5 * "One Ring to rule them all, One Ring to find them..."
8 /* This file contains functions for executing a regular expression. See
9 * also regcomp.c which funnily enough, contains functions for compiling
10 * a regular expression.
12 * This file is also copied at build time to ext/re/re_exec.c, where
13 * it's built with -DPERL_EXT_RE_BUILD -DPERL_EXT_RE_DEBUG -DPERL_EXT.
14 * This causes the main functions to be compiled under new names and with
15 * debugging support added, which makes "use re 'debug'" work.
19 /* NOTE: this is derived from Henry Spencer's regexp code, and should not
20 * confused with the original package (see point 3 below). Thanks, Henry!
23 /* Additional note: this code is very heavily munged from Henry's version
24 * in places. In some spots I've traded clarity for efficiency, so don't
25 * blame Henry for some of the lack of readability.
28 /* The names of the functions have been changed from regcomp and
29 * regexec to pregcomp and pregexec in order to avoid conflicts
30 * with the POSIX routines of the same names.
33 #ifdef PERL_EXT_RE_BUILD
34 /* need to replace pregcomp et al, so enable that */
35 # ifndef PERL_IN_XSUB_RE
36 # define PERL_IN_XSUB_RE
38 /* need access to debugger hooks */
39 # if defined(PERL_EXT_RE_DEBUG) && !defined(DEBUGGING)
44 #ifdef PERL_IN_XSUB_RE
45 /* We *really* need to overwrite these symbols: */
46 # define Perl_regexec_flags my_regexec
47 # define Perl_regdump my_regdump
48 # define Perl_regprop my_regprop
49 # define Perl_re_intuit_start my_re_intuit_start
50 /* *These* symbols are masked to allow static link. */
51 # define Perl_pregexec my_pregexec
52 # define Perl_reginitcolors my_reginitcolors
53 # define Perl_regclass_swash my_regclass_swash
55 # define PERL_NO_GET_CONTEXT
59 * pregcomp and pregexec -- regsub and regerror are not used in perl
61 * Copyright (c) 1986 by University of Toronto.
62 * Written by Henry Spencer. Not derived from licensed software.
64 * Permission is granted to anyone to use this software for any
65 * purpose on any computer system, and to redistribute it freely,
66 * subject to the following restrictions:
68 * 1. The author is not responsible for the consequences of use of
69 * this software, no matter how awful, even if they arise
72 * 2. The origin of this software must not be misrepresented, either
73 * by explicit claim or by omission.
75 * 3. Altered versions must be plainly marked as such, and must not
76 * be misrepresented as being the original software.
78 **** Alterations to Henry's code are...
80 **** Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
81 **** 2000, 2001, 2002, 2003, 2004, 2005, 2006, by Larry Wall and others
83 **** You may distribute under the terms of either the GNU General Public
84 **** License or the Artistic License, as specified in the README file.
86 * Beware that some of this code is subtly aware of the way operator
87 * precedence is structured in regular expressions. Serious changes in
88 * regular-expression syntax might require a total rethink.
91 #define PERL_IN_REGEXEC_C
96 #define RF_tainted 1 /* tainted information used? */
97 #define RF_warned 2 /* warned about big count? */
98 #define RF_evaled 4 /* Did an EVAL with setting? */
99 #define RF_utf8 8 /* String contains multibyte chars? */
101 #define UTF ((PL_reg_flags & RF_utf8) != 0)
103 #define RS_init 1 /* eval environment created */
104 #define RS_set 2 /* replsv value is set */
107 #define STATIC static
110 #define REGINCLASS(p,c) (ANYOF_FLAGS(p) ? reginclass(p,c,0,0) : ANYOF_BITMAP_TEST(p,*(c)))
116 #define CHR_SVLEN(sv) (do_utf8 ? sv_len_utf8(sv) : SvCUR(sv))
117 #define CHR_DIST(a,b) (PL_reg_match_utf8 ? utf8_distance(a,b) : a - b)
119 #define reghop_c(pos,off) ((char*)reghop((U8*)pos, off))
120 #define reghopmaybe_c(pos,off) ((char*)reghopmaybe((U8*)pos, off))
121 #define HOP(pos,off) (PL_reg_match_utf8 ? reghop((U8*)pos, off) : (U8*)(pos + off))
122 #define HOPMAYBE(pos,off) (PL_reg_match_utf8 ? reghopmaybe((U8*)pos, off) : (U8*)(pos + off))
123 #define HOPc(pos,off) ((char*)HOP(pos,off))
124 #define HOPMAYBEc(pos,off) ((char*)HOPMAYBE(pos,off))
126 #define HOPBACK(pos, off) ( \
127 (PL_reg_match_utf8) \
128 ? reghopmaybe((U8*)pos, -off) \
129 : (pos - off >= PL_bostr) \
133 #define HOPBACKc(pos, off) (char*)HOPBACK(pos, off)
135 #define reghop3_c(pos,off,lim) ((char*)reghop3((U8*)pos, off, (U8*)lim))
136 #define reghopmaybe3_c(pos,off,lim) ((char*)reghopmaybe3((U8*)pos, off, (U8*)lim))
137 #define HOP3(pos,off,lim) (PL_reg_match_utf8 ? reghop3((U8*)pos, off, (U8*)lim) : (U8*)(pos + off))
138 #define HOPMAYBE3(pos,off,lim) (PL_reg_match_utf8 ? reghopmaybe3((U8*)pos, off, (U8*)lim) : (U8*)(pos + off))
139 #define HOP3c(pos,off,lim) ((char*)HOP3(pos,off,lim))
140 #define HOPMAYBE3c(pos,off,lim) ((char*)HOPMAYBE3(pos,off,lim))
142 #define LOAD_UTF8_CHARCLASS(class,str) STMT_START { \
143 if (!CAT2(PL_utf8_,class)) { bool ok; ENTER; save_re_context(); ok=CAT2(is_utf8_,class)((const U8*)str); assert(ok); LEAVE; } } STMT_END
144 #define LOAD_UTF8_CHARCLASS_ALNUM() LOAD_UTF8_CHARCLASS(alnum,"a")
145 #define LOAD_UTF8_CHARCLASS_DIGIT() LOAD_UTF8_CHARCLASS(digit,"0")
146 #define LOAD_UTF8_CHARCLASS_SPACE() LOAD_UTF8_CHARCLASS(space," ")
147 #define LOAD_UTF8_CHARCLASS_MARK() LOAD_UTF8_CHARCLASS(mark, "\xcd\x86")
149 /* for use after a quantifier and before an EXACT-like node -- japhy */
150 #define JUMPABLE(rn) ( \
151 OP(rn) == OPEN || OP(rn) == CLOSE || OP(rn) == EVAL || \
152 OP(rn) == SUSPEND || OP(rn) == IFMATCH || \
153 OP(rn) == PLUS || OP(rn) == MINMOD || \
154 (PL_regkind[(U8)OP(rn)] == CURLY && ARG1(rn) > 0) \
157 #define HAS_TEXT(rn) ( \
158 PL_regkind[(U8)OP(rn)] == EXACT || PL_regkind[(U8)OP(rn)] == REF \
162 Search for mandatory following text node; for lookahead, the text must
163 follow but for lookbehind (rn->flags != 0) we skip to the next step.
165 #define FIND_NEXT_IMPT(rn) STMT_START { \
166 while (JUMPABLE(rn)) \
167 if (OP(rn) == SUSPEND || PL_regkind[(U8)OP(rn)] == CURLY) \
168 rn = NEXTOPER(NEXTOPER(rn)); \
169 else if (OP(rn) == PLUS) \
171 else if (OP(rn) == IFMATCH) \
172 rn = (rn->flags == 0) ? NEXTOPER(NEXTOPER(rn)) : rn + ARG(rn); \
173 else rn += NEXT_OFF(rn); \
176 static void restore_pos(pTHX_ void *arg);
179 S_regcppush(pTHX_ I32 parenfloor)
182 const int retval = PL_savestack_ix;
183 #define REGCP_PAREN_ELEMS 4
184 const int paren_elems_to_push = (PL_regsize - parenfloor) * REGCP_PAREN_ELEMS;
187 if (paren_elems_to_push < 0)
188 Perl_croak(aTHX_ "panic: paren_elems_to_push < 0");
190 #define REGCP_OTHER_ELEMS 6
191 SSGROW(paren_elems_to_push + REGCP_OTHER_ELEMS);
192 for (p = PL_regsize; p > parenfloor; p--) {
193 /* REGCP_PARENS_ELEMS are pushed per pairs of parentheses. */
194 SSPUSHINT(PL_regendp[p]);
195 SSPUSHINT(PL_regstartp[p]);
196 SSPUSHPTR(PL_reg_start_tmp[p]);
199 /* REGCP_OTHER_ELEMS are pushed in any case, parentheses or no. */
200 SSPUSHINT(PL_regsize);
201 SSPUSHINT(*PL_reglastparen);
202 SSPUSHINT(*PL_reglastcloseparen);
203 SSPUSHPTR(PL_reginput);
204 #define REGCP_FRAME_ELEMS 2
205 /* REGCP_FRAME_ELEMS are part of the REGCP_OTHER_ELEMS and
206 * are needed for the regexp context stack bookkeeping. */
207 SSPUSHINT(paren_elems_to_push + REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS);
208 SSPUSHINT(SAVEt_REGCONTEXT); /* Magic cookie. */
213 /* These are needed since we do not localize EVAL nodes: */
214 # define REGCP_SET(cp) DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, \
215 " Setting an EVAL scope, savestack=%"IVdf"\n", \
216 (IV)PL_savestack_ix)); cp = PL_savestack_ix
218 # define REGCP_UNWIND(cp) DEBUG_EXECUTE_r(cp != PL_savestack_ix ? \
219 PerlIO_printf(Perl_debug_log, \
220 " Clearing an EVAL scope, savestack=%"IVdf"..%"IVdf"\n", \
221 (IV)(cp), (IV)PL_savestack_ix) : 0); regcpblow(cp)
231 GET_RE_DEBUG_FLAGS_DECL;
233 /* Pop REGCP_OTHER_ELEMS before the parentheses loop starts. */
235 assert(i == SAVEt_REGCONTEXT); /* Check that the magic cookie is there. */
236 i = SSPOPINT; /* Parentheses elements to pop. */
237 input = (char *) SSPOPPTR;
238 *PL_reglastcloseparen = SSPOPINT;
239 *PL_reglastparen = SSPOPINT;
240 PL_regsize = SSPOPINT;
242 /* Now restore the parentheses context. */
243 for (i -= (REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS);
244 i > 0; i -= REGCP_PAREN_ELEMS) {
246 paren = (U32)SSPOPINT;
247 PL_reg_start_tmp[paren] = (char *) SSPOPPTR;
248 PL_regstartp[paren] = SSPOPINT;
250 if (paren <= *PL_reglastparen)
251 PL_regendp[paren] = tmps;
253 PerlIO_printf(Perl_debug_log,
254 " restoring \\%"UVuf" to %"IVdf"(%"IVdf")..%"IVdf"%s\n",
255 (UV)paren, (IV)PL_regstartp[paren],
256 (IV)(PL_reg_start_tmp[paren] - PL_bostr),
257 (IV)PL_regendp[paren],
258 (paren > *PL_reglastparen ? "(no)" : ""));
262 if ((I32)(*PL_reglastparen + 1) <= PL_regnpar) {
263 PerlIO_printf(Perl_debug_log,
264 " restoring \\%"IVdf"..\\%"IVdf" to undef\n",
265 (IV)(*PL_reglastparen + 1), (IV)PL_regnpar);
269 /* It would seem that the similar code in regtry()
270 * already takes care of this, and in fact it is in
271 * a better location to since this code can #if 0-ed out
272 * but the code in regtry() is needed or otherwise tests
273 * requiring null fields (pat.t#187 and split.t#{13,14}
274 * (as of patchlevel 7877) will fail. Then again,
275 * this code seems to be necessary or otherwise
276 * building DynaLoader will fail:
277 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
279 for (paren = *PL_reglastparen + 1; (I32)paren <= PL_regnpar; paren++) {
280 if ((I32)paren > PL_regsize)
281 PL_regstartp[paren] = -1;
282 PL_regendp[paren] = -1;
288 typedef struct re_cc_state
292 struct re_cc_state *prev;
297 #define regcpblow(cp) LEAVE_SCOPE(cp) /* Ignores regcppush()ed data. */
299 #define TRYPAREN(paren, n, input) { \
302 PL_regstartp[paren] = HOPc(input, -1) - PL_bostr; \
303 PL_regendp[paren] = input - PL_bostr; \
306 PL_regendp[paren] = -1; \
308 if (regmatch(next)) \
311 PL_regendp[paren] = -1; \
316 * pregexec and friends
320 - pregexec - match a regexp against a string
323 Perl_pregexec(pTHX_ register regexp *prog, char *stringarg, register char *strend,
324 char *strbeg, I32 minend, SV *screamer, U32 nosave)
325 /* strend: pointer to null at end of string */
326 /* strbeg: real beginning of string */
327 /* minend: end of match must be >=minend after stringarg. */
328 /* nosave: For optimizations. */
331 regexec_flags(prog, stringarg, strend, strbeg, minend, screamer, NULL,
332 nosave ? 0 : REXEC_COPY_STR);
336 S_cache_re(pTHX_ regexp *prog)
339 PL_regprecomp = prog->precomp; /* Needed for FAIL. */
341 PL_regprogram = prog->program;
343 PL_regnpar = prog->nparens;
344 PL_regdata = prog->data;
349 * Need to implement the following flags for reg_anch:
351 * USE_INTUIT_NOML - Useful to call re_intuit_start() first
353 * INTUIT_AUTORITATIVE_NOML - Can trust a positive answer
354 * INTUIT_AUTORITATIVE_ML
355 * INTUIT_ONCE_NOML - Intuit can match in one location only.
358 * Another flag for this function: SECOND_TIME (so that float substrs
359 * with giant delta may be not rechecked).
362 /* Assumptions: if ANCH_GPOS, then strpos is anchored. XXXX Check GPOS logic */
364 /* If SCREAM, then SvPVX_const(sv) should be compatible with strpos and strend.
365 Otherwise, only SvCUR(sv) is used to get strbeg. */
367 /* XXXX We assume that strpos is strbeg unless sv. */
369 /* XXXX Some places assume that there is a fixed substring.
370 An update may be needed if optimizer marks as "INTUITable"
371 RExen without fixed substrings. Similarly, it is assumed that
372 lengths of all the strings are no more than minlen, thus they
373 cannot come from lookahead.
374 (Or minlen should take into account lookahead.) */
376 /* A failure to find a constant substring means that there is no need to make
377 an expensive call to REx engine, thus we celebrate a failure. Similarly,
378 finding a substring too deep into the string means that less calls to
379 regtry() should be needed.
381 REx compiler's optimizer found 4 possible hints:
382 a) Anchored substring;
384 c) Whether we are anchored (beginning-of-line or \G);
385 d) First node (of those at offset 0) which may distingush positions;
386 We use a)b)d) and multiline-part of c), and try to find a position in the
387 string which does not contradict any of them.
390 /* Most of decisions we do here should have been done at compile time.
391 The nodes of the REx which we used for the search should have been
392 deleted from the finite automaton. */
395 Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos,
396 char *strend, U32 flags, re_scream_pos_data *data)
399 register I32 start_shift = 0;
400 /* Should be nonnegative! */
401 register I32 end_shift = 0;
406 const int do_utf8 = sv ? SvUTF8(sv) : 0; /* if no sv we have to assume bytes */
408 register char *other_last = NULL; /* other substr checked before this */
409 char *check_at = NULL; /* check substr found at this pos */
410 const I32 multiline = prog->reganch & PMf_MULTILINE;
412 const char * const i_strpos = strpos;
413 SV * const dsv = PERL_DEBUG_PAD_ZERO(0);
416 GET_RE_DEBUG_FLAGS_DECL;
418 RX_MATCH_UTF8_set(prog,do_utf8);
420 if (prog->reganch & ROPT_UTF8) {
421 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
422 "UTF-8 regex...\n"));
423 PL_reg_flags |= RF_utf8;
427 const char *s = PL_reg_match_utf8 ?
428 sv_uni_display(dsv, sv, 60, UNI_DISPLAY_REGEX) :
430 const int len = PL_reg_match_utf8 ?
431 strlen(s) : strend - strpos;
434 if (PL_reg_match_utf8)
435 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
436 "UTF-8 target...\n"));
437 PerlIO_printf(Perl_debug_log,
438 "%sGuessing start of match, REx%s \"%s%.60s%s%s\" against \"%s%.*s%s%s\"...\n",
439 PL_colors[4], PL_colors[5], PL_colors[0],
442 (strlen(prog->precomp) > 60 ? "..." : ""),
444 (int)(len > 60 ? 60 : len),
446 (len > 60 ? "..." : "")
450 /* CHR_DIST() would be more correct here but it makes things slow. */
451 if (prog->minlen > strend - strpos) {
452 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
453 "String too short... [re_intuit_start]\n"));
456 strbeg = (sv && SvPOK(sv)) ? strend - SvCUR(sv) : strpos;
459 if (!prog->check_utf8 && prog->check_substr)
460 to_utf8_substr(prog);
461 check = prog->check_utf8;
463 if (!prog->check_substr && prog->check_utf8)
464 to_byte_substr(prog);
465 check = prog->check_substr;
467 if (check == &PL_sv_undef) {
468 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
469 "Non-utf string cannot match utf check string\n"));
472 if (prog->reganch & ROPT_ANCH) { /* Match at beg-of-str or after \n */
473 ml_anch = !( (prog->reganch & ROPT_ANCH_SINGLE)
474 || ( (prog->reganch & ROPT_ANCH_BOL)
475 && !multiline ) ); /* Check after \n? */
478 if ( !(prog->reganch & (ROPT_ANCH_GPOS /* Checked by the caller */
479 | ROPT_IMPLICIT)) /* not a real BOL */
480 /* SvCUR is not set on references: SvRV and SvPVX_const overlap */
482 && (strpos != strbeg)) {
483 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Not at start...\n"));
486 if (prog->check_offset_min == prog->check_offset_max &&
487 !(prog->reganch & ROPT_CANY_SEEN)) {
488 /* Substring at constant offset from beg-of-str... */
491 s = HOP3c(strpos, prog->check_offset_min, strend);
493 slen = SvCUR(check); /* >= 1 */
495 if ( strend - s > slen || strend - s < slen - 1
496 || (strend - s == slen && strend[-1] != '\n')) {
497 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "String too long...\n"));
500 /* Now should match s[0..slen-2] */
502 if (slen && (*SvPVX_const(check) != *s
504 && memNE(SvPVX_const(check), s, slen)))) {
506 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "String not equal...\n"));
510 else if (*SvPVX_const(check) != *s
511 || ((slen = SvCUR(check)) > 1
512 && memNE(SvPVX_const(check), s, slen)))
515 goto success_at_start;
518 /* Match is anchored, but substr is not anchored wrt beg-of-str. */
520 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
521 end_shift = prog->minlen - start_shift -
522 CHR_SVLEN(check) + (SvTAIL(check) != 0);
524 const I32 end = prog->check_offset_max + CHR_SVLEN(check)
525 - (SvTAIL(check) != 0);
526 const I32 eshift = CHR_DIST((U8*)strend, (U8*)s) - end;
528 if (end_shift < eshift)
532 else { /* Can match at random position */
535 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
536 /* Should be nonnegative! */
537 end_shift = prog->minlen - start_shift -
538 CHR_SVLEN(check) + (SvTAIL(check) != 0);
541 #ifdef DEBUGGING /* 7/99: reports of failure (with the older version) */
543 Perl_croak(aTHX_ "panic: end_shift");
547 /* Find a possible match in the region s..strend by looking for
548 the "check" substring in the region corrected by start/end_shift. */
549 if (flags & REXEC_SCREAM) {
550 I32 p = -1; /* Internal iterator of scream. */
551 I32 * const pp = data ? data->scream_pos : &p;
553 if (PL_screamfirst[BmRARE(check)] >= 0
554 || ( BmRARE(check) == '\n'
555 && (BmPREVIOUS(check) == SvCUR(check) - 1)
557 s = screaminstr(sv, check,
558 start_shift + (s - strbeg), end_shift, pp, 0);
561 /* we may be pointing at the wrong string */
562 if (s && RX_MATCH_COPIED(prog))
563 s = strbeg + (s - SvPVX_const(sv));
565 *data->scream_olds = s;
567 else if (prog->reganch & ROPT_CANY_SEEN)
568 s = fbm_instr((U8*)(s + start_shift),
569 (U8*)(strend - end_shift),
570 check, multiline ? FBMrf_MULTILINE : 0);
572 s = fbm_instr(HOP3(s, start_shift, strend),
573 HOP3(strend, -end_shift, strbeg),
574 check, multiline ? FBMrf_MULTILINE : 0);
576 /* Update the count-of-usability, remove useless subpatterns,
579 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%s %s substr \"%s%.*s%s\"%s%s",
580 (s ? "Found" : "Did not find"),
581 (check == (do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) ? "anchored" : "floating"),
583 (int)(SvCUR(check) - (SvTAIL(check)!=0)),
585 PL_colors[1], (SvTAIL(check) ? "$" : ""),
586 (s ? " at offset " : "...\n") ) );
593 /* Finish the diagnostic message */
594 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%ld...\n", (long)(s - i_strpos)) );
596 /* Got a candidate. Check MBOL anchoring, and the *other* substr.
597 Start with the other substr.
598 XXXX no SCREAM optimization yet - and a very coarse implementation
599 XXXX /ttx+/ results in anchored="ttx", floating="x". floating will
600 *always* match. Probably should be marked during compile...
601 Probably it is right to do no SCREAM here...
604 if (do_utf8 ? (prog->float_utf8 && prog->anchored_utf8) : (prog->float_substr && prog->anchored_substr)) {
605 /* Take into account the "other" substring. */
606 /* XXXX May be hopelessly wrong for UTF... */
609 if (check == (do_utf8 ? prog->float_utf8 : prog->float_substr)) {
612 char * const last = HOP3c(s, -start_shift, strbeg);
617 t = s - prog->check_offset_max;
618 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
620 || ((t = reghopmaybe3_c(s, -(prog->check_offset_max), strpos))
625 t = HOP3c(t, prog->anchored_offset, strend);
626 if (t < other_last) /* These positions already checked */
628 last2 = last1 = HOP3c(strend, -prog->minlen, strbeg);
631 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
632 /* On end-of-str: see comment below. */
633 must = do_utf8 ? prog->anchored_utf8 : prog->anchored_substr;
634 if (must == &PL_sv_undef) {
636 DEBUG_EXECUTE_r(must = prog->anchored_utf8); /* for debug */
641 HOP3(HOP3(last1, prog->anchored_offset, strend)
642 + SvCUR(must), -(SvTAIL(must)!=0), strbeg),
644 multiline ? FBMrf_MULTILINE : 0
646 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
647 "%s anchored substr \"%s%.*s%s\"%s",
648 (s ? "Found" : "Contradicts"),
651 - (SvTAIL(must)!=0)),
653 PL_colors[1], (SvTAIL(must) ? "$" : "")));
655 if (last1 >= last2) {
656 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
657 ", giving up...\n"));
660 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
661 ", trying floating at offset %ld...\n",
662 (long)(HOP3c(s1, 1, strend) - i_strpos)));
663 other_last = HOP3c(last1, prog->anchored_offset+1, strend);
664 s = HOP3c(last, 1, strend);
668 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
669 (long)(s - i_strpos)));
670 t = HOP3c(s, -prog->anchored_offset, strbeg);
671 other_last = HOP3c(s, 1, strend);
679 else { /* Take into account the floating substring. */
684 t = HOP3c(s, -start_shift, strbeg);
686 HOP3c(strend, -prog->minlen + prog->float_min_offset, strbeg);
687 if (CHR_DIST((U8*)last, (U8*)t) > prog->float_max_offset)
688 last = HOP3c(t, prog->float_max_offset, strend);
689 s = HOP3c(t, prog->float_min_offset, strend);
692 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
693 must = do_utf8 ? prog->float_utf8 : prog->float_substr;
694 /* fbm_instr() takes into account exact value of end-of-str
695 if the check is SvTAIL(ed). Since false positives are OK,
696 and end-of-str is not later than strend we are OK. */
697 if (must == &PL_sv_undef) {
699 DEBUG_EXECUTE_r(must = prog->float_utf8); /* for debug message */
702 s = fbm_instr((unsigned char*)s,
703 (unsigned char*)last + SvCUR(must)
705 must, multiline ? FBMrf_MULTILINE : 0);
706 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%s floating substr \"%s%.*s%s\"%s",
707 (s ? "Found" : "Contradicts"),
709 (int)(SvCUR(must) - (SvTAIL(must)!=0)),
711 PL_colors[1], (SvTAIL(must) ? "$" : "")));
714 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
715 ", giving up...\n"));
718 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
719 ", trying anchored starting at offset %ld...\n",
720 (long)(s1 + 1 - i_strpos)));
722 s = HOP3c(t, 1, strend);
726 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
727 (long)(s - i_strpos)));
728 other_last = s; /* Fix this later. --Hugo */
737 t = s - prog->check_offset_max;
738 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
740 || ((t = reghopmaybe3_c(s, -prog->check_offset_max, strpos))
742 /* Fixed substring is found far enough so that the match
743 cannot start at strpos. */
745 if (ml_anch && t[-1] != '\n') {
746 /* Eventually fbm_*() should handle this, but often
747 anchored_offset is not 0, so this check will not be wasted. */
748 /* XXXX In the code below we prefer to look for "^" even in
749 presence of anchored substrings. And we search even
750 beyond the found float position. These pessimizations
751 are historical artefacts only. */
753 while (t < strend - prog->minlen) {
755 if (t < check_at - prog->check_offset_min) {
756 if (do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) {
757 /* Since we moved from the found position,
758 we definitely contradict the found anchored
759 substr. Due to the above check we do not
760 contradict "check" substr.
761 Thus we can arrive here only if check substr
762 is float. Redo checking for "other"=="fixed".
765 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld, rescanning for anchored from offset %ld...\n",
766 PL_colors[0], PL_colors[1], (long)(strpos - i_strpos), (long)(strpos - i_strpos + prog->anchored_offset)));
767 goto do_other_anchored;
769 /* We don't contradict the found floating substring. */
770 /* XXXX Why not check for STCLASS? */
772 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld...\n",
773 PL_colors[0], PL_colors[1], (long)(s - i_strpos)));
776 /* Position contradicts check-string */
777 /* XXXX probably better to look for check-string
778 than for "\n", so one should lower the limit for t? */
779 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m, restarting lookup for check-string at offset %ld...\n",
780 PL_colors[0], PL_colors[1], (long)(t + 1 - i_strpos)));
781 other_last = strpos = s = t + 1;
786 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Did not find /%s^%s/m...\n",
787 PL_colors[0], PL_colors[1]));
791 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Starting position does not contradict /%s^%s/m...\n",
792 PL_colors[0], PL_colors[1]));
796 ++BmUSEFUL(do_utf8 ? prog->check_utf8 : prog->check_substr); /* hooray/5 */
799 /* The found string does not prohibit matching at strpos,
800 - no optimization of calling REx engine can be performed,
801 unless it was an MBOL and we are not after MBOL,
802 or a future STCLASS check will fail this. */
804 /* Even in this situation we may use MBOL flag if strpos is offset
805 wrt the start of the string. */
806 if (ml_anch && sv && !SvROK(sv) /* See prev comment on SvROK */
807 && (strpos != strbeg) && strpos[-1] != '\n'
808 /* May be due to an implicit anchor of m{.*foo} */
809 && !(prog->reganch & ROPT_IMPLICIT))
814 DEBUG_EXECUTE_r( if (ml_anch)
815 PerlIO_printf(Perl_debug_log, "Position at offset %ld does not contradict /%s^%s/m...\n",
816 (long)(strpos - i_strpos), PL_colors[0], PL_colors[1]);
819 if (!(prog->reganch & ROPT_NAUGHTY) /* XXXX If strpos moved? */
821 prog->check_utf8 /* Could be deleted already */
822 && --BmUSEFUL(prog->check_utf8) < 0
823 && (prog->check_utf8 == prog->float_utf8)
825 prog->check_substr /* Could be deleted already */
826 && --BmUSEFUL(prog->check_substr) < 0
827 && (prog->check_substr == prog->float_substr)
830 /* If flags & SOMETHING - do not do it many times on the same match */
831 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "... Disabling check substring...\n"));
832 SvREFCNT_dec(do_utf8 ? prog->check_utf8 : prog->check_substr);
833 if (do_utf8 ? prog->check_substr : prog->check_utf8)
834 SvREFCNT_dec(do_utf8 ? prog->check_substr : prog->check_utf8);
835 prog->check_substr = prog->check_utf8 = NULL; /* disable */
836 prog->float_substr = prog->float_utf8 = NULL; /* clear */
837 check = NULL; /* abort */
839 /* XXXX This is a remnant of the old implementation. It
840 looks wasteful, since now INTUIT can use many
842 prog->reganch &= ~RE_USE_INTUIT;
849 /* XXXX BmUSEFUL already changed, maybe multiple change is meaningful... */
850 if (prog->regstclass) {
851 /* minlen == 0 is possible if regstclass is \b or \B,
852 and the fixed substr is ''$.
853 Since minlen is already taken into account, s+1 is before strend;
854 accidentally, minlen >= 1 guaranties no false positives at s + 1
855 even for \b or \B. But (minlen? 1 : 0) below assumes that
856 regstclass does not come from lookahead... */
857 /* If regstclass takes bytelength more than 1: If charlength==1, OK.
858 This leaves EXACTF only, which is dealt with in find_byclass(). */
859 const U8* const str = (U8*)STRING(prog->regstclass);
860 const int cl_l = (PL_regkind[(U8)OP(prog->regstclass)] == EXACT
861 ? CHR_DIST(str+STR_LEN(prog->regstclass), str)
863 const char * const endpos = (prog->anchored_substr || prog->anchored_utf8 || ml_anch)
864 ? HOP3c(s, (prog->minlen ? cl_l : 0), strend)
865 : (prog->float_substr || prog->float_utf8
866 ? HOP3c(HOP3c(check_at, -start_shift, strbeg),
872 s = find_byclass(prog, prog->regstclass, s, endpos, 1);
875 const char *what = NULL;
877 if (endpos == strend) {
878 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
879 "Could not match STCLASS...\n") );
882 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
883 "This position contradicts STCLASS...\n") );
884 if ((prog->reganch & ROPT_ANCH) && !ml_anch)
886 /* Contradict one of substrings */
887 if (prog->anchored_substr || prog->anchored_utf8) {
888 if ((do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) == check) {
889 DEBUG_EXECUTE_r( what = "anchored" );
891 s = HOP3c(t, 1, strend);
892 if (s + start_shift + end_shift > strend) {
893 /* XXXX Should be taken into account earlier? */
894 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
895 "Could not match STCLASS...\n") );
900 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
901 "Looking for %s substr starting at offset %ld...\n",
902 what, (long)(s + start_shift - i_strpos)) );
905 /* Have both, check_string is floating */
906 if (t + start_shift >= check_at) /* Contradicts floating=check */
907 goto retry_floating_check;
908 /* Recheck anchored substring, but not floating... */
912 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
913 "Looking for anchored substr starting at offset %ld...\n",
914 (long)(other_last - i_strpos)) );
915 goto do_other_anchored;
917 /* Another way we could have checked stclass at the
918 current position only: */
923 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
924 "Looking for /%s^%s/m starting at offset %ld...\n",
925 PL_colors[0], PL_colors[1], (long)(t - i_strpos)) );
928 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr)) /* Could have been deleted */
930 /* Check is floating subtring. */
931 retry_floating_check:
932 t = check_at - start_shift;
933 DEBUG_EXECUTE_r( what = "floating" );
934 goto hop_and_restart;
937 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
938 "By STCLASS: moving %ld --> %ld\n",
939 (long)(t - i_strpos), (long)(s - i_strpos))
943 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
944 "Does not contradict STCLASS...\n");
949 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%s%s:%s match at offset %ld\n",
950 PL_colors[4], (check ? "Guessed" : "Giving up"),
951 PL_colors[5], (long)(s - i_strpos)) );
954 fail_finish: /* Substring not found */
955 if (prog->check_substr || prog->check_utf8) /* could be removed already */
956 BmUSEFUL(do_utf8 ? prog->check_utf8 : prog->check_substr) += 5; /* hooray */
958 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch rejected by optimizer%s\n",
959 PL_colors[4], PL_colors[5]));
963 /* We know what class REx starts with. Try to find this position... */
965 S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, const char *strend, I32 norun)
968 const I32 doevery = (prog->reganch & ROPT_SKIP) == 0;
972 register STRLEN uskip;
976 register I32 tmp = 1; /* Scratch variable? */
977 register const bool do_utf8 = PL_reg_match_utf8;
979 /* We know what class it must start with. */
983 while (s + (uskip = UTF8SKIP(s)) <= strend) {
984 if ((ANYOF_FLAGS(c) & ANYOF_UNICODE) ||
985 !UTF8_IS_INVARIANT((U8)s[0]) ?
986 reginclass(c, (U8*)s, 0, do_utf8) :
987 REGINCLASS(c, (U8*)s)) {
988 if (tmp && (norun || regtry(prog, s)))
1002 if (REGINCLASS(c, (U8*)s) ||
1003 (ANYOF_FOLD_SHARP_S(c, s, strend) &&
1004 /* The assignment of 2 is intentional:
1005 * for the folded sharp s, the skip is 2. */
1006 (skip = SHARP_S_SKIP))) {
1007 if (tmp && (norun || regtry(prog, s)))
1019 while (s < strend) {
1020 if (tmp && (norun || regtry(prog, s)))
1029 ln = STR_LEN(c); /* length to match in octets/bytes */
1030 lnc = (I32) ln; /* length to match in characters */
1032 STRLEN ulen1, ulen2;
1034 U8 tmpbuf1[UTF8_MAXBYTES_CASE+1];
1035 U8 tmpbuf2[UTF8_MAXBYTES_CASE+1];
1036 const U32 uniflags = ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY;
1038 to_utf8_lower((U8*)m, tmpbuf1, &ulen1);
1039 to_utf8_upper((U8*)m, tmpbuf2, &ulen2);
1041 c1 = utf8n_to_uvchr(tmpbuf1, UTF8_MAXBYTES_CASE,
1043 c2 = utf8n_to_uvchr(tmpbuf2, UTF8_MAXBYTES_CASE,
1046 while (sm < ((U8 *) m + ln)) {
1061 c2 = PL_fold_locale[c1];
1063 e = HOP3c(strend, -((I32)lnc), s);
1066 e = s; /* Due to minlen logic of intuit() */
1068 /* The idea in the EXACTF* cases is to first find the
1069 * first character of the EXACTF* node and then, if
1070 * necessary, case-insensitively compare the full
1071 * text of the node. The c1 and c2 are the first
1072 * characters (though in Unicode it gets a bit
1073 * more complicated because there are more cases
1074 * than just upper and lower: one needs to use
1075 * the so-called folding case for case-insensitive
1076 * matching (called "loose matching" in Unicode).
1077 * ibcmp_utf8() will do just that. */
1081 U8 tmpbuf [UTF8_MAXBYTES+1];
1082 STRLEN len, foldlen;
1083 const U32 uniflags = ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY;
1085 /* Upper and lower of 1st char are equal -
1086 * probably not a "letter". */
1088 c = utf8n_to_uvchr((U8*)s, UTF8_MAXBYTES, &len,
1092 ibcmp_utf8(s, (char **)0, 0, do_utf8,
1093 m, (char **)0, ln, (bool)UTF))
1094 && (norun || regtry(prog, s)) )
1097 U8 foldbuf[UTF8_MAXBYTES_CASE+1];
1098 uvchr_to_utf8(tmpbuf, c);
1099 f = to_utf8_fold(tmpbuf, foldbuf, &foldlen);
1101 && (f == c1 || f == c2)
1102 && (ln == foldlen ||
1103 !ibcmp_utf8((char *) foldbuf,
1104 (char **)0, foldlen, do_utf8,
1106 (char **)0, ln, (bool)UTF))
1107 && (norun || regtry(prog, s)) )
1115 c = utf8n_to_uvchr((U8*)s, UTF8_MAXBYTES, &len,
1118 /* Handle some of the three Greek sigmas cases.
1119 * Note that not all the possible combinations
1120 * are handled here: some of them are handled
1121 * by the standard folding rules, and some of
1122 * them (the character class or ANYOF cases)
1123 * are handled during compiletime in
1124 * regexec.c:S_regclass(). */
1125 if (c == (UV)UNICODE_GREEK_CAPITAL_LETTER_SIGMA ||
1126 c == (UV)UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA)
1127 c = (UV)UNICODE_GREEK_SMALL_LETTER_SIGMA;
1129 if ( (c == c1 || c == c2)
1131 ibcmp_utf8(s, (char **)0, 0, do_utf8,
1132 m, (char **)0, ln, (bool)UTF))
1133 && (norun || regtry(prog, s)) )
1136 U8 foldbuf[UTF8_MAXBYTES_CASE+1];
1137 uvchr_to_utf8(tmpbuf, c);
1138 f = to_utf8_fold(tmpbuf, foldbuf, &foldlen);
1140 && (f == c1 || f == c2)
1141 && (ln == foldlen ||
1142 !ibcmp_utf8((char *) foldbuf,
1143 (char **)0, foldlen, do_utf8,
1145 (char **)0, ln, (bool)UTF))
1146 && (norun || regtry(prog, s)) )
1157 && (ln == 1 || !(OP(c) == EXACTF
1159 : ibcmp_locale(s, m, ln)))
1160 && (norun || regtry(prog, s)) )
1166 if ( (*(U8*)s == c1 || *(U8*)s == c2)
1167 && (ln == 1 || !(OP(c) == EXACTF
1169 : ibcmp_locale(s, m, ln)))
1170 && (norun || regtry(prog, s)) )
1177 PL_reg_flags |= RF_tainted;
1184 U8 * const r = reghop3((U8*)s, -1, (U8*)PL_bostr);
1185 tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, 0);
1187 tmp = ((OP(c) == BOUND ?
1188 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1189 LOAD_UTF8_CHARCLASS_ALNUM();
1190 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1191 if (tmp == !(OP(c) == BOUND ?
1192 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1193 isALNUM_LC_utf8((U8*)s)))
1196 if ((norun || regtry(prog, s)))
1203 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1204 tmp = ((OP(c) == BOUND ? isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1205 while (s < strend) {
1207 !(OP(c) == BOUND ? isALNUM(*s) : isALNUM_LC(*s))) {
1209 if ((norun || regtry(prog, s)))
1215 if ((!prog->minlen && tmp) && (norun || regtry(prog, s)))
1219 PL_reg_flags |= RF_tainted;
1226 U8 * const r = reghop3((U8*)s, -1, (U8*)PL_bostr);
1227 tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, 0);
1229 tmp = ((OP(c) == NBOUND ?
1230 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1231 LOAD_UTF8_CHARCLASS_ALNUM();
1232 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1233 if (tmp == !(OP(c) == NBOUND ?
1234 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1235 isALNUM_LC_utf8((U8*)s)))
1237 else if ((norun || regtry(prog, s)))
1243 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1244 tmp = ((OP(c) == NBOUND ?
1245 isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1246 while (s < strend) {
1248 !(OP(c) == NBOUND ? isALNUM(*s) : isALNUM_LC(*s)))
1250 else if ((norun || regtry(prog, s)))
1255 if ((!prog->minlen && !tmp) && (norun || regtry(prog, s)))
1260 LOAD_UTF8_CHARCLASS_ALNUM();
1261 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1262 if (swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1263 if (tmp && (norun || regtry(prog, s)))
1274 while (s < strend) {
1276 if (tmp && (norun || regtry(prog, s)))
1288 PL_reg_flags |= RF_tainted;
1290 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1291 if (isALNUM_LC_utf8((U8*)s)) {
1292 if (tmp && (norun || regtry(prog, s)))
1303 while (s < strend) {
1304 if (isALNUM_LC(*s)) {
1305 if (tmp && (norun || regtry(prog, s)))
1318 LOAD_UTF8_CHARCLASS_ALNUM();
1319 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1320 if (!swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1321 if (tmp && (norun || regtry(prog, s)))
1332 while (s < strend) {
1334 if (tmp && (norun || regtry(prog, s)))
1346 PL_reg_flags |= RF_tainted;
1348 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1349 if (!isALNUM_LC_utf8((U8*)s)) {
1350 if (tmp && (norun || regtry(prog, s)))
1361 while (s < strend) {
1362 if (!isALNUM_LC(*s)) {
1363 if (tmp && (norun || regtry(prog, s)))
1376 LOAD_UTF8_CHARCLASS_SPACE();
1377 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1378 if (*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8)) {
1379 if (tmp && (norun || regtry(prog, s)))
1390 while (s < strend) {
1392 if (tmp && (norun || regtry(prog, s)))
1404 PL_reg_flags |= RF_tainted;
1406 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1407 if (*s == ' ' || isSPACE_LC_utf8((U8*)s)) {
1408 if (tmp && (norun || regtry(prog, s)))
1419 while (s < strend) {
1420 if (isSPACE_LC(*s)) {
1421 if (tmp && (norun || regtry(prog, s)))
1434 LOAD_UTF8_CHARCLASS_SPACE();
1435 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1436 if (!(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8))) {
1437 if (tmp && (norun || regtry(prog, s)))
1448 while (s < strend) {
1450 if (tmp && (norun || regtry(prog, s)))
1462 PL_reg_flags |= RF_tainted;
1464 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1465 if (!(*s == ' ' || isSPACE_LC_utf8((U8*)s))) {
1466 if (tmp && (norun || regtry(prog, s)))
1477 while (s < strend) {
1478 if (!isSPACE_LC(*s)) {
1479 if (tmp && (norun || regtry(prog, s)))
1492 LOAD_UTF8_CHARCLASS_DIGIT();
1493 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1494 if (swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1495 if (tmp && (norun || regtry(prog, s)))
1506 while (s < strend) {
1508 if (tmp && (norun || regtry(prog, s)))
1520 PL_reg_flags |= RF_tainted;
1522 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1523 if (isDIGIT_LC_utf8((U8*)s)) {
1524 if (tmp && (norun || regtry(prog, s)))
1535 while (s < strend) {
1536 if (isDIGIT_LC(*s)) {
1537 if (tmp && (norun || regtry(prog, s)))
1550 LOAD_UTF8_CHARCLASS_DIGIT();
1551 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1552 if (!swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1553 if (tmp && (norun || regtry(prog, s)))
1564 while (s < strend) {
1566 if (tmp && (norun || regtry(prog, s)))
1578 PL_reg_flags |= RF_tainted;
1580 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1581 if (!isDIGIT_LC_utf8((U8*)s)) {
1582 if (tmp && (norun || regtry(prog, s)))
1593 while (s < strend) {
1594 if (!isDIGIT_LC(*s)) {
1595 if (tmp && (norun || regtry(prog, s)))
1607 Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c));
1616 - regexec_flags - match a regexp against a string
1619 Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char *strend,
1620 char *strbeg, I32 minend, SV *sv, void *data, U32 flags)
1621 /* strend: pointer to null at end of string */
1622 /* strbeg: real beginning of string */
1623 /* minend: end of match must be >=minend after stringarg. */
1624 /* data: May be used for some additional optimizations. */
1625 /* nosave: For optimizations. */
1629 register regnode *c;
1630 register char *startpos = stringarg;
1631 I32 minlen; /* must match at least this many chars */
1632 I32 dontbother = 0; /* how many characters not to try at end */
1633 I32 end_shift = 0; /* Same for the end. */ /* CC */
1634 I32 scream_pos = -1; /* Internal iterator of scream. */
1635 char *scream_olds = NULL;
1636 SV* oreplsv = GvSV(PL_replgv);
1637 const bool do_utf8 = DO_UTF8(sv);
1638 const I32 multiline = prog->reganch & PMf_MULTILINE;
1640 SV * const dsv0 = PERL_DEBUG_PAD_ZERO(0);
1641 SV * const dsv1 = PERL_DEBUG_PAD_ZERO(1);
1644 GET_RE_DEBUG_FLAGS_DECL;
1646 PERL_UNUSED_ARG(data);
1647 RX_MATCH_UTF8_set(prog,do_utf8);
1653 PL_regnarrate = DEBUG_r_TEST;
1656 /* Be paranoid... */
1657 if (prog == NULL || startpos == NULL) {
1658 Perl_croak(aTHX_ "NULL regexp parameter");
1662 minlen = prog->minlen;
1663 if (strend - startpos < minlen) {
1664 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
1665 "String too short [regexec_flags]...\n"));
1669 /* Check validity of program. */
1670 if (UCHARAT(prog->program) != REG_MAGIC) {
1671 Perl_croak(aTHX_ "corrupted regexp program");
1675 PL_reg_eval_set = 0;
1678 if (prog->reganch & ROPT_UTF8)
1679 PL_reg_flags |= RF_utf8;
1681 /* Mark beginning of line for ^ and lookbehind. */
1682 PL_regbol = startpos;
1686 /* Mark end of line for $ (and such) */
1689 /* see how far we have to get to not match where we matched before */
1690 PL_regtill = startpos+minend;
1692 /* We start without call_cc context. */
1695 /* If there is a "must appear" string, look for it. */
1698 if (prog->reganch & ROPT_GPOS_SEEN) { /* Need to have PL_reg_ganch */
1701 if (flags & REXEC_IGNOREPOS) /* Means: check only at start */
1702 PL_reg_ganch = startpos;
1703 else if (sv && SvTYPE(sv) >= SVt_PVMG
1705 && (mg = mg_find(sv, PERL_MAGIC_regex_global))
1706 && mg->mg_len >= 0) {
1707 PL_reg_ganch = strbeg + mg->mg_len; /* Defined pos() */
1708 if (prog->reganch & ROPT_ANCH_GPOS) {
1709 if (s > PL_reg_ganch)
1714 else /* pos() not defined */
1715 PL_reg_ganch = strbeg;
1718 if (!(flags & REXEC_CHECKED) && (prog->check_substr != NULL || prog->check_utf8 != NULL)) {
1719 re_scream_pos_data d;
1721 d.scream_olds = &scream_olds;
1722 d.scream_pos = &scream_pos;
1723 s = re_intuit_start(prog, sv, s, strend, flags, &d);
1725 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Not present...\n"));
1726 goto phooey; /* not present */
1731 const char * const s0 = UTF
1732 ? pv_uni_display(dsv0, (U8*)prog->precomp, prog->prelen, 60,
1735 const int len0 = UTF ? SvCUR(dsv0) : prog->prelen;
1736 const char * const s1 = do_utf8 ? sv_uni_display(dsv1, sv, 60,
1737 UNI_DISPLAY_REGEX) : startpos;
1738 const int len1 = do_utf8 ? SvCUR(dsv1) : strend - startpos;
1741 PerlIO_printf(Perl_debug_log,
1742 "%sMatching REx%s \"%s%*.*s%s%s\" against \"%s%.*s%s%s\"\n",
1743 PL_colors[4], PL_colors[5], PL_colors[0],
1746 len0 > 60 ? "..." : "",
1748 (int)(len1 > 60 ? 60 : len1),
1750 (len1 > 60 ? "..." : "")
1754 /* Simplest case: anchored match need be tried only once. */
1755 /* [unless only anchor is BOL and multiline is set] */
1756 if (prog->reganch & (ROPT_ANCH & ~ROPT_ANCH_GPOS)) {
1757 if (s == startpos && regtry(prog, startpos))
1759 else if (multiline || (prog->reganch & ROPT_IMPLICIT)
1760 || (prog->reganch & ROPT_ANCH_MBOL)) /* XXXX SBOL? */
1765 dontbother = minlen - 1;
1766 end = HOP3c(strend, -dontbother, strbeg) - 1;
1767 /* for multiline we only have to try after newlines */
1768 if (prog->check_substr || prog->check_utf8) {
1772 if (regtry(prog, s))
1777 if (prog->reganch & RE_USE_INTUIT) {
1778 s = re_intuit_start(prog, sv, s + 1, strend, flags, NULL);
1789 if (*s++ == '\n') { /* don't need PL_utf8skip here */
1790 if (regtry(prog, s))
1797 } else if (prog->reganch & ROPT_ANCH_GPOS) {
1798 if (regtry(prog, PL_reg_ganch))
1803 /* Messy cases: unanchored match. */
1804 if ((prog->anchored_substr || prog->anchored_utf8) && prog->reganch & ROPT_SKIP) {
1805 /* we have /x+whatever/ */
1806 /* it must be a one character string (XXXX Except UTF?) */
1811 if (!(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr))
1812 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1813 ch = SvPVX_const(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr)[0];
1816 while (s < strend) {
1818 DEBUG_EXECUTE_r( did_match = 1 );
1819 if (regtry(prog, s)) goto got_it;
1821 while (s < strend && *s == ch)
1828 while (s < strend) {
1830 DEBUG_EXECUTE_r( did_match = 1 );
1831 if (regtry(prog, s)) goto got_it;
1833 while (s < strend && *s == ch)
1839 DEBUG_EXECUTE_r(if (!did_match)
1840 PerlIO_printf(Perl_debug_log,
1841 "Did not find anchored character...\n")
1844 else if (prog->anchored_substr != NULL
1845 || prog->anchored_utf8 != NULL
1846 || ((prog->float_substr != NULL || prog->float_utf8 != NULL)
1847 && prog->float_max_offset < strend - s)) {
1852 char *last1; /* Last position checked before */
1856 if (prog->anchored_substr || prog->anchored_utf8) {
1857 if (!(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr))
1858 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1859 must = do_utf8 ? prog->anchored_utf8 : prog->anchored_substr;
1860 back_max = back_min = prog->anchored_offset;
1862 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr))
1863 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1864 must = do_utf8 ? prog->float_utf8 : prog->float_substr;
1865 back_max = prog->float_max_offset;
1866 back_min = prog->float_min_offset;
1868 if (must == &PL_sv_undef)
1869 /* could not downgrade utf8 check substring, so must fail */
1872 last = HOP3c(strend, /* Cannot start after this */
1873 -(I32)(CHR_SVLEN(must)
1874 - (SvTAIL(must) != 0) + back_min), strbeg);
1877 last1 = HOPc(s, -1);
1879 last1 = s - 1; /* bogus */
1881 /* XXXX check_substr already used to find "s", can optimize if
1882 check_substr==must. */
1884 dontbother = end_shift;
1885 strend = HOPc(strend, -dontbother);
1886 while ( (s <= last) &&
1887 ((flags & REXEC_SCREAM)
1888 ? (s = screaminstr(sv, must, HOP3c(s, back_min, strend) - strbeg,
1889 end_shift, &scream_pos, 0))
1890 : (s = fbm_instr((unsigned char*)HOP3(s, back_min, strend),
1891 (unsigned char*)strend, must,
1892 multiline ? FBMrf_MULTILINE : 0))) ) {
1893 /* we may be pointing at the wrong string */
1894 if ((flags & REXEC_SCREAM) && RX_MATCH_COPIED(prog))
1895 s = strbeg + (s - SvPVX_const(sv));
1896 DEBUG_EXECUTE_r( did_match = 1 );
1897 if (HOPc(s, -back_max) > last1) {
1898 last1 = HOPc(s, -back_min);
1899 s = HOPc(s, -back_max);
1902 char *t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1;
1904 last1 = HOPc(s, -back_min);
1908 while (s <= last1) {
1909 if (regtry(prog, s))
1915 while (s <= last1) {
1916 if (regtry(prog, s))
1922 DEBUG_EXECUTE_r(if (!did_match)
1923 PerlIO_printf(Perl_debug_log,
1924 "Did not find %s substr \"%s%.*s%s\"%s...\n",
1925 ((must == prog->anchored_substr || must == prog->anchored_utf8)
1926 ? "anchored" : "floating"),
1928 (int)(SvCUR(must) - (SvTAIL(must)!=0)),
1930 PL_colors[1], (SvTAIL(must) ? "$" : ""))
1934 else if ((c = prog->regstclass)) {
1936 I32 op = (U8)OP(prog->regstclass);
1937 /* don't bother with what can't match */
1938 if (PL_regkind[op] != EXACT && op != CANY)
1939 strend = HOPc(strend, -(minlen - 1));
1942 SV *prop = sv_newmortal();
1950 pv_uni_display(dsv0, (U8*)SvPVX_const(prop), SvCUR(prop), 60,
1951 UNI_DISPLAY_REGEX) :
1953 len0 = UTF ? SvCUR(dsv0) : SvCUR(prop);
1955 sv_uni_display(dsv1, sv, 60, UNI_DISPLAY_REGEX) : s;
1956 len1 = UTF ? SvCUR(dsv1) : strend - s;
1957 PerlIO_printf(Perl_debug_log,
1958 "Matching stclass \"%*.*s\" against \"%*.*s\"\n",
1962 if (find_byclass(prog, c, s, strend, 0))
1964 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass...\n"));
1968 if (prog->float_substr != NULL || prog->float_utf8 != NULL) {
1973 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr))
1974 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1975 float_real = do_utf8 ? prog->float_utf8 : prog->float_substr;
1977 if (flags & REXEC_SCREAM) {
1978 last = screaminstr(sv, float_real, s - strbeg,
1979 end_shift, &scream_pos, 1); /* last one */
1981 last = scream_olds; /* Only one occurrence. */
1982 /* we may be pointing at the wrong string */
1983 else if (RX_MATCH_COPIED(prog))
1984 s = strbeg + (s - SvPVX_const(sv));
1988 const char * const little = SvPV_const(float_real, len);
1990 if (SvTAIL(float_real)) {
1991 if (memEQ(strend - len + 1, little, len - 1))
1992 last = strend - len + 1;
1993 else if (!multiline)
1994 last = memEQ(strend - len, little, len)
1995 ? strend - len : NULL;
2001 last = rninstr(s, strend, little, little + len);
2003 last = strend; /* matching "$" */
2007 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
2008 "%sCan't trim the tail, match fails (should not happen)%s\n",
2009 PL_colors[4], PL_colors[5]));
2010 goto phooey; /* Should not happen! */
2012 dontbother = strend - last + prog->float_min_offset;
2014 if (minlen && (dontbother < minlen))
2015 dontbother = minlen - 1;
2016 strend -= dontbother; /* this one's always in bytes! */
2017 /* We don't know much -- general case. */
2020 if (regtry(prog, s))
2029 if (regtry(prog, s))
2031 } while (s++ < strend);
2039 RX_MATCH_TAINTED_set(prog, PL_reg_flags & RF_tainted);
2041 if (PL_reg_eval_set) {
2042 /* Preserve the current value of $^R */
2043 if (oreplsv != GvSV(PL_replgv))
2044 sv_setsv(oreplsv, GvSV(PL_replgv));/* So that when GvSV(replgv) is
2045 restored, the value remains
2047 restore_pos(aTHX_ 0);
2050 /* make sure $`, $&, $', and $digit will work later */
2051 if ( !(flags & REXEC_NOT_FIRST) ) {
2052 RX_MATCH_COPY_FREE(prog);
2053 if (flags & REXEC_COPY_STR) {
2054 I32 i = PL_regeol - startpos + (stringarg - strbeg);
2055 #ifdef PERL_OLD_COPY_ON_WRITE
2057 || (SvFLAGS(sv) & CAN_COW_MASK) == CAN_COW_FLAGS)) {
2059 PerlIO_printf(Perl_debug_log,
2060 "Copy on write: regexp capture, type %d\n",
2063 prog->saved_copy = sv_setsv_cow(prog->saved_copy, sv);
2064 prog->subbeg = (char *)SvPVX_const(prog->saved_copy);
2065 assert (SvPOKp(prog->saved_copy));
2069 RX_MATCH_COPIED_on(prog);
2070 s = savepvn(strbeg, i);
2076 prog->subbeg = strbeg;
2077 prog->sublen = PL_regeol - strbeg; /* strend may have been modified */
2084 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n",
2085 PL_colors[4], PL_colors[5]));
2086 if (PL_reg_eval_set)
2087 restore_pos(aTHX_ 0);
2092 - regtry - try match at specific point
2094 STATIC I32 /* 0 failure, 1 success */
2095 S_regtry(pTHX_ regexp *prog, char *startpos)
2102 GET_RE_DEBUG_FLAGS_DECL;
2105 PL_regindent = 0; /* XXXX Not good when matches are reenterable... */
2107 if ((prog->reganch & ROPT_EVAL_SEEN) && !PL_reg_eval_set) {
2110 PL_reg_eval_set = RS_init;
2111 DEBUG_EXECUTE_r(DEBUG_s(
2112 PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %"IVdf"\n",
2113 (IV)(PL_stack_sp - PL_stack_base));
2115 SAVEI32(cxstack[cxstack_ix].blk_oldsp);
2116 cxstack[cxstack_ix].blk_oldsp = PL_stack_sp - PL_stack_base;
2117 /* Otherwise OP_NEXTSTATE will free whatever on stack now. */
2119 /* Apparently this is not needed, judging by wantarray. */
2120 /* SAVEI8(cxstack[cxstack_ix].blk_gimme);
2121 cxstack[cxstack_ix].blk_gimme = G_SCALAR; */
2124 /* Make $_ available to executed code. */
2125 if (PL_reg_sv != DEFSV) {
2130 if (!(SvTYPE(PL_reg_sv) >= SVt_PVMG && SvMAGIC(PL_reg_sv)
2131 && (mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global)))) {
2132 /* prepare for quick setting of pos */
2133 #ifdef PERL_OLD_COPY_ON_WRITE
2135 sv_force_normal_flags(sv, 0);
2137 mg = sv_magicext(PL_reg_sv, (SV*)0, PERL_MAGIC_regex_global,
2138 &PL_vtbl_mglob, NULL, 0);
2142 PL_reg_oldpos = mg->mg_len;
2143 SAVEDESTRUCTOR_X(restore_pos, 0);
2145 if (!PL_reg_curpm) {
2146 Newxz(PL_reg_curpm, 1, PMOP);
2149 SV* repointer = newSViv(0);
2150 /* so we know which PL_regex_padav element is PL_reg_curpm */
2151 SvFLAGS(repointer) |= SVf_BREAK;
2152 av_push(PL_regex_padav,repointer);
2153 PL_reg_curpm->op_pmoffset = av_len(PL_regex_padav);
2154 PL_regex_pad = AvARRAY(PL_regex_padav);
2158 PM_SETRE(PL_reg_curpm, prog);
2159 PL_reg_oldcurpm = PL_curpm;
2160 PL_curpm = PL_reg_curpm;
2161 if (RX_MATCH_COPIED(prog)) {
2162 /* Here is a serious problem: we cannot rewrite subbeg,
2163 since it may be needed if this match fails. Thus
2164 $` inside (?{}) could fail... */
2165 PL_reg_oldsaved = prog->subbeg;
2166 PL_reg_oldsavedlen = prog->sublen;
2167 #ifdef PERL_OLD_COPY_ON_WRITE
2168 PL_nrs = prog->saved_copy;
2170 RX_MATCH_COPIED_off(prog);
2173 PL_reg_oldsaved = NULL;
2174 prog->subbeg = PL_bostr;
2175 prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */
2177 prog->startp[0] = startpos - PL_bostr;
2178 PL_reginput = startpos;
2179 PL_regstartp = prog->startp;
2180 PL_regendp = prog->endp;
2181 PL_reglastparen = &prog->lastparen;
2182 PL_reglastcloseparen = &prog->lastcloseparen;
2183 prog->lastparen = 0;
2184 prog->lastcloseparen = 0;
2186 DEBUG_EXECUTE_r(PL_reg_starttry = startpos);
2187 if (PL_reg_start_tmpl <= prog->nparens) {
2188 PL_reg_start_tmpl = prog->nparens*3/2 + 3;
2189 if(PL_reg_start_tmp)
2190 Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
2192 Newx(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
2195 /* XXXX What this code is doing here?!!! There should be no need
2196 to do this again and again, PL_reglastparen should take care of
2199 /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code.
2200 * Actually, the code in regcppop() (which Ilya may be meaning by
2201 * PL_reglastparen), is not needed at all by the test suite
2202 * (op/regexp, op/pat, op/split), but that code is needed, oddly
2203 * enough, for building DynaLoader, or otherwise this
2204 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
2205 * will happen. Meanwhile, this code *is* needed for the
2206 * above-mentioned test suite tests to succeed. The common theme
2207 * on those tests seems to be returning null fields from matches.
2212 if (prog->nparens) {
2213 for (i = prog->nparens; i > (I32)*PL_reglastparen; i--) {
2220 if (regmatch(prog->program + 1)) {
2221 prog->endp[0] = PL_reginput - PL_bostr;
2224 REGCP_UNWIND(lastcp);
2228 #define RE_UNWIND_BRANCH 1
2229 #define RE_UNWIND_BRANCHJ 2
2233 typedef struct { /* XX: makes sense to enlarge it... */
2237 } re_unwind_generic_t;
2250 } re_unwind_branch_t;
2252 typedef union re_unwind_t {
2254 re_unwind_generic_t generic;
2255 re_unwind_branch_t branch;
2258 #define sayYES goto yes
2259 #define sayNO goto no
2260 #define sayNO_ANYOF goto no_anyof
2261 #define sayYES_FINAL goto yes_final
2262 #define sayYES_LOUD goto yes_loud
2263 #define sayNO_FINAL goto no_final
2264 #define sayNO_SILENT goto do_no
2265 #define saySAME(x) if (x) goto yes; else goto no
2267 #define POSCACHE_SUCCESS 0 /* caching success rather than failure */
2268 #define POSCACHE_SEEN 1 /* we know what we're caching */
2269 #define POSCACHE_START 2 /* the real cache: this bit maps to pos 0 */
2270 #define CACHEsayYES STMT_START { \
2271 if (cache_offset | cache_bit) { \
2272 if (!(PL_reg_poscache[0] & (1<<POSCACHE_SEEN))) \
2273 PL_reg_poscache[0] |= (1<<POSCACHE_SUCCESS) || (1<<POSCACHE_SEEN); \
2274 else if (!(PL_reg_poscache[0] & (1<<POSCACHE_SUCCESS))) { \
2275 /* cache records failure, but this is success */ \
2277 PerlIO_printf(Perl_debug_log, \
2278 "%*s (remove success from failure cache)\n", \
2279 REPORT_CODE_OFF+PL_regindent*2, "") \
2281 PL_reg_poscache[cache_offset] &= ~(1<<cache_bit); \
2286 #define CACHEsayNO STMT_START { \
2287 if (cache_offset | cache_bit) { \
2288 if (!(PL_reg_poscache[0] & (1<<POSCACHE_SEEN))) \
2289 PL_reg_poscache[0] |= (1<<POSCACHE_SEEN); \
2290 else if ((PL_reg_poscache[0] & (1<<POSCACHE_SUCCESS))) { \
2291 /* cache records success, but this is failure */ \
2293 PerlIO_printf(Perl_debug_log, \
2294 "%*s (remove failure from success cache)\n", \
2295 REPORT_CODE_OFF+PL_regindent*2, "") \
2297 PL_reg_poscache[cache_offset] &= ~(1<<cache_bit); \
2303 /* this is used to determine how far from the left messages like
2304 'failed...' are printed. Currently 29 makes these messages line
2305 up with the opcode they refer to. Earlier perls used 25 which
2306 left these messages outdented making reviewing a debug output
2309 #define REPORT_CODE_OFF 29
2312 /* Make sure there is a test for this +1 options in re_tests */
2313 #define TRIE_INITAL_ACCEPT_BUFFLEN 4;
2315 #define TRIE_CHECK_STATE_IS_ACCEPTING STMT_START { \
2316 if ( trie->states[ state ].wordnum ) { \
2317 if ( !accepted ) { \
2320 bufflen = TRIE_INITAL_ACCEPT_BUFFLEN ; \
2321 sv_accept_buff=newSV(bufflen * sizeof(reg_trie_accepted) - 1 );\
2322 SvCUR_set( sv_accept_buff, sizeof(reg_trie_accepted) ); \
2323 SvPOK_on( sv_accept_buff ); \
2324 sv_2mortal( sv_accept_buff ); \
2325 accept_buff = (reg_trie_accepted*)SvPV_nolen( sv_accept_buff );\
2327 if ( accepted >= bufflen ) { \
2329 accept_buff =(reg_trie_accepted*)SvGROW( sv_accept_buff, \
2330 bufflen * sizeof(reg_trie_accepted) ); \
2332 SvCUR_set( sv_accept_buff,SvCUR( sv_accept_buff ) \
2333 + sizeof( reg_trie_accepted ) ); \
2335 accept_buff[ accepted ].wordnum = trie->states[ state ].wordnum; \
2336 accept_buff[ accepted ].endpos = uc; \
2340 #define TRIE_HANDLE_CHAR STMT_START { \
2341 if ( uvc < 256 ) { \
2342 charid = trie->charmap[ uvc ]; \
2345 if( trie->widecharmap ) { \
2346 SV** svpp = (SV**)NULL; \
2347 svpp = hv_fetch( trie->widecharmap, (char*)&uvc, \
2348 sizeof( UV ), 0 ); \
2350 charid = (U16)SvIV( *svpp ); \
2355 ( base + charid > trie->uniquecharcount ) && \
2356 ( base + charid - 1 - trie->uniquecharcount < trie->lasttrans) && \
2357 trie->trans[ base + charid - 1 - trie->uniquecharcount ].check == state ) \
2359 state = trie->trans[ base + charid - 1 - trie->uniquecharcount ].next; \
2367 - regmatch - main matching routine
2369 * Conceptually the strategy is simple: check to see whether the current
2370 * node matches, call self recursively to see whether the rest matches,
2371 * and then act accordingly. In practice we make some effort to avoid
2372 * recursion, in particular by going through "ordinary" nodes (that don't
2373 * need to know whether the rest of the match failed) by a loop instead of
2376 /* [lwall] I've hoisted the register declarations to the outer block in order to
2377 * maybe save a little bit of pushing and popping on the stack. It also takes
2378 * advantage of machines that use a register save mask on subroutine entry.
2380 STATIC I32 /* 0 failure, 1 success */
2381 S_regmatch(pTHX_ regnode *prog)
2384 register regnode *scan; /* Current node. */
2385 regnode *next; /* Next node. */
2386 regnode *inner; /* Next node in internal branch. */
2387 register I32 nextchr; /* renamed nextchr - nextchar colides with
2388 function of same name */
2389 register I32 n; /* no or next */
2390 register I32 ln = 0; /* len or last */
2391 register char *s = NULL; /* operand or save */
2392 register char *locinput = PL_reginput;
2393 register I32 c1 = 0, c2 = 0, paren; /* case fold search, parenth */
2394 int minmod = 0, sw = 0, logical = 0;
2397 /* used by the trie code */
2398 SV *sv_accept_buff = NULL; /* accepting states we have traversed */
2399 reg_trie_accepted *accept_buff = NULL; /* "" */
2400 reg_trie_data *trie; /* what trie are we using right now */
2401 U32 accepted = 0; /* how many accepting states we have seen*/
2404 I32 firstcp = PL_savestack_ix;
2406 register const bool do_utf8 = PL_reg_match_utf8;
2408 SV * const dsv0 = PERL_DEBUG_PAD_ZERO(0);
2409 SV * const dsv1 = PERL_DEBUG_PAD_ZERO(1);
2410 SV * const dsv2 = PERL_DEBUG_PAD_ZERO(2);
2412 SV *re_debug_flags = NULL;
2414 U32 uniflags = ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY;
2423 /* Note that nextchr is a byte even in UTF */
2424 nextchr = UCHARAT(locinput);
2426 while (scan != NULL) {
2429 SV * const prop = sv_newmortal();
2430 const int docolor = *PL_colors[0];
2431 const int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */
2432 int l = (PL_regeol - locinput) > taill ? taill : (PL_regeol - locinput);
2433 /* The part of the string before starttry has one color
2434 (pref0_len chars), between starttry and current
2435 position another one (pref_len - pref0_len chars),
2436 after the current position the third one.
2437 We assume that pref0_len <= pref_len, otherwise we
2438 decrease pref0_len. */
2439 int pref_len = (locinput - PL_bostr) > (5 + taill) - l
2440 ? (5 + taill) - l : locinput - PL_bostr;
2443 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len)))
2445 pref0_len = pref_len - (locinput - PL_reg_starttry);
2446 if (l + pref_len < (5 + taill) && l < PL_regeol - locinput)
2447 l = ( PL_regeol - locinput > (5 + taill) - pref_len
2448 ? (5 + taill) - pref_len : PL_regeol - locinput);
2449 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput + l)))
2453 if (pref0_len > pref_len)
2454 pref0_len = pref_len;
2455 regprop(prop, scan);
2457 const char * const s0 =
2458 do_utf8 && OP(scan) != CANY ?
2459 pv_uni_display(dsv0, (U8*)(locinput - pref_len),
2460 pref0_len, 60, UNI_DISPLAY_REGEX) :
2461 locinput - pref_len;
2462 const int len0 = do_utf8 ? strlen(s0) : pref0_len;
2463 const char * const s1 = do_utf8 && OP(scan) != CANY ?
2464 pv_uni_display(dsv1, (U8*)(locinput - pref_len + pref0_len),
2465 pref_len - pref0_len, 60, UNI_DISPLAY_REGEX) :
2466 locinput - pref_len + pref0_len;
2467 const int len1 = do_utf8 ? strlen(s1) : pref_len - pref0_len;
2468 const char * const s2 = do_utf8 && OP(scan) != CANY ?
2469 pv_uni_display(dsv2, (U8*)locinput,
2470 PL_regeol - locinput, 60, UNI_DISPLAY_REGEX) :
2472 const int len2 = do_utf8 ? strlen(s2) : l;
2473 PerlIO_printf(Perl_debug_log,
2474 "%4"IVdf" <%s%.*s%s%s%.*s%s%s%s%.*s%s>%*s|%3"IVdf":%*s%s\n",
2475 (IV)(locinput - PL_bostr),
2482 (docolor ? "" : "> <"),
2486 15 - l - pref_len + 1,
2488 (IV)(scan - PL_regprogram), PL_regindent*2, "",
2493 next = scan + NEXT_OFF(scan);
2499 if (locinput == PL_bostr)
2501 /* regtill = regbol; */
2506 if (locinput == PL_bostr ||
2507 ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n'))
2513 if (locinput == PL_bostr)
2517 if (locinput == PL_reg_ganch)
2523 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2528 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2530 if (PL_regeol - locinput > 1)
2534 if (PL_regeol != locinput)
2538 if (!nextchr && locinput >= PL_regeol)
2541 locinput += PL_utf8skip[nextchr];
2542 if (locinput > PL_regeol)
2544 nextchr = UCHARAT(locinput);
2547 nextchr = UCHARAT(++locinput);
2550 if (!nextchr && locinput >= PL_regeol)
2552 nextchr = UCHARAT(++locinput);
2555 if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n')
2558 locinput += PL_utf8skip[nextchr];
2559 if (locinput > PL_regeol)
2561 nextchr = UCHARAT(locinput);
2564 nextchr = UCHARAT(++locinput);
2570 traverse the TRIE keeping track of all accepting states
2571 we transition through until we get to a failing node.
2573 we use two slightly different pieces of code to handle
2574 the traversal depending on whether its case sensitive or
2575 not. we reuse the accept code however. (this should probably
2576 be turned into a macro.)
2582 U8 *uc = ( U8* )locinput;
2589 U8 *uscan = (U8*)NULL;
2593 trie = (reg_trie_data*)PL_regdata->data[ ARG( scan ) ];
2595 while ( state && uc <= (U8*)PL_regeol ) {
2597 TRIE_CHECK_STATE_IS_ACCEPTING;
2599 base = trie->states[ state ].trans.base;
2601 DEBUG_TRIE_EXECUTE_r(
2602 PerlIO_printf( Perl_debug_log,
2603 "%*s %sState: %4"UVxf", Base: %4"UVxf", Accepted: %4"UVxf" ",
2604 REPORT_CODE_OFF + PL_regindent * 2, "", PL_colors[4],
2605 (UV)state, (UV)base, (UV)accepted );
2612 uvc = utf8n_to_uvuni( uscan, UTF8_MAXLEN, &len, uniflags );
2617 U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ];
2618 uvc = utf8n_to_uvuni( (U8*)uc, UTF8_MAXLEN, &len, uniflags );
2619 uvc = to_uni_fold( uvc, foldbuf, &foldlen );
2620 foldlen -= UNISKIP( uvc );
2621 uscan = foldbuf + UNISKIP( uvc );
2633 DEBUG_TRIE_EXECUTE_r(
2634 PerlIO_printf( Perl_debug_log,
2635 "Charid:%3x CV:%4"UVxf" After State: %4"UVxf"%s\n",
2636 charid, uvc, (UV)state, PL_colors[5] );
2645 /* unreached codepoint: we jump into the middle of the next case
2646 from previous if blocks */
2649 U8 *uc = (U8*)locinput;
2658 trie = (reg_trie_data*)PL_regdata->data[ ARG( scan ) ];
2660 while ( state && uc <= (U8*)PL_regeol ) {
2662 TRIE_CHECK_STATE_IS_ACCEPTING;
2664 base = trie->states[ state ].trans.base;
2666 DEBUG_TRIE_EXECUTE_r(
2667 PerlIO_printf( Perl_debug_log,
2668 "%*s %sState: %4"UVxf", Base: %4"UVxf", Accepted: %4"UVxf" ",
2669 REPORT_CODE_OFF + PL_regindent * 2, "", PL_colors[4],
2670 (UV)state, (UV)base, (UV)accepted );
2676 uvc = utf8n_to_uvuni( (U8*)uc, UTF8_MAXLEN, &len, uniflags );
2687 DEBUG_TRIE_EXECUTE_r(
2688 PerlIO_printf( Perl_debug_log,
2689 "Charid:%3x CV:%4"UVxf" After State: %4"UVxf"%s\n",
2690 charid, uvc, (UV)state, PL_colors[5] );
2700 There was at least one accepting state that we
2701 transitioned through. Presumably the number of accepting
2702 states is going to be low, typically one or two. So we
2703 simply scan through to find the one with lowest wordnum.
2704 Once we find it, we swap the last state into its place
2705 and decrement the size. We then try to match the rest of
2706 the pattern at the point where the word ends, if we
2707 succeed then we end the loop, otherwise the loop
2708 eventually terminates once all of the accepting states
2715 if ( accepted == 1 ) {
2717 SV **tmp = av_fetch( trie->words, accept_buff[ 0 ].wordnum-1, 0 );
2718 PerlIO_printf( Perl_debug_log,
2719 "%*s %sonly one match : #%d <%s>%s\n",
2720 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4],
2721 accept_buff[ 0 ].wordnum,
2722 tmp ? SvPV_nolen_const( *tmp ) : "not compiled under -Dr",
2725 PL_reginput = (char *)accept_buff[ 0 ].endpos;
2726 /* in this case we free tmps/leave before we call regmatch
2727 as we wont be using accept_buff again. */
2730 gotit = regmatch( scan + NEXT_OFF( scan ) );
2733 PerlIO_printf( Perl_debug_log,"%*s %sgot %"IVdf" possible matches%s\n",
2734 REPORT_CODE_OFF + PL_regindent * 2, "", PL_colors[4], (IV)accepted,
2737 while ( !gotit && accepted-- ) {
2740 for( cur = 1 ; cur <= accepted ; cur++ ) {
2741 DEBUG_TRIE_EXECUTE_r(
2742 PerlIO_printf( Perl_debug_log,
2743 "%*s %sgot %"IVdf" (%d) as best, looking at %"IVdf" (%d)%s\n",
2744 REPORT_CODE_OFF + PL_regindent * 2, "", PL_colors[4],
2745 (IV)best, accept_buff[ best ].wordnum, (IV)cur,
2746 accept_buff[ cur ].wordnum, PL_colors[5] );
2749 if ( accept_buff[ cur ].wordnum < accept_buff[ best ].wordnum )
2753 SV ** const tmp = av_fetch( trie->words, accept_buff[ best ].wordnum - 1, 0 );
2754 PerlIO_printf( Perl_debug_log, "%*s %strying alternation #%d <%s> at 0x%p%s\n",
2755 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4],
2756 accept_buff[best].wordnum,
2757 tmp ? SvPV_nolen_const( *tmp ) : "not compiled under -Dr",scan,
2760 if ( best<accepted ) {
2761 reg_trie_accepted tmp = accept_buff[ best ];
2762 accept_buff[ best ] = accept_buff[ accepted ];
2763 accept_buff[ accepted ] = tmp;
2766 PL_reginput = (char *)accept_buff[ best ].endpos;
2769 as far as I can tell we only need the SAVETMPS/FREETMPS
2770 for re's with EVAL in them but I'm leaving them in for
2771 all until I can be sure.
2774 gotit = regmatch( scan + NEXT_OFF( scan ) ) ;
2787 /* unreached codepoint */
2791 if (do_utf8 != UTF) {
2792 /* The target and the pattern have differing utf8ness. */
2794 const char *e = s + ln;
2797 /* The target is utf8, the pattern is not utf8. */
2802 if (NATIVE_TO_UNI(*(U8*)s) !=
2803 utf8n_to_uvuni((U8*)l, UTF8_MAXBYTES, &ulen,
2811 /* The target is not utf8, the pattern is utf8. */
2816 if (NATIVE_TO_UNI(*((U8*)l)) !=
2817 utf8n_to_uvuni((U8*)s, UTF8_MAXBYTES, &ulen,
2825 nextchr = UCHARAT(locinput);
2828 /* The target and the pattern have the same utf8ness. */
2829 /* Inline the first character, for speed. */
2830 if (UCHARAT(s) != nextchr)
2832 if (PL_regeol - locinput < ln)
2834 if (ln > 1 && memNE(s, locinput, ln))
2837 nextchr = UCHARAT(locinput);
2840 PL_reg_flags |= RF_tainted;
2846 if (do_utf8 || UTF) {
2847 /* Either target or the pattern are utf8. */
2849 char *e = PL_regeol;
2851 if (ibcmp_utf8(s, 0, ln, (bool)UTF,
2852 l, &e, 0, do_utf8)) {
2853 /* One more case for the sharp s:
2854 * pack("U0U*", 0xDF) =~ /ss/i,
2855 * the 0xC3 0x9F are the UTF-8
2856 * byte sequence for the U+00DF. */
2858 toLOWER(s[0]) == 's' &&
2860 toLOWER(s[1]) == 's' &&
2867 nextchr = UCHARAT(locinput);
2871 /* Neither the target and the pattern are utf8. */
2873 /* Inline the first character, for speed. */
2874 if (UCHARAT(s) != nextchr &&
2875 UCHARAT(s) != ((OP(scan) == EXACTF)
2876 ? PL_fold : PL_fold_locale)[nextchr])
2878 if (PL_regeol - locinput < ln)
2880 if (ln > 1 && (OP(scan) == EXACTF
2881 ? ibcmp(s, locinput, ln)
2882 : ibcmp_locale(s, locinput, ln)))
2885 nextchr = UCHARAT(locinput);
2889 STRLEN inclasslen = PL_regeol - locinput;
2891 if (!reginclass(scan, (U8*)locinput, &inclasslen, do_utf8))
2893 if (locinput >= PL_regeol)
2895 locinput += inclasslen ? inclasslen : UTF8SKIP(locinput);
2896 nextchr = UCHARAT(locinput);
2901 nextchr = UCHARAT(locinput);
2902 if (!REGINCLASS(scan, (U8*)locinput))
2904 if (!nextchr && locinput >= PL_regeol)
2906 nextchr = UCHARAT(++locinput);
2910 /* If we might have the case of the German sharp s
2911 * in a casefolding Unicode character class. */
2913 if (ANYOF_FOLD_SHARP_S(scan, locinput, PL_regeol)) {
2914 locinput += SHARP_S_SKIP;
2915 nextchr = UCHARAT(locinput);
2921 PL_reg_flags |= RF_tainted;
2927 LOAD_UTF8_CHARCLASS_ALNUM();
2928 if (!(OP(scan) == ALNUM
2929 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2930 : isALNUM_LC_utf8((U8*)locinput)))
2934 locinput += PL_utf8skip[nextchr];
2935 nextchr = UCHARAT(locinput);
2938 if (!(OP(scan) == ALNUM
2939 ? isALNUM(nextchr) : isALNUM_LC(nextchr)))
2941 nextchr = UCHARAT(++locinput);
2944 PL_reg_flags |= RF_tainted;
2947 if (!nextchr && locinput >= PL_regeol)
2950 LOAD_UTF8_CHARCLASS_ALNUM();
2951 if (OP(scan) == NALNUM
2952 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2953 : isALNUM_LC_utf8((U8*)locinput))
2957 locinput += PL_utf8skip[nextchr];
2958 nextchr = UCHARAT(locinput);
2961 if (OP(scan) == NALNUM
2962 ? isALNUM(nextchr) : isALNUM_LC(nextchr))
2964 nextchr = UCHARAT(++locinput);
2968 PL_reg_flags |= RF_tainted;
2972 /* was last char in word? */
2974 if (locinput == PL_bostr)
2977 const U8 * const r = reghop3((U8*)locinput, -1, (U8*)PL_bostr);
2979 ln = utf8n_to_uvchr(r, UTF8SKIP(r), 0, 0);
2981 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2982 ln = isALNUM_uni(ln);
2983 LOAD_UTF8_CHARCLASS_ALNUM();
2984 n = swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8);
2987 ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(ln));
2988 n = isALNUM_LC_utf8((U8*)locinput);
2992 ln = (locinput != PL_bostr) ?
2993 UCHARAT(locinput - 1) : '\n';
2994 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2996 n = isALNUM(nextchr);
2999 ln = isALNUM_LC(ln);
3000 n = isALNUM_LC(nextchr);
3003 if (((!ln) == (!n)) == (OP(scan) == BOUND ||
3004 OP(scan) == BOUNDL))
3008 PL_reg_flags |= RF_tainted;
3014 if (UTF8_IS_CONTINUED(nextchr)) {
3015 LOAD_UTF8_CHARCLASS_SPACE();
3016 if (!(OP(scan) == SPACE
3017 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
3018 : isSPACE_LC_utf8((U8*)locinput)))
3022 locinput += PL_utf8skip[nextchr];
3023 nextchr = UCHARAT(locinput);
3026 if (!(OP(scan) == SPACE
3027 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
3029 nextchr = UCHARAT(++locinput);
3032 if (!(OP(scan) == SPACE
3033 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
3035 nextchr = UCHARAT(++locinput);
3039 PL_reg_flags |= RF_tainted;
3042 if (!nextchr && locinput >= PL_regeol)
3045 LOAD_UTF8_CHARCLASS_SPACE();
3046 if (OP(scan) == NSPACE
3047 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
3048 : isSPACE_LC_utf8((U8*)locinput))
3052 locinput += PL_utf8skip[nextchr];
3053 nextchr = UCHARAT(locinput);
3056 if (OP(scan) == NSPACE
3057 ? isSPACE(nextchr) : isSPACE_LC(nextchr))
3059 nextchr = UCHARAT(++locinput);
3062 PL_reg_flags |= RF_tainted;
3068 LOAD_UTF8_CHARCLASS_DIGIT();
3069 if (!(OP(scan) == DIGIT
3070 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
3071 : isDIGIT_LC_utf8((U8*)locinput)))
3075 locinput += PL_utf8skip[nextchr];
3076 nextchr = UCHARAT(locinput);
3079 if (!(OP(scan) == DIGIT
3080 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr)))
3082 nextchr = UCHARAT(++locinput);
3085 PL_reg_flags |= RF_tainted;
3088 if (!nextchr && locinput >= PL_regeol)
3091 LOAD_UTF8_CHARCLASS_DIGIT();
3092 if (OP(scan) == NDIGIT
3093 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
3094 : isDIGIT_LC_utf8((U8*)locinput))
3098 locinput += PL_utf8skip[nextchr];
3099 nextchr = UCHARAT(locinput);
3102 if (OP(scan) == NDIGIT
3103 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr))
3105 nextchr = UCHARAT(++locinput);
3108 if (locinput >= PL_regeol)
3111 LOAD_UTF8_CHARCLASS_MARK();
3112 if (swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
3114 locinput += PL_utf8skip[nextchr];
3115 while (locinput < PL_regeol &&
3116 swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
3117 locinput += UTF8SKIP(locinput);
3118 if (locinput > PL_regeol)
3123 nextchr = UCHARAT(locinput);
3126 PL_reg_flags |= RF_tainted;
3130 n = ARG(scan); /* which paren pair */
3131 ln = PL_regstartp[n];
3132 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
3133 if ((I32)*PL_reglastparen < n || ln == -1)
3134 sayNO; /* Do not match unless seen CLOSEn. */
3135 if (ln == PL_regendp[n])
3139 if (do_utf8 && OP(scan) != REF) { /* REF can do byte comparison */
3141 const char *e = PL_bostr + PL_regendp[n];
3143 * Note that we can't do the "other character" lookup trick as
3144 * in the 8-bit case (no pun intended) because in Unicode we
3145 * have to map both upper and title case to lower case.
3147 if (OP(scan) == REFF) {
3149 STRLEN ulen1, ulen2;
3150 U8 tmpbuf1[UTF8_MAXBYTES_CASE+1];
3151 U8 tmpbuf2[UTF8_MAXBYTES_CASE+1];
3155 toLOWER_utf8((U8*)s, tmpbuf1, &ulen1);
3156 toLOWER_utf8((U8*)l, tmpbuf2, &ulen2);
3157 if (ulen1 != ulen2 || memNE((char *)tmpbuf1, (char *)tmpbuf2, ulen1))
3164 nextchr = UCHARAT(locinput);
3168 /* Inline the first character, for speed. */
3169 if (UCHARAT(s) != nextchr &&
3171 (UCHARAT(s) != ((OP(scan) == REFF
3172 ? PL_fold : PL_fold_locale)[nextchr]))))
3174 ln = PL_regendp[n] - ln;
3175 if (locinput + ln > PL_regeol)
3177 if (ln > 1 && (OP(scan) == REF
3178 ? memNE(s, locinput, ln)
3180 ? ibcmp(s, locinput, ln)
3181 : ibcmp_locale(s, locinput, ln))))
3184 nextchr = UCHARAT(locinput);
3195 OP_4tree * const oop = PL_op;
3196 COP * const ocurcop = PL_curcop;
3199 struct regexp * const oreg = PL_reg_re;
3202 PL_op = (OP_4tree*)PL_regdata->data[n];
3203 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) );
3204 PAD_SAVE_LOCAL(old_comppad, (PAD*)PL_regdata->data[n + 2]);
3205 PL_regendp[0] = PL_reg_magic->mg_len = locinput - PL_bostr;
3208 SV ** const before = SP;
3209 CALLRUNOPS(aTHX); /* Scalar context. */
3212 ret = &PL_sv_undef; /* protect against empty (?{}) blocks. */
3220 PAD_RESTORE_LOCAL(old_comppad);
3221 PL_curcop = ocurcop;
3223 if (logical == 2) { /* Postponed subexpression. */
3227 CHECKPOINT cp, lastcp;
3231 if(SvROK(ret) && SvSMAGICAL(sv = SvRV(ret)))
3232 mg = mg_find(sv, PERL_MAGIC_qr);
3233 else if (SvSMAGICAL(ret)) {
3234 if (SvGMAGICAL(ret))
3235 sv_unmagic(ret, PERL_MAGIC_qr);
3237 mg = mg_find(ret, PERL_MAGIC_qr);
3241 re = (regexp *)mg->mg_obj;
3242 (void)ReREFCNT_inc(re);
3246 const char * const t = SvPV_const(ret, len);
3248 char * const oprecomp = PL_regprecomp;
3249 const I32 osize = PL_regsize;
3250 const I32 onpar = PL_regnpar;
3253 if (DO_UTF8(ret)) pm.op_pmdynflags |= PMdf_DYN_UTF8;
3254 re = CALLREGCOMP(aTHX_ (char*)t, (char*)t + len, &pm);
3256 & (SVs_TEMP | SVs_PADTMP | SVf_READONLY
3258 sv_magic(ret,(SV*)ReREFCNT_inc(re),
3260 PL_regprecomp = oprecomp;
3265 PerlIO_printf(Perl_debug_log,
3266 "Entering embedded \"%s%.60s%s%s\"\n",
3270 (strlen(re->precomp) > 60 ? "..." : ""))
3273 state.prev = PL_reg_call_cc;
3274 state.cc = PL_regcc;
3275 state.re = PL_reg_re;
3279 cp = regcppush(0); /* Save *all* the positions. */
3282 state.ss = PL_savestack_ix;
3283 *PL_reglastparen = 0;
3284 *PL_reglastcloseparen = 0;
3285 PL_reg_call_cc = &state;
3286 PL_reginput = locinput;
3287 toggleutf = ((PL_reg_flags & RF_utf8) != 0) ^
3288 ((re->reganch & ROPT_UTF8) != 0);
3289 if (toggleutf) PL_reg_flags ^= RF_utf8;
3291 /* XXXX This is too dramatic a measure... */
3294 if (regmatch(re->program + 1)) {
3295 /* Even though we succeeded, we need to restore
3296 global variables, since we may be wrapped inside
3297 SUSPEND, thus the match may be not finished yet. */
3299 /* XXXX Do this only if SUSPENDed? */
3300 PL_reg_call_cc = state.prev;
3301 PL_regcc = state.cc;
3302 PL_reg_re = state.re;
3303 cache_re(PL_reg_re);
3304 if (toggleutf) PL_reg_flags ^= RF_utf8;
3306 /* XXXX This is too dramatic a measure... */
3309 /* These are needed even if not SUSPEND. */
3315 REGCP_UNWIND(lastcp);
3317 PL_reg_call_cc = state.prev;
3318 PL_regcc = state.cc;
3319 PL_reg_re = state.re;
3320 cache_re(PL_reg_re);
3321 if (toggleutf) PL_reg_flags ^= RF_utf8;
3323 /* XXXX This is too dramatic a measure... */
3333 sv_setsv(save_scalar(PL_replgv), ret);
3339 n = ARG(scan); /* which paren pair */
3340 PL_reg_start_tmp[n] = locinput;
3345 n = ARG(scan); /* which paren pair */
3346 PL_regstartp[n] = PL_reg_start_tmp[n] - PL_bostr;
3347 PL_regendp[n] = locinput - PL_bostr;
3348 if (n > (I32)*PL_reglastparen)
3349 *PL_reglastparen = n;
3350 *PL_reglastcloseparen = n;
3353 n = ARG(scan); /* which paren pair */
3354 sw = ((I32)*PL_reglastparen >= n && PL_regendp[n] != -1);
3357 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
3359 next = NEXTOPER(NEXTOPER(scan));
3361 next = scan + ARG(scan);
3362 if (OP(next) == IFTHEN) /* Fake one. */
3363 next = NEXTOPER(NEXTOPER(next));
3367 logical = scan->flags;
3369 /*******************************************************************
3370 PL_regcc contains infoblock about the innermost (...)* loop, and
3371 a pointer to the next outer infoblock.
3373 Here is how Y(A)*Z is processed (if it is compiled into CURLYX/WHILEM):
3375 1) After matching X, regnode for CURLYX is processed;
3377 2) This regnode creates infoblock on the stack, and calls
3378 regmatch() recursively with the starting point at WHILEM node;
3380 3) Each hit of WHILEM node tries to match A and Z (in the order
3381 depending on the current iteration, min/max of {min,max} and
3382 greediness). The information about where are nodes for "A"
3383 and "Z" is read from the infoblock, as is info on how many times "A"
3384 was already matched, and greediness.
3386 4) After A matches, the same WHILEM node is hit again.
3388 5) Each time WHILEM is hit, PL_regcc is the infoblock created by CURLYX
3389 of the same pair. Thus when WHILEM tries to match Z, it temporarily
3390 resets PL_regcc, since this Y(A)*Z can be a part of some other loop:
3391 as in (Y(A)*Z)*. If Z matches, the automaton will hit the WHILEM node
3392 of the external loop.
3394 Currently present infoblocks form a tree with a stem formed by PL_curcc
3395 and whatever it mentions via ->next, and additional attached trees
3396 corresponding to temporarily unset infoblocks as in "5" above.
3398 In the following picture infoblocks for outer loop of
3399 (Y(A)*?Z)*?T are denoted O, for inner I. NULL starting block
3400 is denoted by x. The matched string is YAAZYAZT. Temporarily postponed
3401 infoblocks are drawn below the "reset" infoblock.
3403 In fact in the picture below we do not show failed matches for Z and T
3404 by WHILEM blocks. [We illustrate minimal matches, since for them it is
3405 more obvious *why* one needs to *temporary* unset infoblocks.]
3407 Matched REx position InfoBlocks Comment
3411 Y A)*?Z)*?T x <- O <- I
3412 YA )*?Z)*?T x <- O <- I
3413 YA A)*?Z)*?T x <- O <- I
3414 YAA )*?Z)*?T x <- O <- I
3415 YAA Z)*?T x <- O # Temporary unset I
3418 YAAZ Y(A)*?Z)*?T x <- O
3421 YAAZY (A)*?Z)*?T x <- O
3424 YAAZY A)*?Z)*?T x <- O <- I
3427 YAAZYA )*?Z)*?T x <- O <- I
3430 YAAZYA Z)*?T x <- O # Temporary unset I
3436 YAAZYAZ T x # Temporary unset O
3443 *******************************************************************/
3446 CHECKPOINT cp = PL_savestack_ix;
3447 /* No need to save/restore up to this paren */
3448 I32 parenfloor = scan->flags;
3450 if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */
3452 cc.oldcc = PL_regcc;
3454 /* XXXX Probably it is better to teach regpush to support
3455 parenfloor > PL_regsize... */
3456 if (parenfloor > (I32)*PL_reglastparen)
3457 parenfloor = *PL_reglastparen; /* Pessimization... */
3458 cc.parenfloor = parenfloor;
3460 cc.min = ARG1(scan);
3461 cc.max = ARG2(scan);
3462 cc.scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
3466 PL_reginput = locinput;
3467 n = regmatch(PREVOPER(next)); /* start on the WHILEM */
3469 PL_regcc = cc.oldcc;
3475 * This is really hard to understand, because after we match
3476 * what we're trying to match, we must make sure the rest of
3477 * the REx is going to match for sure, and to do that we have
3478 * to go back UP the parse tree by recursing ever deeper. And
3479 * if it fails, we have to reset our parent's current state
3480 * that we can try again after backing off.
3483 CHECKPOINT cp, lastcp;
3484 CURCUR* cc = PL_regcc;
3485 char * const lastloc = cc->lastloc; /* Detection of 0-len. */
3486 I32 cache_offset = 0, cache_bit = 0;
3488 n = cc->cur + 1; /* how many we know we matched */
3489 PL_reginput = locinput;
3492 PerlIO_printf(Perl_debug_log,
3493 "%*s %ld out of %ld..%ld cc=%"UVxf"\n",
3494 REPORT_CODE_OFF+PL_regindent*2, "",
3495 (long)n, (long)cc->min,
3496 (long)cc->max, PTR2UV(cc))
3499 /* If degenerate scan matches "", assume scan done. */
3501 if (locinput == cc->lastloc && n >= cc->min) {
3502 PL_regcc = cc->oldcc;
3506 PerlIO_printf(Perl_debug_log,
3507 "%*s empty match detected, try continuation...\n",
3508 REPORT_CODE_OFF+PL_regindent*2, "")
3510 if (regmatch(cc->next))
3518 /* First just match a string of min scans. */
3522 cc->lastloc = locinput;
3523 if (regmatch(cc->scan))
3526 cc->lastloc = lastloc;
3531 /* Check whether we already were at this position.
3532 Postpone detection until we know the match is not
3533 *that* much linear. */
3534 if (!PL_reg_maxiter) {
3535 PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4);
3536 PL_reg_leftiter = PL_reg_maxiter;
3538 if (PL_reg_leftiter-- == 0) {
3539 const I32 size = (PL_reg_maxiter + 7 + POSCACHE_START)/8;
3540 if (PL_reg_poscache) {
3541 if ((I32)PL_reg_poscache_size < size) {
3542 Renew(PL_reg_poscache, size, char);
3543 PL_reg_poscache_size = size;
3545 Zero(PL_reg_poscache, size, char);
3548 PL_reg_poscache_size = size;
3549 Newxz(PL_reg_poscache, size, char);
3552 PerlIO_printf(Perl_debug_log,
3553 "%sDetected a super-linear match, switching on caching%s...\n",
3554 PL_colors[4], PL_colors[5])
3557 if (PL_reg_leftiter < 0) {
3558 cache_offset = locinput - PL_bostr;
3560 cache_offset = (scan->flags & 0xf) - 1 + POSCACHE_START
3561 + cache_offset * (scan->flags>>4);
3562 cache_bit = cache_offset % 8;
3564 if (PL_reg_poscache[cache_offset] & (1<<cache_bit)) {
3566 PerlIO_printf(Perl_debug_log,
3567 "%*s already tried at this position...\n",
3568 REPORT_CODE_OFF+PL_regindent*2, "")
3570 if (PL_reg_poscache[0] & (1<<POSCACHE_SUCCESS))
3571 /* cache records success */
3574 /* cache records failure */
3577 PL_reg_poscache[cache_offset] |= (1<<cache_bit);
3581 /* Prefer next over scan for minimal matching. */
3584 PL_regcc = cc->oldcc;
3587 cp = regcppush(cc->parenfloor);
3589 if (regmatch(cc->next)) {
3591 CACHEsayYES; /* All done. */
3593 REGCP_UNWIND(lastcp);
3599 if (n >= cc->max) { /* Maximum greed exceeded? */
3600 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3601 && !(PL_reg_flags & RF_warned)) {
3602 PL_reg_flags |= RF_warned;
3603 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s limit (%d) exceeded",
3604 "Complex regular subexpression recursion",
3611 PerlIO_printf(Perl_debug_log,
3612 "%*s trying longer...\n",
3613 REPORT_CODE_OFF+PL_regindent*2, "")
3615 /* Try scanning more and see if it helps. */
3616 PL_reginput = locinput;
3618 cc->lastloc = locinput;
3619 cp = regcppush(cc->parenfloor);
3621 if (regmatch(cc->scan)) {
3625 REGCP_UNWIND(lastcp);
3628 cc->lastloc = lastloc;
3632 /* Prefer scan over next for maximal matching. */
3634 if (n < cc->max) { /* More greed allowed? */
3635 cp = regcppush(cc->parenfloor);
3637 cc->lastloc = locinput;
3639 if (regmatch(cc->scan)) {
3643 REGCP_UNWIND(lastcp);
3644 regcppop(); /* Restore some previous $<digit>s? */
3645 PL_reginput = locinput;
3647 PerlIO_printf(Perl_debug_log,
3648 "%*s failed, try continuation...\n",
3649 REPORT_CODE_OFF+PL_regindent*2, "")
3652 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3653 && !(PL_reg_flags & RF_warned)) {
3654 PL_reg_flags |= RF_warned;
3655 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s limit (%d) exceeded",
3656 "Complex regular subexpression recursion",
3660 /* Failed deeper matches of scan, so see if this one works. */
3661 PL_regcc = cc->oldcc;
3664 if (regmatch(cc->next))
3670 cc->lastloc = lastloc;
3675 next = scan + ARG(scan);
3678 inner = NEXTOPER(NEXTOPER(scan));
3681 inner = NEXTOPER(scan);
3685 if (OP(next) != c1) /* No choice. */
3686 next = inner; /* Avoid recursion. */
3688 const I32 lastparen = *PL_reglastparen;
3689 /* Put unwinding data on stack */
3690 const I32 unwind1 = SSNEWt(1,re_unwind_branch_t);
3691 re_unwind_branch_t * const uw = SSPTRt(unwind1,re_unwind_branch_t);
3695 uw->type = ((c1 == BRANCH)
3697 : RE_UNWIND_BRANCHJ);
3698 uw->lastparen = lastparen;
3700 uw->locinput = locinput;
3701 uw->nextchr = nextchr;
3703 uw->regindent = ++PL_regindent;
3706 REGCP_SET(uw->lastcp);
3708 /* Now go into the first branch */
3723 /* We suppose that the next guy does not need
3724 backtracking: in particular, it is of constant non-zero length,
3725 and has no parenths to influence future backrefs. */
3726 ln = ARG1(scan); /* min to match */
3727 n = ARG2(scan); /* max to match */
3728 paren = scan->flags;
3730 if (paren > PL_regsize)
3732 if (paren > (I32)*PL_reglastparen)
3733 *PL_reglastparen = paren;
3735 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3737 scan += NEXT_OFF(scan); /* Skip former OPEN. */
3738 PL_reginput = locinput;
3739 maxwanted = minmod ? ln : n;
3741 while (PL_reginput < PL_regeol && matches < maxwanted) {
3742 if (!regmatch(scan))
3744 /* on first match, determine length, l */
3746 if (PL_reg_match_utf8) {
3748 while (s < PL_reginput) {
3754 l = PL_reginput - locinput;
3757 matches = maxwanted;
3761 locinput = PL_reginput;
3765 PL_reginput = locinput;
3769 if (ln && matches < ln)
3771 if (HAS_TEXT(next) || JUMPABLE(next)) {
3772 regnode *text_node = next;
3774 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3776 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3778 if (PL_regkind[(U8)OP(text_node)] == REF) {
3782 else { c1 = (U8)*STRING(text_node); }
3783 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3785 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3786 c2 = PL_fold_locale[c1];
3795 while (n >= ln || (n == REG_INFTY && ln > 0)) { /* ln overflow ? */
3796 /* If it could work, try it. */
3798 UCHARAT(PL_reginput) == c1 ||
3799 UCHARAT(PL_reginput) == c2)
3803 PL_regstartp[paren] =
3804 HOPc(PL_reginput, -l) - PL_bostr;
3805 PL_regendp[paren] = PL_reginput - PL_bostr;
3808 PL_regendp[paren] = -1;
3812 REGCP_UNWIND(lastcp);
3814 /* Couldn't or didn't -- move forward. */
3815 PL_reginput = locinput;
3816 if (regmatch(scan)) {
3818 locinput = PL_reginput;
3826 PerlIO_printf(Perl_debug_log,
3827 "%*s matched %"IVdf" times, len=%"IVdf"...\n",
3828 (int)(REPORT_CODE_OFF+PL_regindent*2), "",
3829 (IV) matches, (IV)l)
3831 if (matches >= ln) {
3832 if (HAS_TEXT(next) || JUMPABLE(next)) {
3833 regnode *text_node = next;
3835 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3837 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3839 if (PL_regkind[(U8)OP(text_node)] == REF) {
3843 else { c1 = (U8)*STRING(text_node); }
3845 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3847 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3848 c2 = PL_fold_locale[c1];
3858 while (matches >= ln) {
3859 /* If it could work, try it. */
3861 UCHARAT(PL_reginput) == c1 ||
3862 UCHARAT(PL_reginput) == c2)
3865 PerlIO_printf(Perl_debug_log,
3866 "%*s trying tail with matches=%"IVdf"...\n",
3867 (int)(REPORT_CODE_OFF+PL_regindent*2),
3872 PL_regstartp[paren] = HOPc(PL_reginput, -l) - PL_bostr;
3873 PL_regendp[paren] = PL_reginput - PL_bostr;
3876 PL_regendp[paren] = -1;
3880 REGCP_UNWIND(lastcp);
3882 /* Couldn't or didn't -- back up. */
3884 locinput = HOPc(locinput, -l);
3885 PL_reginput = locinput;
3893 paren = scan->flags; /* Which paren to set */
3894 if (paren > PL_regsize)
3896 if (paren > (I32)*PL_reglastparen)
3897 *PL_reglastparen = paren;
3898 ln = ARG1(scan); /* min to match */
3899 n = ARG2(scan); /* max to match */
3900 scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE);
3904 ln = ARG1(scan); /* min to match */
3905 n = ARG2(scan); /* max to match */
3906 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3911 scan = NEXTOPER(scan);
3917 scan = NEXTOPER(scan);
3921 * Lookahead to avoid useless match attempts
3922 * when we know what character comes next.
3926 * Used to only do .*x and .*?x, but now it allows
3927 * for )'s, ('s and (?{ ... })'s to be in the way
3928 * of the quantifier and the EXACT-like node. -- japhy
3931 if (HAS_TEXT(next) || JUMPABLE(next)) {
3933 regnode *text_node = next;
3935 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3937 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3939 if (PL_regkind[(U8)OP(text_node)] == REF) {
3941 goto assume_ok_easy;
3943 else { s = (U8*)STRING(text_node); }
3947 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3949 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3950 c2 = PL_fold_locale[c1];
3953 if (OP(text_node) == EXACTF || OP(text_node) == REFF) {
3954 STRLEN ulen1, ulen2;
3955 U8 tmpbuf1[UTF8_MAXBYTES_CASE+1];
3956 U8 tmpbuf2[UTF8_MAXBYTES_CASE+1];
3958 to_utf8_lower((U8*)s, tmpbuf1, &ulen1);
3959 to_utf8_upper((U8*)s, tmpbuf2, &ulen2);
3961 c1 = utf8n_to_uvuni(tmpbuf1, UTF8_MAXBYTES, 0,
3963 c2 = utf8n_to_uvuni(tmpbuf2, UTF8_MAXBYTES, 0,
3967 c2 = c1 = utf8n_to_uvchr(s, UTF8_MAXBYTES, 0,
3976 PL_reginput = locinput;
3980 if (ln && regrepeat(scan, ln) < ln)
3982 locinput = PL_reginput;
3985 char *e; /* Should not check after this */
3986 char *old = locinput;
3989 if (n == REG_INFTY) {
3992 while (UTF8_IS_CONTINUATION(*(U8*)e))
3998 m >0 && e + UTF8SKIP(e) <= PL_regeol; m--)
4002 e = locinput + n - ln;
4007 /* Find place 'next' could work */
4010 while (locinput <= e &&
4011 UCHARAT(locinput) != c1)
4014 while (locinput <= e
4015 && UCHARAT(locinput) != c1
4016 && UCHARAT(locinput) != c2)
4019 count = locinput - old;
4024 /* count initialised to
4025 * utf8_distance(old, locinput) */
4026 while (locinput <= e &&
4027 utf8n_to_uvchr((U8*)locinput,
4028 UTF8_MAXBYTES, &len,
4029 uniflags) != (UV)c1) {
4035 /* count initialised to
4036 * utf8_distance(old, locinput) */
4037 while (locinput <= e) {
4038 UV c = utf8n_to_uvchr((U8*)locinput,
4039 UTF8_MAXBYTES, &len,
4041 if (c == (UV)c1 || c == (UV)c2)
4050 /* PL_reginput == old now */
4051 if (locinput != old) {
4052 ln = 1; /* Did some */
4053 if (regrepeat(scan, count) < count)
4056 /* PL_reginput == locinput now */
4057 TRYPAREN(paren, ln, locinput);
4058 PL_reginput = locinput; /* Could be reset... */
4059 REGCP_UNWIND(lastcp);
4060 /* Couldn't or didn't -- move forward. */
4063 locinput += UTF8SKIP(locinput);
4070 while (n >= ln || (n == REG_INFTY && ln > 0)) { /* ln overflow ? */
4074 c = utf8n_to_uvchr((U8*)PL_reginput,
4078 c = UCHARAT(PL_reginput);
4079 /* If it could work, try it. */
4080 if (c == (UV)c1 || c == (UV)c2)
4082 TRYPAREN(paren, ln, PL_reginput);
4083 REGCP_UNWIND(lastcp);
4086 /* If it could work, try it. */
4087 else if (c1 == -1000)
4089 TRYPAREN(paren, ln, PL_reginput);
4090 REGCP_UNWIND(lastcp);
4092 /* Couldn't or didn't -- move forward. */
4093 PL_reginput = locinput;
4094 if (regrepeat(scan, 1)) {
4096 locinput = PL_reginput;
4104 n = regrepeat(scan, n);
4105 locinput = PL_reginput;
4106 if (ln < n && PL_regkind[(U8)OP(next)] == EOL &&
4107 (OP(next) != MEOL ||
4108 OP(next) == SEOL || OP(next) == EOS))
4110 ln = n; /* why back off? */
4111 /* ...because $ and \Z can match before *and* after
4112 newline at the end. Consider "\n\n" =~ /\n+\Z\n/.
4113 We should back off by one in this case. */
4114 if (UCHARAT(PL_reginput - 1) == '\n' && OP(next) != EOS)
4123 c = utf8n_to_uvchr((U8*)PL_reginput,
4127 c = UCHARAT(PL_reginput);
4129 /* If it could work, try it. */
4130 if (c1 == -1000 || c == (UV)c1 || c == (UV)c2)
4132 TRYPAREN(paren, n, PL_reginput);
4133 REGCP_UNWIND(lastcp);
4135 /* Couldn't or didn't -- back up. */
4137 PL_reginput = locinput = HOPc(locinput, -1);
4145 c = utf8n_to_uvchr((U8*)PL_reginput,
4149 c = UCHARAT(PL_reginput);
4151 /* If it could work, try it. */
4152 if (c1 == -1000 || c == (UV)c1 || c == (UV)c2)
4154 TRYPAREN(paren, n, PL_reginput);
4155 REGCP_UNWIND(lastcp);
4157 /* Couldn't or didn't -- back up. */
4159 PL_reginput = locinput = HOPc(locinput, -1);
4166 if (PL_reg_call_cc) {
4167 re_cc_state *cur_call_cc = PL_reg_call_cc;
4168 CURCUR *cctmp = PL_regcc;
4169 regexp *re = PL_reg_re;
4173 /* Save *all* the positions. */
4174 const CHECKPOINT cp = regcppush(0);
4177 /* Restore parens of the caller. */
4178 tmp = PL_savestack_ix;
4179 PL_savestack_ix = PL_reg_call_cc->ss;
4181 PL_savestack_ix = tmp;
4183 /* Make position available to the callcc. */
4184 PL_reginput = locinput;
4186 cache_re(PL_reg_call_cc->re);
4187 PL_regcc = PL_reg_call_cc->cc;
4188 PL_reg_call_cc = PL_reg_call_cc->prev;
4189 if (regmatch(cur_call_cc->node)) {
4190 PL_reg_call_cc = cur_call_cc;
4194 REGCP_UNWIND(lastcp);
4196 PL_reg_call_cc = cur_call_cc;
4202 PerlIO_printf(Perl_debug_log,
4203 "%*s continuation failed...\n",
4204 REPORT_CODE_OFF+PL_regindent*2, "")
4208 if (locinput < PL_regtill) {
4209 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
4210 "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n",
4212 (long)(locinput - PL_reg_starttry),
4213 (long)(PL_regtill - PL_reg_starttry),
4215 sayNO_FINAL; /* Cannot match: too short. */
4217 PL_reginput = locinput; /* put where regtry can find it */
4218 sayYES_FINAL; /* Success! */
4220 PL_reginput = locinput; /* put where regtry can find it */
4221 sayYES_LOUD; /* Success! */
4224 PL_reginput = locinput;
4229 s = HOPBACKc(locinput, scan->flags);
4235 PL_reginput = locinput;
4240 s = HOPBACKc(locinput, scan->flags);
4246 PL_reginput = locinput;
4249 inner = NEXTOPER(NEXTOPER(scan));
4250 if (regmatch(inner) != n) {
4265 if (OP(scan) == SUSPEND) {
4266 locinput = PL_reginput;
4267 nextchr = UCHARAT(locinput);
4272 next = scan + ARG(scan);
4277 PerlIO_printf(Perl_error_log, "%"UVxf" %d\n",
4278 PTR2UV(scan), OP(scan));
4279 Perl_croak(aTHX_ "regexp memory corruption");
4286 * We get here only if there's trouble -- normally "case END" is
4287 * the terminating point.
4289 Perl_croak(aTHX_ "corrupted regexp pointers");
4295 PerlIO_printf(Perl_debug_log,
4296 "%*s %scould match...%s\n",
4297 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4], PL_colors[5])
4301 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n",
4302 PL_colors[4], PL_colors[5]));
4308 #if 0 /* Breaks $^R */
4316 PerlIO_printf(Perl_debug_log,
4317 "%*s %sfailed...%s\n",
4318 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4], PL_colors[5])
4324 re_unwind_t * const uw = SSPTRt(unwind,re_unwind_t);
4327 case RE_UNWIND_BRANCH:
4328 case RE_UNWIND_BRANCHJ:
4330 re_unwind_branch_t * const uwb = &(uw->branch);
4331 const I32 lastparen = uwb->lastparen;
4333 REGCP_UNWIND(uwb->lastcp);
4334 for (n = *PL_reglastparen; n > lastparen; n--)
4336 *PL_reglastparen = n;
4337 scan = next = uwb->next;
4339 OP(scan) != (uwb->type == RE_UNWIND_BRANCH
4340 ? BRANCH : BRANCHJ) ) { /* Failure */
4347 /* Have more choice yet. Reuse the same uwb. */
4348 if ((n = (uwb->type == RE_UNWIND_BRANCH
4349 ? NEXT_OFF(next) : ARG(next))))
4352 next = NULL; /* XXXX Needn't unwinding in this case... */
4354 next = NEXTOPER(scan);
4355 if (uwb->type == RE_UNWIND_BRANCHJ)
4356 next = NEXTOPER(next);
4357 locinput = uwb->locinput;
4358 nextchr = uwb->nextchr;
4360 PL_regindent = uwb->regindent;
4367 Perl_croak(aTHX_ "regexp unwind memory corruption");
4378 - regrepeat - repeatedly match something simple, report how many
4381 * [This routine now assumes that it will only match on things of length 1.
4382 * That was true before, but now we assume scan - reginput is the count,
4383 * rather than incrementing count on every character. [Er, except utf8.]]
4386 S_regrepeat(pTHX_ const regnode *p, I32 max)
4389 register char *scan;
4391 register char *loceol = PL_regeol;
4392 register I32 hardcount = 0;
4393 register bool do_utf8 = PL_reg_match_utf8;
4396 if (max == REG_INFTY)
4398 else if (max < loceol - scan)
4399 loceol = scan + max;
4404 while (scan < loceol && hardcount < max && *scan != '\n') {
4405 scan += UTF8SKIP(scan);
4409 while (scan < loceol && *scan != '\n')
4416 while (scan < loceol && hardcount < max) {
4417 scan += UTF8SKIP(scan);
4427 case EXACT: /* length of string is 1 */
4429 while (scan < loceol && UCHARAT(scan) == c)
4432 case EXACTF: /* length of string is 1 */
4434 while (scan < loceol &&
4435 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold[c]))
4438 case EXACTFL: /* length of string is 1 */
4439 PL_reg_flags |= RF_tainted;
4441 while (scan < loceol &&
4442 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold_locale[c]))
4448 while (hardcount < max && scan < loceol &&
4449 reginclass(p, (U8*)scan, 0, do_utf8)) {
4450 scan += UTF8SKIP(scan);
4454 while (scan < loceol && REGINCLASS(p, (U8*)scan))
4461 LOAD_UTF8_CHARCLASS_ALNUM();
4462 while (hardcount < max && scan < loceol &&
4463 swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
4464 scan += UTF8SKIP(scan);
4468 while (scan < loceol && isALNUM(*scan))
4473 PL_reg_flags |= RF_tainted;
4476 while (hardcount < max && scan < loceol &&
4477 isALNUM_LC_utf8((U8*)scan)) {
4478 scan += UTF8SKIP(scan);
4482 while (scan < loceol && isALNUM_LC(*scan))
4489 LOAD_UTF8_CHARCLASS_ALNUM();
4490 while (hardcount < max && scan < loceol &&
4491 !swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
4492 scan += UTF8SKIP(scan);
4496 while (scan < loceol && !isALNUM(*scan))
4501 PL_reg_flags |= RF_tainted;
4504 while (hardcount < max && scan < loceol &&
4505 !isALNUM_LC_utf8((U8*)scan)) {
4506 scan += UTF8SKIP(scan);
4510 while (scan < loceol && !isALNUM_LC(*scan))
4517 LOAD_UTF8_CHARCLASS_SPACE();
4518 while (hardcount < max && scan < loceol &&
4520 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
4521 scan += UTF8SKIP(scan);
4525 while (scan < loceol && isSPACE(*scan))
4530 PL_reg_flags |= RF_tainted;
4533 while (hardcount < max && scan < loceol &&
4534 (*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
4535 scan += UTF8SKIP(scan);
4539 while (scan < loceol && isSPACE_LC(*scan))
4546 LOAD_UTF8_CHARCLASS_SPACE();
4547 while (hardcount < max && scan < loceol &&
4549 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
4550 scan += UTF8SKIP(scan);
4554 while (scan < loceol && !isSPACE(*scan))
4559 PL_reg_flags |= RF_tainted;
4562 while (hardcount < max && scan < loceol &&
4563 !(*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
4564 scan += UTF8SKIP(scan);
4568 while (scan < loceol && !isSPACE_LC(*scan))
4575 LOAD_UTF8_CHARCLASS_DIGIT();
4576 while (hardcount < max && scan < loceol &&
4577 swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
4578 scan += UTF8SKIP(scan);
4582 while (scan < loceol && isDIGIT(*scan))
4589 LOAD_UTF8_CHARCLASS_DIGIT();
4590 while (hardcount < max && scan < loceol &&
4591 !swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
4592 scan += UTF8SKIP(scan);
4596 while (scan < loceol && !isDIGIT(*scan))
4600 default: /* Called on something of 0 width. */
4601 break; /* So match right here or not at all. */
4607 c = scan - PL_reginput;
4611 SV *re_debug_flags = NULL;
4612 SV * const prop = sv_newmortal();
4616 PerlIO_printf(Perl_debug_log,
4617 "%*s %s can match %"IVdf" times out of %"IVdf"...\n",
4618 REPORT_CODE_OFF+1, "", SvPVX_const(prop),(IV)c,(IV)max);
4627 - regclass_swash - prepare the utf8 swash
4631 Perl_regclass_swash(pTHX_ register const regnode* node, bool doinit, SV** listsvp, SV **altsvp)
4638 if (PL_regdata && PL_regdata->count) {
4639 const U32 n = ARG(node);
4641 if (PL_regdata->what[n] == 's') {
4642 SV * const rv = (SV*)PL_regdata->data[n];
4643 AV * const av = (AV*)SvRV((SV*)rv);
4644 SV **const ary = AvARRAY(av);
4647 /* See the end of regcomp.c:S_regclass() for
4648 * documentation of these array elements. */
4651 a = SvROK(ary[1]) ? &ary[1] : 0;
4652 b = SvTYPE(ary[2]) == SVt_PVAV ? &ary[2] : 0;
4656 else if (si && doinit) {
4657 sw = swash_init("utf8", "", si, 1, 0);
4658 (void)av_store(av, 1, sw);
4674 - reginclass - determine if a character falls into a character class
4676 The n is the ANYOF regnode, the p is the target string, lenp
4677 is pointer to the maximum length of how far to go in the p
4678 (if the lenp is zero, UTF8SKIP(p) is used),
4679 do_utf8 tells whether the target string is in UTF-8.
4684 S_reginclass(pTHX_ register const regnode *n, register const U8* p, STRLEN* lenp, register bool do_utf8)
4687 const char flags = ANYOF_FLAGS(n);
4693 if (do_utf8 && !UTF8_IS_INVARIANT(c)) {
4694 c = utf8n_to_uvchr(p, UTF8_MAXBYTES, &len,
4695 ckWARN(WARN_UTF8) ? UTF8_CHECK_ONLY :
4696 UTF8_ALLOW_ANYUV|UTF8_CHECK_ONLY);
4697 if (len == (STRLEN)-1)
4698 Perl_croak(aTHX_ "Malformed UTF-8 character (fatal)");
4701 plen = lenp ? *lenp : UNISKIP(NATIVE_TO_UNI(c));
4702 if (do_utf8 || (flags & ANYOF_UNICODE)) {
4705 if (do_utf8 && !ANYOF_RUNTIME(n)) {
4706 if (len != (STRLEN)-1 && c < 256 && ANYOF_BITMAP_TEST(n, c))
4709 if (!match && do_utf8 && (flags & ANYOF_UNICODE_ALL) && c >= 256)
4713 SV * const sw = regclass_swash(n, TRUE, 0, (SV**)&av);
4716 if (swash_fetch(sw, p, do_utf8))
4718 else if (flags & ANYOF_FOLD) {
4719 if (!match && lenp && av) {
4721 for (i = 0; i <= av_len(av); i++) {
4722 SV* const sv = *av_fetch(av, i, FALSE);
4724 const char * const s = SvPV_const(sv, len);
4726 if (len <= plen && memEQ(s, (char*)p, len)) {
4734 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
4737 to_utf8_fold(p, tmpbuf, &tmplen);
4738 if (swash_fetch(sw, tmpbuf, do_utf8))
4744 if (match && lenp && *lenp == 0)
4745 *lenp = UNISKIP(NATIVE_TO_UNI(c));
4747 if (!match && c < 256) {
4748 if (ANYOF_BITMAP_TEST(n, c))
4750 else if (flags & ANYOF_FOLD) {
4753 if (flags & ANYOF_LOCALE) {
4754 PL_reg_flags |= RF_tainted;
4755 f = PL_fold_locale[c];
4759 if (f != c && ANYOF_BITMAP_TEST(n, f))
4763 if (!match && (flags & ANYOF_CLASS)) {
4764 PL_reg_flags |= RF_tainted;
4766 (ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) ||
4767 (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) ||
4768 (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) ||
4769 (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) ||
4770 (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) ||
4771 (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) ||
4772 (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) ||
4773 (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) ||
4774 (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) ||
4775 (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) ||
4776 (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII(c)) ||
4777 (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII(c)) ||
4778 (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) ||
4779 (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) ||
4780 (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) ||
4781 (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) ||
4782 (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) ||
4783 (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) ||
4784 (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) ||
4785 (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) ||
4786 (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) ||
4787 (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) ||
4788 (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) ||
4789 (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) ||
4790 (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) ||
4791 (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) ||
4792 (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) ||
4793 (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) ||
4794 (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK(c)) ||
4795 (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK(c))
4796 ) /* How's that for a conditional? */
4803 return (flags & ANYOF_INVERT) ? !match : match;
4807 S_reghop(pTHX_ U8 *s, I32 off)
4810 return S_reghop3(s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4814 S_reghop3(U8 *s, I32 off, U8* lim)
4818 while (off-- && s < lim) {
4819 /* XXX could check well-formedness here */
4827 if (UTF8_IS_CONTINUED(*s)) {
4828 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4831 /* XXX could check well-formedness here */
4839 S_reghopmaybe(pTHX_ U8 *s, I32 off)
4842 return S_reghopmaybe3(s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4846 S_reghopmaybe3(U8* s, I32 off, U8* lim)
4850 while (off-- && s < lim) {
4851 /* XXX could check well-formedness here */
4861 if (UTF8_IS_CONTINUED(*s)) {
4862 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4865 /* XXX could check well-formedness here */
4877 restore_pos(pTHX_ void *arg)
4880 PERL_UNUSED_ARG(arg);
4881 if (PL_reg_eval_set) {
4882 if (PL_reg_oldsaved) {
4883 PL_reg_re->subbeg = PL_reg_oldsaved;
4884 PL_reg_re->sublen = PL_reg_oldsavedlen;
4885 #ifdef PERL_OLD_COPY_ON_WRITE
4886 PL_reg_re->saved_copy = PL_nrs;
4888 RX_MATCH_COPIED_on(PL_reg_re);
4890 PL_reg_magic->mg_len = PL_reg_oldpos;
4891 PL_reg_eval_set = 0;
4892 PL_curpm = PL_reg_oldcurpm;
4897 S_to_utf8_substr(pTHX_ register regexp *prog)
4899 if (prog->float_substr && !prog->float_utf8) {
4901 prog->float_utf8 = sv = newSVsv(prog->float_substr);
4902 sv_utf8_upgrade(sv);
4903 if (SvTAIL(prog->float_substr))
4905 if (prog->float_substr == prog->check_substr)
4906 prog->check_utf8 = sv;
4908 if (prog->anchored_substr && !prog->anchored_utf8) {
4910 prog->anchored_utf8 = sv = newSVsv(prog->anchored_substr);
4911 sv_utf8_upgrade(sv);
4912 if (SvTAIL(prog->anchored_substr))
4914 if (prog->anchored_substr == prog->check_substr)
4915 prog->check_utf8 = sv;
4920 S_to_byte_substr(pTHX_ register regexp *prog)
4923 if (prog->float_utf8 && !prog->float_substr) {
4925 prog->float_substr = sv = newSVsv(prog->float_utf8);
4926 if (sv_utf8_downgrade(sv, TRUE)) {
4927 if (SvTAIL(prog->float_utf8))
4931 prog->float_substr = sv = &PL_sv_undef;
4933 if (prog->float_utf8 == prog->check_utf8)
4934 prog->check_substr = sv;
4936 if (prog->anchored_utf8 && !prog->anchored_substr) {
4938 prog->anchored_substr = sv = newSVsv(prog->anchored_utf8);
4939 if (sv_utf8_downgrade(sv, TRUE)) {
4940 if (SvTAIL(prog->anchored_utf8))
4944 prog->anchored_substr = sv = &PL_sv_undef;
4946 if (prog->anchored_utf8 == prog->check_utf8)
4947 prog->check_substr = sv;
4953 * c-indentation-style: bsd
4955 * indent-tabs-mode: t
4958 * ex: set ts=8 sts=4 sw=4 noet: