5 * One Ring to rule them all, One Ring to find them
7 * [p.v of _The Lord of the Rings_, opening poem]
8 * [p.50 of _The Lord of the Rings_, I/iii: "The Shadow of the Past"]
9 * [p.254 of _The Lord of the Rings_, II/ii: "The Council of Elrond"]
12 /* This file contains functions for executing a regular expression. See
13 * also regcomp.c which funnily enough, contains functions for compiling
14 * a regular expression.
16 * This file is also copied at build time to ext/re/re_exec.c, where
17 * it's built with -DPERL_EXT_RE_BUILD -DPERL_EXT_RE_DEBUG -DPERL_EXT.
18 * This causes the main functions to be compiled under new names and with
19 * debugging support added, which makes "use re 'debug'" work.
22 /* NOTE: this is derived from Henry Spencer's regexp code, and should not
23 * confused with the original package (see point 3 below). Thanks, Henry!
26 /* Additional note: this code is very heavily munged from Henry's version
27 * in places. In some spots I've traded clarity for efficiency, so don't
28 * blame Henry for some of the lack of readability.
31 /* The names of the functions have been changed from regcomp and
32 * regexec to pregcomp and pregexec in order to avoid conflicts
33 * with the POSIX routines of the same names.
36 #ifdef PERL_EXT_RE_BUILD
41 * pregcomp and pregexec -- regsub and regerror are not used in perl
43 * Copyright (c) 1986 by University of Toronto.
44 * Written by Henry Spencer. Not derived from licensed software.
46 * Permission is granted to anyone to use this software for any
47 * purpose on any computer system, and to redistribute it freely,
48 * subject to the following restrictions:
50 * 1. The author is not responsible for the consequences of use of
51 * this software, no matter how awful, even if they arise
54 * 2. The origin of this software must not be misrepresented, either
55 * by explicit claim or by omission.
57 * 3. Altered versions must be plainly marked as such, and must not
58 * be misrepresented as being the original software.
60 **** Alterations to Henry's code are...
62 **** Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
63 **** 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008
64 **** by Larry Wall and others
66 **** You may distribute under the terms of either the GNU General Public
67 **** License or the Artistic License, as specified in the README file.
69 * Beware that some of this code is subtly aware of the way operator
70 * precedence is structured in regular expressions. Serious changes in
71 * regular-expression syntax might require a total rethink.
74 #define PERL_IN_REGEXEC_C
77 #ifdef PERL_IN_XSUB_RE
83 #include "inline_invlist.c"
84 #include "unicode_constants.h"
87 /* At least one required character in the target string is expressible only in
89 static const char* const non_utf8_target_but_utf8_required
90 = "Can't match, because target string needs to be in UTF-8\n";
93 #define NON_UTF8_TARGET_BUT_UTF8_REQUIRED(target) STMT_START { \
94 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%s", non_utf8_target_but_utf8_required));\
98 #define HAS_NONLATIN1_FOLD_CLOSURE(i) _HAS_NONLATIN1_FOLD_CLOSURE_ONLY_FOR_USE_BY_REGCOMP_DOT_C_AND_REGEXEC_DOT_C(i)
101 #define STATIC static
104 /* Valid only for non-utf8 strings: avoids the reginclass
105 * call if there are no complications: i.e., if everything matchable is
106 * straight forward in the bitmap */
107 #define REGINCLASS(prog,p,c) (ANYOF_FLAGS(p) ? reginclass(prog,p,c,c+1,0) \
108 : ANYOF_BITMAP_TEST(p,*(c)))
114 #define CHR_SVLEN(sv) (utf8_target ? sv_len_utf8(sv) : SvCUR(sv))
115 #define CHR_DIST(a,b) (reginfo->is_utf8_target ? utf8_distance(a,b) : a - b)
117 #define HOPc(pos,off) \
118 (char *)(reginfo->is_utf8_target \
119 ? reghop3((U8*)pos, off, \
120 (U8*)(off >= 0 ? reginfo->strend : reginfo->strbeg)) \
123 #define HOPBACKc(pos, off) \
124 (char*)(reginfo->is_utf8_target \
125 ? reghopmaybe3((U8*)pos, -off, (U8*)(reginfo->strbeg)) \
126 : (pos - off >= reginfo->strbeg) \
130 #define HOP3(pos,off,lim) (reginfo->is_utf8_target ? reghop3((U8*)(pos), off, (U8*)(lim)) : (U8*)(pos + off))
131 #define HOP3c(pos,off,lim) ((char*)HOP3(pos,off,lim))
133 /* lim must be +ve. Returns NULL on overshoot */
134 #define HOPMAYBE3(pos,off,lim) \
135 (reginfo->is_utf8_target \
136 ? reghopmaybe3((U8*)pos, off, (U8*)(lim)) \
137 : ((U8*)pos + off <= lim) \
141 /* like HOP3, but limits the result to <= lim even for the non-utf8 case.
142 * off must be >=0; args should be vars rather than expressions */
143 #define HOP3lim(pos,off,lim) (reginfo->is_utf8_target \
144 ? reghop3((U8*)(pos), off, (U8*)(lim)) \
145 : (U8*)((pos + off) > lim ? lim : (pos + off)))
147 #define HOP4(pos,off,llim, rlim) (reginfo->is_utf8_target \
148 ? reghop4((U8*)(pos), off, (U8*)(llim), (U8*)(rlim)) \
150 #define HOP4c(pos,off,llim, rlim) ((char*)HOP4(pos,off,llim, rlim))
152 #define NEXTCHR_EOS -10 /* nextchr has fallen off the end */
153 #define NEXTCHR_IS_EOS (nextchr < 0)
155 #define SET_nextchr \
156 nextchr = ((locinput < reginfo->strend) ? UCHARAT(locinput) : NEXTCHR_EOS)
158 #define SET_locinput(p) \
163 #define LOAD_UTF8_CHARCLASS(swash_ptr, property_name, invlist) STMT_START { \
165 U8 flags = _CORE_SWASH_INIT_ACCEPT_INVLIST; \
166 swash_ptr = _core_swash_init("utf8", property_name, &PL_sv_undef, \
167 1, 0, invlist, &flags); \
172 /* If in debug mode, we test that a known character properly matches */
174 # define LOAD_UTF8_CHARCLASS_DEBUG_TEST(swash_ptr, \
177 utf8_char_in_property) \
178 LOAD_UTF8_CHARCLASS(swash_ptr, property_name, invlist); \
179 assert(swash_fetch(swash_ptr, (U8 *) utf8_char_in_property, TRUE));
181 # define LOAD_UTF8_CHARCLASS_DEBUG_TEST(swash_ptr, \
184 utf8_char_in_property) \
185 LOAD_UTF8_CHARCLASS(swash_ptr, property_name, invlist)
188 #define LOAD_UTF8_CHARCLASS_ALNUM() LOAD_UTF8_CHARCLASS_DEBUG_TEST( \
189 PL_utf8_swash_ptrs[_CC_WORDCHAR], \
191 PL_XPosix_ptrs[_CC_WORDCHAR], \
192 LATIN_CAPITAL_LETTER_SHARP_S_UTF8);
194 #define LOAD_UTF8_CHARCLASS_GCB() /* Grapheme cluster boundaries */ \
196 LOAD_UTF8_CHARCLASS_DEBUG_TEST(PL_utf8_X_regular_begin, \
197 "_X_regular_begin", \
199 LATIN_CAPITAL_LETTER_SHARP_S_UTF8); \
200 LOAD_UTF8_CHARCLASS_DEBUG_TEST(PL_utf8_X_extend, \
203 COMBINING_GRAVE_ACCENT_UTF8); \
206 #define PLACEHOLDER /* Something for the preprocessor to grab onto */
207 /* TODO: Combine JUMPABLE and HAS_TEXT to cache OP(rn) */
209 /* for use after a quantifier and before an EXACT-like node -- japhy */
210 /* it would be nice to rework regcomp.sym to generate this stuff. sigh
212 * NOTE that *nothing* that affects backtracking should be in here, specifically
213 * VERBS must NOT be included. JUMPABLE is used to determine if we can ignore a
214 * node that is in between two EXACT like nodes when ascertaining what the required
215 * "follow" character is. This should probably be moved to regex compile time
216 * although it may be done at run time beause of the REF possibility - more
217 * investigation required. -- demerphq
219 #define JUMPABLE(rn) ( \
221 (OP(rn) == CLOSE && (!cur_eval || cur_eval->u.eval.close_paren != ARG(rn))) || \
223 OP(rn) == SUSPEND || OP(rn) == IFMATCH || \
224 OP(rn) == PLUS || OP(rn) == MINMOD || \
226 (PL_regkind[OP(rn)] == CURLY && ARG1(rn) > 0) \
228 #define IS_EXACT(rn) (PL_regkind[OP(rn)] == EXACT)
230 #define HAS_TEXT(rn) ( IS_EXACT(rn) || PL_regkind[OP(rn)] == REF )
233 /* Currently these are only used when PL_regkind[OP(rn)] == EXACT so
234 we don't need this definition. XXX These are now out-of-sync*/
235 #define IS_TEXT(rn) ( OP(rn)==EXACT || OP(rn)==REF || OP(rn)==NREF )
236 #define IS_TEXTF(rn) ( OP(rn)==EXACTFU || OP(rn)==EXACTFU_SS || OP(rn)==EXACTFA || OP(rn)==EXACTFA_NO_TRIE || OP(rn)==EXACTF || OP(rn)==REFF || OP(rn)==NREFF )
237 #define IS_TEXTFL(rn) ( OP(rn)==EXACTFL || OP(rn)==REFFL || OP(rn)==NREFFL )
240 /* ... so we use this as its faster. */
241 #define IS_TEXT(rn) ( OP(rn)==EXACT || OP(rn)==EXACTL )
242 #define IS_TEXTFU(rn) ( OP(rn)==EXACTFU || OP(rn)==EXACTFLU8 || OP(rn)==EXACTFU_SS || OP(rn) == EXACTFA || OP(rn) == EXACTFA_NO_TRIE)
243 #define IS_TEXTF(rn) ( OP(rn)==EXACTF )
244 #define IS_TEXTFL(rn) ( OP(rn)==EXACTFL )
249 Search for mandatory following text node; for lookahead, the text must
250 follow but for lookbehind (rn->flags != 0) we skip to the next step.
252 #define FIND_NEXT_IMPT(rn) STMT_START { \
253 while (JUMPABLE(rn)) { \
254 const OPCODE type = OP(rn); \
255 if (type == SUSPEND || PL_regkind[type] == CURLY) \
256 rn = NEXTOPER(NEXTOPER(rn)); \
257 else if (type == PLUS) \
259 else if (type == IFMATCH) \
260 rn = (rn->flags == 0) ? NEXTOPER(NEXTOPER(rn)) : rn + ARG(rn); \
261 else rn += NEXT_OFF(rn); \
265 /* These constants are for finding GCB=LV and GCB=LVT in the CLUMP regnode.
266 * These are for the pre-composed Hangul syllables, which are all in a
267 * contiguous block and arranged there in such a way so as to facilitate
268 * alorithmic determination of their characteristics. As such, they don't need
269 * a swash, but can be determined by simple arithmetic. Almost all are
270 * GCB=LVT, but every 28th one is a GCB=LV */
271 #define SBASE 0xAC00 /* Start of block */
272 #define SCount 11172 /* Length of block */
275 #define SLAB_FIRST(s) (&(s)->states[0])
276 #define SLAB_LAST(s) (&(s)->states[PERL_REGMATCH_SLAB_SLOTS-1])
278 static void S_setup_eval_state(pTHX_ regmatch_info *const reginfo);
279 static void S_cleanup_regmatch_info_aux(pTHX_ void *arg);
280 static regmatch_state * S_push_slab(pTHX);
282 #define REGCP_PAREN_ELEMS 3
283 #define REGCP_OTHER_ELEMS 3
284 #define REGCP_FRAME_ELEMS 1
285 /* REGCP_FRAME_ELEMS are not part of the REGCP_OTHER_ELEMS and
286 * are needed for the regexp context stack bookkeeping. */
289 S_regcppush(pTHX_ const regexp *rex, I32 parenfloor, U32 maxopenparen)
291 const int retval = PL_savestack_ix;
292 const int paren_elems_to_push =
293 (maxopenparen - parenfloor) * REGCP_PAREN_ELEMS;
294 const UV total_elems = paren_elems_to_push + REGCP_OTHER_ELEMS;
295 const UV elems_shifted = total_elems << SAVE_TIGHT_SHIFT;
297 GET_RE_DEBUG_FLAGS_DECL;
299 PERL_ARGS_ASSERT_REGCPPUSH;
301 if (paren_elems_to_push < 0)
302 Perl_croak(aTHX_ "panic: paren_elems_to_push, %i < 0, maxopenparen: %i parenfloor: %i REGCP_PAREN_ELEMS: %u",
303 (int)paren_elems_to_push, (int)maxopenparen,
304 (int)parenfloor, (unsigned)REGCP_PAREN_ELEMS);
306 if ((elems_shifted >> SAVE_TIGHT_SHIFT) != total_elems)
307 Perl_croak(aTHX_ "panic: paren_elems_to_push offset %"UVuf
308 " out of range (%lu-%ld)",
310 (unsigned long)maxopenparen,
313 SSGROW(total_elems + REGCP_FRAME_ELEMS);
316 if ((int)maxopenparen > (int)parenfloor)
317 PerlIO_printf(Perl_debug_log,
318 "rex=0x%"UVxf" offs=0x%"UVxf": saving capture indices:\n",
323 for (p = parenfloor+1; p <= (I32)maxopenparen; p++) {
324 /* REGCP_PARENS_ELEMS are pushed per pairs of parentheses. */
325 SSPUSHIV(rex->offs[p].end);
326 SSPUSHIV(rex->offs[p].start);
327 SSPUSHINT(rex->offs[p].start_tmp);
328 DEBUG_BUFFERS_r(PerlIO_printf(Perl_debug_log,
329 " \\%"UVuf": %"IVdf"(%"IVdf")..%"IVdf"\n",
331 (IV)rex->offs[p].start,
332 (IV)rex->offs[p].start_tmp,
336 /* REGCP_OTHER_ELEMS are pushed in any case, parentheses or no. */
337 SSPUSHINT(maxopenparen);
338 SSPUSHINT(rex->lastparen);
339 SSPUSHINT(rex->lastcloseparen);
340 SSPUSHUV(SAVEt_REGCONTEXT | elems_shifted); /* Magic cookie. */
345 /* These are needed since we do not localize EVAL nodes: */
346 #define REGCP_SET(cp) \
348 PerlIO_printf(Perl_debug_log, \
349 " Setting an EVAL scope, savestack=%"IVdf"\n", \
350 (IV)PL_savestack_ix)); \
353 #define REGCP_UNWIND(cp) \
355 if (cp != PL_savestack_ix) \
356 PerlIO_printf(Perl_debug_log, \
357 " Clearing an EVAL scope, savestack=%"IVdf"..%"IVdf"\n", \
358 (IV)(cp), (IV)PL_savestack_ix)); \
361 #define UNWIND_PAREN(lp, lcp) \
362 for (n = rex->lastparen; n > lp; n--) \
363 rex->offs[n].end = -1; \
364 rex->lastparen = n; \
365 rex->lastcloseparen = lcp;
369 S_regcppop(pTHX_ regexp *rex, U32 *maxopenparen_p)
373 GET_RE_DEBUG_FLAGS_DECL;
375 PERL_ARGS_ASSERT_REGCPPOP;
377 /* Pop REGCP_OTHER_ELEMS before the parentheses loop starts. */
379 assert((i & SAVE_MASK) == SAVEt_REGCONTEXT); /* Check that the magic cookie is there. */
380 i >>= SAVE_TIGHT_SHIFT; /* Parentheses elements to pop. */
381 rex->lastcloseparen = SSPOPINT;
382 rex->lastparen = SSPOPINT;
383 *maxopenparen_p = SSPOPINT;
385 i -= REGCP_OTHER_ELEMS;
386 /* Now restore the parentheses context. */
388 if (i || rex->lastparen + 1 <= rex->nparens)
389 PerlIO_printf(Perl_debug_log,
390 "rex=0x%"UVxf" offs=0x%"UVxf": restoring capture indices to:\n",
395 paren = *maxopenparen_p;
396 for ( ; i > 0; i -= REGCP_PAREN_ELEMS) {
398 rex->offs[paren].start_tmp = SSPOPINT;
399 rex->offs[paren].start = SSPOPIV;
401 if (paren <= rex->lastparen)
402 rex->offs[paren].end = tmps;
403 DEBUG_BUFFERS_r( PerlIO_printf(Perl_debug_log,
404 " \\%"UVuf": %"IVdf"(%"IVdf")..%"IVdf"%s\n",
406 (IV)rex->offs[paren].start,
407 (IV)rex->offs[paren].start_tmp,
408 (IV)rex->offs[paren].end,
409 (paren > rex->lastparen ? "(skipped)" : ""));
414 /* It would seem that the similar code in regtry()
415 * already takes care of this, and in fact it is in
416 * a better location to since this code can #if 0-ed out
417 * but the code in regtry() is needed or otherwise tests
418 * requiring null fields (pat.t#187 and split.t#{13,14}
419 * (as of patchlevel 7877) will fail. Then again,
420 * this code seems to be necessary or otherwise
421 * this erroneously leaves $1 defined: "1" =~ /^(?:(\d)x)?\d$/
422 * --jhi updated by dapm */
423 for (i = rex->lastparen + 1; i <= rex->nparens; i++) {
424 if (i > *maxopenparen_p)
425 rex->offs[i].start = -1;
426 rex->offs[i].end = -1;
427 DEBUG_BUFFERS_r( PerlIO_printf(Perl_debug_log,
428 " \\%"UVuf": %s ..-1 undeffing\n",
430 (i > *maxopenparen_p) ? "-1" : " "
436 /* restore the parens and associated vars at savestack position ix,
437 * but without popping the stack */
440 S_regcp_restore(pTHX_ regexp *rex, I32 ix, U32 *maxopenparen_p)
442 I32 tmpix = PL_savestack_ix;
443 PL_savestack_ix = ix;
444 regcppop(rex, maxopenparen_p);
445 PL_savestack_ix = tmpix;
448 #define regcpblow(cp) LEAVE_SCOPE(cp) /* Ignores regcppush()ed data. */
451 S_isFOO_lc(pTHX_ const U8 classnum, const U8 character)
453 /* Returns a boolean as to whether or not 'character' is a member of the
454 * Posix character class given by 'classnum' that should be equivalent to a
455 * value in the typedef '_char_class_number'.
457 * Ideally this could be replaced by a just an array of function pointers
458 * to the C library functions that implement the macros this calls.
459 * However, to compile, the precise function signatures are required, and
460 * these may vary from platform to to platform. To avoid having to figure
461 * out what those all are on each platform, I (khw) am using this method,
462 * which adds an extra layer of function call overhead (unless the C
463 * optimizer strips it away). But we don't particularly care about
464 * performance with locales anyway. */
466 switch ((_char_class_number) classnum) {
467 case _CC_ENUM_ALPHANUMERIC: return isALPHANUMERIC_LC(character);
468 case _CC_ENUM_ALPHA: return isALPHA_LC(character);
469 case _CC_ENUM_ASCII: return isASCII_LC(character);
470 case _CC_ENUM_BLANK: return isBLANK_LC(character);
471 case _CC_ENUM_CASED: return isLOWER_LC(character)
472 || isUPPER_LC(character);
473 case _CC_ENUM_CNTRL: return isCNTRL_LC(character);
474 case _CC_ENUM_DIGIT: return isDIGIT_LC(character);
475 case _CC_ENUM_GRAPH: return isGRAPH_LC(character);
476 case _CC_ENUM_LOWER: return isLOWER_LC(character);
477 case _CC_ENUM_PRINT: return isPRINT_LC(character);
478 case _CC_ENUM_PSXSPC: return isPSXSPC_LC(character);
479 case _CC_ENUM_PUNCT: return isPUNCT_LC(character);
480 case _CC_ENUM_SPACE: return isSPACE_LC(character);
481 case _CC_ENUM_UPPER: return isUPPER_LC(character);
482 case _CC_ENUM_WORDCHAR: return isWORDCHAR_LC(character);
483 case _CC_ENUM_XDIGIT: return isXDIGIT_LC(character);
484 default: /* VERTSPACE should never occur in locales */
485 Perl_croak(aTHX_ "panic: isFOO_lc() has an unexpected character class '%d'", classnum);
488 NOT_REACHED; /* NOTREACHED */
493 S_isFOO_utf8_lc(pTHX_ const U8 classnum, const U8* character)
495 /* Returns a boolean as to whether or not the (well-formed) UTF-8-encoded
496 * 'character' is a member of the Posix character class given by 'classnum'
497 * that should be equivalent to a value in the typedef
498 * '_char_class_number'.
500 * This just calls isFOO_lc on the code point for the character if it is in
501 * the range 0-255. Outside that range, all characters use Unicode
502 * rules, ignoring any locale. So use the Unicode function if this class
503 * requires a swash, and use the Unicode macro otherwise. */
505 PERL_ARGS_ASSERT_ISFOO_UTF8_LC;
507 if (UTF8_IS_INVARIANT(*character)) {
508 return isFOO_lc(classnum, *character);
510 else if (UTF8_IS_DOWNGRADEABLE_START(*character)) {
511 return isFOO_lc(classnum,
512 TWO_BYTE_UTF8_TO_NATIVE(*character, *(character + 1)));
515 _CHECK_AND_OUTPUT_WIDE_LOCALE_UTF8_MSG(character, character + UTF8SKIP(character));
517 if (classnum < _FIRST_NON_SWASH_CC) {
519 /* Initialize the swash unless done already */
520 if (! PL_utf8_swash_ptrs[classnum]) {
521 U8 flags = _CORE_SWASH_INIT_ACCEPT_INVLIST;
522 PL_utf8_swash_ptrs[classnum] =
523 _core_swash_init("utf8",
526 PL_XPosix_ptrs[classnum], &flags);
529 return cBOOL(swash_fetch(PL_utf8_swash_ptrs[classnum], (U8 *)
531 TRUE /* is UTF */ ));
534 switch ((_char_class_number) classnum) {
536 case _CC_ENUM_PSXSPC: return is_XPERLSPACE_high(character);
538 case _CC_ENUM_BLANK: return is_HORIZWS_high(character);
539 case _CC_ENUM_XDIGIT: return is_XDIGIT_high(character);
540 case _CC_ENUM_VERTSPACE: return is_VERTWS_high(character);
544 return FALSE; /* Things like CNTRL are always below 256 */
548 * pregexec and friends
551 #ifndef PERL_IN_XSUB_RE
553 - pregexec - match a regexp against a string
556 Perl_pregexec(pTHX_ REGEXP * const prog, char* stringarg, char *strend,
557 char *strbeg, SSize_t minend, SV *screamer, U32 nosave)
558 /* stringarg: the point in the string at which to begin matching */
559 /* strend: pointer to null at end of string */
560 /* strbeg: real beginning of string */
561 /* minend: end of match must be >= minend bytes after stringarg. */
562 /* screamer: SV being matched: only used for utf8 flag, pos() etc; string
563 * itself is accessed via the pointers above */
564 /* nosave: For optimizations. */
566 PERL_ARGS_ASSERT_PREGEXEC;
569 regexec_flags(prog, stringarg, strend, strbeg, minend, screamer, NULL,
570 nosave ? 0 : REXEC_COPY_STR);
576 /* re_intuit_start():
578 * Based on some optimiser hints, try to find the earliest position in the
579 * string where the regex could match.
581 * rx: the regex to match against
582 * sv: the SV being matched: only used for utf8 flag; the string
583 * itself is accessed via the pointers below. Note that on
584 * something like an overloaded SV, SvPOK(sv) may be false
585 * and the string pointers may point to something unrelated to
587 * strbeg: real beginning of string
588 * strpos: the point in the string at which to begin matching
589 * strend: pointer to the byte following the last char of the string
590 * flags currently unused; set to 0
591 * data: currently unused; set to NULL
593 * The basic idea of re_intuit_start() is to use some known information
594 * about the pattern, namely:
596 * a) the longest known anchored substring (i.e. one that's at a
597 * constant offset from the beginning of the pattern; but not
598 * necessarily at a fixed offset from the beginning of the
600 * b) the longest floating substring (i.e. one that's not at a constant
601 * offset from the beginning of the pattern);
602 * c) Whether the pattern is anchored to the string; either
603 * an absolute anchor: /^../, or anchored to \n: /^.../m,
604 * or anchored to pos(): /\G/;
605 * d) A start class: a real or synthetic character class which
606 * represents which characters are legal at the start of the pattern;
608 * to either quickly reject the match, or to find the earliest position
609 * within the string at which the pattern might match, thus avoiding
610 * running the full NFA engine at those earlier locations, only to
611 * eventually fail and retry further along.
613 * Returns NULL if the pattern can't match, or returns the address within
614 * the string which is the earliest place the match could occur.
616 * The longest of the anchored and floating substrings is called 'check'
617 * and is checked first. The other is called 'other' and is checked
618 * second. The 'other' substring may not be present. For example,
620 * /(abc|xyz)ABC\d{0,3}DEFG/
624 * check substr (float) = "DEFG", offset 6..9 chars
625 * other substr (anchored) = "ABC", offset 3..3 chars
628 * Be aware that during the course of this function, sometimes 'anchored'
629 * refers to a substring being anchored relative to the start of the
630 * pattern, and sometimes to the pattern itself being anchored relative to
631 * the string. For example:
633 * /\dabc/: "abc" is anchored to the pattern;
634 * /^\dabc/: "abc" is anchored to the pattern and the string;
635 * /\d+abc/: "abc" is anchored to neither the pattern nor the string;
636 * /^\d+abc/: "abc" is anchored to neither the pattern nor the string,
637 * but the pattern is anchored to the string.
641 Perl_re_intuit_start(pTHX_
644 const char * const strbeg,
648 re_scream_pos_data *data)
650 struct regexp *const prog = ReANY(rx);
651 SSize_t start_shift = prog->check_offset_min;
652 /* Should be nonnegative! */
653 SSize_t end_shift = 0;
654 /* current lowest pos in string where the regex can start matching */
655 char *rx_origin = strpos;
657 const bool utf8_target = (sv && SvUTF8(sv)) ? 1 : 0; /* if no sv we have to assume bytes */
658 U8 other_ix = 1 - prog->substrs->check_ix;
660 char *other_last = strpos;/* latest pos 'other' substr already checked to */
661 char *check_at = NULL; /* check substr found at this pos */
662 const I32 multiline = prog->extflags & RXf_PMf_MULTILINE;
663 RXi_GET_DECL(prog,progi);
664 regmatch_info reginfo_buf; /* create some info to pass to find_byclass */
665 regmatch_info *const reginfo = ®info_buf;
666 GET_RE_DEBUG_FLAGS_DECL;
668 PERL_ARGS_ASSERT_RE_INTUIT_START;
669 PERL_UNUSED_ARG(flags);
670 PERL_UNUSED_ARG(data);
672 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
673 "Intuit: trying to determine minimum start position...\n"));
675 /* for now, assume that all substr offsets are positive. If at some point
676 * in the future someone wants to do clever things with look-behind and
677 * -ve offsets, they'll need to fix up any code in this function
678 * which uses these offsets. See the thread beginning
679 * <20140113145929.GF27210@iabyn.com>
681 assert(prog->substrs->data[0].min_offset >= 0);
682 assert(prog->substrs->data[0].max_offset >= 0);
683 assert(prog->substrs->data[1].min_offset >= 0);
684 assert(prog->substrs->data[1].max_offset >= 0);
685 assert(prog->substrs->data[2].min_offset >= 0);
686 assert(prog->substrs->data[2].max_offset >= 0);
688 /* for now, assume that if both present, that the floating substring
689 * doesn't start before the anchored substring.
690 * If you break this assumption (e.g. doing better optimisations
691 * with lookahead/behind), then you'll need to audit the code in this
692 * function carefully first
695 ! ( (prog->anchored_utf8 || prog->anchored_substr)
696 && (prog->float_utf8 || prog->float_substr))
697 || (prog->float_min_offset >= prog->anchored_offset));
699 /* byte rather than char calculation for efficiency. It fails
700 * to quickly reject some cases that can't match, but will reject
701 * them later after doing full char arithmetic */
702 if (prog->minlen > strend - strpos) {
703 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
704 " String too short...\n"));
708 RX_MATCH_UTF8_set(rx,utf8_target);
709 reginfo->is_utf8_target = cBOOL(utf8_target);
710 reginfo->info_aux = NULL;
711 reginfo->strbeg = strbeg;
712 reginfo->strend = strend;
713 reginfo->is_utf8_pat = cBOOL(RX_UTF8(rx));
715 /* not actually used within intuit, but zero for safety anyway */
716 reginfo->poscache_maxiter = 0;
719 if (!prog->check_utf8 && prog->check_substr)
720 to_utf8_substr(prog);
721 check = prog->check_utf8;
723 if (!prog->check_substr && prog->check_utf8) {
724 if (! to_byte_substr(prog)) {
725 NON_UTF8_TARGET_BUT_UTF8_REQUIRED(fail);
728 check = prog->check_substr;
731 /* dump the various substring data */
732 DEBUG_OPTIMISE_MORE_r({
734 for (i=0; i<=2; i++) {
735 SV *sv = (utf8_target ? prog->substrs->data[i].utf8_substr
736 : prog->substrs->data[i].substr);
740 PerlIO_printf(Perl_debug_log,
741 " substrs[%d]: min=%"IVdf" max=%"IVdf" end shift=%"IVdf
742 " useful=%"IVdf" utf8=%d [%s]\n",
744 (IV)prog->substrs->data[i].min_offset,
745 (IV)prog->substrs->data[i].max_offset,
746 (IV)prog->substrs->data[i].end_shift,
753 if (prog->intflags & PREGf_ANCH) { /* Match at \G, beg-of-str or after \n */
755 /* ml_anch: check after \n?
757 * A note about IMPLICIT: on an un-anchored pattern beginning
758 * with /.*.../, these flags will have been added by the
760 * /.*abc/, /.*abc/m: PREGf_IMPLICIT | PREGf_ANCH_MBOL
761 * /.*abc/s: PREGf_IMPLICIT | PREGf_ANCH_SBOL
763 ml_anch = (prog->intflags & PREGf_ANCH_MBOL)
764 && !(prog->intflags & PREGf_IMPLICIT);
766 if (!ml_anch && !(prog->intflags & PREGf_IMPLICIT)) {
767 /* we are only allowed to match at BOS or \G */
769 /* trivially reject if there's a BOS anchor and we're not at BOS.
771 * Note that we don't try to do a similar quick reject for
772 * \G, since generally the caller will have calculated strpos
773 * based on pos() and gofs, so the string is already correctly
774 * anchored by definition; and handling the exceptions would
775 * be too fiddly (e.g. REXEC_IGNOREPOS).
777 if ( strpos != strbeg
778 && (prog->intflags & PREGf_ANCH_SBOL))
780 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
781 " Not at start...\n"));
785 /* in the presence of an anchor, the anchored (relative to the
786 * start of the regex) substr must also be anchored relative
787 * to strpos. So quickly reject if substr isn't found there.
788 * This works for \G too, because the caller will already have
789 * subtracted gofs from pos, and gofs is the offset from the
790 * \G to the start of the regex. For example, in /.abc\Gdef/,
791 * where substr="abcdef", pos()=3, gofs=4, offset_min=1:
792 * caller will have set strpos=pos()-4; we look for the substr
793 * at position pos()-4+1, which lines up with the "a" */
795 if (prog->check_offset_min == prog->check_offset_max
796 && !(prog->intflags & PREGf_CANY_SEEN))
798 /* Substring at constant offset from beg-of-str... */
799 SSize_t slen = SvCUR(check);
800 char *s = HOP3c(strpos, prog->check_offset_min, strend);
802 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
803 " Looking for check substr at fixed offset %"IVdf"...\n",
804 (IV)prog->check_offset_min));
807 /* In this case, the regex is anchored at the end too.
808 * Unless it's a multiline match, the lengths must match
809 * exactly, give or take a \n. NB: slen >= 1 since
810 * the last char of check is \n */
812 && ( strend - s > slen
813 || strend - s < slen - 1
814 || (strend - s == slen && strend[-1] != '\n')))
816 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
817 " String too long...\n"));
820 /* Now should match s[0..slen-2] */
823 if (slen && (*SvPVX_const(check) != *s
824 || (slen > 1 && memNE(SvPVX_const(check), s, slen))))
826 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
827 " String not equal...\n"));
832 goto success_at_start;
837 end_shift = prog->check_end_shift;
839 #ifdef DEBUGGING /* 7/99: reports of failure (with the older version) */
841 Perl_croak(aTHX_ "panic: end_shift: %"IVdf" pattern:\n%s\n ",
842 (IV)end_shift, RX_PRECOMP(prog));
847 /* This is the (re)entry point of the main loop in this function.
848 * The goal of this loop is to:
849 * 1) find the "check" substring in the region rx_origin..strend
850 * (adjusted by start_shift / end_shift). If not found, reject
852 * 2) If it exists, look for the "other" substr too if defined; for
853 * example, if the check substr maps to the anchored substr, then
854 * check the floating substr, and vice-versa. If not found, go
855 * back to (1) with rx_origin suitably incremented.
856 * 3) If we find an rx_origin position that doesn't contradict
857 * either of the substrings, then check the possible additional
858 * constraints on rx_origin of /^.../m or a known start class.
859 * If these fail, then depending on which constraints fail, jump
860 * back to here, or to various other re-entry points further along
861 * that skip some of the first steps.
862 * 4) If we pass all those tests, update the BmUSEFUL() count on the
863 * substring. If the start position was determined to be at the
864 * beginning of the string - so, not rejected, but not optimised,
865 * since we have to run regmatch from position 0 - decrement the
866 * BmUSEFUL() count. Otherwise increment it.
870 /* first, look for the 'check' substring */
876 DEBUG_OPTIMISE_MORE_r({
877 PerlIO_printf(Perl_debug_log,
878 " At restart: rx_origin=%"IVdf" Check offset min: %"IVdf
879 " Start shift: %"IVdf" End shift %"IVdf
880 " Real end Shift: %"IVdf"\n",
881 (IV)(rx_origin - strpos),
882 (IV)prog->check_offset_min,
885 (IV)prog->check_end_shift);
888 if (prog->intflags & PREGf_CANY_SEEN) {
889 start_point= (U8*)(rx_origin + start_shift);
890 end_point= (U8*)(strend - end_shift);
891 if (start_point > end_point)
894 end_point = HOP3(strend, -end_shift, strbeg);
895 start_point = HOPMAYBE3(rx_origin, start_shift, end_point);
901 /* If the regex is absolutely anchored to either the start of the
902 * string (SBOL) or to pos() (ANCH_GPOS), then
903 * check_offset_max represents an upper bound on the string where
904 * the substr could start. For the ANCH_GPOS case, we assume that
905 * the caller of intuit will have already set strpos to
906 * pos()-gofs, so in this case strpos + offset_max will still be
907 * an upper bound on the substr.
910 && prog->intflags & PREGf_ANCH
911 && prog->check_offset_max != SSize_t_MAX)
913 SSize_t len = SvCUR(check) - !!SvTAIL(check);
914 const char * const anchor =
915 (prog->intflags & PREGf_ANCH_GPOS ? strpos : strbeg);
917 /* do a bytes rather than chars comparison. It's conservative;
918 * so it skips doing the HOP if the result can't possibly end
919 * up earlier than the old value of end_point.
921 if ((char*)end_point - anchor > prog->check_offset_max) {
922 end_point = HOP3lim((U8*)anchor,
923 prog->check_offset_max,
929 DEBUG_OPTIMISE_MORE_r({
930 PerlIO_printf(Perl_debug_log, " fbm_instr len=%d str=<%.*s>\n",
931 (int)(end_point - start_point),
932 (int)(end_point - start_point) > 20 ? 20 : (int)(end_point - start_point),
936 check_at = fbm_instr( start_point, end_point,
937 check, multiline ? FBMrf_MULTILINE : 0);
939 /* Update the count-of-usability, remove useless subpatterns,
943 RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0),
944 SvPVX_const(check), RE_SV_DUMPLEN(check), 30);
945 PerlIO_printf(Perl_debug_log, " %s %s substr %s%s%s",
946 (check_at ? "Found" : "Did not find"),
947 (check == (utf8_target ? prog->anchored_utf8 : prog->anchored_substr)
948 ? "anchored" : "floating"),
951 (check_at ? " at offset " : "...\n") );
956 /* Finish the diagnostic message */
957 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%ld...\n", (long)(check_at - strpos)) );
959 /* set rx_origin to the minimum position where the regex could start
960 * matching, given the constraint of the just-matched check substring.
961 * But don't set it lower than previously.
964 if (check_at - rx_origin > prog->check_offset_max)
965 rx_origin = HOP3c(check_at, -prog->check_offset_max, rx_origin);
969 /* now look for the 'other' substring if defined */
971 if (utf8_target ? prog->substrs->data[other_ix].utf8_substr
972 : prog->substrs->data[other_ix].substr)
974 /* Take into account the "other" substring. */
978 struct reg_substr_datum *other;
981 other = &prog->substrs->data[other_ix];
983 /* if "other" is anchored:
984 * we've previously found a floating substr starting at check_at.
985 * This means that the regex origin must lie somewhere
986 * between min (rx_origin): HOP3(check_at, -check_offset_max)
987 * and max: HOP3(check_at, -check_offset_min)
988 * (except that min will be >= strpos)
989 * So the fixed substr must lie somewhere between
990 * HOP3(min, anchored_offset)
991 * HOP3(max, anchored_offset) + SvCUR(substr)
994 /* if "other" is floating
995 * Calculate last1, the absolute latest point where the
996 * floating substr could start in the string, ignoring any
997 * constraints from the earlier fixed match. It is calculated
1000 * strend - prog->minlen (in chars) is the absolute latest
1001 * position within the string where the origin of the regex
1002 * could appear. The latest start point for the floating
1003 * substr is float_min_offset(*) on from the start of the
1004 * regex. last1 simply combines thee two offsets.
1006 * (*) You might think the latest start point should be
1007 * float_max_offset from the regex origin, and technically
1008 * you'd be correct. However, consider
1010 * Here, float min, max are 3,5 and minlen is 7.
1011 * This can match either
1015 * In the first case, the regex matches minlen chars; in the
1016 * second, minlen+1, in the third, minlen+2.
1017 * In the first case, the floating offset is 3 (which equals
1018 * float_min), in the second, 4, and in the third, 5 (which
1019 * equals float_max). In all cases, the floating string bcd
1020 * can never start more than 4 chars from the end of the
1021 * string, which equals minlen - float_min. As the substring
1022 * starts to match more than float_min from the start of the
1023 * regex, it makes the regex match more than minlen chars,
1024 * and the two cancel each other out. So we can always use
1025 * float_min - minlen, rather than float_max - minlen for the
1026 * latest position in the string.
1028 * Note that -minlen + float_min_offset is equivalent (AFAIKT)
1029 * to CHR_SVLEN(must) - !!SvTAIL(must) + prog->float_end_shift
1032 assert(prog->minlen >= other->min_offset);
1033 last1 = HOP3c(strend,
1034 other->min_offset - prog->minlen, strbeg);
1036 if (other_ix) {/* i.e. if (other-is-float) */
1037 /* last is the latest point where the floating substr could
1038 * start, *given* any constraints from the earlier fixed
1039 * match. This constraint is that the floating string starts
1040 * <= float_max_offset chars from the regex origin (rx_origin).
1041 * If this value is less than last1, use it instead.
1043 assert(rx_origin <= last1);
1045 /* this condition handles the offset==infinity case, and
1046 * is a short-cut otherwise. Although it's comparing a
1047 * byte offset to a char length, it does so in a safe way,
1048 * since 1 char always occupies 1 or more bytes,
1049 * so if a string range is (last1 - rx_origin) bytes,
1050 * it will be less than or equal to (last1 - rx_origin)
1051 * chars; meaning it errs towards doing the accurate HOP3
1052 * rather than just using last1 as a short-cut */
1053 (last1 - rx_origin) < other->max_offset
1055 : (char*)HOP3lim(rx_origin, other->max_offset, last1);
1058 assert(strpos + start_shift <= check_at);
1059 last = HOP4c(check_at, other->min_offset - start_shift,
1063 s = HOP3c(rx_origin, other->min_offset, strend);
1064 if (s < other_last) /* These positions already checked */
1067 must = utf8_target ? other->utf8_substr : other->substr;
1068 assert(SvPOK(must));
1071 (unsigned char*)last + SvCUR(must) - (SvTAIL(must)!=0),
1073 multiline ? FBMrf_MULTILINE : 0
1076 RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0),
1077 SvPVX_const(must), RE_SV_DUMPLEN(must), 30);
1078 PerlIO_printf(Perl_debug_log, " %s %s substr %s%s",
1079 s ? "Found" : "Contradicts",
1080 other_ix ? "floating" : "anchored",
1081 quoted, RE_SV_TAIL(must));
1086 /* last1 is latest possible substr location. If we didn't
1087 * find it before there, we never will */
1088 if (last >= last1) {
1089 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
1090 ", giving up...\n"));
1094 /* try to find the check substr again at a later
1095 * position. Maybe next time we'll find the "other" substr
1097 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
1098 ", trying %s at offset %ld...\n",
1099 (other_ix ? "floating" : "anchored"),
1100 (long)(HOP3c(check_at, 1, strend) - strpos)));
1102 other_last = HOP3c(last, 1, strend) /* highest failure */;
1104 other_ix /* i.e. if other-is-float */
1105 ? HOP3c(rx_origin, 1, strend)
1106 : HOP4c(last, 1 - other->min_offset, strbeg, strend);
1110 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
1111 (long)(s - strpos)));
1113 if (other_ix) { /* if (other-is-float) */
1114 /* other_last is set to s, not s+1, since its possible for
1115 * a floating substr to fail first time, then succeed
1116 * second time at the same floating position; e.g.:
1117 * "-AB--AABZ" =~ /\wAB\d*Z/
1118 * The first time round, anchored and float match at
1119 * "-(AB)--AAB(Z)" then fail on the initial \w character
1120 * class. Second time round, they match at "-AB--A(AB)(Z)".
1125 rx_origin = HOP3c(s, -other->min_offset, strbeg);
1126 other_last = HOP3c(s, 1, strend);
1131 DEBUG_OPTIMISE_MORE_r(
1132 PerlIO_printf(Perl_debug_log,
1133 " Check-only match: offset min:%"IVdf" max:%"IVdf
1134 " check_at:%"IVdf" rx_origin:%"IVdf" rx_origin-check_at:%"IVdf
1135 " strend-strpos:%"IVdf"\n",
1136 (IV)prog->check_offset_min,
1137 (IV)prog->check_offset_max,
1138 (IV)(check_at-strpos),
1139 (IV)(rx_origin-strpos),
1140 (IV)(rx_origin-check_at),
1146 postprocess_substr_matches:
1148 /* handle the extra constraint of /^.../m if present */
1150 if (ml_anch && rx_origin != strbeg && rx_origin[-1] != '\n') {
1153 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
1154 " looking for /^/m anchor"));
1156 /* we have failed the constraint of a \n before rx_origin.
1157 * Find the next \n, if any, even if it's beyond the current
1158 * anchored and/or floating substrings. Whether we should be
1159 * scanning ahead for the next \n or the next substr is debatable.
1160 * On the one hand you'd expect rare substrings to appear less
1161 * often than \n's. On the other hand, searching for \n means
1162 * we're effectively flipping been check_substr and "\n" on each
1163 * iteration as the current "rarest" string candidate, which
1164 * means for example that we'll quickly reject the whole string if
1165 * hasn't got a \n, rather than trying every substr position
1169 s = HOP3c(strend, - prog->minlen, strpos);
1170 if (s <= rx_origin ||
1171 ! ( rx_origin = (char *)memchr(rx_origin, '\n', s - rx_origin)))
1173 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
1174 " Did not find /%s^%s/m...\n",
1175 PL_colors[0], PL_colors[1]));
1179 /* earliest possible origin is 1 char after the \n.
1180 * (since *rx_origin == '\n', it's safe to ++ here rather than
1181 * HOP(rx_origin, 1)) */
1184 if (prog->substrs->check_ix == 0 /* check is anchored */
1185 || rx_origin >= HOP3c(check_at, - prog->check_offset_min, strpos))
1187 /* Position contradicts check-string; either because
1188 * check was anchored (and thus has no wiggle room),
1189 * or check was float and rx_origin is above the float range */
1190 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
1191 " Found /%s^%s/m, restarting lookup for check-string at offset %ld...\n",
1192 PL_colors[0], PL_colors[1], (long)(rx_origin - strpos)));
1196 /* if we get here, the check substr must have been float,
1197 * is in range, and we may or may not have had an anchored
1198 * "other" substr which still contradicts */
1199 assert(prog->substrs->check_ix); /* check is float */
1201 if (utf8_target ? prog->anchored_utf8 : prog->anchored_substr) {
1202 /* whoops, the anchored "other" substr exists, so we still
1203 * contradict. On the other hand, the float "check" substr
1204 * didn't contradict, so just retry the anchored "other"
1206 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
1207 " Found /%s^%s/m at offset %ld, rescanning for anchored from offset %ld...\n",
1208 PL_colors[0], PL_colors[1],
1209 (long)(rx_origin - strpos),
1210 (long)(rx_origin - strpos + prog->anchored_offset)));
1211 goto do_other_substr;
1214 /* success: we don't contradict the found floating substring
1215 * (and there's no anchored substr). */
1216 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
1217 " Found /%s^%s/m at offset %ld...\n",
1218 PL_colors[0], PL_colors[1], (long)(rx_origin - strpos)));
1221 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
1222 " (multiline anchor test skipped)\n"));
1228 /* if we have a starting character class, then test that extra constraint.
1229 * (trie stclasses are too expensive to use here, we are better off to
1230 * leave it to regmatch itself) */
1232 if (progi->regstclass && PL_regkind[OP(progi->regstclass)]!=TRIE) {
1233 const U8* const str = (U8*)STRING(progi->regstclass);
1235 /* XXX this value could be pre-computed */
1236 const int cl_l = (PL_regkind[OP(progi->regstclass)] == EXACT
1237 ? (reginfo->is_utf8_pat
1238 ? utf8_distance(str + STR_LEN(progi->regstclass), str)
1239 : STR_LEN(progi->regstclass))
1243 /* latest pos that a matching float substr constrains rx start to */
1244 char *rx_max_float = NULL;
1246 /* if the current rx_origin is anchored, either by satisfying an
1247 * anchored substring constraint, or a /^.../m constraint, then we
1248 * can reject the current origin if the start class isn't found
1249 * at the current position. If we have a float-only match, then
1250 * rx_origin is constrained to a range; so look for the start class
1251 * in that range. if neither, then look for the start class in the
1252 * whole rest of the string */
1254 /* XXX DAPM it's not clear what the minlen test is for, and why
1255 * it's not used in the floating case. Nothing in the test suite
1256 * causes minlen == 0 here. See <20140313134639.GS12844@iabyn.com>.
1257 * Here are some old comments, which may or may not be correct:
1259 * minlen == 0 is possible if regstclass is \b or \B,
1260 * and the fixed substr is ''$.
1261 * Since minlen is already taken into account, rx_origin+1 is
1262 * before strend; accidentally, minlen >= 1 guaranties no false
1263 * positives at rx_origin + 1 even for \b or \B. But (minlen? 1 :
1264 * 0) below assumes that regstclass does not come from lookahead...
1265 * If regstclass takes bytelength more than 1: If charlength==1, OK.
1266 * This leaves EXACTF-ish only, which are dealt with in
1270 if (prog->anchored_substr || prog->anchored_utf8 || ml_anch)
1271 endpos= HOP3c(rx_origin, (prog->minlen ? cl_l : 0), strend);
1272 else if (prog->float_substr || prog->float_utf8) {
1273 rx_max_float = HOP3c(check_at, -start_shift, strbeg);
1274 endpos= HOP3c(rx_max_float, cl_l, strend);
1279 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
1280 " looking for class: start_shift: %"IVdf" check_at: %"IVdf
1281 " rx_origin: %"IVdf" endpos: %"IVdf"\n",
1282 (IV)start_shift, (IV)(check_at - strbeg),
1283 (IV)(rx_origin - strbeg), (IV)(endpos - strbeg)));
1285 s = find_byclass(prog, progi->regstclass, rx_origin, endpos,
1288 if (endpos == strend) {
1289 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
1290 " Could not match STCLASS...\n") );
1293 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
1294 " This position contradicts STCLASS...\n") );
1295 if ((prog->intflags & PREGf_ANCH) && !ml_anch
1296 && !(prog->intflags & PREGf_IMPLICIT))
1299 /* Contradict one of substrings */
1300 if (prog->anchored_substr || prog->anchored_utf8) {
1301 if (prog->substrs->check_ix == 1) { /* check is float */
1302 /* Have both, check_string is floating */
1303 assert(rx_origin + start_shift <= check_at);
1304 if (rx_origin + start_shift != check_at) {
1305 /* not at latest position float substr could match:
1306 * Recheck anchored substring, but not floating.
1307 * The condition above is in bytes rather than
1308 * chars for efficiency. It's conservative, in
1309 * that it errs on the side of doing 'goto
1310 * do_other_substr', where a more accurate
1311 * char-based calculation will be done */
1312 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
1313 " Looking for anchored substr starting at offset %ld...\n",
1314 (long)(other_last - strpos)) );
1315 goto do_other_substr;
1323 /* In the presence of ml_anch, we might be able to
1324 * find another \n without breaking the current float
1327 /* strictly speaking this should be HOP3c(..., 1, ...),
1328 * but since we goto a block of code that's going to
1329 * search for the next \n if any, its safe here */
1331 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
1332 " Looking for /%s^%s/m starting at offset %ld...\n",
1333 PL_colors[0], PL_colors[1],
1334 (long)(rx_origin - strpos)) );
1335 goto postprocess_substr_matches;
1338 /* strictly speaking this can never be true; but might
1339 * be if we ever allow intuit without substrings */
1340 if (!(utf8_target ? prog->float_utf8 : prog->float_substr))
1343 rx_origin = rx_max_float;
1346 /* at this point, any matching substrings have been
1347 * contradicted. Start again... */
1349 rx_origin = HOP3c(rx_origin, 1, strend);
1351 /* uses bytes rather than char calculations for efficiency.
1352 * It's conservative: it errs on the side of doing 'goto restart',
1353 * where there is code that does a proper char-based test */
1354 if (rx_origin + start_shift + end_shift > strend) {
1355 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
1356 " Could not match STCLASS...\n") );
1359 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
1360 " Looking for %s substr starting at offset %ld...\n",
1361 (prog->substrs->check_ix ? "floating" : "anchored"),
1362 (long)(rx_origin + start_shift - strpos)) );
1368 if (rx_origin != s) {
1369 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
1370 " By STCLASS: moving %ld --> %ld\n",
1371 (long)(rx_origin - strpos), (long)(s - strpos))
1375 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
1376 " Does not contradict STCLASS...\n");
1381 /* Decide whether using the substrings helped */
1383 if (rx_origin != strpos) {
1384 /* Fixed substring is found far enough so that the match
1385 cannot start at strpos. */
1387 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " try at offset...\n"));
1388 ++BmUSEFUL(utf8_target ? prog->check_utf8 : prog->check_substr); /* hooray/5 */
1391 /* The found rx_origin position does not prohibit matching at
1392 * strpos, so calling intuit didn't gain us anything. Decrement
1393 * the BmUSEFUL() count on the check substring, and if we reach
1395 if (!(prog->intflags & PREGf_NAUGHTY)
1397 prog->check_utf8 /* Could be deleted already */
1398 && --BmUSEFUL(prog->check_utf8) < 0
1399 && (prog->check_utf8 == prog->float_utf8)
1401 prog->check_substr /* Could be deleted already */
1402 && --BmUSEFUL(prog->check_substr) < 0
1403 && (prog->check_substr == prog->float_substr)
1406 /* If flags & SOMETHING - do not do it many times on the same match */
1407 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " ... Disabling check substring...\n"));
1408 /* XXX Does the destruction order has to change with utf8_target? */
1409 SvREFCNT_dec(utf8_target ? prog->check_utf8 : prog->check_substr);
1410 SvREFCNT_dec(utf8_target ? prog->check_substr : prog->check_utf8);
1411 prog->check_substr = prog->check_utf8 = NULL; /* disable */
1412 prog->float_substr = prog->float_utf8 = NULL; /* clear */
1413 check = NULL; /* abort */
1414 /* XXXX This is a remnant of the old implementation. It
1415 looks wasteful, since now INTUIT can use many
1416 other heuristics. */
1417 prog->extflags &= ~RXf_USE_INTUIT;
1421 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
1422 "Intuit: %sSuccessfully guessed:%s match at offset %ld\n",
1423 PL_colors[4], PL_colors[5], (long)(rx_origin - strpos)) );
1427 fail_finish: /* Substring not found */
1428 if (prog->check_substr || prog->check_utf8) /* could be removed already */
1429 BmUSEFUL(utf8_target ? prog->check_utf8 : prog->check_substr) += 5; /* hooray */
1431 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch rejected by optimizer%s\n",
1432 PL_colors[4], PL_colors[5]));
1437 #define DECL_TRIE_TYPE(scan) \
1438 const enum { trie_plain, trie_utf8, trie_utf8_fold, trie_latin_utf8_fold, \
1439 trie_utf8_exactfa_fold, trie_latin_utf8_exactfa_fold, \
1440 trie_utf8l, trie_flu8 } \
1441 trie_type = ((scan->flags == EXACT) \
1442 ? (utf8_target ? trie_utf8 : trie_plain) \
1443 : (scan->flags == EXACTL) \
1444 ? (utf8_target ? trie_utf8l : trie_plain) \
1445 : (scan->flags == EXACTFA) \
1447 ? trie_utf8_exactfa_fold \
1448 : trie_latin_utf8_exactfa_fold) \
1449 : (scan->flags == EXACTFLU8 \
1453 : trie_latin_utf8_fold)))
1455 #define REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc, uscan, len, uvc, charid, foldlen, foldbuf, uniflags) \
1458 U8 flags = FOLD_FLAGS_FULL; \
1459 switch (trie_type) { \
1461 _CHECK_AND_WARN_PROBLEMATIC_LOCALE; \
1462 if (utf8_target && UTF8_IS_ABOVE_LATIN1(*uc)) { \
1463 _CHECK_AND_OUTPUT_WIDE_LOCALE_UTF8_MSG(uc, uc + UTF8SKIP(uc)); \
1465 goto do_trie_utf8_fold; \
1466 case trie_utf8_exactfa_fold: \
1467 flags |= FOLD_FLAGS_NOMIX_ASCII; \
1469 case trie_utf8_fold: \
1470 do_trie_utf8_fold: \
1471 if ( foldlen>0 ) { \
1472 uvc = utf8n_to_uvchr( (const U8*) uscan, UTF8_MAXLEN, &len, uniflags ); \
1477 uvc = _to_utf8_fold_flags( (const U8*) uc, foldbuf, &foldlen, flags); \
1478 len = UTF8SKIP(uc); \
1479 skiplen = UNISKIP( uvc ); \
1480 foldlen -= skiplen; \
1481 uscan = foldbuf + skiplen; \
1484 case trie_latin_utf8_exactfa_fold: \
1485 flags |= FOLD_FLAGS_NOMIX_ASCII; \
1487 case trie_latin_utf8_fold: \
1488 if ( foldlen>0 ) { \
1489 uvc = utf8n_to_uvchr( (const U8*) uscan, UTF8_MAXLEN, &len, uniflags ); \
1495 uvc = _to_fold_latin1( (U8) *uc, foldbuf, &foldlen, flags); \
1496 skiplen = UNISKIP( uvc ); \
1497 foldlen -= skiplen; \
1498 uscan = foldbuf + skiplen; \
1502 _CHECK_AND_WARN_PROBLEMATIC_LOCALE; \
1503 if (utf8_target && UTF8_IS_ABOVE_LATIN1(*uc)) { \
1504 _CHECK_AND_OUTPUT_WIDE_LOCALE_UTF8_MSG(uc, uc + UTF8SKIP(uc)); \
1508 uvc = utf8n_to_uvchr( (const U8*) uc, UTF8_MAXLEN, &len, uniflags ); \
1515 charid = trie->charmap[ uvc ]; \
1519 if (widecharmap) { \
1520 SV** const svpp = hv_fetch(widecharmap, \
1521 (char*)&uvc, sizeof(UV), 0); \
1523 charid = (U16)SvIV(*svpp); \
1528 #define DUMP_EXEC_POS(li,s,doutf8) \
1529 dump_exec_pos(li,s,(reginfo->strend),(reginfo->strbeg), \
1532 #define REXEC_FBC_EXACTISH_SCAN(COND) \
1536 && (ln == 1 || folder(s, pat_string, ln)) \
1537 && (reginfo->intuit || regtry(reginfo, &s)) )\
1543 #define REXEC_FBC_UTF8_SCAN(CODE) \
1545 while (s < strend) { \
1551 #define REXEC_FBC_SCAN(CODE) \
1553 while (s < strend) { \
1559 #define REXEC_FBC_UTF8_CLASS_SCAN(COND) \
1560 REXEC_FBC_UTF8_SCAN( /* Loops while (s < strend) */ \
1562 if (tmp && (reginfo->intuit || regtry(reginfo, &s))) \
1571 #define REXEC_FBC_CLASS_SCAN(COND) \
1572 REXEC_FBC_SCAN( /* Loops while (s < strend) */ \
1574 if (tmp && (reginfo->intuit || regtry(reginfo, &s))) \
1583 #define REXEC_FBC_CSCAN(CONDUTF8,COND) \
1584 if (utf8_target) { \
1585 REXEC_FBC_UTF8_CLASS_SCAN(CONDUTF8); \
1588 REXEC_FBC_CLASS_SCAN(COND); \
1591 /* The three macros below are slightly different versions of the same logic.
1593 * The first is for /a and /aa when the target string is UTF-8. This can only
1594 * match ascii, but it must advance based on UTF-8. The other two handle the
1595 * non-UTF-8 and the more generic UTF-8 cases. In all three, we are looking
1596 * for the boundary (or non-boundary) between a word and non-word character.
1597 * The utf8 and non-utf8 cases have the same logic, but the details must be
1598 * different. Find the "wordness" of the character just prior to this one, and
1599 * compare it with the wordness of this one. If they differ, we have a
1600 * boundary. At the beginning of the string, pretend that the previous
1601 * character was a new-line.
1603 * All these macros uncleanly have side-effects with each other and outside
1604 * variables. So far it's been too much trouble to clean-up
1606 * TEST_NON_UTF8 is the macro or function to call to test if its byte input is
1607 * a word character or not.
1608 * IF_SUCCESS is code to do if it finds that we are at a boundary between
1610 * IF_FAIL is code to do if we aren't at a boundary between word/non-word
1612 * Exactly one of the two IF_FOO parameters is a no-op, depending on whether we
1613 * are looking for a boundary or for a non-boundary. If we are looking for a
1614 * boundary, we want IF_FAIL to be the no-op, and for IF_SUCCESS to go out and
1615 * see if this tentative match actually works, and if so, to quit the loop
1616 * here. And vice-versa if we are looking for a non-boundary.
1618 * 'tmp' below in the next three macros in the REXEC_FBC_SCAN and
1619 * REXEC_FBC_UTF8_SCAN loops is a loop invariant, a bool giving the return of
1620 * TEST_NON_UTF8(s-1). To see this, note that that's what it is defined to be
1621 * at entry to the loop, and to get to the IF_FAIL branch, tmp must equal
1622 * TEST_NON_UTF8(s), and in the opposite branch, IF_SUCCESS, tmp is that
1623 * complement. But in that branch we complement tmp, meaning that at the
1624 * bottom of the loop tmp is always going to be equal to TEST_NON_UTF8(s),
1625 * which means at the top of the loop in the next iteration, it is
1626 * TEST_NON_UTF8(s-1) */
1627 #define FBC_UTF8_A(TEST_NON_UTF8, IF_SUCCESS, IF_FAIL) \
1628 tmp = (s != reginfo->strbeg) ? UCHARAT(s - 1) : '\n'; \
1629 tmp = TEST_NON_UTF8(tmp); \
1630 REXEC_FBC_UTF8_SCAN( /* advances s while s < strend */ \
1631 if (tmp == ! TEST_NON_UTF8((U8) *s)) { \
1633 IF_SUCCESS; /* Is a boundary if values for s-1 and s differ */ \
1640 /* Like FBC_UTF8_A, but TEST_UV is a macro which takes a UV as its input, and
1641 * TEST_UTF8 is a macro that for the same input code points returns identically
1642 * to TEST_UV, but takes a pointer to a UTF-8 encoded string instead */
1643 #define FBC_UTF8(TEST_UV, TEST_UTF8, IF_SUCCESS, IF_FAIL) \
1644 if (s == reginfo->strbeg) { \
1647 else { /* Back-up to the start of the previous character */ \
1648 U8 * const r = reghop3((U8*)s, -1, (U8*)reginfo->strbeg); \
1649 tmp = utf8n_to_uvchr(r, (U8*) reginfo->strend - r, \
1650 0, UTF8_ALLOW_DEFAULT); \
1652 tmp = TEST_UV(tmp); \
1653 LOAD_UTF8_CHARCLASS_ALNUM(); \
1654 REXEC_FBC_UTF8_SCAN( /* advances s while s < strend */ \
1655 if (tmp == ! (TEST_UTF8((U8 *) s))) { \
1664 /* Like the above two macros. UTF8_CODE is the complete code for handling
1665 * UTF-8. Common to the BOUND and NBOUND cases, set-up by the FBC_BOUND, etc
1667 #define FBC_BOUND_COMMON(UTF8_CODE, TEST_NON_UTF8, IF_SUCCESS, IF_FAIL) \
1668 if (utf8_target) { \
1671 else { /* Not utf8 */ \
1672 tmp = (s != reginfo->strbeg) ? UCHARAT(s - 1) : '\n'; \
1673 tmp = TEST_NON_UTF8(tmp); \
1674 REXEC_FBC_SCAN( /* advances s while s < strend */ \
1675 if (tmp == ! TEST_NON_UTF8((U8) *s)) { \
1684 /* Here, things have been set up by the previous code so that tmp is the \
1685 * return of TEST_NON_UTF(s-1) or TEST_UTF8(s-1) (depending on the \
1686 * utf8ness of the target). We also have to check if this matches against \
1687 * the EOS, which we treat as a \n (which is the same value in both UTF-8 \
1688 * or non-UTF8, so can use the non-utf8 test condition even for a UTF-8 \
1690 if (tmp == ! TEST_NON_UTF8('\n')) { \
1697 /* This is the macro to use when we want to see if something that looks like it
1698 * could match, actually does, and if so exits the loop */
1699 #define REXEC_FBC_TRYIT \
1700 if ((reginfo->intuit || regtry(reginfo, &s))) \
1703 /* The only difference between the BOUND and NBOUND cases is that
1704 * REXEC_FBC_TRYIT is called when matched in BOUND, and when non-matched in
1705 * NBOUND. This is accomplished by passing it as either the if or else clause,
1706 * with the other one being empty (PLACEHOLDER is defined as empty).
1708 * The TEST_FOO parameters are for operating on different forms of input, but
1709 * all should be ones that return identically for the same underlying code
1711 #define FBC_BOUND(TEST_NON_UTF8, TEST_UV, TEST_UTF8) \
1713 FBC_UTF8(TEST_UV, TEST_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER), \
1714 TEST_NON_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER)
1716 #define FBC_BOUND_A(TEST_NON_UTF8, TEST_UV, TEST_UTF8) \
1718 FBC_UTF8_A(TEST_NON_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER), \
1719 TEST_NON_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER)
1721 #define FBC_NBOUND(TEST_NON_UTF8, TEST_UV, TEST_UTF8) \
1723 FBC_UTF8(TEST_UV, TEST_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT), \
1724 TEST_NON_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT)
1726 #define FBC_NBOUND_A(TEST_NON_UTF8, TEST_UV, TEST_UTF8) \
1728 FBC_UTF8_A(TEST_NON_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT), \
1729 TEST_NON_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT)
1732 /* We know what class REx starts with. Try to find this position... */
1733 /* if reginfo->intuit, its a dryrun */
1734 /* annoyingly all the vars in this routine have different names from their counterparts
1735 in regmatch. /grrr */
1737 S_find_byclass(pTHX_ regexp * prog, const regnode *c, char *s,
1738 const char *strend, regmatch_info *reginfo)
1741 const I32 doevery = (prog->intflags & PREGf_SKIP) == 0;
1742 char *pat_string; /* The pattern's exactish string */
1743 char *pat_end; /* ptr to end char of pat_string */
1744 re_fold_t folder; /* Function for computing non-utf8 folds */
1745 const U8 *fold_array; /* array for folding ords < 256 */
1751 I32 tmp = 1; /* Scratch variable? */
1752 const bool utf8_target = reginfo->is_utf8_target;
1753 UV utf8_fold_flags = 0;
1754 const bool is_utf8_pat = reginfo->is_utf8_pat;
1755 bool to_complement = FALSE; /* Invert the result? Taking the xor of this
1756 with a result inverts that result, as 0^1 =
1758 _char_class_number classnum;
1760 RXi_GET_DECL(prog,progi);
1762 PERL_ARGS_ASSERT_FIND_BYCLASS;
1764 /* We know what class it must start with. */
1767 _CHECK_AND_WARN_PROBLEMATIC_LOCALE;
1771 REXEC_FBC_UTF8_CLASS_SCAN(
1772 reginclass(prog, c, (U8*)s, (U8*) strend, utf8_target));
1775 REXEC_FBC_CLASS_SCAN(REGINCLASS(prog, c, (U8*)s));
1780 if (tmp && (reginfo->intuit || regtry(reginfo, &s)))
1787 case EXACTFA_NO_TRIE: /* This node only generated for non-utf8 patterns */
1788 assert(! is_utf8_pat);
1791 if (is_utf8_pat || utf8_target) {
1792 utf8_fold_flags = FOLDEQ_UTF8_NOMIX_ASCII;
1793 goto do_exactf_utf8;
1795 fold_array = PL_fold_latin1; /* Latin1 folds are not affected by */
1796 folder = foldEQ_latin1; /* /a, except the sharp s one which */
1797 goto do_exactf_non_utf8; /* isn't dealt with by these */
1799 case EXACTF: /* This node only generated for non-utf8 patterns */
1800 assert(! is_utf8_pat);
1802 utf8_fold_flags = 0;
1803 goto do_exactf_utf8;
1805 fold_array = PL_fold;
1807 goto do_exactf_non_utf8;
1810 _CHECK_AND_WARN_PROBLEMATIC_LOCALE;
1811 if (is_utf8_pat || utf8_target || IN_UTF8_CTYPE_LOCALE) {
1812 utf8_fold_flags = FOLDEQ_LOCALE;
1813 goto do_exactf_utf8;
1815 fold_array = PL_fold_locale;
1816 folder = foldEQ_locale;
1817 goto do_exactf_non_utf8;
1821 utf8_fold_flags = FOLDEQ_S2_ALREADY_FOLDED;
1823 goto do_exactf_utf8;
1826 if (! utf8_target) { /* All code points in this node require
1827 UTF-8 to express. */
1830 utf8_fold_flags = FOLDEQ_LOCALE | FOLDEQ_S2_ALREADY_FOLDED
1831 | FOLDEQ_S2_FOLDS_SANE;
1832 goto do_exactf_utf8;
1835 if (is_utf8_pat || utf8_target) {
1836 utf8_fold_flags = is_utf8_pat ? FOLDEQ_S2_ALREADY_FOLDED : 0;
1837 goto do_exactf_utf8;
1840 /* Any 'ss' in the pattern should have been replaced by regcomp,
1841 * so we don't have to worry here about this single special case
1842 * in the Latin1 range */
1843 fold_array = PL_fold_latin1;
1844 folder = foldEQ_latin1;
1848 do_exactf_non_utf8: /* Neither pattern nor string are UTF8, and there
1849 are no glitches with fold-length differences
1850 between the target string and pattern */
1852 /* The idea in the non-utf8 EXACTF* cases is to first find the
1853 * first character of the EXACTF* node and then, if necessary,
1854 * case-insensitively compare the full text of the node. c1 is the
1855 * first character. c2 is its fold. This logic will not work for
1856 * Unicode semantics and the german sharp ss, which hence should
1857 * not be compiled into a node that gets here. */
1858 pat_string = STRING(c);
1859 ln = STR_LEN(c); /* length to match in octets/bytes */
1861 /* We know that we have to match at least 'ln' bytes (which is the
1862 * same as characters, since not utf8). If we have to match 3
1863 * characters, and there are only 2 availabe, we know without
1864 * trying that it will fail; so don't start a match past the
1865 * required minimum number from the far end */
1866 e = HOP3c(strend, -((SSize_t)ln), s);
1868 if (reginfo->intuit && e < s) {
1869 e = s; /* Due to minlen logic of intuit() */
1873 c2 = fold_array[c1];
1874 if (c1 == c2) { /* If char and fold are the same */
1875 REXEC_FBC_EXACTISH_SCAN(*(U8*)s == c1);
1878 REXEC_FBC_EXACTISH_SCAN(*(U8*)s == c1 || *(U8*)s == c2);
1886 /* If one of the operands is in utf8, we can't use the simpler folding
1887 * above, due to the fact that many different characters can have the
1888 * same fold, or portion of a fold, or different- length fold */
1889 pat_string = STRING(c);
1890 ln = STR_LEN(c); /* length to match in octets/bytes */
1891 pat_end = pat_string + ln;
1892 lnc = is_utf8_pat /* length to match in characters */
1893 ? utf8_length((U8 *) pat_string, (U8 *) pat_end)
1896 /* We have 'lnc' characters to match in the pattern, but because of
1897 * multi-character folding, each character in the target can match
1898 * up to 3 characters (Unicode guarantees it will never exceed
1899 * this) if it is utf8-encoded; and up to 2 if not (based on the
1900 * fact that the Latin 1 folds are already determined, and the
1901 * only multi-char fold in that range is the sharp-s folding to
1902 * 'ss'. Thus, a pattern character can match as little as 1/3 of a
1903 * string character. Adjust lnc accordingly, rounding up, so that
1904 * if we need to match at least 4+1/3 chars, that really is 5. */
1905 expansion = (utf8_target) ? UTF8_MAX_FOLD_CHAR_EXPAND : 2;
1906 lnc = (lnc + expansion - 1) / expansion;
1908 /* As in the non-UTF8 case, if we have to match 3 characters, and
1909 * only 2 are left, it's guaranteed to fail, so don't start a
1910 * match that would require us to go beyond the end of the string
1912 e = HOP3c(strend, -((SSize_t)lnc), s);
1914 if (reginfo->intuit && e < s) {
1915 e = s; /* Due to minlen logic of intuit() */
1918 /* XXX Note that we could recalculate e to stop the loop earlier,
1919 * as the worst case expansion above will rarely be met, and as we
1920 * go along we would usually find that e moves further to the left.
1921 * This would happen only after we reached the point in the loop
1922 * where if there were no expansion we should fail. Unclear if
1923 * worth the expense */
1926 char *my_strend= (char *)strend;
1927 if (foldEQ_utf8_flags(s, &my_strend, 0, utf8_target,
1928 pat_string, NULL, ln, is_utf8_pat, utf8_fold_flags)
1929 && (reginfo->intuit || regtry(reginfo, &s)) )
1933 s += (utf8_target) ? UTF8SKIP(s) : 1;
1939 _CHECK_AND_WARN_PROBLEMATIC_LOCALE;
1940 FBC_BOUND(isWORDCHAR_LC, isWORDCHAR_LC_uvchr, isWORDCHAR_LC_utf8);
1943 _CHECK_AND_WARN_PROBLEMATIC_LOCALE;
1944 FBC_NBOUND(isWORDCHAR_LC, isWORDCHAR_LC_uvchr, isWORDCHAR_LC_utf8);
1947 FBC_BOUND(isWORDCHAR, isWORDCHAR_uni, isWORDCHAR_utf8);
1950 FBC_BOUND_A(isWORDCHAR_A, isWORDCHAR_A, isWORDCHAR_A);
1953 FBC_NBOUND(isWORDCHAR, isWORDCHAR_uni, isWORDCHAR_utf8);
1956 FBC_NBOUND_A(isWORDCHAR_A, isWORDCHAR_A, isWORDCHAR_A);
1959 FBC_BOUND(isWORDCHAR_L1, isWORDCHAR_uni, isWORDCHAR_utf8);
1962 FBC_NBOUND(isWORDCHAR_L1, isWORDCHAR_uni, isWORDCHAR_utf8);
1965 REXEC_FBC_CSCAN(is_LNBREAK_utf8_safe(s, strend),
1966 is_LNBREAK_latin1_safe(s, strend)
1970 /* The argument to all the POSIX node types is the class number to pass to
1971 * _generic_isCC() to build a mask for searching in PL_charclass[] */
1978 _CHECK_AND_WARN_PROBLEMATIC_LOCALE;
1979 REXEC_FBC_CSCAN(to_complement ^ cBOOL(isFOO_utf8_lc(FLAGS(c), (U8 *) s)),
1980 to_complement ^ cBOOL(isFOO_lc(FLAGS(c), *s)));
1995 /* The complement of something that matches only ASCII matches all
1996 * non-ASCII, plus everything in ASCII that isn't in the class. */
1997 REXEC_FBC_UTF8_CLASS_SCAN(! isASCII_utf8(s)
1998 || ! _generic_isCC_A(*s, FLAGS(c)));
2007 /* Don't need to worry about utf8, as it can match only a single
2008 * byte invariant character. */
2009 REXEC_FBC_CLASS_SCAN(
2010 to_complement ^ cBOOL(_generic_isCC_A(*s, FLAGS(c))));
2018 if (! utf8_target) {
2019 REXEC_FBC_CLASS_SCAN(to_complement ^ cBOOL(_generic_isCC(*s,
2025 classnum = (_char_class_number) FLAGS(c);
2026 if (classnum < _FIRST_NON_SWASH_CC) {
2027 while (s < strend) {
2029 /* We avoid loading in the swash as long as possible, but
2030 * should we have to, we jump to a separate loop. This
2031 * extra 'if' statement is what keeps this code from being
2032 * just a call to REXEC_FBC_UTF8_CLASS_SCAN() */
2033 if (UTF8_IS_ABOVE_LATIN1(*s)) {
2034 goto found_above_latin1;
2036 if ((UTF8_IS_INVARIANT(*s)
2037 && to_complement ^ cBOOL(_generic_isCC((U8) *s,
2039 || (UTF8_IS_DOWNGRADEABLE_START(*s)
2040 && to_complement ^ cBOOL(
2041 _generic_isCC(TWO_BYTE_UTF8_TO_NATIVE(*s,
2045 if (tmp && (reginfo->intuit || regtry(reginfo, &s)))
2057 else switch (classnum) { /* These classes are implemented as
2059 case _CC_ENUM_SPACE: /* XXX would require separate code if we
2060 revert the change of \v matching this */
2063 case _CC_ENUM_PSXSPC:
2064 REXEC_FBC_UTF8_CLASS_SCAN(
2065 to_complement ^ cBOOL(isSPACE_utf8(s)));
2068 case _CC_ENUM_BLANK:
2069 REXEC_FBC_UTF8_CLASS_SCAN(
2070 to_complement ^ cBOOL(isBLANK_utf8(s)));
2073 case _CC_ENUM_XDIGIT:
2074 REXEC_FBC_UTF8_CLASS_SCAN(
2075 to_complement ^ cBOOL(isXDIGIT_utf8(s)));
2078 case _CC_ENUM_VERTSPACE:
2079 REXEC_FBC_UTF8_CLASS_SCAN(
2080 to_complement ^ cBOOL(isVERTWS_utf8(s)));
2083 case _CC_ENUM_CNTRL:
2084 REXEC_FBC_UTF8_CLASS_SCAN(
2085 to_complement ^ cBOOL(isCNTRL_utf8(s)));
2089 Perl_croak(aTHX_ "panic: find_byclass() node %d='%s' has an unexpected character class '%d'", OP(c), PL_reg_name[OP(c)], classnum);
2090 NOT_REACHED; /* NOTREACHED */
2095 found_above_latin1: /* Here we have to load a swash to get the result
2096 for the current code point */
2097 if (! PL_utf8_swash_ptrs[classnum]) {
2098 U8 flags = _CORE_SWASH_INIT_ACCEPT_INVLIST;
2099 PL_utf8_swash_ptrs[classnum] =
2100 _core_swash_init("utf8",
2103 PL_XPosix_ptrs[classnum], &flags);
2106 /* This is a copy of the loop above for swash classes, though using the
2107 * FBC macro instead of being expanded out. Since we've loaded the
2108 * swash, we don't have to check for that each time through the loop */
2109 REXEC_FBC_UTF8_CLASS_SCAN(
2110 to_complement ^ cBOOL(_generic_utf8(
2113 swash_fetch(PL_utf8_swash_ptrs[classnum],
2121 /* what trie are we using right now */
2122 reg_ac_data *aho = (reg_ac_data*)progi->data->data[ ARG( c ) ];
2123 reg_trie_data *trie = (reg_trie_data*)progi->data->data[ aho->trie ];
2124 HV *widecharmap = MUTABLE_HV(progi->data->data[ aho->trie + 1 ]);
2126 const char *last_start = strend - trie->minlen;
2128 const char *real_start = s;
2130 STRLEN maxlen = trie->maxlen;
2132 U8 **points; /* map of where we were in the input string
2133 when reading a given char. For ASCII this
2134 is unnecessary overhead as the relationship
2135 is always 1:1, but for Unicode, especially
2136 case folded Unicode this is not true. */
2137 U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ];
2141 GET_RE_DEBUG_FLAGS_DECL;
2143 /* We can't just allocate points here. We need to wrap it in
2144 * an SV so it gets freed properly if there is a croak while
2145 * running the match */
2148 sv_points=newSV(maxlen * sizeof(U8 *));
2149 SvCUR_set(sv_points,
2150 maxlen * sizeof(U8 *));
2151 SvPOK_on(sv_points);
2152 sv_2mortal(sv_points);
2153 points=(U8**)SvPV_nolen(sv_points );
2154 if ( trie_type != trie_utf8_fold
2155 && (trie->bitmap || OP(c)==AHOCORASICKC) )
2158 bitmap=(U8*)trie->bitmap;
2160 bitmap=(U8*)ANYOF_BITMAP(c);
2162 /* this is the Aho-Corasick algorithm modified a touch
2163 to include special handling for long "unknown char" sequences.
2164 The basic idea being that we use AC as long as we are dealing
2165 with a possible matching char, when we encounter an unknown char
2166 (and we have not encountered an accepting state) we scan forward
2167 until we find a legal starting char.
2168 AC matching is basically that of trie matching, except that when
2169 we encounter a failing transition, we fall back to the current
2170 states "fail state", and try the current char again, a process
2171 we repeat until we reach the root state, state 1, or a legal
2172 transition. If we fail on the root state then we can either
2173 terminate if we have reached an accepting state previously, or
2174 restart the entire process from the beginning if we have not.
2177 while (s <= last_start) {
2178 const U32 uniflags = UTF8_ALLOW_DEFAULT;
2186 U8 *uscan = (U8*)NULL;
2187 U8 *leftmost = NULL;
2189 U32 accepted_word= 0;
2193 while ( state && uc <= (U8*)strend ) {
2195 U32 word = aho->states[ state ].wordnum;
2199 DEBUG_TRIE_EXECUTE_r(
2200 if ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) {
2201 dump_exec_pos( (char *)uc, c, strend, real_start,
2202 (char *)uc, utf8_target );
2203 PerlIO_printf( Perl_debug_log,
2204 " Scanning for legal start char...\n");
2208 while ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) {
2212 while ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) {
2218 if (uc >(U8*)last_start) break;
2222 U8 *lpos= points[ (pointpos - trie->wordinfo[word].len) % maxlen ];
2223 if (!leftmost || lpos < leftmost) {
2224 DEBUG_r(accepted_word=word);
2230 points[pointpos++ % maxlen]= uc;
2231 if (foldlen || uc < (U8*)strend) {
2232 REXEC_TRIE_READ_CHAR(trie_type, trie,
2234 uscan, len, uvc, charid, foldlen,
2236 DEBUG_TRIE_EXECUTE_r({
2237 dump_exec_pos( (char *)uc, c, strend,
2238 real_start, s, utf8_target);
2239 PerlIO_printf(Perl_debug_log,
2240 " Charid:%3u CP:%4"UVxf" ",
2252 word = aho->states[ state ].wordnum;
2254 base = aho->states[ state ].trans.base;
2256 DEBUG_TRIE_EXECUTE_r({
2258 dump_exec_pos( (char *)uc, c, strend, real_start,
2260 PerlIO_printf( Perl_debug_log,
2261 "%sState: %4"UVxf", word=%"UVxf,
2262 failed ? " Fail transition to " : "",
2263 (UV)state, (UV)word);
2269 ( ((offset = base + charid
2270 - 1 - trie->uniquecharcount)) >= 0)
2271 && ((U32)offset < trie->lasttrans)
2272 && trie->trans[offset].check == state
2273 && (tmp=trie->trans[offset].next))
2275 DEBUG_TRIE_EXECUTE_r(
2276 PerlIO_printf( Perl_debug_log," - legal\n"));
2281 DEBUG_TRIE_EXECUTE_r(
2282 PerlIO_printf( Perl_debug_log," - fail\n"));
2284 state = aho->fail[state];
2288 /* we must be accepting here */
2289 DEBUG_TRIE_EXECUTE_r(
2290 PerlIO_printf( Perl_debug_log," - accepting\n"));
2299 if (!state) state = 1;
2302 if ( aho->states[ state ].wordnum ) {
2303 U8 *lpos = points[ (pointpos - trie->wordinfo[aho->states[ state ].wordnum].len) % maxlen ];
2304 if (!leftmost || lpos < leftmost) {
2305 DEBUG_r(accepted_word=aho->states[ state ].wordnum);
2310 s = (char*)leftmost;
2311 DEBUG_TRIE_EXECUTE_r({
2313 Perl_debug_log,"Matches word #%"UVxf" at position %"IVdf". Trying full pattern...\n",
2314 (UV)accepted_word, (IV)(s - real_start)
2317 if (reginfo->intuit || regtry(reginfo, &s)) {
2323 DEBUG_TRIE_EXECUTE_r({
2324 PerlIO_printf( Perl_debug_log,"Pattern failed. Looking for new start point...\n");
2327 DEBUG_TRIE_EXECUTE_r(
2328 PerlIO_printf( Perl_debug_log,"No match.\n"));
2337 Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c));
2344 /* set RX_SAVED_COPY, RX_SUBBEG etc.
2345 * flags have same meanings as with regexec_flags() */
2348 S_reg_set_capture_string(pTHX_ REGEXP * const rx,
2355 struct regexp *const prog = ReANY(rx);
2357 if (flags & REXEC_COPY_STR) {
2361 PerlIO_printf(Perl_debug_log,
2362 "Copy on write: regexp capture, type %d\n",
2365 /* Create a new COW SV to share the match string and store
2366 * in saved_copy, unless the current COW SV in saved_copy
2367 * is valid and suitable for our purpose */
2368 if (( prog->saved_copy
2369 && SvIsCOW(prog->saved_copy)
2370 && SvPOKp(prog->saved_copy)
2373 && SvPVX(sv) == SvPVX(prog->saved_copy)))
2375 /* just reuse saved_copy SV */
2376 if (RXp_MATCH_COPIED(prog)) {
2377 Safefree(prog->subbeg);
2378 RXp_MATCH_COPIED_off(prog);
2382 /* create new COW SV to share string */
2383 RX_MATCH_COPY_FREE(rx);
2384 prog->saved_copy = sv_setsv_cow(prog->saved_copy, sv);
2386 prog->subbeg = (char *)SvPVX_const(prog->saved_copy);
2387 assert (SvPOKp(prog->saved_copy));
2388 prog->sublen = strend - strbeg;
2389 prog->suboffset = 0;
2390 prog->subcoffset = 0;
2395 SSize_t max = strend - strbeg;
2398 if ( (flags & REXEC_COPY_SKIP_POST)
2399 && !(prog->extflags & RXf_PMf_KEEPCOPY) /* //p */
2400 && !(PL_sawampersand & SAWAMPERSAND_RIGHT)
2401 ) { /* don't copy $' part of string */
2404 /* calculate the right-most part of the string covered
2405 * by a capture. Due to look-ahead, this may be to
2406 * the right of $&, so we have to scan all captures */
2407 while (n <= prog->lastparen) {
2408 if (prog->offs[n].end > max)
2409 max = prog->offs[n].end;
2413 max = (PL_sawampersand & SAWAMPERSAND_LEFT)
2414 ? prog->offs[0].start
2416 assert(max >= 0 && max <= strend - strbeg);
2419 if ( (flags & REXEC_COPY_SKIP_PRE)
2420 && !(prog->extflags & RXf_PMf_KEEPCOPY) /* //p */
2421 && !(PL_sawampersand & SAWAMPERSAND_LEFT)
2422 ) { /* don't copy $` part of string */
2425 /* calculate the left-most part of the string covered
2426 * by a capture. Due to look-behind, this may be to
2427 * the left of $&, so we have to scan all captures */
2428 while (min && n <= prog->lastparen) {
2429 if ( prog->offs[n].start != -1
2430 && prog->offs[n].start < min)
2432 min = prog->offs[n].start;
2436 if ((PL_sawampersand & SAWAMPERSAND_RIGHT)
2437 && min > prog->offs[0].end
2439 min = prog->offs[0].end;
2443 assert(min >= 0 && min <= max && min <= strend - strbeg);
2446 if (RX_MATCH_COPIED(rx)) {
2447 if (sublen > prog->sublen)
2449 (char*)saferealloc(prog->subbeg, sublen+1);
2452 prog->subbeg = (char*)safemalloc(sublen+1);
2453 Copy(strbeg + min, prog->subbeg, sublen, char);
2454 prog->subbeg[sublen] = '\0';
2455 prog->suboffset = min;
2456 prog->sublen = sublen;
2457 RX_MATCH_COPIED_on(rx);
2459 prog->subcoffset = prog->suboffset;
2460 if (prog->suboffset && utf8_target) {
2461 /* Convert byte offset to chars.
2462 * XXX ideally should only compute this if @-/@+
2463 * has been seen, a la PL_sawampersand ??? */
2465 /* If there's a direct correspondence between the
2466 * string which we're matching and the original SV,
2467 * then we can use the utf8 len cache associated with
2468 * the SV. In particular, it means that under //g,
2469 * sv_pos_b2u() will use the previously cached
2470 * position to speed up working out the new length of
2471 * subcoffset, rather than counting from the start of
2472 * the string each time. This stops
2473 * $x = "\x{100}" x 1E6; 1 while $x =~ /(.)/g;
2474 * from going quadratic */
2475 if (SvPOKp(sv) && SvPVX(sv) == strbeg)
2476 prog->subcoffset = sv_pos_b2u_flags(sv, prog->subcoffset,
2477 SV_GMAGIC|SV_CONST_RETURN);
2479 prog->subcoffset = utf8_length((U8*)strbeg,
2480 (U8*)(strbeg+prog->suboffset));
2484 RX_MATCH_COPY_FREE(rx);
2485 prog->subbeg = strbeg;
2486 prog->suboffset = 0;
2487 prog->subcoffset = 0;
2488 prog->sublen = strend - strbeg;
2496 - regexec_flags - match a regexp against a string
2499 Perl_regexec_flags(pTHX_ REGEXP * const rx, char *stringarg, char *strend,
2500 char *strbeg, SSize_t minend, SV *sv, void *data, U32 flags)
2501 /* stringarg: the point in the string at which to begin matching */
2502 /* strend: pointer to null at end of string */
2503 /* strbeg: real beginning of string */
2504 /* minend: end of match must be >= minend bytes after stringarg. */
2505 /* sv: SV being matched: only used for utf8 flag, pos() etc; string
2506 * itself is accessed via the pointers above */
2507 /* data: May be used for some additional optimizations.
2508 Currently unused. */
2509 /* flags: For optimizations. See REXEC_* in regexp.h */
2512 struct regexp *const prog = ReANY(rx);
2516 SSize_t minlen; /* must match at least this many chars */
2517 SSize_t dontbother = 0; /* how many characters not to try at end */
2518 const bool utf8_target = cBOOL(DO_UTF8(sv));
2520 RXi_GET_DECL(prog,progi);
2521 regmatch_info reginfo_buf; /* create some info to pass to regtry etc */
2522 regmatch_info *const reginfo = ®info_buf;
2523 regexp_paren_pair *swap = NULL;
2525 GET_RE_DEBUG_FLAGS_DECL;
2527 PERL_ARGS_ASSERT_REGEXEC_FLAGS;
2528 PERL_UNUSED_ARG(data);
2530 /* Be paranoid... */
2531 if (prog == NULL || stringarg == NULL) {
2532 Perl_croak(aTHX_ "NULL regexp parameter");
2536 debug_start_match(rx, utf8_target, stringarg, strend,
2540 startpos = stringarg;
2542 if (prog->intflags & PREGf_GPOS_SEEN) {
2545 /* set reginfo->ganch, the position where \G can match */
2548 (flags & REXEC_IGNOREPOS)
2549 ? stringarg /* use start pos rather than pos() */
2550 : (sv && (mg = mg_find_mglob(sv)) && mg->mg_len >= 0)
2551 /* Defined pos(): */
2552 ? strbeg + MgBYTEPOS(mg, sv, strbeg, strend-strbeg)
2553 : strbeg; /* pos() not defined; use start of string */
2555 DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log,
2556 "GPOS ganch set to strbeg[%"IVdf"]\n", (IV)(reginfo->ganch - strbeg)));
2558 /* in the presence of \G, we may need to start looking earlier in
2559 * the string than the suggested start point of stringarg:
2560 * if prog->gofs is set, then that's a known, fixed minimum
2563 * /ab|c\G/: gofs = 1
2564 * or if the minimum offset isn't known, then we have to go back
2565 * to the start of the string, e.g. /w+\G/
2568 if (prog->intflags & PREGf_ANCH_GPOS) {
2569 startpos = reginfo->ganch - prog->gofs;
2571 ((flags & REXEC_FAIL_ON_UNDERFLOW) ? stringarg : strbeg))
2573 DEBUG_r(PerlIO_printf(Perl_debug_log,
2574 "fail: ganch-gofs before earliest possible start\n"));
2578 else if (prog->gofs) {
2579 if (startpos - prog->gofs < strbeg)
2582 startpos -= prog->gofs;
2584 else if (prog->intflags & PREGf_GPOS_FLOAT)
2588 minlen = prog->minlen;
2589 if ((startpos + minlen) > strend || startpos < strbeg) {
2590 DEBUG_r(PerlIO_printf(Perl_debug_log,
2591 "Regex match can't succeed, so not even tried\n"));
2595 /* at the end of this function, we'll do a LEAVE_SCOPE(oldsave),
2596 * which will call destuctors to reset PL_regmatch_state, free higher
2597 * PL_regmatch_slabs, and clean up regmatch_info_aux and
2598 * regmatch_info_aux_eval */
2600 oldsave = PL_savestack_ix;
2604 if ((prog->extflags & RXf_USE_INTUIT)
2605 && !(flags & REXEC_CHECKED))
2607 s = re_intuit_start(rx, sv, strbeg, startpos, strend,
2612 if (prog->extflags & RXf_CHECK_ALL) {
2613 /* we can match based purely on the result of INTUIT.
2614 * Set up captures etc just for $& and $-[0]
2615 * (an intuit-only match wont have $1,$2,..) */
2616 assert(!prog->nparens);
2618 /* s/// doesn't like it if $& is earlier than where we asked it to
2619 * start searching (which can happen on something like /.\G/) */
2620 if ( (flags & REXEC_FAIL_ON_UNDERFLOW)
2623 /* this should only be possible under \G */
2624 assert(prog->intflags & PREGf_GPOS_SEEN);
2625 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
2626 "matched, but failing for REXEC_FAIL_ON_UNDERFLOW\n"));
2630 /* match via INTUIT shouldn't have any captures.
2631 * Let @-, @+, $^N know */
2632 prog->lastparen = prog->lastcloseparen = 0;
2633 RX_MATCH_UTF8_set(rx, utf8_target);
2634 prog->offs[0].start = s - strbeg;
2635 prog->offs[0].end = utf8_target
2636 ? (char*)utf8_hop((U8*)s, prog->minlenret) - strbeg
2637 : s - strbeg + prog->minlenret;
2638 if ( !(flags & REXEC_NOT_FIRST) )
2639 S_reg_set_capture_string(aTHX_ rx,
2641 sv, flags, utf8_target);
2647 multiline = prog->extflags & RXf_PMf_MULTILINE;
2649 if (strend - s < (minlen+(prog->check_offset_min<0?prog->check_offset_min:0))) {
2650 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
2651 "String too short [regexec_flags]...\n"));
2655 /* Check validity of program. */
2656 if (UCHARAT(progi->program) != REG_MAGIC) {
2657 Perl_croak(aTHX_ "corrupted regexp program");
2660 RX_MATCH_TAINTED_off(rx);
2661 RX_MATCH_UTF8_set(rx, utf8_target);
2663 reginfo->prog = rx; /* Yes, sorry that this is confusing. */
2664 reginfo->intuit = 0;
2665 reginfo->is_utf8_target = cBOOL(utf8_target);
2666 reginfo->is_utf8_pat = cBOOL(RX_UTF8(rx));
2667 reginfo->warned = FALSE;
2668 reginfo->strbeg = strbeg;
2670 reginfo->poscache_maxiter = 0; /* not yet started a countdown */
2671 reginfo->strend = strend;
2672 /* see how far we have to get to not match where we matched before */
2673 reginfo->till = stringarg + minend;
2675 if (prog->extflags & RXf_EVAL_SEEN && SvPADTMP(sv)) {
2676 /* SAVEFREESV, not sv_mortalcopy, as this SV must last until after
2677 S_cleanup_regmatch_info_aux has executed (registered by
2678 SAVEDESTRUCTOR_X below). S_cleanup_regmatch_info_aux modifies
2679 magic belonging to this SV.
2680 Not newSVsv, either, as it does not COW.
2682 reginfo->sv = newSV(0);
2683 SvSetSV_nosteal(reginfo->sv, sv);
2684 SAVEFREESV(reginfo->sv);
2687 /* reserve next 2 or 3 slots in PL_regmatch_state:
2688 * slot N+0: may currently be in use: skip it
2689 * slot N+1: use for regmatch_info_aux struct
2690 * slot N+2: use for regmatch_info_aux_eval struct if we have (?{})'s
2691 * slot N+3: ready for use by regmatch()
2695 regmatch_state *old_regmatch_state;
2696 regmatch_slab *old_regmatch_slab;
2697 int i, max = (prog->extflags & RXf_EVAL_SEEN) ? 2 : 1;
2699 /* on first ever match, allocate first slab */
2700 if (!PL_regmatch_slab) {
2701 Newx(PL_regmatch_slab, 1, regmatch_slab);
2702 PL_regmatch_slab->prev = NULL;
2703 PL_regmatch_slab->next = NULL;
2704 PL_regmatch_state = SLAB_FIRST(PL_regmatch_slab);
2707 old_regmatch_state = PL_regmatch_state;
2708 old_regmatch_slab = PL_regmatch_slab;
2710 for (i=0; i <= max; i++) {
2712 reginfo->info_aux = &(PL_regmatch_state->u.info_aux);
2714 reginfo->info_aux_eval =
2715 reginfo->info_aux->info_aux_eval =
2716 &(PL_regmatch_state->u.info_aux_eval);
2718 if (++PL_regmatch_state > SLAB_LAST(PL_regmatch_slab))
2719 PL_regmatch_state = S_push_slab(aTHX);
2722 /* note initial PL_regmatch_state position; at end of match we'll
2723 * pop back to there and free any higher slabs */
2725 reginfo->info_aux->old_regmatch_state = old_regmatch_state;
2726 reginfo->info_aux->old_regmatch_slab = old_regmatch_slab;
2727 reginfo->info_aux->poscache = NULL;
2729 SAVEDESTRUCTOR_X(S_cleanup_regmatch_info_aux, reginfo->info_aux);
2731 if ((prog->extflags & RXf_EVAL_SEEN))
2732 S_setup_eval_state(aTHX_ reginfo);
2734 reginfo->info_aux_eval = reginfo->info_aux->info_aux_eval = NULL;
2737 /* If there is a "must appear" string, look for it. */
2739 if (PL_curpm && (PM_GETRE(PL_curpm) == rx)) {
2740 /* We have to be careful. If the previous successful match
2741 was from this regex we don't want a subsequent partially
2742 successful match to clobber the old results.
2743 So when we detect this possibility we add a swap buffer
2744 to the re, and switch the buffer each match. If we fail,
2745 we switch it back; otherwise we leave it swapped.
2748 /* do we need a save destructor here for eval dies? */
2749 Newxz(prog->offs, (prog->nparens + 1), regexp_paren_pair);
2750 DEBUG_BUFFERS_r(PerlIO_printf(Perl_debug_log,
2751 "rex=0x%"UVxf" saving offs: orig=0x%"UVxf" new=0x%"UVxf"\n",
2758 /* Simplest case: anchored match need be tried only once. */
2759 /* [unless only anchor is MBOL - implying multiline is set] */
2760 if (prog->intflags & (PREGf_ANCH & ~PREGf_ANCH_GPOS)) {
2761 if (s == startpos && regtry(reginfo, &s))
2763 else if (multiline || (prog->intflags & (PREGf_IMPLICIT | PREGf_ANCH_MBOL))) /* XXXX SBOL? */
2768 dontbother = minlen - 1;
2769 end = HOP3c(strend, -dontbother, strbeg) - 1;
2770 /* for multiline we only have to try after newlines */
2771 if (prog->check_substr || prog->check_utf8) {
2772 /* because of the goto we can not easily reuse the macros for bifurcating the
2773 unicode/non-unicode match modes here like we do elsewhere - demerphq */
2776 goto after_try_utf8;
2778 if (regtry(reginfo, &s)) {
2785 if (prog->extflags & RXf_USE_INTUIT) {
2786 s = re_intuit_start(rx, sv, strbeg,
2787 s + UTF8SKIP(s), strend, flags, NULL);
2796 } /* end search for check string in unicode */
2798 if (s == startpos) {
2799 goto after_try_latin;
2802 if (regtry(reginfo, &s)) {
2809 if (prog->extflags & RXf_USE_INTUIT) {
2810 s = re_intuit_start(rx, sv, strbeg,
2811 s + 1, strend, flags, NULL);
2820 } /* end search for check string in latin*/
2821 } /* end search for check string */
2822 else { /* search for newline */
2824 /*XXX: The s-- is almost definitely wrong here under unicode - demeprhq*/
2827 /* We can use a more efficient search as newlines are the same in unicode as they are in latin */
2828 while (s <= end) { /* note it could be possible to match at the end of the string */
2829 if (*s++ == '\n') { /* don't need PL_utf8skip here */
2830 if (regtry(reginfo, &s))
2834 } /* end search for newline */
2835 } /* end anchored/multiline check string search */
2837 } else if (prog->intflags & PREGf_ANCH_GPOS)
2839 /* PREGf_ANCH_GPOS should never be true if PREGf_GPOS_SEEN is not true */
2840 assert(prog->intflags & PREGf_GPOS_SEEN);
2841 /* For anchored \G, the only position it can match from is
2842 * (ganch-gofs); we already set startpos to this above; if intuit
2843 * moved us on from there, we can't possibly succeed */
2844 assert(startpos == reginfo->ganch - prog->gofs);
2845 if (s == startpos && regtry(reginfo, &s))
2850 /* Messy cases: unanchored match. */
2851 if ((prog->anchored_substr || prog->anchored_utf8) && prog->intflags & PREGf_SKIP) {
2852 /* we have /x+whatever/ */
2853 /* it must be a one character string (XXXX Except is_utf8_pat?) */
2859 if (! prog->anchored_utf8) {
2860 to_utf8_substr(prog);
2862 ch = SvPVX_const(prog->anchored_utf8)[0];
2865 DEBUG_EXECUTE_r( did_match = 1 );
2866 if (regtry(reginfo, &s)) goto got_it;
2868 while (s < strend && *s == ch)
2875 if (! prog->anchored_substr) {
2876 if (! to_byte_substr(prog)) {
2877 NON_UTF8_TARGET_BUT_UTF8_REQUIRED(phooey);
2880 ch = SvPVX_const(prog->anchored_substr)[0];
2883 DEBUG_EXECUTE_r( did_match = 1 );
2884 if (regtry(reginfo, &s)) goto got_it;
2886 while (s < strend && *s == ch)
2891 DEBUG_EXECUTE_r(if (!did_match)
2892 PerlIO_printf(Perl_debug_log,
2893 "Did not find anchored character...\n")
2896 else if (prog->anchored_substr != NULL
2897 || prog->anchored_utf8 != NULL
2898 || ((prog->float_substr != NULL || prog->float_utf8 != NULL)
2899 && prog->float_max_offset < strend - s)) {
2904 char *last1; /* Last position checked before */
2908 if (prog->anchored_substr || prog->anchored_utf8) {
2910 if (! prog->anchored_utf8) {
2911 to_utf8_substr(prog);
2913 must = prog->anchored_utf8;
2916 if (! prog->anchored_substr) {
2917 if (! to_byte_substr(prog)) {
2918 NON_UTF8_TARGET_BUT_UTF8_REQUIRED(phooey);
2921 must = prog->anchored_substr;
2923 back_max = back_min = prog->anchored_offset;
2926 if (! prog->float_utf8) {
2927 to_utf8_substr(prog);
2929 must = prog->float_utf8;
2932 if (! prog->float_substr) {
2933 if (! to_byte_substr(prog)) {
2934 NON_UTF8_TARGET_BUT_UTF8_REQUIRED(phooey);
2937 must = prog->float_substr;
2939 back_max = prog->float_max_offset;
2940 back_min = prog->float_min_offset;
2946 last = HOP3c(strend, /* Cannot start after this */
2947 -(SSize_t)(CHR_SVLEN(must)
2948 - (SvTAIL(must) != 0) + back_min), strbeg);
2950 if (s > reginfo->strbeg)
2951 last1 = HOPc(s, -1);
2953 last1 = s - 1; /* bogus */
2955 /* XXXX check_substr already used to find "s", can optimize if
2956 check_substr==must. */
2958 strend = HOPc(strend, -dontbother);
2959 while ( (s <= last) &&
2960 (s = fbm_instr((unsigned char*)HOP4c(s, back_min, strbeg, strend),
2961 (unsigned char*)strend, must,
2962 multiline ? FBMrf_MULTILINE : 0)) ) {
2963 DEBUG_EXECUTE_r( did_match = 1 );
2964 if (HOPc(s, -back_max) > last1) {
2965 last1 = HOPc(s, -back_min);
2966 s = HOPc(s, -back_max);
2969 char * const t = (last1 >= reginfo->strbeg)
2970 ? HOPc(last1, 1) : last1 + 1;
2972 last1 = HOPc(s, -back_min);
2976 while (s <= last1) {
2977 if (regtry(reginfo, &s))
2980 s++; /* to break out of outer loop */
2987 while (s <= last1) {
2988 if (regtry(reginfo, &s))
2994 DEBUG_EXECUTE_r(if (!did_match) {
2995 RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0),
2996 SvPVX_const(must), RE_SV_DUMPLEN(must), 30);
2997 PerlIO_printf(Perl_debug_log, "Did not find %s substr %s%s...\n",
2998 ((must == prog->anchored_substr || must == prog->anchored_utf8)
2999 ? "anchored" : "floating"),
3000 quoted, RE_SV_TAIL(must));
3004 else if ( (c = progi->regstclass) ) {
3006 const OPCODE op = OP(progi->regstclass);
3007 /* don't bother with what can't match */
3008 if (PL_regkind[op] != EXACT && op != CANY && PL_regkind[op] != TRIE)
3009 strend = HOPc(strend, -(minlen - 1));
3012 SV * const prop = sv_newmortal();
3013 regprop(prog, prop, c, reginfo, NULL);
3015 RE_PV_QUOTED_DECL(quoted,utf8_target,PERL_DEBUG_PAD_ZERO(1),
3017 PerlIO_printf(Perl_debug_log,
3018 "Matching stclass %.*s against %s (%d bytes)\n",
3019 (int)SvCUR(prop), SvPVX_const(prop),
3020 quoted, (int)(strend - s));
3023 if (find_byclass(prog, c, s, strend, reginfo))
3025 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass... [regexec_flags]\n"));
3029 if (prog->float_substr != NULL || prog->float_utf8 != NULL) {
3037 if (! prog->float_utf8) {
3038 to_utf8_substr(prog);
3040 float_real = prog->float_utf8;
3043 if (! prog->float_substr) {
3044 if (! to_byte_substr(prog)) {
3045 NON_UTF8_TARGET_BUT_UTF8_REQUIRED(phooey);
3048 float_real = prog->float_substr;
3051 little = SvPV_const(float_real, len);
3052 if (SvTAIL(float_real)) {
3053 /* This means that float_real contains an artificial \n on
3054 * the end due to the presence of something like this:
3055 * /foo$/ where we can match both "foo" and "foo\n" at the
3056 * end of the string. So we have to compare the end of the
3057 * string first against the float_real without the \n and
3058 * then against the full float_real with the string. We
3059 * have to watch out for cases where the string might be
3060 * smaller than the float_real or the float_real without
3062 char *checkpos= strend - len;
3064 PerlIO_printf(Perl_debug_log,
3065 "%sChecking for float_real.%s\n",
3066 PL_colors[4], PL_colors[5]));
3067 if (checkpos + 1 < strbeg) {
3068 /* can't match, even if we remove the trailing \n
3069 * string is too short to match */
3071 PerlIO_printf(Perl_debug_log,
3072 "%sString shorter than required trailing substring, cannot match.%s\n",
3073 PL_colors[4], PL_colors[5]));
3075 } else if (memEQ(checkpos + 1, little, len - 1)) {
3076 /* can match, the end of the string matches without the
3078 last = checkpos + 1;
3079 } else if (checkpos < strbeg) {
3080 /* cant match, string is too short when the "\n" is
3083 PerlIO_printf(Perl_debug_log,
3084 "%sString does not contain required trailing substring, cannot match.%s\n",
3085 PL_colors[4], PL_colors[5]));
3087 } else if (!multiline) {
3088 /* non multiline match, so compare with the "\n" at the
3089 * end of the string */
3090 if (memEQ(checkpos, little, len)) {
3094 PerlIO_printf(Perl_debug_log,
3095 "%sString does not contain required trailing substring, cannot match.%s\n",
3096 PL_colors[4], PL_colors[5]));
3100 /* multiline match, so we have to search for a place
3101 * where the full string is located */
3107 last = rninstr(s, strend, little, little + len);
3109 last = strend; /* matching "$" */
3112 /* at one point this block contained a comment which was
3113 * probably incorrect, which said that this was a "should not
3114 * happen" case. Even if it was true when it was written I am
3115 * pretty sure it is not anymore, so I have removed the comment
3116 * and replaced it with this one. Yves */
3118 PerlIO_printf(Perl_debug_log,
3119 "%sString does not contain required substring, cannot match.%s\n",
3120 PL_colors[4], PL_colors[5]
3124 dontbother = strend - last + prog->float_min_offset;
3126 if (minlen && (dontbother < minlen))
3127 dontbother = minlen - 1;
3128 strend -= dontbother; /* this one's always in bytes! */
3129 /* We don't know much -- general case. */
3132 if (regtry(reginfo, &s))
3141 if (regtry(reginfo, &s))
3143 } while (s++ < strend);
3151 /* s/// doesn't like it if $& is earlier than where we asked it to
3152 * start searching (which can happen on something like /.\G/) */
3153 if ( (flags & REXEC_FAIL_ON_UNDERFLOW)
3154 && (prog->offs[0].start < stringarg - strbeg))
3156 /* this should only be possible under \G */
3157 assert(prog->intflags & PREGf_GPOS_SEEN);
3158 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
3159 "matched, but failing for REXEC_FAIL_ON_UNDERFLOW\n"));
3165 PerlIO_printf(Perl_debug_log,
3166 "rex=0x%"UVxf" freeing offs: 0x%"UVxf"\n",
3173 /* clean up; this will trigger destructors that will free all slabs
3174 * above the current one, and cleanup the regmatch_info_aux
3175 * and regmatch_info_aux_eval sructs */
3177 LEAVE_SCOPE(oldsave);
3179 if (RXp_PAREN_NAMES(prog))
3180 (void)hv_iterinit(RXp_PAREN_NAMES(prog));
3182 /* make sure $`, $&, $', and $digit will work later */
3183 if ( !(flags & REXEC_NOT_FIRST) )
3184 S_reg_set_capture_string(aTHX_ rx,
3185 strbeg, reginfo->strend,
3186 sv, flags, utf8_target);
3191 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n",
3192 PL_colors[4], PL_colors[5]));
3194 /* clean up; this will trigger destructors that will free all slabs
3195 * above the current one, and cleanup the regmatch_info_aux
3196 * and regmatch_info_aux_eval sructs */
3198 LEAVE_SCOPE(oldsave);
3201 /* we failed :-( roll it back */
3202 DEBUG_BUFFERS_r(PerlIO_printf(Perl_debug_log,
3203 "rex=0x%"UVxf" rolling back offs: freeing=0x%"UVxf" restoring=0x%"UVxf"\n",
3208 Safefree(prog->offs);
3215 /* Set which rex is pointed to by PL_reg_curpm, handling ref counting.
3216 * Do inc before dec, in case old and new rex are the same */
3217 #define SET_reg_curpm(Re2) \
3218 if (reginfo->info_aux_eval) { \
3219 (void)ReREFCNT_inc(Re2); \
3220 ReREFCNT_dec(PM_GETRE(PL_reg_curpm)); \
3221 PM_SETRE((PL_reg_curpm), (Re2)); \
3226 - regtry - try match at specific point
3228 STATIC I32 /* 0 failure, 1 success */
3229 S_regtry(pTHX_ regmatch_info *reginfo, char **startposp)
3232 REGEXP *const rx = reginfo->prog;
3233 regexp *const prog = ReANY(rx);
3235 RXi_GET_DECL(prog,progi);
3236 GET_RE_DEBUG_FLAGS_DECL;
3238 PERL_ARGS_ASSERT_REGTRY;
3240 reginfo->cutpoint=NULL;
3242 prog->offs[0].start = *startposp - reginfo->strbeg;
3243 prog->lastparen = 0;
3244 prog->lastcloseparen = 0;
3246 /* XXXX What this code is doing here?!!! There should be no need
3247 to do this again and again, prog->lastparen should take care of
3250 /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code.
3251 * Actually, the code in regcppop() (which Ilya may be meaning by
3252 * prog->lastparen), is not needed at all by the test suite
3253 * (op/regexp, op/pat, op/split), but that code is needed otherwise
3254 * this erroneously leaves $1 defined: "1" =~ /^(?:(\d)x)?\d$/
3255 * Meanwhile, this code *is* needed for the
3256 * above-mentioned test suite tests to succeed. The common theme
3257 * on those tests seems to be returning null fields from matches.
3258 * --jhi updated by dapm */
3260 if (prog->nparens) {
3261 regexp_paren_pair *pp = prog->offs;
3263 for (i = prog->nparens; i > (I32)prog->lastparen; i--) {
3271 result = regmatch(reginfo, *startposp, progi->program + 1);
3273 prog->offs[0].end = result;
3276 if (reginfo->cutpoint)
3277 *startposp= reginfo->cutpoint;
3278 REGCP_UNWIND(lastcp);
3283 #define sayYES goto yes
3284 #define sayNO goto no
3285 #define sayNO_SILENT goto no_silent
3287 /* we dont use STMT_START/END here because it leads to
3288 "unreachable code" warnings, which are bogus, but distracting. */
3289 #define CACHEsayNO \
3290 if (ST.cache_mask) \
3291 reginfo->info_aux->poscache[ST.cache_offset] |= ST.cache_mask; \
3294 /* this is used to determine how far from the left messages like
3295 'failed...' are printed. It should be set such that messages
3296 are inline with the regop output that created them.
3298 #define REPORT_CODE_OFF 32
3301 #define CHRTEST_UNINIT -1001 /* c1/c2 haven't been calculated yet */
3302 #define CHRTEST_VOID -1000 /* the c1/c2 "next char" test should be skipped */
3303 #define CHRTEST_NOT_A_CP_1 -999
3304 #define CHRTEST_NOT_A_CP_2 -998
3306 /* grab a new slab and return the first slot in it */
3308 STATIC regmatch_state *
3311 #if PERL_VERSION < 9 && !defined(PERL_CORE)
3314 regmatch_slab *s = PL_regmatch_slab->next;
3316 Newx(s, 1, regmatch_slab);
3317 s->prev = PL_regmatch_slab;
3319 PL_regmatch_slab->next = s;
3321 PL_regmatch_slab = s;
3322 return SLAB_FIRST(s);
3326 /* push a new state then goto it */
3328 #define PUSH_STATE_GOTO(state, node, input) \
3329 pushinput = input; \
3331 st->resume_state = state; \
3334 /* push a new state with success backtracking, then goto it */
3336 #define PUSH_YES_STATE_GOTO(state, node, input) \
3337 pushinput = input; \
3339 st->resume_state = state; \
3340 goto push_yes_state;
3347 regmatch() - main matching routine
3349 This is basically one big switch statement in a loop. We execute an op,
3350 set 'next' to point the next op, and continue. If we come to a point which
3351 we may need to backtrack to on failure such as (A|B|C), we push a
3352 backtrack state onto the backtrack stack. On failure, we pop the top
3353 state, and re-enter the loop at the state indicated. If there are no more
3354 states to pop, we return failure.
3356 Sometimes we also need to backtrack on success; for example /A+/, where
3357 after successfully matching one A, we need to go back and try to
3358 match another one; similarly for lookahead assertions: if the assertion
3359 completes successfully, we backtrack to the state just before the assertion
3360 and then carry on. In these cases, the pushed state is marked as
3361 'backtrack on success too'. This marking is in fact done by a chain of
3362 pointers, each pointing to the previous 'yes' state. On success, we pop to
3363 the nearest yes state, discarding any intermediate failure-only states.
3364 Sometimes a yes state is pushed just to force some cleanup code to be
3365 called at the end of a successful match or submatch; e.g. (??{$re}) uses
3366 it to free the inner regex.
3368 Note that failure backtracking rewinds the cursor position, while
3369 success backtracking leaves it alone.
3371 A pattern is complete when the END op is executed, while a subpattern
3372 such as (?=foo) is complete when the SUCCESS op is executed. Both of these
3373 ops trigger the "pop to last yes state if any, otherwise return true"
3376 A common convention in this function is to use A and B to refer to the two
3377 subpatterns (or to the first nodes thereof) in patterns like /A*B/: so A is
3378 the subpattern to be matched possibly multiple times, while B is the entire
3379 rest of the pattern. Variable and state names reflect this convention.
3381 The states in the main switch are the union of ops and failure/success of
3382 substates associated with with that op. For example, IFMATCH is the op
3383 that does lookahead assertions /(?=A)B/ and so the IFMATCH state means
3384 'execute IFMATCH'; while IFMATCH_A is a state saying that we have just
3385 successfully matched A and IFMATCH_A_fail is a state saying that we have
3386 just failed to match A. Resume states always come in pairs. The backtrack
3387 state we push is marked as 'IFMATCH_A', but when that is popped, we resume
3388 at IFMATCH_A or IFMATCH_A_fail, depending on whether we are backtracking
3389 on success or failure.
3391 The struct that holds a backtracking state is actually a big union, with
3392 one variant for each major type of op. The variable st points to the
3393 top-most backtrack struct. To make the code clearer, within each
3394 block of code we #define ST to alias the relevant union.
3396 Here's a concrete example of a (vastly oversimplified) IFMATCH
3402 #define ST st->u.ifmatch
3404 case IFMATCH: // we are executing the IFMATCH op, (?=A)B
3405 ST.foo = ...; // some state we wish to save
3407 // push a yes backtrack state with a resume value of
3408 // IFMATCH_A/IFMATCH_A_fail, then continue execution at the
3410 PUSH_YES_STATE_GOTO(IFMATCH_A, A, newinput);
3413 case IFMATCH_A: // we have successfully executed A; now continue with B
3415 bar = ST.foo; // do something with the preserved value
3418 case IFMATCH_A_fail: // A failed, so the assertion failed
3419 ...; // do some housekeeping, then ...
3420 sayNO; // propagate the failure
3427 For any old-timers reading this who are familiar with the old recursive
3428 approach, the code above is equivalent to:
3430 case IFMATCH: // we are executing the IFMATCH op, (?=A)B
3439 ...; // do some housekeeping, then ...
3440 sayNO; // propagate the failure
3443 The topmost backtrack state, pointed to by st, is usually free. If you
3444 want to claim it, populate any ST.foo fields in it with values you wish to
3445 save, then do one of
3447 PUSH_STATE_GOTO(resume_state, node, newinput);
3448 PUSH_YES_STATE_GOTO(resume_state, node, newinput);
3450 which sets that backtrack state's resume value to 'resume_state', pushes a
3451 new free entry to the top of the backtrack stack, then goes to 'node'.
3452 On backtracking, the free slot is popped, and the saved state becomes the
3453 new free state. An ST.foo field in this new top state can be temporarily
3454 accessed to retrieve values, but once the main loop is re-entered, it
3455 becomes available for reuse.
3457 Note that the depth of the backtrack stack constantly increases during the
3458 left-to-right execution of the pattern, rather than going up and down with
3459 the pattern nesting. For example the stack is at its maximum at Z at the
3460 end of the pattern, rather than at X in the following:
3462 /(((X)+)+)+....(Y)+....Z/
3464 The only exceptions to this are lookahead/behind assertions and the cut,
3465 (?>A), which pop all the backtrack states associated with A before
3468 Backtrack state structs are allocated in slabs of about 4K in size.
3469 PL_regmatch_state and st always point to the currently active state,
3470 and PL_regmatch_slab points to the slab currently containing
3471 PL_regmatch_state. The first time regmatch() is called, the first slab is
3472 allocated, and is never freed until interpreter destruction. When the slab
3473 is full, a new one is allocated and chained to the end. At exit from
3474 regmatch(), slabs allocated since entry are freed.
3479 #define DEBUG_STATE_pp(pp) \
3481 DUMP_EXEC_POS(locinput, scan, utf8_target); \
3482 PerlIO_printf(Perl_debug_log, \
3483 " %*s"pp" %s%s%s%s%s\n", \
3485 PL_reg_name[st->resume_state], \
3486 ((st==yes_state||st==mark_state) ? "[" : ""), \
3487 ((st==yes_state) ? "Y" : ""), \
3488 ((st==mark_state) ? "M" : ""), \
3489 ((st==yes_state||st==mark_state) ? "]" : "") \
3494 #define REG_NODE_NUM(x) ((x) ? (int)((x)-prog) : -1)
3499 S_debug_start_match(pTHX_ const REGEXP *prog, const bool utf8_target,
3500 const char *start, const char *end, const char *blurb)
3502 const bool utf8_pat = RX_UTF8(prog) ? 1 : 0;
3504 PERL_ARGS_ASSERT_DEBUG_START_MATCH;
3509 RE_PV_QUOTED_DECL(s0, utf8_pat, PERL_DEBUG_PAD_ZERO(0),
3510 RX_PRECOMP_const(prog), RX_PRELEN(prog), 60);
3512 RE_PV_QUOTED_DECL(s1, utf8_target, PERL_DEBUG_PAD_ZERO(1),
3513 start, end - start, 60);
3515 PerlIO_printf(Perl_debug_log,
3516 "%s%s REx%s %s against %s\n",
3517 PL_colors[4], blurb, PL_colors[5], s0, s1);
3519 if (utf8_target||utf8_pat)
3520 PerlIO_printf(Perl_debug_log, "UTF-8 %s%s%s...\n",
3521 utf8_pat ? "pattern" : "",
3522 utf8_pat && utf8_target ? " and " : "",
3523 utf8_target ? "string" : ""
3529 S_dump_exec_pos(pTHX_ const char *locinput,
3530 const regnode *scan,
3531 const char *loc_regeol,
3532 const char *loc_bostr,
3533 const char *loc_reg_starttry,
3534 const bool utf8_target)
3536 const int docolor = *PL_colors[0] || *PL_colors[2] || *PL_colors[4];
3537 const int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */
3538 int l = (loc_regeol - locinput) > taill ? taill : (loc_regeol - locinput);
3539 /* The part of the string before starttry has one color
3540 (pref0_len chars), between starttry and current
3541 position another one (pref_len - pref0_len chars),
3542 after the current position the third one.
3543 We assume that pref0_len <= pref_len, otherwise we
3544 decrease pref0_len. */
3545 int pref_len = (locinput - loc_bostr) > (5 + taill) - l
3546 ? (5 + taill) - l : locinput - loc_bostr;
3549 PERL_ARGS_ASSERT_DUMP_EXEC_POS;
3551 while (utf8_target && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len)))
3553 pref0_len = pref_len - (locinput - loc_reg_starttry);
3554 if (l + pref_len < (5 + taill) && l < loc_regeol - locinput)
3555 l = ( loc_regeol - locinput > (5 + taill) - pref_len
3556 ? (5 + taill) - pref_len : loc_regeol - locinput);
3557 while (utf8_target && UTF8_IS_CONTINUATION(*(U8*)(locinput + l)))
3561 if (pref0_len > pref_len)
3562 pref0_len = pref_len;
3564 const int is_uni = (utf8_target && OP(scan) != CANY) ? 1 : 0;
3566 RE_PV_COLOR_DECL(s0,len0,is_uni,PERL_DEBUG_PAD(0),
3567 (locinput - pref_len),pref0_len, 60, 4, 5);
3569 RE_PV_COLOR_DECL(s1,len1,is_uni,PERL_DEBUG_PAD(1),
3570 (locinput - pref_len + pref0_len),
3571 pref_len - pref0_len, 60, 2, 3);
3573 RE_PV_COLOR_DECL(s2,len2,is_uni,PERL_DEBUG_PAD(2),
3574 locinput, loc_regeol - locinput, 10, 0, 1);
3576 const STRLEN tlen=len0+len1+len2;
3577 PerlIO_printf(Perl_debug_log,
3578 "%4"IVdf" <%.*s%.*s%s%.*s>%*s|",
3579 (IV)(locinput - loc_bostr),
3582 (docolor ? "" : "> <"),
3584 (int)(tlen > 19 ? 0 : 19 - tlen),
3591 /* reg_check_named_buff_matched()
3592 * Checks to see if a named buffer has matched. The data array of
3593 * buffer numbers corresponding to the buffer is expected to reside
3594 * in the regexp->data->data array in the slot stored in the ARG() of
3595 * node involved. Note that this routine doesn't actually care about the
3596 * name, that information is not preserved from compilation to execution.
3597 * Returns the index of the leftmost defined buffer with the given name
3598 * or 0 if non of the buffers matched.
3601 S_reg_check_named_buff_matched(const regexp *rex, const regnode *scan)
3604 RXi_GET_DECL(rex,rexi);
3605 SV *sv_dat= MUTABLE_SV(rexi->data->data[ ARG( scan ) ]);
3606 I32 *nums=(I32*)SvPVX(sv_dat);
3608 PERL_ARGS_ASSERT_REG_CHECK_NAMED_BUFF_MATCHED;
3610 for ( n=0; n<SvIVX(sv_dat); n++ ) {
3611 if ((I32)rex->lastparen >= nums[n] &&
3612 rex->offs[nums[n]].end != -1)
3622 S_setup_EXACTISH_ST_c1_c2(pTHX_ const regnode * const text_node, int *c1p,
3623 U8* c1_utf8, int *c2p, U8* c2_utf8, regmatch_info *reginfo)
3625 /* This function determines if there are one or two characters that match
3626 * the first character of the passed-in EXACTish node <text_node>, and if
3627 * so, returns them in the passed-in pointers.
3629 * If it determines that no possible character in the target string can
3630 * match, it returns FALSE; otherwise TRUE. (The FALSE situation occurs if
3631 * the first character in <text_node> requires UTF-8 to represent, and the
3632 * target string isn't in UTF-8.)
3634 * If there are more than two characters that could match the beginning of
3635 * <text_node>, or if more context is required to determine a match or not,
3636 * it sets both *<c1p> and *<c2p> to CHRTEST_VOID.
3638 * The motiviation behind this function is to allow the caller to set up
3639 * tight loops for matching. If <text_node> is of type EXACT, there is
3640 * only one possible character that can match its first character, and so
3641 * the situation is quite simple. But things get much more complicated if
3642 * folding is involved. It may be that the first character of an EXACTFish
3643 * node doesn't participate in any possible fold, e.g., punctuation, so it
3644 * can be matched only by itself. The vast majority of characters that are
3645 * in folds match just two things, their lower and upper-case equivalents.
3646 * But not all are like that; some have multiple possible matches, or match
3647 * sequences of more than one character. This function sorts all that out.
3649 * Consider the patterns A*B or A*?B where A and B are arbitrary. In a
3650 * loop of trying to match A*, we know we can't exit where the thing
3651 * following it isn't a B. And something can't be a B unless it is the
3652 * beginning of B. By putting a quick test for that beginning in a tight
3653 * loop, we can rule out things that can't possibly be B without having to
3654 * break out of the loop, thus avoiding work. Similarly, if A is a single
3655 * character, we can make a tight loop matching A*, using the outputs of
3658 * If the target string to match isn't in UTF-8, and there aren't
3659 * complications which require CHRTEST_VOID, *<c1p> and *<c2p> are set to
3660 * the one or two possible octets (which are characters in this situation)
3661 * that can match. In all cases, if there is only one character that can
3662 * match, *<c1p> and *<c2p> will be identical.
3664 * If the target string is in UTF-8, the buffers pointed to by <c1_utf8>
3665 * and <c2_utf8> will contain the one or two UTF-8 sequences of bytes that
3666 * can match the beginning of <text_node>. They should be declared with at
3667 * least length UTF8_MAXBYTES+1. (If the target string isn't in UTF-8, it is
3668 * undefined what these contain.) If one or both of the buffers are
3669 * invariant under UTF-8, *<c1p>, and *<c2p> will also be set to the
3670 * corresponding invariant. If variant, the corresponding *<c1p> and/or
3671 * *<c2p> will be set to a negative number(s) that shouldn't match any code
3672 * point (unless inappropriately coerced to unsigned). *<c1p> will equal
3673 * *<c2p> if and only if <c1_utf8> and <c2_utf8> are the same. */
3675 const bool utf8_target = reginfo->is_utf8_target;
3677 UV c1 = (UV)CHRTEST_NOT_A_CP_1;
3678 UV c2 = (UV)CHRTEST_NOT_A_CP_2;
3679 bool use_chrtest_void = FALSE;
3680 const bool is_utf8_pat = reginfo->is_utf8_pat;
3682 /* Used when we have both utf8 input and utf8 output, to avoid converting
3683 * to/from code points */
3684 bool utf8_has_been_setup = FALSE;
3688 U8 *pat = (U8*)STRING(text_node);
3689 U8 folded[UTF8_MAX_FOLD_CHAR_EXPAND * UTF8_MAXBYTES_CASE + 1] = { '\0' };
3691 if (OP(text_node) == EXACT || OP(text_node) == EXACTL) {
3693 /* In an exact node, only one thing can be matched, that first
3694 * character. If both the pat and the target are UTF-8, we can just
3695 * copy the input to the output, avoiding finding the code point of
3700 else if (utf8_target) {
3701 Copy(pat, c1_utf8, UTF8SKIP(pat), U8);
3702 Copy(pat, c2_utf8, UTF8SKIP(pat), U8);
3703 utf8_has_been_setup = TRUE;
3706 c2 = c1 = valid_utf8_to_uvchr(pat, NULL);
3709 else { /* an EXACTFish node */
3710 U8 *pat_end = pat + STR_LEN(text_node);
3712 /* An EXACTFL node has at least some characters unfolded, because what
3713 * they match is not known until now. So, now is the time to fold
3714 * the first few of them, as many as are needed to determine 'c1' and
3715 * 'c2' later in the routine. If the pattern isn't UTF-8, we only need
3716 * to fold if in a UTF-8 locale, and then only the Sharp S; everything
3717 * else is 1-1 and isn't assumed to be folded. In a UTF-8 pattern, we
3718 * need to fold as many characters as a single character can fold to,
3719 * so that later we can check if the first ones are such a multi-char
3720 * fold. But, in such a pattern only locale-problematic characters
3721 * aren't folded, so we can skip this completely if the first character
3722 * in the node isn't one of the tricky ones */
3723 if (OP(text_node) == EXACTFL) {
3725 if (! is_utf8_pat) {
3726 if (IN_UTF8_CTYPE_LOCALE && *pat == LATIN_SMALL_LETTER_SHARP_S)
3728 folded[0] = folded[1] = 's';
3730 pat_end = folded + 2;
3733 else if (is_PROBLEMATIC_LOCALE_FOLDEDS_START_utf8(pat)) {
3738 for (i = 0; i < UTF8_MAX_FOLD_CHAR_EXPAND && s < pat_end; i++) {
3740 *(d++) = (U8) toFOLD_LC(*s);
3745 _to_utf8_fold_flags(s,
3748 FOLD_FLAGS_FULL | FOLD_FLAGS_LOCALE);
3759 if ((is_utf8_pat && is_MULTI_CHAR_FOLD_utf8_safe(pat, pat_end))
3760 || (!is_utf8_pat && is_MULTI_CHAR_FOLD_latin1_safe(pat, pat_end)))
3762 /* Multi-character folds require more context to sort out. Also
3763 * PL_utf8_foldclosures used below doesn't handle them, so have to
3764 * be handled outside this routine */
3765 use_chrtest_void = TRUE;
3767 else { /* an EXACTFish node which doesn't begin with a multi-char fold */
3768 c1 = is_utf8_pat ? valid_utf8_to_uvchr(pat, NULL) : *pat;
3770 /* Load the folds hash, if not already done */
3772 if (! PL_utf8_foldclosures) {
3773 _load_PL_utf8_foldclosures();
3776 /* The fold closures data structure is a hash with the keys
3777 * being the UTF-8 of every character that is folded to, like
3778 * 'k', and the values each an array of all code points that
3779 * fold to its key. e.g. [ 'k', 'K', KELVIN_SIGN ].
3780 * Multi-character folds are not included */
3781 if ((! (listp = hv_fetch(PL_utf8_foldclosures,
3786 /* Not found in the hash, therefore there are no folds
3787 * containing it, so there is only a single character that
3791 else { /* Does participate in folds */
3792 AV* list = (AV*) *listp;
3793 if (av_tindex(list) != 1) {
3795 /* If there aren't exactly two folds to this, it is
3796 * outside the scope of this function */
3797 use_chrtest_void = TRUE;
3799 else { /* There are two. Get them */
3800 SV** c_p = av_fetch(list, 0, FALSE);
3802 Perl_croak(aTHX_ "panic: invalid PL_utf8_foldclosures structure");
3806 c_p = av_fetch(list, 1, FALSE);
3808 Perl_croak(aTHX_ "panic: invalid PL_utf8_foldclosures structure");
3812 /* Folds that cross the 255/256 boundary are forbidden
3813 * if EXACTFL (and isnt a UTF8 locale), or EXACTFA and
3814 * one is ASCIII. Since the pattern character is above
3815 * 255, and its only other match is below 256, the only
3816 * legal match will be to itself. We have thrown away
3817 * the original, so have to compute which is the one
3819 if ((c1 < 256) != (c2 < 256)) {
3820 if ((OP(text_node) == EXACTFL
3821 && ! IN_UTF8_CTYPE_LOCALE)
3822 || ((OP(text_node) == EXACTFA
3823 || OP(text_node) == EXACTFA_NO_TRIE)
3824 && (isASCII(c1) || isASCII(c2))))
3837 else /* Here, c1 is <= 255 */
3839 && HAS_NONLATIN1_FOLD_CLOSURE(c1)
3840 && ( ! (OP(text_node) == EXACTFL && ! IN_UTF8_CTYPE_LOCALE))
3841 && ((OP(text_node) != EXACTFA
3842 && OP(text_node) != EXACTFA_NO_TRIE)
3845 /* Here, there could be something above Latin1 in the target
3846 * which folds to this character in the pattern. All such
3847 * cases except LATIN SMALL LETTER Y WITH DIAERESIS have more
3848 * than two characters involved in their folds, so are outside
3849 * the scope of this function */
3850 if (UNLIKELY(c1 == LATIN_SMALL_LETTER_Y_WITH_DIAERESIS)) {
3851 c2 = LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS;
3854 use_chrtest_void = TRUE;
3857 else { /* Here nothing above Latin1 can fold to the pattern
3859 switch (OP(text_node)) {
3861 case EXACTFL: /* /l rules */
3862 c2 = PL_fold_locale[c1];
3865 case EXACTF: /* This node only generated for non-utf8
3867 assert(! is_utf8_pat);
3868 if (! utf8_target) { /* /d rules */
3873 /* /u rules for all these. This happens to work for
3874 * EXACTFA as nothing in Latin1 folds to ASCII */
3875 case EXACTFA_NO_TRIE: /* This node only generated for
3876 non-utf8 patterns */
3877 assert(! is_utf8_pat);
3882 c2 = PL_fold_latin1[c1];
3886 Perl_croak(aTHX_ "panic: Unexpected op %u", OP(text_node));
3887 NOT_REACHED; /* NOTREACHED */
3893 /* Here have figured things out. Set up the returns */
3894 if (use_chrtest_void) {
3895 *c2p = *c1p = CHRTEST_VOID;
3897 else if (utf8_target) {
3898 if (! utf8_has_been_setup) { /* Don't have the utf8; must get it */
3899 uvchr_to_utf8(c1_utf8, c1);
3900 uvchr_to_utf8(c2_utf8, c2);
3903 /* Invariants are stored in both the utf8 and byte outputs; Use
3904 * negative numbers otherwise for the byte ones. Make sure that the
3905 * byte ones are the same iff the utf8 ones are the same */
3906 *c1p = (UTF8_IS_INVARIANT(*c1_utf8)) ? *c1_utf8 : CHRTEST_NOT_A_CP_1;
3907 *c2p = (UTF8_IS_INVARIANT(*c2_utf8))
3910 ? CHRTEST_NOT_A_CP_1
3911 : CHRTEST_NOT_A_CP_2;
3913 else if (c1 > 255) {
3914 if (c2 > 255) { /* both possibilities are above what a non-utf8 string
3919 *c1p = *c2p = c2; /* c2 is the only representable value */
3921 else { /* c1 is representable; see about c2 */
3923 *c2p = (c2 < 256) ? c2 : c1;
3929 /* returns -1 on failure, $+[0] on success */
3931 S_regmatch(pTHX_ regmatch_info *reginfo, char *startpos, regnode *prog)
3933 #if PERL_VERSION < 9 && !defined(PERL_CORE)
3937 const bool utf8_target = reginfo->is_utf8_target;
3938 const U32 uniflags = UTF8_ALLOW_DEFAULT;
3939 REGEXP *rex_sv = reginfo->prog;
3940 regexp *rex = ReANY(rex_sv);
3941 RXi_GET_DECL(rex,rexi);
3942 /* the current state. This is a cached copy of PL_regmatch_state */
3944 /* cache heavy used fields of st in registers */
3947 U32 n = 0; /* general value; init to avoid compiler warning */
3948 SSize_t ln = 0; /* len or last; init to avoid compiler warning */
3949 char *locinput = startpos;
3950 char *pushinput; /* where to continue after a PUSH */
3951 I32 nextchr; /* is always set to UCHARAT(locinput) */
3953 bool result = 0; /* return value of S_regmatch */
3954 int depth = 0; /* depth of backtrack stack */
3955 U32 nochange_depth = 0; /* depth of GOSUB recursion with nochange */
3956 const U32 max_nochange_depth =
3957 (3 * rex->nparens > MAX_RECURSE_EVAL_NOCHANGE_DEPTH) ?
3958 3 * rex->nparens : MAX_RECURSE_EVAL_NOCHANGE_DEPTH;
3959 regmatch_state *yes_state = NULL; /* state to pop to on success of
3961 /* mark_state piggy backs on the yes_state logic so that when we unwind
3962 the stack on success we can update the mark_state as we go */
3963 regmatch_state *mark_state = NULL; /* last mark state we have seen */
3964 regmatch_state *cur_eval = NULL; /* most recent EVAL_AB state */
3965 struct regmatch_state *cur_curlyx = NULL; /* most recent curlyx */
3967 bool no_final = 0; /* prevent failure from backtracking? */
3968 bool do_cutgroup = 0; /* no_final only until next branch/trie entry */
3969 char *startpoint = locinput;
3970 SV *popmark = NULL; /* are we looking for a mark? */
3971 SV *sv_commit = NULL; /* last mark name seen in failure */
3972 SV *sv_yes_mark = NULL; /* last mark name we have seen
3973 during a successful match */
3974 U32 lastopen = 0; /* last open we saw */
3975 bool has_cutgroup = RX_HAS_CUTGROUP(rex) ? 1 : 0;
3976 SV* const oreplsv = GvSVn(PL_replgv);
3977 /* these three flags are set by various ops to signal information to
3978 * the very next op. They have a useful lifetime of exactly one loop
3979 * iteration, and are not preserved or restored by state pushes/pops
3981 bool sw = 0; /* the condition value in (?(cond)a|b) */
3982 bool minmod = 0; /* the next "{n,m}" is a "{n,m}?" */
3983 int logical = 0; /* the following EVAL is:
3987 or the following IFMATCH/UNLESSM is:
3988 false: plain (?=foo)
3989 true: used as a condition: (?(?=foo))
3991 PAD* last_pad = NULL;
3993 I32 gimme = G_SCALAR;
3994 CV *caller_cv = NULL; /* who called us */
3995 CV *last_pushed_cv = NULL; /* most recently called (?{}) CV */
3996 CHECKPOINT runops_cp; /* savestack position before executing EVAL */
3997 U32 maxopenparen = 0; /* max '(' index seen so far */
3998 int to_complement; /* Invert the result? */
3999 _char_class_number classnum;
4000 bool is_utf8_pat = reginfo->is_utf8_pat;
4003 GET_RE_DEBUG_FLAGS_DECL;
4006 /* protect against undef(*^R) */
4007 SAVEFREESV(SvREFCNT_inc_simple_NN(oreplsv));
4009 /* shut up 'may be used uninitialized' compiler warnings for dMULTICALL */
4010 multicall_oldcatch = 0;
4011 multicall_cv = NULL;
4013 PERL_UNUSED_VAR(multicall_cop);
4014 PERL_UNUSED_VAR(newsp);
4017 PERL_ARGS_ASSERT_REGMATCH;
4019 DEBUG_OPTIMISE_r( DEBUG_EXECUTE_r({
4020 PerlIO_printf(Perl_debug_log,"regmatch start\n");
4023 st = PL_regmatch_state;
4025 /* Note that nextchr is a byte even in UTF */
4028 while (scan != NULL) {
4031 SV * const prop = sv_newmortal();
4032 regnode *rnext=regnext(scan);
4033 DUMP_EXEC_POS( locinput, scan, utf8_target );
4034 regprop(rex, prop, scan, reginfo, NULL);
4036 PerlIO_printf(Perl_debug_log,
4037 "%3"IVdf":%*s%s(%"IVdf")\n",
4038 (IV)(scan - rexi->program), depth*2, "",
4040 (PL_regkind[OP(scan)] == END || !rnext) ?
4041 0 : (IV)(rnext - rexi->program));
4044 next = scan + NEXT_OFF(scan);
4047 state_num = OP(scan);
4053 assert(nextchr < 256 && (nextchr >= 0 || nextchr == NEXTCHR_EOS));
4055 switch (state_num) {
4056 case SBOL: /* /^../ and /\A../ */
4057 if (locinput == reginfo->strbeg)
4061 case MBOL: /* /^../m */
4062 if (locinput == reginfo->strbeg ||
4063 (!NEXTCHR_IS_EOS && locinput[-1] == '\n'))
4070 if (locinput == reginfo->ganch)
4074 case KEEPS: /* \K */
4075 /* update the startpoint */
4076 st->u.keeper.val = rex->offs[0].start;
4077 rex->offs[0].start = locinput - reginfo->strbeg;
4078 PUSH_STATE_GOTO(KEEPS_next, next, locinput);
4082 case KEEPS_next_fail:
4083 /* rollback the start point change */
4084 rex->offs[0].start = st->u.keeper.val;
4089 case MEOL: /* /..$/m */
4090 if (!NEXTCHR_IS_EOS && nextchr != '\n')
4094 case SEOL: /* /..$/ */
4095 if (!NEXTCHR_IS_EOS && nextchr != '\n')
4097 if (reginfo->strend - locinput > 1)
4102 if (!NEXTCHR_IS_EOS)
4106 case SANY: /* /./s */
4109 goto increment_locinput;
4117 case REG_ANY: /* /./ */
4118 if ((NEXTCHR_IS_EOS) || nextchr == '\n')
4120 goto increment_locinput;
4124 #define ST st->u.trie
4125 case TRIEC: /* (ab|cd) with known charclass */
4126 /* In this case the charclass data is available inline so
4127 we can fail fast without a lot of extra overhead.
4129 if(!NEXTCHR_IS_EOS && !ANYOF_BITMAP_TEST(scan, nextchr)) {
4131 PerlIO_printf(Perl_debug_log,
4132 "%*s %sfailed to match trie start class...%s\n",
4133 REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5])
4140 case TRIE: /* (ab|cd) */
4141 /* the basic plan of execution of the trie is:
4142 * At the beginning, run though all the states, and
4143 * find the longest-matching word. Also remember the position
4144 * of the shortest matching word. For example, this pattern:
4147 * when matched against the string "abcde", will generate
4148 * accept states for all words except 3, with the longest
4149 * matching word being 4, and the shortest being 2 (with
4150 * the position being after char 1 of the string).
4152 * Then for each matching word, in word order (i.e. 1,2,4,5),
4153 * we run the remainder of the pattern; on each try setting
4154 * the current position to the character following the word,
4155 * returning to try the next word on failure.
4157 * We avoid having to build a list of words at runtime by
4158 * using a compile-time structure, wordinfo[].prev, which
4159 * gives, for each word, the previous accepting word (if any).
4160 * In the case above it would contain the mappings 1->2, 2->0,
4161 * 3->0, 4->5, 5->1. We can use this table to generate, from
4162 * the longest word (4 above), a list of all words, by
4163 * following the list of prev pointers; this gives us the
4164 * unordered list 4,5,1,2. Then given the current word we have
4165 * just tried, we can go through the list and find the
4166 * next-biggest word to try (so if we just failed on word 2,
4167 * the next in the list is 4).
4169 * Since at runtime we don't record the matching position in
4170 * the string for each word, we have to work that out for
4171 * each word we're about to process. The wordinfo table holds
4172 * the character length of each word; given that we recorded
4173 * at the start: the position of the shortest word and its
4174 * length in chars, we just need to move the pointer the
4175 * difference between the two char lengths. Depending on
4176 * Unicode status and folding, that's cheap or expensive.
4178 * This algorithm is optimised for the case where are only a
4179 * small number of accept states, i.e. 0,1, or maybe 2.
4180 * With lots of accepts states, and having to try all of them,
4181 * it becomes quadratic on number of accept states to find all
4186 /* what type of TRIE am I? (utf8 makes this contextual) */
4187 DECL_TRIE_TYPE(scan);
4189 /* what trie are we using right now */
4190 reg_trie_data * const trie
4191 = (reg_trie_data*)rexi->data->data[ ARG( scan ) ];
4192 HV * widecharmap = MUTABLE_HV(rexi->data->data[ ARG( scan ) + 1 ]);
4193 U32 state = trie->startstate;
4195 if (scan->flags == EXACTL || scan->flags == EXACTFLU8) {
4196 _CHECK_AND_WARN_PROBLEMATIC_LOCALE;
4198 && UTF8_IS_ABOVE_LATIN1(nextchr)
4199 && scan->flags == EXACTL)
4201 /* We only output for EXACTL, as we let the folder
4202 * output this message for EXACTFLU8 to avoid
4204 _CHECK_AND_OUTPUT_WIDE_LOCALE_UTF8_MSG(locinput,
4209 && (NEXTCHR_IS_EOS || !TRIE_BITMAP_TEST(trie, nextchr)))
4211 if (trie->states[ state ].wordnum) {
4213 PerlIO_printf(Perl_debug_log,
4214 "%*s %smatched empty string...%s\n",
4215 REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5])
4221 PerlIO_printf(Perl_debug_log,
4222 "%*s %sfailed to match trie start class...%s\n",
4223 REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5])
4230 U8 *uc = ( U8* )locinput;
4234 U8 *uscan = (U8*)NULL;
4235 U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ];
4236 U32 charcount = 0; /* how many input chars we have matched */
4237 U32 accepted = 0; /* have we seen any accepting states? */
4239 ST.jump = trie->jump;
4242 ST.longfold = FALSE; /* char longer if folded => it's harder */
4245 /* fully traverse the TRIE; note the position of the
4246 shortest accept state and the wordnum of the longest
4249 while ( state && uc <= (U8*)(reginfo->strend) ) {
4250 U32 base = trie->states[ state ].trans.base;
4254 wordnum = trie->states[ state ].wordnum;
4256 if (wordnum) { /* it's an accept state */
4259 /* record first match position */
4261 ST.firstpos = (U8*)locinput;
4266 ST.firstchars = charcount;
4269 if (!ST.nextword || wordnum < ST.nextword)
4270 ST.nextword = wordnum;
4271 ST.topword = wordnum;
4274 DEBUG_TRIE_EXECUTE_r({
4275 DUMP_EXEC_POS( (char *)uc, scan, utf8_target );
4276 PerlIO_printf( Perl_debug_log,
4277 "%*s %sState: %4"UVxf" Accepted: %c ",
4278 2+depth * 2, "", PL_colors[4],
4279 (UV)state, (accepted ? 'Y' : 'N'));
4282 /* read a char and goto next state */
4283 if ( base && (foldlen || uc < (U8*)(reginfo->strend))) {
4285 REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc,
4286 uscan, len, uvc, charid, foldlen,
4293 base + charid - 1 - trie->uniquecharcount)) >= 0)
4295 && ((U32)offset < trie->lasttrans)
4296 && trie->trans[offset].check == state)
4298 state = trie->trans[offset].next;
4309 DEBUG_TRIE_EXECUTE_r(
4310 PerlIO_printf( Perl_debug_log,
4311 "Charid:%3x CP:%4"UVxf" After State: %4"UVxf"%s\n",
4312 charid, uvc, (UV)state, PL_colors[5] );
4318 /* calculate total number of accept states */
4323 w = trie->wordinfo[w].prev;
4326 ST.accepted = accepted;
4330 PerlIO_printf( Perl_debug_log,
4331 "%*s %sgot %"IVdf" possible matches%s\n",
4332 REPORT_CODE_OFF + depth * 2, "",
4333 PL_colors[4], (IV)ST.accepted, PL_colors[5] );
4335 goto trie_first_try; /* jump into the fail handler */
4340 case TRIE_next_fail: /* we failed - try next alternative */
4344 REGCP_UNWIND(ST.cp);
4345 UNWIND_PAREN(ST.lastparen, ST.lastcloseparen);
4347 if (!--ST.accepted) {
4349 PerlIO_printf( Perl_debug_log,
4350 "%*s %sTRIE failed...%s\n",
4351 REPORT_CODE_OFF+depth*2, "",
4358 /* Find next-highest word to process. Note that this code
4359 * is O(N^2) per trie run (O(N) per branch), so keep tight */
4362 U16 const nextword = ST.nextword;
4363 reg_trie_wordinfo * const wordinfo
4364 = ((reg_trie_data*)rexi->data->data[ARG(ST.me)])->wordinfo;
4365 for (word=ST.topword; word; word=wordinfo[word].prev) {
4366 if (word > nextword && (!min || word < min))
4379 ST.lastparen = rex->lastparen;
4380 ST.lastcloseparen = rex->lastcloseparen;
4384 /* find start char of end of current word */
4386 U32 chars; /* how many chars to skip */
4387 reg_trie_data * const trie
4388 = (reg_trie_data*)rexi->data->data[ARG(ST.me)];
4390 assert((trie->wordinfo[ST.nextword].len - trie->prefixlen)
4392 chars = (trie->wordinfo[ST.nextword].len - trie->prefixlen)
4397 /* the hard option - fold each char in turn and find
4398 * its folded length (which may be different */
4399 U8 foldbuf[UTF8_MAXBYTES_CASE + 1];
4407 uvc = utf8n_to_uvchr((U8*)uc, UTF8_MAXLEN, &len,
4415 uvc = to_uni_fold(uvc, foldbuf, &foldlen);
4420 uvc = utf8n_to_uvchr(uscan, UTF8_MAXLEN, &len,
4436 scan = ST.me + ((ST.jump && ST.jump[ST.nextword])
4437 ? ST.jump[ST.nextword]
4441 PerlIO_printf( Perl_debug_log,
4442 "%*s %sTRIE matched word #%d, continuing%s\n",
4443 REPORT_CODE_OFF+depth*2, "",
4450 if (ST.accepted > 1 || has_cutgroup) {
4451 PUSH_STATE_GOTO(TRIE_next, scan, (char*)uc);
4455 /* only one choice left - just continue */
4457 AV *const trie_words
4458 = MUTABLE_AV(rexi->data->data[ARG(ST.me)+TRIE_WORDS_OFFSET]);
4459 SV ** const tmp = av_fetch( trie_words,
4461 SV *sv= tmp ? sv_newmortal() : NULL;
4463 PerlIO_printf( Perl_debug_log,
4464 "%*s %sonly one match left, short-circuiting: #%d <%s>%s\n",
4465 REPORT_CODE_OFF+depth*2, "", PL_colors[4],
4467 tmp ? pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), 0,
4468 PL_colors[0], PL_colors[1],
4469 (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0)|PERL_PV_ESCAPE_NONASCII
4471 : "not compiled under -Dr",
4475 locinput = (char*)uc;
4476 continue; /* execute rest of RE */
4481 case EXACTL: /* /abc/l */
4482 _CHECK_AND_WARN_PROBLEMATIC_LOCALE;
4484 /* Complete checking would involve going through every character
4485 * matched by the string to see if any is above latin1. But the
4486 * comparision otherwise might very well be a fast assembly
4487 * language routine, and I (khw) don't think slowing things down
4488 * just to check for this warning is worth it. So this just checks
4489 * the first character */
4490 if (utf8_target && UTF8_IS_ABOVE_LATIN1(*locinput)) {
4491 _CHECK_AND_OUTPUT_WIDE_LOCALE_UTF8_MSG(locinput, reginfo->strend);
4494 case EXACT: { /* /abc/ */
4495 char *s = STRING(scan);
4497 if (utf8_target != is_utf8_pat) {
4498 /* The target and the pattern have differing utf8ness. */
4500 const char * const e = s + ln;
4503 /* The target is utf8, the pattern is not utf8.
4504 * Above-Latin1 code points can't match the pattern;
4505 * invariants match exactly, and the other Latin1 ones need
4506 * to be downgraded to a single byte in order to do the
4507 * comparison. (If we could be confident that the target
4508 * is not malformed, this could be refactored to have fewer
4509 * tests by just assuming that if the first bytes match, it
4510 * is an invariant, but there are tests in the test suite
4511 * dealing with (??{...}) which violate this) */
4513 if (l >= reginfo->strend
4514 || UTF8_IS_ABOVE_LATIN1(* (U8*) l))
4518 if (UTF8_IS_INVARIANT(*(U8*)l)) {
4525 if (TWO_BYTE_UTF8_TO_NATIVE(*l, *(l+1)) != * (U8*) s)
4535 /* The target is not utf8, the pattern is utf8. */
4537 if (l >= reginfo->strend
4538 || UTF8_IS_ABOVE_LATIN1(* (U8*) s))
4542 if (UTF8_IS_INVARIANT(*(U8*)s)) {
4549 if (TWO_BYTE_UTF8_TO_NATIVE(*s, *(s+1)) != * (U8*) l)
4561 /* The target and the pattern have the same utf8ness. */
4562 /* Inline the first character, for speed. */
4563 if (reginfo->strend - locinput < ln
4564 || UCHARAT(s) != nextchr
4565 || (ln > 1 && memNE(s, locinput, ln)))
4574 case EXACTFL: { /* /abc/il */
4576 const U8 * fold_array;
4578 U32 fold_utf8_flags;
4580 _CHECK_AND_WARN_PROBLEMATIC_LOCALE;
4581 folder = foldEQ_locale;
4582 fold_array = PL_fold_locale;
4583 fold_utf8_flags = FOLDEQ_LOCALE;
4586 case EXACTFLU8: /* /abc/il; but all 'abc' are above 255, so
4587 is effectively /u; hence to match, target
4589 if (! utf8_target) {
4592 fold_utf8_flags = FOLDEQ_LOCALE | FOLDEQ_S1_ALREADY_FOLDED
4593 | FOLDEQ_S1_FOLDS_SANE;
4594 folder = foldEQ_latin1;
4595 fold_array = PL_fold_latin1;
4598 case EXACTFU_SS: /* /\x{df}/iu */
4599 case EXACTFU: /* /abc/iu */
4600 folder = foldEQ_latin1;
4601 fold_array = PL_fold_latin1;
4602 fold_utf8_flags = is_utf8_pat ? FOLDEQ_S1_ALREADY_FOLDED : 0;
4605 case EXACTFA_NO_TRIE: /* This node only generated for non-utf8
4607 assert(! is_utf8_pat);
4609 case EXACTFA: /* /abc/iaa */
4610 folder = foldEQ_latin1;
4611 fold_array = PL_fold_latin1;
4612 fold_utf8_flags = FOLDEQ_UTF8_NOMIX_ASCII;
4615 case EXACTF: /* /abc/i This node only generated for
4616 non-utf8 patterns */
4617 assert(! is_utf8_pat);
4619 fold_array = PL_fold;
4620 fold_utf8_flags = 0;
4628 || state_num == EXACTFU_SS
4629 || (state_num == EXACTFL && IN_UTF8_CTYPE_LOCALE))
4631 /* Either target or the pattern are utf8, or has the issue where
4632 * the fold lengths may differ. */
4633 const char * const l = locinput;
4634 char *e = reginfo->strend;
4636 if (! foldEQ_utf8_flags(s, 0, ln, is_utf8_pat,
4637 l, &e, 0, utf8_target, fold_utf8_flags))
4645 /* Neither the target nor the pattern are utf8 */
4646 if (UCHARAT(s) != nextchr
4648 && UCHARAT(s) != fold_array[nextchr])
4652 if (reginfo->strend - locinput < ln)
4654 if (ln > 1 && ! folder(s, locinput, ln))
4660 /* XXX At that point regcomp.c would no longer * have to set the FLAGS fields of these */
4661 case NBOUNDL: /* /\B/l */
4665 case BOUNDL: /* /\b/l */
4666 _CHECK_AND_WARN_PROBLEMATIC_LOCALE;
4668 if (locinput == reginfo->strbeg)
4669 ln = isWORDCHAR_LC('\n');
4671 ln = isWORDCHAR_LC_utf8(reghop3((U8*)locinput, -1,
4672 (U8*)(reginfo->strbeg)));
4674 n = (NEXTCHR_IS_EOS)
4675 ? isWORDCHAR_LC('\n')
4676 : isWORDCHAR_LC_utf8((U8*)locinput);
4678 else { /* Here the string isn't utf8 */
4679 ln = (locinput == reginfo->strbeg)
4680 ? isWORDCHAR_LC('\n')
4681 : isWORDCHAR_LC(UCHARAT(locinput - 1));
4682 n = (NEXTCHR_IS_EOS)
4683 ? isWORDCHAR_LC('\n')
4684 : isWORDCHAR_LC(nextchr);
4686 if (to_complement ^ (ln == n)) {
4691 case NBOUND: /* /\B/ */
4695 case BOUND: /* /\b/ */
4699 goto bound_ascii_match_only;
4701 case NBOUNDA: /* /\B/a */
4705 case BOUNDA: /* /\b/a */
4707 bound_ascii_match_only:
4708 /* Here the string isn't utf8, or is utf8 and only ascii characters
4709 * are to match \w. In the latter case looking at the byte just
4710 * prior to the current one may be just the final byte of a
4711 * multi-byte character. This is ok. There are two cases:
4712 * 1) it is a single byte character, and then the test is doing
4713 * just what it's supposed to.
4714 * 2) it is a multi-byte character, in which case the final byte is
4715 * never mistakable for ASCII, and so the test will say it is
4716 * not a word character, which is the correct answer. */
4717 ln = (locinput == reginfo->strbeg)
4718 ? isWORDCHAR_A('\n')
4719 : isWORDCHAR_A(UCHARAT(locinput - 1));
4720 n = (NEXTCHR_IS_EOS)
4721 ? isWORDCHAR_A('\n')
4722 : isWORDCHAR_A(nextchr);
4723 if (to_complement ^ (ln == n)) {
4728 case NBOUNDU: /* /\B/u */
4732 case BOUNDU: /* /\b/u */
4736 ln = (locinput == reginfo->strbeg)
4737 ? isWORDCHAR_L1('\n')
4738 : isWORDCHAR_utf8(reghop3((U8*)locinput, -1,
4739 (U8*)(reginfo->strbeg)));
4740 n = (NEXTCHR_IS_EOS)
4741 ? isWORDCHAR_L1('\n')
4742 : isWORDCHAR_utf8((U8*)locinput);
4745 ln = (locinput == reginfo->strbeg)
4746 ? isWORDCHAR_L1('\n')
4747 : isWORDCHAR_L1(UCHARAT(locinput - 1));
4748 n = (NEXTCHR_IS_EOS)
4749 ? isWORDCHAR_L1('\n')
4750 : isWORDCHAR_L1(nextchr);
4754 if (to_complement ^ (ln == n)) {
4759 case ANYOFL: /* /[abc]/l */
4760 _CHECK_AND_WARN_PROBLEMATIC_LOCALE;
4762 case ANYOF: /* /[abc]/ */
4766 if (!reginclass(rex, scan, (U8*)locinput, (U8*)reginfo->strend,
4769 locinput += UTF8SKIP(locinput);
4772 if (!REGINCLASS(rex, scan, (U8*)locinput))
4778 /* The argument (FLAGS) to all the POSIX node types is the class number
4781 case NPOSIXL: /* \W or [:^punct:] etc. under /l */
4785 case POSIXL: /* \w or [:punct:] etc. under /l */
4786 _CHECK_AND_WARN_PROBLEMATIC_LOCALE;
4790 /* Use isFOO_lc() for characters within Latin1. (Note that
4791 * UTF8_IS_INVARIANT works even on non-UTF-8 strings, or else
4792 * wouldn't be invariant) */
4793 if (UTF8_IS_INVARIANT(nextchr) || ! utf8_target) {
4794 if (! (to_complement ^ cBOOL(isFOO_lc(FLAGS(scan), (U8) nextchr)))) {
4798 else if (UTF8_IS_DOWNGRADEABLE_START(nextchr)) {
4799 if (! (to_complement ^ cBOOL(isFOO_lc(FLAGS(scan),
4800 (U8) TWO_BYTE_UTF8_TO_NATIVE(nextchr,
4801 *(locinput + 1))))))
4806 else { /* Here, must be an above Latin-1 code point */
4807 _CHECK_AND_OUTPUT_WIDE_LOCALE_UTF8_MSG(locinput, reginfo->strend);
4808 goto utf8_posix_above_latin1;
4811 /* Here, must be utf8 */
4812 locinput += UTF8SKIP(locinput);
4815 case NPOSIXD: /* \W or [:^punct:] etc. under /d */
4819 case POSIXD: /* \w or [:punct:] etc. under /d */
4825 case NPOSIXA: /* \W or [:^punct:] etc. under /a */
4827 if (NEXTCHR_IS_EOS) {
4831 /* All UTF-8 variants match */
4832 if (! UTF8_IS_INVARIANT(nextchr)) {
4833 goto increment_locinput;
4839 case POSIXA: /* \w or [:punct:] etc. under /a */
4842 /* We get here through POSIXD, NPOSIXD, and NPOSIXA when not in
4843 * UTF-8, and also from NPOSIXA even in UTF-8 when the current
4844 * character is a single byte */
4847 || ! (to_complement ^ cBOOL(_generic_isCC_A(nextchr,
4853 /* Here we are either not in utf8, or we matched a utf8-invariant,
4854 * so the next char is the next byte */
4858 case NPOSIXU: /* \W or [:^punct:] etc. under /u */
4862 case POSIXU: /* \w or [:punct:] etc. under /u */
4864 if (NEXTCHR_IS_EOS) {
4868 /* Use _generic_isCC() for characters within Latin1. (Note that
4869 * UTF8_IS_INVARIANT works even on non-UTF-8 strings, or else
4870 * wouldn't be invariant) */
4871 if (UTF8_IS_INVARIANT(nextchr) || ! utf8_target) {
4872 if (! (to_complement ^ cBOOL(_generic_isCC(nextchr,
4879 else if (UTF8_IS_DOWNGRADEABLE_START(nextchr)) {
4880 if (! (to_complement
4881 ^ cBOOL(_generic_isCC(TWO_BYTE_UTF8_TO_NATIVE(nextchr,
4889 else { /* Handle above Latin-1 code points */
4890 utf8_posix_above_latin1:
4891 classnum = (_char_class_number) FLAGS(scan);
4892 if (classnum < _FIRST_NON_SWASH_CC) {
4894 /* Here, uses a swash to find such code points. Load if if
4895 * not done already */
4896 if (! PL_utf8_swash_ptrs[classnum]) {
4897 U8 flags = _CORE_SWASH_INIT_ACCEPT_INVLIST;
4898 PL_utf8_swash_ptrs[classnum]
4899 = _core_swash_init("utf8",
4902 PL_XPosix_ptrs[classnum], &flags);
4904 if (! (to_complement
4905 ^ cBOOL(swash_fetch(PL_utf8_swash_ptrs[classnum],
4906 (U8 *) locinput, TRUE))))
4911 else { /* Here, uses macros to find above Latin-1 code points */
4913 case _CC_ENUM_SPACE: /* XXX would require separate
4914 code if we revert the change
4915 of \v matching this */
4916 case _CC_ENUM_PSXSPC:
4917 if (! (to_complement
4918 ^ cBOOL(is_XPERLSPACE_high(locinput))))
4923 case _CC_ENUM_BLANK:
4924 if (! (to_complement
4925 ^ cBOOL(is_HORIZWS_high(locinput))))
4930 case _CC_ENUM_XDIGIT:
4931 if (! (to_complement
4932 ^ cBOOL(is_XDIGIT_high(locinput))))
4937 case _CC_ENUM_VERTSPACE:
4938 if (! (to_complement
4939 ^ cBOOL(is_VERTWS_high(locinput))))
4944 default: /* The rest, e.g. [:cntrl:], can't match
4946 if (! to_complement) {
4952 locinput += UTF8SKIP(locinput);
4956 case CLUMP: /* Match \X: logical Unicode character. This is defined as
4957 a Unicode extended Grapheme Cluster */
4958 /* From http://www.unicode.org/reports/tr29 (5.2 version). An
4959 extended Grapheme Cluster is:
4962 | Prepend* Begin Extend*
4965 Begin is: ( Special_Begin | ! Control )
4966 Special_Begin is: ( Regional-Indicator+ | Hangul-syllable )
4967 Extend is: ( Grapheme_Extend | Spacing_Mark )
4968 Control is: [ GCB_Control | CR | LF ]
4969 Hangul-syllable is: ( T+ | ( L* ( L | ( LVT | ( V | LV ) V* ) T* ) ))
4971 If we create a 'Regular_Begin' = Begin - Special_Begin, then
4974 Begin is ( Regular_Begin + Special Begin )
4976 It turns out that 98.4% of all Unicode code points match
4977 Regular_Begin. Doing it this way eliminates a table match in
4978 the previous implementation for almost all Unicode code points.
4980 There is a subtlety with Prepend* which showed up in testing.
4981 Note that the Begin, and only the Begin is required in:
4982 | Prepend* Begin Extend*
4983 Also, Begin contains '! Control'. A Prepend must be a
4984 '! Control', which means it must also be a Begin. What it
4985 comes down to is that if we match Prepend* and then find no
4986 suitable Begin afterwards, that if we backtrack the last
4987 Prepend, that one will be a suitable Begin.
4992 if (! utf8_target) {
4994 /* Match either CR LF or '.', as all the other possibilities
4996 locinput++; /* Match the . or CR */
4997 if (nextchr == '\r' /* And if it was CR, and the next is LF,
4999 && locinput < reginfo->strend
5000 && UCHARAT(locinput) == '\n')
5007 /* Utf8: See if is ( CR LF ); already know that locinput <
5008 * reginfo->strend, so locinput+1 is in bounds */
5009 if ( nextchr == '\r' && locinput+1 < reginfo->strend
5010 && UCHARAT(locinput + 1) == '\n')
5017 /* In case have to backtrack to beginning, then match '.' */
5018 char *starting = locinput;
5020 /* In case have to backtrack the last prepend */
5021 char *previous_prepend = NULL;
5023 LOAD_UTF8_CHARCLASS_GCB();
5025 /* Match (prepend)* */
5026 while (locinput < reginfo->strend
5027 && (len = is_GCB_Prepend_utf8(locinput)))
5029 previous_prepend = locinput;
5033 /* As noted above, if we matched a prepend character, but
5034 * the next thing won't match, back off the last prepend we
5035 * matched, as it is guaranteed to match the begin */
5036 if (previous_prepend
5037 && (locinput >= reginfo->strend
5038 || (! swash_fetch(PL_utf8_X_regular_begin,
5039 (U8*)locinput, utf8_target)
5040 && ! is_GCB_SPECIAL_BEGIN_START_utf8(locinput)))
5043 locinput = previous_prepend;
5046 /* Note that here we know reginfo->strend > locinput, as we
5047 * tested that upon input to this switch case, and if we
5048 * moved locinput forward, we tested the result just above
5049 * and it either passed, or we backed off so that it will
5051 if (swash_fetch(PL_utf8_X_regular_begin,
5052 (U8*)locinput, utf8_target)) {
5053 locinput += UTF8SKIP(locinput);
5055 else if (! is_GCB_SPECIAL_BEGIN_START_utf8(locinput)) {
5057 /* Here did not match the required 'Begin' in the
5058 * second term. So just match the very first
5059 * character, the '.' of the final term of the regex */
5060 locinput = starting + UTF8SKIP(starting);
5064 /* Here is a special begin. It can be composed of
5065 * several individual characters. One possibility is
5067 if ((len = is_GCB_RI_utf8(locinput))) {
5069 while (locinput < reginfo->strend
5070 && (len = is_GCB_RI_utf8(locinput)))
5074 } else if ((len = is_GCB_T_utf8(locinput))) {
5075 /* Another possibility is T+ */
5077 while (locinput < reginfo->strend
5078 && (len = is_GCB_T_utf8(locinput)))
5084 /* Here, neither RI+ nor T+; must be some other
5085 * Hangul. That means it is one of the others: L,
5086 * LV, LVT or V, and matches:
5087 * L* (L | LVT T* | V * V* T* | LV V* T*) */
5090 while (locinput < reginfo->strend
5091 && (len = is_GCB_L_utf8(locinput)))
5096 /* Here, have exhausted L*. If the next character
5097 * is not an LV, LVT nor V, it means we had to have
5098 * at least one L, so matches L+ in the original
5099 * equation, we have a complete hangul syllable.
5102 if (locinput < reginfo->strend
5103 && is_GCB_LV_LVT_V_utf8(locinput))
5105 /* Otherwise keep going. Must be LV, LVT or V.
5106 * See if LVT, by first ruling out V, then LV */
5107 if (! is_GCB_V_utf8(locinput)
5108 /* All but every TCount one is LV */
5109 && (valid_utf8_to_uvchr((U8 *) locinput,
5114 locinput += UTF8SKIP(locinput);
5117 /* Must be V or LV. Take it, then match
5119 locinput += UTF8SKIP(locinput);
5120 while (locinput < reginfo->strend
5121 && (len = is_GCB_V_utf8(locinput)))
5127 /* And any of LV, LVT, or V can be followed
5129 while (locinput < reginfo->strend
5130 && (len = is_GCB_T_utf8(locinput)))
5138 /* Match any extender */
5139 while (locinput < reginfo->strend
5140 && swash_fetch(PL_utf8_X_extend,
5141 (U8*)locinput, utf8_target))
5143 locinput += UTF8SKIP(locinput);
5147 if (locinput > reginfo->strend) sayNO;
5151 case NREFFL: /* /\g{name}/il */
5152 { /* The capture buffer cases. The ones beginning with N for the
5153 named buffers just convert to the equivalent numbered and
5154 pretend they were called as the corresponding numbered buffer
5156 /* don't initialize these in the declaration, it makes C++
5161 const U8 *fold_array;
5164 _CHECK_AND_WARN_PROBLEMATIC_LOCALE;
5165 folder = foldEQ_locale;
5166 fold_array = PL_fold_locale;
5168 utf8_fold_flags = FOLDEQ_LOCALE;
5171 case NREFFA: /* /\g{name}/iaa */
5172 folder = foldEQ_latin1;
5173 fold_array = PL_fold_latin1;
5175 utf8_fold_flags = FOLDEQ_UTF8_NOMIX_ASCII;
5178 case NREFFU: /* /\g{name}/iu */
5179 folder = foldEQ_latin1;
5180 fold_array = PL_fold_latin1;
5182 utf8_fold_flags = 0;
5185 case NREFF: /* /\g{name}/i */
5187 fold_array = PL_fold;
5189 utf8_fold_flags = 0;
5192 case NREF: /* /\g{name}/ */
5196 utf8_fold_flags = 0;
5199 /* For the named back references, find the corresponding buffer
5201 n = reg_check_named_buff_matched(rex,scan);
5206 goto do_nref_ref_common;
5208 case REFFL: /* /\1/il */
5209 _CHECK_AND_WARN_PROBLEMATIC_LOCALE;
5210 folder = foldEQ_locale;
5211 fold_array = PL_fold_locale;
5212 utf8_fold_flags = FOLDEQ_LOCALE;
5215 case REFFA: /* /\1/iaa */
5216 folder = foldEQ_latin1;
5217 fold_array = PL_fold_latin1;
5218 utf8_fold_flags = FOLDEQ_UTF8_NOMIX_ASCII;
5221 case REFFU: /* /\1/iu */
5222 folder = foldEQ_latin1;
5223 fold_array = PL_fold_latin1;
5224 utf8_fold_flags = 0;
5227 case REFF: /* /\1/i */
5229 fold_array = PL_fold;
5230 utf8_fold_flags = 0;
5233 case REF: /* /\1/ */
5236 utf8_fold_flags = 0;
5240 n = ARG(scan); /* which paren pair */
5243 ln = rex->offs[n].start;
5244 reginfo->poscache_iter = reginfo->poscache_maxiter; /* Void cache */
5245 if (rex->lastparen < n || ln == -1)
5246 sayNO; /* Do not match unless seen CLOSEn. */
5247 if (ln == rex->offs[n].end)
5250 s = reginfo->strbeg + ln;
5251 if (type != REF /* REF can do byte comparison */
5252 && (utf8_target || type == REFFU || type == REFFL))
5254 char * limit = reginfo->strend;
5256 /* This call case insensitively compares the entire buffer
5257 * at s, with the current input starting at locinput, but
5258 * not going off the end given by reginfo->strend, and
5259 * returns in <limit> upon success, how much of the
5260 * current input was matched */
5261 if (! foldEQ_utf8_flags(s, NULL, rex->offs[n].end - ln, utf8_target,
5262 locinput, &limit, 0, utf8_target, utf8_fold_flags))
5270 /* Not utf8: Inline the first character, for speed. */
5271 if (!NEXTCHR_IS_EOS &&
5272 UCHARAT(s) != nextchr &&
5274 UCHARAT(s) != fold_array[nextchr]))
5276 ln = rex->offs[n].end - ln;
5277 if (locinput + ln > reginfo->strend)
5279 if (ln > 1 && (type == REF
5280 ? memNE(s, locinput, ln)
5281 : ! folder(s, locinput, ln)))
5287 case NOTHING: /* null op; e.g. the 'nothing' following
5288 * the '*' in m{(a+|b)*}' */
5290 case TAIL: /* placeholder while compiling (A|B|C) */
5294 #define ST st->u.eval
5299 regexp_internal *rei;
5300 regnode *startpoint;
5302 case GOSTART: /* (?R) */
5303 case GOSUB: /* /(...(?1))/ /(...(?&foo))/ */
5304 if (cur_eval && cur_eval->locinput==locinput) {
5305 if (cur_eval->u.eval.close_paren == (U32)ARG(scan))
5306 Perl_croak(aTHX_ "Infinite recursion in regex");
5307 if ( ++nochange_depth > max_nochange_depth )
5309 "Pattern subroutine nesting without pos change"
5310 " exceeded limit in regex");
5317 if (OP(scan)==GOSUB) {
5318 startpoint = scan + ARG2L(scan);
5319 ST.close_paren = ARG(scan);
5321 startpoint = rei->program+1;
5325 /* Save all the positions seen so far. */
5326 ST.cp = regcppush(rex, 0, maxopenparen);
5327 REGCP_SET(ST.lastcp);
5329 /* and then jump to the code we share with EVAL */
5330 goto eval_recurse_doit;
5333 case EVAL: /* /(?{A})B/ /(??{A})B/ and /(?(?{A})X|Y)B/ */
5334 if (cur_eval && cur_eval->locinput==locinput) {
5335 if ( ++nochange_depth > max_nochange_depth )
5336 Perl_croak(aTHX_ "EVAL without pos change exceeded limit in regex");
5341 /* execute the code in the {...} */
5345 OP * const oop = PL_op;
5346 COP * const ocurcop = PL_curcop;
5350 /* save *all* paren positions */
5351 regcppush(rex, 0, maxopenparen);
5352 REGCP_SET(runops_cp);
5355 caller_cv = find_runcv(NULL);
5359 if (rexi->data->what[n] == 'r') { /* code from an external qr */
5361 (REGEXP*)(rexi->data->data[n])
5364 nop = (OP*)rexi->data->data[n+1];
5366 else if (rexi->data->what[n] == 'l') { /* literal code */
5368 nop = (OP*)rexi->data->data[n];
5369 assert(CvDEPTH(newcv));
5372 /* literal with own CV */
5373 assert(rexi->data->what[n] == 'L');
5374 newcv = rex->qr_anoncv;
5375 nop = (OP*)rexi->data->data[n];
5378 /* normally if we're about to execute code from the same
5379 * CV that we used previously, we just use the existing
5380 * CX stack entry. However, its possible that in the
5381 * meantime we may have backtracked, popped from the save
5382 * stack, and undone the SAVECOMPPAD(s) associated with
5383 * PUSH_MULTICALL; in which case PL_comppad no longer
5384 * points to newcv's pad. */
5385 if (newcv != last_pushed_cv || PL_comppad != last_pad)
5387 U8 flags = (CXp_SUB_RE |
5388 ((newcv == caller_cv) ? CXp_SUB_RE_FAKE : 0));
5389 if (last_pushed_cv) {
5390 CHANGE_MULTICALL_FLAGS(newcv, flags);
5393 PUSH_MULTICALL_FLAGS(newcv, flags);
5395 last_pushed_cv = newcv;
5398 /* these assignments are just to silence compiler
5400 multicall_cop = NULL;
5403 last_pad = PL_comppad;
5405 /* the initial nextstate you would normally execute
5406 * at the start of an eval (which would cause error
5407 * messages to come from the eval), may be optimised
5408 * away from the execution path in the regex code blocks;
5409 * so manually set PL_curcop to it initially */
5411 OP *o = cUNOPx(nop)->op_first;
5412 assert(o->op_type == OP_NULL);
5413 if (o->op_targ == OP_SCOPE) {
5414 o = cUNOPo->op_first;
5417 assert(o->op_targ == OP_LEAVE);
5418 o = cUNOPo->op_first;
5419 assert(o->op_type == OP_ENTER);
5423 if (o->op_type != OP_STUB) {
5424 assert( o->op_type == OP_NEXTSTATE
5425 || o->op_type == OP_DBSTATE
5426 || (o->op_type == OP_NULL
5427 && ( o->op_targ == OP_NEXTSTATE
5428 || o->op_targ == OP_DBSTATE
5432 PL_curcop = (COP*)o;
5437 DEBUG_STATE_r( PerlIO_printf(Perl_debug_log,
5438 " re EVAL PL_op=0x%"UVxf"\n", PTR2UV(nop)) );
5440 rex->offs[0].end = locinput - reginfo->strbeg;
5441 if (reginfo->info_aux_eval->pos_magic)
5442 MgBYTEPOS_set(reginfo->info_aux_eval->pos_magic,
5443 reginfo->sv, reginfo->strbeg,
5444 locinput - reginfo->strbeg);
5447 SV *sv_mrk = get_sv("REGMARK", 1);
5448 sv_setsv(sv_mrk, sv_yes_mark);
5451 /* we don't use MULTICALL here as we want to call the
5452 * first op of the block of interest, rather than the
5453 * first op of the sub */
5454 before = (IV)(SP-PL_stack_base);
5456 CALLRUNOPS(aTHX); /* Scalar context. */
5458 if ((IV)(SP-PL_stack_base) == before)
5459 ret = &PL_sv_undef; /* protect against empty (?{}) blocks. */
5465 /* before restoring everything, evaluate the returned
5466 * value, so that 'uninit' warnings don't use the wrong
5467 * PL_op or pad. Also need to process any magic vars
5468 * (e.g. $1) *before* parentheses are restored */
5473 if (logical == 0) /* (?{})/ */
5474 sv_setsv(save_scalar(PL_replgv), ret); /* $^R */
5475 else if (logical == 1) { /* /(?(?{...})X|Y)/ */
5476 sw = cBOOL(SvTRUE(ret));
5479 else { /* /(??{}) */
5480 /* if its overloaded, let the regex compiler handle
5481 * it; otherwise extract regex, or stringify */
5482 if (SvGMAGICAL(ret))
5483 ret = sv_mortalcopy(ret);
5484 if (!SvAMAGIC(ret)) {
5488 if (SvTYPE(sv) == SVt_REGEXP)
5489 re_sv = (REGEXP*) sv;
5490 else if (SvSMAGICAL(ret)) {
5491 MAGIC *mg = mg_find(ret, PERL_MAGIC_qr);
5493 re_sv = (REGEXP *) mg->mg_obj;
5496 /* force any undef warnings here */
5497 if (!re_sv && !SvPOK(ret) && !SvNIOK(ret)) {
5498 ret = sv_mortalcopy(ret);
5499 (void) SvPV_force_nolen(ret);
5505 /* *** Note that at this point we don't restore
5506 * PL_comppad, (or pop the CxSUB) on the assumption it may
5507 * be used again soon. This is safe as long as nothing
5508 * in the regexp code uses the pad ! */
5510 PL_curcop = ocurcop;
5511 S_regcp_restore(aTHX_ rex, runops_cp, &maxopenparen);
5512 PL_curpm = PL_reg_curpm;
5518 /* only /(??{})/ from now on */
5521 /* extract RE object from returned value; compiling if
5525 re_sv = reg_temp_copy(NULL, re_sv);
5530 if (SvUTF8(ret) && IN_BYTES) {
5531 /* In use 'bytes': make a copy of the octet
5532 * sequence, but without the flag on */
5534 const char *const p = SvPV(ret, len);
5535 ret = newSVpvn_flags(p, len, SVs_TEMP);
5537 if (rex->intflags & PREGf_USE_RE_EVAL)
5538 pm_flags |= PMf_USE_RE_EVAL;
5540 /* if we got here, it should be an engine which
5541 * supports compiling code blocks and stuff */
5542 assert(rex->engine && rex->engine->op_comp);
5543 assert(!(scan->flags & ~RXf_PMf_COMPILETIME));
5544 re_sv = rex->engine->op_comp(aTHX_ &ret, 1, NULL,
5545 rex->engine, NULL, NULL,
5546 /* copy /msixn etc to inner pattern */
5551 & (SVs_TEMP | SVs_GMG | SVf_ROK))
5552 && (!SvPADTMP(ret) || SvREADONLY(ret))) {
5553 /* This isn't a first class regexp. Instead, it's
5554 caching a regexp onto an existing, Perl visible
5556 sv_magic(ret, MUTABLE_SV(re_sv), PERL_MAGIC_qr, 0, 0);
5562 RXp_MATCH_COPIED_off(re);
5563 re->subbeg = rex->subbeg;
5564 re->sublen = rex->sublen;
5565 re->suboffset = rex->suboffset;
5566 re->subcoffset = rex->subcoffset;
5568 re->lastcloseparen = 0;
5571 debug_start_match(re_sv, utf8_target, locinput,
5572 reginfo->strend, "Matching embedded");
5574 startpoint = rei->program + 1;
5575 ST.close_paren = 0; /* only used for GOSUB */
5576 /* Save all the seen positions so far. */
5577 ST.cp = regcppush(rex, 0, maxopenparen);
5578 REGCP_SET(ST.lastcp);
5579 /* and set maxopenparen to 0, since we are starting a "fresh" match */
5581 /* run the pattern returned from (??{...}) */
5583 eval_recurse_doit: /* Share code with GOSUB below this line
5584 * At this point we expect the stack context to be
5585 * set up correctly */
5587 /* invalidate the S-L poscache. We're now executing a
5588 * different set of WHILEM ops (and their associated
5589 * indexes) against the same string, so the bits in the
5590 * cache are meaningless. Setting maxiter to zero forces
5591 * the cache to be invalidated and zeroed before reuse.
5592 * XXX This is too dramatic a measure. Ideally we should
5593 * save the old cache and restore when running the outer
5595 reginfo->poscache_maxiter = 0;
5597 /* the new regexp might have a different is_utf8_pat than we do */
5598 is_utf8_pat = reginfo->is_utf8_pat = cBOOL(RX_UTF8(re_sv));
5600 ST.prev_rex = rex_sv;
5601 ST.prev_curlyx = cur_curlyx;
5603 SET_reg_curpm(rex_sv);
5608 ST.prev_eval = cur_eval;
5610 /* now continue from first node in postoned RE */
5611 PUSH_YES_STATE_GOTO(EVAL_AB, startpoint, locinput);
5616 case EVAL_AB: /* cleanup after a successful (??{A})B */
5617 /* note: this is called twice; first after popping B, then A */
5618 rex_sv = ST.prev_rex;
5619 is_utf8_pat = reginfo->is_utf8_pat = cBOOL(RX_UTF8(rex_sv));
5620 SET_reg_curpm(rex_sv);
5621 rex = ReANY(rex_sv);
5622 rexi = RXi_GET(rex);
5624 /* preserve $^R across LEAVE's. See Bug 121070. */
5625 SV *save_sv= GvSV(PL_replgv);
5626 SvREFCNT_inc(save_sv);
5627 regcpblow(ST.cp); /* LEAVE in disguise */
5628 sv_setsv(GvSV(PL_replgv), save_sv);
5629 SvREFCNT_dec(save_sv);
5631 cur_eval = ST.prev_eval;
5632 cur_curlyx = ST.prev_curlyx;
5634 /* Invalidate cache. See "invalidate" comment above. */
5635 reginfo->poscache_maxiter = 0;
5636 if ( nochange_depth )
5641 case EVAL_AB_fail: /* unsuccessfully ran A or B in (??{A})B */
5642 /* note: this is called twice; first after popping B, then A */
5643 rex_sv = ST.prev_rex;
5644 is_utf8_pat = reginfo->is_utf8_pat = cBOOL(RX_UTF8(rex_sv));
5645 SET_reg_curpm(rex_sv);
5646 rex = ReANY(rex_sv);
5647 rexi = RXi_GET(rex);
5649 REGCP_UNWIND(ST.lastcp);
5650 regcppop(rex, &maxopenparen);
5651 cur_eval = ST.prev_eval;
5652 cur_curlyx = ST.prev_curlyx;
5653 /* Invalidate cache. See "invalidate" comment above. */
5654 reginfo->poscache_maxiter = 0;
5655 if ( nochange_depth )
5661 n = ARG(scan); /* which paren pair */
5662 rex->offs[n].start_tmp = locinput - reginfo->strbeg;
5663 if (n > maxopenparen)
5665 DEBUG_BUFFERS_r(PerlIO_printf(Perl_debug_log,
5666 "rex=0x%"UVxf" offs=0x%"UVxf": \\%"UVuf": set %"IVdf" tmp; maxopenparen=%"UVuf"\n",
5670 (IV)rex->offs[n].start_tmp,
5676 /* XXX really need to log other places start/end are set too */
5677 #define CLOSE_CAPTURE \
5678 rex->offs[n].start = rex->offs[n].start_tmp; \
5679 rex->offs[n].end = locinput - reginfo->strbeg; \
5680 DEBUG_BUFFERS_r(PerlIO_printf(Perl_debug_log, \
5681 "rex=0x%"UVxf" offs=0x%"UVxf": \\%"UVuf": set %"IVdf"..%"IVdf"\n", \
5683 PTR2UV(rex->offs), \
5685 (IV)rex->offs[n].start, \
5686 (IV)rex->offs[n].end \
5690 n = ARG(scan); /* which paren pair */
5692 if (n > rex->lastparen)
5694 rex->lastcloseparen = n;
5695 if (cur_eval && cur_eval->u.eval.close_paren == n) {
5700 case ACCEPT: /* (*ACCEPT) */
5704 cursor && OP(cursor)!=END;
5705 cursor=regnext(cursor))
5707 if ( OP(cursor)==CLOSE ){
5709 if ( n <= lastopen ) {
5711 if (n > rex->lastparen)
5713 rex->lastcloseparen = n;
5714 if ( n == ARG(scan) || (cur_eval &&
5715 cur_eval->u.eval.close_paren == n))
5724 case GROUPP: /* (?(1)) */
5725 n = ARG(scan); /* which paren pair */
5726 sw = cBOOL(rex->lastparen >= n && rex->offs[n].end != -1);
5729 case NGROUPP: /* (?(<name>)) */
5730 /* reg_check_named_buff_matched returns 0 for no match */
5731 sw = cBOOL(0 < reg_check_named_buff_matched(rex,scan));
5734 case INSUBP: /* (?(R)) */
5736 sw = (cur_eval && (!n || cur_eval->u.eval.close_paren == n));
5739 case DEFINEP: /* (?(DEFINE)) */
5743 case IFTHEN: /* (?(cond)A|B) */
5744 reginfo->poscache_iter = reginfo->poscache_maxiter; /* Void cache */
5746 next = NEXTOPER(NEXTOPER(scan));
5748 next = scan + ARG(scan);
5749 if (OP(next) == IFTHEN) /* Fake one. */
5750 next = NEXTOPER(NEXTOPER(next));
5754 case LOGICAL: /* modifier for EVAL and IFMATCH */
5755 logical = scan->flags;
5758 /*******************************************************************
5760 The CURLYX/WHILEM pair of ops handle the most generic case of the /A*B/
5761 pattern, where A and B are subpatterns. (For simple A, CURLYM or
5762 STAR/PLUS/CURLY/CURLYN are used instead.)
5764 A*B is compiled as <CURLYX><A><WHILEM><B>
5766 On entry to the subpattern, CURLYX is called. This pushes a CURLYX
5767 state, which contains the current count, initialised to -1. It also sets
5768 cur_curlyx to point to this state, with any previous value saved in the
5771 CURLYX then jumps straight to the WHILEM op, rather than executing A,
5772 since the pattern may possibly match zero times (i.e. it's a while {} loop
5773 rather than a do {} while loop).
5775 Each entry to WHILEM represents a successful match of A. The count in the
5776 CURLYX block is incremented, another WHILEM state is pushed, and execution
5777 passes to A or B depending on greediness and the current count.
5779 For example, if matching against the string a1a2a3b (where the aN are
5780 substrings that match /A/), then the match progresses as follows: (the
5781 pushed states are interspersed with the bits of strings matched so far):
5784 <CURLYX cnt=0><WHILEM>
5785 <CURLYX cnt=1><WHILEM> a1 <WHILEM>
5786 <CURLYX cnt=2><WHILEM> a1 <WHILEM> a2 <WHILEM>
5787 <CURLYX cnt=3><WHILEM> a1 <WHILEM> a2 <WHILEM> a3 <WHILEM>
5788 <CURLYX cnt=3><WHILEM> a1 <WHILEM> a2 <WHILEM> a3 <WHILEM> b
5790 (Contrast this with something like CURLYM, which maintains only a single
5794 a1 <CURLYM cnt=1> a2
5795 a1 a2 <CURLYM cnt=2> a3
5796 a1 a2 a3 <CURLYM cnt=3> b
5799 Each WHILEM state block marks a point to backtrack to upon partial failure
5800 of A or B, and also contains some minor state data related to that
5801 iteration. The CURLYX block, pointed to by cur_curlyx, contains the
5802 overall state, such as the count, and pointers to the A and B ops.
5804 This is complicated slightly by nested CURLYX/WHILEM's. Since cur_curlyx
5805 must always point to the *current* CURLYX block, the rules are:
5807 When executing CURLYX, save the old cur_curlyx in the CURLYX state block,
5808 and set cur_curlyx to point the new block.
5810 When popping the CURLYX block after a successful or unsuccessful match,
5811 restore the previous cur_curlyx.
5813 When WHILEM is about to execute B, save the current cur_curlyx, and set it
5814 to the outer one saved in the CURLYX block.
5816 When popping the WHILEM block after a successful or unsuccessful B match,
5817 restore the previous cur_curlyx.
5819 Here's an example for the pattern (AI* BI)*BO
5820 I and O refer to inner and outer, C and W refer to CURLYX and WHILEM:
5823 curlyx backtrack stack
5824 ------ ---------------
5826 CO <CO prev=NULL> <WO>
5827 CI <CO prev=NULL> <WO> <CI prev=CO> <WI> ai
5828 CO <CO prev=NULL> <WO> <CI prev=CO> <WI> ai <WI prev=CI> bi
5829 NULL <CO prev=NULL> <WO> <CI prev=CO> <WI> ai <WI prev=CI> bi <WO prev=CO> bo
5831 At this point the pattern succeeds, and we work back down the stack to
5832 clean up, restoring as we go:
5834 CO <CO prev=NULL> <WO> <CI prev=CO> <WI> ai <WI prev=CI> bi
5835 CI <CO prev=NULL> <WO> <CI prev=CO> <WI> ai
5836 CO <CO prev=NULL> <WO>
5839 *******************************************************************/
5841 #define ST st->u.curlyx
5843 case CURLYX: /* start of /A*B/ (for complex A) */
5845 /* No need to save/restore up to this paren */
5846 I32 parenfloor = scan->flags;
5848 assert(next); /* keep Coverity happy */
5849 if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */
5852 /* XXXX Probably it is better to teach regpush to support
5853 parenfloor > maxopenparen ... */
5854 if (parenfloor > (I32)rex->lastparen)
5855 parenfloor = rex->lastparen; /* Pessimization... */
5857 ST.prev_curlyx= cur_curlyx;
5859 ST.cp = PL_savestack_ix;
5861 /* these fields contain the state of the current curly.
5862 * they are accessed by subsequent WHILEMs */
5863 ST.parenfloor = parenfloor;
5868 ST.count = -1; /* this will be updated by WHILEM */
5869 ST.lastloc = NULL; /* this will be updated by WHILEM */
5871 PUSH_YES_STATE_GOTO(CURLYX_end, PREVOPER(next), locinput);
5876 case CURLYX_end: /* just finished matching all of A*B */
5877 cur_curlyx = ST.prev_curlyx;
5882 case CURLYX_end_fail: /* just failed to match all of A*B */
5884 cur_curlyx = ST.prev_curlyx;
5891 #define ST st->u.whilem
5893 case WHILEM: /* just matched an A in /A*B/ (for complex A) */
5895 /* see the discussion above about CURLYX/WHILEM */
5900 assert(cur_curlyx); /* keep Coverity happy */
5902 min = ARG1(cur_curlyx->u.curlyx.me);
5903 max = ARG2(cur_curlyx->u.curlyx.me);
5904 A = NEXTOPER(cur_curlyx->u.curlyx.me) + EXTRA_STEP_2ARGS;
5905 n = ++cur_curlyx->u.curlyx.count; /* how many A's matched */
5906 ST.save_lastloc = cur_curlyx->u.curlyx.lastloc;
5907 ST.cache_offset = 0;
5911 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
5912 "%*s whilem: matched %ld out of %d..%d\n",
5913 REPORT_CODE_OFF+depth*2, "", (long)n, min, max)
5916 /* First just match a string of min A's. */
5919 ST.cp = regcppush(rex, cur_curlyx->u.curlyx.parenfloor,
5921 cur_curlyx->u.curlyx.lastloc = locinput;
5922 REGCP_SET(ST.lastcp);
5924 PUSH_STATE_GOTO(WHILEM_A_pre, A, locinput);
5929 /* If degenerate A matches "", assume A done. */
5931 if (locinput == cur_curlyx->u.curlyx.lastloc) {
5932 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
5933 "%*s whilem: empty match detected, trying continuation...\n",
5934 REPORT_CODE_OFF+depth*2, "")
5936 goto do_whilem_B_max;
5939 /* super-linear cache processing.
5941 * The idea here is that for certain types of CURLYX/WHILEM -
5942 * principally those whose upper bound is infinity (and
5943 * excluding regexes that have things like \1 and other very
5944 * non-regular expresssiony things), then if a pattern like
5945 * /....A*.../ fails and we backtrack to the WHILEM, then we
5946 * make a note that this particular WHILEM op was at string
5947 * position 47 (say) when the rest of pattern failed. Then, if
5948 * we ever find ourselves back at that WHILEM, and at string
5949 * position 47 again, we can just fail immediately rather than
5950 * running the rest of the pattern again.
5952 * This is very handy when patterns start to go
5953 * 'super-linear', like in (a+)*(a+)*(a+)*, where you end up
5954 * with a combinatorial explosion of backtracking.
5956 * The cache is implemented as a bit array, with one bit per
5957 * string byte position per WHILEM op (up to 16) - so its
5958 * between 0.25 and 2x the string size.
5960 * To avoid allocating a poscache buffer every time, we do an
5961 * initially countdown; only after we have executed a WHILEM
5962 * op (string-length x #WHILEMs) times do we allocate the
5965 * The top 4 bits of scan->flags byte say how many different
5966 * relevant CURLLYX/WHILEM op pairs there are, while the
5967 * bottom 4-bits is the identifying index number of this
5973 if (!reginfo->poscache_maxiter) {
5974 /* start the countdown: Postpone detection until we
5975 * know the match is not *that* much linear. */
5976 reginfo->poscache_maxiter
5977 = (reginfo->strend - reginfo->strbeg + 1)
5979 /* possible overflow for long strings and many CURLYX's */
5980 if (reginfo->poscache_maxiter < 0)
5981 reginfo->poscache_maxiter = I32_MAX;
5982 reginfo->poscache_iter = reginfo->poscache_maxiter;
5985 if (reginfo->poscache_iter-- == 0) {
5986 /* initialise cache */
5987 const SSize_t size = (reginfo->poscache_maxiter + 7)/8;
5988 regmatch_info_aux *const aux = reginfo->info_aux;
5989 if (aux->poscache) {
5990 if ((SSize_t)reginfo->poscache_size < size) {
5991 Renew(aux->poscache, size, char);
5992 reginfo->poscache_size = size;
5994 Zero(aux->poscache, size, char);
5997 reginfo->poscache_size = size;
5998 Newxz(aux->poscache, size, char);
6000 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
6001 "%swhilem: Detected a super-linear match, switching on caching%s...\n",
6002 PL_colors[4], PL_colors[5])
6006 if (reginfo->poscache_iter < 0) {
6007 /* have we already failed at this position? */
6008 SSize_t offset, mask;
6010 reginfo->poscache_iter = -1; /* stop eventual underflow */
6011 offset = (scan->flags & 0xf) - 1
6012 + (locinput - reginfo->strbeg)
6014 mask = 1 << (offset % 8);
6016 if (reginfo->info_aux->poscache[offset] & mask) {
6017 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
6018 "%*s whilem: (cache) already tried at this position...\n",
6019 REPORT_CODE_OFF+depth*2, "")
6021 sayNO; /* cache records failure */
6023 ST.cache_offset = offset;
6024 ST.cache_mask = mask;
6028 /* Prefer B over A for minimal matching. */
6030 if (cur_curlyx->u.curlyx.minmod) {
6031 ST.save_curlyx = cur_curlyx;
6032 cur_curlyx = cur_curlyx->u.curlyx.prev_curlyx;
6033 ST.cp = regcppush(rex, ST.save_curlyx->u.curlyx.parenfloor,
6035 REGCP_SET(ST.lastcp);
6036 PUSH_YES_STATE_GOTO(WHILEM_B_min, ST.save_curlyx->u.curlyx.B,
6042 /* Prefer A over B for maximal matching. */
6044 if (n < max) { /* More greed allowed? */
6045 ST.cp = regcppush(rex, cur_curlyx->u.curlyx.parenfloor,
6047 cur_curlyx->u.curlyx.lastloc = locinput;
6048 REGCP_SET(ST.lastcp);
6049 PUSH_STATE_GOTO(WHILEM_A_max, A, locinput);
6053 goto do_whilem_B_max;
6058 case WHILEM_B_min: /* just matched B in a minimal match */
6059 case WHILEM_B_max: /* just matched B in a maximal match */
6060 cur_curlyx = ST.save_curlyx;
6065 case WHILEM_B_max_fail: /* just failed to match B in a maximal match */
6066 cur_curlyx = ST.save_curlyx;
6067 cur_curlyx->u.curlyx.lastloc = ST.save_lastloc;
6068 cur_curlyx->u.curlyx.count--;
6073 case WHILEM_A_min_fail: /* just failed to match A in a minimal match */
6075 case WHILEM_A_pre_fail: /* just failed to match even minimal A */
6076 REGCP_UNWIND(ST.lastcp);
6077 regcppop(rex, &maxopenparen);
6078 cur_curlyx->u.curlyx.lastloc = ST.save_lastloc;
6079 cur_curlyx->u.curlyx.count--;
6084 case WHILEM_A_max_fail: /* just failed to match A in a maximal match */
6085 REGCP_UNWIND(ST.lastcp);
6086 regcppop(rex, &maxopenparen); /* Restore some previous $<digit>s? */
6087 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
6088 "%*s whilem: failed, trying continuation...\n",
6089 REPORT_CODE_OFF+depth*2, "")
6092 if (cur_curlyx->u.curlyx.count >= REG_INFTY
6093 && ckWARN(WARN_REGEXP)
6094 && !reginfo->warned)
6096 reginfo->warned = TRUE;
6097 Perl_warner(aTHX_ packWARN(WARN_REGEXP),
6098 "Complex regular subexpression recursion limit (%d) "
6104 ST.save_curlyx = cur_curlyx;
6105 cur_curlyx = cur_curlyx->u.curlyx.prev_curlyx;
6106 PUSH_YES_STATE_GOTO(WHILEM_B_max, ST.save_curlyx->u.curlyx.B,
6111 case WHILEM_B_min_fail: /* just failed to match B in a minimal match */
6112 cur_curlyx = ST.save_curlyx;
6113 REGCP_UNWIND(ST.lastcp);
6114 regcppop(rex, &maxopenparen);
6116 if (cur_curlyx->u.curlyx.count >= /*max*/ARG2(cur_curlyx->u.curlyx.me)) {
6117 /* Maximum greed exceeded */
6118 if (cur_curlyx->u.curlyx.count >= REG_INFTY
6119 && ckWARN(WARN_REGEXP)
6120 && !reginfo->warned)
6122 reginfo->warned = TRUE;
6123 Perl_warner(aTHX_ packWARN(WARN_REGEXP),
6124 "Complex regular subexpression recursion "
6125 "limit (%d) exceeded",
6128 cur_curlyx->u.curlyx.count--;
6132 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
6133 "%*s trying longer...\n", REPORT_CODE_OFF+depth*2, "")
6135 /* Try grabbing another A and see if it helps. */
6136 cur_curlyx->u.curlyx.lastloc = locinput;
6137 ST.cp = regcppush(rex, cur_curlyx->u.curlyx.parenfloor,
6139 REGCP_SET(ST.lastcp);
6140 PUSH_STATE_GOTO(WHILEM_A_min,
6141 /*A*/ NEXTOPER(ST.save_curlyx->u.curlyx.me) + EXTRA_STEP_2ARGS,
6147 #define ST st->u.branch
6149 case BRANCHJ: /* /(...|A|...)/ with long next pointer */
6150 next = scan + ARG(scan);
6153 scan = NEXTOPER(scan);
6156 case BRANCH: /* /(...|A|...)/ */
6157 scan = NEXTOPER(scan); /* scan now points to inner node */
6158 ST.lastparen = rex->lastparen;
6159 ST.lastcloseparen = rex->lastcloseparen;
6160 ST.next_branch = next;
6163 /* Now go into the branch */
6165 PUSH_YES_STATE_GOTO(BRANCH_next, scan, locinput);
6167 PUSH_STATE_GOTO(BRANCH_next, scan, locinput);
6172 case CUTGROUP: /* /(*THEN)/ */
6173 sv_yes_mark = st->u.mark.mark_name = scan->flags ? NULL :
6174 MUTABLE_SV(rexi->data->data[ ARG( scan ) ]);
6175 PUSH_STATE_GOTO(CUTGROUP_next, next, locinput);
6179 case CUTGROUP_next_fail:
6182 if (st->u.mark.mark_name)
6183 sv_commit = st->u.mark.mark_name;
6193 case BRANCH_next_fail: /* that branch failed; try the next, if any */
6198 REGCP_UNWIND(ST.cp);
6199 UNWIND_PAREN(ST.lastparen, ST.lastcloseparen);
6200 scan = ST.next_branch;
6201 /* no more branches? */
6202 if (!scan || (OP(scan) != BRANCH && OP(scan) != BRANCHJ)) {
6204 PerlIO_printf( Perl_debug_log,
6205 "%*s %sBRANCH failed...%s\n",
6206 REPORT_CODE_OFF+depth*2, "",
6212 continue; /* execute next BRANCH[J] op */
6215 case MINMOD: /* next op will be non-greedy, e.g. A*? */
6220 #define ST st->u.curlym
6222 case CURLYM: /* /A{m,n}B/ where A is fixed-length */
6224 /* This is an optimisation of CURLYX that enables us to push
6225 * only a single backtracking state, no matter how many matches
6226 * there are in {m,n}. It relies on the pattern being constant
6227 * length, with no parens to influence future backrefs
6231 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
6233 ST.lastparen = rex->lastparen;
6234 ST.lastcloseparen = rex->lastcloseparen;
6236 /* if paren positive, emulate an OPEN/CLOSE around A */
6238 U32 paren = ST.me->flags;
6239 if (paren > maxopenparen)
6240 maxopenparen = paren;
6241 scan += NEXT_OFF(scan); /* Skip former OPEN. */
6249 ST.c1 = CHRTEST_UNINIT;
6252 if (!(ST.minmod ? ARG1(ST.me) : ARG2(ST.me))) /* min/max */
6255 curlym_do_A: /* execute the A in /A{m,n}B/ */
6256 PUSH_YES_STATE_GOTO(CURLYM_A, ST.A, locinput); /* match A */
6260 case CURLYM_A: /* we've just matched an A */
6262 /* after first match, determine A's length: u.curlym.alen */
6263 if (ST.count == 1) {
6264 if (reginfo->is_utf8_target) {
6265 char *s = st->locinput;
6266 while (s < locinput) {
6272 ST.alen = locinput - st->locinput;
6275 ST.count = ST.minmod ? ARG1(ST.me) : ARG2(ST.me);
6278 PerlIO_printf(Perl_debug_log,
6279 "%*s CURLYM now matched %"IVdf" times, len=%"IVdf"...\n",
6280 (int)(REPORT_CODE_OFF+(depth*2)), "",
6281 (IV) ST.count, (IV)ST.alen)
6284 if (cur_eval && cur_eval->u.eval.close_paren &&
6285 cur_eval->u.eval.close_paren == (U32)ST.me->flags)
6289 I32 max = (ST.minmod ? ARG1(ST.me) : ARG2(ST.me));
6290 if ( max == REG_INFTY || ST.count < max )
6291 goto curlym_do_A; /* try to match another A */
6293 goto curlym_do_B; /* try to match B */
6295 case CURLYM_A_fail: /* just failed to match an A */
6296 REGCP_UNWIND(ST.cp);
6298 if (ST.minmod || ST.count < ARG1(ST.me) /* min*/
6299 || (cur_eval && cur_eval->u.eval.close_paren &&
6300 cur_eval->u.eval.close_paren == (U32)ST.me->flags))
6303 curlym_do_B: /* execute the B in /A{m,n}B/ */
6304 if (ST.c1 == CHRTEST_UNINIT) {
6305 /* calculate c1 and c2 for possible match of 1st char
6306 * following curly */
6307 ST.c1 = ST.c2 = CHRTEST_VOID;
6309 if (HAS_TEXT(ST.B) || JUMPABLE(ST.B)) {
6310 regnode *text_node = ST.B;
6311 if (! HAS_TEXT(text_node))
6312 FIND_NEXT_IMPT(text_node);
6315 (HAS_TEXT(text_node) && PL_regkind[OP(text_node)] == EXACT)
6317 But the former is redundant in light of the latter.
6319 if this changes back then the macro for
6320 IS_TEXT and friends need to change.
6322 if (PL_regkind[OP(text_node)] == EXACT) {
6323 if (! S_setup_EXACTISH_ST_c1_c2(aTHX_
6324 text_node, &ST.c1, ST.c1_utf8, &ST.c2, ST.c2_utf8,
6334 PerlIO_printf(Perl_debug_log,
6335 "%*s CURLYM trying tail with matches=%"IVdf"...\n",
6336 (int)(REPORT_CODE_OFF+(depth*2)),
6339 if (! NEXTCHR_IS_EOS && ST.c1 != CHRTEST_VOID) {
6340 if (! UTF8_IS_INVARIANT(nextchr) && utf8_target) {
6341 if (memNE(locinput, ST.c1_utf8, UTF8SKIP(locinput))
6342 && memNE(locinput, ST.c2_utf8, UTF8SKIP(locinput)))
6344 /* simulate B failing */
6346 PerlIO_printf(Perl_debug_log,
6347 "%*s CURLYM Fast bail next target=0x%"UVXf" c1=0x%"UVXf" c2=0x%"UVXf"\n",
6348 (int)(REPORT_CODE_OFF+(depth*2)),"",
6349 valid_utf8_to_uvchr((U8 *) locinput, NULL),
6350 valid_utf8_to_uvchr(ST.c1_utf8, NULL),
6351 valid_utf8_to_uvchr(ST.c2_utf8, NULL))
6353 state_num = CURLYM_B_fail;
6354 goto reenter_switch;
6357 else if (nextchr != ST.c1 && nextchr != ST.c2) {
6358 /* simulate B failing */
6360 PerlIO_printf(Perl_debug_log,
6361 "%*s CURLYM Fast bail next target=0x%X c1=0x%X c2=0x%X\n",
6362 (int)(REPORT_CODE_OFF+(depth*2)),"",
6363 (int) nextchr, ST.c1, ST.c2)
6365 state_num = CURLYM_B_fail;
6366 goto reenter_switch;
6371 /* emulate CLOSE: mark current A as captured */
6372 I32 paren = ST.me->flags;
6374 rex->offs[paren].start
6375 = HOPc(locinput, -ST.alen) - reginfo->strbeg;
6376 rex->offs[paren].end = locinput - reginfo->strbeg;
6377 if ((U32)paren > rex->lastparen)
6378 rex->lastparen = paren;
6379 rex->lastcloseparen = paren;
6382 rex->offs[paren].end = -1;
6383 if (cur_eval && cur_eval->u.eval.close_paren &&
6384 cur_eval->u.eval.close_paren == (U32)ST.me->flags)
6393 PUSH_STATE_GOTO(CURLYM_B, ST.B, locinput); /* match B */
6397 case CURLYM_B_fail: /* just failed to match a B */
6398 REGCP_UNWIND(ST.cp);
6399 UNWIND_PAREN(ST.lastparen, ST.lastcloseparen);
6401 I32 max = ARG2(ST.me);
6402 if (max != REG_INFTY && ST.count == max)
6404 goto curlym_do_A; /* try to match a further A */
6406 /* backtrack one A */
6407 if (ST.count == ARG1(ST.me) /* min */)
6410 SET_locinput(HOPc(locinput, -ST.alen));
6411 goto curlym_do_B; /* try to match B */
6414 #define ST st->u.curly
6416 #define CURLY_SETPAREN(paren, success) \
6419 rex->offs[paren].start = HOPc(locinput, -1) - reginfo->strbeg; \
6420 rex->offs[paren].end = locinput - reginfo->strbeg; \
6421 if (paren > rex->lastparen) \
6422 rex->lastparen = paren; \
6423 rex->lastcloseparen = paren; \
6426 rex->offs[paren].end = -1; \
6427 rex->lastparen = ST.lastparen; \
6428 rex->lastcloseparen = ST.lastcloseparen; \
6432 case STAR: /* /A*B/ where A is width 1 char */
6436 scan = NEXTOPER(scan);
6439 case PLUS: /* /A+B/ where A is width 1 char */
6443 scan = NEXTOPER(scan);
6446 case CURLYN: /* /(A){m,n}B/ where A is width 1 char */
6447 ST.paren = scan->flags; /* Which paren to set */
6448 ST.lastparen = rex->lastparen;
6449 ST.lastcloseparen = rex->lastcloseparen;
6450 if (ST.paren > maxopenparen)
6451 maxopenparen = ST.paren;
6452 ST.min = ARG1(scan); /* min to match */
6453 ST.max = ARG2(scan); /* max to match */
6454 if (cur_eval && cur_eval->u.eval.close_paren &&
6455 cur_eval->u.eval.close_paren == (U32)ST.paren) {
6459 scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE);
6462 case CURLY: /* /A{m,n}B/ where A is width 1 char */
6464 ST.min = ARG1(scan); /* min to match */
6465 ST.max = ARG2(scan); /* max to match */
6466 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
6469 * Lookahead to avoid useless match attempts
6470 * when we know what character comes next.
6472 * Used to only do .*x and .*?x, but now it allows
6473 * for )'s, ('s and (?{ ... })'s to be in the way
6474 * of the quantifier and the EXACT-like node. -- japhy
6477 assert(ST.min <= ST.max);
6478 if (! HAS_TEXT(next) && ! JUMPABLE(next)) {
6479 ST.c1 = ST.c2 = CHRTEST_VOID;
6482 regnode *text_node = next;
6484 if (! HAS_TEXT(text_node))
6485 FIND_NEXT_IMPT(text_node);
6487 if (! HAS_TEXT(text_node))
6488 ST.c1 = ST.c2 = CHRTEST_VOID;
6490 if ( PL_regkind[OP(text_node)] != EXACT ) {
6491 ST.c1 = ST.c2 = CHRTEST_VOID;
6495 /* Currently we only get here when
6497 PL_rekind[OP(text_node)] == EXACT
6499 if this changes back then the macro for IS_TEXT and
6500 friends need to change. */
6501 if (! S_setup_EXACTISH_ST_c1_c2(aTHX_
6502 text_node, &ST.c1, ST.c1_utf8, &ST.c2, ST.c2_utf8,
6514 char *li = locinput;
6517 regrepeat(rex, &li, ST.A, reginfo, ST.min, depth)
6523 if (ST.c1 == CHRTEST_VOID)
6524 goto curly_try_B_min;
6526 ST.oldloc = locinput;
6528 /* set ST.maxpos to the furthest point along the
6529 * string that could possibly match */
6530 if (ST.max == REG_INFTY) {
6531 ST.maxpos = reginfo->strend - 1;
6533 while (UTF8_IS_CONTINUATION(*(U8*)ST.maxpos))
6536 else if (utf8_target) {
6537 int m = ST.max - ST.min;
6538 for (ST.maxpos = locinput;
6539 m >0 && ST.maxpos < reginfo->strend; m--)
6540 ST.maxpos += UTF8SKIP(ST.maxpos);
6543 ST.maxpos = locinput + ST.max - ST.min;
6544 if (ST.maxpos >= reginfo->strend)
6545 ST.maxpos = reginfo->strend - 1;
6547 goto curly_try_B_min_known;
6551 /* avoid taking address of locinput, so it can remain
6553 char *li = locinput;
6554 ST.count = regrepeat(rex, &li, ST.A, reginfo, ST.max, depth);
6555 if (ST.count < ST.min)
6558 if ((ST.count > ST.min)
6559 && (PL_regkind[OP(ST.B)] == EOL) && (OP(ST.B) != MEOL))
6561 /* A{m,n} must come at the end of the string, there's
6562 * no point in backing off ... */
6564 /* ...except that $ and \Z can match before *and* after
6565 newline at the end. Consider "\n\n" =~ /\n+\Z\n/.
6566 We may back off by one in this case. */
6567 if (UCHARAT(locinput - 1) == '\n' && OP(ST.B) != EOS)
6571 goto curly_try_B_max;
6576 case CURLY_B_min_known_fail:
6577 /* failed to find B in a non-greedy match where c1,c2 valid */
6579 REGCP_UNWIND(ST.cp);
6581 UNWIND_PAREN(ST.lastparen, ST.lastcloseparen);
6583 /* Couldn't or didn't -- move forward. */
6584 ST.oldloc = locinput;
6586 locinput += UTF8SKIP(locinput);
6590 curly_try_B_min_known:
6591 /* find the next place where 'B' could work, then call B */
6595 n = (ST.oldloc == locinput) ? 0 : 1;
6596 if (ST.c1 == ST.c2) {
6597 /* set n to utf8_distance(oldloc, locinput) */
6598 while (locinput <= ST.maxpos
6599 && memNE(locinput, ST.c1_utf8, UTF8SKIP(locinput)))
6601 locinput += UTF8SKIP(locinput);
6606 /* set n to utf8_distance(oldloc, locinput) */
6607 while (locinput <= ST.maxpos
6608 && memNE(locinput, ST.c1_utf8, UTF8SKIP(locinput))
6609 && memNE(locinput, ST.c2_utf8, UTF8SKIP(locinput)))
6611 locinput += UTF8SKIP(locinput);
6616 else { /* Not utf8_target */
6617 if (ST.c1 == ST.c2) {
6618 while (locinput <= ST.maxpos &&
6619 UCHARAT(locinput) != ST.c1)
6623 while (locinput <= ST.maxpos
6624 && UCHARAT(locinput) != ST.c1
6625 && UCHARAT(locinput) != ST.c2)
6628 n = locinput - ST.oldloc;
6630 if (locinput > ST.maxpos)
6633 /* In /a{m,n}b/, ST.oldloc is at "a" x m, locinput is
6634 * at b; check that everything between oldloc and
6635 * locinput matches */
6636 char *li = ST.oldloc;
6638 if (regrepeat(rex, &li, ST.A, reginfo, n, depth) < n)
6640 assert(n == REG_INFTY || locinput == li);
6642 CURLY_SETPAREN(ST.paren, ST.count);
6643 if (cur_eval && cur_eval->u.eval.close_paren &&
6644 cur_eval->u.eval.close_paren == (U32)ST.paren) {
6647 PUSH_STATE_GOTO(CURLY_B_min_known, ST.B, locinput);
6652 case CURLY_B_min_fail:
6653 /* failed to find B in a non-greedy match where c1,c2 invalid */
6655 REGCP_UNWIND(ST.cp);
6657 UNWIND_PAREN(ST.lastparen, ST.lastcloseparen);
6659 /* failed -- move forward one */
6661 char *li = locinput;
6662 if (!regrepeat(rex, &li, ST.A, reginfo, 1, depth)) {
6669 if (ST.count <= ST.max || (ST.max == REG_INFTY &&
6670 ST.count > 0)) /* count overflow ? */
6673 CURLY_SETPAREN(ST.paren, ST.count);
6674 if (cur_eval && cur_eval->u.eval.close_paren &&
6675 cur_eval->u.eval.close_paren == (U32)ST.paren) {
6678 PUSH_STATE_GOTO(CURLY_B_min, ST.B, locinput);
6686 /* a successful greedy match: now try to match B */
6687 if (cur_eval && cur_eval->u.eval.close_paren &&
6688 cur_eval->u.eval.close_paren == (U32)ST.paren) {
6692 bool could_match = locinput < reginfo->strend;
6694 /* If it could work, try it. */
6695 if (ST.c1 != CHRTEST_VOID && could_match) {
6696 if (! UTF8_IS_INVARIANT(UCHARAT(locinput)) && utf8_target)
6698 could_match = memEQ(locinput,
6703 UTF8SKIP(locinput));
6706 could_match = UCHARAT(locinput) == ST.c1
6707 || UCHARAT(locinput) == ST.c2;
6710 if (ST.c1 == CHRTEST_VOID || could_match) {
6711 CURLY_SETPAREN(ST.paren, ST.count);
6712 PUSH_STATE_GOTO(CURLY_B_max, ST.B, locinput);
6719 case CURLY_B_max_fail:
6720 /* failed to find B in a greedy match */
6722 REGCP_UNWIND(ST.cp);
6724 UNWIND_PAREN(ST.lastparen, ST.lastcloseparen);
6727 if (--ST.count < ST.min)
6729 locinput = HOPc(locinput, -1);
6730 goto curly_try_B_max;
6734 case END: /* last op of main pattern */
6737 /* we've just finished A in /(??{A})B/; now continue with B */
6739 st->u.eval.prev_rex = rex_sv; /* inner */
6741 /* Save *all* the positions. */
6742 st->u.eval.cp = regcppush(rex, 0, maxopenparen);
6743 rex_sv = cur_eval->u.eval.prev_rex;
6744 is_utf8_pat = reginfo->is_utf8_pat = cBOOL(RX_UTF8(rex_sv));
6745 SET_reg_curpm(rex_sv);
6746 rex = ReANY(rex_sv);
6747 rexi = RXi_GET(rex);
6748 cur_curlyx = cur_eval->u.eval.prev_curlyx;
6750 REGCP_SET(st->u.eval.lastcp);
6752 /* Restore parens of the outer rex without popping the
6754 S_regcp_restore(aTHX_ rex, cur_eval->u.eval.lastcp,
6757 st->u.eval.prev_eval = cur_eval;
6758 cur_eval = cur_eval->u.eval.prev_eval;
6760 PerlIO_printf(Perl_debug_log, "%*s EVAL trying tail ... %"UVxf"\n",
6761 REPORT_CODE_OFF+depth*2, "",PTR2UV(cur_eval)););
6762 if ( nochange_depth )
6765 PUSH_YES_STATE_GOTO(EVAL_AB, st->u.eval.prev_eval->u.eval.B,
6766 locinput); /* match B */
6769 if (locinput < reginfo->till) {
6770 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
6771 "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n",
6773 (long)(locinput - startpos),
6774 (long)(reginfo->till - startpos),
6777 sayNO_SILENT; /* Cannot match: too short. */
6779 sayYES; /* Success! */
6781 case SUCCEED: /* successful SUSPEND/UNLESSM/IFMATCH/CURLYM */
6783 PerlIO_printf(Perl_debug_log,
6784 "%*s %ssubpattern success...%s\n",
6785 REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5]));
6786 sayYES; /* Success! */
6789 #define ST st->u.ifmatch
6794 case SUSPEND: /* (?>A) */
6796 newstart = locinput;
6799 case UNLESSM: /* -ve lookaround: (?!A), or with flags, (?<!A) */
6801 goto ifmatch_trivial_fail_test;
6803 case IFMATCH: /* +ve lookaround: (?=A), or with flags, (?<=A) */
6805 ifmatch_trivial_fail_test:
6807 char * const s = HOPBACKc(locinput, scan->flags);
6812 sw = 1 - cBOOL(ST.wanted);
6816 next = scan + ARG(scan);
6824 newstart = locinput;
6828 ST.logical = logical;
6829 logical = 0; /* XXX: reset state of logical once it has been saved into ST */
6831 /* execute body of (?...A) */
6832 PUSH_YES_STATE_GOTO(IFMATCH_A, NEXTOPER(NEXTOPER(scan)), newstart);
6837 case IFMATCH_A_fail: /* body of (?...A) failed */
6838 ST.wanted = !ST.wanted;
6841 case IFMATCH_A: /* body of (?...A) succeeded */
6843 sw = cBOOL(ST.wanted);
6845 else if (!ST.wanted)
6848 if (OP(ST.me) != SUSPEND) {
6849 /* restore old position except for (?>...) */
6850 locinput = st->locinput;
6852 scan = ST.me + ARG(ST.me);
6855 continue; /* execute B */
6859 case LONGJMP: /* alternative with many branches compiles to
6860 * (BRANCHJ; EXACT ...; LONGJMP ) x N */
6861 next = scan + ARG(scan);
6866 case COMMIT: /* (*COMMIT) */
6867 reginfo->cutpoint = reginfo->strend;
6870 case PRUNE: /* (*PRUNE) */
6872 sv_yes_mark = sv_commit = MUTABLE_SV(rexi->data->data[ ARG( scan ) ]);
6873 PUSH_STATE_GOTO(COMMIT_next, next, locinput);
6877 case COMMIT_next_fail:
6881 case OPFAIL: /* (*FAIL) */
6886 #define ST st->u.mark
6887 case MARKPOINT: /* (*MARK:foo) */
6888 ST.prev_mark = mark_state;
6889 ST.mark_name = sv_commit = sv_yes_mark
6890 = MUTABLE_SV(rexi->data->data[ ARG( scan ) ]);
6892 ST.mark_loc = locinput;
6893 PUSH_YES_STATE_GOTO(MARKPOINT_next, next, locinput);
6897 case MARKPOINT_next:
6898 mark_state = ST.prev_mark;
6903 case MARKPOINT_next_fail:
6904 if (popmark && sv_eq(ST.mark_name,popmark))
6906 if (ST.mark_loc > startpoint)
6907 reginfo->cutpoint = HOPBACKc(ST.mark_loc, 1);
6908 popmark = NULL; /* we found our mark */
6909 sv_commit = ST.mark_name;
6912 PerlIO_printf(Perl_debug_log,
6913 "%*s %ssetting cutpoint to mark:%"SVf"...%s\n",
6914 REPORT_CODE_OFF+depth*2, "",
6915 PL_colors[4], SVfARG(sv_commit), PL_colors[5]);
6918 mark_state = ST.prev_mark;
6919 sv_yes_mark = mark_state ?
6920 mark_state->u.mark.mark_name : NULL;
6925 case SKIP: /* (*SKIP) */
6927 /* (*SKIP) : if we fail we cut here*/
6928 ST.mark_name = NULL;
6929 ST.mark_loc = locinput;
6930 PUSH_STATE_GOTO(SKIP_next,next, locinput);
6932 /* (*SKIP:NAME) : if there is a (*MARK:NAME) fail where it was,
6933 otherwise do nothing. Meaning we need to scan
6935 regmatch_state *cur = mark_state;
6936 SV *find = MUTABLE_SV(rexi->data->data[ ARG( scan ) ]);
6939 if ( sv_eq( cur->u.mark.mark_name,
6942 ST.mark_name = find;
6943 PUSH_STATE_GOTO( SKIP_next, next, locinput);
6945 cur = cur->u.mark.prev_mark;
6948 /* Didn't find our (*MARK:NAME) so ignore this (*SKIP:NAME) */
6951 case SKIP_next_fail:
6953 /* (*CUT:NAME) - Set up to search for the name as we
6954 collapse the stack*/
6955 popmark = ST.mark_name;
6957 /* (*CUT) - No name, we cut here.*/
6958 if (ST.mark_loc > startpoint)
6959 reginfo->cutpoint = HOPBACKc(ST.mark_loc, 1);
6960 /* but we set sv_commit to latest mark_name if there
6961 is one so they can test to see how things lead to this
6964 sv_commit=mark_state->u.mark.mark_name;
6972 case LNBREAK: /* \R */
6973 if ((n=is_LNBREAK_safe(locinput, reginfo->strend, utf8_target))) {
6980 PerlIO_printf(Perl_error_log, "%"UVxf" %d\n",
6981 PTR2UV(scan), OP(scan));
6982 Perl_croak(aTHX_ "regexp memory corruption");
6984 /* this is a point to jump to in order to increment
6985 * locinput by one character */
6987 assert(!NEXTCHR_IS_EOS);
6989 locinput += PL_utf8skip[nextchr];
6990 /* locinput is allowed to go 1 char off the end, but not 2+ */
6991 if (locinput > reginfo->strend)
7000 /* switch break jumps here */
7001 scan = next; /* prepare to execute the next op and ... */
7002 continue; /* ... jump back to the top, reusing st */
7006 /* push a state that backtracks on success */
7007 st->u.yes.prev_yes_state = yes_state;
7011 /* push a new regex state, then continue at scan */
7013 regmatch_state *newst;
7016 regmatch_state *cur = st;
7017 regmatch_state *curyes = yes_state;
7019 regmatch_slab *slab = PL_regmatch_slab;
7020 for (;curd > -1;cur--,curd--) {
7021 if (cur < SLAB_FIRST(slab)) {
7023 cur = SLAB_LAST(slab);
7025 PerlIO_printf(Perl_error_log, "%*s#%-3d %-10s %s\n",
7026 REPORT_CODE_OFF + 2 + depth * 2,"",
7027 curd, PL_reg_name[cur->resume_state],
7028 (curyes == cur) ? "yes" : ""
7031 curyes = cur->u.yes.prev_yes_state;
7034 DEBUG_STATE_pp("push")
7037 st->locinput = locinput;
7039 if (newst > SLAB_LAST(PL_regmatch_slab))
7040 newst = S_push_slab(aTHX);
7041 PL_regmatch_state = newst;
7043 locinput = pushinput;
7051 * We get here only if there's trouble -- normally "case END" is
7052 * the terminating point.
7054 Perl_croak(aTHX_ "corrupted regexp pointers");
7061 /* we have successfully completed a subexpression, but we must now
7062 * pop to the state marked by yes_state and continue from there */
7063 assert(st != yes_state);
7065 while (st != yes_state) {
7067 if (st < SLAB_FIRST(PL_regmatch_slab)) {
7068 PL_regmatch_slab = PL_regmatch_slab->prev;
7069 st = SLAB_LAST(PL_regmatch_slab);
7073 DEBUG_STATE_pp("pop (no final)");
7075 DEBUG_STATE_pp("pop (yes)");
7081 while (yes_state < SLAB_FIRST(PL_regmatch_slab)
7082 || yes_state > SLAB_LAST(PL_regmatch_slab))
7084 /* not in this slab, pop slab */
7085 depth -= (st - SLAB_FIRST(PL_regmatch_slab) + 1);
7086 PL_regmatch_slab = PL_regmatch_slab->prev;
7087 st = SLAB_LAST(PL_regmatch_slab);
7089 depth -= (st - yes_state);
7092 yes_state = st->u.yes.prev_yes_state;
7093 PL_regmatch_state = st;
7096 locinput= st->locinput;
7097 state_num = st->resume_state + no_final;
7098 goto reenter_switch;
7101 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n",
7102 PL_colors[4], PL_colors[5]));
7104 if (reginfo->info_aux_eval) {
7105 /* each successfully executed (?{...}) block does the equivalent of
7106 * local $^R = do {...}
7107 * When popping the save stack, all these locals would be undone;
7108 * bypass this by setting the outermost saved $^R to the latest
7110 /* I dont know if this is needed or works properly now.
7111 * see code related to PL_replgv elsewhere in this file.
7114 if (oreplsv != GvSV(PL_replgv))
7115 sv_setsv(oreplsv, GvSV(PL_replgv));
7122 PerlIO_printf(Perl_debug_log,
7123 "%*s %sfailed...%s\n",
7124 REPORT_CODE_OFF+depth*2, "",
7125 PL_colors[4], PL_colors[5])
7137 /* there's a previous state to backtrack to */
7139 if (st < SLAB_FIRST(PL_regmatch_slab)) {
7140 PL_regmatch_slab = PL_regmatch_slab->prev;
7141 st = SLAB_LAST(PL_regmatch_slab);
7143 PL_regmatch_state = st;
7144 locinput= st->locinput;
7146 DEBUG_STATE_pp("pop");
7148 if (yes_state == st)
7149 yes_state = st->u.yes.prev_yes_state;
7151 state_num = st->resume_state + 1; /* failure = success + 1 */
7152 goto reenter_switch;
7157 if (rex->intflags & PREGf_VERBARG_SEEN) {
7158 SV *sv_err = get_sv("REGERROR", 1);
7159 SV *sv_mrk = get_sv("REGMARK", 1);
7161 sv_commit = &PL_sv_no;
7163 sv_yes_mark = &PL_sv_yes;
7166 sv_commit = &PL_sv_yes;
7167 sv_yes_mark = &PL_sv_no;
7171 sv_setsv(sv_err, sv_commit);
7172 sv_setsv(sv_mrk, sv_yes_mark);
7176 if (last_pushed_cv) {
7179 PERL_UNUSED_VAR(SP);
7182 assert(!result || locinput - reginfo->strbeg >= 0);
7183 return result ? locinput - reginfo->strbeg : -1;
7187 - regrepeat - repeatedly match something simple, report how many
7189 * What 'simple' means is a node which can be the operand of a quantifier like
7192 * startposp - pointer a pointer to the start position. This is updated
7193 * to point to the byte following the highest successful
7195 * p - the regnode to be repeatedly matched against.
7196 * reginfo - struct holding match state, such as strend
7197 * max - maximum number of things to match.
7198 * depth - (for debugging) backtracking depth.
7201 S_regrepeat(pTHX_ regexp *prog, char **startposp, const regnode *p,
7202 regmatch_info *const reginfo, I32 max, int depth)
7204 char *scan; /* Pointer to current position in target string */
7206 char *loceol = reginfo->strend; /* local version */
7207 I32 hardcount = 0; /* How many matches so far */
7208 bool utf8_target = reginfo->is_utf8_target;
7209 int to_complement = 0; /* Invert the result? */
7211 _char_class_number classnum;
7213 PERL_UNUSED_ARG(depth);
7216 PERL_ARGS_ASSERT_REGREPEAT;
7219 if (max == REG_INFTY)
7221 else if (! utf8_target && loceol - scan > max)
7222 loceol = scan + max;
7224 /* Here, for the case of a non-UTF-8 target we have adjusted <loceol> down
7225 * to the maximum of how far we should go in it (leaving it set to the real
7226 * end, if the maximum permissible would take us beyond that). This allows
7227 * us to make the loop exit condition that we haven't gone past <loceol> to
7228 * also mean that we haven't exceeded the max permissible count, saving a
7229 * test each time through the loop. But it assumes that the OP matches a
7230 * single byte, which is true for most of the OPs below when applied to a
7231 * non-UTF-8 target. Those relatively few OPs that don't have this
7232 * characteristic will have to compensate.
7234 * There is no adjustment for UTF-8 targets, as the number of bytes per
7235 * character varies. OPs will have to test both that the count is less
7236 * than the max permissible (using <hardcount> to keep track), and that we
7237 * are still within the bounds of the string (using <loceol>. A few OPs
7238 * match a single byte no matter what the encoding. They can omit the max
7239 * test if, for the UTF-8 case, they do the adjustment that was skipped
7242 * Thus, the code above sets things up for the common case; and exceptional
7243 * cases need extra work; the common case is to make sure <scan> doesn't
7244 * go past <loceol>, and for UTF-8 to also use <hardcount> to make sure the
7245 * count doesn't exceed the maximum permissible */
7250 while (scan < loceol && hardcount < max && *scan != '\n') {
7251 scan += UTF8SKIP(scan);
7255 while (scan < loceol && *scan != '\n')
7261 while (scan < loceol && hardcount < max) {
7262 scan += UTF8SKIP(scan);
7269 case CANY: /* Move <scan> forward <max> bytes, unless goes off end */
7270 if (utf8_target && loceol - scan > max) {
7272 /* <loceol> hadn't been adjusted in the UTF-8 case */
7280 _CHECK_AND_WARN_PROBLEMATIC_LOCALE;
7281 if (utf8_target && UTF8_IS_ABOVE_LATIN1(*scan)) {
7282 _CHECK_AND_OUTPUT_WIDE_LOCALE_UTF8_MSG(scan, loceol);
7286 assert(STR_LEN(p) == reginfo->is_utf8_pat ? UTF8SKIP(STRING(p)) : 1);
7290 /* Can use a simple loop if the pattern char to match on is invariant
7291 * under UTF-8, or both target and pattern aren't UTF-8. Note that we
7292 * can use UTF8_IS_INVARIANT() even if the pattern isn't UTF-8, as it's
7293 * true iff it doesn't matter if the argument is in UTF-8 or not */
7294 if (UTF8_IS_INVARIANT(c) || (! utf8_target && ! reginfo->is_utf8_pat)) {
7295 if (utf8_target && loceol - scan > max) {
7296 /* We didn't adjust <loceol> because is UTF-8, but ok to do so,
7297 * since here, to match at all, 1 char == 1 byte */
7298 loceol = scan + max;
7300 while (scan < loceol && UCHARAT(scan) == c) {
7304 else if (reginfo->is_utf8_pat) {
7306 STRLEN scan_char_len;
7308 /* When both target and pattern are UTF-8, we have to do
7310 while (hardcount < max
7312 && (scan_char_len = UTF8SKIP(scan)) <= STR_LEN(p)
7313 && memEQ(scan, STRING(p), scan_char_len))
7315 scan += scan_char_len;
7319 else if (! UTF8_IS_ABOVE_LATIN1(c)) {
7321 /* Target isn't utf8; convert the character in the UTF-8
7322 * pattern to non-UTF8, and do a simple loop */
7323 c = TWO_BYTE_UTF8_TO_NATIVE(c, *(STRING(p) + 1));
7324 while (scan < loceol && UCHARAT(scan) == c) {
7327 } /* else pattern char is above Latin1, can't possibly match the
7332 /* Here, the string must be utf8; pattern isn't, and <c> is
7333 * different in utf8 than not, so can't compare them directly.
7334 * Outside the loop, find the two utf8 bytes that represent c, and
7335 * then look for those in sequence in the utf8 string */
7336 U8 high = UTF8_TWO_BYTE_HI(c);
7337 U8 low = UTF8_TWO_BYTE_LO(c);
7339 while (hardcount < max
7340 && scan + 1 < loceol
7341 && UCHARAT(scan) == high
7342 && UCHARAT(scan + 1) == low)
7350 case EXACTFA_NO_TRIE: /* This node only generated for non-utf8 patterns */
7351 assert(! reginfo->is_utf8_pat);
7354 utf8_flags = FOLDEQ_UTF8_NOMIX_ASCII;
7358 _CHECK_AND_WARN_PROBLEMATIC_LOCALE;
7359 utf8_flags = FOLDEQ_LOCALE;
7362 case EXACTF: /* This node only generated for non-utf8 patterns */
7363 assert(! reginfo->is_utf8_pat);
7368 if (! utf8_target) {
7371 utf8_flags = FOLDEQ_LOCALE | FOLDEQ_S2_ALREADY_FOLDED
7372 | FOLDEQ_S2_FOLDS_SANE;
7377 utf8_flags = reginfo->is_utf8_pat ? FOLDEQ_S2_ALREADY_FOLDED : 0;
7381 U8 c1_utf8[UTF8_MAXBYTES+1], c2_utf8[UTF8_MAXBYTES+1];
7383 assert(STR_LEN(p) == reginfo->is_utf8_pat ? UTF8SKIP(STRING(p)) : 1);
7385 if (S_setup_EXACTISH_ST_c1_c2(aTHX_ p, &c1, c1_utf8, &c2, c2_utf8,
7388 if (c1 == CHRTEST_VOID) {
7389 /* Use full Unicode fold matching */
7390 char *tmpeol = reginfo->strend;
7391 STRLEN pat_len = reginfo->is_utf8_pat ? UTF8SKIP(STRING(p)) : 1;
7392 while (hardcount < max
7393 && foldEQ_utf8_flags(scan, &tmpeol, 0, utf8_target,
7394 STRING(p), NULL, pat_len,
7395 reginfo->is_utf8_pat, utf8_flags))
7398 tmpeol = reginfo->strend;
7402 else if (utf8_target) {
7404 while (scan < loceol
7406 && memEQ(scan, c1_utf8, UTF8SKIP(scan)))
7408 scan += UTF8SKIP(scan);
7413 while (scan < loceol
7415 && (memEQ(scan, c1_utf8, UTF8SKIP(scan))
7416 || memEQ(scan, c2_utf8, UTF8SKIP(scan))))
7418 scan += UTF8SKIP(scan);
7423 else if (c1 == c2) {
7424 while (scan < loceol && UCHARAT(scan) == c1) {
7429 while (scan < loceol &&
7430 (UCHARAT(scan) == c1 || UCHARAT(scan) == c2))
7439 _CHECK_AND_WARN_PROBLEMATIC_LOCALE;
7443 while (hardcount < max
7445 && reginclass(prog, p, (U8*)scan, (U8*) loceol, utf8_target))
7447 scan += UTF8SKIP(scan);
7451 while (scan < loceol && REGINCLASS(prog, p, (U8*)scan))
7456 /* The argument (FLAGS) to all the POSIX node types is the class number */
7463 _CHECK_AND_WARN_PROBLEMATIC_LOCALE;
7464 if (! utf8_target) {
7465 while (scan < loceol && to_complement ^ cBOOL(isFOO_lc(FLAGS(p),
7471 while (hardcount < max && scan < loceol
7472 && to_complement ^ cBOOL(isFOO_utf8_lc(FLAGS(p),
7475 scan += UTF8SKIP(scan);
7488 if (utf8_target && loceol - scan > max) {
7490 /* We didn't adjust <loceol> at the beginning of this routine
7491 * because is UTF-8, but it is actually ok to do so, since here, to
7492 * match, 1 char == 1 byte. */
7493 loceol = scan + max;
7495 while (scan < loceol && _generic_isCC_A((U8) *scan, FLAGS(p))) {
7508 if (! utf8_target) {
7509 while (scan < loceol && ! _generic_isCC_A((U8) *scan, FLAGS(p))) {
7515 /* The complement of something that matches only ASCII matches all
7516 * non-ASCII, plus everything in ASCII that isn't in the class. */
7517 while (hardcount < max && scan < loceol
7518 && (! isASCII_utf8(scan)
7519 || ! _generic_isCC_A((U8) *scan, FLAGS(p))))
7521 scan += UTF8SKIP(scan);
7532 if (! utf8_target) {
7533 while (scan < loceol && to_complement
7534 ^ cBOOL(_generic_isCC((U8) *scan, FLAGS(p))))
7541 classnum = (_char_class_number) FLAGS(p);
7542 if (classnum < _FIRST_NON_SWASH_CC) {
7544 /* Here, a swash is needed for above-Latin1 code points.
7545 * Process as many Latin1 code points using the built-in rules.
7546 * Go to another loop to finish processing upon encountering
7547 * the first Latin1 code point. We could do that in this loop
7548 * as well, but the other way saves having to test if the swash
7549 * has been loaded every time through the loop: extra space to
7551 while (hardcount < max && scan < loceol) {
7552 if (UTF8_IS_INVARIANT(*scan)) {
7553 if (! (to_complement ^ cBOOL(_generic_isCC((U8) *scan,
7560 else if (UTF8_IS_DOWNGRADEABLE_START(*scan)) {
7561 if (! (to_complement
7562 ^ cBOOL(_generic_isCC(TWO_BYTE_UTF8_TO_NATIVE(*scan,
7571 goto found_above_latin1;
7578 /* For these character classes, the knowledge of how to handle
7579 * every code point is compiled in to Perl via a macro. This
7580 * code is written for making the loops as tight as possible.
7581 * It could be refactored to save space instead */
7583 case _CC_ENUM_SPACE: /* XXX would require separate code
7584 if we revert the change of \v
7587 case _CC_ENUM_PSXSPC:
7588 while (hardcount < max
7590 && (to_complement ^ cBOOL(isSPACE_utf8(scan))))
7592 scan += UTF8SKIP(scan);
7596 case _CC_ENUM_BLANK:
7597 while (hardcount < max
7599 && (to_complement ^ cBOOL(isBLANK_utf8(scan))))
7601 scan += UTF8SKIP(scan);
7605 case _CC_ENUM_XDIGIT:
7606 while (hardcount < max
7608 && (to_complement ^ cBOOL(isXDIGIT_utf8(scan))))
7610 scan += UTF8SKIP(scan);
7614 case _CC_ENUM_VERTSPACE:
7615 while (hardcount < max
7617 && (to_complement ^ cBOOL(isVERTWS_utf8(scan))))
7619 scan += UTF8SKIP(scan);
7623 case _CC_ENUM_CNTRL:
7624 while (hardcount < max
7626 && (to_complement ^ cBOOL(isCNTRL_utf8(scan))))
7628 scan += UTF8SKIP(scan);
7633 Perl_croak(aTHX_ "panic: regrepeat() node %d='%s' has an unexpected character class '%d'", OP(p), PL_reg_name[OP(p)], classnum);
7639 found_above_latin1: /* Continuation of POSIXU and NPOSIXU */
7641 /* Load the swash if not already present */
7642 if (! PL_utf8_swash_ptrs[classnum]) {
7643 U8 flags = _CORE_SWASH_INIT_ACCEPT_INVLIST;
7644 PL_utf8_swash_ptrs[classnum] = _core_swash_init(
7648 PL_XPosix_ptrs[classnum], &flags);
7651 while (hardcount < max && scan < loceol
7652 && to_complement ^ cBOOL(_generic_utf8(
7655 swash_fetch(PL_utf8_swash_ptrs[classnum],
7659 scan += UTF8SKIP(scan);
7666 while (hardcount < max && scan < loceol &&
7667 (c=is_LNBREAK_utf8_safe(scan, loceol))) {
7672 /* LNBREAK can match one or two latin chars, which is ok, but we
7673 * have to use hardcount in this situation, and throw away the
7674 * adjustment to <loceol> done before the switch statement */
7675 loceol = reginfo->strend;
7676 while (scan < loceol && (c=is_LNBREAK_latin1_safe(scan, loceol))) {
7685 _CHECK_AND_WARN_PROBLEMATIC_LOCALE;
7699 /* These are all 0 width, so match right here or not at all. */
7703 Perl_croak(aTHX_ "panic: regrepeat() called with unrecognized node type %d='%s'", OP(p), PL_reg_name[OP(p)]);
7712 c = scan - *startposp;
7716 GET_RE_DEBUG_FLAGS_DECL;
7718 SV * const prop = sv_newmortal();
7719 regprop(prog, prop, p, reginfo, NULL);
7720 PerlIO_printf(Perl_debug_log,
7721 "%*s %s can match %"IVdf" times out of %"IVdf"...\n",
7722 REPORT_CODE_OFF + depth*2, "", SvPVX_const(prop),(IV)c,(IV)max);
7730 #if !defined(PERL_IN_XSUB_RE) || defined(PLUGGABLE_RE_EXTENSION)
7732 - regclass_swash - prepare the utf8 swash. Wraps the shared core version to
7733 create a copy so that changes the caller makes won't change the shared one.
7734 If <altsvp> is non-null, will return NULL in it, for back-compat.
7737 Perl_regclass_swash(pTHX_ const regexp *prog, const regnode* node, bool doinit, SV** listsvp, SV **altsvp)
7739 PERL_ARGS_ASSERT_REGCLASS_SWASH;
7745 return newSVsv(_get_regclass_nonbitmap_data(prog, node, doinit, listsvp, NULL, NULL));
7748 #endif /* !defined(PERL_IN_XSUB_RE) || defined(PLUGGABLE_RE_EXTENSION) */
7751 - reginclass - determine if a character falls into a character class
7753 n is the ANYOF-type regnode
7754 p is the target string
7755 p_end points to one byte beyond the end of the target string
7756 utf8_target tells whether p is in UTF-8.
7758 Returns true if matched; false otherwise.
7760 Note that this can be a synthetic start class, a combination of various
7761 nodes, so things you think might be mutually exclusive, such as locale,
7762 aren't. It can match both locale and non-locale
7767 S_reginclass(pTHX_ regexp * const prog, const regnode * const n, const U8* const p, const U8* const p_end, const bool utf8_target)
7770 const char flags = ANYOF_FLAGS(n);
7774 PERL_ARGS_ASSERT_REGINCLASS;
7776 /* If c is not already the code point, get it. Note that
7777 * UTF8_IS_INVARIANT() works even if not in UTF-8 */
7778 if (! UTF8_IS_INVARIANT(c) && utf8_target) {
7780 c = utf8n_to_uvchr(p, p_end - p, &c_len,
7781 (UTF8_ALLOW_DEFAULT & UTF8_ALLOW_ANYUV)
7782 | UTF8_ALLOW_FFFF | UTF8_CHECK_ONLY);
7783 /* see [perl #37836] for UTF8_ALLOW_ANYUV; [perl #38293] for
7784 * UTF8_ALLOW_FFFF */
7785 if (c_len == (STRLEN)-1)
7786 Perl_croak(aTHX_ "Malformed UTF-8 character (fatal)");
7787 if (c > 255 && OP(n) == ANYOFL && ! is_ANYOF_SYNTHETIC(n)) {
7788 _CHECK_AND_OUTPUT_WIDE_LOCALE_CP_MSG(c);
7792 /* If this character is potentially in the bitmap, check it */
7793 if (c < NUM_ANYOF_CODE_POINTS) {
7794 if (ANYOF_BITMAP_TEST(n, c))
7796 else if ((flags & ANYOF_MATCHES_ALL_NON_UTF8_NON_ASCII)
7802 else if (flags & ANYOF_LOCALE_FLAGS) {
7803 if ((flags & ANYOF_LOC_FOLD)
7805 && ANYOF_BITMAP_TEST(n, PL_fold_locale[c]))
7809 else if (ANYOF_POSIXL_TEST_ANY_SET(n)
7813 /* The data structure is arranged so bits 0, 2, 4, ... are set
7814 * if the class includes the Posix character class given by
7815 * bit/2; and 1, 3, 5, ... are set if the class includes the
7816 * complemented Posix class given by int(bit/2). So we loop
7817 * through the bits, each time changing whether we complement
7818 * the result or not. Suppose for the sake of illustration
7819 * that bits 0-3 mean respectively, \w, \W, \s, \S. If bit 0
7820 * is set, it means there is a match for this ANYOF node if the
7821 * character is in the class given by the expression (0 / 2 = 0
7822 * = \w). If it is in that class, isFOO_lc() will return 1,
7823 * and since 'to_complement' is 0, the result will stay TRUE,
7824 * and we exit the loop. Suppose instead that bit 0 is 0, but
7825 * bit 1 is 1. That means there is a match if the character
7826 * matches \W. We won't bother to call isFOO_lc() on bit 0,
7827 * but will on bit 1. On the second iteration 'to_complement'
7828 * will be 1, so the exclusive or will reverse things, so we
7829 * are testing for \W. On the third iteration, 'to_complement'
7830 * will be 0, and we would be testing for \s; the fourth
7831 * iteration would test for \S, etc.
7833 * Note that this code assumes that all the classes are closed
7834 * under folding. For example, if a character matches \w, then
7835 * its fold does too; and vice versa. This should be true for
7836 * any well-behaved locale for all the currently defined Posix
7837 * classes, except for :lower: and :upper:, which are handled
7838 * by the pseudo-class :cased: which matches if either of the
7839 * other two does. To get rid of this assumption, an outer
7840 * loop could be used below to iterate over both the source
7841 * character, and its fold (if different) */
7844 int to_complement = 0;
7846 while (count < ANYOF_MAX) {
7847 if (ANYOF_POSIXL_TEST(n, count)
7848 && to_complement ^ cBOOL(isFOO_lc(count/2, (U8) c)))
7861 /* If the bitmap didn't (or couldn't) match, and something outside the
7862 * bitmap could match, try that. */
7864 if (c >= NUM_ANYOF_CODE_POINTS
7865 && (flags & ANYOF_MATCHES_ALL_ABOVE_BITMAP))
7867 match = TRUE; /* Everything above the bitmap matches */
7869 else if ((flags & ANYOF_HAS_NONBITMAP_NON_UTF8_MATCHES)
7870 || (utf8_target && (flags & ANYOF_HAS_UTF8_NONBITMAP_MATCHES))
7871 || ((flags & ANYOF_LOC_FOLD)
7872 && IN_UTF8_CTYPE_LOCALE
7873 && ARG(n) != ANYOF_ONLY_HAS_BITMAP))
7875 SV* only_utf8_locale = NULL;
7876 SV * const sw = _get_regclass_nonbitmap_data(prog, n, TRUE, 0,
7877 &only_utf8_locale, NULL);
7883 } else { /* Convert to utf8 */
7884 utf8_p = utf8_buffer;
7885 append_utf8_from_native_byte(*p, &utf8_p);
7886 utf8_p = utf8_buffer;
7889 if (swash_fetch(sw, utf8_p, TRUE)) {
7893 if (! match && only_utf8_locale && IN_UTF8_CTYPE_LOCALE) {
7894 match = _invlist_contains_cp(only_utf8_locale, c);
7898 if (UNICODE_IS_SUPER(c)
7899 && (flags & ANYOF_WARN_SUPER)
7900 && ckWARN_d(WARN_NON_UNICODE))
7902 Perl_warner(aTHX_ packWARN(WARN_NON_UNICODE),
7903 "Matched non-Unicode code point 0x%04"UVXf" against Unicode property; may not be portable", c);
7907 #if ANYOF_INVERT != 1
7908 /* Depending on compiler optimization cBOOL takes time, so if don't have to
7910 # error ANYOF_INVERT needs to be set to 1, or guarded with cBOOL below,
7913 /* The xor complements the return if to invert: 1^1 = 0, 1^0 = 1 */
7914 return (flags & ANYOF_INVERT) ^ match;
7918 S_reghop3(U8 *s, SSize_t off, const U8* lim)
7920 /* return the position 'off' UTF-8 characters away from 's', forward if
7921 * 'off' >= 0, backwards if negative. But don't go outside of position
7922 * 'lim', which better be < s if off < 0 */
7924 PERL_ARGS_ASSERT_REGHOP3;
7927 while (off-- && s < lim) {
7928 /* XXX could check well-formedness here */
7933 while (off++ && s > lim) {
7935 if (UTF8_IS_CONTINUED(*s)) {
7936 while (s > lim && UTF8_IS_CONTINUATION(*s))
7939 /* XXX could check well-formedness here */
7946 S_reghop4(U8 *s, SSize_t off, const U8* llim, const U8* rlim)
7948 PERL_ARGS_ASSERT_REGHOP4;
7951 while (off-- && s < rlim) {
7952 /* XXX could check well-formedness here */
7957 while (off++ && s > llim) {
7959 if (UTF8_IS_CONTINUED(*s)) {
7960 while (s > llim && UTF8_IS_CONTINUATION(*s))
7963 /* XXX could check well-formedness here */
7969 /* like reghop3, but returns NULL on overrun, rather than returning last
7973 S_reghopmaybe3(U8* s, SSize_t off, const U8* lim)
7975 PERL_ARGS_ASSERT_REGHOPMAYBE3;
7978 while (off-- && s < lim) {
7979 /* XXX could check well-formedness here */
7986 while (off++ && s > lim) {
7988 if (UTF8_IS_CONTINUED(*s)) {
7989 while (s > lim && UTF8_IS_CONTINUATION(*s))
7992 /* XXX could check well-formedness here */
8001 /* when executing a regex that may have (?{}), extra stuff needs setting
8002 up that will be visible to the called code, even before the current
8003 match has finished. In particular:
8005 * $_ is localised to the SV currently being matched;
8006 * pos($_) is created if necessary, ready to be updated on each call-out
8008 * a fake PMOP is created that can be set to PL_curpm (normally PL_curpm
8009 isn't set until the current pattern is successfully finished), so that
8010 $1 etc of the match-so-far can be seen;
8011 * save the old values of subbeg etc of the current regex, and set then
8012 to the current string (again, this is normally only done at the end
8017 S_setup_eval_state(pTHX_ regmatch_info *const reginfo)
8020 regexp *const rex = ReANY(reginfo->prog);
8021 regmatch_info_aux_eval *eval_state = reginfo->info_aux_eval;
8023 eval_state->rex = rex;
8026 /* Make $_ available to executed code. */
8027 if (reginfo->sv != DEFSV) {
8029 DEFSV_set(reginfo->sv);
8032 if (!(mg = mg_find_mglob(reginfo->sv))) {
8033 /* prepare for quick setting of pos */
8034 mg = sv_magicext_mglob(reginfo->sv);
8037 eval_state->pos_magic = mg;
8038 eval_state->pos = mg->mg_len;
8039 eval_state->pos_flags = mg->mg_flags;
8042 eval_state->pos_magic = NULL;
8044 if (!PL_reg_curpm) {
8045 /* PL_reg_curpm is a fake PMOP that we can attach the current
8046 * regex to and point PL_curpm at, so that $1 et al are visible
8047 * within a /(?{})/. It's just allocated once per interpreter the
8048 * first time its needed */
8049 Newxz(PL_reg_curpm, 1, PMOP);
8052 SV* const repointer = &PL_sv_undef;
8053 /* this regexp is also owned by the new PL_reg_curpm, which
8054 will try to free it. */
8055 av_push(PL_regex_padav, repointer);
8056 PL_reg_curpm->op_pmoffset = av_tindex(PL_regex_padav);
8057 PL_regex_pad = AvARRAY(PL_regex_padav);
8061 SET_reg_curpm(reginfo->prog);
8062 eval_state->curpm = PL_curpm;
8063 PL_curpm = PL_reg_curpm;
8064 if (RXp_MATCH_COPIED(rex)) {
8065 /* Here is a serious problem: we cannot rewrite subbeg,
8066 since it may be needed if this match fails. Thus
8067 $` inside (?{}) could fail... */
8068 eval_state->subbeg = rex->subbeg;
8069 eval_state->sublen = rex->sublen;
8070 eval_state->suboffset = rex->suboffset;
8071 eval_state->subcoffset = rex->subcoffset;
8073 eval_state->saved_copy = rex->saved_copy;
8075 RXp_MATCH_COPIED_off(rex);
8078 eval_state->subbeg = NULL;
8079 rex->subbeg = (char *)reginfo->strbeg;
8081 rex->subcoffset = 0;
8082 rex->sublen = reginfo->strend - reginfo->strbeg;
8086 /* destructor to clear up regmatch_info_aux and regmatch_info_aux_eval */
8089 S_cleanup_regmatch_info_aux(pTHX_ void *arg)
8091 regmatch_info_aux *aux = (regmatch_info_aux *) arg;
8092 regmatch_info_aux_eval *eval_state = aux->info_aux_eval;
8095 Safefree(aux->poscache);
8099 /* undo the effects of S_setup_eval_state() */
8101 if (eval_state->subbeg) {
8102 regexp * const rex = eval_state->rex;
8103 rex->subbeg = eval_state->subbeg;
8104 rex->sublen = eval_state->sublen;
8105 rex->suboffset = eval_state->suboffset;
8106 rex->subcoffset = eval_state->subcoffset;
8108 rex->saved_copy = eval_state->saved_copy;
8110 RXp_MATCH_COPIED_on(rex);
8112 if (eval_state->pos_magic)
8114 eval_state->pos_magic->mg_len = eval_state->pos;
8115 eval_state->pos_magic->mg_flags =
8116 (eval_state->pos_magic->mg_flags & ~MGf_BYTES)
8117 | (eval_state->pos_flags & MGf_BYTES);
8120 PL_curpm = eval_state->curpm;
8123 PL_regmatch_state = aux->old_regmatch_state;
8124 PL_regmatch_slab = aux->old_regmatch_slab;
8126 /* free all slabs above current one - this must be the last action
8127 * of this function, as aux and eval_state are allocated within
8128 * slabs and may be freed here */
8130 s = PL_regmatch_slab->next;
8132 PL_regmatch_slab->next = NULL;
8134 regmatch_slab * const osl = s;
8143 S_to_utf8_substr(pTHX_ regexp *prog)
8145 /* Converts substr fields in prog from bytes to UTF-8, calling fbm_compile
8146 * on the converted value */
8150 PERL_ARGS_ASSERT_TO_UTF8_SUBSTR;
8153 if (prog->substrs->data[i].substr
8154 && !prog->substrs->data[i].utf8_substr) {
8155 SV* const sv = newSVsv(prog->substrs->data[i].substr);
8156 prog->substrs->data[i].utf8_substr = sv;
8157 sv_utf8_upgrade(sv);
8158 if (SvVALID(prog->substrs->data[i].substr)) {
8159 if (SvTAIL(prog->substrs->data[i].substr)) {
8160 /* Trim the trailing \n that fbm_compile added last
8162 SvCUR_set(sv, SvCUR(sv) - 1);
8163 /* Whilst this makes the SV technically "invalid" (as its
8164 buffer is no longer followed by "\0") when fbm_compile()
8165 adds the "\n" back, a "\0" is restored. */
8166 fbm_compile(sv, FBMcf_TAIL);
8170 if (prog->substrs->data[i].substr == prog->check_substr)
8171 prog->check_utf8 = sv;
8177 S_to_byte_substr(pTHX_ regexp *prog)
8179 /* Converts substr fields in prog from UTF-8 to bytes, calling fbm_compile
8180 * on the converted value; returns FALSE if can't be converted. */
8184 PERL_ARGS_ASSERT_TO_BYTE_SUBSTR;
8187 if (prog->substrs->data[i].utf8_substr
8188 && !prog->substrs->data[i].substr) {
8189 SV* sv = newSVsv(prog->substrs->data[i].utf8_substr);
8190 if (! sv_utf8_downgrade(sv, TRUE)) {
8193 if (SvVALID(prog->substrs->data[i].utf8_substr)) {
8194 if (SvTAIL(prog->substrs->data[i].utf8_substr)) {
8195 /* Trim the trailing \n that fbm_compile added last
8197 SvCUR_set(sv, SvCUR(sv) - 1);
8198 fbm_compile(sv, FBMcf_TAIL);
8202 prog->substrs->data[i].substr = sv;
8203 if (prog->substrs->data[i].utf8_substr == prog->check_utf8)
8204 prog->check_substr = sv;
8213 * c-indentation-style: bsd
8215 * indent-tabs-mode: nil
8218 * ex: set ts=8 sts=4 sw=4 et:
https://perl5.git.perl.org / perl5.git / blob