PL_colors[5]));
1981 goto phooey; /* Should not happen! */
1983 dontbother = strend - last + prog->float_min_offset;
1985 if (minlen && (dontbother < minlen))
1986 dontbother = minlen - 1;
1987 strend -= dontbother; /* this one's always in bytes! */
1988 /* We don't know much -- general case. */
1991 if (regtry(prog, s))
2000 if (regtry(prog, s))
2002 } while (s++ < strend);
2010 RX_MATCH_TAINTED_set(prog, PL_reg_flags & RF_tainted);
2012 if (PL_reg_eval_set) {
2013 /* Preserve the current value of $^R */
2014 if (oreplsv != GvSV(PL_replgv))
2015 sv_setsv(oreplsv, GvSV(PL_replgv));/* So that when GvSV(replgv) is
2016 restored, the value remains
2018 restore_pos(aTHX_ 0);
2021 /* make sure $`, $&, $', and $digit will work later */
2022 if ( !(flags & REXEC_NOT_FIRST) ) {
2023 if (RX_MATCH_COPIED(prog)) {
2024 Safefree(prog->subbeg);
2025 RX_MATCH_COPIED_off(prog);
2027 if (flags & REXEC_COPY_STR) {
2028 I32 i = PL_regeol - startpos + (stringarg - strbeg);
2030 s = savepvn(strbeg, i);
2033 RX_MATCH_COPIED_on(prog);
2036 prog->subbeg = strbeg;
2037 prog->sublen = PL_regeol - strbeg; /* strend may have been modified */
2044 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n",
2045 PL_colors[4],PL_colors[5]));
2046 if (PL_reg_eval_set)
2047 restore_pos(aTHX_ 0);
2052 - regtry - try match at specific point
2054 STATIC I32 /* 0 failure, 1 success */
2055 S_regtry(pTHX_ regexp *prog, char *startpos)
2063 PL_regindent = 0; /* XXXX Not good when matches are reenterable... */
2065 if ((prog->reganch & ROPT_EVAL_SEEN) && !PL_reg_eval_set) {
2068 PL_reg_eval_set = RS_init;
2070 PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %"IVdf"\n",
2071 (IV)(PL_stack_sp - PL_stack_base));
2073 SAVEI32(cxstack[cxstack_ix].blk_oldsp);
2074 cxstack[cxstack_ix].blk_oldsp = PL_stack_sp - PL_stack_base;
2075 /* Otherwise OP_NEXTSTATE will free whatever on stack now. */
2077 /* Apparently this is not needed, judging by wantarray. */
2078 /* SAVEI8(cxstack[cxstack_ix].blk_gimme);
2079 cxstack[cxstack_ix].blk_gimme = G_SCALAR; */
2082 /* Make $_ available to executed code. */
2083 if (PL_reg_sv != DEFSV) {
2084 /* SAVE_DEFSV does *not* suffice here for USE_5005THREADS */
2089 if (!(SvTYPE(PL_reg_sv) >= SVt_PVMG && SvMAGIC(PL_reg_sv)
2090 && (mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global)))) {
2091 /* prepare for quick setting of pos */
2092 sv_magic(PL_reg_sv, (SV*)0,
2093 PERL_MAGIC_regex_global, Nullch, 0);
2094 mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global);
2098 PL_reg_oldpos = mg->mg_len;
2099 SAVEDESTRUCTOR_X(restore_pos, 0);
2101 if (!PL_reg_curpm) {
2102 Newz(22,PL_reg_curpm, 1, PMOP);
2105 SV* repointer = newSViv(0);
2106 /* so we know which PL_regex_padav element is PL_reg_curpm */
2107 SvFLAGS(repointer) |= SVf_BREAK;
2108 av_push(PL_regex_padav,repointer);
2109 PL_reg_curpm->op_pmoffset = av_len(PL_regex_padav);
2110 PL_regex_pad = AvARRAY(PL_regex_padav);
2114 PM_SETRE(PL_reg_curpm, prog);
2115 PL_reg_oldcurpm = PL_curpm;
2116 PL_curpm = PL_reg_curpm;
2117 if (RX_MATCH_COPIED(prog)) {
2118 /* Here is a serious problem: we cannot rewrite subbeg,
2119 since it may be needed if this match fails. Thus
2120 $` inside (?{}) could fail... */
2121 PL_reg_oldsaved = prog->subbeg;
2122 PL_reg_oldsavedlen = prog->sublen;
2123 RX_MATCH_COPIED_off(prog);
2126 PL_reg_oldsaved = Nullch;
2127 prog->subbeg = PL_bostr;
2128 prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */
2130 prog->startp[0] = startpos - PL_bostr;
2131 PL_reginput = startpos;
2132 PL_regstartp = prog->startp;
2133 PL_regendp = prog->endp;
2134 PL_reglastparen = &prog->lastparen;
2135 PL_reglastcloseparen = &prog->lastcloseparen;
2136 prog->lastparen = 0;
2137 prog->lastcloseparen = 0;
2139 DEBUG_r(PL_reg_starttry = startpos);
2140 if (PL_reg_start_tmpl <= prog->nparens) {
2141 PL_reg_start_tmpl = prog->nparens*3/2 + 3;
2142 if(PL_reg_start_tmp)
2143 Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
2145 New(22,PL_reg_start_tmp, PL_reg_start_tmpl, char*);
2148 /* XXXX What this code is doing here?!!! There should be no need
2149 to do this again and again, PL_reglastparen should take care of
2152 /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code.
2153 * Actually, the code in regcppop() (which Ilya may be meaning by
2154 * PL_reglastparen), is not needed at all by the test suite
2155 * (op/regexp, op/pat, op/split), but that code is needed, oddly
2156 * enough, for building DynaLoader, or otherwise this
2157 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
2158 * will happen. Meanwhile, this code *is* needed for the
2159 * above-mentioned test suite tests to succeed. The common theme
2160 * on those tests seems to be returning null fields from matches.
2165 if (prog->nparens) {
2166 for (i = prog->nparens; i > (I32)*PL_reglastparen; i--) {
2173 if (regmatch(prog->program + 1)) {
2174 prog->endp[0] = PL_reginput - PL_bostr;
2177 REGCP_UNWIND(lastcp);
2181 #define RE_UNWIND_BRANCH 1
2182 #define RE_UNWIND_BRANCHJ 2
2186 typedef struct { /* XX: makes sense to enlarge it... */
2190 } re_unwind_generic_t;
2203 } re_unwind_branch_t;
2205 typedef union re_unwind_t {
2207 re_unwind_generic_t generic;
2208 re_unwind_branch_t branch;
2211 #define sayYES goto yes
2212 #define sayNO goto no
2213 #define sayNO_ANYOF goto no_anyof
2214 #define sayYES_FINAL goto yes_final
2215 #define sayYES_LOUD goto yes_loud
2216 #define sayNO_FINAL goto no_final
2217 #define sayNO_SILENT goto do_no
2218 #define saySAME(x) if (x) goto yes; else goto no
2220 #define REPORT_CODE_OFF 24
2223 - regmatch - main matching routine
2225 * Conceptually the strategy is simple: check to see whether the current
2226 * node matches, call self recursively to see whether the rest matches,
2227 * and then act accordingly. In practice we make some effort to avoid
2228 * recursion, in particular by going through "ordinary" nodes (that don't
2229 * need to know whether the rest of the match failed) by a loop instead of
2232 /* [lwall] I've hoisted the register declarations to the outer block in order to
2233 * maybe save a little bit of pushing and popping on the stack. It also takes
2234 * advantage of machines that use a register save mask on subroutine entry.
2236 STATIC I32 /* 0 failure, 1 success */
2237 S_regmatch(pTHX_ regnode *prog)
2239 register regnode *scan; /* Current node. */
2240 regnode *next; /* Next node. */
2241 regnode *inner; /* Next node in internal branch. */
2242 register I32 nextchr; /* renamed nextchr - nextchar colides with
2243 function of same name */
2244 register I32 n; /* no or next */
2245 register I32 ln = 0; /* len or last */
2246 register char *s = Nullch; /* operand or save */
2247 register char *locinput = PL_reginput;
2248 register I32 c1 = 0, c2 = 0, paren; /* case fold search, parenth */
2249 int minmod = 0, sw = 0, logical = 0;
2252 I32 firstcp = PL_savestack_ix;
2254 register bool do_utf8 = PL_reg_match_utf8;
2256 SV *dsv0 = PERL_DEBUG_PAD_ZERO(0);
2257 SV *dsv1 = PERL_DEBUG_PAD_ZERO(1);
2258 SV *dsv2 = PERL_DEBUG_PAD_ZERO(2);
2265 /* Note that nextchr is a byte even in UTF */
2266 nextchr = UCHARAT(locinput);
2268 while (scan != NULL) {
2271 SV *prop = sv_newmortal();
2272 int docolor = *PL_colors[0];
2273 int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */
2274 int l = (PL_regeol - locinput) > taill ? taill : (PL_regeol - locinput);
2275 /* The part of the string before starttry has one color
2276 (pref0_len chars), between starttry and current
2277 position another one (pref_len - pref0_len chars),
2278 after the current position the third one.
2279 We assume that pref0_len <= pref_len, otherwise we
2280 decrease pref0_len. */
2281 int pref_len = (locinput - PL_bostr) > (5 + taill) - l
2282 ? (5 + taill) - l : locinput - PL_bostr;
2285 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len)))
2287 pref0_len = pref_len - (locinput - PL_reg_starttry);
2288 if (l + pref_len < (5 + taill) && l < PL_regeol - locinput)
2289 l = ( PL_regeol - locinput > (5 + taill) - pref_len
2290 ? (5 + taill) - pref_len : PL_regeol - locinput);
2291 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput + l)))
2295 if (pref0_len > pref_len)
2296 pref0_len = pref_len;
2297 regprop(prop, scan);
2300 do_utf8 && OP(scan) != CANY ?
2301 pv_uni_display(dsv0, (U8*)(locinput - pref_len),
2302 pref0_len, 60, UNI_DISPLAY_REGEX) :
2303 locinput - pref_len;
2304 int len0 = do_utf8 ? strlen(s0) : pref0_len;
2305 char *s1 = do_utf8 && OP(scan) != CANY ?
2306 pv_uni_display(dsv1, (U8*)(locinput - pref_len + pref0_len),
2307 pref_len - pref0_len, 60, UNI_DISPLAY_REGEX) :
2308 locinput - pref_len + pref0_len;
2309 int len1 = do_utf8 ? strlen(s1) : pref_len - pref0_len;
2310 char *s2 = do_utf8 && OP(scan) != CANY ?
2311 pv_uni_display(dsv2, (U8*)locinput,
2312 PL_regeol - locinput, 60, UNI_DISPLAY_REGEX) :
2314 int len2 = do_utf8 ? strlen(s2) : l;
2315 PerlIO_printf(Perl_debug_log,
2316 "%4"IVdf" <%s%.*s%s%s%.*s%s%s%s%.*s%s>%*s|%3"IVdf":%*s%s\n",
2317 (IV)(locinput - PL_bostr),
2324 (docolor ? "" : "> <"),
2328 15 - l - pref_len + 1,
2330 (IV)(scan - PL_regprogram), PL_regindent*2, "",
2335 next = scan + NEXT_OFF(scan);
2341 if (locinput == PL_bostr || (PL_multiline &&
2342 (nextchr || locinput < PL_regeol) && locinput[-1] == '\n') )
2344 /* regtill = regbol; */
2349 if (locinput == PL_bostr ||
2350 ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n'))
2356 if (locinput == PL_bostr)
2360 if (locinput == PL_reg_ganch)
2370 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2375 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2377 if (PL_regeol - locinput > 1)
2381 if (PL_regeol != locinput)
2385 if (!nextchr && locinput >= PL_regeol)
2388 locinput += PL_utf8skip[nextchr];
2389 if (locinput > PL_regeol)
2391 nextchr = UCHARAT(locinput);
2394 nextchr = UCHARAT(++locinput);
2397 if (!nextchr && locinput >= PL_regeol)
2399 nextchr = UCHARAT(++locinput);
2402 if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n')
2405 locinput += PL_utf8skip[nextchr];
2406 if (locinput > PL_regeol)
2408 nextchr = UCHARAT(locinput);
2411 nextchr = UCHARAT(++locinput);
2416 if (do_utf8 != UTF) {
2417 /* The target and the pattern have differing utf8ness. */
2423 /* The target is utf8, the pattern is not utf8. */
2427 if (NATIVE_TO_UNI(*(U8*)s) !=
2428 utf8n_to_uvuni((U8*)l, UTF8_MAXLEN, &ulen,
2430 0 : UTF8_ALLOW_ANY))
2437 /* The target is not utf8, the pattern is utf8. */
2441 if (NATIVE_TO_UNI(*((U8*)l)) !=
2442 utf8n_to_uvuni((U8*)s, UTF8_MAXLEN, &ulen,
2444 0 : UTF8_ALLOW_ANY))
2451 nextchr = UCHARAT(locinput);
2454 /* The target and the pattern have the same utf8ness. */
2455 /* Inline the first character, for speed. */
2456 if (UCHARAT(s) != nextchr)
2458 if (PL_regeol - locinput < ln)
2460 if (ln > 1 && memNE(s, locinput, ln))
2463 nextchr = UCHARAT(locinput);
2466 PL_reg_flags |= RF_tainted;
2472 if (do_utf8 || UTF) {
2473 /* Either target or the pattern are utf8. */
2475 char *e = PL_regeol;
2477 if (ibcmp_utf8(s, 0, ln, (bool)UTF,
2478 l, &e, 0, do_utf8)) {
2479 /* One more case for the sharp s:
2480 * pack("U0U*", 0xDF) =~ /ss/i,
2481 * the 0xC3 0x9F are the UTF-8
2482 * byte sequence for the U+00DF. */
2484 toLOWER(s[0]) == 's' &&
2486 toLOWER(s[1]) == 's' &&
2493 nextchr = UCHARAT(locinput);
2497 /* Neither the target and the pattern are utf8. */
2499 /* Inline the first character, for speed. */
2500 if (UCHARAT(s) != nextchr &&
2501 UCHARAT(s) != ((OP(scan) == EXACTF)
2502 ? PL_fold : PL_fold_locale)[nextchr])
2504 if (PL_regeol - locinput < ln)
2506 if (ln > 1 && (OP(scan) == EXACTF
2507 ? ibcmp(s, locinput, ln)
2508 : ibcmp_locale(s, locinput, ln)))
2511 nextchr = UCHARAT(locinput);
2515 STRLEN inclasslen = PL_regeol - locinput;
2517 if (!reginclass(scan, (U8*)locinput, &inclasslen, do_utf8))
2519 if (locinput >= PL_regeol)
2521 locinput += inclasslen ? inclasslen : UTF8SKIP(locinput);
2522 nextchr = UCHARAT(locinput);
2527 nextchr = UCHARAT(locinput);
2528 if (!REGINCLASS(scan, (U8*)locinput))
2530 if (!nextchr && locinput >= PL_regeol)
2532 nextchr = UCHARAT(++locinput);
2536 /* If we might have the case of the German sharp s
2537 * in a casefolding Unicode character class. */
2539 if (ANYOF_FOLD_SHARP_S(scan, locinput, PL_regeol)) {
2540 locinput += SHARP_S_SKIP;
2541 nextchr = UCHARAT(locinput);
2547 PL_reg_flags |= RF_tainted;
2553 LOAD_UTF8_CHARCLASS(alnum,"a");
2554 if (!(OP(scan) == ALNUM
2555 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2556 : isALNUM_LC_utf8((U8*)locinput)))
2560 locinput += PL_utf8skip[nextchr];
2561 nextchr = UCHARAT(locinput);
2564 if (!(OP(scan) == ALNUM
2565 ? isALNUM(nextchr) : isALNUM_LC(nextchr)))
2567 nextchr = UCHARAT(++locinput);
2570 PL_reg_flags |= RF_tainted;
2573 if (!nextchr && locinput >= PL_regeol)
2576 LOAD_UTF8_CHARCLASS(alnum,"a");
2577 if (OP(scan) == NALNUM
2578 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2579 : isALNUM_LC_utf8((U8*)locinput))
2583 locinput += PL_utf8skip[nextchr];
2584 nextchr = UCHARAT(locinput);
2587 if (OP(scan) == NALNUM
2588 ? isALNUM(nextchr) : isALNUM_LC(nextchr))
2590 nextchr = UCHARAT(++locinput);
2594 PL_reg_flags |= RF_tainted;
2598 /* was last char in word? */
2600 if (locinput == PL_bostr)
2603 U8 *r = reghop3((U8*)locinput, -1, (U8*)PL_bostr);
2605 ln = utf8n_to_uvchr(r, UTF8SKIP(r), 0, 0);
2607 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2608 ln = isALNUM_uni(ln);
2609 LOAD_UTF8_CHARCLASS(alnum,"a");
2610 n = swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8);
2613 ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(ln));
2614 n = isALNUM_LC_utf8((U8*)locinput);
2618 ln = (locinput != PL_bostr) ?
2619 UCHARAT(locinput - 1) : '\n';
2620 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2622 n = isALNUM(nextchr);
2625 ln = isALNUM_LC(ln);
2626 n = isALNUM_LC(nextchr);
2629 if (((!ln) == (!n)) == (OP(scan) == BOUND ||
2630 OP(scan) == BOUNDL))
2634 PL_reg_flags |= RF_tainted;
2640 if (UTF8_IS_CONTINUED(nextchr)) {
2641 LOAD_UTF8_CHARCLASS(space," ");
2642 if (!(OP(scan) == SPACE
2643 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2644 : isSPACE_LC_utf8((U8*)locinput)))
2648 locinput += PL_utf8skip[nextchr];
2649 nextchr = UCHARAT(locinput);
2652 if (!(OP(scan) == SPACE
2653 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2655 nextchr = UCHARAT(++locinput);
2658 if (!(OP(scan) == SPACE
2659 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2661 nextchr = UCHARAT(++locinput);
2665 PL_reg_flags |= RF_tainted;
2668 if (!nextchr && locinput >= PL_regeol)
2671 LOAD_UTF8_CHARCLASS(space," ");
2672 if (OP(scan) == NSPACE
2673 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2674 : isSPACE_LC_utf8((U8*)locinput))
2678 locinput += PL_utf8skip[nextchr];
2679 nextchr = UCHARAT(locinput);
2682 if (OP(scan) == NSPACE
2683 ? isSPACE(nextchr) : isSPACE_LC(nextchr))
2685 nextchr = UCHARAT(++locinput);
2688 PL_reg_flags |= RF_tainted;
2694 LOAD_UTF8_CHARCLASS(digit,"0");
2695 if (!(OP(scan) == DIGIT
2696 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2697 : isDIGIT_LC_utf8((U8*)locinput)))
2701 locinput += PL_utf8skip[nextchr];
2702 nextchr = UCHARAT(locinput);
2705 if (!(OP(scan) == DIGIT
2706 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr)))
2708 nextchr = UCHARAT(++locinput);
2711 PL_reg_flags |= RF_tainted;
2714 if (!nextchr && locinput >= PL_regeol)
2717 LOAD_UTF8_CHARCLASS(digit,"0");
2718 if (OP(scan) == NDIGIT
2719 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2720 : isDIGIT_LC_utf8((U8*)locinput))
2724 locinput += PL_utf8skip[nextchr];
2725 nextchr = UCHARAT(locinput);
2728 if (OP(scan) == NDIGIT
2729 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr))
2731 nextchr = UCHARAT(++locinput);
2734 if (locinput >= PL_regeol)
2737 LOAD_UTF8_CHARCLASS(mark,"~");
2738 if (swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2740 locinput += PL_utf8skip[nextchr];
2741 while (locinput < PL_regeol &&
2742 swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2743 locinput += UTF8SKIP(locinput);
2744 if (locinput > PL_regeol)
2749 nextchr = UCHARAT(locinput);
2752 PL_reg_flags |= RF_tainted;
2756 n = ARG(scan); /* which paren pair */
2757 ln = PL_regstartp[n];
2758 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2759 if ((I32)*PL_reglastparen < n || ln == -1)
2760 sayNO; /* Do not match unless seen CLOSEn. */
2761 if (ln == PL_regendp[n])
2765 if (do_utf8 && OP(scan) != REF) { /* REF can do byte comparison */
2767 char *e = PL_bostr + PL_regendp[n];
2769 * Note that we can't do the "other character" lookup trick as
2770 * in the 8-bit case (no pun intended) because in Unicode we
2771 * have to map both upper and title case to lower case.
2773 if (OP(scan) == REFF) {
2774 STRLEN ulen1, ulen2;
2775 U8 tmpbuf1[UTF8_MAXLEN_UCLC+1];
2776 U8 tmpbuf2[UTF8_MAXLEN_UCLC+1];
2780 toLOWER_utf8((U8*)s, tmpbuf1, &ulen1);
2781 toLOWER_utf8((U8*)l, tmpbuf2, &ulen2);
2782 if (ulen1 != ulen2 || memNE((char *)tmpbuf1, (char *)tmpbuf2, ulen1))
2789 nextchr = UCHARAT(locinput);
2793 /* Inline the first character, for speed. */
2794 if (UCHARAT(s) != nextchr &&
2796 (UCHARAT(s) != ((OP(scan) == REFF
2797 ? PL_fold : PL_fold_locale)[nextchr]))))
2799 ln = PL_regendp[n] - ln;
2800 if (locinput + ln > PL_regeol)
2802 if (ln > 1 && (OP(scan) == REF
2803 ? memNE(s, locinput, ln)
2805 ? ibcmp(s, locinput, ln)
2806 : ibcmp_locale(s, locinput, ln))))
2809 nextchr = UCHARAT(locinput);
2820 OP_4tree *oop = PL_op;
2821 COP *ocurcop = PL_curcop;
2826 PL_op = (OP_4tree*)PL_regdata->data[n];
2827 DEBUG_r( PerlIO_printf(Perl_debug_log, " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) );
2828 PAD_SAVE_LOCAL(old_comppad, (PAD*)PL_regdata->data[n + 2]);
2829 PL_regendp[0] = PL_reg_magic->mg_len = locinput - PL_bostr;
2833 CALLRUNOPS(aTHX); /* Scalar context. */
2836 ret = &PL_sv_undef; /* protect against empty (?{}) blocks. */
2844 PAD_RESTORE_LOCAL(old_comppad);
2845 PL_curcop = ocurcop;
2847 if (logical == 2) { /* Postponed subexpression. */
2849 MAGIC *mg = Null(MAGIC*);
2851 CHECKPOINT cp, lastcp;
2854 if(SvROK(ret) || SvRMAGICAL(ret)) {
2855 SV *sv = SvROK(ret) ? SvRV(ret) : ret;
2858 mg = mg_find(sv, PERL_MAGIC_qr);
2861 re = (regexp *)mg->mg_obj;
2862 (void)ReREFCNT_inc(re);
2866 char *t = SvPV(ret, len);
2868 char *oprecomp = PL_regprecomp;
2869 I32 osize = PL_regsize;
2870 I32 onpar = PL_regnpar;
2873 if (DO_UTF8(ret)) pm.op_pmdynflags |= PMdf_DYN_UTF8;
2874 re = CALLREGCOMP(aTHX_ t, t + len, &pm);
2876 & (SVs_TEMP | SVs_PADTMP | SVf_READONLY)))
2877 sv_magic(ret,(SV*)ReREFCNT_inc(re),
2879 PL_regprecomp = oprecomp;
2884 PerlIO_printf(Perl_debug_log,
2885 "Entering embedded `%s%.60s%s%s'\n",
2889 (strlen(re->precomp) > 60 ? "..." : ""))
2892 state.prev = PL_reg_call_cc;
2893 state.cc = PL_regcc;
2894 state.re = PL_reg_re;
2898 cp = regcppush(0); /* Save *all* the positions. */
2901 state.ss = PL_savestack_ix;
2902 *PL_reglastparen = 0;
2903 *PL_reglastcloseparen = 0;
2904 PL_reg_call_cc = &state;
2905 PL_reginput = locinput;
2906 toggleutf = ((PL_reg_flags & RF_utf8) != 0) ^
2907 ((re->reganch & ROPT_UTF8) != 0);
2908 if (toggleutf) PL_reg_flags ^= RF_utf8;
2910 /* XXXX This is too dramatic a measure... */
2913 if (regmatch(re->program + 1)) {
2914 /* Even though we succeeded, we need to restore
2915 global variables, since we may be wrapped inside
2916 SUSPEND, thus the match may be not finished yet. */
2918 /* XXXX Do this only if SUSPENDed? */
2919 PL_reg_call_cc = state.prev;
2920 PL_regcc = state.cc;
2921 PL_reg_re = state.re;
2922 cache_re(PL_reg_re);
2923 if (toggleutf) PL_reg_flags ^= RF_utf8;
2925 /* XXXX This is too dramatic a measure... */
2928 /* These are needed even if not SUSPEND. */
2934 REGCP_UNWIND(lastcp);
2936 PL_reg_call_cc = state.prev;
2937 PL_regcc = state.cc;
2938 PL_reg_re = state.re;
2939 cache_re(PL_reg_re);
2940 if (toggleutf) PL_reg_flags ^= RF_utf8;
2942 /* XXXX This is too dramatic a measure... */
2952 sv_setsv(save_scalar(PL_replgv), ret);
2956 n = ARG(scan); /* which paren pair */
2957 PL_reg_start_tmp[n] = locinput;
2962 n = ARG(scan); /* which paren pair */
2963 PL_regstartp[n] = PL_reg_start_tmp[n] - PL_bostr;
2964 PL_regendp[n] = locinput - PL_bostr;
2965 if (n > (I32)*PL_reglastparen)
2966 *PL_reglastparen = n;
2967 *PL_reglastcloseparen = n;
2970 n = ARG(scan); /* which paren pair */
2971 sw = ((I32)*PL_reglastparen >= n && PL_regendp[n] != -1);
2974 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2976 next = NEXTOPER(NEXTOPER(scan));
2978 next = scan + ARG(scan);
2979 if (OP(next) == IFTHEN) /* Fake one. */
2980 next = NEXTOPER(NEXTOPER(next));
2984 logical = scan->flags;
2986 /*******************************************************************
2987 PL_regcc contains infoblock about the innermost (...)* loop, and
2988 a pointer to the next outer infoblock.
2990 Here is how Y(A)*Z is processed (if it is compiled into CURLYX/WHILEM):
2992 1) After matching X, regnode for CURLYX is processed;
2994 2) This regnode creates infoblock on the stack, and calls
2995 regmatch() recursively with the starting point at WHILEM node;
2997 3) Each hit of WHILEM node tries to match A and Z (in the order
2998 depending on the current iteration, min/max of {min,max} and
2999 greediness). The information about where are nodes for "A"
3000 and "Z" is read from the infoblock, as is info on how many times "A"
3001 was already matched, and greediness.
3003 4) After A matches, the same WHILEM node is hit again.
3005 5) Each time WHILEM is hit, PL_regcc is the infoblock created by CURLYX
3006 of the same pair. Thus when WHILEM tries to match Z, it temporarily
3007 resets PL_regcc, since this Y(A)*Z can be a part of some other loop:
3008 as in (Y(A)*Z)*. If Z matches, the automaton will hit the WHILEM node
3009 of the external loop.
3011 Currently present infoblocks form a tree with a stem formed by PL_curcc
3012 and whatever it mentions via ->next, and additional attached trees
3013 corresponding to temporarily unset infoblocks as in "5" above.
3015 In the following picture infoblocks for outer loop of
3016 (Y(A)*?Z)*?T are denoted O, for inner I. NULL starting block
3017 is denoted by x. The matched string is YAAZYAZT. Temporarily postponed
3018 infoblocks are drawn below the "reset" infoblock.
3020 In fact in the picture below we do not show failed matches for Z and T
3021 by WHILEM blocks. [We illustrate minimal matches, since for them it is
3022 more obvious *why* one needs to *temporary* unset infoblocks.]
3024 Matched REx position InfoBlocks Comment
3028 Y A)*?Z)*?T x <- O <- I
3029 YA )*?Z)*?T x <- O <- I
3030 YA A)*?Z)*?T x <- O <- I
3031 YAA )*?Z)*?T x <- O <- I
3032 YAA Z)*?T x <- O # Temporary unset I
3035 YAAZ Y(A)*?Z)*?T x <- O
3038 YAAZY (A)*?Z)*?T x <- O
3041 YAAZY A)*?Z)*?T x <- O <- I
3044 YAAZYA )*?Z)*?T x <- O <- I
3047 YAAZYA Z)*?T x <- O # Temporary unset I
3053 YAAZYAZ T x # Temporary unset O
3060 *******************************************************************/
3063 CHECKPOINT cp = PL_savestack_ix;
3064 /* No need to save/restore up to this paren */
3065 I32 parenfloor = scan->flags;
3067 if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */
3069 cc.oldcc = PL_regcc;
3071 /* XXXX Probably it is better to teach regpush to support
3072 parenfloor > PL_regsize... */
3073 if (parenfloor > (I32)*PL_reglastparen)
3074 parenfloor = *PL_reglastparen; /* Pessimization... */
3075 cc.parenfloor = parenfloor;
3077 cc.min = ARG1(scan);
3078 cc.max = ARG2(scan);
3079 cc.scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
3083 PL_reginput = locinput;
3084 n = regmatch(PREVOPER(next)); /* start on the WHILEM */
3086 PL_regcc = cc.oldcc;
3092 * This is really hard to understand, because after we match
3093 * what we're trying to match, we must make sure the rest of
3094 * the REx is going to match for sure, and to do that we have
3095 * to go back UP the parse tree by recursing ever deeper. And
3096 * if it fails, we have to reset our parent's current state
3097 * that we can try again after backing off.
3100 CHECKPOINT cp, lastcp;
3101 CURCUR* cc = PL_regcc;
3102 char *lastloc = cc->lastloc; /* Detection of 0-len. */
3104 n = cc->cur + 1; /* how many we know we matched */
3105 PL_reginput = locinput;
3108 PerlIO_printf(Perl_debug_log,
3109 "%*s %ld out of %ld..%ld cc=%"UVxf"\n",
3110 REPORT_CODE_OFF+PL_regindent*2, "",
3111 (long)n, (long)cc->min,
3112 (long)cc->max, PTR2UV(cc))
3115 /* If degenerate scan matches "", assume scan done. */
3117 if (locinput == cc->lastloc && n >= cc->min) {
3118 PL_regcc = cc->oldcc;
3122 PerlIO_printf(Perl_debug_log,
3123 "%*s empty match detected, try continuation...\n",
3124 REPORT_CODE_OFF+PL_regindent*2, "")
3126 if (regmatch(cc->next))
3134 /* First just match a string of min scans. */
3138 cc->lastloc = locinput;
3139 if (regmatch(cc->scan))
3142 cc->lastloc = lastloc;
3147 /* Check whether we already were at this position.
3148 Postpone detection until we know the match is not
3149 *that* much linear. */
3150 if (!PL_reg_maxiter) {
3151 PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4);
3152 PL_reg_leftiter = PL_reg_maxiter;
3154 if (PL_reg_leftiter-- == 0) {
3155 I32 size = (PL_reg_maxiter + 7)/8;
3156 if (PL_reg_poscache) {
3157 if ((I32)PL_reg_poscache_size < size) {
3158 Renew(PL_reg_poscache, size, char);
3159 PL_reg_poscache_size = size;
3161 Zero(PL_reg_poscache, size, char);
3164 PL_reg_poscache_size = size;
3165 Newz(29, PL_reg_poscache, size, char);
3168 PerlIO_printf(Perl_debug_log,
3169 "%sDetected a super-linear match, switching on caching%s...\n",
3170 PL_colors[4], PL_colors[5])
3173 if (PL_reg_leftiter < 0) {
3174 I32 o = locinput - PL_bostr, b;
3176 o = (scan->flags & 0xf) - 1 + o * (scan->flags>>4);
3179 if (PL_reg_poscache[o] & (1<<b)) {
3181 PerlIO_printf(Perl_debug_log,
3182 "%*s already tried at this position...\n",
3183 REPORT_CODE_OFF+PL_regindent*2, "")
3187 PL_reg_poscache[o] |= (1<<b);
3191 /* Prefer next over scan for minimal matching. */
3194 PL_regcc = cc->oldcc;
3197 cp = regcppush(cc->parenfloor);
3199 if (regmatch(cc->next)) {
3201 sayYES; /* All done. */
3203 REGCP_UNWIND(lastcp);
3209 if (n >= cc->max) { /* Maximum greed exceeded? */
3210 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3211 && !(PL_reg_flags & RF_warned)) {
3212 PL_reg_flags |= RF_warned;
3213 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s limit (%d) exceeded",
3214 "Complex regular subexpression recursion",
3221 PerlIO_printf(Perl_debug_log,
3222 "%*s trying longer...\n",
3223 REPORT_CODE_OFF+PL_regindent*2, "")
3225 /* Try scanning more and see if it helps. */
3226 PL_reginput = locinput;
3228 cc->lastloc = locinput;
3229 cp = regcppush(cc->parenfloor);
3231 if (regmatch(cc->scan)) {
3235 REGCP_UNWIND(lastcp);
3238 cc->lastloc = lastloc;
3242 /* Prefer scan over next for maximal matching. */
3244 if (n < cc->max) { /* More greed allowed? */
3245 cp = regcppush(cc->parenfloor);
3247 cc->lastloc = locinput;
3249 if (regmatch(cc->scan)) {
3253 REGCP_UNWIND(lastcp);
3254 regcppop(); /* Restore some previous $<digit>s? */
3255 PL_reginput = locinput;
3257 PerlIO_printf(Perl_debug_log,
3258 "%*s failed, try continuation...\n",
3259 REPORT_CODE_OFF+PL_regindent*2, "")
3262 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3263 && !(PL_reg_flags & RF_warned)) {
3264 PL_reg_flags |= RF_warned;
3265 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s limit (%d) exceeded",
3266 "Complex regular subexpression recursion",
3270 /* Failed deeper matches of scan, so see if this one works. */
3271 PL_regcc = cc->oldcc;
3274 if (regmatch(cc->next))
3280 cc->lastloc = lastloc;
3285 next = scan + ARG(scan);
3288 inner = NEXTOPER(NEXTOPER(scan));
3291 inner = NEXTOPER(scan);
3295 if (OP(next) != c1) /* No choice. */
3296 next = inner; /* Avoid recursion. */
3298 I32 lastparen = *PL_reglastparen;
3300 re_unwind_branch_t *uw;
3302 /* Put unwinding data on stack */
3303 unwind1 = SSNEWt(1,re_unwind_branch_t);
3304 uw = SSPTRt(unwind1,re_unwind_branch_t);
3307 uw->type = ((c1 == BRANCH)
3309 : RE_UNWIND_BRANCHJ);
3310 uw->lastparen = lastparen;
3312 uw->locinput = locinput;
3313 uw->nextchr = nextchr;
3315 uw->regindent = ++PL_regindent;
3318 REGCP_SET(uw->lastcp);
3320 /* Now go into the first branch */
3333 /* We suppose that the next guy does not need
3334 backtracking: in particular, it is of constant length,
3335 and has no parenths to influence future backrefs. */
3336 ln = ARG1(scan); /* min to match */
3337 n = ARG2(scan); /* max to match */
3338 paren = scan->flags;
3340 if (paren > PL_regsize)
3342 if (paren > (I32)*PL_reglastparen)
3343 *PL_reglastparen = paren;
3345 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3347 scan += NEXT_OFF(scan); /* Skip former OPEN. */
3348 PL_reginput = locinput;
3351 if (ln && regrepeat_hard(scan, ln, &l) < ln)
3353 /* if we matched something zero-length we don't need to
3354 backtrack - capturing parens are already defined, so
3355 the caveat in the maximal case doesn't apply
3357 XXXX if ln == 0, we can redo this check first time
3358 through the following loop
3361 n = ln; /* don't backtrack */
3362 locinput = PL_reginput;
3363 if (HAS_TEXT(next) || JUMPABLE(next)) {
3364 regnode *text_node = next;
3366 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3368 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3370 if (PL_regkind[(U8)OP(text_node)] == REF) {
3372 n = ARG(text_node); /* which paren pair */
3373 ln = PL_regstartp[n];
3374 /* assume yes if we haven't seen CLOSEn */
3376 (I32)*PL_reglastparen < n ||
3383 c1 = *(PL_bostr + ln);
3385 else { c1 = (U8)*STRING(text_node); }
3386 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3388 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3389 c2 = PL_fold_locale[c1];
3398 /* This may be improved if l == 0. */
3399 while (n >= ln || (n == REG_INFTY && ln > 0 && l)) { /* ln overflow ? */
3400 /* If it could work, try it. */
3402 UCHARAT(PL_reginput) == c1 ||
3403 UCHARAT(PL_reginput) == c2)
3407 PL_regstartp[paren] =
3408 HOPc(PL_reginput, -l) - PL_bostr;
3409 PL_regendp[paren] = PL_reginput - PL_bostr;
3412 PL_regendp[paren] = -1;
3416 REGCP_UNWIND(lastcp);
3418 /* Couldn't or didn't -- move forward. */
3419 PL_reginput = locinput;
3420 if (regrepeat_hard(scan, 1, &l)) {
3422 locinput = PL_reginput;
3429 n = regrepeat_hard(scan, n, &l);
3430 /* if we matched something zero-length we don't need to
3431 backtrack, unless the minimum count is zero and we
3432 are capturing the result - in that case the capture
3433 being defined or not may affect later execution
3435 if (n != 0 && l == 0 && !(paren && ln == 0))
3436 ln = n; /* don't backtrack */
3437 locinput = PL_reginput;
3439 PerlIO_printf(Perl_debug_log,
3440 "%*s matched %"IVdf" times, len=%"IVdf"...\n",
3441 (int)(REPORT_CODE_OFF+PL_regindent*2), "",
3445 if (HAS_TEXT(next) || JUMPABLE(next)) {
3446 regnode *text_node = next;
3448 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3450 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3452 if (PL_regkind[(U8)OP(text_node)] == REF) {
3454 n = ARG(text_node); /* which paren pair */
3455 ln = PL_regstartp[n];
3456 /* assume yes if we haven't seen CLOSEn */
3458 (I32)*PL_reglastparen < n ||
3465 c1 = *(PL_bostr + ln);
3467 else { c1 = (U8)*STRING(text_node); }
3469 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3471 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3472 c2 = PL_fold_locale[c1];
3483 /* If it could work, try it. */
3485 UCHARAT(PL_reginput) == c1 ||
3486 UCHARAT(PL_reginput) == c2)
3489 PerlIO_printf(Perl_debug_log,
3490 "%*s trying tail with n=%"IVdf"...\n",
3491 (int)(REPORT_CODE_OFF+PL_regindent*2), "", (IV)n)
3495 PL_regstartp[paren] = HOPc(PL_reginput, -l) - PL_bostr;
3496 PL_regendp[paren] = PL_reginput - PL_bostr;
3499 PL_regendp[paren] = -1;
3503 REGCP_UNWIND(lastcp);
3505 /* Couldn't or didn't -- back up. */
3507 locinput = HOPc(locinput, -l);
3508 PL_reginput = locinput;
3515 paren = scan->flags; /* Which paren to set */
3516 if (paren > PL_regsize)
3518 if (paren > (I32)*PL_reglastparen)
3519 *PL_reglastparen = paren;
3520 ln = ARG1(scan); /* min to match */
3521 n = ARG2(scan); /* max to match */
3522 scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE);
3526 ln = ARG1(scan); /* min to match */
3527 n = ARG2(scan); /* max to match */
3528 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3533 scan = NEXTOPER(scan);
3539 scan = NEXTOPER(scan);
3543 * Lookahead to avoid useless match attempts
3544 * when we know what character comes next.
3548 * Used to only do .*x and .*?x, but now it allows
3549 * for )'s, ('s and (?{ ... })'s to be in the way
3550 * of the quantifier and the EXACT-like node. -- japhy
3553 if (HAS_TEXT(next) || JUMPABLE(next)) {
3555 regnode *text_node = next;
3557 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3559 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3561 if (PL_regkind[(U8)OP(text_node)] == REF) {
3563 n = ARG(text_node); /* which paren pair */
3564 ln = PL_regstartp[n];
3565 /* assume yes if we haven't seen CLOSEn */
3567 (I32)*PL_reglastparen < n ||
3572 goto assume_ok_easy;
3574 s = (U8*)PL_bostr + ln;
3576 else { s = (U8*)STRING(text_node); }
3580 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3582 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3583 c2 = PL_fold_locale[c1];
3586 if (OP(text_node) == EXACTF || OP(text_node) == REFF) {
3587 STRLEN ulen1, ulen2;
3588 U8 tmpbuf1[UTF8_MAXLEN_UCLC+1];
3589 U8 tmpbuf2[UTF8_MAXLEN_UCLC+1];
3591 to_utf8_lower((U8*)s, tmpbuf1, &ulen1);
3592 to_utf8_upper((U8*)s, tmpbuf2, &ulen2);
3594 c1 = utf8n_to_uvuni(tmpbuf1, UTF8_MAXLEN, 0,
3596 0 : UTF8_ALLOW_ANY);
3597 c2 = utf8n_to_uvuni(tmpbuf2, UTF8_MAXLEN, 0,
3599 0 : UTF8_ALLOW_ANY);
3602 c2 = c1 = utf8n_to_uvchr(s, UTF8_MAXLEN, 0,
3604 0 : UTF8_ALLOW_ANY);
3612 PL_reginput = locinput;
3616 if (ln && regrepeat(scan, ln) < ln)
3618 locinput = PL_reginput;
3621 char *e; /* Should not check after this */
3622 char *old = locinput;
3625 if (n == REG_INFTY) {
3628 while (UTF8_IS_CONTINUATION(*(U8*)e))
3634 m >0 && e + UTF8SKIP(e) <= PL_regeol; m--)
3638 e = locinput + n - ln;
3643 /* Find place 'next' could work */
3646 while (locinput <= e &&
3647 UCHARAT(locinput) != c1)
3650 while (locinput <= e
3651 && UCHARAT(locinput) != c1
3652 && UCHARAT(locinput) != c2)
3655 count = locinput - old;
3660 /* count initialised to
3661 * utf8_distance(old, locinput) */
3662 while (locinput <= e &&
3663 utf8n_to_uvchr((U8*)locinput,
3666 0 : UTF8_ALLOW_ANY) != (UV)c1) {
3671 /* count initialised to
3672 * utf8_distance(old, locinput) */
3673 while (locinput <= e) {
3674 UV c = utf8n_to_uvchr((U8*)locinput,
3677 0 : UTF8_ALLOW_ANY);
3678 if (c == (UV)c1 || c == (UV)c2)
3687 /* PL_reginput == old now */
3688 if (locinput != old) {
3689 ln = 1; /* Did some */
3690 if (regrepeat(scan, count) < count)
3693 /* PL_reginput == locinput now */
3694 TRYPAREN(paren, ln, locinput);
3695 PL_reginput = locinput; /* Could be reset... */
3696 REGCP_UNWIND(lastcp);
3697 /* Couldn't or didn't -- move forward. */
3700 locinput += UTF8SKIP(locinput);
3707 while (n >= ln || (n == REG_INFTY && ln > 0)) { /* ln overflow ? */
3711 c = utf8n_to_uvchr((U8*)PL_reginput,
3714 0 : UTF8_ALLOW_ANY);
3716 c = UCHARAT(PL_reginput);
3717 /* If it could work, try it. */
3718 if (c == (UV)c1 || c == (UV)c2)
3720 TRYPAREN(paren, ln, PL_reginput);
3721 REGCP_UNWIND(lastcp);
3724 /* If it could work, try it. */
3725 else if (c1 == -1000)
3727 TRYPAREN(paren, ln, PL_reginput);
3728 REGCP_UNWIND(lastcp);
3730 /* Couldn't or didn't -- move forward. */
3731 PL_reginput = locinput;
3732 if (regrepeat(scan, 1)) {
3734 locinput = PL_reginput;
3742 n = regrepeat(scan, n);
3743 locinput = PL_reginput;
3744 if (ln < n && PL_regkind[(U8)OP(next)] == EOL &&
3745 ((!PL_multiline && OP(next) != MEOL) ||
3746 OP(next) == SEOL || OP(next) == EOS))
3748 ln = n; /* why back off? */
3749 /* ...because $ and \Z can match before *and* after
3750 newline at the end. Consider "\n\n" =~ /\n+\Z\n/.
3751 We should back off by one in this case. */
3752 if (UCHARAT(PL_reginput - 1) == '\n' && OP(next) != EOS)
3761 c = utf8n_to_uvchr((U8*)PL_reginput,
3764 0 : UTF8_ALLOW_ANY);
3766 c = UCHARAT(PL_reginput);
3768 /* If it could work, try it. */
3769 if (c1 == -1000 || c == (UV)c1 || c == (UV)c2)
3771 TRYPAREN(paren, n, PL_reginput);
3772 REGCP_UNWIND(lastcp);
3774 /* Couldn't or didn't -- back up. */
3776 PL_reginput = locinput = HOPc(locinput, -1);
3784 c = utf8n_to_uvchr((U8*)PL_reginput,
3787 0 : UTF8_ALLOW_ANY);
3789 c = UCHARAT(PL_reginput);
3791 /* If it could work, try it. */
3792 if (c1 == -1000 || c == (UV)c1 || c == (UV)c2)
3794 TRYPAREN(paren, n, PL_reginput);
3795 REGCP_UNWIND(lastcp);
3797 /* Couldn't or didn't -- back up. */
3799 PL_reginput = locinput = HOPc(locinput, -1);
3806 if (PL_reg_call_cc) {
3807 re_cc_state *cur_call_cc = PL_reg_call_cc;
3808 CURCUR *cctmp = PL_regcc;
3809 regexp *re = PL_reg_re;
3810 CHECKPOINT cp, lastcp;
3812 cp = regcppush(0); /* Save *all* the positions. */
3814 regcp_set_to(PL_reg_call_cc->ss); /* Restore parens of
3816 PL_reginput = locinput; /* Make position available to
3818 cache_re(PL_reg_call_cc->re);
3819 PL_regcc = PL_reg_call_cc->cc;
3820 PL_reg_call_cc = PL_reg_call_cc->prev;
3821 if (regmatch(cur_call_cc->node)) {
3822 PL_reg_call_cc = cur_call_cc;
3826 REGCP_UNWIND(lastcp);
3828 PL_reg_call_cc = cur_call_cc;
3834 PerlIO_printf(Perl_debug_log,
3835 "%*s continuation failed...\n",
3836 REPORT_CODE_OFF+PL_regindent*2, "")
3840 if (locinput < PL_regtill) {
3841 DEBUG_r(PerlIO_printf(Perl_debug_log,
3842 "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n",
3844 (long)(locinput - PL_reg_starttry),
3845 (long)(PL_regtill - PL_reg_starttry),
3847 sayNO_FINAL; /* Cannot match: too short. */
3849 PL_reginput = locinput; /* put where regtry can find it */
3850 sayYES_FINAL; /* Success! */
3852 PL_reginput = locinput; /* put where regtry can find it */
3853 sayYES_LOUD; /* Success! */
3856 PL_reginput = locinput;
3861 s = HOPBACKc(locinput, scan->flags);
3867 PL_reginput = locinput;
3872 s = HOPBACKc(locinput, scan->flags);
3878 PL_reginput = locinput;
3881 inner = NEXTOPER(NEXTOPER(scan));
3882 if (regmatch(inner) != n) {
3897 if (OP(scan) == SUSPEND) {
3898 locinput = PL_reginput;
3899 nextchr = UCHARAT(locinput);
3904 next = scan + ARG(scan);
3909 PerlIO_printf(Perl_error_log, "%"UVxf" %d\n",
3910 PTR2UV(scan), OP(scan));
3911 Perl_croak(aTHX_ "regexp memory corruption");
3918 * We get here only if there's trouble -- normally "case END" is
3919 * the terminating point.
3921 Perl_croak(aTHX_ "corrupted regexp pointers");
3927 PerlIO_printf(Perl_debug_log,
3928 "%*s %scould match...%s\n",
3929 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4],PL_colors[5])
3933 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n",
3934 PL_colors[4],PL_colors[5]));
3940 #if 0 /* Breaks $^R */
3948 PerlIO_printf(Perl_debug_log,
3949 "%*s %sfailed...%s\n",
3950 REPORT_CODE_OFF+PL_regindent*2, "",PL_colors[4],PL_colors[5])
3956 re_unwind_t *uw = SSPTRt(unwind,re_unwind_t);
3959 case RE_UNWIND_BRANCH:
3960 case RE_UNWIND_BRANCHJ:
3962 re_unwind_branch_t *uwb = &(uw->branch);
3963 I32 lastparen = uwb->lastparen;
3965 REGCP_UNWIND(uwb->lastcp);
3966 for (n = *PL_reglastparen; n > lastparen; n--)
3968 *PL_reglastparen = n;
3969 scan = next = uwb->next;
3971 OP(scan) != (uwb->type == RE_UNWIND_BRANCH
3972 ? BRANCH : BRANCHJ) ) { /* Failure */
3979 /* Have more choice yet. Reuse the same uwb. */
3981 if ((n = (uwb->type == RE_UNWIND_BRANCH
3982 ? NEXT_OFF(next) : ARG(next))))
3985 next = NULL; /* XXXX Needn't unwinding in this case... */
3987 next = NEXTOPER(scan);
3988 if (uwb->type == RE_UNWIND_BRANCHJ)
3989 next = NEXTOPER(next);
3990 locinput = uwb->locinput;
3991 nextchr = uwb->nextchr;
3993 PL_regindent = uwb->regindent;
4000 Perl_croak(aTHX_ "regexp unwind memory corruption");
4011 - regrepeat - repeatedly match something simple, report how many
4014 * [This routine now assumes that it will only match on things of length 1.
4015 * That was true before, but now we assume scan - reginput is the count,
4016 * rather than incrementing count on every character. [Er, except utf8.]]
4019 S_regrepeat(pTHX_ regnode *p, I32 max)
4021 register char *scan;
4023 register char *loceol = PL_regeol;
4024 register I32 hardcount = 0;
4025 register bool do_utf8 = PL_reg_match_utf8;
4028 if (max == REG_INFTY)
4030 else if (max < loceol - scan)
4031 loceol = scan + max;
4036 while (scan < loceol && hardcount < max && *scan != '\n') {
4037 scan += UTF8SKIP(scan);
4041 while (scan < loceol && *scan != '\n')
4048 while (scan < loceol && hardcount < max) {
4049 scan += UTF8SKIP(scan);
4059 case EXACT: /* length of string is 1 */
4061 while (scan < loceol && UCHARAT(scan) == c)
4064 case EXACTF: /* length of string is 1 */
4066 while (scan < loceol &&
4067 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold[c]))
4070 case EXACTFL: /* length of string is 1 */
4071 PL_reg_flags |= RF_tainted;
4073 while (scan < loceol &&
4074 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold_locale[c]))
4080 while (hardcount < max && scan < loceol) {
4082 if (ANYOF_FLAGS(p) & ANYOF_UNICODE) {
4083 if (reginclass(p, (U8*)scan, 0, do_utf8))
4089 if (UTF8_IS_INVARIANT(c)) {
4090 if (ANYOF_BITMAP_TEST(p, c))
4094 if (reginclass(p, (U8*)scan, 0, do_utf8))
4100 scan += UTF8SKIP(scan);
4104 while (scan < loceol && REGINCLASS(p, (U8*)scan))
4111 LOAD_UTF8_CHARCLASS(alnum,"a");
4112 while (hardcount < max && scan < loceol &&
4113 swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
4114 scan += UTF8SKIP(scan);
4118 while (scan < loceol && isALNUM(*scan))
4123 PL_reg_flags |= RF_tainted;
4126 while (hardcount < max && scan < loceol &&
4127 isALNUM_LC_utf8((U8*)scan)) {
4128 scan += UTF8SKIP(scan);
4132 while (scan < loceol && isALNUM_LC(*scan))
4139 LOAD_UTF8_CHARCLASS(alnum,"a");
4140 while (hardcount < max && scan < loceol &&
4141 !swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
4142 scan += UTF8SKIP(scan);
4146 while (scan < loceol && !isALNUM(*scan))
4151 PL_reg_flags |= RF_tainted;
4154 while (hardcount < max && scan < loceol &&
4155 !isALNUM_LC_utf8((U8*)scan)) {
4156 scan += UTF8SKIP(scan);
4160 while (scan < loceol && !isALNUM_LC(*scan))
4167 LOAD_UTF8_CHARCLASS(space," ");
4168 while (hardcount < max && scan < loceol &&
4170 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
4171 scan += UTF8SKIP(scan);
4175 while (scan < loceol && isSPACE(*scan))
4180 PL_reg_flags |= RF_tainted;
4183 while (hardcount < max && scan < loceol &&
4184 (*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
4185 scan += UTF8SKIP(scan);
4189 while (scan < loceol && isSPACE_LC(*scan))
4196 LOAD_UTF8_CHARCLASS(space," ");
4197 while (hardcount < max && scan < loceol &&
4199 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
4200 scan += UTF8SKIP(scan);
4204 while (scan < loceol && !isSPACE(*scan))
4209 PL_reg_flags |= RF_tainted;
4212 while (hardcount < max && scan < loceol &&
4213 !(*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
4214 scan += UTF8SKIP(scan);
4218 while (scan < loceol && !isSPACE_LC(*scan))
4225 LOAD_UTF8_CHARCLASS(digit,"0");
4226 while (hardcount < max && scan < loceol &&
4227 swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
4228 scan += UTF8SKIP(scan);
4232 while (scan < loceol && isDIGIT(*scan))
4239 LOAD_UTF8_CHARCLASS(digit,"0");
4240 while (hardcount < max && scan < loceol &&
4241 !swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
4242 scan += UTF8SKIP(scan);
4246 while (scan < loceol && !isDIGIT(*scan))
4250 default: /* Called on something of 0 width. */
4251 break; /* So match right here or not at all. */
4257 c = scan - PL_reginput;
4262 SV *prop = sv_newmortal();
4265 PerlIO_printf(Perl_debug_log,
4266 "%*s %s can match %"IVdf" times out of %"IVdf"...\n",
4267 REPORT_CODE_OFF+1, "", SvPVX(prop),(IV)c,(IV)max);
4274 - regrepeat_hard - repeatedly match something, report total lenth and length
4276 * The repeater is supposed to have constant length.
4280 S_regrepeat_hard(pTHX_ regnode *p, I32 max, I32 *lp)
4282 register char *scan = Nullch;
4283 register char *start;
4284 register char *loceol = PL_regeol;
4286 I32 count = 0, res = 1;
4291 start = PL_reginput;
4292 if (PL_reg_match_utf8) {
4293 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
4296 while (start < PL_reginput) {
4298 start += UTF8SKIP(start);
4309 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
4311 *lp = l = PL_reginput - start;
4312 if (max != REG_INFTY && l*max < loceol - scan)
4313 loceol = scan + l*max;
4326 - regclass_swash - prepare the utf8 swash
4330 Perl_regclass_swash(pTHX_ register regnode* node, bool doinit, SV** listsvp, SV **altsvp)
4336 if (PL_regdata && PL_regdata->count) {
4339 if (PL_regdata->what[n] == 's') {
4340 SV *rv = (SV*)PL_regdata->data[n];
4341 AV *av = (AV*)SvRV((SV*)rv);
4342 SV **ary = AvARRAY(av);
4345 /* See the end of regcomp.c:S_reglass() for
4346 * documentation of these array elements. */
4349 a = SvTYPE(ary[1]) == SVt_RV ? &ary[1] : 0;
4350 b = SvTYPE(ary[2]) == SVt_PVAV ? &ary[2] : 0;
4354 else if (si && doinit) {
4355 sw = swash_init("utf8", "", si, 1, 0);
4356 (void)av_store(av, 1, sw);
4372 - reginclass - determine if a character falls into a character class
4374 The n is the ANYOF regnode, the p is the target string, lenp
4375 is pointer to the maximum length of how far to go in the p
4376 (if the lenp is zero, UTF8SKIP(p) is used),
4377 do_utf8 tells whether the target string is in UTF-8.
4382 S_reginclass(pTHX_ register regnode *n, register U8* p, STRLEN* lenp, register bool do_utf8)
4384 char flags = ANYOF_FLAGS(n);
4390 if (do_utf8 && !UTF8_IS_INVARIANT(c))
4391 c = utf8n_to_uvchr(p, UTF8_MAXLEN, &len,
4392 ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
4394 plen = lenp ? *lenp : UNISKIP(NATIVE_TO_UNI(c));
4395 if (do_utf8 || (flags & ANYOF_UNICODE)) {
4398 if (do_utf8 && !ANYOF_RUNTIME(n)) {
4399 if (len != (STRLEN)-1 && c < 256 && ANYOF_BITMAP_TEST(n, c))
4402 if (!match && do_utf8 && (flags & ANYOF_UNICODE_ALL) && c >= 256)
4406 SV *sw = regclass_swash(n, TRUE, 0, (SV**)&av);
4409 if (swash_fetch(sw, p, do_utf8))
4411 else if (flags & ANYOF_FOLD) {
4412 if (!match && lenp && av) {
4415 for (i = 0; i <= av_len(av); i++) {
4416 SV* sv = *av_fetch(av, i, FALSE);
4418 char *s = SvPV(sv, len);
4420 if (len <= plen && memEQ(s, (char*)p, len)) {
4428 U8 tmpbuf[UTF8_MAXLEN_FOLD+1];
4431 to_utf8_fold(p, tmpbuf, &tmplen);
4432 if (swash_fetch(sw, tmpbuf, do_utf8))
4438 if (match && lenp && *lenp == 0)
4439 *lenp = UNISKIP(NATIVE_TO_UNI(c));
4441 if (!match && c < 256) {
4442 if (ANYOF_BITMAP_TEST(n, c))
4444 else if (flags & ANYOF_FOLD) {
4447 if (flags & ANYOF_LOCALE) {
4448 PL_reg_flags |= RF_tainted;
4449 f = PL_fold_locale[c];
4453 if (f != c && ANYOF_BITMAP_TEST(n, f))
4457 if (!match && (flags & ANYOF_CLASS)) {
4458 PL_reg_flags |= RF_tainted;
4460 (ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) ||
4461 (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) ||
4462 (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) ||
4463 (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) ||
4464 (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) ||
4465 (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) ||
4466 (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) ||
4467 (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) ||
4468 (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) ||
4469 (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) ||
4470 (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII(c)) ||
4471 (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII(c)) ||
4472 (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) ||
4473 (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) ||
4474 (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) ||
4475 (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) ||
4476 (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) ||
4477 (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) ||
4478 (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) ||
4479 (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) ||
4480 (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) ||
4481 (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) ||
4482 (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) ||
4483 (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) ||
4484 (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) ||
4485 (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) ||
4486 (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) ||
4487 (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) ||
4488 (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK(c)) ||
4489 (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK(c))
4490 ) /* How's that for a conditional? */
4497 return (flags & ANYOF_INVERT) ? !match : match;
4501 S_reghop(pTHX_ U8 *s, I32 off)
4503 return S_reghop3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4507 S_reghop3(pTHX_ U8 *s, I32 off, U8* lim)
4510 while (off-- && s < lim) {
4511 /* XXX could check well-formedness here */
4519 if (UTF8_IS_CONTINUED(*s)) {
4520 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4523 /* XXX could check well-formedness here */
4531 S_reghopmaybe(pTHX_ U8 *s, I32 off)
4533 return S_reghopmaybe3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4537 S_reghopmaybe3(pTHX_ U8* s, I32 off, U8* lim)
4540 while (off-- && s < lim) {
4541 /* XXX could check well-formedness here */
4551 if (UTF8_IS_CONTINUED(*s)) {
4552 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4555 /* XXX could check well-formedness here */
4567 restore_pos(pTHX_ void *arg)
4569 if (PL_reg_eval_set) {
4570 if (PL_reg_oldsaved) {
4571 PL_reg_re->subbeg = PL_reg_oldsaved;
4572 PL_reg_re->sublen = PL_reg_oldsavedlen;
4573 RX_MATCH_COPIED_on(PL_reg_re);
4575 PL_reg_magic->mg_len = PL_reg_oldpos;
4576 PL_reg_eval_set = 0;
4577 PL_curpm = PL_reg_oldcurpm;
4582 S_to_utf8_substr(pTHX_ register regexp *prog)
4585 if (prog->float_substr && !prog->float_utf8) {
4586 prog->float_utf8 = sv = NEWSV(117, 0);
4587 SvSetSV(sv, prog->float_substr);
4588 sv_utf8_upgrade(sv);
4589 if (SvTAIL(prog->float_substr))
4591 if (prog->float_substr == prog->check_substr)
4592 prog->check_utf8 = sv;
4594 if (prog->anchored_substr && !prog->anchored_utf8) {
4595 prog->anchored_utf8 = sv = NEWSV(118, 0);
4596 SvSetSV(sv, prog->anchored_substr);
4597 sv_utf8_upgrade(sv);
4598 if (SvTAIL(prog->anchored_substr))
4600 if (prog->anchored_substr == prog->check_substr)
4601 prog->check_utf8 = sv;
4606 S_to_byte_substr(pTHX_ register regexp *prog)
4609 if (prog->float_utf8 && !prog->float_substr) {
4610 prog->float_substr = sv = NEWSV(117, 0);
4611 SvSetSV(sv, prog->float_utf8);
4612 if (sv_utf8_downgrade(sv, TRUE)) {
4613 if (SvTAIL(prog->float_utf8))
4617 prog->float_substr = sv = &PL_sv_undef;
4619 if (prog->float_utf8 == prog->check_utf8)
4620 prog->check_substr = sv;
4622 if (prog->anchored_utf8 && !prog->anchored_substr) {
4623 prog->anchored_substr = sv = NEWSV(118, 0);
4624 SvSetSV(sv, prog->anchored_utf8);
4625 if (sv_utf8_downgrade(sv, TRUE)) {
4626 if (SvTAIL(prog->anchored_utf8))
4630 prog->anchored_substr = sv = &PL_sv_undef;
4632 if (prog->anchored_utf8 == prog->check_utf8)
4633 prog->check_substr = sv;